English Colour Terms

in Context

Anders Steinvall

Skrifter frn moderna sprk 3

Institutionen fr moderna sprk
Ume universitet 2002

Institutionen fr moderna sprk

Ume universitet
SE-901 87 Ume
Tfn. +46 90 786 51 38
Fax. +46 90 786 60 23
http://www.mos.umu.se/forskning/publikationer/

Skrifter frn moderna sprk 3

Ume universitet ISSN 1650-304X

2002 Anders Steinvall

Omslag: Magnus Olofsson
Tryckt av UmU Tryckerier
ISBN 91-7305-226-4
ISSN 1650-304X

In memory of my father

Contents
Introduction............................................................................................................. 1
0.1
0.2
0.3

Preliminaries..................................................................................................... 1
Aims.................................................................................................................. 2
Method and material......................................................................................... 3
0.3.1
Introduction .......................................................................................... 3
0.3.2
The use of computerised text corpora problems and solutions ......... 5
0.3.3
Statistics................................................................................................ 6
0.3.4
The terms .............................................................................................. 7
0.3.5
Corpora ............................................................................................... 10
0.4 On the terminology of the dimensions of colour............................................ 11
0.5 Overview of the book ..................................................................................... 12

Chapter 1 Setting the Scene: Thirty Years of Studies of Colour Terminology 15

1.1
1.2

1.3
1.4
1.5
1.6
1.7

Introduction .................................................................................................... 15
The dominant theory: Berlin and Kays legacy.............................................. 15
1.2.1
Basic Color Terms .............................................................................. 15
1.2.2
The later development of the B&K theory ......................................... 19
1.2.3
Vantage theory cognitive modelling at the fringe of the paradigm. 22
1.2.4
The quest for verification outside anthropology................................. 23
An alternative position some criticism of the BCT theory.......................... 25
My own position............................................................................................. 27
Colour semantics based on texts: historical studies of colour terminology ... 29
The study of non-basic colour terms .............................................................. 31
Summary......................................................................................................... 32

Chapter 2

Providing the Tools: Meaning in Cognitive Linguistics .............. 33

2.1
2.2
2.3
2.4

Introduction .................................................................................................... 33
Principles of categorisation ............................................................................ 34
Situation, context and mental spaces .............................................................. 38
Characterising meaning .................................................................................. 41
2.4.1
Domain, network and attribute ........................................................... 42
2.5 Conceptual processes...................................................................................... 49
2.5.1
Metaphor............................................................................................. 50
2.5.2
Metonymy........................................................................................... 51
2.6 Two perspectives of meaning: semasiology and onomasiology .................... 56
2.6.1
The semasiological perspective: important sense relations ................ 57
2.6.2
The onomasiological perspective ....................................................... 59
2.7 Summary......................................................................................................... 63

Chapter 3
3.1
3.2

Colour Terms in the Bank of English: Some Statistics................ 65

Introduction .................................................................................................... 65
Word frequency .............................................................................................. 65
3.2.1
The problem of yellow........................................................................ 68
3.2.2
The proportions of colour term categories ......................................... 69
3.2.3
Frequency and age .............................................................................. 71

3.3
3.4
3.5

Colour terms as nouns .................................................................................... 73

Colour colour combinations ........................................................................ 78
Colour terms and bound morphemes ............................................................. 82
3.5.1
Derivational morphology ................................................................... 82
3.5.2
Inflectional morphology..................................................................... 89
3.6 Colour terms in different subcorpora ............................................................. 92
3.7 Summary ........................................................................................................ 95

Chapter 4
4.1
4.2
4.3
4.4
4.5

4.6
4.7
4.8

Chapter 5
5.1
5.2

5.3
5.4

5.5
5.6

Elaborate Colour Terms: Definition, Formation and Usage..... 133

Introduction .................................................................................................. 133

Defining ECTs.............................................................................................. 134
5.2.1
The standardisation approach........................................................... 134
5.2.2
Dictionary definitions....................................................................... 135
5.2.3
Summary .......................................................................................... 138
The Emergence of ECTs. ............................................................................. 140
5.3.1
The motivation ................................................................................. 140
5.3.2
The process of creation .................................................................... 142
The sense relations of ECTs......................................................................... 145
5.4.1
Polysemous relations........................................................................ 145
5.4.2
Hyponymy........................................................................................ 149
5.4.3
Synonymy......................................................................................... 150
A case study of ECT synonymy domain preference and salience ............ 150
5.5.1
The semasiological perspective........................................................ 154
5.5.2
The onomasiological perspective ..................................................... 157
Summary ...................................................................................................... 159

Chapter 6
6.1
6.2
6.3

Different Planes: Colour Terms as Adjectival Type Modifiers... 97

Introduction .................................................................................................... 97
Classifying adjectives..................................................................................... 98
4.2.1
Three classes of attribute modifiers ................................................... 99
Planes and valence relations......................................................................... 102
Compounds, phrases and lexicalisation ....................................................... 106
Type modification in two different corpora ................................................. 110
4.5.1
The OED .......................................................................................... 111
4.5.2
The BoE............................................................................................ 116
4.5.3
Summary: a comparison between the two approaches..................... 121
Decoding the process ................................................................................... 121
Wider implications another look at previous investigations ..................... 130
Summary ...................................................................................................... 131

In Search of Extra Meaning: ECTs and Markedness................ 161

Introduction .................................................................................................. 161

Markedness and specificity .......................................................................... 162
The marked use of saturation terms ............................................................. 164
6.3.1
The pattern........................................................................................ 166
6.3.2
A look at some individual examples ................................................ 170
6.3.3
Theorising the effects ....................................................................... 176
6.4 Facial colour and mood ................................................................................ 179
6.4.1
Colour terms and emotions in the BoE. ........................................... 180

6.4.2
Results .............................................................................................. 181
6.4.3
ECTs and mood ................................................................................ 184
6.5 Summary....................................................................................................... 185

Chapter 7

Figurative Use of Colour Terms................................................... 187

7.1
7.2
7.3
7.4
7.5

Introduction .................................................................................................. 187

Previous studies ............................................................................................ 189
Basic assumption .......................................................................................... 194
Figurative colour expressions and classificatory use ................................... 195
Metonymical processes................................................................................. 196
7.5.1
Part-whole metonymy....................................................................... 197
7.5.2
Part part metonymy ....................................................................... 201
7.6 Post-mapping development: local networks and domain logic .................... 206
7.6.1
Local networks and valence structure .............................................. 206
7.6.2
Domain logic .................................................................................... 210
7.7 Figurative use of ECTs ................................................................................. 212
7.8 Summary....................................................................................................... 215

Concluding Remarks .......................................................................................... 217

Prospects for further research.................................................................................. 221

Bibliography ........................................................................................................ 223

Dictionaries (with abbreviations) ............................................................................ 244
Internet Sources....................................................................................................... 244

Appendix 1. Description of the Bank of English Corpus ............................... 245

Appendix 2. Regular Expressions Eliminated via Automatic Commands... 247
Appendix 3. Frequency of Colour Terms in the Bank of English................. 251
Appendix 4. Definitions of Colour Terms ....................................................... 255
Appendix 5. English Colour Terms and their Age ......................................... 263

Typographical conventions
Linguistic forms:

Italics: dog, black

Meanings of linguistic forms:

Single quotes: dog, written record

Citations:

Double quotes:

Emphasis:

Bold face: emphasis

Cognitive categories/concepts:

Small caps: DOG, VEHICLE

(Semantic) attributes:

normal type in square brackets:

[has legs], [can swim]

Cognitive metaphors/metonymies: + signs and small caps:

+LIFE IS A JOURNEY+,
+PLACE FOR INHABITANTS+
Domains/nominal fields:

small italicised caps: DOG, CHAIR

Acknowledgements
Despite my colourful subject, I can confess that writing a thesis could be both grey
and dull at times. Fortunately, I have been privileged to meet and work with a
number of people who have reduced the number and length of those grey moments
considerably.
First I wish to express my gratitude to my supervisor, Professor Gunnar
Persson. Gunnar suggested this subject to me, and although working at a distance
of 300 kilometres, he has always been close; his help, engagement and support
have been invaluable to me. I am grateful to Professor Henryk Kardela, Maria
Curie-Skodowska University, for his very constructive suggestions for improving
this text. I also owe much of my interest in cognitive linguistics to Henryk. Another
person important for my theoretical development has been Dr Piotr Twardzisz, with
whom I have had many interesting discussions concerning Langackerian
constructs.
Special thanks to Dr Johan Nordlander and Morgan Lundberg for their
meticulous readings of many versions of my manuscript. Their comments have
forced me to clarify obscure thoughts, and sharpen my analytical tools. As true
friends they provided me with endless help in the last stage of my thesis work. I can
truthfully say that without your help, guys, I would never have completed this
thesis. Neville Shrimpton lured me into post-graduate studies and although I gave
up my initial creole project he has continued to give me moral support. Pat
Shrimptons well-trained eyes spotted many places where stylistic improvements
could be made in my manuscript. I am grateful for your kind help, Pat. Together
with Mia Svensson I have shared the frustration and laughter of trying to
understand how to draw images in a computer program. My understanding of some
theoretical concepts has deepened in spiritual discussions with Mia.
The Department of Modern Languages at Ume University is a truly inspiring
environment to work in, and the continuous encouragement I have received from
practically all my colleagues has meant a lot to me. I wish to thank Professor Raoul
Granqvist for giving me valuable comments from the point of view of cultural
studies and Assistant Professor Mark Troy, now, alas, in Karlstad, for teaching me
the importance of passion. Professor Pat Poussa, and my former teacher Berit Wik
have time and again provided me with those little words of encouragement that are
so important. On a personal level, there is a group of people whose importance to
me cannot be overestimated. The humour, warmth and friendship of Katarina
Gregersdotter, Elias Schwieler, Dr Heidi Hansson, Dr Patrik Svensson, Mats
Deutschmann, Janet French and Philip Grey mean more to me than any words can
express. In the last few hectic days in the completion of the thesis Philips
contributions in terms of language advice and support were crucial. Thanks to Gerd
Lilljegren, Christina Karlberg and Gunn-Marie Forsberg for always supplying
practical help whenever I needed it.
Many thanks are due to Magnus Olofsson for helping me with the cover, and
Per Arnqvist for explaining statistics to me.
In May of 1999, I spent a couple of weeks at the Centre for Advanced
Research in English, University of Birmingham, working with the Bank of English.
Jeremy Clear, Michelle Devereux and Dr Susan Hunston helped make my stay

extremely fruitful. Dr Alan Cruse, University of Manchester, accepted the

challenge of discussing a twenty-year old article with a soaking wet Swedish postgraduate student for several hours, and provided me with many facts of language to
consider.
For a while, I terrorised almost any English speakers I could find with
questions about their usage of terms such as turquoise, almond, cinnamon, emerald
etc. Many thanks to Van Leavenworth, Dot strm and Mary Sderstrm for
patiently answering such questions.
I am greatly indebted to Knut and Alice Wallenbergs Stiftelse for awarding
me a full-time grant for two years.
Finally, and most important of all, I wish to thank my family: my mother
Birgitta for always helping out whenever we need a babysitter or any other
assistance, and for her belief in me; my sons Hjalmar and Konrad for accepting that
their father might have to go to work yet another Saturday or Sunday; my LOVE
sa, for her unflagging support, for helping me understand statistics and, not least,
for telling me to take a break occasionally. I hope that this will mark the start of our
new life. My father, Martin Steinvall, did not live to see me complete my thesis, but
he knew that if I had any of his genes I would fight to the bitter end. I dedicate this
book to his memory. A working class hero is something to be.

Introduction
0.1 Preliminaries
This work is a contextual study of a collection of English colour terms. Here
contextual is used loosely in the sense that the colour terms are not analysed in
isolation but in conjunction with other linguistic units. The overall aim of the study
is to describe and analyse a number of aspects of the semantics of English colour
terms within the framework of cognitive linguistics. The material on which this
work is based is an extensive language corpus the Bank of English. The focus of
the study is not, however, on the numbers that can be extracted from the corpus, but
on the processes involved when the meaning of colour terms is extended from the
prototype.
It may perhaps be considered both a bold and foolish task to choose a subject
such as colour semantics, which has attracted such a great interest in the last thirty
years. Since the publication of Berlin and Kays (1969) epoch-making study of
Basic Colour Terms, few areas in semantics have been studied so thoroughly;
Maffis (1991) bibliography of colour categorisation research between 1970 and
1990 exceeds two hundred books and articles. Moreover, MacLaury (1997) claims
that since 1858 more than 3000 works have been written where colour terms are the
major issue. In the light of such numbers, one may justifiably enquire whether there
is actually anything left to uncover, especially in a language like English; a
language which is often used as a metalanguage in this particular field of research
and which, as the lingua franca of the (western) world, is perhaps more researched
than any other language. Nevertheless, my answer to such an enquiry would
definitely be yes; in the present work I hope to demonstrate that there is still room
for a study of English colour terms.
There are, in particular, three aspects of this dissertation which render it
supplementary to the existing body of research. They are intimately related,
referring to different planes, as it were, of the study. First, the method and the
material used here could be seen as suggesting a new and complementary way of
studying colour semantics. To the best of my knowledge, no one has previously
systematically used an extensive computerised language corpus for the description
of colour term usage.1 The majority of articles and books that are devoted to colour
semantics use the anthropological method of naming and mapping, i.e., they are
essentially context-free studies. The relatively few studies that are more
linguistically oriented have most often used dictionaries or introspection as their
sources. Here I show that language corpora can reveal data that will allow a more
comprehensive account to be made of the semantics of English colour terms.
A second feature of this study that complements previous research is its focus
on extensions from the prototype. Whereas the lions share of earlier work has been
devoted to prototype phenomena and the size of categories, the present study is
1

However, very late in my work with this dissertation I was made aware of a similar method being
employed by a Finnish postgraduate student, Seija Kerttula. Although we both use language corpora we
focus on different semantic issues.
1

Introduction

primarily concerned with the peripheral references of colour terms. This is actually
a consequence of the choice of method: whereas mapping and naming are excellent
tools for the study of prototypes and categorial size in the colour domain, corpus
studies, although they cannot be used to this end, can instead point to other
semantic phenomena.
The third feature of this study, making it a complement to previous research,
is its consistent use of linguistic theory for the description and analysis of the
corpus material. It would seem that much colour term theory is primarily devoted
to the cognitive and/or anthropological fields. Having said this, I should
acknowledge the inter-disciplinary character of colour term research, and I myself
also move freely among the fields, whilst maintaining a focus on linguistic aspects.
In this dissertation, I try to explain and discuss facets and patterns of colour term
usage, viewing them through the theoretical framework of cognitive linguistics.
Thus, the primary aim is not to develop this theory further but to see whether its
constructs can be fruitfully exploited in the analysis of colour term semantics. The
version of cognitive linguistics that I draw most heavily on is that proposed and
developed by Langacker (1987, 1991a, 1991b, 1999, etc.). This means that I prefer
his terms rather than similar constructs with different names suggested by other
authors. However, what is presented here is my own conception of Langackers
theory, and I have occasionally taken the liberty of slightly modifying his
terminology. My aim is always to present as coherent a descriptive model as
possible.
In this introduction, I present some explicit objectives of the study as well as
my material and research method. The chapter ends with a brief outline of the
book.

0.2 Aims
In recent years many authors dealing with colour terminology have expressed their
dissatisfaction with the fact that most studies of colour terminology have followed
similar lines. Maffi and Hardin (1997:352) discuss the merits of the type of
research performed in the tradition of Berlin and Kay (1969), but also note that
the study of whole color-term systems has the potential to be very fruitful.
Furthermore, they (p 354) observe that [a]nother point concerning the study of
whole color-systems that warrants consideration is [] the study of the
grammatical (morphological, syntactic) context of color words. Levinson
(forthcoming) makes a similar point. What some critics lack in the approach of the
Berlin and Kay school is a proper linguistic analysis of each language before interlanguage comparisons are made. As an example of what such an analysis should
include, Levinson (p 7) mentions that (among other things)
the typical use and full referential range of each expression should be investigated
without restricting oneself to a pre-set stimulus array, noting e.g. how and in what
contexts color (and other surface) contrasts are made.

It is the overall aim of the present study to respond to the needs of such a linguistic
description. Applying a corpus-based method, it is hoped that the present study can

Introduction

shed some light on specific syntactic, semantic and morphological features of

English colour terms. More specifically, a number of issues will be addressed
which relate to the extension and delimitation of colour terms in context. Many of
these issues seem to have been ignored previously, partly, no doubt, because the
applied methodology could not unravel relevant information. In the present study I
pay special attention to the following questions regarding colour terminology:

What patterns of colour term distribution can be found in a large

corpus such as the Bank of English?
Can colour terms assume different functions and thereby be used
outside their normal delimitations?
How well defined are specific colour terms and are there terms
which can be regarded as synonymous?
Can the specificity of a colour term create readings which
introduce additional meanings?
How do figurative readings of colour terms arise?
What linguistic or conceptual processes can be identified as
participating in the extension of colour terms from the prototype?

Since very little descriptive corpus work has been done on these issues, parts of this
study are mainly descriptive. However, as the final question above emphasises, one
major objective is to explain the observed patterns in a framework built on
cognitive grammar.

0.3 Method and material

0.3.1 Introduction
This section presents the type of method and approach employed in the present
work. First, I begin with a general discussion of the use of linguistic analysis in the
study of colour terminology. This is followed by brief comments concerning corpus
methodology and statistics, and the section ends with a presentation of the colour
terms under investigation, and the corpora used for the study.
Over the years a number of different methods have been developed for the
study of colour terminology. As might be expected, different research traditions
employ different methods: Work with colour arrays (naming, mapping, identifying
focal point) is a traditional anthropological method; response time, consistency and
consensus of naming, and elicited lists are often used by psychologists; frequency
in texts, morphology and word length have been part of the linguistic approach. A
good insight into the various fields of colour studies and the methodologies used is
offered by Hardin and Maffis (1997) anthology on colour categorisation.
Regarding the linguistic analysis of colour terms, the prime concern has
frequently been to correlate various patterns with the evolutionary hierarchy of
colour terms suggested by Berlin and Kay (1969). Little effort has been made to
investigate why the figures presented in their study are as they are. Nor have
researchers exploited the tools of linguistic theory to the extent that they could have

Introduction

done, in my opinion. Corbett and Davies (1995, 1997) detailed study of a variety
of linguistic and behavioural methods is a good example of this tendency. First,
their point of departure illustrates the strong position Berlin and Kays theory
occupies: the methods are tested, not the theory. Second, although they deal with an
impressively large amount of linguistic material, they do not take the analysis
beyond the surface figures. However, to do Corbett and Davies justice, it should be
added that such an analysis would lie outside the scope of their investigation, but
this is also precisely the point: the scope of linguistic studies of this type has been
defined far too narrowly.
There are, however, researchers who have ventured to find their own
analyses, and who have followed a different track in looking beyond mere
frequency. Two researchers, Isabel Forbes and C.P. Biggam, have been the
inspiration for my study. Forbes (1979, 1986) examined the relation between the
two French words for BROWN: brun and marron. Crucially, in her 1979 article,
Forbes took the investigation one step further than is usually the case by
introducing a contextual element to colour designation. By asking her informants
which colour word they would use to describe a collection of objects, she was able
to demonstrate that it was not only the nuance of the colour that determined the use
of one or the other seemingly synonymous forms, brun and marron, but also the
domain of the described object. Thus, marron was only rarely used about hair,
irrespective of the nuance, whereas brun was clearly the preferred term in this
domain. A clear pattern emerged which was also confirmed by the study of a text
corpus. What is interesting here is that neither marron nor brun can be said to be
restricted to a narrow class of objects as is English blond for example, although
there is still a clear pattern of domain preference. Needless to say, merely looking
at frequency would not have revealed this pattern. Forbes does use frequency in her
article, but then she applies a diachronic perspective to demonstrate that the
marron/brun ratio is increasing, which may be taken as a sign of an ongoing
change through which marron may succeed brun as a basic term. In her later study,
Forbes (1986) shows that there are also dialectal differences to be found in the
distribution of brun and marron.
Another notable exception in the linguistic analyses of colour terms is
Biggams (1997, 1998) work on Old English colour terms. In her meticulous
studies of colour terms for GREY and BLUE, Biggam offers an impressively detailed
analysis of the sense relations that obtain for each colour term. Among other things,
she considers collocations, referents and contrasts (see, e.g., Biggam 1997: 81-85)
to be able to draw conclusions about the extension of the word in terms of colour
nuances. Furthermore, she uses these facts to determine the hyponymous status of
the word, which, in turn, helps her decide whether a term is a Basic Colour Term2
or not. Thus, Biggams work illustrates that detailed contextual studies of a
collection of individual tokens can be useful in the overall semantic
characterisation of a given term.
It seems that the types of approach taken by Biggam and Forbes have not
been fully explored in the general study of colour terminology. In the present study
2

The notion of Basic Colour Term is described in detail in Chapter 1.

Introduction

of corpus material, I try to combine all three approaches described above; following
Corbett and Davies, I use the statistical method to be able to uncover general
patterns for further and closer investigation. However, to be able to find interesting
patterns beyond frequencies, I use Forbes domain approach and Biggams token
approach.
A corpus-based approach implies that an important aspect of this study is
what Svensson (1998: 3) called the real-data commitment: Whenever authentic
linguistic data is available it should be used, in context, cautiously and
systematically. Following this directive, I strive to ground my analytical claims in
the patterns that my corpus material provides. Occasionally, complementary
methods such as native speaker intuition, lexicographic descriptions as well as
previous psycholinguistic and anthropological research are used to add further
weight to a particular piece of evidence. However, the groundwork is essentially
confined to the limits of the chosen corpus. In this sense, this study is close to that
of Biggam; it is based on contextual material to which the investigator has indirect
access only. However, there are some marked differences between my position as a
researcher and that of Biggam. The most obvious difference is that, investigating
facets of a contemporary language, I can interview native speakers, which makes it
possible to check any claim against the intuition of the speakers. Another respect in
which this study is different from that of the historical linguist is the richness of my
material. Whereas Biggams analysis of BLUE in Old English is based on a total of
78 tokens, almost the same number of tokens figures in my study in connection
with the highly infrequent term carmine. This is not to say that big is necessarily
beautiful, but it gives a vivid illustration of the different conditions we work under.
0.3.2 The use of computerised text corpora problems and solutions
As I have already pointed out, one reason for using the method employed here is to
obtain a richer and more reliable material than is usually used in this type of
analysis. I also want to take the linguistic analysis one step further, beyond the
aspects of frequency. Having these aspirations, I had to confront a number of
methodological problems. One such problem was how to collect my material.
In an eye-opening article Ball (1994) discusses some methodological
problems concerning the use of computerised text corpora. One aspect of the
method, to which she gives prominence, is the problem of recall and precision.
Precision can be defined as the proportion of the retrieved material that is relevant.
Recall, on the other hand, measures the proportion of the relevant material that was
retrieved. To put it more simply: precision defines how much extra, irrelevant
material you get and recall covers how much relevant material you miss. Precision
is often easier to estimate than recall since it is generally very difficult to have a
good picture of what may be missed.
In my work these aspects are intimately linked to the sense relations of
homonymy and polysemy.3 It seems that most colour terms, if not all, have
homonyms or are polysemous. Basic terms such as green, black, and white have
identical forms functioning as surnames, which are homonyms. Non-basic terms,
3

The issue of polysemy and homonymy is discussed in greater detail in Chapter 2, p 57.

Introduction

on the other hand, are typically derived from flowers, metals and dyestuffs, so they
exhibit polysemous relations with other senses. The form < rose > is a case in
point. Apart from its colour sense, the form is also used as a proper name
(homonymy), a plant name (polysemy), and as past tense of the verb rise
(homonymy). Consequently, a search based on the form < rose > would yield very
low precision but perfect recall. A use of the automatic tag for adjectives, on the
other hand, would dramatically increase the precision, but would lower the recall.
All nominal uses of the colour term rose, as in (1) below would be excluded.
(1) Have you done my old room in rose, as you planned it? asked Channing as
offhandedly as she could. (BoE: usbooks)

Ball (1994: 296) makes a case for favouring poor precision: To read a corpus is to
encounter the unexpected and the undocumented at every turn. [] [W]ith perfect
precision, we find exactly what we said we were looking for, and no more. I have
made a point of following Balls advice in this respect: when searching the corpus I
did not use the tag option. Accordingly, my initial material encompassed perfect
recall but low precision. However, the precision could be increased with the help of
available computer commands. I eliminated the proper name problem by excluding
all forms that contain capitals. This necessarily led to a decrease in the recall, but it
would seem that nominal phrases containing colour words typically contain some
article, and therefore I believe that this adjustment should not have markedly
affected the reliability of the material. When it comes to highly frequent forms of a
polysemous or homonymous character like rose, cream and navy, a grep-v
command4 was used to remove the most frequent collocations containing a noncolour sense (e.g. double cream). The coordinated phrases that were automatically
eliminated with this command can be found in Appendix 2. The final refinement of
the material, however, was conducted manually, where the individual sentences
were examined to determine whether the term had a colour sense. Even here
mistakes are bound have to occurred,5 but given the richness of the material, I feel
confident that any such mistakes should not affect the general conclusions of this
work.
0.3.3 Statistics
Since it is unrealistic to analyse each individual example, many researchers in
view of the ever-increasing practice of using large text corpora see the need to
employ statistical tools. On the other hand, it might be argued that figures by
themselves have little explanatory value, and they have to be examined and
problematized further by the human analyst. Furthermore, as observed by Ball
(1994), it seems that statistics lends itself best to certain types of investigation. As
an example of the type of problem that can occur when statistics and analysis are
not really compatible, Ball mentions hidden variables. The problem is that the
4

Grep is a UNIX command for selecting lines matching a regular expression, on which the lookup
commands g and v are based.
5
One such difficulty even by hand is to determine whether a form refers to dyestuffs or colour. In cases
of doubt I have preferred colour classification.

Introduction

relative frequency of a linguistic phenomenon has to be calculated in relation to the

number of possibilities of occurrence, not in relation to the number of words. Ball
also calls attention to the temptation that the text analysis tools of the software
package represent. These programmes are usually designed to process word level
phenomena and in particular various collocational patterns.
In the present study, the majority of statistical calculations are of the simple
and well-established type, such as chi-square goodness of fit, various correlation
calculations and collocational strength.6 As far as the last one is concerned, I used
the software available in the Bank of English. Occasionally, I used other types of
statistical tools because of the character of the material, and on these occasions I
give a short description of the tool.
In my use of collocational data, I have been influenced by Stubbs (1995)
critical discussion of different ways of calculating collocational strength. There are
three relatively well-established measures: raw frequency, t-score and MI-score.
Stubbs argues that naive use of these measures can be questioned since it is not
entirely clear what the last two actually measure. Underlying t-score and MI-score
is the assumption that the expected frequency and the measured frequency can be
contrasted. The problem is that the expected frequency is calculated as if words
occur totally randomly in a corpus, which, of course, is not the case. Church et al.
(1991) discuss the usefulness of these tools, giving a thorough description of how
the t-score equation is derived. Despite the reservations that can be made as to what
exactly the numeric value of these methods indicate, we should acknowledge their
usefulness in the identification of collocations which appear more often than can be
attributed to chance.
The approach I use in my work in this context is similar to the functional
account suggested by Stubbs (1995: 40-41):
They [MI-score and t-score; Stubbs uses the terms I and T-value] can be used as
filters which catch collocates likely to be of linguistic interest. [---] The cases which
survive the filters provide a set of words, based on solid quantitative evidence, for
further human interpretation.

In my study, I have occasionally used collocational calculations as filters, but at no

point are the calculations seen as anything more than tools for identifying
potentially interesting areas. Another researcher using a similar method is Deignan
(1999a, 1999b). She used collocational patterns as a means of identifying relevant
senses in her analysis of linguistic metaphors in corpus data.
To sum up, statistical calculations are useful tools when the analyst has to
process a large amount of data. However, because at the moment statistical
methods present many pitfalls, these tools should best be treated solely as filters for
identifying potential research areas.
0.3.4 The terms
One of the aims of the present study is to approach English colour terms without
any preconceived idea of basicness. Given this, the choice of colour terms to be
6

I have used Woods et al. (1986) and Zar (1996) as my guides in finding suitable statistical tools.

Introduction

investigated is of great importance all the more so since the number of colour
terms in English, as in any other language, is truly gigantic. The human eye can
distinguish millions of shades7 and theoretically there could be the same number of
corresponding colour terms. In 1930, Maerz and Paul listed 4000 English colour
terms (Wyler 1992), and the number has undoubtedly increased since then. It
would therefore be an impossible task to analyse all terms, and some sort of
sampling has to be made. This study restricts the number of terms to 50 basic
forms including their derivative forms (cf. Chapter 3.5).
In establishing the sample, I followed the assumption that salient terms are
probably more interesting than other terms. Thus, rather than taking a random
sample of terms my aim was to find the most salient ones. One hypothesis
concerning salience was suggested by Berlin and Kay (1969: 6), who remarked that
psychological salience could be indicated by a tendency to occur at the beginning
of elicited lists of colour terms. The above-mentioned studies by Corbett and
Davies (1995, 1997) seem to confirm this hypothesis.
In the light of these considerations, I have based my sample of terms on three
studies: Battig and Montague (1969), Brown (1972), and Taft and Sivik (1997).
Included in my study are those terms in Brown (1972: 139) and Battig and
Montague (1969: 10) with a response frequency of more than ten, and all the terms
listed in Taft and Sivik (1997: 31).8 I hope that this procedure will ensure that the
terms under investigation are among the most salient in British English and
American English.
In Battig and Montagues study, performed in 1965, fifty-six categories were
investigated, and the responses were given by students from the Universities of
Maryland (270 subjects) and Illinois (172 subjects). The students were presented
with notebooks and were asked to write down, in 30 seconds, as many items as they
could for the specified category. Twenty-six colour terms got a higher score than
ten. Browns study was also carried out in 1965, asking 200 Scottish
undergraduates, 100 women and 100 men, to write down items in 28 categories in a
booklet. The students were given one minute to write down as many items as they
could for each category. Thirty-six colour terms got a score higher than ten. The
similarity between the Scottish and the American tests of course invites a crossexamination, and one such examination was performed by Brown (1978). He found
that there was substantial agreement between the American and Scottish studies in
the field of colour terms, both as regards frequency and serial position.9 Neither
study was performed in relation to other types of colour studies.
Taft and Siviks (1997) study is somewhat different from those previously
mentioned. The only task that the respondents were asked to do was to write down
as many colour terms as possible, in the order they thought of them. They were
given one piece of paper with 50 numbered spaces and were told that if they needed
7

Taft (1997: 11), referring to Nickerson and Newhall (1943), claims that we can distinguish among as many
as seven million differences in the psychological color solid.
8
Taft and Siviks list only includes terms which scored above 50% (i.e. more than 50% of the subjects
mentioned the term).
9
Serial position or rank order means the order in which the terms were listed. For example, in the case of
colour, the terms red, blue, green and yellow had the lowest mean serial position in Brown (1972), i.e. they
usually occurred among the first five colour terms listed.

Introduction

more space they could use the back side of the sheet. Furthermore, there was no
time limit given for the accomplishment of the task. This made it possible for the
subjects to focus more effectively on this one task. The study is also different from
the previous ones in that the group was composed of people of different ages,
between 17 and 70 (the mean age of the Americans was 36). All in all, 31 colour
terms were mentioned by more than 50% of the subjects.
It should be mentioned that I have included one wild-card colour term, azure.
The inclusion of this colour term, which did not reach the critical number in any of
the above studies, was based on its semantic quality as a saturation term and as
such it plays a role in one of my case studies of specificity (Chapter 6). The colour
terms considered in this work are listed in Table 0:1 together with an indication of
the study they appeared in.
Table 0:1. Colour terms under investigation
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.

Amber
Aqua(marine)10
Azure
Beige
Black
Blue
Brown
Carmine
Charcoal (grey)
Chartreuse
Cream
Crimson
Emerald
Fawn
Fuchsia
Gold
Golden11
Green
Grey/Gray
Indigo
Lavender
Lemon
Lilac
Lime
Magenta

B
B&M, T&S
*
B&M, B, T&S
B&M, B, T&S
B&M, B, T&S
B&M, B, T&S
B
T&S
B&M, T&S
B
B
B
B
T&S
B&M, B, T&S
B&M, B, T&S
B&M, B, T&S
B&M, B, T&S
B&M, B
B&M, T&S
B
B
B, T&S
B&M, B, T&S

26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.

Maroon
Mauve
Mustard
Navy (blue)
Olive
Orange
Peach
Pink
Plum
Puce
Purple
Red
Rose
Rust
Scarlet
Silver
Tan
Tangerine
Teal
Turquoise
Ultramarine
Vermilion
Violet
White
Yellow

B&M, B, T&S
B&M, B
B
B, T&S
B&M
B&M, B, T&S
T&S
B&M, B, T&S
T&S
B
B&M, B, T&S
B&M, B, T&S
B&M, T&S
T&S
B
B&M, B, T&S
B&M, T&S
B
T&S
B&M, B, T&S
B
B
B&M, B, T&S
B&M, B, T&S
B&M, B, T&S

There are a few things to be noted in connection with this table. First, although they
are fairly frequent terms in a text corpus, highly domain-specific colour terms such
as blond and auburn did not reach the critical score in any list. Second, no
derivation of the type rosy or pinkish was mentioned. Neither possibility was
excluded by the formal instructions concerning the listing tasks. Consequently, it is
tempting to conclude that these terms are not considered to be proper colour terms
10

Both aqua and aquamarine are taken into consideration whenever the form aqua(marine) is used.
Gold and golden are given separate entries since they do not behave similarly in terms of syntax.
Semantically they are essentially similar of course. However, the difficulty of isolating the colour sense of
gold means that it is mainly golden that has been taken into account.
11

10

Introduction

by a majority of English speakers. Nor were luminosity terms such as dark or

light mentioned. Arguably, this is because they are not felt to be colour terms at all,
although from the point of view of reference, technically speaking, they are closely
related to the established colour terms black and white.
0.3.5 Corpora
The data used in my study come from the Bank of English corpus (henceforth
BoE). The BoE is owned by Collins COBUILD, a division of HarperCollins
Publishers, and is presently kept at the University of Birmingham, England. The
corpus, which can perhaps best be characterised as a monitor corpus (cf. Sinclair
1991: 24-26), contained approximately 323 million words at the time of my initial
sampling, February to June 1998, and it was one of the largest computerised
corpora of English in the world, if not the largest. Since then the corpus has
increased in size; with the latest addition in January 2002, it now contains some
450 million words.12 Three varieties of English are represented: British English
(70%), American English (20%) and Australian English (10%).13 The BoE contains
a number of subcorpora, representing different text genres. A complete list of the
subcorpora (with abbreviations) and their size is given in Appendix 1. It should be
noted, however, that the corpus could be said to be unbalanced in the sense that
most subcorpora represent written varieties, among which newspaper language
seems to dominate. It should also be emphasised that within some subcorpora there
are groups of genres which are not controlled and which may display a great
stylistic span; for example, in a given newspaper corpus we find both sports pages
and editorials, presumably representing different registers14 and styles.15 Generally,
I do not take genre into consideration in my analyses.
Following Clear (1992), we can say that the BoE corpus more accurately
Although the
represents language reception16 than language production.
characteristics of the corpus may affect aspects of my study, this should only be in
a minor way given the nature of my analysis.
The reason for choosing the BoE rather than any other corpus, such as the
British National Corpus for example, was largely because of its size. As one of my
aims of this study was to describe and analyse the use of non-basic colour terms, I
found that size should be regarded as of prime interest: unless a reasonable number
of tokens could be found, the reliability of the analysis would be questionable. Thus
I agree with Clear (1992: 30), who maintains that [t]he importance of the overall
size of the corpus upon which claims about a language are based cannot be ignored
more is definitely better. Additional factors favouring the BoE rather than the
BNC were its inclusion of American material and the fact that the software
12

This is of course slightly unfortunate, from the point of view of my research. Because the content of some
subcorpora has been altered, my study cannot be copied. However, as far as it has been possible, I have
chosen examples which are still available in the new corpus.
13
These figures represent the proportion at the time of my data collection.
14
The validity of this term is frequently discussed. Here it is used loosely.
15
Cf. Deignan (1999b) for an insightful discussion in greater detail concerning this fact.
16
Clear (1992: 24) suggests that this perspective is reflected by the question What is the likelihood that a
native speaker has encountered this word recently? An alternative label could be language exposure.

Introduction

11

available facilitated the collection of the material. Where the use of English colour
terms is compared with those in Swedish, the Swedish examples were retrieved
from the Swedish language corpus Sprkbanken17 in Gothenburg.

0.4 On the terminology of the dimensions of colour

Most speakers of a language are aware that there are different qualities or
dimensions intrinsic to the concept of colour. First, there is colour itself we can
distinguish red from yellow, for instance. But the fact is that we often qualify
colour terms to account for the other dimensions. Thus, we can talk about a dark
red but also a dull red; a light red but also a bright red. The above phrases
reflect a three-way distinction which is traditionally made between hue, lightness,
and saturation. These can be illustrated by the three-dimensional colour solid (cf.
e.g., Boynton 1997 and Sivik 1997 for presentations of two different systems).
Although it is not uncontroversial to reduce colour to these three dimensions (cf.
section 1.2), I think one has to admit that the colour terms of English can readily be
analysed in this way. Even so, it is clear, and this will be demonstrated later on, that
other dimensions can be evoked through context in English. However, for the
moment let us confine the discussion to these three dimensions. As this is an
entirely corpus-based study, it is important to realise how these dimensions are
coded linguistically, and, thus, address the question of appropriate terminology.
The terminology is particularly relevant to the discussion of the semantics of more
precise colour terms (Chapter 5).
Consider first the term hue, normally associated with colour. Munsell (1961:
15) defines hue as follows:
It is that quality by which we distinguish one color family from another, as red from
yellow, green from blue or purple. It is specifically and technically that distinctive
quality of coloring in an object or on a surface; the respect in which red, yellow,
green, blue, and purple differ from one another; that quality in which colours of equal
luminosity and chroma may differ. [italics original]

Since hue corresponds to what we normally associate with colour, modification of

a colour in respect to hue is usually achieved with the help of another colour term,
often a derived form, e.g. reddish purple, or yellowish green.
A second dimension of colour is lightness. Lightness is the quality by which
we distinguish a light color from a dark one (Munsell 1961: 15, italics original).
There is a wealth of terms currently used to refer to this dimension; an unfortunate
circumstance which may cause confusion. Frequently used terms are value (e.g.
Munsell 1961), lightness (e.g. Maffi and Hardin 1997, Kay 1999a), brightness (e.g.
Berlin and Kay 1969, MacLaury (1992)) and tone (e.g. Biggam (1997)). As
Biggam (1997) observes, the use of the term brightness to label this dimension is
particularly unfortunate since the term bright as in, say, bright red in ordinary
language, in fact denotes the third dimension saturation. This is a potential source
of ambiguity. Throughout this book therefore, I use the term lightness to indicate
17

Sprkbanken can be accessed via the Internet at: http://spraakbanken.gu.se/

12

Introduction

this dimension. The term brightness only occurs in quotations from other authors.
In everyday language emphasis on lightness is usually indicated by the terms dark
and pale.
The third dimension of colour is that of saturation or chroma. It is the
quality of a color by which we distinguish a strong colour from a weak one; the
degree of departure of a color sensation from that of white or gray; the intensity of a
distinctive Hue; color intensity. (Munsell, 1961: 16)

As the term saturation is rather uncontroversial and usually preferred, it is used in

this study. Dull or greyish and vivid or bright are used in every-day speech to refer
to low and high saturation, respectively.

0.5 Overview of the book

The book falls into two major sections: the first part, containing two chapters,
provides a general background picture; in the second part, Chapters 3-6 deal with
aspects of the literal use of colour terms, whereas Chapter 7 deals exclusively with
the figurative use of colour terms.
Chapter 1 provides a comprehensive overview of the study of colour
semantics since 1969. The development within the dominant theoretical framework
as well as critical voices are discussed, and my own position with regard to the
various theories is presented.
Chapter 2 contains a short description of some basic notions within the
framework of cognitive linguistics. This chapter is mainly aimed at the uninitiated
reader, who hopefully will be able to follow the reasoning in the later chapters with
the help of this presentation. Some crucial concepts from the Langackerian
framework are also presented and discussed.
Chapter 3 gives a comprehensive view of the frequencies of the colour terms
under investigation in the BoE. Some general aspects of morphology and minisyntax of colour terms are treated mainly on numerical bases; thus the chapter is to
a large extent descriptive.
Chapter 4 deals with the issue of extension of colour terms outside their
normal area of reference. It is argued that the majority of these instances can be
attributed to the classificatory function of the colour terms in these instances. A
case study provides a detailed theoretical account of the processes involved from
the point of view of cognitive grammar.
Chapter 5 gives an account of the non-basic colour terms, here called
elaborate colour terms. Questions of transparency and polysemy are discussed
and the alleged synonymy and free variation of some terms are investigated in the
corpus.
Chapter 6 discusses how the use of specific terms can highlight senses outside
the colour domain. This phenomenon is discussed with reference to the linguistic
construct of markedness, which is analysed within the framework of cognitive
linguistics.

Introduction

13

Chapter 7 examines the types of processes involved in the formation of

figurative senses of colour terms. Instances of figurative use found in the corpus are
discussed in relation to the theoretical account.
The Conclusion discusses how the results obtained in this work may affect the
interpretation of previous research. Structures of the two categories of colour terms
identified here, BCTs and ECTs, are suggested on the basis of usage patterns.

Chapter 1 Setting the Scene:

Thirty Years of Studies of Colour Terminology
1.1 Introduction
This chapter surveys the literature on colour term semantics in order to create a
suitable frame for the present study. As mentioned in the Introduction, the amount
of written material that deals with colour terminology is so large that a full account
would constitute a book in itself. For that reason, the survey is necessarily selective
and only the most influential books and articles are presented.1 The chapter also
examines some different methods and approaches that have been used in the study
of colour terminology.
The time frame in the survey is confined to literature produced after 1969, the
publication date of Basic Color Terms. This is an obvious starting point since every
piece of work in colour semantics after the publication of this book has been
forced, in one way or another, to relate to the results presented there. Broadly
speaking, the research after 1969 can be divided into two categories.

Research accepting the general basis of the Berlin and Kay theory.
This research typically tests the hypothesis on new material or
applies new methods to confirm the theory.
Research rejecting the whole premise of the Berlin and Kay theory.

The first category is considerably larger and is given much more space in my work.
This dichotomy structures the present chapter in that the first part of the
chapter is concerned with research within the frame of the Basic Colour Theory.
Special emphasis is given to the work of the original authors (especially Paul Kay)
since they seem to be the ones moving the theoretical framework forward. After
this, the main critique posed by the opposing camp is presented. Subsequently, I try
to formulate my own position in relation to the two categories. The final part of the
chapter briefly surveys the historical research of colour terminology and points to
similarities to and differences from the present work. The present chapter does not
include any references to previous studies of figurative use of colour terms; these
are treated in Chapter 7.

1.2 The dominant theory: Berlin and Kays legacy

1.2.1 Basic Color Terms
The privileged position of Basic Color Terms should become clear if we describe
the prevailing doctrine at the time of its publication. A cornerstone of postSaussurean linguistics is that of the arbitrary nature of meaning. This idea had its
1

The Bibliography, however, is fairly comprehensive as regards recent materials.

15

16

Chapter 1

sequel in the doctrine of relativism, the prime example of which is the SapirWhorf hypothesis. In linguistics, it was long argued that the domain of colour was
the area par excellence for illustration of the arbitrary nature of language.
Bloomfields (1933:140) statement, cited below, was a commonplace of the time as
we find similar claims made by other authors, including Gleason (1962) in his
classic textbook.2
Physicists view the color spectrum as a continuous scale of light waves of different
lengths, ranging from 40 to 72 hundred-thousandths of a millimeter, but languages
mark off different parts of this scale quite arbitrarily and without precise limits, in the
meaning of such color-names as violet, blue, green, yellow, orange, red and the color
names of different languages do not embrace the same gradations.

The view expressed by Bloomfield was widely accepted in linguistics for many
decades. Studies performed in this tradition are, for example, Conklin (1964) and
Lenneberg and Roberts (1956).
It was in fact the aim of Berlin and Kay (1969) (henceforth B&K) to refute
this belief in relativity. Despite many studies demonstrating that the colour
spectrum was divided in various ways, Berlin and Kay hypothesised that there was
an underlying pattern of universality. They wrote (1969:2): We suspect that this
allegation of total arbitrariness in the way languages segment the color space is a
gross overstatement. In order to demonstrate this, they designed an experiment
based on a number of assumptions which were given axiomatic status.

Not all colour terms have equal status; there are basic colour terms
(henceforth BCTs), and such a term can be defined by a few criteria. These
criteria are: a) it is monolexemic, i.e. the meaning is not predictable from its
parts; b) its signification is not included in that of any other colour term; c) its
application is not restricted to a narrow class of objects; d) it must be
psychologically salient for the informants e.g. show a tendency to occur
early in elicited lists, stability of reference across informants and occasions of
use, and occurrence in the idiolects of all informants.3 (B&K: 6)

There is an internal gradation in colour categories, which means that it is

possible to isolate a focus of the category a best example.

The denotatum of a colour term is fully described by the three dimensions

hue, lightness and saturation.

Although not explicitly stated, these axioms were in fact already part of an
established tradition of studies in colour terminology. As far as the first axiom is
concerned, B&K pointed out that the notion of basic colour terms had long been a
working assumption in linguistics and anthropology, although not properly defined.
2

For more evidence and references, see Berlin and Kay, 1969: 159-160 (footnote 1), and MacLaury 1997:
18-20.
3
There are four additional criteria which can be used for doubtful cases, cf. B&K (1969: 6)

Setting the Scene: Studies of Colour Terminology

17

The notion appeared in the work of Lenneberg and his associates (e.g. Brown and
Lenneberg 1954; Lenneberg and Roberts 1956) and also in the work of Conklin
(1964). Consequently, as they saw it, their achievement in this respect was only to
give it a unique operational definition. Regarding the second axiom, Lenneberg
and Roberts (1956: 16) make the following observation about the existence of a
focus of colour categories:
[I]t appears that some stimuli are more likely to elicit a given verbal response than
others. One can think, for example, of a color which is more typically yellow than
another [] We shall use the term focus to describe that cluster of stimuli which has
an extremely high probability of eliciting one distinct verbal response.

Concerning the third axiom, B&K were well aware that Conklin (1964), for
instance, had demonstrated that there were other dimensions of meaning associated
with colour terms, e.g., succulence and desiccation. However, in their quest for the
underlying pattern they argued (p 160, footnote 2) that associations of that kind
would have to be ignored. Again, this approach was anticipated by Lenneberg and
Roberts (1956:14) who make similar reservations:
Since we are proposing to use the three perceptual dimensions as metalanguage in
terms of which the referent of any color term can be described, we might pause to ask
to what extent the coordinate system arising from the use of these dimensions might
itself be culture bound. Two questions are involved: the first is whether the
dimensions (hue, brightness, and saturation) are universally applicable []. For the
time being, we may leave the first question unanswered, recognizing, of course, that
there is nothing natural, logical, or necessary in these dimensions. For our purposes,
they are measuring sticks which can be conveniently used in describing cross-cultural
similarities or differences.

Although Lenneberg and Roberts suggest that there is nothing natural in the
dimensions, they have been treated exactly in that way since then. B&K (1969:160,
footnote 2), for instance, refer to them as the three psychophysical dimensions:
hue, saturation, and brightness. Today, most researchers seem to treat these
dimensions as natural and universal dimensions grounded in physiological
perception.
On the basis of the above-mentioned axiomatic assumptions, B&K asked
native speakers of twenty languages to map the focal points and outer boundaries of
each BCT in their language on an array of 329 Munsell colour chips. This
procedure and material had been used in an almost identical form in another classic
study, the aforementioned Lenneberg and Roberts (1956). They, however, had used
a much larger inventory of colour terms (52 terms) and did not use achromatic
chips (cf. B&K, p 103.). Subsequent to the empirical study and the analysis of the
results, a third, heavily criticised, step in the study was added a study of the
literature concerning previously documented colour studies. On the basis of such
studies, combined with studies of various types of lexicons and dictionaries, B&K
established the BCTs of an additional 78 languages, reaching a total of 98
languages.

18

Chapter 1

In sum, it can be claimed that B&Ks achievement was not based on novel
methodology, nor on the postulation of radical axioms. Their study was quite
traditional, in these respects, but what was actually new was the wide contrastive
approach and, in particular, the equation of colour category to focus. B&K (1969:
13) write: [W]henever we speak of colour categories, we refer to the foci of
categories, rather than to their boundaries or total area [italics original]. Even
though the focal element had been recognised earlier no one had previously taken
this radical step.
The results of their investigation were revolutionary in at least three respects.
First, B&K succeeded in confirming their hypothesis; they claim that they have
been able to show that color categorisation is not random and the foci of
basic color terms are similar in all languages. (1969: 10)
Second, which came as a surprise to B&K, there appears to be a universal
process, which stipulates a fixed order in which BCTs are encoded in a
language. B&K suggested that this process could be reduced to a typology of
seven stages (cf. Figure 1:1 below). They observed that of a possible 2,048
(i.e. 211) patterns only twenty-two combinations occurred in their material.4
Third, apart from some doubtful cases (notably, Hungarian and Russian) no
language seems to have more than eleven BCTs.
I

II

WHITE

and
BLACK

III
GREEN

RED

IV

VI

GREEN

or

and

YELLOW

YELLOW

BLUE

BROWN

VII
PURPLE
PINK
ORANGE
GREY

Figure 1:1. Berlin and Kays 1969 typology of seven stages.

Berlin and Kays (1969: 104) own summary of these results provides us further
indication of what they claimed to have found:
Our research to date points to three main conclusions. First, there exist universally for
humans eleven basic perceptual color categories, which serve as the psychophysical
referents of the eleven or fewer basic color terms in any language. Second, in the
history of a given language, encoding of perceptual categories into basic color terms
follows a fixed partial order. [] Third, the overall temporal order is properly
considered an evolutionary one; color lexicons with few terms tend to occur in
association with relatively simple cultures and simple technologies, while color
lexicons with many terms tend to occur in association with complex cultures and
complex technologies. [Emphasis added]

It is important to understand that, for instance, RED in Figure 1:1 indicates that the focal colour of this third
colour term is red. A stage II language with only three BCTs will have one term whose focus is red, but
whose extension, of course, is much wider than that of English red.

Setting the Scene: Studies of Colour Terminology

19

As MacLaury (1997: 20-21) points out in his summary of Basic Color Terms, the
view of the authors would seem to be that BCTs are universal because neurology
determines that there is no other way to see. The use of the terms psychophysical
and perceptual category above indicates this. In this view, the illusion of crosslinguistic relativity is created through the different number of BCTs, which
frequently lead to differences among boundaries. Other elements adding to the
perceived picture of relativism are the presence of non-basic terms and
connotations. Instead we have a theory which could be given the label evolutionary
determinism, since it suggests (cf. the quote above) that there are universal
perceptual colour categories just waiting to be lexicalised.
1.2.2 The later development of the B&K theory
As mentioned earlier, the publication of Basic Color Terms caused a considerable
stir among linguists and anthropologists. Generally, the theory received positive
reviews despite its revolutionary content. A noteworthy spin-off effect of the book
is that, ever since the publication of Basic Color Terms, the study of the colour
field has become the scene of interdisciplinary activity including anthropology,
behavioural psychology, linguistics, and neurophysiology.5 In this section I briefly
discuss the developments that have taken place inside the B&K paradigm,6 as a
result of new facts, criticism and speculation.7
In my description of the development of the B&K universalist paradigm, I
start with the main stream, which can be said to be represented by an
anthropological method largely in line with the landmark work. Although the
majority of the reviews were positive, there were also a few reviewers who
criticised different aspects of the study (e.g. Hickerson 1971, Durbin 1972, McNeill
1972 and Conklin 1973).8 In response to some new facts and to problems already
addressed in Basic Color Terms (the evolution of GREY, cf. B&K 1969: p 41-45),
Kay (1975) suggested some modifications: GREY should have a wild-card status,
meaning that it may occur earlier than was proposed in the original sequence (cf.
Figure 1:1). Furthermore, a few of the earliest colour categories were renamed to
avoid ambiguity: the category GREEN-and-BLUE was called GRUE and it was
acknowledged that this category might have its focus in either GREEN or BLUE.
Another important contribution was Kays observation that individual speakers in a
language community could have a different number of BCTs.
Even in Basic Color Terms, B&K observed that contemporary formal theories
of lexical definitions could not handle the structure of colour categories, as the
structures were perceived in their work. A very important theoretical step in the
development of the universalist theory was Kay and McDaniels (1978) attempt to
5

However, in the American tradition the affinity between linguistics and anthropology has always been
close, cf. the work of such people as Boas and Sapir.
6
I use the term paradigm here in a loose Kuhnian way, but this use should not be taken as programmatic.
For an interesting and illuminating discussion concerning prototype theory as a paradigm in Kuhns (1962)
sense, see Geeraerts 1997.
7
MacLaury (1997: Chapter 2) provides an excellent account of the history of colour ethnography from 1858
to the present. He also (468-469, footnote 16) provides a detailed list of studies 1970-1985 dealing with
various related subjects.
8
Sampson (1980: 95-102) provides a good summary of early critique, and points to some questionable
details in B&Ks methodology.

20

Chapter 1

model an explanation of the facts observed using another framework. Drawing on

recent work on colour perception (De Valois et al. 1966) and mathematical
formalism, fuzzy set theory (Zadeh 1965), they tried to show that the evolutionary
sequence of BCTs could be successfully modelled if fuzzy set logic was applied to
opponent neural responses. According to De Valois and his associates, there are six
types of cells, four of which determine hue and two of which determine lightness.
The four cells which determine hue are called opponent response cells. According
to the theory, the cells are organised in two pairs: Blue-Yellow and Green-Red.
Kay and McDaniel argued that all basic colour categories can be seen as being
formed from the human visual systems six fundamental neural response categories
(FNRs) by one of three fuzzy-logical operations:
Identity: green (= neural response GREEN)
Fuzzy union: grue (GREEN OR BLUE)
Fuzzy intersection: orange (RED+YELLOW)
From this perspective it was possible to view the entire sequence as containing two
movements:
i)
ii)

The progressive splitting of composite categories (such as macro-RED

and GRUE) into neural categories: RED, GREEN, YELLOW and BLUE.
The formation of derived categories from the neural ones. (i.e. the
formation of BROWN, PINK and PURPLE)

Although the biological aspect of colour categorisation was mentioned in Basic

Color Terms, it was Kay and McDaniels article that made an attempt to formulate
a fully fledged theory of neurophysiological reductionism.9
Interestingly enough, Kay and McDaniels account accords BCTs unequal
status: some BCTs, namely those corresponding to neural categories, seem to be
more basic than the other colour terms. They are usually referred to as Primary
Basic Terms (cf. for example Corbett and Davies 1995). Another aspect of Kay
and McDaniels theory is that it does not restrict the number of BCTs to eleven.
Kay and McDaniel note that there is room for at least fifteen BCTs: apart from the
six Primary Basic Terms there are nine potential terms (Secondary Basic Terms)
which can be modelled through fuzzy intersection for instance, maroon (BLACK
+ RED) could very well develop into a BCT.
In response to the criticism of the data in Basic Color Terms, Kay, Berlin and
Merrifield started the World Colour Survey (hereafter WCS) in 1976, whose main
objectives were twofold (Kay, Berlin, Maffi and Merrifield 1997):
i)
ii)

To collect reliable data from a large variety of languages of the world.

To enhance knowledge of universals, variation and historical
development in the field of colour terminology.

9
Other studies taking this neurological predisposition approach are Ember (1978), von Wattenwyl and
Zollinger (1979) and Lumsden (1985).

Setting the Scene: Studies of Colour Terminology

21

The WCS, but also other independent surveys (notably the work of MacLaury, see
below) found some unexpected data, especially concerning the early categories.
Kay, Berlin and Merrifield (1991), when summarising material from the WCS,
devote their attention primarily to the early stages of the evolutionary sequence. To
meet new data, the original evolutionary sequence was considerably elaborated, but
the authors claimed that the original tenets were still valid:
The wider range of data in the WCS has confirmed the main lines of the original
hypotheses of Berlin and Kay regarding the existence of semantic universals in basic
color lexicons and a system of partial constraints on the evolutionary development of
basic color vocabularies. [---] More broadly, we may observe that the cultural
construction of simple categories is in several respects constrained by biology. [---]
[I]t is clear that the color system of a given language at a certain moment can assume
a large number of forms while remaining within the limits set by biology.
(Kay, Berlin and Merrifield, 1991: 23-24)

What should be stressed is that, in accordance with Kay and McDaniels (1978)
speculations, the constraints on the evolutionary development of colour lexicon are
believed to be predetermined by biology. However, as the last line of the quote
above illustrates, the authors have moderated their claim considerably compared to
that of Kay and McDaniel. In particular it appears that the frames set by biology
allow a greater freedom to construct alternative colour categories.
The B&K model has been reformulated one more time, and, as was the case
in Kay, Berlin and Merrifields (1991) article, it is the early categories which have
attracted interest. In direct response to data collected by Levinson (forthcoming)
but also in response to an alternative theory of colour term evolution, the
Emergence Hypothesis (see below), Kay and Maffi (1999) presented a new model
whose basis is still firmly placed within the paradigm. The model introduces four
principles of colour categorisation, which account for the different patterns of
categorisation that can be found in different languages up to Stage V (see Figure
1:1 above) 10. The four principles suggested by Kay and Maffi (1999: 745) are the
following:
(0) Partition
(1) Black & White
(2) Warm and Cold
(3) Red
The first principle, Partition, is said to be language based; that is, it constitutes a
generalisation of a tendency in language use to partition salient notional domains
lexically. A similar idea is expressed by Clark (1992:371) under the name the
exhaustive constraint. The other three principles pertain solely to the colour
domain, and are based on Kay and Maffis analysis of available data. The
principles reflect the tendencies in languages to partition certain categories early.
On the basis of these four principles, Kay and Maffi can elegantly model, and thus
10
It is at Stage V all Primary categories have been established according to the original B&K theory, see
above.

22

Chapter 1

resolve, the origin of unorthodox categories such as YELLOW/GREEN/BLUE and

YELLOW/GREEN, which could not previously be accommodated into the standard
theory. Kay and Maffi demonstrate that, in terms of evolution, we can distinguish
between a number of different trajectories reflecting different orders in which
these principles have been applied. The most common model is the application of
these principles in the order of their indexing presented above. However, different
application will yield other, slightly unorthodox, colour categories which have been
recorded in some languages. The Emergence Hypothesis states that principle (0)
may not be universally first, i.e. there are languages which do not partition the
entire colour domain. Levinson (forthcoming) demonstrates that this is the case in
Yl Dnye.11
It would seem, then, that the B&K model, although challenged by a great
amount of empirical data, can still accommodate most new facts and many of its
basic assumptions are still treated as valid, including
1) the elemental nature of four hue and two achromatic sensations, and
2) the existence of an evolutionary process.
At the present time, there does not seem to be any evidence which may threaten to
overturn the rule of the B&K paradigm.
1.2.3 Vantage theory cognitive modelling at the fringe of the paradigm
During the past fifteen years another, alternative, model of colour categorisation
Vantage theory12 has emerged as a result of material collected by Robert
MacLaury in the Mesoamerican Color Survey (hereafter MCS). MacLaury was the
first researcher to identify empirically exotic categories which did not comply
with the original B&K sequence. In a number of articles (e.g. 1987, 1992), he
reported patterns previously unrecorded, notably a YELLOW-WITH-GREEN category.
Although distinctly different, vantage theory still adheres to the basic assumptions
laid down by the B&K paradigm, and this explains why I place it here in my
account: He postulates the existence of BCTs (cf. MacLaury 1997: Appendix IV,
419-429), an elemental colour domain which can be studied in isolation, and claims
that colour categorisation is constrained by the physiology of colour perception.13
It should also be mentioned that MacLaury has contributed to the
improvement of the methodology used in eliciting colour terms. Aside from
establishing the focus of a category and naming individual chips, he asks his
subjects to map the range of a specific colour term onto a chart.14 Furthermore,
MacLaury avoids the use of a preconceived notion of BCTs; the subjects are
allowed to use any terms they like in the naming task. These two methodological

11

Yl Dnye is spoken on Rossel Island, Papua New Guinea. (Levinson, forthcoming: 10)
For an exhaustive presentation of all aspects of MacLaurys theories, see MacLaury, 1997.
13
MacLaury (1997: 87) writes that [t]he sameness [i.e. shared physiology of perception] imposes strong
constraints on color categorization in every language and in every individual system.
14
In his 1997 book Color and Cognition in Mesoamerica, MacLaury gives a thorough description of his
methodology in Chapter 3.
12

Setting the Scene: Studies of Colour Terminology

23

novelties are prerequisites for the modelling of the data in vantage theory.15
MacLaury (1992: 141) describes vantage theory as
[a] model of category dynamics according to which people create, maintain and
change a color category by analogy to the manner in which they keep track of their
own position in physical space. People form the analogy specifically at the level of
fixed and mobile co-ordinates, not merely between a space and a category as general
entities.

It would seem that MacLaurys fixed and mobile co-ordinates are closely related to,
if not identical with, the concepts of figure and ground, which are key notions
used in psychology and cognitive linguistics (cf. Langacker 1987). Thus,
MacLaurys theory tries to explain change of colour categories in a quite different
way from fuzzy logic la Kay and McDaniel. Furthermore, MacLaury (1997: 93)
introduces an important cognitive axiom in his model according to which paying
attention to similarities and differences is a crucial cognitive quality. Similar
thoughts have been expressed by Bolinger (1980) among others. In MacLaurys
theory, the introduction of similarity and distinctiveness as fixed or mobile
coordinates (cf. the quotation above) determines the size of the category. Although
initially there were major differences between B&Ks theories and vantage theory,
it would seem that Kay and Maffis identification of colour categorisation
principles (see 21-22 above) has moved the theory closer to vantage theory, to the
extent that the principles stated seem to resemble MacLaurys insights. However,
vantage theory is not confined to the domain of colour, but, as MacLaury indicates
(1997: chapter 6), may have wider implications in a general field of
categorisation.16 I return to MacLaurys theory in Chapter 4 in the context of type
modification.
1.2.4 The quest for verification outside anthropology
Outside anthropology proper, but still within the paradigm, Berlin and Kays theory
has been tested by disciplines such as psychology and linguistics. One area on
which attention has been focused has been that of the criteria for the identification
of BCTs. Several commentators have pointed out that some criteria lack theoretical
justification and that they constitute a hotchpotch of various aspects. Crawford
(1982) tested the validity of the criteria and suggested some modification. His new
definition (Crawford, 1982: 342) runs as follows:
A basic color term occurs in the idiolects of all informants. It has stability of
reference across informants and across occasions of use. Its signification is not
15

One interesting feature of MacLaurys theory, a spin-off effect from both his theory and methodology, is
that he has been able to identify a number of different semantic relations in the domain of colour: near
synonymy, coextension, inclusion and complementation (1997:150-152). Of these relations, near
synonymy and complementation should be familiar to the reader. This is also true of inclusion which
corresponds to the linguistic term hyponymy. Coextension, however, has not formerly been described in
linguistics. Coextension in the domain of colour is characterised as a relation in which the ranges of two
colour terms encompass each others focus. One term is dominant and larger, whereas the other is smaller
and skewed.
16
Taylor and MacLaury (1995) is a volume which addresses the application of vantage outside the colour
domain.

24

Chapter 1

included in that of any other color term. Its application is not restricted to a narrow
class of objects.

Quite a few researchers have subsequently used Crawfords definition of a BCT

rather than the one first suggested by B&K, see, for example, Biggam (1997: 17).
Boynton and Olson (1990) tested the validity of Crawfords criteria experimentally,
examining three measures: consistency (i.e. within the subject), consensus
(agreement among speakers) and response time. Boynton and Olson (1990: 1316)
were able to show that there was a complete dichotomy [] between the use of
basic and nonbasic chromatic terms to denote a set of samples []. They
interpreted this as a verification of Crawfords criteria, which they (p 1316) found
both substantial and sound.
Another area that has attracted interest is the number of colour terms in
various languages, and especially the question of whether there can be more than
eleven BCTs. A group of researchers at the University of Surrey have discussed in
a number of articles (Corbett and Morgan 1988, Moss 1989, Moss et al. 1990,
Morgan 1993 etc.) the status of Russian and French colour terms, in particular
words for BLUE, PURPLE and BROWN,17 reaching the conclusion that Russian has
indeed two basic terms in the blue category: goluboj and sinij. One appealing
aspect of the Surrey groups work is that they have employed a number of different
tests belonging to two broad categories: behavioural and linguistic tests. In a
meta-analysis mentioned in 0.3.1, Corbett and Davis (1995, 1997) evaluated the
adequacy of some of these tests with respect to the theory of BCTs, focussing on
the validity of the results as measures of basicness. They demonstrated that certain
tests, notably elicitation lists and frequency in texts were the most useful measures:
elicitation lists discriminate reliably between basic and non-basic terms, while
frequency correlates with the suggested evolutionary sequence and makes it
possible to distinguish between Primary and Secondary basic terms. Corbett and
Davies linguistic material is re-examined in greater detail in Chapter 3.
There have been other analyses of terms which are thought to be on the verge
of becoming Basic Colour Terms. As suggested by Morgan (1993), beige in French
might be one such term.18 Analysis of the status of the two French terms for
BROWN: marron and brun (Forbes 1979, 1986), showed that there might be an ongoing change in French through which marron is becoming a BCT at the expense
of brun. A somewhat different approach to the entire field of colour in French can
be found in Spence (1989). Zimmer (1982) suggested that German trkis,
turquoise, could be bordering on becoming a BCT. Boynton (1997) pointed out
that there is an area near the centre of the colour solid which is poorly covered by
English BCTs and that this area (corresponding to tan/peach) might be the focus
for a new BCT.
To sum up, the original suggestion put forward by Berlin and Kay has
undergone substantial development, but the basic assumptions of their initial study
remain valid. However, in its latest version, presented by Kay and Maffi (1999), at
17

The Surrey groups research scope in the colour field is much wider. It includes many more languages and
terms.
18
Lyons (1999: 38) identifies beige as a basic term in French: beige [] is a level-1 word in French.

Setting the Scene: Studies of Colour Terminology

25

least one of the cornerstones of the theory has been abandoned that of exhaustive
partition of the colour domain. The basics of the theory have been confirmed
outside anthropological methods, which further adds to its strength. It would also
seem to be the case that most attention is now paid to details in various languages.
Moreover, in general cognitive research the basics of the theory have been given
canonical status, and its present position can be well illustrated by the following
quotation from Valera, Thompson and Rosch (1991: 171):
[C]olor categorization in its entirety depends upon a tangled hierarchy of perceptual
and cognitive processes, some species specific and others culture specific. [][C]olor
categories are not to be found in some pregiven world that is independent of our
perceptual and cognitive capacities. The categories red, green, yellow, blue, purple,
orangeas well as light/warm, dark/cool, yellow-with-green, etc.are experiential,
consensual, and embodied: they depend on our biological and cultural history of
structural coupling.

1.3 An alternative position some criticism of the BCT theory.

Although the position of the B&K paradigm seems to be unchallenged, there are a
few dissenting voices present; researchers who adamantly reject the theory of BCTs
and the evolution of these terms. These voices suggest that the whole project is
doomed to failure and to producing misleading conclusions since it rests on
erroneous assumptions. According to a number of authors, cf. Sahlins (1976), Lucy
(1992a, 1997), van Brakel (1991), Saunders (1995), Saunders and van Brakel
(1997), Lyons (1995, 1999) and Wierzbicka (1990, 1993, 1996), the patterns of
universality that have been detected are really artefacts of the method. The general
claim of these critics is that the axiomatic assumptions that are part of the study
(see above) predict the results.19 The gist of this criticism can perhaps be captured
in the following quote by Sahlins (1976: 8-9).
[T]he problem is precisely what is here being called semantics. And it is surprising
nonetheless to find anthropology conspiring with a certain cognitive psychology to
collapse the problem of meaning into an act of pointing, that is, the act of naming
objective differences present to the senses. It is curious also that the entire discussion
of color categories [] has chosen to relegate true ethnographic existence of color
terms and percepts their actual cultural significance as codes of social, economic,
and ritual value to a secondary place of connotation. [---] And as Saussure himself
foresaw, when language is thus taken for a mere nomenclature rather than a
differential system of meaningful values, cognition will be reduced to recognition,
concept to percept, sign to signal and in the end, culture to nature.

According to these critics, by restricting the study of colour terms solely to the
dimensions of hue, lightness and saturation, the Western concept of colour has been
superimposed on terms which may have much wider reference. On closer
inspection, we can identify the following points as the targets of the criticism:
19

For an insight into the sometimes heated debate between van Brakel and Saunders and the champions of
the B&K paradigm, see Saunders and van Brakel (1997) with peer commentary. See also van Brakels
(1993) and Hardins (1993) debate in The British Journal for the Philosophy of Science.

26

Chapter 1

First we have the issue of whether there is a basic domain of colour

mentioned in Sahlins quote above. Lyons (1995:212), for instance,
claims that Conklin (1964) showed that succulence and desiccation
can be inherent in the meaning of a colour word, and should not be
reduced to the status of connotation. A similar view is expressed by
Lucy (1997).

Second, the question is whether the reduction of meaning to focus

and to neurophysiological responses tells us anything about colour
semantics (Wierzbicka, Lyons). Lucy (1992a: 180) describes the
reductionism as follows: But, in the end, the meaning is the
denotation and the denotation is the focal color, and therefore the
focal color is the semantic meaning of the basic color terms.
Lyons (1999: 52) maintains that a distinction has to be drawn
between the central, or focal, denotation and its total denotation.

Third, the use of English terminology as metalanguage leads us to

make inferences, i.e. see patterns where there are none to be found
(Lucy 1992a).

Finally, there is the issue of context-free mapping. Lyons (1999:50)

stresses that the B&K methodology favours context-independent
(second order) reference [of colour terms], rather than their use for
the (first order) description of entities and substances of this or that
color. Lyons also discusses this in terms of referential and
attributive use of colour terms. In his opinion, the B&K
methodology is only concerned with the first type of use.

It should be mentioned, however, that most of these critical points have been
answered by the proponents of the B&K paradigm. See for example Kays (1999a)
reply to Lyons (1999).
Some of these critics have adopted a position on the issue of meaning which
can be said to be a radicalised version of that expressed by Wittgenstein (1958).
Wittgenstein (1958: 20 43) observes that [f]or a large class of casesthough not
for allin which we employ the word meaning it can be defined thus: the
meaning of a word is its use in the language. In the context of colour terms, van
Brakel20 (1993: 132) rephrases Wittgenstein in the following way:
So perhaps [] what all green objects have in common is that we have learnt to call
them green and what all kwaalt [yellow-green, a Shuswap term] objects have in
common is that Shuswap speakers have learnt to call them kwaalt and can teach us
which things are kwaalt, just as we can teach them which objects are green.

20
van Brakel (1991) evaluates the whole prototype paradigm from this Wittgensteinian point of view. In
this article, he describes his own position as that of open-ended holism (p 248).

Setting the Scene: Studies of Colour Terminology

27

Wierzbicka (1996: 299) argues along similar lines when she points to the
possibility that colour terms are learnt ostensively: Blue means what people call
blue. However, Wierzbicka adds another dimension to the meaning of colour
terms by claiming that we use experiential reference points for our definition of a
primary colour concept. Like Kay and McDaniel she believes that colour terms like
orange, pink and purple are conceptualised as mixtures: e.g. ORANGE = YELLOW +
RED and PURPLE = BLUE + RED. Wierzbicka accepts B&Ks results but refuses to
see neurophysiology as the source of the universality. Rather she points to our
shared experience. Thus, she argues that the definition of red could have the
following appearance (Wierzbicka 1996: 317):
X is red. =
when one sees things like X one can think of fire
when one sees things like X one can think of blood
one can see things like X at times when one cannot see other things.

The difference in extension of colour concepts would then be due to differences in

reference points. In a similar fashion all early colour categories (i.e. before stage
VII, cf. Figure 1:1) are given experiential definitions by Wierzbicka. The aim of
her definitions is to identify the total denotation of a term, which is why some of
her phrases can be rather lengthy. Although her theory is thought provoking, few
people dealing with colour semantics seem to follow Wierzbickas tracks.
To sum up, we can see that the critics of the B&K paradigm are particularly
concerned with two aspects of the research: first, the method context free
mapping on standardised charts and, second, the reduction of meaning to focus.
As stated by Sahlins above, the critics believe that these procedures guarantee the
result. Proponents of the B&K research do not deny that meaning has wider
implications that these foci (see, e.g., Maffi and Hardin 1997). They argue,
however, that the aim has always been to uncover the underlying pattern of
universality, and this can only be achieved through the restriction of colour studies
to colour space.

1.4 My own position

Zlatev (1997) makes an interesting attempt at synthesising the differences between
the Wittgensteinian position which, in its extreme, may lead to total relativism and
a cognitive position which tends to emphasise universalism. His model situated
embodiment is meant to combine the strong points of the two theories and
eliminate the shortcomings. In Zlatevs view, the cognitive paradigm gives the
categorisation of an individual too much credit, whereas the relativists give too
much prominence to culture/society.
My own view is similar to Zlatevs. I fully subscribe to the view that, as
human beings, we are obviously constrained by the nature of our bodies. In this
sense, we can say that perception constrains categorisation. However, I am far more
doubtful whether the connection between perception and categorisation can be
reduced to fuzzy set modelling, as Kay and McDaniel would have us believe. Thus,
I want to state my reservations against neurophysiological reductionism.

28

Chapter 1

Furthermore, I also feel sympathetic towards Lyons view that it is important

to distinguish between (context-independent) referential use and descriptive
(contextual) use. A moments reflection will certainly provide the reader with a
number of examples which illustrate that the meaning of colour terms goes far
beyond their foci. Below I have collected a few sentences which underline the need
for contextual aspects to be taken into account in the description of the meaning of
colour terms.
(1)

a. Sgt Murdoch also said two cups of white coffee were found in Schwabs
room but he drank only black. (BoE: oznews)
b. One of the Irishmen was blonde-haired and drove a white Toyota Celica and
another used to drive the front of a lorry. (BoE: today)

(2)

a. By the same token, when I whizz all over the globe, caring and sharing, there
are bound to be a few under-privileged folk amongst my viewers, green with
envy over my glamorous jet-set existence. (BoE: ukmags)
b. It took an hour or more for the looting to be completed, and for the booty to
be collected, in neat piles, on the green grass of the meadow. (BoE: ukbooks)

(3)

a. But Arriet Arman sent er son to a grammah school proof that what that lot
over there (Labour) were saying to the voters was he paused for breath, his
face now puce. (BoE: times)
b. His shirt was puce and his shorts khaki. (BoE: ukbooks)

It is quite clear that we have to speak of different meanings (in the sense of hue) of
white, green and puce here. Whereas the (b) examples are likely to be near the
centre of the category or at least interpreted as such the (a) examples are
definitely not. Still, such use of white, green and puce has to be part of the meaning
of the terms, and a theory of colour semantics should be able to handle such
systematic usage.
My own position can be summarised as follows: I fully acknowledge the
importance of the results and facts that have been compiled and processed from
most of the worlds languages and that seem to testify to universal patterns in
colour categorisation. I see no reason to question these results; instead these
patterns should be taken as established facts within reasonable limits. One such
example is the cognitive salience of the six landmark colours: WHITE, BLACK, RED,
GREEN, YELLOW and BLUE. I also find it reasonable to assume that there is a colour
domain, universal in character, which is an irreducible basic domain. Langacker
(1987:149) points to the link between our sensory capacities and basic domains and
claims that [a]ll human conception is presumably grounded in basic domains.
On the other hand, as hinted at above, it is my conviction that a
comprehensive semantic description of the colour domain has to take into account
usage in a wider sense. Because, as pointed out by Lyons (1999) and other critics,
the naming and mapping procedures of the anthropological method do not reflect
normal use, a contextual study of colour terms is likely to broaden our

Setting the Scene: Studies of Colour Terminology

29

understanding of colour semantics. It would seem clear that for a full picture to
emerge, both the structure of the conceptual domain and the lexical field should be
thoroughly examined. In creating such a picture it should be possible to combine
the results of the mapping procedures with linguistic theories to explain extensions
of the kind presented in the (a) examples above. Chapters 4-7 in this book represent
such an attempt. It could perhaps be said that my main objection against traditional
B&K research is that the scope of colour semantics has been too narrowly defined.
It is now time to widen the picture. Furthermore, I believe that a study which takes
colour term usage into consideration might be able to shed new light on debated
issues and from a slightly different angle. One example of a debated issue of this
kind is Conklins (1964) observation that succulence is an inherent dimension of
malatuy (GREEN in the B&K classification).

1.5 Colour semantics based on texts: historical studies of colour

terminology
As was pointed out above, the method which dominates the study of colour terms
has been that of mapping and naming in colour charts. Even if Kay (1999a: 78)
refuses to see this as context-independent reference to the colour domain, it can be
argued that this procedure hardly reflects normal usage of colour terms. In this
sense, we can regard Lyons and Lucys criticism mentioned above as justified.
One conspicuous exception to the conformity of method is the historical study
of colour terminology. A number of articles and books have been published
discussing the colour terminology in ancient languages in the context of Berlin and
Kays theory.21 The most thorough investigations in this respect are the
monographs by Maxwell-Stuart (1981a and 1981b), Brenner (1982) and Biggam
(1997 and 1998). Historical research of the colour systems cannot, per definition,
employ colour charts so the researchers have had to resort to other methods. I
discuss these below focussing on those methodological aspects which have
influenced me, but I also point to some important differences between these
historical studies and my own.
One noticeable feature of these studies is that they are all corpus studies in a
strict sense, i.e. they explore the usage of certain colour terms in well-defined text
corpora. Thus, Maxwell-Stuart examines the use of the terms (glaukos
blue-grey22) and (charopos23) in Classical Greek on the basis of listed
textual sources, Brenner analyses the meaning of the colour terms appearing in the
Old Testament, and Biggam investigates how the colour areas BLUE and GREY were
designated in Old English on the basis of, again, well-defined textual evidence.
Another methodological element that these studies share is that they contain
meticulous analyses of individual examples in context, allowing for conclusions to
21

Cf., for example, Barley (1974), Burnley (1976, 1992), Krieg (1976), Moonwomon (1994), Shields (1979)
and Wescott (1975).
22
According to Moonwomon (1994). Maxwell-Stuart (1981a:124) defines the most likely meaning of
in prose as light-blue, tending to a touch of milkiness or greyishness, as in steel grey.
23
Maxwell-Stuart (1981b:3) describes the meaning of this terms thus: It has been the convention to
translate , either as bright or grim or fierce, or to treat it as more or less a synonym of and
so render it by blue-grey or grey-blue.

30

Chapter 1

be drawn as to the extension of the terms. The investigators have taken a number of
parameters into account including the genre of the texts (for instance verse vs.
prose) and the entity that a particular colour word describes in a particular context.
This interdisciplinary semantics as Biggam (1997: 14) calls her approach
involves
a wealth of helpful evidence from disciplines such as archaeology, botany, zoology,
geology and many more, as well as from more structured methods of linguistic
research. Evidence on colour semantics should be garnered from any discipline which
can shed light on a particular case.

Biggams (1998: 312-323) clear summary of the vocabulary for GREY gives a good
picture of the merits of this kind of approach. All three authors give attention to the
etymology of the terms considered, but only as a marginal tool. As mentioned
previously (Section 0.3.1.), Biggams insights have influenced my own study,
although I have found it necessary to modify and simplify her methods due to the
nature of my material.
There are some other characteristics of these studies which should also be
mentioned. I discuss these briefly below merely to give a picture of the problems
that historical semanticists face.
It seems that the nature of the historical material can constitute a major
problem. If we look at the studies mentioned above, it turns out that colour terms
are not very frequent in old manuscripts. This obviously affects the reliability of the
conclusions drawn by the researchers. As examples of the low frequency, we can
mention that in Maxwell-Stuarts study of , the word appears 133 times in
prose, that Brenners most frequent colour term, (lbn bright/white),
occurs only 29 times, and that Biggams analysis of the field of BLUE is based on a
total of 79 tokens of five different terms (hwen being the most frequent one with
64 tokens). Brenner is also frequently faced with the problem of hapax legomena in
her categorisation of secondary and tertiary24 colour terms. Thus, the possibility of
exploring the domain of colour systematically is very limited in historical studies.
What further complicates the picture for the historical linguist is the great span in
years between the different texts in his or her corpus. As many as several hundred
years can separate the youngest and the oldest texts. All three researchers are aware
of this complication and try to adjust their conclusions accordingly, but it does,
nevertheless, affect the reliability of the results. Another characteristic that these
three historical studies share is that they are primarily descriptive and little
semantic theory is used although Ullmann (1957 and 1962) and Lyons (1963) are
mentioned and used by Brenner and Biggam. In relation to the B&K paradigm, it is
an explicit objective of Biggams (1997, 1998) to establish whether Old English
had BCTs for BLUE and GREY.
To sum up, these three studies show that there are other ways of examining the
semantics of colour vocabulary than that of the traditional colour charts. The
historical researchers were forced to explore new methods since the languages they
24
Brenner (1982: 33-34) makes a categorial distinction between primary, secondary and tertiary colour
terms. The first category appears to be identical with BCTs.

Setting the Scene: Studies of Colour Terminology

31

studied were no longer spoken. Although the corpora set some limitations as to the
reliability of the results, these methods are in my view appealing and should be
explored by semanticists studying present-day languages.

1.6 The study of non-basic colour terms

Let me finally point out that work on non-basic colour terms is much less frequent
when it comes to the study of colour nomenclatures. Generally speaking, we can
distinguish between two types of research: the study of the development of nonbasic terms, and the study of usage of non-basic terms in relation to gender. I
discuss non-basic terms in greater detail in Chapter 5.
The category of non-basic terms is named Elaborate Colour Terms (ECTs)
in the present study. ECT is preferred to Cassons (1994, 1997) label secondary
term since the latter, technically speaking, could include derivative forms (e.g.
yellowish and pinky) and compound colour terms (e.g. red-brown and leaf-green);
the term could also easily be confused with the notion of secondary basic terms
(see 1.2.2. and 3.2). The term ECT is here understood as referring to conventional
simplex lexemes (cf. Bauer 1983). The term has previously been used by
Nowaczyk (1982) and Simpson and Tarrant (1991) in a similar fashion.
In the study of the origin and development of secondary terms, Casson (1994,
1997) has contributed with two important studies. He could show that there seems
to have been a change in the semantics of ECTs in the history of English. The
development, which took place in the Middle English period, meant that the
emphasis was changed from lightness to hue. It appears that most secondary terms
in Old English may have been lightness terms rather than hue terms, whereas hue
became increasingly important during the Middle English period. Furthermore,
Casson demonstrates that the development of new ECTs was normally based on a
metonymic process in which entity stands for entitys colour. I return to Cassons
discussions in Chapter 5. In another article, Smith et al. (1995) analysed the
salience of basic and non-basic terms on the basis of a free-listing task, and
suggested a typology of colour terms.
Most research dealing with ECTs, however, has dealt with these terms and
gender. Following Robin Lakoffs (1975) claim that women use specific colour
terms more often than men, quite a substantial amount of research has been devoted
to this particular issue. Rich (1977), Swaringen et al. (1978), Nowaczyk (1982),
Simpson and Tarrant (1991) and Green and Gynther (1995) are among the
researchers who have tested Lakoffs (1975) claim empirically. There appears to be
general agreement from this research that women do indeed use ECTs more often
and also more accurately. Although the evidence is convincing, there are some
aspects which deserve closer inspection. One such aspect is method. I have been
critical elsewhere (Steinvall 2000) of the unproblematic way in which these studies
have used one particular method: context free mapping and naming. To my mind,
this method does not reflect the circumstances of usage with sufficient accuracy.
Therefore I think that there is room and need to look at this phenomenon through
corpus studies, especially in order to validate the claim that women use ECTs more
often. In their handbook to the BNC, Aston and Burnard (1998) give an example of

32

Chapter 1

how such an approach could be made, asking the question whether mauve is a
particularly female term to use. My own investigation of Victorian poetry referred
to above was also meant to test whether such preferences could actually be studied
in corpora. It is my belief that studies along those lines may add to our knowledge
of ECTs. The issue here is whether we should regard gender25 as an immediate or
the ultimate explanation. It would seem that certain types of discourse have become
gendered, and that these are characterised by fine colour distinctions. If this is true,
then, ultimately, the focus on fine colour designation is likely to be a product of the
nominal domain rather than gender. In Chapter 5, I look more closely at the
distribution of some ECTs in relation to nominal domains.

1.7 Summary
The aim of the present chapter has been to give a brief summary of the current state
of affairs in colour terminology research. Due to lack of space, I have been forced
to focus on what I have called the B&K paradigm, which means that some aspects
of present-day research have been ignored. The development of the paradigm has
been presented and it has been demonstrated that most aspects of the original
theory are still valid. I have also pointed to some important criticism and tried to
identify my own position in relation to the paradigm and its critics. Finally, I have
indicated that there are close affinities between the present study and those carried
out in historical linguistics: in particular the use of a well-defined textual corpus
and the inclusion of contextual elements. Important differences are the size of
corpus and the systematic application of recent linguistic theory. In the next
chapter, I give a presentation of the linguistic framework within which I develop
my analysis.

25

Technically, the actual parameter that was considered in these studies is that of sex, and, indeed, all but
Greene and Gynther title their studies as investigating sex. However, it would seem likely that this
phenomenon is linked to gender roles rather than biology.

Chapter 2

Providing the Tools:

Meaning in Cognitive Linguistics

2.1 Introduction
Chapter 1 described some controversies and results in colour terminology research
over the last thirty years, pointing out that certain aspects of the subject have been
infrequently investigated or been completely ignored despite the large number of
publications. Very few studies have examined the use of colour terms outside the
colour charts of the anthropological method that is, little effort has been made to
analyse phenomena related to usage from the point of view of semantic theory.
The prime objective of my dissertation is to fill this vacuum, and this chapter offers
a short presentation of the theory of semantics that is used in my study. The
theoretical framework within which my linguistic analysis is carried out is that of
cognitive linguistics. As a school of linguistics, cognitive linguistics has developed
quickly during the last twenty years and is becoming increasingly influential, at
least in terms of the number of publications. This relatively fast development has
not been unproblematic, however. It soon becomes clear to the reader of almost any
cognitive linguistics textbook that, although its central tenets are not disputed,
cognitive linguistics is not a homogeneous school. What is particularly striking is
the great variety of terms that characterises this theoretical model. (For brief
summaries in this respect, see Taylor 1995, Clausner and Croft 1999.) The same
term may be used differently by different writers, or different writers may use
different terms to describe apparently very similar phenomena.
The purpose of this chapter is twofold. First, it aims at giving a coherent but
condensed description of how meaning is analysed and conceived of within the
framework of cognitive linguistics. In doing so, I discuss in passing the relevance
of the theory in relation to colour studies. However, such mentions should only be
viewed as appetisers in relation to the chapters to follow, and not as premature
analyses. Furthermore, it should be emphasised that my account of cognitive
linguistics is necessarily fragmentary rather than comprehensive. Second, despite
the risk of adding to the terminological confusion, I also try to define the terms that
are central in my own analyses. I owe much of my conception of semantics to the
work of Langacker (1987, 1991b, 1999), but in some crucial respects, I prefer to
use a terminology which differs from his.
The chapter is structured in the following way. It starts with an account of the
notion of categorisation since it is a key concept in cognitive linguistics. This is
followed by a presentation of how cognitive linguistics deals with other phenomena
closely related to categorisation, such as context and situation. The third section
presents an analysis of conceptual and lexical meaning from a cognitive
perspective; in this section I define key elements and introduce a schematic figure
of meaning that recurs throughout my dissertation. Finally, the last section gives an
account of sense relations. These are discussed from two perspectives, which, it is
argued, reflect the speaker hearer situation.

33

34

Chapter 2

2.2 Principles of categorisation

To be able to understand the world and interact with it, we sort our experiences, as
it were, into boxes. This on-going sorting process, which for the most part is
unconscious, is usually termed categorisation. As Lakoff (1987: 5-6) points out,
categorisation is the basis for our ability to think, act and speak:
Every time we see something as a kind of thing, for example, a tree, we are
categorizing. Whenever we reason about kinds of thingschairs, nations, illnesses,
emotions, any kind of thing at allwe are employing categories. Whenever we
intentionally perform any kind of action, say something as mundane as writing with a
pencil, hammering with a hammer, or ironing clothes, we are using categories. The
particular action we perform on that occasion is a kind of motor activity (e.g., writing,
hammering, ironing), that is, it is in a particular category of motor actions. They are
never done in exactly the same way, yet despite the differences in particular
movements, they are all movements of a kind, and we know how to make movements
of that kind.

Thus, from this point of view, categorisation is pivotal to any understanding of how
our minds work and in particular how language works since there is an explicit link
between categorisation on the one hand and thinking and understanding on the
other. A crucial claim made by cognitive linguistics is that categories are formed
through our interaction with the world. This interaction is experienced through our
bodies and as a result categories are ultimately grounded in bodily experiences (cf.
Varela et al. 1991). This highly egocentric position has been named experiential
realism by Lakoff (1987:xv).
According to experiential realism, our conceptual system grows out of and is
constrained by the limitations set by our bodies to experience in terms of
perception, movement, and physical and social interaction. Consequently, abstract
thinking and imagination, which are not directly grounded in bodily experience, are
considered to be based on bodily experiences mediated through metaphor,
metonymy or analogy. As an example, consider our understanding of time, which is
frequently metaphorically understood as movement; thus we can talk about the
conceptual metaphor +TIME IS A MOVING OBJECT+ (Lakoff and Johnson 1980: 42)
our notion of moving objects being experientially based.
Another consequence of the theory of experiential realism is that structural
principles are also viewed as being ultimately grounded in bodily experiences.
Johnson (1987) suggests that there is a set of image schemata which serves as
structures for the organisation of our experience and of our understanding of the
world. He (1987: 29) defines an image schema as consisting of
a small number of parts and relations, by virtue of which it can structure indefinitely
many perceptions, images and events. In sum, image schemata operate at a level of
mental organisation that falls between abstract propositional structures, on the one
side, and particular concrete images, on the other.
The view I am proposing is this: in order for us to have meaningful, connected
experiences that we can comprehend and reason about, there must be pattern and
order to our actions, perceptions, and conceptions. A schema is a recurrent pattern,
shape, and regularity in, or of, these ongoing ordering activities. These patterns
emerge as meaningful structures for us chiefly at the level of our bodily movements

Providing the Tools: Meaning in Cognitive Linguistics

35

through space, our manipulation of objects, and our perceptual interactions. [emphasis
original]

Accordingly, image schemata form important structural principles, and Lakoff

(1987) suggests that whenever we talk about structures in abstract domains, these
structures should be understood as image schemata. An important characteristic of
image schemata is that they are gestalt structures, meaning that, although they
consist of smaller parts, they form organised and unified wholes. Important image
schemata that will recur in my presentation are the containment schema (Johnson
1987: 22-23) and the centre-periphery schema (Johnson 1987: 124-25). Since
both categories and concepts rely on structure, it follows that image schemata are
preconceptually grounded in physical experience.1 It should now be clear that, in
cognitive linguistics, categorisation is a mental phenomenon grounded in bodily
experience and structured by abstract image schemata. Figure 2:1 below attempts to
capture the essential features of this position. Structuring principles (image
schemata) are the result of bodily experiences and the bodys interaction with the
real world. Thus, categorisation is ultimately grounded in bodily experiences, and
the body can be seen as a mediator, supplying the structures (schemata) for our
conception of the real world. The most essential aspect of this view is that indicated
by the small-headed arrows: categories are only indirectly linked to the real world
they are mental phenomena whose contact with the world is based on the body.
Analogous with this, imagined alternative worlds are structured according to our
embodied conception of the real world.

MIND

Cognitive categories
(concepts)

Mental categorisation

BODY

Structuring principles

Perceptual and sensory-motor abilities

Physical interaction

WORLD

Bodily experience

Real world objects

Figure 2:1 Categorisation in experiential realism

1

For a critical discussion of the employment of image schemata in cognitive linguistics, see Zlatev 1997.

36

Chapter 2

How, then, do we categorise? In their pioneering work, Rosch and her associates
(e.g. Rosch 1975b, Rosch et al. 1976), have demonstrated that categories are
typically graded. The idea of categorical focus in fact started with Berlin and Kays
(1969) study discussed in Chapter 1. As we have seen, colour categories do indeed
have focal points and fuzzy boundaries and, subsequently, Rosch demonstrated that
this is true of many other categories. Or, to be more precise, what she was actually
able to demonstrate was that prototype effects occur in categorisation tasks. How
these effects should be interpreted in terms of category structure and mental
representation is discussed by Lakoff (1987: Chapter 4). As we noted in Chapter 1,
Berlin and Kays universal pattern of colour categorisation hinges upon their
equation of focus with category. However, this is hardly an accurate description of
what categories are. It seems that the general idea within cognitive linguistics is
that prototype effects should not be taken to be a characterisation of category
structure. Furthermore, it has also been claimed that we cannot consider prototypes
to be mental representations in a strict image sense.
In what follows I will attempt to give a summary of Langackers view of
categorisation, a view to which I will adhere in my discussion of categorisation.
Langacker makes a distinction between two types of categorisation: by prototype
and by schema. The two modes of categorisation are viewed as intimately linked,
but emphasise different aspects of the same phenomenon. It is important not to
confuse Langackers schema with Johnsons image schemata. Langacker (1987:
371) defines a schema as
an abstract characterization that is fully compatible with all the members of the
category it defines (so membership is not a matter of degree); it is an integrated
structure that embodies the commonality of its members, which are conceptions of
greater specificity and detail that elaborate the schema in contrasting ways.

Figure 2:2, below, illustrates categorising principles in the Langackerian

framework.
SCHEMA

PROTOTYPE

EXTENSION

Figure 2:2. Categorisation by prototype and schema.

(Based on Langacker 1993: 2)

Langacker (1993:2) observes that although their relative importance varies from
one category to the next, schemas and prototypes are both essential to category
structure, reflecting different aspects of a unified phenomenon. Extensions
typically occur when we categorise a new experience that deviates slightly from
previous experiences of a category as belonging to that category. Langacker claims

Providing the Tools: Meaning in Cognitive Linguistics

37

that extensions are characteristically made from the prototype. Extensions may be
based on metaphor, metonymy, or other processes. From this extension a schema is
abstracted, containing the common features of the prototype and the extension.
The solid arrows in Figure 2:2 illustrate the specialisation or elaboration of
the schema that the prototype and any extension represent. The dotted lines
upwards indicate the abstraction or schematization that is taking place as the
common features of the instantiations are schematised. If new extensions are made,
a new schema is extracted and thus, according to this model, any extension will
lead to new schemata being formed on a more abstract level, so that the widening
of a category will necessarily lead to a higher structure (cf. Langacker 1987:
chapter 10). This means that complex categories usually constitute quite advanced
structures which Langacker describes as schematic networks of nodes representing
schemata, prototypes and extensions. The network model is meant to capture the
same kind of phenomenon as Lakoffs radial category and both terms will appear
in my analysis. Langacker (1999:101) observes that there are similarities between
the models: [t]o the extent that the network consists of chains of extensions
radiating outward from [the prototype], it constitutes a radial category. Lakoff
(1987:91) describes a radial category as follows:
There is a central subcategory, defined by a cluster of converging cognitive models
[]; in addition, there are noncentral extensions which are not specialized instances
of the central subcategory, but rather are variants of it []. These variants are not
generated from the central model by general rules; instead they are extended by
convention and must be learned one by one. But the extensions are by no means
random. []We will describe the extensions of a central model as being motivated by
the central model plus certain general principles of extension. [italics original]

Some researchers, for example Dirven and Verspoor (1998), seem to have blended
the ideas of radial structure and network into the term radial network.
Another feature of the network model is that a network is supposed to be
dynamic and not static it is constantly elaborated through the input of usage and
experience. We are concerned here with the cognitive notion of entrenchment.
Langacker (1999:93) describes it as a general psychological phenomenon
characterised by the fact that
[t]he occurrence of psychological events leaves some kind of trace that facilitates
their re-occurrence. Through repetition, even a highly complex event can coalesce
into a well-rehearsed routine that is easily elicited and reliably executed.

Thus, continued activation of certain parts of the network (certain schemas,

extensions or prototypes) will entrench these and may, as a consequence,
restructure the network. For instance, the category COMPUTER has probably been
restructured continually during the past twenty years, for the language users who
have been part of the development.
It should be pointed out that, at first sight, Langackers idea of network does
not seem to be entirely compatible with Wittgensteins (1958) and later Rosch and
Mervis (1975) idea of family resemblances. Discussing the category Spiel
(game), Wittgenstein (1958: 31-33: 66-70) could demonstrate that there seems

38

Chapter 2

to be no property that is shared by all the instances of this category. Instead the
category is kept together through a chain of properties: the structure could be
described as forming
a set of items of the form AB, BC, CD, DE. That is, each item has at least one, and
probably several, elements in common with one or more items, but no, or few,
elements are common to all items.
(Rosch and Mervis 1975:575)

It is important to note that the highest order schema of any network is supposed to
represent the commonalities of the entire concept of game. Given the lack of
common properties, such a schema would have to be very general. Taylor (1995)
rejects Langackers notion of schema for this very reason claiming that it is not
meaningful to talk about abstract schemata which are almost empty. However, to
my mind, the dynamic character of the network (cf. Langacker 1987: 381-386)
makes it possible to accommodate family resemblance structures within the
network metaphor. For instance, the network model can take other forms than that
of a radial category.
In conclusion, the gist of categorisation in cognitive linguistics can be
summarised as follows:

Categorisation is a mental phenomenon through which we organise

our understanding of the world.

Categorisation is grounded in our bodily experience of the world.

Categorisation creates cognitive categories or concepts which are

typically complex structures.

Concepts can be viewed as forming networks of any complexity in

which the nodes can represent prototypes, extensions and
schemas of any order.

Cognitive categories (structures) are dynamic and under constant

elaboration through our interaction with the real world.

The repeated activation of certain categories or structures will

entrench these and make them more salient from a cognitive
point of view.

2.3 Situation, context and mental spaces

One of the most important aspects of the cognitive framework and the position of
experiential realism is that there is no Gods eye view (cf. Johnson 1987). This
suggests that things become meaningful only through the experiencing person.
Accordingly, as mentioned above, semantic phenomena are always mental
phenomena, created through the conceptualisation of the person. This, in turn,
suggests that we cannot reduce meaning to a reference phenomenon in the real

Providing the Tools: Meaning in Cognitive Linguistics

39

world but have to consider it a mental phenomenon. As pointed out by Langacker

(1987: 113-14), most of us are familiar with conceptual worlds, in terms of
fantasies and dreams, so we know that we can create them. The important thing
here is to realise that our understanding of the real world itself is also a conceptual
world. Thus it makes sense to claim that it is this conceptual world of reality and
not the real world that is relevant for a discussion of meaning. This difference
makes it possible to explain why people in earlier times actually did see trolls, for
instance. Their understanding of the world shaped certain perceptions into these
concepts, and gave them reference.2
For the purpose of the present study, there are a few key concepts that need to
be presented and defined. Let us first consider situation and context. Since it is
important to distinguish between our mental construal of the world and the world
itself, this distinction is also necessary in terms of interpretation. Here, I follow
Ungerer and Schmid (1996) and consider context a mental phenomenon, whereas
situation refers to interaction between things in the real world. This suggests that
context is linked to other key notions such as vantage point and construal, which
are presented in 2.6. Situation, on the other hand, being a real world phenomenon
entails no perspective or focus.
Intimately linked with the idea of context is the theoretical construct of
mental spaces (Fauconnier 1994, 1997). Fauconnier (1994: 16) describes metal
spaces as
structured, incrementable sets [] with elements (a, b, c, ) and relations holding
between them (R1ab, R2a, R3cbf, ), such that new elements can be added to them
and new relations established between their elements.

Less technically, we can view a mental space as our mental representation of

ongoing discourse, or possibly of an ongoing situation. The mental space will
continue to be built as the discourse proceeds, and occasionally new parallel spaces
are built with the help of space builders. Thus, a mental space can be said to be our
image (context) of what is being described (discourse) or being perceived
(situation). Consequently, the current mental space will constitute the context for
any mental representation.3 Fauconnier (1997: 39) observes that mental spaces
[] are internally structured by frames and cognitive models [italics original].4
This structure is created by the background knowledge provided by lexemes.
Frames and cognitive models will be treated in the next section.
Langacker (1991b, 1999) suggests that we can view the mental space that is
created in an on-going conversation as the joint construction of the two (or more)
participants a current discourse space (CDS). He (1991b: 97) defines it as a
mental space which comprises those elements and relations construed as being
2

For further discussions of the consequences of this particular perspective, see Lakoff (1987) and Johnson
(1987).
3
Lakoff (1987:281) claims that A mental space is a medium for conceptualization and thought. Thus any
fixed or ongoing state of affairs as we conceptualize it is represented by a mental space. As an example
Lakoff mentions (p 281) our immediate reality, as understood.
4
Frame in Fauconniers use appears to be very similar to that of Ungerer and Schmid (1996), i.e. it is
centred around a verbal situation. For details, see Fauconnier 1997: 10-13.

40

Chapter 2

shared by the speaker and hearer as a basis for communication at a given moment
in the flow of discourse. This space is constantly changing as the conversation
proceeds but relies to a great extent on the fact that the structuring knowledge is
shared by the speaker and the hearer. Following Langacker (1999), we can
illustrate the situation as in Figure 2:3 below.

Speaker knowledge

CDS

Hearer knowledge

Figure 2:3. Current discourse space and participant knowledge.

(Based on Langacker 1999:263)

Given the structure of the space and the (believed) shared knowledge, the hearer
will expect the speaker to stay within the created space unless there is a formal sign
of the creation of a new space. If such a sign is missing, the hearer tends to draw on
situational and/or wider contextual knowledge to make sense out of what is being
said. The shared knowledge of the space and its structure will also guide
expectation in terms of level of specificity, an issue which will be readdressed in
Chapter 6.
A new space other than default space can easily be built with the help of a
space builder. Let us consider one example, based on Fauconnier (1997), sentence
(1) below.
(1)

I think that Bill is in love with Hillary

As pointed out by Fauconnier, a sentence of this type brings to the fore our
background knowledge of what x is in love with y means. In Fauconniers terms
the phrase I think functions as a space builder creating a mental space of my
belief in this particular case. Thus we can illustrate the process as in Figure 2:4
below. 5

F (in Fauconnier 1994) symbolises the connector (Fauconnier 1997:43), whose function is to link the
objects in Base space with the roles in Belief space. In this particular example the connector is that of
identity, i.e., b corresponds to b.

Providing the Tools: Meaning in Cognitive Linguistics

a: Bill
b: Hillary

41

LOVE ab

b
a

Base space

Belief space

Figure 2:4. Space building. I think that Bill is in love with Hillary
(Based on Fauconnier 1997:43.)

Belief space is structured according to the background knowledge of the cognitive

model of LOVE. Included in the structure of the space is also our background
knowledge of the participants, in this case two people called Bill and Hillary.
Exactly who these people are (i.e. to which individuals these names refer) will be
determined by mutual assumptions relating to the context or situation and how we
then interpret the statement in a wider sense will be affected by our knowledge of
these people. Consequently, the meaning of love might be slightly restructured in
accordance with our knowledge of the two people and their history etc.
To sum up, in this section the importance of distinguishing between the real
world and our conceptual understanding of it has been pointed out. To codify this
difference it was posited that a distinction be made between situation and context.
Mental spaces were identified as important constructs for the understanding of
conceptualisation and thought.

2.4 Characterising meaning

Cognitive semantics embraces a conceptualist view of meaning, where meaning is
claimed to be encyclopaedic in nature. An important consequence of this view is
that the meaning of any lexeme is in principle open-ended, since it may, in the
individual speaker, evoke any kind of association. However, in practice, some
meanings are conventionalised in the speech community, and it is these that I refer
to when I talk about meaning. This is also true on the level of the individual:
meanings are entrenched to a greater or lesser degree, due to frequency of usage.

42

Chapter 2

Following Langacker (1987), I will call the conceptual content of a lexical category
a predication; however, frequently I will use the term concept loosely in the same
sense. In cognitive semantics, the meaning of a lexical item is usually regarded as
being best described in terms of a figure-ground relationship. Any linguistic
concept (predication) is understood against some sort of background knowledge
and the associations that are linked to the concept.
2.4.1 Domain, network and attribute
The background knowledge which serves as the ground for a predication (the
figure) is called a cognitive domain, or shorter, domain. In Langackers view a
domain is a conceptualisation of any sort and of any kind of complexity against
which a meaning is created. Langacker (1987) makes a fine distinction between
different types of domains. Most domains presuppose other underlying domains, so
that a particular predication can be built on an advanced structure of domains. This
is succinctly illustrated by Croft (1993), who discusses the underlying structures of
the letter T. Croft demonstrates that the domain structure of this relatively simple
concept can be quite deep. However, there are primitive domains which cannot be
reduced to more basic ones these are called basic domains (Langacker 1987).
Examples of such domains are COLOUR, SMELL and TWO- and THREE-DIMENSIONAL
SPACE. Non-basic domains are called abstract domains by Langacker. An abstract
domain seems to correspond essentially to what Lakoff (1987) calls ICM
Idealised Cognitive Model.
It would seem that a predication is frequently characterised against a
background of several domains, not just one. Langacker calls this cluster of
domains presupposed by a concept a domain matrix or matrix.6 Again we can find
alternative terminology. For instance, Taylor (1995: 87) calls this construct frame:
In the following, frame will refer to the knowledge network linking the multiple
domains associated with a given linguistic form.7 This use of frame seems to
correspond closely to that of Fillmore (1985) and Barsalou (1992). As a technical
definition of frame, Barsalou (1992: 28) offers the one used in artificial
intelligence: a fixed set of named slots whose values vary across applications.8
Thus, although there seems to be a consensus as to how meaning should be defined,
the great variety of terminology can be confusing. In this work the term domain
will be used for this construct, irrespective of its complexity.
In a matrix there is usually one domain which is the most important for the
characterisation of the predication. This domain is termed primary domain by
Langacker (1987: 163-64). It seems that the connection between the primary
domain and the figure is typically a part-whole or a subordinate-superordinate
relation. These relations correspond to what in lexical semantics are termed
meronymy and hyponymy. Clausner and Croft (1999) claim that the connection is
6

Ungerer and Schmid (1996: 190) appear to believe, erroneously in my view, that the matrix contains only
basic domains.
7
However, it seems that the notion of frame is being increasingly used to denote the background of a verbal
situation. In this use it seems to constitute a special kind of domain. An example of this approach can be
found in Ungerer and Schmid (1996). They (p 211) define frame as a type of cognitive model which
represents the knowledge and beliefs pertaining to specific and frequently recurring situations.
8
The original italics have been normalised.

Providing the Tools: Meaning in Cognitive Linguistics

43

primarily of a part-whole kind. However, while this appears to be true of many

basic level predications, it is certainly not true at a subordinate level; it seems
reasonable to argue that the meaning of German Shepherd is primarily defined
against the background of DOG. This is in line with views held in Gestalt
psychology, where it is suggested that the whole is conceptualised before the parts
(cf. the discussion of image schemata above). As mentioned earlier, the basic idea
of figure-ground forms the foundation of Langackers theories. In addition, it
should be stressed that this part-whole idea is completely different from the
classical approach embodying necessary and sufficient conditions. The domains
evoked by a lexical item are ranked according to their centrality or weight, to use
another metaphor. The most central domain is the primary domain. Langacker
(1999: 5) suggests that centrality can be determined in terms of likelihood and
strength of activation. The notions of primary domain and centrality of domains
explain the difference in meaning among concepts which evoke more or less the
same kind of domains. Langacker mentions roe and caviar. Although, technically
speaking, they denote the same thing the words have different meanings. In the
framework described here, we can say that the primary domain of roe is FISH
9
REPRODUCTION, whereas in the case of caviar this domain is marginally ranked.
The idea of background knowledge may need some further explanation and I
use an example established in the literature for my presentation; that of a weekday,
in my case Friday. The meaning of Friday can only be understood against our
conception of the seven-day week, divided and organised according to certain
principles. As pointed out by Lakoff (1987), this is an idealised model according to
western principles. Time can be divided in numerous other ways; Aitchison (1994)
mentions the Incas, who organised the week as a nine-day cycle. The basic idea,
however, is that without the concept WEEK, we cannot understand the meaning of a
weekday. If we return to our example Friday, it is clear that it is defined in relation
to our concept of WEEK. As a domain, WEEK has to be considered an abstract
domain since it presupposes a number of other domains, the day and night cycle
being the most important one. (Cf. Taylor 1995: 84.) It is also clear that there are
many other factors that affect the meaning of a word like Friday. However, the
most important one is the week and its structure, so WEEK can be seen as the
primary domain of FRIDAY.
If we want to maintain an encyclopaedic and open-ended view of meaning,
we also have to acknowledge that the meaning of Friday can go beyond the mere
designation of the sixth day of the week.10 People know a lot more than this about
Fridays and to the degree that this is general knowledge, and culturally significant,
we can say that it is part of the meaning of Friday. Such associations can be seen as
other domains being part of the matrix.

Other examples of the same phenomenon are the classic example of the morning star, evening star and
Venus. (Mentioned by Langacker, 1987: 165).
10
As an aside, we can observe that there is a difference here between some European languages, in that for
instance Swedish considers fredag, Friday, the fifth day of the week. However, in practice this does not
lead to any communicative difficulties. Incidentally, there is also another model in English-speaking
societies which can create a conflict: the division of the week into workdays and weekend suggests a
different position for Friday.

44

Chapter 2

In my presentation, I essentially follow Langackers terminology, but with

some important alterations. First and most importantly, I use the term attribute
(following Barsalou 1992 and several others) for the units of association outside the
primary domain mentioned above. The primary domain itself contains attributes, of
course, but since these are at the very centre of the characterisation, there is no need
to draw further attention to these properties. Consider, for example, German
Shepard: the attributes of dogginess are presupposed since the predication is
defined against the primary domain of DOG. It seems to me that attribute defined
thus basically corresponds to highlighted aspects of peripheral domains and also to
what Langacker frequently mentions as specifications, associations and
properties (cf. e.g. Langacker 1987: 158-161). It is essential to understand that
attributes are not to be taken to be semantic features in the classical sense of
necessary and sufficient conditions. An attribute is not in any sense an atomic
feature, but it constitutes in itself a complex structure of meaning (cf. Smith and
Medin 1981:17, Barsalou 1992: 30, Taylor 1995). Furthermore, attribute formation
can be highly productive: Barsalou (1992: 31) argues that people are highly
creative in their construction of attributes, often producing new ones relevant to
specific contexts. For a technical definition of attribute I suggest a solution along
the lines offered by Taylor (1995: 63), namely that attributes are dimensions
(highlighted aspects of domains) in relation to which entities (abstract/concrete,
thing/relation) are regarded as similar (or different).11
The fact that attributes have an internal structure, i.e. that they are concepts, is
very interesting from a structural point of view, as it suggests that we can talk about
a kind of limited degree of recursiveness in cognitive structures. Furthermore, the
conventional link between two cognitive structures creates a relationship of
reciprocity between the structures. If a concept functions as an attribute of another
concept, then the converse should also be true. Langacker (1987: 164) argues that if
[mouse] is an attribute (he does not use this term) of CAT, then [cat] is an attribute
of MOUSE. It should be borne in mind, however, that this reciprocity is not
necessarily symmetrical. On the contrary, it would seem that a normal occurrence
would suggest asymmetrical reciprocity. For example, a typical (and thus central)
attribute of swans, notably the Whooper Swan (Cygnus cygnus) and Mute Swan
(Cygnus olor) is [white] thus, in accordance with the above argumentation, white
is part of the meaning of swan.12 The centrality of the attribute [white] can be
demonstrated by the fact that most English speakers will mention white very early,
if not first, when asked to mention a few attributes of swan. However, the
reciprocity of the structures in this case is only true to a limited extent. It seems
very unlikely that [swan] would be ranked very highly as an attribute13 of white, at
11
The cognitive status of attributes is debatable. Medin and Barsalou (1987:460-61) commit themselves to a
position in which they see attributes as being convenient fictions that allow one to bring out relationships
among concepts. A similar view is expressed by Ungerer and Schmid (1996).
12
In fact, it could be argued that the attribute [white] is part of our overall model of SWAN in the northern
hemisphere, ignoring the fact that e.g. the Mute Swan (Cygnus olor) may have tame black varieties. In
Australia, where the Black Swan (Chenopsis atratus) is frequent, the conception of SWAN is likely to be
slightly different.
13
Some readers may find it awkward to accept [swan] as an attribute of WHITE, from a technical point of
view. However, if we are to take a maximalist view of meaning seriously, then we should include this type

Providing the Tools: Meaning in Cognitive Linguistics

45

least out of context. This asymmetrical reciprocity can be explained in terms of

entrenchment and degrees of activation (cf. previous section). Although the colour
white frequently occurs in our every-day life, very few indeed of these occasions
are linked with the presence of swans. However, every time (more or less) we see a
swan, we also perceive whiteness. Of course, even in the case of whiteness, context
may restructure the category so that [swan] becomes more central e.g. white
being discussed in the context of bird watching.
The presence of recursive structures begs the question whether the meaning of
a unit is supposed to include all possible underlying knowledge. Langacker (1987:
165-166) does not exclude this possibility but he questions the usefulness of such
an assumption. Instead he suggests that we can
delimit the semantic pole [i.e. the predication, or conceptual structure (A.S)] of a
symbolic unit in the following way: a structure [X] figures directly in the
encyclopedic characterization of a symbolic unit with access node [A] only if there is
a structure of the form [AX] that has unit status; otherwise it figures only
indirectly and is not an established part of the units meaning. [emphasis original]
(Langacker 1987: 165-166)

Expressed differently, we can say that an attribute 1 is part of a concept C1 only if

1 directly contributes to the meaning of C1. However, if 1 is an attribute of a
concept C2, which, in turn, constitutes an attribute 2 to C1, then 1 cannot be seen
as part of the meaning of C1. But, as Langacker points out, such relations may
easily become activated and hence reach attribute status.14 For a slightly different
view on this matter see Croft (1993), who argues that entire domain structures are
activated.
It appears, then, that predication, domain and attribute defined in the fashion
suggested above constitute different facets of the same conceptualisation. In terms
of figure ground organisation, domain is a conceptualisation that represents
ground. Predication represents the figure and attributes are concepts that function
as peripheral dimensions or units which can be abstracted from the domain matrix,
thus being part of the ground.15
At this point it should also be mentioned that a distinction has to be made
between conceptualisations per se and linguistic categories. This is necessary since
not all concepts are linguistically coded. As demonstrated by Barsalou (1983,
1985), people can readily create ad hoc concepts for new categories which do not
have direct linguistic encoding e.g. THINGS THAT BABIES DO. In the domain of
colour, it has often been pointed out that there is no adjectival lexeme to represent
the entire domain; coloured, the best candidate, does not include black, white and

of association, and I see no a priori reason to treat this particular meaning differently from, say, lightness.
They do differ, of course, in terms of centrality, which is precisely the point I am trying to make.
14
An example of this could be the fact that swans are occasionally used to represent innocence. Although
there might be several motivations for this, one conspicuous feature is the whiteness. White, in Western
society is intimately linked with innocence. However, this motivation may not be clear to all members of the
speech community and this kind of symbolism is frequently not understood.
15
Barsalou (1992:30) offers a similar view of attributes in relation to concepts. He defines an attribute as a
concept that describes an aspect of at least some category members.

46

Chapter 2

grey. Nevertheless, we can assume the existence of the colour domain.16 In this
study, the terms conceptual category (all concepts) and lexical category
(lexicalised concepts only) serve to clarify this distinction, whenever necessary.
Having defined a few key concepts, we are now in a position to present a
figure of meaning which will recur in this dissertation. Figure 2:5 is an attempt to
illustrate a simplified representation of meaning, according to the theory sketched
above. The figure should be seen as a metaphor, and it is important to stress that I
fully sympathise with Langackers ideas of a maximalist, open-ended view of
meaning, which may be viewed as residing in processes (networks of activated
nodes) rather than in static structures. However, for the sake of convenience this
figure is preferred to other more complex and hence less transparent illustrations.
Concept:
Concrete

Abstract

6
2

Primary
domain:

10

Attribute list:
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.

Figure 2:5. Conceptual structure: attributes and primary domain

The dotted line surrounding the structure is meant to illustrate the fact that it is not
possible to draw a distinct line between essential attributes and those which are less
important in the definition of a concept. The shaded centre of the figure represents
the primary domain. Croft (1993: 344) describes the primary domain as the
domain in which the most central facts about the concept are defined. These facts
can reasonably be translated into attributes, but given the function of the domains
as backgrounds against which the figure is defined, I have found it rational to
ignore these attributes in this figure. The octagons are meant to represent noncentral attributes which are part of the meaning. The distance from the centre is
meant to be suggestive of the centrality of the attribute, and thereby its importance.
16
However, consider the views of Lyons (1995, 1999) and Lucy (1997) mentioned in Chapter 1, as far as
this assumption is concerned.

Providing the Tools: Meaning in Cognitive Linguistics

47

However, attribute structures can be affected by context and a specific context may
shift the attribute structure considerably (cf. Ungerer and Schmid 1996 and Cruse
1986). Finally, I have divided the concept into two halves, one concrete and one
abstract. This division is not to be interpreted as an indication that I believe all
concepts can be divided in this fashion. On the contrary, reason suggests that it is
more or less impossible to make a sharp division any abstract conceptualisation is
grounded in physical, embodied experience. However, for the purpose of the
present investigation a simplification has been made. As an example we might
claim that an abstract attribute such as [dangerous] may be part of the meaning of
knife. Possibly the formation of this abstract attribute relates to experiences of other
attributes of knife, plausibly [sharp], [can cut] and others. Thus, we can presume
that the attributes themselves are connected in an intricate network model.
Furthermore, it seems reasonable to assume that these connections may lead to joint
centralisation, so that, for instance, a context which will centralise [sharp] as an
attribute of knife may also centralise [dangerous]. This network is indicated by the
dotted lines; however, these lines will not appear as regular features of this model.
It should also be stressed that, although my figure here is symmetrical, there is no
reason to assume that attribute structures are symmetrical.
The model may become clearer if we explain it by way of an example. Let us
reconsider our example Friday. As mentioned above, the primary domain of
FRIDAY is the seven-day week. Typically, this is the idea we evoke when we use the
linguistic form Friday, cf. example (2) below.
(2)

His arrest was followed by two more on Friday. (BoE: times)

In such a sentence, no other attributes apart from those residing in the primary
domain are highlighted. However, as mentioned previously, we know many more
things about Fridays. These aspects are usually culturally coded facets of
experience. Thus, they are likely to change over time, whereas we can confidently
assume that the primary domain of FRIDAY will not change that is Friday, or the
meaning of Friday out of context, is likely to be maintained. It should be observed
that the stability of Friday, as far as the primary domain is concerned, is
presumably much less characteristic of other concepts.
In the case of our example, many of the additional features that we associate
with the meaning of Friday have arisen as a result of the fact that it is the last day
of the working week. However, this cultural phenomenon is an idealisation17 since
it is not true of all occupational groups. Furthermore this is a fairly new habit.
Nevertheless, the attribute that this knowledge forms it may be labelled
tentatively as [free tomorrow] is centralised in many usages of Friday. Consider,
for instance, (3) below.
(3)

Every night is Friday night when youre a showbiz gadabout like me, but I do
try to enjoy that special Friday mood that ordinary people only get at the end of
the week. (BoE: sunnow)

17
Cf. Lakoffs (1987) notion of ICM (Idealized Cognitive Model); the point is exactly that mental models
that serve as backgrounds are to a certain degree idealised.

48

Chapter 2

It is quite clear that, contrary to (2) above, additional attributes are highlighted in
our understanding of Friday. In addition to the obvious [free tomorrow], we might
include attributes which are based on our experience of peoples habit of having a
nice dinner and a few drinks on a Friday. Since the conceptual structure is based on
experiential aspects, we can assume that there might be differences here among
people. However, the general cultural model of a Friday evening can probably be
expected to be reasonably salient to most language users.
Intriguingly, the importance of the attribute [free tomorrow] in FRIDAY can be
demonstrated by the supposed synonym of Friday, the (jocular) phrase POETS day,
cf. (4) below. Humorous though it is, it is also evidence of the conception of Friday
in (parts of) the British English speech community.
(4)

National Poetry Day was celebrated yesterday a day early. After all, every one
knows that Friday is POETS day. P Off Early Tomorrows Saturday. (BoE:
today)

This aspect of Friday is also highlighted in an expression like a Friday afternoon

car designating a car with which one experiences endless annoying problems. The
cultural model (cf. Holland and Quinn 1987 for a detailed discussion of cultural
model) behind such an expression is that the car was put together in a rush on a
Friday afternoon when people were no longer concentrating on their work and just
wanted to go home.18
The importance of the attribute structure for our understanding of a word, and
the dynamics and cultural dependence of conceptual structures can be illustrated by
(5) below.
(5)

What makes you look so sad, and moodily? with such a Friday face. (1681 W.
Robertson Phraseol. Gen. (1693) 1092) (The OED)

To the modern reader the collocation of sad and moodily, on the one hand, and
Friday face, on the other, may come as a surprise. Given our own understanding of
Fridays, we may be tempted to view this as an idiosyncratic formation. However,
the now extinct usage of the expression Friday-faced had the meaning of sadlooking. Since the cultural model behind this meaning has vanished, we have great
difficulty interpreting such a phrase correctly without extensive context. The
underlying motivation for this phrase can be found in religious customs.
Encyclopaedia Britannica tells us that
Until relatively recently, the most notable dietary law in Christianity was the Roman
Catholic prescription to abstain from eating meat on Friday.

dietary law Encyclopdia Britannica Online.

<http://search.eb.com/bol/topic?eu=117420&sctn=16> [Accessed February 2 2000].

18
It is interesting to note that the corresponding Swedish phrase is mndagsexemplar, a Monday product;
the lack of concentration appears to be central in both these models.

Providing the Tools: Meaning in Cognitive Linguistics

49

Apparently, this was a tradition which had some influence on the meaning of
Friday in England even after Catholicism had been abandoned. So much, in fact,
that it could serve as the most conspicuous feature in an expression like Fridayfaced. Accordingly, it may be claimed, with good reason, that culturally based
attributes may be of considerable importance for the interpretation of lexemes even
if the latter appear to be static from a technical-definitional point of view.19 This
old meaning of Friday presupposed a completely different attribute structure than
the current one. However, since the primary domain is the same, we have no
difficulty understanding the word on most occasions. In a similar fashion, we can
assume that colour terms, although their designata are firmly established within the
colour domain, may take on and centralise attributes outside this domain. This is
studied in greater detail in Chapters 4, 5, 6 and 7.

2.5 Conceptual processes

It was argued above that categories are frequently quite complex structures. It was
further demonstrated that a category grows through extension; normally from a
prototype of some sort. This extension may take place from a local prototype, i.e.
a non-central member of a category, in which case the overall picture of the
category may be seen as having the characteristics of family resemblances. A
pivotal concept in cognitive linguistics as far as category extension is concerned is
that of motivation (cf. Lakoff 1987). Motivation creates a non-arbitrary link
between a concept and its extensions and it is related to the background of other
existing concepts and models within a speech community. Lakoff (1987: 113)
observes that [m]otivation depends on overall characteristics of the conceptual
system, not just characteristics of the category at hand. One can justifiably include
in the notion of motivation the conventions of form-content relations that have been
established in a particular speech community. However, motivation is a very broad
concept, and could indeed be deemed a prerequisite for any kind of extension. In
view of its general character, we can argue that the concept itself has very low
explanatory power. Philosophically, on the other hand, it is important since it marks
a position radically different from a traditional and disembodied view. The
motivation for the extension of colour categories is analysed in greater detail here,
primarily in terms of two conceptual processes metaphor and metonymy.
Cognitive linguistics assumes that these two conceptual processes are of
special importance. Theories of metaphor and metonymy go back as far as ancient
Greece and Rome (cf. Warren 1995 and Koch 1999) and traditionally the processes
have been regarded as rhetorical tropes. They were regarded as merely figures of
speech, deviating from normal language usage. However, through the pioneering
work of Lakoff and Johnson (1980)20 the conceptual nature of these processes has
become apparent and has received a great deal of attention. Their importance goes

19

For a detailed study of attribute evolution in a concept, see Perssons (1994) article on the development of
the meaning of the word spinster.
20
The work of Lakoff and Johnson was partly inspired by Reddys (1979) observations about the existence
of a conduit metaphor.

50

Chapter 2

beyond the mere extension of categories. Being conceptual processes they are
crucial to our understanding and structuring of reality.
The purpose of the present section is to provide the uninitiated reader with a
brief introduction to the nature of metaphor and metonymy; the references are
meant to guide the reader to more substantial analyses of these processes. Of the
two, metaphor attracted much more interest initially, but in the last few years the
study of conceptual metonymy has developed rapidly. In my study, metonymy is of
central importance, hence the section devoted to it is considerably longer. This
section also includes two Langackerian notions closely related to metonymy:
reference-point construction and active zone. Recently, several scholars (e.g.
Goossens 1995, Grady 1999 and Barcelona 2000) have observed that the relation
between metaphor and metonymy can be quite problematic and that there is often
an intricate interplay between the two. In my presentation below I ignore these
observations for the sake of brevity.
2.5.1 Metaphor
Although not the first to observe that there may be more to metaphor and
metonymy than their merely breaking the rules of grammar,21 Lakoff and Johnson
(1980) (henceforth L&J) were the first to systematise linguistic material into a
coherent theory of conceptual metaphors. They showed that systematic analysis of
language use could indicate the presence of an underlying metaphoric conceptual
process which can explain whole collections of expressions. For instance, they
suggest (1980:4) that the expressions in (6) below point to an underlying
conceptual metaphor of +ARGUMENT IS WAR+.
(6)

Your claims are indefensible.

He attacked every weak point in my argument.
His criticisms were right on target.
I demolished his argument.

However, in the view advocated by L&J, we not only use metaphors of war to
account for aspects of an argument, we in fact understand an argument as being
structured according to the same principles as obtain in war, or even more radically
(L&J: 5): the essence of metaphor is understanding and experiencing one kind of
thing in terms of another [italics original].
From a cognitive perspective, our understanding of the world is based on
bodily experience. Whenever there is a conceived similarity of some sort between
two types of experience we may categorise them as being similar (cf. above).
However, in the case of metaphor, there is a marked difference from normal
categorisation in that the notion of similarity crosses different domains of
experience. From the point of view of experiential realism described above, it
follows that abstract reasoning has to be grounded in bodily, concrete experience
and thus metaphorical structuring of abstract entities is a very common device.
Technically, metaphor has been described as the mapping of one domain onto
another. The source or donor domain is the domain that is mapped, and the target
21

Jakobson (1956) is mentioned by Koch (1999) as one of the first scholars to do so.

Providing the Tools: Meaning in Cognitive Linguistics

51

(domain), as the name suggests is the recipient of the mapping.22 In the case of the
+ARGUMENT IS WAR+ metaphor, ARGUMENTATION is the target domain and WAR is
the source domain. Through the metaphorical mapping the target domain is
structured in accordance with the source domain, which means that certain aspects
of the source domain are hidden whereas others are emphasised. Thus, the effect of
a conceptual metaphor is that the mapping transfers aspects of the source domain
onto the target. For instance, the +ARGUMENT IS WAR+ metaphor imposes on us a
picture of the other party involved as an opponent/enemy. We understand our
verbal interaction as attacking or defending positions. A crucial feature, however, is
that metaphoric mapping is always partial, not total. Were it not so, the source
concept/domain would actually be the other (cf. L&J: 13) and we could talk about
normal categorisation. In the above-mentioned metaphor there are numerous
aspects of argumentation which have no correspondence in the WAR domain and,
conversely, the structure of the WAR domain is not fully covered by aspects of the
ARGUMENTATION domain. The metaphorical process is highly creative but, as far as
conceptual metaphors are concerned, there is one important characteristic which
should be mentioned that constrains the metaphorical process in some fashion. The
Invariance Hypothesis (Lakoff 1990: 54) claims that [m]etaphorical mappings
preserve the cognitive topology (that is, the image-schema structure) of the source
domain.23 This principle or hypothesis, depending on the weight we want to give
it, thus imposes some constraints on the mapping not every domain is suitable as
source for, say, TIME and at the same time it provides a structural explanation of
metaphoric structures.
Once a metaphoric mapping becomes established in a speech community in
terms of systematic exploitation, we can say that the target domain has been
metaphorically structured in accordance with the source domain. As demonstrated
by L&J and numerous other researchers, a substantial number of these
metaphorical mappings are so well established and entrenched dead in
traditional terminology that the average language user makes use of them
unaware that they are metaphoric in nature. At this stage we can truly say that we
have conceptualised one thing in terms of another.
2.5.2 Metonymy
As mentioned above, much of the early research in cognitive linguistics was
centred around metaphor and categorisation. Metonymy was largely given a
marginal position. However, in recent years, metonymy has attracted more interest.
The publication of two new edited volumes devoted to metonymy (Panther and
Radden 1999) and the interplay between metaphor and metonymy (Barcelona
2000) testify to the recently awakened interest in this conceptual process. From the
large number of articles that have discussed the basic pattern of metonymy, I draw
mostly on the analyses performed by Kvecses and Radden (1998) and Radden and
22

In the discussion of an individual linguistic metaphor it is also customary to identify the vehicle, i.e. the
form that instigates the metaphorical process in the interpreter, cf. Fauconnier (1994), Warren (1995). The
italicised words in (6) are vehicles.
23
For a critical view, see Brugman (1990).

52

Chapter 2

Kvecses (1999). However, other influential articles and books are those by Croft
(1993), Dirven (1993), Warren (1995), and Fass (1997). A common denominator in
these latter works is their focus on the difference between metonymy and metaphor.
Lakoff and Johnson (1980:36) stress that they are different processes and define the
difference between metaphor and metonymy in the following way:
Metaphor and metonymy are different kinds of processes. Metaphor is principally a
way of conceiving of one thing in terms of another, and its primary function is
understanding. Metonymy, on the other hand, has primarily a referential function, that
is, it allows us to use one entity to stand for another. But metonymy is not merely a
referential device. It also serves the function of providing understanding. [italics
original]

The referential function of metonymy was long considered to be of primary

importance, cf. for example, Nunberg (1978), but an increasing amount of interest
has come to be focussed on the non-referential aspects of metonymy. Nevertheless,
many still consider the referential function to be central, e.g., Warren (1999).
The traditional definition of metonymy, as in the quotation from the OED
below, hinges on the two things being closely related.
A figure of speech which consists in substituting for the name of a thing the name of
an attribute of it or of something closely related. (OED: metonymy)

Closely related needs further definition and some writers have used the term
contiguity to account for a metonymic relation, for instance Ullmann (1957) and
Dirven (1993). However, the introduction of a new term does not solve the problem
unless it is properly defined. In the last few years it seems that general agreement
has been reached that metonymy can be seen as mapping within a single conceptual
cognitive unit. Lakoff (1987)24 and Lakoff and Turner (1989) suggest domain,
Croft (1993) domain matrix, Radden and Kvecses (1999) ICM and Blank (1999)
and Koch (1999) frame. These terminological differences should not obscure the
fact that basically the terms refer to the same idea.25 The idea, however we name
the unit in which metonymy occurs, specifies that metonymy is a conceptual
phenomenon. Radden and Kvecses (1999: 21) give the following definition of
metonymy, and presumably it can be used with the appropriate terminological
changes to account for similar notions of metonymy within cognitive linguistics:
Metonymy is a cognitive process in which one conceptual entity, the vehicle,
provides mental access to another conceptual entity, the target, within the same
idealized cognitive model.

24

To be more precise, Lakoff (1987: 288) says: A metonymic mapping occurs within a single conceptual
domain, which is structured by an ICM. It seems that some authors use domain and ICM synonymously
whereas Lakoff himself seems to consider ICM a structuring principle of a domain. Whatever the difference,
it must be a subtle one.
25
Rather, they are symptomatic of the terminological chaos that plagues some aspects of cognitive grammar.
However, Koch (1999: 152-153) points to some important differences between the frame approach and the
domain-matrix approach.

Providing the Tools: Meaning in Cognitive Linguistics

53

From this point of view, the difference between metonymy and metaphor can be
said to reside in the conceptual distance between target and source: metaphor is
mapping between domains, whereas metonymy is a process which operates within
a conceptual domain.
Langacker (1993) sees metonymy as being grounded in one general
phenomenon of cognitive processing: that of reference-point constructions.
According to him (1993: 30), metonymy is prevalent because of the fundamental
nature of the reference-point construction, and, consequently, metonymy serves a
communicative as well as a cognitive function:
By virtue of our reference-point ability, a well-chosen metonymic expression lets us
mention one entity that is salient and easily coded, and thereby evoke essentially
automatically a target that is either of lesser interest or harder to name.

As Figure 2:6 shows, the reference-point is used to locate another entity which is
somehow related to it.

R
D

C = conceptualiser
R = reference point
T = target
D = dominion
= mental path

Figure 2:6. Aspects of reference point abilities. (Based on Langacker 1993: 6)

Reference-point reasoning is dynamic in Langackers model; the initial referencepoint (R) retreats in favour of the target (T), which may then serve as the referencepoint for reaching a new target. To exemplify of the dynamics, Langacker
(1993:26) mentions nested locatives as in (7) below, where a series of reference
points are used to reach the target.
(7)

Your copy of Women, Fire, and Dangerous Things is downstairs in the study in
the bookcase on the bottom shelf next to the Illustrated Encyclopedia of
Glottochronology.

The wider implication of reference-point constructions is demonstrated by

Langackers analysis of possessives and the notion of topic using this construct.
The aspect of salience is also assumed by Croft (1993: 348) who suggests that
the conceptual effect of metonymy can be described as domain highlighting.
Barcelona (2000: 4) advocates a similar view, but sees the process as a special case
of what Langacker has called activation. If we go back to Langackers (1993)

54

Chapter 2

article, he argues that there are certain principles of salience to be identified, for
instance human over non-human. This seems to make sense: if the function is to
facilitate communication and highlight important features, then it seems logical that
they are salient features. However, the contextual salience of a given situation may
override these general principles. An instance of this is the notorious ham sandwich
(Nunberg 1978: 41, Lakoff and Johnson 1980: 35, etc.). Its salience can arguably
be seen as markedly high, from the point of view of the waitresses.
(8)

The ham sandwich is getting impatient for his check.

However, the situational boundedness of this type of expression makes it a less

good candidate for lexicalisation (cf. Seto 1999: 105). Obviously, the situation is
not stable or typical enough to yield so much entrenchment. This does not preclude
the fact that this metonymy +DISH FOR CUSTOMER+ and similar ones are quite
productive in various contexts (cf. Clark 1992). Radden and Kvecses (1999),
following Langacker (1993), list a number of principles, cognitive and
communicative, which account for the choice of vehicle in cases of metonymy. As
important factors they identify human experience, perceptual selectivity, cultural
preferences and the principle of relevance, all of which can explain the principles
underlying the choice of vehicles.
As was argued above, metonymy requires a relationship of contiguity
between the target and the vehicle. Furthermore, it was argued that contiguity could
be defined as a relationship existing inside an ICM or domain. Radden and
Kvecses (1999) take this one step further and identify two general configurations
to which most metonymies can be transferred:
Whole ICM and its parts
Parts of an ICM
The first type includes examples such as
(9)
(10)
(11)

England for Great Britain (+PART OF A THING FOR THE WHOLE THING+)
They went to the altar for they got married (+SUBEVENT FOR WHOLE EVENT+)
Hoover for any vacuum cleaner (+MEMBER OF A CATEGORY FOR THE
CATEGORY+)

These few examples do not do justice to Radden and Kvecses intricate analysis,
but they serve as good starting-points for a discussion. What is at the centre here is
the relationship between a whole and its parts. Typically, a relationship of this kind
applies to things and their parts but as examples (10) and (11) illustrate, this is also
applicable to events and categories. Radden and Kvecses list 21 different
relationships of ten general types belonging to the part-whole configuration.
The second configuration that is metonymy-producing is characterised by
conceptual entities being related with respect to the whole ICM. This
characteristically applies to event situations and the participants therein. Radden
and Kvecses (1999: 36) observe that [p]art for part metonymies tend to build on
the interaction between a relation and one of the things participating in the

Providing the Tools: Meaning in Cognitive Linguistics

55

relation. Typical types of ICMs are the Action ICM, the Causation ICM, the
Location ICM, etc. Examples of these include
(12)
(13)
(14)

To blanket the bed (+OBJECT FOR ACTION+)

Slow road for slow traffic resulting from the poor state of the road (+EFFECT
FOR CAUSE+)
The whole town showed up for the people (+PLACE FOR INHABITANTS+)

Radden and Kvecses identify ten such types of part-part metonymies and list 28
examples of such relations. In my analysis of colour term usage I will essentially
follow the ideas outlined above. However, one terminological difference is that I
prefer the term domain to ICM.
Finally, let me introduce another construct of Langackers which can be
viewed as a special case of metonymy active zone.26 The active zone of an entity
is described by Langacker as representing the parts of it that most directly
participate in the relationship that is depicted. Consider the sentences below.
(15)

I could hear a piano in the background. (BoE: usbooks)

(16)

Lawrence was always out polishing his car. (BoE: today)

As observed by Langacker, there is a discrepancy here between the meaning of the

lexeme and what is actually referred to. We cannot actually hear an object (like a
piano); we hear sounds, and when we polish a car it is not the whole car, but the car
enamel that we polish. As these examples demonstrate this situation is quite
normal.27 From this observation, an active zone can be described as the portion of
an evoked predication that participates directly in some relation. In other words, it
could be claimed that the whole stands for some, in the specified context, essential
part, which, thus, is only referred to indirectly; i.e., a conceptual metonymy of the
type +THE WHOLE STANDS FOR A PART+.28 The active zone phenomenon is
particularly relevant in the analysis of the naming function of colour terms. The
phrase blue pencil is a case in point. In a normal reading blue does not refer to the
colour of the whole pencil but to the colour of the lead, its active zone. Thus, it is
only the active zone that is modified by the colour adjective. This type of usage is
considered more closely in Chapter 4.
In colour semantics, metonymy is a frequently explored means of extension.
Metonymy in the shape of reference point construction is extremely important for
the understanding of extensions of colour terms in phrases such as red hair, where
the term refers to very poor examples of the colour (Chapter 4). Metonymy is also
the basic process through which colour terms acquire figurative meaning (Chapter
7).

26

The construct active zone is treated by Langacker in a number of articles and books: Langacker (1984,
1987: 271-274, 1991a, 1991b, 1993 and 1999). My summary of this notion draws on the last four in
particular.
27
Other researchers have addressed this phenomenon; Cruse (1986) treats it as contextual modulation.
28
Traditionally this type of metonymy has been called synecdoche. Cf. also Seto (1999).

56

Chapter 2

2.6 Two perspectives of meaning: semasiology and onomasiology

Intimately linked to observable phenomena in the studies of linguistic and
conceptual categories are the two perspectives from which we approach studies of
meaning. As pointed out by Geeraerts et al. (1994), the two perspectives present the
investigator with different types of phenomena.

The semasiological perspective takes a linguistic form as its

starting point and is concerned with the extension and
delimitation of it. This suggests that questions regarding
prototypes, polysemy, vagueness and homonymy are dealt with
from this point of view.

The onomasiological perspective, on the other hand, takes a

concept as its starting point, and is concerned with how it can be
expressed/codified. Thus, questions of synonymy, hyponymy,
and levels of representation are of prime interest.

Grlach (1997: 120-121) observes that these perspectives correspond to the

speaker-hearer situation:
The contrast between onomasiology and semasiology can be seen as reflecting the
achievement of the speakers in expressing in linguistic form what they wish to
communicate, and that of listeners in decoding the message.

The speakers view in a conversation is that of the onomasiological perspective;

she/he has to determine how best to dress a particular thought. The hearer, in
contrast, has a semasiological perspective; he/she will have to find the best
interpretation of a presented form. Using the semiotic triangle of Ogden and
Richards (1972), we can summarise the perspectives as in Figure 2:7 below.29
concept or meaning
B

SIGN
A
form

C
entity in conceptual and
experiential world

Figure 2:7. Two perspectives of meaning seen in the semiotic triangle.

29

Here I have used Dirven and Verspoors (1998:28) modification of the semiotic triangle, as this version is
more suitable in the context of cognitive linguistics. Ogden and Richards (1972:11) use the following labels:
A = symbol, B = thought or reference, C = referent.

Providing the Tools: Meaning in Cognitive Linguistics

57

As was mentioned in Chapter 1, in the realm of colour terminology, traditional

research has primarily been concerned with naming, i.e., onomasiology. However,
MacLaury (e.g. 1997) has developed a methodology according to which both
aspects are accounted for. His improved methodology has made it possible for him
to establish a lexical relation which has not previously been identified in the
literature that of coextension (cf. footnote 15 p.23). Another study that has
successfully explored the two perspectives in combination is the above-mentioned
work by Geeraerts et al. (1994), in which the domain of clothes is analysed.
In what follows, I consider some aspects of these perspectives from the point
of view of cognitive grammar. These aspects recur later in my analysis of colour
terms usage. For the present, my presentation is restricted to some general
observations.
2.6.1 The semasiological perspective: important sense relations
As stated above the semasiological perspective is that of the hearer, and is
characterised by his/her encounter with a form. One immediate problem which
appears is that of vagueness and ambiguity. Vagueness and ambiguity are related
phenomena but are distinguished in so far as vagueness is considered to be related
to one meaning (monosemy), whereas ambiguity is related to several senses, in
particular the accidental phonological merging of two unrelated forms i.e.
homonymy. Figure 2:8 represents and summarises some crucial aspects of the
semasiological perspective.
FORM

One form

PROBLEM

SENSE RELATIONS
Homonymy
(Unrelated senses)

LEXEME STATUS
Two words or more

Ambiguity

Vagueness

Polysemy

Monosemy

(Related senses)

(One sense)

One word

Figure 2:8. Aspects of the semasiological perspective.

Traditionally, efforts have been made to distinguish between ambiguity and

vagueness. Taylor (1995) discusses the difference between polysemy and
monosemy, and claims initially that, typically, polysemous senses refer to different

58

Chapter 2

domains, and by virtue of that a polysemous word can be said to be ambiguous.

However, he later qualifies this statement and notes that (p 100) [a]n item can still
be polysemous even if its different meanings need to be characterized against the
same domain. Using the classical instrument of zeugma, Taylor tries to
demonstrate this latter point. However, he (p 102) is forced to confess that it takes
a sophisticated (and cooperative!) informant to detect the zeugma in a sentence
like (17) below.
(17)

Both the ceiling and the bookcase are high.

A question which naturally arises is: Is it really a matter of polysemy if only

sophisticated informants can detect it? To my mind, the answer has to be no.
Tuggy (1993) demonstrates that, in a cognitive framework, we can best
consider the question of ambiguity and vagueness as a continuum or cline in which
polysemy occupies a middle position. Following Tuggys example, we can
consider the examples bank, paint and aunt. In terms of the Langackerian notions
of schema and entrenchment, the extreme case of ambiguity homonymy will
have a very abstract schema which is hardly entrenched at all, whereas in the
prototypical case of vagueness monosemy the schema will be well entrenched
and very specific. On the other hand, the elaborations will be well entrenched in the
case of homonymy, but only marginally entrenched in the case of monosemy.
APPLY PAINT

THING

TO SURFACE

PARENTS
SISTER

FATHERS
SISTER

ARTISTIC
PAINT

RIVER
EDGE

MOTHERS
SISTER

UTILITARIAN PAINT

FINANCIAL
INSTITUTION

a. bank

b. paint

c. aunt

Figure 2:9. Examples of the continuum of ambiguity and vagueness. Based on

Tuggy (1993: 283). Paint is construed as polysemous following
Tuggys suggestion.

Providing the Tools: Meaning in Cognitive Linguistics

59

Polysemy will form a middle ground in this view. Figure 2:9 above, adapted from
Tuggy (1993:283), illustrates the cline.
The entrenchment of a unit is indicated by the thickness of the lines. We can
say that bank is homonymous, paint is polysemous and aunt monosemous. Thus
polysemy can be seen as filling out the continuum between the two extremes.
Tuggys model gives a more dynamic account of the classical problem of
vagueness vs. ambiguity than the traditional one which entailed these categories
being viewed as well-defined.
One characteristic of polysemous words is that several senses are similar
enough to create a well-entrenched schema, but at the same time the individual
senses are salient enough to afford them unit status. Possibly this can be related to a
difference in the structure of the primary domain. Accordingly, in the case of a
polysemous lexeme we have two or more senses which are well entrenched but are
profiled against different domains. However, as a rule there is one sense which has
a privileged position, the prototype. Since prototypes are experientially based, it is
fairly obvious that they are intimately linked with cultural practices and other
similar phenomena. For example, paint (cf. Figure 2:9 above) can be construed so
that either sense may be the prototype, but because of the well-entrenched schema
this does not significantly affect the structure of the category.
As far as colour terms are concerned, we saw in Chapter 1 that the prototypes,
or focal colours, are to a certain extent determined by our biology, especially in
respect to the primary basic terms: black, white, red, yellow, green and blue. Kay
and McDaniel (1978) tried to demonstrate that foci, and indeed entire categories,
could be derived from neurological responses to perception.30 An alternative way
of explaining the universal pattern presented by Berlin and Kay is that suggested by
Wierzbicka. In a number of publications (Wierzbicka 1990, 1992, 1996), she does
not question the original results of Berlin and Kays study, but she claims that the
landmarks are experientially grounded and that the universal pattern is originates in
universal landmarks, for instance the sky (in the case of BLUE).
However, my prime concern in this work is not the prototypes but the
extensions, that is, precisely the kind of phenomena that Tuggy discusses. The issue
of monosemy, polysemy and homonymy is therefore readdressed in the chapters
that follow. Briefly, Chapter 4 is concerned with vagueness as this appears to be a
phenomenon related to certain uses of Primary BCTs. Chapter 5, addressing ECTs,
is particularly concerned with polysemy, and finally Chapter 7, dealing with
figurative usage, questions whether some of the figurative uses have reached a
stage where we should talk about homonymy rather than polysemy. As a final note
it should be mentioned that one reason the issue of polysemy is only a marginal
problem in the case of BCTs, is probably inherent in the definition of BCTs: one of
B&Ks original criteria explicitly eliminates all cases of polysemous terms.
2.6.2 The onomasiological perspective
As mentioned earlier, the onomasiological perspective is closely linked to the
speaker situation. Starting with a concept or, more vaguely, an idea, the speaker can
30

Dedrick (1998) gives a detailed presentation from a philosophical perspective.

60

Chapter 2

choose to refer to it in a number of different ways. Thus, this perspective is related

to the ideas of sense and reference31, in that we can denote the referent through a
number of different expressions, each with a different sense. From a cognitive
perspective, this phenomenon is frequently analysed in terms of construal.
Construal means the way we choose to view a particular situation or entity. In
Langackers analysis, construal is a complex mental phenomenon which includes a
number of cognitive abilities (Langacker 1999: 5). One such ability is zooming in,
which correlates with specificity. We can choose to portray a situation or a thing at
different degrees of specificity, which will include a corresponding inclusion of
details. Consider the chains of specificity below (from Langacker 1999).
(18)
(19)

Substance liquid beverage wine [ red wine] Chianti.

Do act move run sprint.

Viewed in terms of vagueness, one can reasonably claim that a general term is
usually much more vague than a specific one: Chianti is arguably less vague than
wine. It seems, however, that the preferred level of specificity in most types of
communication is not that which is most specific. In an early influential article,
Brown (1958) suggests that most conversation takes place at the middle level of
specificity. This preferred level has later been termed the basic level. Rosch et al.
(1976) tested the existence of this notion experimentally and found that it can be
defined in terms of a number of different parameters. The most important aspect
was that of attributes: They could demonstrate that the number of attributes shared
by members of a category is maximised at the basic level and, at the same time, the
basic level minimises the number of attributes shared with other categories. Lakoff
(1987: 47) claims that basic level categories are basic in four respects: perception
(a single mental image), function (general motor program), communication
(shortest, most commonly used) and knowledge organisation (maximise
attributes). In Lakoffs view, basic level knowledge is organised primarily around
part-whole divisions; attributes of basic-level categories are primarily of the partwhole type.
The sense relation that holds between the various levels is usually termed
hyponymy. The general term in a hyponymous relationship is called
superordinate, the specific term hyponym. In this thesis I will use the general
term hyponymy for any kind of inclusion relation. In the domain of colour terms we
have clear cases of hyponymy structures. Consider the diagram below, based on
some of the colour terms under investigation.
RED

MAROON

31

MAGENTA

CRIMSON

Reference and referent are here taken to be mental phenomena.

VERMILION

SCARLET

Providing the Tools: Meaning in Cognitive Linguistics

61

Hyponymy is a non-reversible, unidirectional relation in that example (20) entails

(21) but not vice versa.
(20)
(21)

He painted the car scarlet.

He painted the car red.

However, the nature of the colour domain creates some difficulty in the otherwise
straightforward relation of hyponymy. Hues that are lexicalised as ECTs and occur
frequently are typically borderline cases and occupy slots between two BCTs. The
colour maroon is a case in point. Consider these definitions of maroon taken from
the OED and Longman:
A particular kind of brownish-crimson or claret colour. (The OED)
A very dark red-brown colour. (Longman)
Although maroon is readily categorised as a type of red under normal
circumstances, it is not true that every nuance of maroon is red. Consider example
(22) below:
(22)

Introduced to Britain as long ago as 1835, bestowed with an Award of Merit

1938, this plant was all but lost to cultivation until fairly recently. Often
referred to as Chocolate Cosmos because the deep maroon, almost black,
velvety flowers have more than a hint of the aroma of chocolate.
(BoE: ukephem)

It seems that we are concerned here with a hue that few people would call red. One
way of solving this problem would be to take the same view as Biggam (1997: 31)
does, who argues that a term can be the hyponym of two superordinates. This
seems to be a reasonable solution in the domain of colour, and is probably valid in
other domains.
In the domain of colour, it is frequently assumed that BCTs correspond to
basic level terms. Although this might appear to be the case, it must be borne in
mind that there seem to be other principles involved in the case of colour notably
the (supposed) biological predestination of this categorisation. However, the picture
appears to be even more complex. To my mind, it seems that a term like blond(e)
could be said to be basic level and a colour term, although it does not qualify as a
BCT. Arguably, the term is frequent enough (which implies deep entrenchment) to
be considered basic level. Dedrick (1998) and Lyons (1999) discuss this paradox as
a general issue. It is also interesting to note that Rosch et al. (1976) suggest that the
expertise that a person may possess can affect basic level categorisation so that the
basic level may not be the same for two people. However, in the domain of colour
terms, some researchers, notably MacLaury,32 do not allow such dynamism,
probably on the grounds of the assumed biological constraint.
Another sense relation which falls within the onomasiological perspective is
that of synonymy. It has long been recognised that synonymy is a problematic
32

Cf. MacLaury (1997, Appendix IV, 419-429).

62

Chapter 2

concept as there seem to be very few words which are truly synonymous. Cruse
(1986) proposes that synonymy can be conceived of as a scale. The zero point
would be what Cruse (1986: 268) terms absolute synonymity the definition of
which would be that two lexical units would be absolute synonyms (i.e. would
have identical meanings) if and only if all their contextual relations [] were
identical. Given this definition, we can assume that it is virtually impossible to
find absolute synonyms unless we have a loose interpretation of all their
contextual relations. This is clearly demonstrated by Cruse. Thus, we can safely
claim that most cases of synonymy can be considered to be partial.
From an encyclopaedic perspective this is obvious, since there will always be
attributes, if peripheral, distinguishing the meaning of one form from another.
Persson (1990) suggests an alternative analysis of partial synonyms. On the basis of
corpus research, he advocates a position in which synonymy is seen as a special
case of hyponymy. Persson observes that synonyms typically have a shared central
meaning, something that is also characteristic of co-hyponyms. As with hyponymy,
this suggests that different forms may describe a given entity the choice of form
that has to be made by the speaker. In the case of colour terms this problem arises
particularly in the choice of specific terms. For instance, a particular purplish-red
nuance may possibly be described as crimson, magenta, maroon or purplish-red.
The question to be asked is whether these terms are to be treated as synonymous. In
Chapter 5 we will investigate further the extent to which Elaborate Colour Terms
can be said to be synonymous, and also whether there are contextual factors that
determine the choice of term.33
In a wide sense it can be argued that the onomasiological perspective is
concerned with lexical fields. Dirven and Verspoor (1998: 37) loosely define a
lexical field as a collection of words that all name things in the same conceptual
domain. This definition appears to result in a double structure:
A conceptual domain against which concepts are defined.
A lexical field in which lexemes name these entities.
The important difference would appear to be one which has caused such great
controversy in colour terms studies. Although colour terms typically refer to the
colour domain, they may have much wider designations. However, Berlin and Kay
(1969) were not interested in the designation of colour terms (the lexical field) as
much as the structure of the conceptual domain COLOUR.
From the point of view of the onomasiological perspective, the relevant
terminology can be summarised tentatively in the way suggested in Figure 2:10
below. The circle represents the idea that the speaker wants to express. The
rectangles represent predications of different levels of specificity; the two smallest
rectangles are almost synonymous and include the particular idea to be expressed.
33
Kay (1999: 81-82) discusses synonymy in colour space and observes that a colour word may be replaced
by another word for the same category. In the transition time these terms will appear to be synonymous. He
further claims that in his own speech tan, beige; chartreuse, lime and turquoise, aqua represent such
synonymous pairs in free variation. The question is whether this could be regarded as evidence of
synonymy. This is discussed further in Chapter 5. Another problem to be discussed later is the relation
between colour definitions as such and usage.

Providing the Tools: Meaning in Cognitive Linguistics

63

The growing size of the rectangles illustrates a hyponymy structure. From the
domain of colour terms we may let coloured represent a superordinate term, red a
basic level term and scarlet and crimson hyponyms. If there were two rectangles
which had perfect overlap, they would be considered absolute synonyms.

Hyponym1
Concept

Hyponym2
Basic level term
Superordinate

Figure 2:10. The onomasiological perspective.

2.7 Summary
The aim of the present chapter was to provide a brief introduction to the analysis of
meaning in the theory of cognitive linguistics. Many of these concepts are given
further clarification in the chapters that follow. The basic terminology used in this
dissertation has been presented, and the stage is now set for the study proper.

Chapter 3

Colour Terms in the Bank of English:

Some Statistics

3.1 Introduction
The primary aim of this chapter is to present some general observations pertaining
to the usage of colour terms in English on the basis of frequencies obtained in the
BoE. Since the figures presented in this chapter receive closer scrutiny and
theoretical treatment in the chapters to follow, we can regard this chapter as a point
of departure for the analysis of colour terms in context. Most of this chapter is
concerned with the frequency of colour terms in relation to various parameters.
Thus, primarily my observations are subjected to statistical analysis, using some
elementary tools. However, some brief discussions of semantic issues are also
offered. The use of statistical calculations in linguistic studies related to semantics
is a rather recent phenomenon and the value of the method was briefly discussed in
Section 0.3.3. Statistics are useful in that they can serve as good reference points
for comparisons with other researchers results. This is one of my aims here, and as
far as possible, I compare my material with previous research in the area of colour
semantics or other studies which mention related observations.

3.2 Word frequency

The frequency of colour terms in texts has repeatedly been proposed as one useful
criterion in judging whether or not a term can be regarded as a BCT. It has been
argued that the more frequent the occurrence the greater the salience of the term.
Hays et al. (1972) were the first to make this claim, and it has been repeated by
several others. Corbett and Davies (1995), in their evaluation of methods employed
in identifying BCTs, consider word frequency in texts and they analyse material
from a number of different languages. In the case of English, they use the figures
provided by Johansson and Hoflands (1989) analysis of the LOB corpus. It should
be pointed out that, while doing so, they only consider colour words in their
adjectival function. The advantage of such an approach, Corbett and Davies (1995:
329) claim, is that it allows more reliable comparisons with other languages. I am
far from convinced that this is necessarily the case. Although colour terms are
prototypically adjectival and are said to be prototypical adjectives (cf. Dixon 1982:
46), it is quite clear that colour terms are also nominal in character, and my view in
the matter is that it would be wise to consider colour terms irrespective of their
formal label if we want to achieve a clear picture of this particular semantic
domain. From their examination of the LOB figures, Corbett and Davies report a
high correlation between frequency and the evolutionary hierarchy suggested by
Berlin and Kay (tau = 0.77, p < 0.001).1 Similar high correlations between
1

In their article, Corbett and Davies report correlation in terms of Kendalls tau, rather than the more
common Spearmans rho. They argue that Kendalls tau is more appropriate in view of the many ties that
appear in Berlin and Kays hierarchical order of colour terms (Corbett and Davies 1995: 309-312). However,
65

66

Chapter 3

frequency and the B&K hierarchy have been observed by McManus (1983, 1997)
in his studies of poetry and various computerised databases.2 Generally, the
analysts have seen this as evidence of the accuracy of Berlin and Kays theory.3
Whereas the frequencies of BCTs have been studied previously, very few
studies have been devoted to ECTs regarding their frequency in texts. To the best
of my knowledge, the only two studies addressing this subject have been Franks
(1990) on advertising, and my own, Steinvall (2000), on Victorian poetry. Thus this
is very much unexplored territory, and consequently there is no other material with
which comparisons can be made. This is also one reason why the focus in this
chapter is on BCTs. ECTs are dealt with in greater detail in Chapters 5 and 6.
In Figures 3:1 and 3:2 below the data of the colour terms under investigation
are presented. It should be pointed out that the frequencies of the terms gold,
golden and silver have been calculated on the basis of a random sample of one
thousand tokens.4 An asterisk beside the estimated terms indicates this in the
diagram.

apart from the first comparison, in this dissertation I will use Spearmans rho, which can be corrected for
tied data (Zar 1996: 390).
2
There are some aspects of McManus (1997) which, to my mind, drastically reduce the value of his study.
First, the corpora are ill-defined and therefore it is difficult to determine what they represent. For instance, it
appears that some corpora contain abstracts rather than real texts. Second, and much more seriously,
McManus does not seem to be bothered by the problem of polysemy and homonymy. In an off-hand manner
he (1997: 369) notes that [i]t is possible that grey/gray is anomalous [i.e. more frequent than its position in
the B&K hierarchy would predict] because of its usage as a proper name, although Black, White, Green, and
Brown are also common proper names. This failure suggests that large question marks have to be placed
over the results McManus presents.
3
As mentioned earlier (Chapter 1), Corbett and Davies do not test the validity of the theory, but only the
suitability of the methods.
4
The very high frequencies of gold, golden and silver and their polysemous character precluded a manual
counting of the tokens that have colour designation. Instead the frequencies of these terms in their colour
sense were calculated on the basis of a random sample of a thousand tokens which was analysed in order to
identify the number of colour designations (and from there derived meanings). The proportion of colour
terms in the population was estimated by the proportion of colour terms in the sample ( 95 % confidence
interval), assuming a binomial distribution. The following procedure was applied, cf. Zar (1996: 521-526).
X
The proportion was calculated p =
, where n = sample size (here 1000), and X = the size of the colour
n
category in the sample. The size of the entire colour category (Y) was then calculated Y = p N , where N =
the entire population (i.e. number of all tokens of the word form). A determination of a 95% confidence
interval for the categories was also calculated. The lower confidence limit for p is
L1 =
L2 =

X
, where v1 = 2(n X + 1) , and v2 = 2 X . The upper confidence limit for p is
X + ( n X + 1) F ( 2),v1 ,v2
( X + 1) F ( 2),v ' v '

1 2

n X + ( X + 1) F ( 2),v ' ,v '

1

, where v1' = 2( X + 1) = v 2 + 2 , and v 2' = 2( n X ) = v1 2 .

My samplings gave gold: N = 43286; X = 84, Y= 3636; 95 % confidence interval = 2934 4499.
golden: N = 14646; X = 272, Y = 3984; confidence interval = 3612 3994.
silver: N = 14354; X = 253, Y= 3632; confidence interval = 3284 3642.

67

68037
3983

3779

3636

3631

3553

1384

1208

1126

*Golden

Purple

*Gold

*Silver

Orange

Cream

Scarlet

Navy (blue)

Brown

Pink

11927

Yellow

8924

13055

Grey/Gray

Blue

13113

27102

Green

Red

White

Black

28664

38912

80000
70000
60000
50000
40000
30000
20000
10000
0

74895

Tokens

Colour Terms in the Bank of English

Colour terms

748
35
35
75
76
80
93
98
108
111
124
134
136
151
174
186
194
200
231
275
302
326
339
340
427
449
459
501
518
583
632
640

1000
900
800
700
600
500
400
300
200
100
0

878

Tokens

Figure 3:1. Frequency of colour terms in the BoE, n > 1000.

Ultramarine
Chartreuse
Teal
Puce
Carmine
Fuchsia
Vermilion
Mustard
Tangerine
Fawn
Plum
Azure
Charcoal
Rust
Aqua(marine)
Magenta
Indigo
Lemon
Lavender
Peach
Emerald
Lime
Lilac
Amber
Olive
Tan
Mauve
Maroon
Violet
Rose
Turquoise
Beige
Crimson

Colour terms
Figure 3:2. Frequency of colour terms in the BoE, n < 1000.

As Figure 3:1 suggests, the same pattern of frequency materialises in the BoE as in
the studies mentioned above; that is, there is a high correlation between the
frequencies of BCTs and their position in the B&K hierarchy. In the BoE the
correlation is very high (tau = 0.834, p < 0.001), higher than that in LOB.
A few other observations can also be made here. One methodologically
interesting fact is that it is not possible to use frequency as a criterion for
distinguishing between BCTs and non-basic colour terms. The three metal colours

68

Chapter 3

golden, gold and silver are as frequent as purple and orange. Corbett and Davies
(1997: 208) make a similar remark, but they argue that frequency is a strong
separator for primary [i.e., in the case of English, black, white, red, green, yellow
and blue] versus secondary basics. In my view, this claim is slightly misleading.
Corbett and Davies arrive at this conclusion since they take the means of the token
frequency of the two categories into consideration. The use of means however,
presupposes that a categorisation in Primary BCTs and Secondary BCTs has
already been made, and if that is so there is no need for a separator. On the other
hand, we can also see that only yellow clearly violates Corbett and Davies claim in
my material and, incidentally, the same phenomenon can be detected in their own
study.
3.2.1 The problem of yellow
It seems that yellow frequently causes problems for investigators who would like to
see a correspondence between frequency and the B&K hierarchy; see, for example,
Corbett and Davies (1995) and McManus (1997). It is interesting to note that this
phenomenon is not confined to English, but seems to have wider significance: in
the lists of Hays et al. (1972) we find the words representing YELLOW deviating
from the expected pattern in French, German, Spanish, Russian and Romanian.
Corbett and Davies (1995) using different corpora confirm the patterns for Russian
and French. In a later version of their article, Corbett and Davies (1997: 206)
observe the problem of the yellow category term and concede that [w]e have no
explanation for this, but it seems to be a common problem. If we take a closer
look at Figure 3:1, however, I think that we can identify a plausible reason for the
relatively low frequency of yellow. It can be presumed that the large number of
occurrences of the colour terms gold and golden should affect the frequency of
yellow.5 Indeed, there is also the term blond(e) signifying a yellow colour in the
domains of HAIR and, more rarely, BEER.
Strangely enough, Corbett and Davies seem to have overlooked these colour
words; golden at least should have been mentioned in their presentation
Johansson and Hofland (1989: 170) report 46 instances of golden. True, not all of
these are necessarily instances of colour representation, but the same problem
arises in the cases of silver and bronze, which are considered by Corbett and
Davies. Blond(e) and gold have low scores as adjectives and could justifiably be
excluded.
If we summarise the scores for yellow, gold, golden and blond(e)6 in BoE, we
get a figure for the yellow category which is on a par with the figures for the other
elemental colour terms green and blue: 26 317. Thus it would seem that the
5
McManus (1997:370) notes that taking non-basic colour terms into consideration may explain the low
frequency of yellow. However, it is not clear how we are to understand McManuss view of semantics, as he
claims that non-basic terms are only of interest if they [non-basic colour terms] are being used as synonyms
for the eleven basic colour terms This is an odd claim given the fact that we would normally describe the
relation between BCTs and ECTs as one of hyponymy, admittedly with some special cases (see Section
2.6.2). But how is one to determine whether, for instance, gold is synonymous with yellow? And under what
circumstances is it not? Needless to say, this claim does not make sense at all from a cognitive linguistic
perspective; the terms gold and yellow would never be considered synonymous, again cf. Section 2.6.2.
6
There are 5 643 instances of blonde and blond without a capital B in the BoE.

Colour Terms in the Bank of English

69

unexpectedly low frequency of the primary BCT yellow and corresponding terms in
other languages could be due to the fact that this category, YELLOW, is not
structured in the same way as the other primary colour categories. Whereas we find
a clear predominance of the BCT in other categories, this is not true of YELLOW,
where terms denoting gold and its colour are very frequent and where there is a
special term for yellowish hair. I feel confident that we will find similar patterns in,
for instance, Russian and French. We can only speculate why this is so, but the
great importance of gold in European cultures, and elsewhere, would no doubt
provide a good explanation.
3.2.2 The proportions of colour term categories
One feature of basic level terms is that they constitute the type of words preferred
in everyday language use (cf. Brown 1958, Lakoff 1987 and Chapter 2). Although,
as pointed out previously (2.6.2), there is no immediate relation between BCTs and
basic level, it is still interesting to note that the overall pattern of BCT usage is
overwhelming in the BoE. Approximately 92 % of all tokens involving the colour
terms under investigation are BCTs.7 Figure 3:3 below gives a picture of the
domination of BCTs. It might be claimed that this picture is slightly misleading
since phrases such as charcoal grey and azure blue are counted twice, both as
ECTs (charcoal and azure) and as BCTs (grey and blue), although the actual
specific designata (the nuances the phrases refer to) occur only once. Thus it could
be argued that the figures show a slightly more dominant picture for BCTs than is
actually true. On the other hand, the aim of the figures is specifically to illustrate
the frequency of the terms, and in view of this I think it is justifiable to treat
phrases in this way.8
ECTs
7.78%

BCTs
92.22%
Figure 3:3. Proportions of tokens in the BoE: BCTs and ECTs
7

If we take into account the confidence intervals for the calculated tokens of gold, golden and silver, we get
a maximum of 92.62 % and a minimum of 91.95 % for the BCT category.
8
It should be stressed that the frequency of BCTs gives us no idea of the specificity as such of the denotata,
since many of the BCTs may be qualified by other phrases, as in, for example, bright blue or pea-green.

70

Chapter 3

There are other categories that can be explored in a similar manner, including two
categories which are easily comparable: chromatic colour terms and achromatic
colour terms. Figure 3:4 illustrates such a comparison. In the second category, that
of achromatic colour terms, I have included the terms black, white, grey/gray,
charcoal9 and silver.10 As Figure 3:4 demonstrates, every second mention of a
colour term in the BoE corpus is an achromatic colour term.11

Chromatic terms
49.52%

Achromatic
terms
50.48%

Figure 3:4. Proportions of tokens in the BoE: chromatic and achromatic terms

Clearly, the very high frequency of the terms black and white is the main reason for
this pattern, cf. Figure 3:1. This may seem somewhat surprising given the fact that
we can hardly be said to live in a world where achromatic colours have a great
impact. These figures could apparently be taken as clear indications that colour
term usage goes well beyond the merely descriptive. Consequently, it is crucial to
distinguish between achromatic colours and achromatic colour terms. Although
achromatic colour terms typically designate achromatic colours, and are defined in
this way, they are not confined to this territory. On the contrary, Figure 3:4 is a
good indication that this is not the case. Intriguingly, this seems to be a
unidirectional phenomenon; I cannot think of one occasion on which a chromatic
colour term would denote an achromatic colour nuance. It is tempting to see this as
an indication of the primacy of lightness in terms of communication.
Whether or not the pattern in Figure 3:4 is culturally grounded is open to
speculation and will remain so until we receive a vast amount of data for other
languages. However, an analysis of the material for Russian provided by Corbett
and Davies (1995) indicates that we can find a similar pattern in that language. In
Corbett and Davies list, 48.18 % of the tokens are achromatic colour terms.
Although the figure is slightly lower than that for my English material (50.48%), it
is still very high, and we should bear in mind that Russian texts are likely to include
9

The problem of double representation of phrases like charcoal grey and azure blue is of course also present
in these figures.
10
Although cream is usually defined as yellowish white (see Appendix 4), and may be treated as a
hyponym of white, I have not included it among the achromatic colour terms since the yellow element
clearly indicates that we are concerned with a chromatic colour term.
11
If we take the confidence interval of the calculated terms into account we get a maximum of 49.72 % for
chromatic terms and a minimum of 49.35 %.

Colour Terms in the Bank of English

71

far fewer references to the black and white skin dichotomy that we can assume is
fairly frequent in British and American texts. The figures are not really equivalent
since Corbett and Davies present very few non-basic terms, so a figure comparable
to mine concerning achromatic colour terms is probably slightly lower than that
given above.
There are a few possible explanations for this pattern. One explanation would
suggest that we are concerned here with a diachronic phenomenon. That is, some
collocations of black and white were formed at a time when there were few colour
terms around. Thus, they represent, as it were, relics or fossils from the past.
Another explanation suggests that we take the contrast of lightness as primary
when we categorise the colour of objects. It has been argued (MacLaury 1997, Kay
and Maffi 1999) that lightness may be more basic from a conceptual and perceptual
point of view,12 and these frequencies may support that claim. Furthermore, the
antithetical character of this dimension (white vs. black) makes it particularly
suitable in terms of categorisation of things. It would then seem that, although we
can zoom in and be more precise in our designation, there is no communicative
pressure to do so and thus we get a high frequency of achromatic terms. In Chapter
4 I try to suggest a model which may explain the pattern preference of black and
white.
3.2.3 Frequency and age
One aspect of Berlin and Kays hierarchical order of colour categorisation is that it
entails a temporal perspective. Thus, the RED category is assumed to be older than
the PURPLE one, and consequently the same assumption is usually made about the
terms, i.e. red should be older than purple. However, occasionally a BCT can lose
its basicness for some reason and be replaced by a new BCT. Forbes (1979)
demonstrates that this seems to be an on-going process in French, in which the
older term, brun, is currently losing ground to marron. Few studies have been
carried out in this area so little is known about the mechanisms involved.13
In this section, I investigate whether there is a connection between frequency
and age. As a general rule, most basic terms are as old as the oldest written records,
but there are BCTs which have developed more recently. In so far as ECTs are
concerned, Casson (1994) suggests that English went through a change in Early
Middle English during which hue became more dominant at the expense of
brightness. Thus most of the ECTs studied here developed after AD 1300. He
suggests that the quick increase in colour terms can be related to developments in
the textile and dyeing industries. However, fashion is an ever-changing industry,
and if terms were confined to this domain quite a few of them would probably die
out. In Steinvall (2000), I showed that old terms were preferred to newer,
transparent, terms by poets in the nineteenth century. Thus, the picture might be
quite complex.

12

MacLaury (1992, 1997) has claimed that colour categorisation on the basis of lightness very often
precedes colour categorisation by hue. Indeed, Casson (1997) suggests that this could be the case in English.
13
However, apart from the above-mentioned Forbes (1979, 1986), I should also mention Biggam (1997,
1998) and Kristol (1980).

72

Chapter 3

The null hypothesis suggests that there is no connection between frequency

and age, whereas my working hypothesis is that there is. As regards establishing
the age of the terms, I have used the same method as Casson; that is to say, it is
determined from the historical quotations in the OED. The full list of terms and
their age can be found in Appendix 5. A scatter plot for my variables is shown in
Figure 3:5 below. The null hypothesis could be rejected as the Spearman rank
correlation test shows that there is a significant correlation between frequency and
age (rs = 0.682, N=50, p< 0.001). Within the category of ECTs, there is also a
significant correlation between age and frequency, if not as high (rs = 0.422, N=39,
p < 0.01). Thus, we can establish that, among the terms I have chosen to
investigate, there is a clear pattern suggesting that those terms which are fairly
frequent in the BoE tend to be older than less frequent ones. Before we draw any
far-reaching conclusions we have to remember the basis for the selection of terms
elicitation lists. Therefore, we must not conclude that older terms are necessarily
more frequent than younger terms. There is no direct causal relationship: many
quite old colour terms are only marginally used in English (cf. Casson and Gardner
1992, Casson 1997). Nevertheless, this correlation does indicate that there might be
something interesting going on here, and this tendency should be borne in mind
during the more detailed study of ECTs in Chapter 5.14

14

Without anticipating later discussions, I would like to suggest that one likely explanation is that terms
which show high frequency are not confined to a single domain, say, horticulture, but can be used in many
different domains. Such an expansion of usage could reasonably correlate with age.

Colour Terms in the Bank of English

73

100000

Number of tokens in the BoE

10000

1000

100

10

1
0

200

400

600

800

1000

1200

1400

Age of the term

Figure 3:5. Scatter plot for frequency and age of terms. (y-axis logarithmic)

3.3 Colour terms as nouns

It is generally agreed that colour terms are bona fide adjectives. However, in
contrast to other such adjectives (e.g. the adjectival classes Dimension, Age and
Value, cf. Dixon 1982: 55), colour words are readily conceived of as nouns. We
have no problem in understanding the meaning of a colour term as relating to the
profile of a certain area in the colour domain. Clausner and Croft (1999: 12) make
the interesting observation that [c]olor concepts function like proper names for
different locations in the COLOR domain (the focal colors). We can say that
whenever we refer directly to the colour domain, we use colour terms in their
nominal form. This fact has had methodological implications, as mentioned earlier
in Chapter 1 (pp 25-28). Lyons (1999: 48) points out that one has to distinguish
between referential use (involving reference to the colour domain itself, i.e.
nominal use) and descriptive or attributive use (involving adjectival use). Since
context-free studies, naming colour chips for instance, imply referential use, Lyons
claims that traditional colour term studies have only explored the first type of use.
Regular language use on the other hand is primarily a matter of the second type.

74

Chapter 3

Kay (1999: 78) rejects this analysis emphatically, stating that [w]orkers in the
B&K tradition have not, to my knowledge, studied color nouns The discussion
was presented in some detail in Chapter 1, and I shall not reiterate here. My aim is
not to resolve this issue, it is much more modest: as this seems to be an area which
has been ignored (even deliberately so, recall that Corbett and Davies (1995: 329)
omit nominal usage from their account for the sake of allowing more reliable
comparisons with other languages), I intend to present the most obvious aspects
that could be obtained from the BoE in this respect.
This section, then, explores nominal usage of a sample of colour terms, to
estimate the frequency of occurrence of colour terms as nouns. For this particular
study, I did not examine all tokens, of course the size of my material precluded
such an approach. In fact, this would have been possible if I had relied on the tags
that the BoE supplies. However, although some people claim that machine tags
have a high degree of accuracy Manning and Schltze (1999: 371) mention a
figure as high as 95-97% my own impression is that the polysemous nature of
many colour terms severely affects the reliability of machine tags.15 Instead, I
chose to examine a random sample of 200 instances of a collection of 25 colour
terms. The terms that I looked at were the eleven BCTs, the six most frequent ECTs
(with the exception of golden (which cannot be used as a noun) and gold (my
sample only contained 84 occurrences), plus eight more terms, which were
randomly chosen from my selection of colour terms.16
The study demonstrated that the proportions of nominal usage differed
somewhat. In Table 3:1 below, the colour terms are ordered according to their
overall frequency in the BoE, chartreuse being the least frequent term and black
the most frequent (cf. Figures 3:1 and 3:2). Simply put, it would appear that the
more frequent a term is the lower the proportion of nominal usage. In fact this was
quite the opposite to what I had expected my working hypothesis was that
figurative usage of the most common terms would mean they were more frequently
used as nouns. One factor that may have affected this picture is the exclusion of
plurals. On the other hand, my impression is that plurals referring to colour space
are fairly infrequent.

15

As examples supporting my view, I can mention that all instances of chartreuse and mustard are tagged as
common singular nouns (NN) in the BoE, whereas all instances of lilac are tagged as adjectives (JJ).
Needless to say this is not correct from any reasonable linguistic point of view.
16
For the terms with a score lower than 200, all examples were taken into consideration. For the terms with
higher frequencies than 200, the procedure was more complicated. To obtain a random sample of instances,
the following procedure was decided on: For the BCTs which are not polysemous, i.e. in effect all but
orange, I used the BoE software, which allows the investigator to choose a random sample of any size of a
term. Words with initial capitals or all in caps were eliminated. As far as the other terms are concerned, the
material was saved in files in the Wordsmith Concordance program, so that each sentence containing a
colour instance was enumerated. With the help of the random number generation analysis tool in the
spreadsheet program Microsoft Excel, a random sample of 200 numbers was collected for each colour term
each number obtained from the tool corresponding to a particular sentence, i.e. an instance.

Colour Terms in the Bank of English

75

Table 3:1. Colour terms as nouns.

COLOUR TERM

FREQUENCY

SAMPLE SIZE

TOKENS AS NOUN

Black
White
Red
Green
Blue
Grey/Gray
Yellow
Brown
Pink
Purple
*Silver
Orange
Cream
Scarlet
Navy (blue)
Crimson
Beige
Turquoise
Amber
Lilac
Lime (green)
Peach
Indigo
Mustard
Chartreuse

74895
68037
38912
28664
27102
13113
13055
11927
8924
3779
3631
3553
1384
1208
1126
878
748
639
428
340
339
299
200
108
35

200
200
200
200
200
200
200
200
200
200
200
200
200
200
200
200
200
200
200
200
200
200
200
108
35

14
18
32
29
35
24
24
25
30
33
19
31
36
42
55
32
45
44
37
57
59
60
85
29
15

PERCENTAGE
NOUNS

7
9
16
14.5
17.5
12
12
12.5
15
16.5
9,5
15.5
18
21
27.5
16
22.5
22
18.5
28.5
29.5
30
42.5
26.9
42.9

In order to see whether the trend was an actual one, the significance of the
hypothesised correlation was tested. Before this was done, the data were
transformed to facilitate the calculation in accordance with established methods.
The X values represent the overall frequency of each colour term and the nature of
this distribution (cf. Figures 3:1 and 3:2) implied that a logarithmic transformation
was suitable.17 The Y data represent proportions and thus arcsine transformation is
called for (cf. Zar 1996: 346, and also Woods et al. 1986: 220, 245-46).18 The
calculation showed that there is a significant negative correlation, Pearsons r =
-0.85 (p < 0.001). Figure 3:6 below illustrates the distribution after the
transformations. Proportionally, colour terms with low frequency appear more
often as nouns than terms with high frequency occurrence.

17

Thus X = log X

18

Thus Y ' = arcsin Y

76

Chapter 3

0,8

Y' = arcsin square root Y

0,7
0,6
0,5
0,4
0,3
0,2
0,1
0
0

X' = log X

Figure 3:6.

The relationship between frequency of a colour term (x-axis) and

its proportion as a colour noun (y-axis). Transformed values.

The correlation between low frequency and nominal usage deserves closer
inspection. Some preliminary observations show that we can distinguish between a
number of typical instances of the nominal usage of colour terms. By far the most
common constructions are prepositional phrases, notably those containing in and
of. These two account for more than forty per cent of all instances of colour terms
used as nouns. If we take all prepositional phrases into account we get a figure as
high as sixty per cent. The examples below illustrate some typical constructions.
(1)

Knits are the big news, worn hugging the body in fine crochet or rib dresses,
cropped well above the belt in fluffy mohair and Fair Isle sweaters, or all
thrown
together for true sweater dressing in shades of chocolate, charcoal and cream.
(BoE: brmags)

(2)

then on for a swim in what Princess Margaret, during her honeymoon,

baptised the Nylon Pool, where the water, cradled by a sandbar, is the colour
and smoothness of pale chartreuse. (BoE: times)

(3)

Priced at 24.99, it is also available in navy and black (BoE: guard)

(4)

She let it ring for the machine to pick up and continued the delicate work of
spraying on the paint so that it made a seamless gradient from the lightest cobalt
on the horizon to the deepest indigo at the top. (BoE: brbooks)

As we can see in these examples, the prepositional phrases typically evoke the
colour domain more or less overtly. It would seem that the zooming in that
colour specificity entails leads in many cases to a direct reference to colour space.
The precision, as it were, decontextualises the colour from the object. As one might

Colour Terms in the Bank of English

77

have expected, phrases of this type appear to be much more common in the context
of fashion and gardening, when one is likely to discuss similar instances of an
object which differ only in respect to nuances. This is also a type of context in
which the nuance itself is of prime importance since one is trying to create a whole
of matching nuances. It appears that specific terms which are fairly infrequent seem
to be primarily used in such contexts. BCTs and some frequent ECTs, on the other
hand, are used much more freely and occur in many more nominal domains.
The relatively high frequency of the in constructions is interesting from a
theoretical point of view. Langacker (1987: 227), discussing atemporal relations,
posits that there is an IN-relation between the head and the modifier in a phrase like
red book. Arguably, an object cannot occupy a place in the colour domain, and so
Langackers solution is that the colour serves as a landmark with which the
trajectory (the object) is connected through an IN-relation. The adjectival meaning
then arises in the profiling of the relationship between a colour sensation and an
object (cf. Langacker 1991b). In a footnote, Langacker (1987: 228, fn 9) states that
[t]his analysis predicts that some languages might express a similar concept by
means of an oblique phrase (literally, in red) rather than an adjective. The
relatively high proportion of in-constructions in the noun material can be
interpreted as evidence supporting this hypothesis. However, more detailed
investigations have to be carried out in this area.
Another fact worth mentioning is that colour terms occasionally represent
count noun usage, thus designating a bounded region in the colour domain (cf.
Langacker 1991b: 29). Our normal conception of colour terms is that, in their
nominal interpretation, they represent mass nouns as in the examples above. In (5)
and (6) below, however, there can be no doubt that we are dealing with count noun
instances. It is typical of these instances that, as in examples below, they are
modified by some attribute. As shown by Svensson (1998: 166), this is quite a
common phenomenon among mass nouns when they appear as count nouns. The
function of the attribute is to provide a type specification of the mass entity, thus
creating a bounded region of the normally unbounded substance. In the examples
below, the construction leads us to believe that there are other types of amber and
peach, respectively.
(5)

Bundaberg rum is one of the worlds best dark rums. It glows a deep amber in
the glass with a rich and penetrating aroma. (BoE: oznews)

(6)

Kim teamed this with a blue floral border and sunny peach paper above the
dado, while the stair rails and hall cupboard have been dragged in a soft peach,
and the paintwork is offset by the mid-blue carpet. (BoE: brmags)

By and large, this type of construction is infrequent. Without a premodifying

adjective the number of instances of colour terms as count nouns is very low
indeed; in the pilot sample of 5 000 instances (25 * 200) there were only two (2)
such constructions, one of which is given in (7).

78

Chapter 3

(7)

And you also know that you have to make your colors evolve so that if you
have an indigo, but you want indigo to continue, you cant show the same
indigo. Well, youyou could, but you really should try to show a fresh indigo.
So you may want to make it a bit more purple. (BoE: npr)

This clearly indicates that people very seldom find reason to construe parts of
colour space as bounded without first further specifying it in terms of a type.
To summarise the section on nominal use of colour terms, we can state that
the prima facie relation between low frequency and nominal usage seems to be
related to the type of discourse in which these rare terms are used. However, there
is room and need for more detailed research in this particular field.

3.4 Colour colour combinations

It has been observed by some authors, e.g. Levinson (forthcoming), that there is a
kind of micro-syntax of colour terms, stipulating the order in which the terms
should occur when combined. In this view, the much higher frequency of bluegreen than green-blue in English is not based on some perverse fondness of
designating bluish green areas with slightly more green in them, but is based on
other principles. A large corpus is of course the perfect setting for uncovering such
patterns, and below I present the patterns detectable in the BoE. I have restricted
my investigation to BCTs, mainly because ECTs have a habit of appearing together
with their superordinate BCT, as in, for example, charcoal grey.
When processing the corpus, we have to be aware that not all colour term
combinations can be viewed in terms of a micro-syntax. Very often we find
enumerations of colour terms where more than just two colour terms appear:
blue/green/red/yellow or blue, green, red, yellow. Consequently a search for blue
followed by green in BoE is likely to contain such combinations. The figures below
are based on eliminations of co-occurrences in which the colour terms are separated
by a full stop, a comma or a slash. Only spaces and dashes have been allowed
between the terms. Despite these precautions there are some further factors which
may affect the result. First, we may still end up with an example like (8) below in
which the colour terms follow one another but they do not qualify the same noun,
and cannot possibly be seen as constituting an instance of a micro-syntactic pattern
in the sense sketched above. Without reading every sentence, it is not possible to
eliminate instances like that exemplified in (8) and eventually I decided to let such
sentences remain in the material. A sample test showed that their impact was very
marginal indeed.
(8)

And quotas is a great issue because it separates blacks from white blue-collar
workers. (BoE: npr)

Second, quite often we find combinations in which it is difficult to judge whether

the colour terms are meant to designate a blend or two parallel instances. In cases
like black-white it is obvious that the latter is intended, but in the case of blue-grey
both interpretations are possible occasionally. However, to my mind these two

Colour Terms in the Bank of English

79

possible interpretations do not alter the fact that these collocations instantiate
micro-syntactic patterns the reading does not affect the syntax significantly.
To determine whether there was an existing micro-syntactic pattern,
combinations of the type C1-C2 and C2-C1 were tested for chi-square goodness-offit. The null hypothesis was that the two combinations are equally common in the
English language (i.e. a 1:1 ratio).19 A combination was deemed to suggest a
pattern when 2 > 3.841, p < 0.05, i.e. a strong deviation from the original
hypothesis. In addition to this standard procedure, I have chosen to stipulate a size
filter and exclude all combinations where n < 25 in order to avoid patterns which
may be idiosyncratic and negligible.20 In the end, this filter only excluded four
combinations, all but one fitting the general structure outlined below.
In Table 3:2 below, I have indicated in size and bold figures the combinations
that showed significant dominance, i.e. those which could not be accounted for
within the null hypothesis. There were seventeen combinations that met the criteria
of significance outlined above: blue-black, black-white, grey-black, blue-green,
blue-grey, blue-white, purple-blue, red-brown, grey-brown, yellow-brown, greygreen, yellow-green, red-green, grey-white, orange-red, red-white and yellowwhite. There is a clear pattern where most of these phrases contain colour terms
denoting adjacent colours in the colour domain. This could be interpreted as
showing that this type of construction most often refers to a nuance somewhere
between the two colours mentioned in the phrase. Conspicuous exceptions are
black-white and red-green. In these two cases, it is quite clear that we cannot talk
about modification or a nuance between the colours, since they are opposites (cf.
Wittgenstein 1978 and Westphal 1987). We can preferably view them as attributive
dvanda compounds (cf. Bauer 1983) or as attributive analytic compounds
(Downing 1977: 824).21 An important characteristic of these two phrases is that a
majority of the instances do not refer to the colour domain at all. Red-green refers
frequently to politics as in (9), whereas a majority of the black-white instances refer
to racial issues, as in (10). Although they constitute special cases, their patterns
conform to the general picture.

19

(9)

The low turnout in what was the first direct election of a mayor of Frankfurt
does not negate the belief that the red-green alliance has lost out. (BoE: guard)

(10)

Bill Clintons Texas address was promised after the acquittal of OJ Simpson
exposed the countrys black-white divide. (BoE: today)

Data were analysed by a 2 goodness-of-fit using Yates correction for continuity. Since the null

hypothesis here assumes a 1:1 ratio, the following equation was used: 2 =

( f1 f 2 1)2

(cf. Zar 1996:

n
465); n indicates the total number of tokens of the two combinations, C1-C2 and C2-C1.
20
The choice of 25 tokens is purely arbitrary and may perhaps justifiably be challenged. The recommended
minimum value is 5 per cell, but this would allow single texts with a few tokens to make an impact.
21
As an example Downing (1977: 824) mentions love-hate relationships.

21

79

17
74
0
4

21
5
8

16

32

21

14

17

232

Green

Grey/gray

Orange

Pink

Purple

Red

White

Yellow

Sum: Head

140

Brown

628

87

40
225

193

301

21

39

14

115

Blue

12

14

POSITION

Black

Green

Brown

Blue

Black

MODIFIER

TERMS IN

COLOUR

231

12

16

172

Grey/
Gray

62

26

24

Orange

41

13

12

Pink

COLOUR TERMS IN HEAD POSITION

47

19

10

Purple

157

19

269

26

37

86
4

45

79

38

11

87
58

Yellow

White

11

11

Red

2111

210

53

216

93

31

166

395

87

33

674

153

MOD

SUM:

80
Chapter 3

Table 3:2. Micro-syntactic colour-colour patterns

Colour Terms in the Bank of English

81

A closer inspection of the combinations reveals that a pattern of preferred slots can
be constructed. In Table 3:2, I have indicated this pattern in shades of grey. Some
terms prefer to take the modifier slot, whereas other terms occur much more often
in the head slot. The shadowed areas indicate that blue, grey/gray, red and yellow
are the preferred terms in the modifier slot whereas black, brown, green and white
occur more often as the second element (i.e. the head slot). As we can see below,
this systemisation accounts for almost all significant instances, thirteen out of
seventeen.
Of the four not covered by the system described above, two contain colour
terms which have the same preference, blue-grey (modifier position) and blackwhite (head position). One could, perhaps, expect that, under such circumstances,
we would not find a significant difference from the null hypothesis. However, the
frequent occurrence of these two phrases seems to have carved out a pattern. The
other two patterns which reached significant preference for one order, orange-red
and purple-blue, contain terms outside the observed patterns. The modifier slot in
these phrases is taken by the Secondary BCT, rather than the Primary BCT; it may
make sense that a chronologically more recent and less frequent term would not
take the position associated with the head. However, this is pure speculation; it
should be observed that in the case of orange-yellow and yellow-orange no
significant difference22 could be found although the sum of the two combinations
was fairly high. Furthermore, brown, although a secondary BCT, has the head
position as its preferred slot.
Finally, it should be pointed out that there are three patterns (in the suggested
systemisation in grey tone in Table 3:2) which have not been realised in actual
usage, these are blue-brown, yellow-black and red-black. One plausible reason for
this is that we have difficulty in imagining any nuance between them.23
How should we understand these patterns? From the point of view of colour
semantics, it would seem that there is very little that blue, grey, red and yellow
have in common. The same is true of black, brown, green and white. And maybe
the answer is as simple as that terms which denote adjacent areas do not pick the
same syntactic slot. Consider blue and green for example. Green occurs far more
often in the head position and blue in the modifier position. It is no wonder then
that speakers of English say blue-green (301 instances), rather than green-blue (17
instances). The colour terms in modifying positions very seldom co-occur since
their denotational areas in reality prevent this in the sense of designation of an
intermediate area. If this is so, then we may expect a similar pattern in other
languages. However, if there is no deeper explanation, we could theoretically
assume that the reverse pattern could also occur, meaning that terms which can be
found in the modifier slot in English might equally well prefer the head position.
To give this hypothesis further support I made a pilot study of the equivalent

22

There is a difference, favouring yellow as a head, but since 2 = 1.8906, 0.25 > p > 0.10, the null
hypothesis cannot be rejected.
23
However, technically a shade like maroon could be viewed as BLACK + RED (cf. Kay and McDaniel 1978:
641).

82

Chapter 3

Swedish colour terms.24 The pilot revealed almost exactly the same pattern as that
of English. There were two exceptions: the pattern containing BLUE and GREY
showed no significant preference25, and in the case of BLUE and PURPLE, bl-lila,
blue-purple, was the preferred construction. In the second case, it seems likely
that the disyllabic structure of lila could be of decisive importance, leading to its
favouring the second position. As far as general conclusions are concerned, we may
have to find languages which are further apart geographically, culturally and
genealogically to be able to reveal other patterns.
The only similar study to my knowledge is Conley and Coopers (1981) study
of conjoined ordering of colour terms. However, conjoined phrases of the type
black and blue are both semantically and syntactically different from the
construction considered here. Nevertheless, there are two patterns which are shared
by the studies: the preference for white in the second position and the preference of
red in the first position.
To sum up, in this section we have been able to verify what many researchers
have long suspected to be the case; the existence of a pattern of preferred positions
in colour-colour phrases in English. It was shown that the pattern could be broken
down into a small but clear system of preferences. I have argued that the system
may have been arbitrarily carved out from usage; one conspicuous feature of the
pattern is that the terms which favour the same slots do not denote adjacent colours.
A similarity to the unveiled system can be demonstrated in Swedish, but more
languages will have to be investigated for more far-reaching conclusions to be
drawn.

3.5 Colour terms and bound morphemes

When we look at English colour term morphology, we can observe that there are
actually very few bound morphemes that can be attached to colour terms. There are
a few derivational morphemes and the comparative and superlative forms, no more.
In my treatment of morphemes I have chosen to exclude the zero morpheme the
nominal use of colour terms has already been dealt with and the plural -s.26
3.5.1 Derivational morphology
It was one of Berlin and Kays (1969: 6) subsidiary criteria that a dubious
candidate for classification as a BCT should have the same distributional potential
as the previously established terms. Dixon (1982: 23-24) makes a similar point
while claiming that it is unlikely that a word low in the hierarchy would have
morphological/syntactic possibilities that a higher term lacks. On the basis of
24

My Swedish study only concerned the patterns which showed significant preferences in English. For this
study, I used the corpus Sprkbanken and the following subcorpora: p65, p76, dn 87, p95, p96, p97, p98,
rom i, rom ii.
25
As argued above, this would seem to be an expected outcome in view of the fact that both terms prefer the
modifier slot.
26
The plural morpheme, s, was excluded mainly for practical reasons. It occurs most often in figurative
constructions (e.g. whites in the sense of white people), and, furthermore, in the case of ECTs, it usually
denotes a plural form of the object from which the colour term was derived (e.g. peaches in the sense of
fruits).

Colour Terms in the Bank of English

83

these two statements, Corbett and Davies (1995) investigate the derivational
possibilities in French and Russian. They report that both Russian and French
primary BCTs have significantly more derivatives, and that there is a correlation
between B&Ks suggested hierarchy and the number of derivational possibilities.
They ignore English, probably because of the relatively few possibilities there are
in this language. Corbett and Davies study differs from the present one in that they
are concerned with the number of derivational possibilities open to each term and
in that they treat each possible morpheme as an either-or case on the grounds of the
appearance of a form in derivational dictionaries. Although it served a purpose in
Corbett and Davies (1995) study, I find this way of treating morphological
possibilities most doubtful. Certainly, this category, like so many other phenomena
in linguistics, is likely to have fuzzy boundaries; the failure to list a morphological
form in a dictionary entry does not necessarily mean such a form is not possible.
This problem is discussed further below.
In this section, I present the frequency of derivational forms in the BoE. By
doing so, I show that morphological possibilities and productivity in this respect are
characterised by the fuzziness that we can find in so many aspects of language
usage. The derivational possibilities that are open to English colour words are: ish,
y, ly, ness, en, ed and ing. There are also such possibilities as ishly,
ishness, ished, ily, iness, ened and ening, but these are secondary derivations,
presupposing the existence of the previously mentioned suffixes ish, y and en.
Thus, in English, we have a small number of derivational possibilities, most of
which can still be considered to be productive. This means that, in theory, if need
be they can be used with any colour term. In view of this fact, the frequency of the
derivations is more telling than their occurrence in dictionaries.
In my discussion below I focus my attention on the first two derivative forms
ish and y. My reason for doing so is that these appear to be competing forms with
more or less the same meaning. Thus, a corpus study may reveal interesting
patterns. The rareness of the other morphemes, and the fact that en is restricted to
black, white and red, have led me to exclude them from further analysis. An
additional reason for concentrating on ish and y is that these morphemes have
been discussed recently in the context of colour terms by Wyler (1992) and
Frankowska (1995).
Let us first establish that the meaning of these two suffixes is usually regarded
as virtually the same. The OED states that ish has the following history in the
domain of colour:
Added to adjs. with the sense Of the nature of, approaching the quality of,
somewhat, apparently first with words of colour (which may have been treated as
ns., and so have originally come under 2): e.g. bluish (a 1400), blackish (a 1500),
brownish, reddish, whitish, yellowish, etc.

For y, the OED states that:

84

Chapter 3

In the 15th cent., if not earlier, certain monosyllabic adjs. were extended by means of
this suffix, app. with the design of giving them a more adjectival appearance []. In
this application the suffix has not infrequently come to express much the same notion
as -ish; this is particularly so with colour-epithets, as blacky, yellowy, and esp. when
these are used quasi-advb., as greeny-blue, bluey-green, reddy-brown.

In a similar fashion, the CCELD (Collins COBUILD) defines the meaning and
function of ish as added to [] describe something as having a particular []
colour, but only to a limited extent [f]or example reddish means slightly red,
and y as added to colour terms in order to form adjectives that describe
something as being roughly that colour or having some of that colour in it.
Marchand (1969) makes the same observation, and, in addition, he (p 353) claims
in reference to y that [t]he type is weak, ish being the stronger rival of the
suffix. Berlin and Kay (1969) and Dixon (1982) say that these derived forms
occur more frequently and freely with the BCTs than with the non-basic terms.
However, these claims appear to be based entirely on impressionistic data as no
evidence is presented.
Wyler (1992) illustrates another approach. He (1992: 134-135) provides a list
of derivations based on the entries in the Shorter Oxford English Dictionary
(edition 1980) (henceforth, SOED). His list is restricted to the following forms:
blackish, reddish, yellowish, greenish, bluish, brownish, greyish, pinkish, purplish
and yellowy. Since he star-marks forms like whitish and greeny, he would appear to
suggest that these forms are ungrammatical, something which may come as a
surprise to some speakers of English. Wyler gives no motivation for his choice of
dictionary, but the flaw in his method can be illustrated by consulting of the OED,
the very dictionary on which the SOED is founded. Almost all of his star-marked
forms appear in the OED as attested. There are other aspects of the study which
raise questions. For example, he (1992: 135-136) claims that [w]ith the exception
of rose [] none of these colour terms [i.e. what I here call ECTs] occur other
than as in the representation given above [i.e. the root form].27 This statement is
much stronger than those of Berlin and Kay (1969) and Dixon (1982), and to my
mind this seems to be far too rigid a statement to make given Wylers weak
empirical foundation. In fact, it could be claimed that Wylers reliance on the forms
listed in one dictionary leads him to stray way off the mark in his discussions on
morphology and colour terminology, and he is demonstrably wrong on many
occasions.
Frankowska (1995) devotes one section of her essay on the suffix ish to the
domain of colour terms. Her study exemplifies another type of approach in that she
bases it on examples taken from newspapers and books (in other words, a type of
corpus material) and the intuition of a few native speakers. Furthermore, her
wording is much more careful than Wylers as she points out that the nature of her
material may suggest that the acceptability of some forms can be questioned.
Indeed, she found that, occasionally, there was no agreement among the native

27

Somewhat confusingly, Wyler does not use the traditional way of indicating word forms, italics, e.g. rose.
Instead he uses inverted commas which would normally indicate meaning. I have retained his notation in the
quotation.

Colour Terms in the Bank of English

85

speakers as to the acceptability of certain forms. The general pattern that

Frankowska (1995: 41-42) was able to uncover was that
the application of the suffix [] is completely unlimited when the base is a focal
term. Consequently, we are free to derive the following complex forms: whitish,
blackish, yellowish, brownish, pinkish, reddish, greenish, bluish, violetish, greyish.
However, when the base is a shade, a secondary colour or a very specific colour, the
derivation appears to be blocked quite frequently.

Among the acceptable forms, apart from those mentioned above, she lists khakish,
amberish, goldish, silverish and marblish and among doubtful forms (i.e. cases
where the native speakers were not in agreement) ?fawnish, ?peachish,
?crimsonish, ?mauvish and ?maroonish. The unacceptable forms listed include
*orangish, *emeraldish, *olivish, *lilacish, *scarletish, *beigeish and *fuchsiaish.
As an overall conclusion Frankowska suggests that the more specific a colour
term is the less acceptable is a derivation of that term containing the suffix ish.
This conclusion is in broad agreement with Dixons proposals. However, although
this is certainly true as a general characterisation, I think the examples given above
also demonstrate that the picture is quite complex. The above conclusion cannot
explain why amberish would be acceptable but not olivish. In fact, it would appear
to be quite difficult on many occasions to establish to what degree one colour term
is more specific than another. It may be the case that a term can be said to be more
specific in the sense that it is only used in certain types of discourse, say, that of
fashion. However, there seems to be no reason why it could not still take a
derivational suffix given the right context.
One important source for the lack of certain forms, overlooked by the two
above-mentioned researchers, could be phonological constraints. Such constraints
could plausibly explain why informants are less happy with peachish, orangish,
beigeish and fuchsiaish. Should there be a semantic constraint we would also
expect forms such as peachy, orangey, beigey and fuchsiay to be unacceptable and
not to occur in texts.

86

Chapter 3

600

568

500
287

242

200

198

191

149

124
57

100

1
tannish

305

300

maroonish

354

400

limeish

creamish

orangish

fawnish

goldish

mauvish

blackish

whitish

purplish

brownish

pinkish

greyish/grayish

bluish/blueish

yellowish

greenish

reddish

Figure 3:7. Derivations of ish, number of occurrences.

In Figures 3:7 above and Figure 3:8 below we get an overall picture of the number
of tokens of the derivations of y and ish that occur in the BoE.As Figure 3:7
clearly illustrates, it seems that the derivation of ish is largely confined to BCTs.
This fits well with Frankowskas observation above, and with Dixons prediction.
The rank of the colour terms does not correlate with Berlin and Kays evolutionary
hierarchy (rs = 0.247, p > 0.05), nor with the frequency rank of the BCTs in the
corpus (rs = 0.227, p > 0.05). However, this appears to be strongly linked to the low
frequency of blackish and whitish. A post hoc correlation test without blackish and
whitish showed a significant correlation with the Berlin and Kay hierarchy (rs =
0.866, p < 0.05). The position of blackish and whitish is further discussed below.
1200

1104

1000
800
591

600

504

400

lilacy/lilacky

limy

tanny

blacky

turquoisey

olivey

maroony

mauvey

goldy

94 79 69 55 44 44 41
28 28 17 10 6

beigey/beigy

113

200

mustardy

lemony

plummy

reddy

purply

bluey

whitey

peachy

browny

orangey

yellowy

rusty

greeny

pinky

creamy

rosy

silvery

Figure 3:8. Derivations of y, number of occurrences.

Colour Terms in the Bank of English

87

The picture we get when we consider the y suffix is very different from that of the
ish suffix. The most conspicuous feature of the y derivations is perhaps the
predominance of the ECT derivations rosy, silvery and creamy. A rank correlation
test of the BCTs showed no significance correlation to the Berlin and Kay
hierarchy (rs = 0.206, p > 0.05), nor to the rank frequency of the BoE (rs = -0.427, p
> 0.05). It seems that y occurs much more freely with non-basic colour terms than
does the ish suffix. This last statement may not be statistically significant due to
the low frequency of each type, but it is worth noting that, apart from fawnish and
greyish, all forms containing ish also occur with the y form, whereas no less than
thirteen forms containing y do not occur in the ish form.28 It can also be observed
that Marchands (1969) claim of the weaker character of y, mentioned above,
seems to be confirmed here. In those cases where a term occurs with both ish and
y, the ish form is much more frequent.
As in the case of ish, the y suffix is only rarely attached to black and white.
The low frequency of blacky, blackish, whitey and whitish can reasonably be
explained by the fact that there is little need for these terms. The sense of of the
nature of, approaching the quality of, somewhat black/white is lexicalised in the
words dark and light when reference is made to lightness. If reference is made to
colour, these terms seem to be preferred together with some colour word as a head,
as in, say, dark grey rather than blackish. The meanings of these expressions are
not of course the same, although they may refer to the same type of nuance.
Blackish and whitish take the focus of BLACK and WHITE as their reference points,
whereas modifications of grey start from GREY. The low frequency of these terms
suggests that this is quite an unusual perspective, possibly due to the absoluteness
of black and white.
How should we understand the different pictures that the two diagrams
present? A closer inspection of Figures 3:7 and 3:8 suggests that, although similar
in meaning, the two suffixes use different reference points. Given the character of
Figure 3:8, it might be proposed that terms such as rosy, silvery, creamy, rusty and
peachy do not derive their sense from the colour terms rose, cream, silver, rust and
peach, respectively. Instead they should be considered derivations from the entity
that the nominal form designates. Apart from the BCTs, most of the terms that are
attached to y could be either objects or colours. Significantly, it is pink (also a
flower) which occurs most often with y of the BCTs.
The fact that rosy (1 104 tokens) is much more frequent than the term rose
(632 tokens, cf. Figure 3:2) in reference to colour suggests that their semantic
relation is not the same as that between, say, brown and browny. Indeed, if we
assume that there is a link between frequency and conventionalisation, rosy is the
more conventionalised colour term in the speech community. Furthermore, from a
diachronic point of view, we can see that rosy is the older of the two in the sense of
colour; the OED lists 1374 as the first quote for rosy, and 1530 for rose.29 In
addition to this, it should also be mentioned that a quarter of the instances of rose
28

Some preferences may, as suggested above, be explained by phonological or, indeed, graphical
constraints. Thus the occurrence of peachy and beigey but not beigish and peachish may have phonological
grounds, whereas the lack of greyy or greiy may be due to graphical preferences.
29
In the case of creamy, peachy and rusty, these forms are older than the zero-forms in reference to colour.

88

Chapter 3

are linked to the phrase rose-tinted/coloured glasses/spectacles. Rosy, on the other

hand, seems to be used much more freely, although there is a preference for phrases
such as rosy cheeks and rosy future as in (11).30
(11)

After poorly attended, but decisive local government elections, some predict a
rosy future for Nigeria, under a civilian government, benefitting from booming
oil revenues. (BoE: bbc)

On the basis of the above facts, it is tempting to suggest that rosy and rose (and
similarly cream creamy, silver silvery, rust rusty etc.) might be viewed as two
alternative derivations, y and zero, available in English for denoting a colour
nuance similar to that typical of the object from which they are derived. In view of
the high frequency of rosy, it is somewhat surprising that rose rather than rosy
appeared in the elicitation studies on which I base my choice of terms. One
plausible answer may be that derived terms are not regarded as proper colour
terms.
The ish suffix appears to be derived from a colour term which refers to a
salient focal point in the colour domain. This could explain the closed character of
the ish category. However, a few ECTs occur with ish and Frankowskas (1995)
informants were willing to accept forms like amberish and khakish. It seems
however that the majority of these derivations are used to modify another term as in
(12) below.
(12)

Concrete used to cover any large surface will very quickly craze unless you
allow for expansion joints. Many precast coloured concrete slabs come in rather
sickly greens and mauvish pinks, colours which will swear with any planting.
(BoE: brbooks)

In fact, most colour terms with the suffix ish and y are used in this way.
However, as far as the y suffix is concerned, this appears to be its main function
when attached to a BCT; this usage is proportionally lower in the case of BCTish.
One factor which no doubt affects the different patterns of these morphemes is their
age as productive morphemes in the colour domain. Predictably, ish is the older
suffix according to the OED; derivations from colour terms are found as early as
1379 (yellowish and whitish31), whereas the first colour term derivations of y
found in the OED comes two hundred years later, in 1594 (blacky).
To sum up, we can say that both ish and y, being productive morphemes,
are free to form derivative forms with most, if not all colour terms. However, the
functional need for an approximate form of a very precise term is strictly limited so
they do not occur very often in texts. Moreover, it appears that ish is the more
frequent derivative form when it comes to BCTs, suggesting that it refers to a
30

Whether or not a phrase like rosy future is closely related to the colour sense is debatable. However, I
have chosen to treat it as such. Both rose and rosy have been taken into account when they have figurative
senses of good. For a more detailed discussion of this particular sense, see Section 7.7.
31
In fact, the OED lists the first quote of whitish as dating from 1398, but in the quotation for yellowish,
whitish also occurs: 1379 Glouc. Cath. MS. 19 No. I. i. iv. lf. 11 It ys evirmare whityssh or olowyssh.
[Emphasis added].

Colour Terms in the Bank of English

89

nuance which deviates from a salient reference point in the colour domain. In the
case of ECTs, y seems to be preferred and occurs most often with transparent
terms i.e. terms which also have an object sense. This implies that it is the object
that serves as the reference point and not a nuance in the colour domain.
3.5.2 Inflectional morphology
The only inflections I consider here are the comparative and superlative forms. In
theory there should be no constraints apart from phonology on this formation.
However, as Figure 3:9 illustrates, it is quite unusual to inflect colour adjectives.
And, as is so often the case, we find that it is the terms commonly referred to as
primary BCTs that are inflected most often, with the exception of yellow. Another
conspicuous feature is that the comparative form occurs much more often.
However, this seems to be the case for most, if not all adjectives,32 and makes sense
semantically. Whereas the comparative form merely expresses that something
exceeds a reference point, the superlative form identifies a particular instance as the
most extreme instance within a defined group of reference. In other words, we can
say that comparatives are comparisons in which an individual token serves as
ground, whereas superlatives are comparisons in which a group of individuals
(arbitrarily delimited for the purpose of comparison) serves as ground.33
Dixon (1982: 19) observes that the semantic structure of the colour field
affects the meaning of the comparative form the domain precludes converse
relations between the forms. He even suggests that the term comparative, although
morphologically apposite, is perhaps semantically inappropriate. However, I see
no reason why converse relations should be part of the notion of comparativeness.
As Dixon correctly points out, a converse relation is the result of the semantic
structure of a field (it allows antonymy), and consequently it is not directly related
to the semantic notion of comparativeness. To say that X is redder than Y is an act
of comparison. The fact that we cannot construe this to mean that Y is
bluer/yellower/ browner than X does not appear to me to be really relevant.

32

A random check of the following typical adjectives in the BoE gave the frequency pattern below: old
(comp.: 23674, sup.: 5782), young (comp.: 16718, sup.: 4323) tall (comp.: 1415, sup.: 594), short (comp.:
4188, sup.: 895) angry (comp.: 200, sup.: 50)
33
The description of comparative forms of adjectives is a much greater issue, of course, cf. Rusiecki (1985).

90

Chapter 3

450

419

400

Comparative form

350

Superlative forms

300

250

250

180
113
25

53

20

Grey/Gray

Blue

Red

White

Black

Green

25
1

18 7

6 2

1 4

0 1

Orange

42

50

100

Puce

75

Purple

115

Yellow

85

100

Brown

150

Pink

200

Figure 3:9. Inflections with colour terms, number of occurrences.

When we look at Figure 3:9, we see that, unpredictably, the superlative form is
most frequent in the case of black. However, a closer examination of the actual text
data reveals that the use of blackest is primarily used in a figurative sense, as in
(13) and (14) below.
(13)

Monday was one of Sarajevos blackest days since the Bosnian war erupted in
April 1992, and it was Bosnian Serb forces who carried out this atrocious attack
on civilians shopping at the citys main market. (BoE: indy)

(14)

His blackest humour is devoted to the complexity of rules he has to cope with.
(BoE: econ)

Indeed, the most frequent collocations are of the first type, blackest day(s) and
blackest moment, where the meaning of black can be paraphrased as sad or
worst. This particular meaning of black does not seem to occur in the comparative
form; intriguingly, it is also the case that saddest is a more frequent form than
sadder in BoE, so this pattern may be due to the structure of this particular domain.
One idea is that this type of meaning can frequently be exposed to hyperbole.
On the whole, an analysis of the individual examples reveals that the use of
comparative forms is either confined to certain domains, such as HAIR, FACE, and
SKY, or is purely figurative. It is typical of the preferred domains that the colour
term does not represent the best example of the colour (as viewed in the domain of
colour), but a generalised use (extension from the focal colour, i.e. metonymy); for
example, red hair can more correctly be described as orange or rust-coloured on
most occasions, and someone who is described as red in the face very seldom has
the colour of a traditional fire engine; it is more like a pinkish red hue. It is also
interesting to note that, in the case of facial colour, the majority of comparisons are
made with the same object. Thus we find examples of what might be called

Colour Terms in the Bank of English

91

temporal or type-token comparisons rather than comparisons between two present

objects, as in (15) below.
(15)

His big white face was whiter than usual. (BoE: npr)

Here, the white face of a person is compared with the same persons normal facial
colour. This type of comparison can also be construed as taking place between two
mental spaces (cf. Section 2.3 and Fauconnier 1997). We compare the facial colour
that this person normally has (this knowledge resides in a generic space, G) with
that expressed in the current discourse space (CDS).
a

CDS
[face] whiter than usual
a

G space
[normal] white face

Figure 3:10. Comparison between mental spaces.

It is also significant that the only ECT that appears in the comparative, puce, is
most frequent in the domain of facial colour (see Chapter 5) and can be regarded as
polysemous, with the figurative meaning of anger. Both interpretations are
possible in (16).
(16)

And people were furious? Well, William Rees-Mogg went puce and Paul
Johnson went pucer: he asserted that Hitchens had slurred the saintly one as a
lesbian; (BoE:guard)

In conclusion we can say that the use of inflectional forms of colour words is very
infrequent when comparing two objects present in the same mental space. One
possible explanation for this may be that, in cases where we are comparing and
focussing on different nuances of, say, red, we employ more definite colour terms.
This usage will be given closer attention in Chapter 5.

92

Chapter 3

To sum up, it can be noted that the overall pattern of morphology bears a
close resemblance to the evolution of colour terms, as suggested by Berlin and Kay.
Previous studies in the area have shown that these are the forms that informants
deem to be acceptable, and that they are the forms one tends to find in dictionaries.
I have here demonstrated that the frequency of these terms also correlates with the
sequence.

3.6 Colour terms in different subcorpora

The overall frequency of the colour words I have chosen to investigate (cf. Figures
3:1 and 3:2) is approximately 979 tokens per million words.34 This figure is very
similar to that obtained by Corbett and Davies (1995) for the LOB corpus: 974.
However, there are a few differences that should be indicated. First, as mentioned
earlier, Corbett and Davies consider colour terms only in their adjectival function
and, as demonstrated in Section 3.3, nominal usage can play a significant part.
Second, apart from the BCTs, Corbett and Davies include no more than three
colour terms: silver, bronze and scarlet. On the other hand, Johansson and Hofland
(1989), on whom Corbett and Davies base their frequencies, do not specify how
they treat polysemous words, so Corbett and Davies inclusion of orange, silver
and bronze could be called into question.
My own analysis of LOB shows that the score for orange is correct (the fruit
sense is invariably tagged as nominal) whereas that of silver includes three
instances of silver medal(list) which I find very doubtful as references to colour;
such phrases have not been included in my material. Accordingly, in my view,
bronze should not have been included at all since all tokens in LOB occur in the
context of medal/medallist.35 If we adjust for these facts, we get a figure of 961
colour words per million words in the LOB corpus. Given the other differences,
this figure seems to be comparable to my own.
There are additional differences which make comparison difficult. One such
difference is the composition of the corpora. Whereas LOB is composed of written
British English texts, BoE includes both written and spoken texts, from Britain, the
USA and Australia. Corbett and Davies (1995) observe that colour terms are much
more frequent in the Russian corpus they use for their analysis, 2261 tokens per
million words, and they suggest that this may be connected with the composition of
the corpora.36 In order to make a more thorough investigation of the frequency of
colour terms in different text types, an analysis was made of the frequency of terms
in some of the subcorpora of the BoE. Five different British subcorpora were
considered, each representing a medial genre of some sort: brspok, brmags, bbc,
guard and brbooks. I specifically chose British texts to eliminate the problem of
dialectal preferences. However, having said that, I should also acknowledge that
34

If we take the confidence interval into account we get a figure between 982 and 975 per million words.
To do Corbett and Davies (1995: 329) justice, it should be pointed out that they are aware of the problem
of polysemy. They state that: [t]he frequencies of silver and bronze are probably over-estimates of their use
as colour adjectives, however.
36
The characteristics of this corpus are described in greater detail in Corbett and Morgan (1988). It is a
frequency dictionary edited by Zasorina (1977) and it is based on a corpus containing 1 056 382 words. It
contains journalism, drama, non-fiction and literature.
35

Colour Terms in the Bank of English

93

Occurences per million word

there is considerable variation within these subcorpora and that they contain in fact
several different text types themselves. Nevertheless, the clear picture that Figure
3:11 presents can give us some general ideas of colour frequency in different types
of media. The only similar study I know of is that carried out by Corbett and
Morgan (1988) for Russian. However, their study does not include spoken material,
nor broadcasting media, and the text types are not readily comparable with those of
the BoE.
Before we look at the figures I should also point out that the frequencies of
the calculated terms gold, golden and silver are not included in these figures. This,
of course, means that the figures could be slightly higher. However, it is the
relation between the genres which I find interesting, and I feel confident that the
exclusion of these terms has not affected the overall picture in any vital way.
2000
1800
1600
1400
1200
1000
800
600
400
200
0

1795.65
1363.29
846.29

405.23

brspok

384.45

brmags

bbc

guard

brbooks

Figure 3:11. Relative frequency of words in five subcorpora of the BoE.

Figure 3:11 demonstrates clearly the differences between the various genres. Most
conspicuous is the markedly lower frequency in the spoken corpora brspok and
bbc. There is a formal difference between these two corpora, in that brspok
contains spontaneous spoken material, whereas the bbc corpus contains broadcast
material, much of which may have been written down before being spoken.
Nevertheless, it appears that the transient character of speech makes people less
keen to use colour terms. Another important characteristic of the radio medium,
which may explain the low frequency of colour terms in the bbc corpus, is the fact
that the radio listener, i.e. the addressee (directly or indirectly) of what is said,
cannot see what the speaker is talking about, and under such circumstances
mentioning the colour of a particular object may seem unnecessary. This is of
course a feature which radio shares with the written medium (unless pictures are
provided). It would appear, however, that the permanent character of the written
material allows it to be more detailed with respect to descriptions.

94

Chapter 3

Another striking feature is the very high frequency of occurrence in

magazines. It seems reasonable to relate this high figure to the typical content of
magazines, both in terms of topics and layout. Many magazines pay great attention
to subjects which are intimately linked to colours and thus colour terms: e.g.
fashion, cosmetics and horticulture. Advertisements containing colour references
may also be more frequent in magazines than in a newspaper like The Guardian
(guard corpus). Finally a very important factor is the presence of colour pictures in
many magazines and books. It would seem that an interesting and possibly fruitful
approach could be to study the correlation between use of colour terms and pictures
in various media.
If we consider the proportion of BCTs and ECTs, we find once more great
differences between the spoken and the written texts. Table 3:3 gives the precise
figures. When we read these figures we should, again, bear in mind that these do
not include the terms gold, golden and silver, so comparisons should only be made
with the other subcorpora.
Table 3:3.

Elaborate colour terms in five subcorpora

BRSPOK

BRMAGS

BBC

GUARD

BRBOOKS

BCTs tokens

8013

50531

7091

19792

54343

ECTs tokens

165

3587

30

740

3090

2.02 %

6.63 %

0.42 %

3.60 %

5.38 %

10

36

32

34

Per cent ECTs

ECT types ( n 5 )

There is a very noticeable difference between spoken and written corpora. This
difference is manifested more in terms of types than proportion, although the
proportion in the bbc corpus is also markedly low. The colour terms that were
mentioned more than five times in the bbc corpus were maroon and scarlet, and the
ten mentioned in the spoken corpus were amber, beige, cream, emerald, maroon,
navy, olive, rose, scarlet and turquoise. All of these colour terms are among the
most frequent ECTs (cf. Figures 3:1 and 3:2), as might have been expected. If we
look at the terms mentioned in the spoken corpus we can also see that no particular
area seems to be favoured. Most frequent are ECTs in the brmags corpus and, in
fact, all colour terms under investigation occur more than five times in this
subcorpus. It seems likely that these patterns can be explained along the lines
outlined above. Fashion, for instance, is a subject which has a need for specificity.
Furthermore, magazines and books frequently contain colour pictures in relation to
fashion articles. In the context of a picture, it makes sense to be more precise in
colour designation. The relation between specificity and different domains of the
qualified noun will be further explored in Chapter 5.37
37

Stoeva-Holm (1996) gives a thorough description of colour term usage in German fashion magazines.
Interestingly, she gives a diachronic perspective, demonstrating that the type/token ratio has hardly changed
at all over a hundred years.

Colour Terms in the Bank of English

95

3.7 Summary
The aim of this chapter was to give a general description of colour term usage in
the BoE. Some of the frequency patterns may provide starting points for a closer
semantic analysis of colour terms. Through statistical calculations based on the
corpus material mainly frequencies it was possible to identify a number of
patterns in relation to age, nominal usage, morphology and colour-colour
combinations. The picture formed by these patterns does not unequivocally support
a theory of BCTs in its most rigid expression. On the whole, clear category
boundaries have not appeared. Instead the notion of fuzziness dominates. However,
if we consider the internal structure of the BCT category, we can note that
frequency of occurrence, age and the use of ish correlated fairly well with the
hierarchy suggested by B&K. This may reflect a conceptually privileged position
of the early terms, something which will be readdressed in Chapter 4.
There are also signs which imply that the domain of the qualified noun may
be of some importance when it comes to the meaning and behaviour of typical
ECTs. Chapters 5 and 6 investigate these patterns more closely from a semantic
point of view.

Chapter 4

Different Planes:
Colour Terms as Adjectival Type Modifiers

4.1 Introduction
Consider the sentences below, focusing on the word red.
(1)
(2)
(3)
(4)
(5)

Red beets make a good garnish.

The boy with red hair stood out in the crowd.
The red paint spilled onto the floor.
Her red dress made her the center of attention.
He saw red when his secretary came in an hour late.

Would you say that red means the same thing in these sentences?
In a classic study, Halff et al. (1976) asked 19 subjects to judge whether the
redness of red was the same in these sentences (and 15 more of this type), or
whether some reds were redder than others. The result supported their hypothesis
that (p 382) context places bounds on the internal representations of a concept.
Furthermore, they claim to have shown that there is little agreement even in our
own culture concerning the meaning of any particular color word. Although their
first conclusion seems to be valid, the second one appears to be much more
controversial. It would seem to be a gross overstatement to claim that there is little
agreement as to the meaning of a term like red. Indeed, it was demonstrated by
Berlin and Kay (1969) that even the same speaker may classify borderline cases
differently from time to time, but they showed convincingly that there is general
agreement as far as the focal colour is concerned (cf. Chapter 1). However, the
methodological approach of Halff et al. is quite different from that of B&K and
they certainly did discover something; the question is, what?
On closer inspection we can see that red in these sentences has what we might
call different functions. Red in red beet and red hair is not just a description of
some beets and some hair; its function is to point to some particular type of beets
and a certain type of hair. In sentences (3) and (4), on the other hand, the function
of red is precisely descriptive. Finally, in (5), we are concerned with a non-literal
use of red. The main function in this case could perhaps be described as the transfer
of meaning to some other domain (here ANGER). It is tempting to suggest that the
result Halff et al. were able to reveal is closely linked to these different functions. It
seems that a detailed study of the functions will give us additional knowledge
regarding the meaning of colour terms. It may also be the case that some of the
issues that have characterised debates on colour semantics are due to the failure to
acknowledge the existence of these different functions.
This chapter considers in some detail the functions that a colour term may
have when it appears in the position of an adjectival attribute modifier in a noun
phrase. The focus is on what can be called type modification or classifying function

97

98

Chapter 4

(i.e. that represented by examples (1) and (2) above). This function is of special
interest for three reasons:
it seems to be restricted to only a few terms, whereas any colour term can
be used for descriptive purposes.
it would appear that this function allows the colour term to be used outside
the domain of its normal designation; consider, for example, the phrase
red hair in (2). On many occasions red hair is as close to orange or rust as
to red and would possibly be more correctly described as such.
finally, it would seem that type modification is intimately linked with
some aspects of figurative usage.
The aim of the chapter is primarily to investigate both descriptively and
analytically the first two phenomena described above. An empirical study of two
corpora (the OED and the BoE) will test the truth of the first claim, whereas the
nature of the second phenomenon will be analysed on the basis of some individual
examples. The third observation above will be treated in Chapter 7.
The present chapter is organised in the following way. First, there is a general
description of some suggested ways of classifying adjectival functions, followed by
a discussion of some adjectival functions of special interest for the present study
and their placing within the framework of cognitive grammar. In the latter part of
the chapter, an empirical study is presented together with a theoretical analysis of
the phenomenon. The chapter ends with a discussion of the possible implications of
the results of the present study for the interpretation of previous works.

4.2 Classifying adjectives

Adjectives can be classified in a number of different ways depending on the
perspective adopted. We can, for example, distinguish between attributes and
predicate complements on the basis of their syntactical position. From a semantic
point of view, other distinctions can be made: one such common distinction
between adjectives is that between absolute and relative (or object-related)
adjectives. Absolute adjectives are characterised by the fact that their meaning is
not dependent on the character of the noun, compare lively (absolute) vs. large
(relative) in the sentences below.
(6)
(7)

The lively Chihuahua was unusually large.

The lively Irish Wolfhound was unusually large.

Whereas the meaning of lively remains more or less the same, that of large is
changed considerably due to the character of the nouns. More technically, it could
be argued that it is not the actual meaning of the word that changes so much as the
reading (interpretation) or referential scope. Colour terms have usually been
considered absolute adjectives our understanding of red is not affected whether it

Colour Terms as Adjectical Type Modifiers

99

qualifies car or flower. However, as we could see above, this is not entirely true
there is a certain amount of vagueness even among absolute adjectives. It would
seem that many absolute adjectives do not have antonyms; instead the semantic
opposition is one of complementarity (cf. Dixon 1982). Relative adjectives, on the
other hand, are typically characterised by having an antonym; e.g. dimension
adjectives (big, small, thin), age adjectives (old, young), and a few more.1 And, as
was demonstrated in (6) and (7), the interpretation of relative adjectives is largely
dependent on the meaning of the head noun.
Other classifications are possible. As early as the 1920s and 1930s, Jespersen
suggested a classification of adjuncts, i.e. modifiers in attributive positions, on
syntactic grounds. He (1924: 108-144) distinguished between restrictive and nonrestrictive adjuncts on the basis of their syntactic relation to the noun. The two
types correspond to the two types of relative clauses with the same name. Thus a
restrictive adjective can usually be rephrased in terms of a restrictive relative
clause, as demonstrated in (8) and (9) below.
(8)
(9)

Can you give me the red jersey?

Can you give me the jersey that is red?

Similarly, a non-restrictive adjunct corresponds to a non-restrictive clause.

Jespersen makes some notes about the function, arguing that that of restrictive
adjectives is to help the speaker identify the object, by specifying it. The function
of non-restrictive adjectives, on the other hand, is merely ornamental (p 112) in
his view.
A different type of distinction, and one that I will try to explore in the context
of colour terms, has been suggested by Teyssier (1968), Warren (1984, 1988) and
Magnusson and Persson (1986). They advocate a tripartite functional distinction
between identifying, classifying and characterising functions.2 The last two
roughly correspond to what other authors have discussed in terms of referencemodifying and referent-modifying functions (Bolinger 1967), intensional and
extensional (Siegel 1980), non-predicating and predicating (Levi 1976,
1978).3
4.2.1 Three classes of attribute modifiers
The basic function of adjectives, most researchers would probably agree, is to
ascribe, connect, a property to a thing. However, the purpose of this ascription can
be analysed in terms of the three classes suggested above. To my knowledge, the
first person to make a distinction between identifying, classifying and descriptive
adjectives was Teyssier (1968). As far as identification is concerned, he (1968:
227) observes that

See Dixon (1982: 15 ff) for a thorough semantically based account of different types of adjectives.
A functional distinction of adjectives is also made by Halliday (for instance, 1994: 184-186). He
distinguishes between epithets and classifiers, the latter identical with the classifying function outlined here.
3
Additional types of classification are intersective vs. non-intersective (Siegel 1980), synthetic vs. absolute
(Taylor 1992).
2

100

Chapter 4

the adjective is assimilated to a defining function of a syntactic order. It then points to

ONE particular instance of the object thus determined, its main function being [] to
substantiate (or expand) the definiteness implicitly contained in the determinative.
[emphasis original]

In the terminology of Jespersen, this means that the identifying function produces a
restrictive meaning. Teyssier mainly discusses adjectives which have only this
specific function; words such as right (not left), first, same and only. However, as
pointed out by Warren (1984, 1988) and Magnusson and Persson (1986),
practically any adjective can be used to this end, provided it creates a contrast
between the intended referents and other possible referents within the same frame
of discourse, or mental space (cf. Chapter 2). This means that permanent as well as
temporary properties may be highlighted for this purpose. Since the intention is to
identify a referent, in whatever mental space, obviously the noun phrase has to be
specific. An important point is that the converse is not necessarily true; it does not
follow that specific reference entails identifying function. Consider (10) below.
(10)

Soon after the maroon-clad waitress took one of our orders, food started to
arrive, and it never seemed to stop. (BoE: usnews)

Whether we should regard maroon-clad as identifying or descriptive cannot be

determined on the basis of this sentence alone, but has to be left to wider context. A
restrictive reading with identifying function is possible in the event of there being
another waitress with slightly different clothing. However, a descriptive
interpretation would appear to be more likely. Our knowledge of restaurants tells us
that waitresses usually wear some kind of uniform and that their job is to pick up
orders, so we have no problem understanding the definite article here. In fact, the
wider context (which is not supplied) confirms the speculation that this is the first
mentioning of a waitress: no other waitress is ever mentioned and thus context will
direct the reading of the attribute into having a descriptive function i.e. a nonrestrictive reading. What we have learnt is that waitresses at this particular
restaurant usually wear maroon dresses. The fact that quite a substantial amount of
context has to be considered to be able to distinguish between an identifying and a
descriptive reading makes it difficult to study these functions carefully in a large
language corpus. I have therefore decided against investigating these functions
thoroughly. However, I will return to them from time to time.
When it comes to the classifying function, Teyssier (1968: 228) points out
that
in the case of CLASSIFICATION the adjective is assimilated to a categorizing function
of a semantic order. It then points to A SPECIMEN of a class of objects, its main
function being to substantiate (or expand) the inherent indefiniteness (i.e. a degree
of extension) of the noun it applies to. [emphasis original]

Furthermore, he observes that most adjectives can be used this way, but claims (p
228) that some types of adjectives are intrinsically more classifying than others,
namely those denoting nationality, age, size, and colour. Intriguingly, these belong
to the core adjective classes, as defined by Dixon (1982: 54-55). Prima facie,

Colour Terms as Adjectical Type Modifiers

101

Teyssiers statement appears to be contradicted by Warrens (1988: 127) assertion

that [s]ome premodifiers are basically descriptors, viz. unaffixed bona fide
adjectives (old, wise, sad, red, tall) However, on closer examination, there is no
contradiction here. Classification is concerned with more or less permanent (or
perceived as such) properties4 through which several tokens can be typified.5
Consequently, these properties can make us view these tokens as forming a subtype of their own. Magnusson and Persson (1986: 207) take this one step further
when they argue that description precedes classification from a conceptual point of
view. For the cognitively oriented linguist, this makes perfect sense; only from our
situational, token-related, experiences can we classify objects, cf. Figure 2:2, which
illustrates schematisation from the point of view of cognitive grammar.
Teyssier makes another interesting observation in connection with classifying
adjectives. He points out that they are usually used together with nouns which have
general meaning, such as (Teyssiers examples) man, woman, child, girl, and bird.
Nouns with prototypically unique reference, like father and mother do not co-occur
with adjectives with classifying function. Since the classifying function of
adjectives creates subcategories it borders on compounding, especially endocentric
compounds (cf. Bauer 1983), a fact which will be given further attention below.
Let us finally consider descriptive adjectives, which form an intermediate
category in Teyssiers (1968: 229) view, but, he observes, the additional
qualification half way between identifying and classifying adjuncts is necessarily of
a different nature. It merely describes the noun, the object, by the indication of
some, at a particular point in time, relevant features of a thing. This means that the
relation between the noun and the attribute differs in character from the cases
discussed above in Jespersens terminology we can say that descriptive attributes
are non-restrictive. Contrary to the classifying function, the descriptive function
focuses on specific instances and, accordingly, we can describe this function as
referent-modifying, in the terminology of Bolinger (1967). Descriptive attributive
adjectives are also necessarily predicating, and can usually be paraphrased in a
complement structure (copula + predicative complement) as in a red book a book
which is red. Magnusson and Persson (1986: 207) suggest that the aim of
description
is to bring an entity into focus either for the main purpose of drawing attention to
some of its properties or to do so as a piece of side-information as something else is
predicated about it. The entity described is perfectly independent in the context, and
there is no reason to be interested in classifying or identifying it. [emphasis added]

Thus, it is the property that attracts the focus of attention, not the entity. If we go
back to example (10) above (the maroon-clad waitress), we could argue that it was
the colour maroon that was the main interest. The intention was probably to tell us
that all waitresses in this restaurant or bar wore an outfit of this colour. As we shall
4

The properties should last long enough or be repeated with such frequency that they can serve as reference
points for the identification of the type.
5
This is also clarified by Warren (1988: 130): Although, admittedly, there are some adjectives that are
basically descriptive [] and some tend to be classifying [] these are in no way restricted as to these
functions. The former may very well serve as classifiers or identifiers and thus become non-predicating

102

Chapter 4

see in the next chapter, the focus on the colour itself is of essential importance for
the use of many ECTs. This is also to say that any type of property can be used in
descriptions, irrespective of whether or not the property is permanent outside the
current discourse space.
To sum up this account of adjectival functions, we can note that they operate
on different levels of abstraction and with different purposes. Whereas identifying
and descriptive use refers to token entities with specific reference, classifying use is
more abstract in that it is concerned with type entities with non-specific reference.
Furthermore, we have observed that there can be said to be an underlying syntactic
difference codified in Jespersens terms restrictive non-restrictive, which
describes the relation between the modifier and the head. These differences are
summarised in Table 4:1 below.
Table 4:1.

Discourse functions of attributive adjectives

LEVEL OF REFERENCE

MODIFIER-HEAD RELATION

TYPE

TOKEN

RESTRICTIVE

Classifying

Identifying

NON-RESTRICTIVE

Descriptive

4.3 Planes and valence relations

In order to make a proper description of adjective-noun phrases we need to
establish how the functions that were identified above can be described in the
framework of cognitive grammar. Let us first consider the nature of the ascription
that adjectives typically perform. In Warrens (1984, 1988) view, the semantic
content of adjectives can be divided into two components: referential content and
a connecting link.6 As the term suggests, the referential content constitutes the
substantial meaning of an adjective. The connecting link, on the other hand,
organises the relation between the modifier and the head. Warren suggests that the
relation can and should be analysed in terms of semantic roles.
We can illustrate this type of analysis by looking at her (1984: 22) treatment
of criminal in criminal assault, criminal case and criminal court. The referential
meaning of criminal can be identified as crime in this view. However, we get the
full adjectival meaning of criminal only by considering whole noun phrases. These
meanings are paraphrased below; the connecting links are given in small caps:

6
As Warren herself points out, this mode of analysis was also used by Aarts and Calbert (1979). They,
however, use the term (predicational) relator, rather than connecting link.

Colour Terms as Adjectical Type Modifiers

103

Criminal assault: assault CONSTITUTING a crime; Role combination: RESULT

SOURCE.
Criminal case: case CONSTITUTED BY a crime; Role combination: SOURCE
RESULT
Criminal court: court DEALING WITH crimes; Role combination: AFFECTED
OBJECT ACTOR.
This account is extremely convincing, especially when Warrens rich material is
taken into consideration. However, it should be pointed out that Warren is
primarily concerned with denominal adjectives. Nevertheless, in a later article
(Warren 1988), she suggests that central adjectives may also show the same
patterns of referential content and connecting links. As regards the status of the
connecting links, Warren argues that there is a limited set of covert links
structural default devices in a language but she acknowledges that there are also
other, idiosyncratic, formations. Downing (1977) in her study of nonce compounds
strongly rejects the idea of a limited number of connecting links, and she (1977:
840-41) points to the fact that
because of the important differences in the functions served by compounds, as
opposed to the sentential structures which more or less accurately paraphrase them,
attempts to characterize compounds as derived from a limited set of such structures
can only be considered misguided. A paraphrase relationship need not imply a
derivational one.

As we observed above there is a close affinity between compounds and the

classificatory usage of adjectives; thus there is little reason to believe that there
should be any marked differences, if any at all, between these two processes.
In Langackers (1987, 1991a, 1991b, 1999) conception of cognitive grammar,
he accommodates Downings claim that it is the nature of the entity that determines
the relationship. Consequently, he prefers to talk about the formation of complex
structures in the metaphor of valency. In chemistry, a valence relation means the
sharing of electrons among atoms to form molecules. In linguistics we can
understand it as the sharing of elements, or structures. Langacker (1987: 278)
points out that [i]t is only by virtue of having certain substructures in common that
two component expressions can be integrated to form a coherent composite
expression. Although he stresses the importance of the internal structure of the
components and the, in principle, very rich number of potential combinations,
Langacker acknowledges the fact that established conventions usually sanction
only a fraction of the available combinations, but they dominate if we look at
instances. Of course, Warrens well-documented work clearly testifies to the truth
of such an assumption. However, the important contribution of Downing and
Langacker is that their approach takes the focus off the connecting link itself and
puts it instead on the nature of the shared structures. As an example on how valence
relations work we can mention old in various combinations. As demonstrated by
Taylor (1992), we can interpret an expression like old friend in two ways: that the
person (the friend) is old, or (more likely) that there has been a fairly close relation
with this person for a long time. These two possible readings are linked to the two

104

Chapter 4

possible valence relations that can be established between old and friend. Both
interpretations reside in the structure of friend: either it is the FRIENDSHIP structure
that determines the valence relation or it is the HUMAN structure (a friend usually
being human).
If we look more carefully at colour term usage we can observe that there is
little controversy here. Colour terms normally ascribe colour to the surface of an
entity, as in (11) below. Accordingly, in Warrens terminology, the connecting link
can be identified as HAVING and the role combination as PART-WHOLE.
(11)

The Senator is the ideal host. You feel that, when you arrive in front of the
green house in Mount Vernon Avenue with the wide porch that runs around the
corner. (BoE: usbooks)

However, on closer examination we can find that the picture is somewhat more
complex than this. Unless we are prepared to give some very loose interpretations
to the notions of connecting link and the role combinations, we cannot readily say
that the above mode of analysis is able to explain the nominal phrase red pencil, in
the sense a pencil which leaves red marks when used. In cognitive grammar we
can explain the formation of such a structure through the concept of active zones.
The notion of active zone was described in Section 2.5.2. It seems that type
modification involving colour terms frequently exhibits active zone features.
Another example of an active zone phenomenon in a classifying usage is black
eye.7
What we also need is a concept of reference which can accommodate the two
levels mentioned above (cf. Table 4:1) type and token. In his most recent book,
Langacker (1999) offers the construct of plane for treating the levels of reference
observed above. A crucial distinction that has to be made is that between type and
instance in Langackers view. This division has in fact greater significance within
cognitive grammar, but for the present purpose I restrict my attention to the noun
phrase. Langacker (1991b: 53) claims that [t]he semantic content of a simple noun
like site amounts to nothing more than a type specification: it specifies the basis
for identifying various entities as being representatives of the same class but not
tied to any particular instance of that class [emphasis original]. This means that a
type specification directs our attention to certain objects but makes us exclude
others in our mental activities. The instance level, which is represented by a full
nominal, the site, presupposes instantiation of the type in question and designates
one or more instances [emphasis original]. (1991b: 53)
Thus, a nominal adds additional information to the type specifications;
essentially two pieces of information are present in a nominal:

Technically, it is not the eye that is black, but the area around the eye. For another interesting discussion of
categorization and type modification, but within a theory of a slightly different nature (conceptual blending),
see Turner and Fauconnier (1995), who, among other things, discuss red pencil.

Colour Terms as Adjectical Type Modifiers

105

1) there is some indication as to the quantity of the type;

2) there is some indication as to how the type relates to the speech event and
the participants therein this second feature is called grounding by
Langacker.
The differences we have established between type and instance can be codified
metaphorically in Langackers suggestion of planes. Type plane is far more abstract
and can be instantiated by several instances as demonstrated in Figure 4:1 below.
Figure 4:1b illustrates that grounding is an essential and necessary part of the
instance plane. Since generics usually lack grounding a precise connection with
the speech event it follows, as observed by Magnusson and Persson (1986: 205),
that the attribute has a classifying function.

ti

Type
Plane

tj

tk
tl

a) Four instances of the single type T

oak

oi

Instance
Plane

this
oak
G

b) Grounding of an instance

Figure 4:1. Type and instance plane.

(Based on Langacker 1999: 271) G = grounding

Another central aspect is the fact that type specification can be quite complex. As
Langacker (1999: 272) points out, there is clearly no limit, as we can go on to
form ungrounded compounds evoking type specifications of indefinite complexity
(e.g. cat-lover hater behavior modification school instructor). Relating to Figure
4:1a, we may for instance assume that there is a subclass TI which contains
instances ti and tk, which may be represented as in Figure 4:2 below. The figure

106

Chapter 4

illustrates how a quality Q shared by some instances can be used for modification
in type plane, thus creating a subtype, TI.
In the rest of the chapter, I try to explore whether, as concerns colour terms,
there are any restrictions on what colour terms can be used as type modifiers. First,
however, a discussion of the difference between compounds and phrases.

Type
Plane
TI

ti

Figure 4:2.

tj

tk

tl

Instance
Plane

Type modification and subclassification.

4.4 Compounds, phrases and lexicalisation

Before we approach the empirical part of this chapter, we should stop and consider
some aspects which in a sense are peripheral, but are nevertheless relevant to this
study. One such aspect concerns the issue of whether it is possible to make a clear
distinction between compounds and adjective-noun phrases that represent type
modification. Since type modification, as described above, creates subtypes of an
established class of nouns it would seem that this process is very close to that of
compounding. Bolinger (1967: 32) notes that any attempt to make such a
distinction would yield an idiosyncratic result: It appears that there is no way to

Colour Terms as Adjectical Type Modifiers

107

draw a line between reference modification [= type modification] and compounds.

A similar observation is made by Halliday (1994: 185), discussing the difference
between compounds and phrases. He points out that the line [] is very fuzzy and
shifting, which is why people are often uncertain how to write such sequences.
From the point of view of cognitive grammar, the difference between
compounds and syntactic phrases is really a non-issue. Semantically, both phrases
and compounds form complex composite structures, the analysability and
compositionality of which may show degrees of difference. The concept of
composite structure is defined by Langacker (1987: 487) as [a] structure that
results when two or more structures in a given domain (phonological, semantic, or
symbolic) combine in a valence relation. Traditionally, two criteria are mentioned
as rules of thumb for the identification of compounds (cf., for instance, Fromkin
and Rodman (1998) or Dirven and Verspoor (1998)); they are stress and meaning:

a compound has one main stress usually on the first syllable whereas
phrases have two stresses, the main one falling on the second element.

The sum of the meaning of the forming elements is not equal to the
meaning of the compound.

However, as might be expected, these criteria are by no means as clear-cut as they

appear. Bauer (1983) discusses the issue of stress at some length and concludes that
stress is not a defining criterion for compounds.
Turning to the other criterion, the sum of the meaning, it would seem that
we can translate this into a question of the compositionality of the structure, in
cognitive terminology. Compositionality is usually understood as referring to the
composition of an entity; that is, to what degree we can predict the meaning of the
whole from the meaning of the parts. Full compositionality thus entails that the
meaning of the parts of an expression fully covers the meaning of the whole.
Langacker (1987) points out that the underlying problem of compositionality, as it
is usually discussed, is the building block metaphor. This metaphor forces upon us
a conception of meaning as building blocks which are put together in accordance
with some design (= syntactic rule, or, to use Warrens (1984) vocabulary,
connecting link), so that [A] + [B] = [AB] = [C]. This is the pattern we expect from
the most common nominal phrases so that green + house = green house. However,
in cognitive grammar, there is a much more flexible view of meaning in that it
comprises encyclopaedic knowledge. Langacker (1987: 453) therefore suggests that
the composite structure should be viewed
as a coherent structure in its own right: component structures are not the building
blocks out of which it is assembled, but function instead to motivate various aspects
of it. From this perspective there is nothing problematic about a composite structure
evoking a knowledge system (abstract domain) to which neither of its components
provides direct access; nor is it surprising that in many instances the component
structures motivate and highlight selected facets of the composite meaning, but
nonetheless fail to exhaust its content.

108

Chapter 4

The notion of motivation is central in cognitive linguistics and Lakoff (1987)

discusses compound formation in a similar fashion. He too observes that the parts
in a compound motivate the meaning of the whole but, he (1987: 147) stresses,
more is required: a relevant ICM where each part of the compound fits some
element of the ICM. Recall (Chapter 2) that Lakoffs ICM and Langackers
abstract domain appear to be more or less the same type of construct. As an
example, Lakoff mentions chains of compounds such as topless dress, topless
waitress, topless dancer, topless bar, topless district. Arguably, to understand an
expression like topless bar, we need a knowledge structure which goes well beyond
that provided by the two components in isolation. We cannot understand or
successfully analyse the phrase in the old building block metaphor.
Nevertheless, the formation of this compound is well motivated if one has
acquired the necessary knowledge. In the traditional building block metaphor we
can illustrate the meaning of [C] of such a compound as [ABX], where [X]
represents some additional meaning; that is, the meaning of C > [AB] (cf.
Langacker 1987: 450). Furthermore, it may also be the case that the meaning of the
composite structure builds on stretched meanings of the components so that [C] =
[AB], as in the case of greenhouse. This compound follows the established pattern
of boathouse and henhouse; the difference being that [A] represents a special case
of [A],8 here a metonymic extension.9
A related issue is that of lexicalisation. From the point of view of cognitive
linguistics this issue is best discussed in terms of the concept of entrenchment,
which was described in Section 2.2. In short it suggests that repeated occurrences
of some concept or combination of concepts make it easier to access them again. In
the context of lexicalisation, we can note that Langacker (1999: 93) gives the
following description of what the entrenchment of such structures leads to:
When a complex structure comes to be manipulable as a pre-packaged assembly, no
longer requiring conscious attention to its parts or their arrangement, I say that it has
the status as a unit.

If such a unit has symbolic representation we can reasonably equate this with
lexicalisation in its traditional sense. It is important, however, to appreciate that we
are once again concerned with fuzziness; both Downing (1977: 839) and Langacker
(1987: 59-60) emphasise that it is in reality impossible to decide exactly when unit
status is acquired.
To recapitulate this section, we can tentatively summarise the fuzzy notions of
compounding and syntactic phrases by means of a figure. We have seen that
compositionality and entrenchment provide important dimensions for our
categorisation of a phrase as a compound. It is tempting to see these notions as
representing two mutually independent axes, as in Figure 4:3 below.

It is interesting to note that on those occasions when [B] represents a special case of [B], this leads to a
different class of compounds; for instance, a turncoat is not a type of coat. Cf. Bauer (1983) on the
traditional division between endocentric and exocentric compounds.
9
In line with Levis (1978) analysis of noun-noun compounds, we can identify the covert predicate in these
expressions as FOR.

Colour Terms as Adjectical Type Modifiers

Syntactic
phrases

109

High compositionality

Low
degree of
entrenchment/
conventionalisation

High degree of
entrenchment/conventionalisation

Compounds

Low compositionality

Figure 4:3. Typical characteristics of compounds and syntactic phrases in

relation to two cognitive dimensions.

The diagram is intended to illustrate the general tendency in the process by which a
syntactic phrase becomes a compound. If a composite structure, which a syntactic
phrase constitutes, is repeatedly used with reference to a particular type, it will
become more deeply entrenched, and in so far as the type becomes more
specifically defined, the phrase will lose in degrees of compositionality. If we relate
this more specifically to type modification, we can observe that generics (such as
red roses in Red roses are rare these days) will typically belong to the upper lefthand corner, whereas other instances of type modification can be more deeply
entrenched and decompositional to a lesser degree.
One such example could be beige book, which does not refer to any beige
book, but to a periodical report of a survey of the economic conditions around the
USA conducted by the Federal Reserve. Moreover, compounds normally have a
fairly high degree of conventionalisation and can show, in extreme cases, close to
no compositionality.10 One such extreme is greenmail,11 which means The
practice of purchasing enough shares in a firm or trading company to threaten a
take-over, thereby forcing the owners to buy them back at a premium in order to
10

Gunnar Persson (pc) points out that the final stage in this development is when assimilation processes
have obscured the original compound on the phonological side as well. One such example is daisy (Bellis
perennis), Old English: des ae, days eye. Such fusions appear to be fewer after the advent of
printing.
11
Admittedly, the conventionality of greenmail in the whole British-American speech community can be
questioned. It is, however, conventional in certain types of discourse.

110

Chapter 4

retain control of the business. (OED definition) However, it should be emphasised

that Figure 4:3 is only meant to be suggestive; thus compounds may very well
exhibit close to full compositionality and still be perceived as compounds.
Similarly, as demonstrated by Downing (1977), compounds may be formed
for the nonce. One characteristic which seems to be obligatory is that, at an initial
stage, a composite structure is well motivated within the mental space (usually,
Current Discourse Space) contextually, situationally or linguistically. But, as
described earlier, if the phrase becomes conventionalised within a speech
community, it will be defined in relation to some domain matrix just like any other
expression, and in this process of abstraction, additional specific attributes may be
added (or centralised) or, indeed, taken away (decentralised). The fact that lipstick
nowadays can be bought in liquid form suggests that the traditional definition
meaning of the word lipstick, as that in the OED: A stick of cosmetic for colouring
the lips, is now slightly misleading. Hence, from the point of view of
compositionality, we can reasonably say that the compound now exhibits this to a
lesser extent (cf. Downing 1977).
To sum up, the purpose of the present section was to problematise two
established linguistic concepts: compounds and type modification. It was suggested
that these notions exhibit fuzziness in relation to important dimensions, and it is
therefore argued that we can best understand these notions as representing the poles
of a continuum.

4.5 Type modification in two different corpora

As stated above, the main aim of this chapter is to establish aspects of colour term
usage for the purpose of type modification. Having established some theoretical
frames, it is now time to turn to the language material. First, however, we will have
to formulate clearly what we are actually looking for, as this may have
methodological implications.
A number of questions seem relevant to the present study:

What colour terms are most frequently used for type modification and
compounding?
Can we identify a specific class of colour terms that performs this
function?
In what nominal domains do we find type modification involving colour
terms?

On a second level we may enquire

What mechanisms are involved in the apparent extension of terms (cf.

red hair)?
What level of specificity do the heads of the compound exhibit?
Does the colour term modify the whole area indicated by the head, or do
we find that some other linguistic phenomena are involved?

Colour Terms as Adjectical Type Modifiers

111

In view of our observation that compounds and type modification phrases cannot
be viably distinguished, it follows that a comprehensive study should include both
compounds and phrases. This creates a methodological problem. How do we obtain
representative material that contains both types? Some compounds are written as
one word, others and all phrases as two words. A second problem is how to
distinguish type modification from token modification in a corpus as this
distinction can be said to be cryptotypic it is functionally and semantically
important but hidden, lacking formal representation.12
To be able to give as broad a picture as possible I decided to make two
separate investigations, probing two different types of language material: the OED
and the BoE. By dividing my investigation in this way, I hope to be able to describe
and identify different aspects of this usage. The OED represents the broad, codified
aspect with a historical dimension, whereas the BoE represents preferences in
present-day usage. This means that while the OED provides figures which give us
an estimation of the width of type modification for a term (how widely a term is
used for type modification), the BoE gives us a picture of the salience of particular
kinds of type modification (what sorts of type modification occur most frequently).
The two investigations will be presented separately, but a summary will provide a
comparison.
4.5.1 The OED
In this part of my study, I counted and classified all entries (compounds and
lexicalised phrases) in the OED according to the colour term and the domain of the
noun. Phrases or compounds marked as obsolete were not included. My approach
more or less parallels that of Verspoor (1998),13 the major differences being the
source (Verspoor used Websters), the size (Verspoor confined her study to blue,
red and black) and the number of categories of nouns (I have opted for a more
refined categorisation). Another similar study is that of Bennett (1988). He
provides very rich lists of colour collocations; collocations which he collected from
a number of different dictionaries (including the OED), but also from newspapers
and magazines and through personal communication. However, the looseness of his
method and the fact that he also includes place names occasionally (for instance,
Greenland) have led me to choose my own, more well-defined, material.
Even if the material is well-defined, a dictionary approach can be problematic.
Verspoor (1998) claims that the confinement to one dictionary is preferable since
one can assume that there is a coherent strategy within the dictionary. True as this
may seem, I have come to doubt it during my work with the OED. Only words with
an entry of their own were collected, but occasionally I saw other forms as in the

12

The notions of cryptotypes and covert categories were introduced by Whorf. Lee (1996: Chapter 4) gives
an excellent account of Whorfs development of these notions and how other scholars have used them
thereafter.
13
I would like to thank Marjolijn Verspoor for kindly sending me her article.

112

Chapter 4

quote below, which, for some obscure reason, were not given their own entry
there is no explicit comment by the editors concerning the choices.14
2. a. In names of varieties of fruits or plants, as violet clover, maize, plum, etc.
b. In names of birds, insects, etc., as violet bee, cormorant, crab, creeper, heron, etc.
(The OED: violet)

This means that the figures presented below do not necessarily represent a coherent
strategy. Nevertheless, I feel confident that they provide a clear picture of the
general tendency.
Moreover, it should be mentioned that each entry was only counted once,
irrespective of the number of reported senses related or not. This procedure
probably disfavoured the most frequent terms in particular. My reason for
proceeding in this way was based on the notorious difficulty that the questions of
polysemy, homonymy and vagueness can cause (cf. Chapter 2). This approach
eliminated that problem altogether. The drawback is, of course, that I allowed the
lexicographers make my decisions for me. Figure 4:4 below gives an overview of
the colour terms which are preferred for the purpose of creating subclasses in the
OED.
The distribution of colour terms may not come as a great surprise in view of
the patterns presented in Chapter 3. We can observe that the Primary Basic terms
are those most frequently used for this purpose. In fact this group (with the possible
inclusion of grey) is fairly well-defined in terms of frequency. This is not the case
with the Secondary BCTs, which are mixed with non-basic terms.

400
288
218

200

139 130

150

117 105

100
50

77

62 53 49
44 38 33 32
21 8

Turquois e

250

Lilac
Crimson

300

Cream

350

Emerald
Amber

Number of phrases

450 410

Violet

Lemon
Peach

Olive
Indigo

Pink
Gold
Orange

Scarlet

Rose
Brown
Purple

Silver
Golden

Grey/Gray

Blue
Green
Yellow

Black

Red
White

Figure 4:4. Lexicalised phrases and compounds in the OED

14

For example, we find that red plum, black plum, grey plum and yellow plum have their own entry in the
OED, but not violet plum (mentioned in the quote above). One possible explanation for the patterns observed
above could be the fact that the OED has a long history involving several editors.

Colour Terms as Adjectical Type Modifiers

113

The null hypothesis suggests that there is no correlation between lexicalisation and
the B&K hierarchy. However, a rank correlation test showed that there was a
significant correlation between the lexicalised pattern above and the B&K
hierarchy (rs = 0.893, p > 0.01). There was also a high correlation between the
lexicalised patterns of the OED and the frequency rank of the BCTs in the BoE (rs
= 0.936, p > 0.001). Taken together this may be a strong indication that a term used
for type modification has a conceptually privileged position. This is further
discussed below.
On the individual level, we can somewhat surprisingly observe that rose has a
slightly higher figure than its supposed superordinate pink. One plausible
explanation for this is the age of the terms. Rose (although the hyponym) is older
than pink in its colour sense; rose was first recorded as a colour in 1530, and pink
not until 1720. It seems likely that rose may have been the more salient of the terms
until quite late. If we look at the nominal domains to which the compounds and
phrases refer, we can also detect a great difference. Out of the 53 instances of rose
no less than 48 (90%) refer to natural domains such as ANIMALS and PLANTS. The
figures in the case of pink are 33 and 20 (60%). On the other hand, 11 instances of
pink refer to artefacts or have figurative meanings whereas, in the case of rose, we
can only find two instances in artefactual domains and none of figurative use.
Again, this may be indicative of the lateness of the acquired salience of pink, as we
can assume that these artefactual subcategories were created later: all the first
instances listed in the OED date from the 19th and 20th centuries.
Another conspicuous fact is the privileged status of red among the chromatic
colour terms it occurs more than three times as often in compounds and
classifying phrases than any other chromatic term. A closer examination reveals
that red is used for this purpose particularly in the domains of plants and animals. It
is tempting to see the high frequency of red as supporting Sahlins (1976: 4) claim
that red stands out in relation to all other hues by virtue of a reciprocal
heightening effect between saturation and brightness. It should be borne in mind,
however, that the frequency pattern reflects primarily the great salience of the term
red. Since quite a few of these instances with red actually refer to objects whose
colour is only marginally red, we need a theoretical model to be able to
accommodate Sahlins claim. One example is red onion, whose colour, I think
many people will agree, is closer to purple than to red. I return to this issue in
Section 4.6 below.
If we then take a look at the domains, we can see that it is above all in natural
domains that type modification and compounding occur in the OED material. We
find much more infrequent use in domains related to humans or artefacts. Table 4:2
below presents the figures from my investigation for a collection of terms.
However, the presentation is restricted to the most frequent terms only. For most
colour terms listed here, the proportion between natural objects and the other
categories was that more than 60% of the instances referred to the natural category.
It should also be mentioned that among artefacts I have included everything

114

Chapter 4

manipulated by human beings, so, for instance, food and beverages are part of this
category, thus white meat and white wine are categorised as artefacts.15
Table 4:2. Domains of compounds and phrases in the OED.
Number of instances.
CATEGORY
Type of
Domain
Red

NATURAL OBJECTS/PHENOMENA
Diseases

Plants

Minerals
Animals
etc
142
42

HUMAN RELATED

Other
31

Parts of
Clothes
body
13
9

ARTEFACTS

Other

Concrete

Abstract

FIGURA
TIVE TOTALS
USE

105

38

410

White

15

62

49

21

20

14

10

16

52

14

15

288

Black

16

51

36

13

15

10

13

35

16

218

Blue

30

33

19

12

17

11

139

Green

20

36

10

20

24

130

Yellow

37

31

10

117

Grey/Gray

11

45

14

14

105

Silver

28

32

77

Golden

40

17

62

Rose

19

15

53

Brown

11

15

49

Purple

18

14

44

Scarlet

14

16

38

Pink

33

Gold

21

32

Orange

21

TOTALS

70

463

511

151

116

56

56

46

233

42

72

1816

How can we explain this clear tendency? The fact that the majority of instances
serve to form classes (or rather, subclasses) of natural objects animals, plants,
minerals and diseases seems to indicate that we can find a close affinity between
the function of colour terms here and the naming principle of taxonomic structures
of biology and geology. In a frame semantic study of names for fungi, Persson
(2000) demonstrates that [colour] is one of a limited number of naming attributes.
This makes sense since colour is a very important signal tool in nature. Consider,
for example, how male birds change their plumage for the mating season so as to
attract females. Similarly, poisonous frogs in the rainforest are very brightly
coloured to give clear signals. It is logical then, that these patches of colour would
serve as reference points for naming. The same logic applies to minerals and plants.
One aspect that may favour colour over many other attributes is that it is an
attribute which is very often accessible at a distance. Other attributes such as
[smell], [taste] and [locality] (mentioned by Persson 2000: 306) are not as easily
accessible, and not as transparent in the context of, say, a walk in the forest.

15

Note that modifications of colour terms, e.g. blue-grey were not included. Furthermore, I only included
instances in which the colour term serves as the modifier; thus a phrase like Coventry blue was not taken
into account.

Colour Terms as Adjectical Type Modifiers

115

This salience of colour makes it an excellent naming device for creating

subcategories of a more general category. Consider, for example, naming habits in
relation to various varieties of trees. In the case of oak, we can find mentioned in
the OED: blue oak, black oak, golden oak, grey oak, red oak, scarlet oak and white
oak. Indeed, there is also rose oak, which, strictly biologically speaking, is not an
oak but some Indian species of rhododendron, according to the OED.16 As Table
4:2 illustrates, this kind of subcategorisation is quite common among plants as well
as animals and minerals. It would be tempting to see the head in such compounds
(here oak) as representing the basic level. However, Rosch et al. (1976)
demonstrated that this does not seem to be the case. In their study they could
demonstrate that what they had assumed to be superordinate categories (TREE, FISH
and BIRD) in fact showed basic level characteristics, and that OAK, BIRCH, EAGLE
and SPARROW were subordinates.17 Rosch et al. discuss this phenomenon at length
and conclude that the basic level might previously have been at a lower level of
abstraction in English-speaking societies. They (1976: 431-432) point out that
[t]hat English speakers once made basic level distinctions at the level of the genus is
suggested by the number of monolexemic terms for biological genuses available in
English (see the bird, fish, and tree names in Battig and Montague, 1969) Such lists
of terms are not similarly available for subordinates of chair, shirt, hammer or for any
other nonbiological basic level terms in our taxonomies. Thus, biological taxonomies
are probably of the type in which two basic level groupings are possible, but our city
dwelling subjects appeared to be ignorant of the attributes characteristic of the generic
level biological categories in their environment.

Irrespective of whether or not we treat the head as basic level, the figures above
testify to the usefulness of colour as a naming attribute for subclasses. In view of
these observations, it is interesting to look at Kay and Maffis (1999) brief
discussion of the impact of colour in different societies. They (1999: 746) claim
that
in a technologically simple society, color is a more predictable, hence less
informative, property of things than in a technologically complex one. Except perhaps
for a few pairs of closely related species of birds or of fish, it is rare that naturally
occurring objects or the artifacts of technologically simple societies are
distinguishable only by color. In technologically complex societies, on the other hand,
artifacts are frequently to be told apart only by color. [] [A]lmost every kind of
material thing we encounter in daily lifeclothing, books, cars, housespresents us
with the possibility that two tokens of the same type will be distinguishable only, or
most easily, by their colors.

This is perfectly true, of course. The question, however, is whether Kay and
Maffis conclusion is correct (p 746): As technology develops, the increased
importance of color as a distinguishing property of objects appears to be an
important factor in causing languages to add basic color terms, i.e., to refine the
16

The OED is not the best of reference in this context, of course. However, for the present purpose an exact
designation of the species is not necessary.
17
It should be pointed out that here we are primarily concerned with folk taxonomy; the scientific
classification in classes, orders, families, genera and species is of minor interest in this context.

116

Chapter 4

lexical partition of the color domain (Casson 1997). This speculation is based on
the pattern that Berlin and Kay found, which suggested that there was a correlation
between technological complexity and the number of BCTs in the language.
However, what Kay and Maffi actually discuss here is instance modification and
to be more precise the type which Warren (1984) calls identifying function. To my
mind, we could turn this argument on its head if we take a wider view: precisely
because colour is predictable in nature it is very informative and useful as a naming
device for the identification of species or subspecies types. For this purpose wellentrenched, basic terms are needed. Consider again Table 4:2, and the frequency of
type modification and compounds in the categories of natural things. Since there is
no other difference between a white and a brown (or say, a beige) raincoat except
the colour, this distinction has very little informative value in terms of
classification. Of course, the value increases if we pay special interest to the colour
itself, due to fashion trends, for instance. However, for this purpose there is no need
for a basic colour vocabulary. I shall return to this issue in greater detail in the next
chapter as I discuss ECT usage, i.e. non-basic terms. My main point here is that
type modification is likely to be of great importance from a functional perspective,
and that colour serving as a central naming attribute should stimulate the
development of BCTs. However, as Table 4:2 indicates, it may be the case that
only a few basic terms are needed for this purpose. It could also be that the method
of identifying BCTs in primitive cultures has, in itself, blocked the recording of
important type modification terms. This of course is mere speculation.
Finally we can note that another interesting aspect of type modification is that
it frequently allows the colour term to be stretched outside its normal boundaries,
cf. for example white coffee, blue oak and red onion, whose colours would not be
designated by these terms had they been presented in the context of a colour chart.
We shall take a closer look at the processes involved in this pattern, but first we
shall consider the other method outlined above a study of the BoE.
4.5.2 The BoE
When looking for type modification in the BoE, we are faced with the
methodological problem of finding instances of this function. How can we
distinguish these from instance modification? Finding all cases of type
modification in the BoE would be a Herculean task outstripping even the Augean
stables, so obviously we have to set some limits to our study.
Below I present a study of the most frequent collocations of each colour term
in my sample. The thirty colour adjective + noun phrases that had the highest tscore values18 for each colour term were considered. Combinations that had a tscore value lower than 2 were omitted (in reality, this means n < 4). Since type
modification creates subtypes, it follows that this type of adjective is placed close
to the noun, cf. Teyssier (1968) and Warren (1984). This is of course a position
open to instance modification too. However, we would expect type modification to
show strong collocational ties, since it creates subtypes. After the thirty phrases had
18

In principle, I could just as well have used frequency, but the t-score method has the advantage of
eliminating very frequent words such as grammatical words. Section 0.3.3 gives a very brief summary of the
methods of calculating collocational strength.

Colour Terms as Adjectical Type Modifiers

117

been collected, I tried to determine whether or not each individual phrase could be
regarded as representing type modification. For this purpose I used the context that
could be provided in the BoE and two test devices introduced by Warren (1984:
86). Type modification is not gradable and thus the adjective cannot take the
premodifier very without some change in abstraction. Consider the less felicitous a
very black olive, which can only be interpreted as token modification. Furthermore,
as a category has been formed, negation is performed by the prefix non- rather than
un-. Consider (12) below; *unwhite does not exist of course, but the use of non
indicates the status of white washes as a case of type modification.
(12)

For the uninitiated, micro and ultra refer to concentrated products.

Colour products are best for non-white washes, as they do not contain bleach
or whitening agent. (BoE: today)

Despite these tools and a vast context, determining whether a phrase represents an
actual instance of type modification can still be quite problematic. One such
example could be the phrase the amber liquid, referring to beer19 as in (13) below.
(13)

A good site for connoisseurs of the amber liquid is the Beer Masters Tasting
Society at # Beermasters # (BoE: oznews)

The phrase is used here in a way that would give amber type modification status
although a phrase such as amber liquid would not be seen as a subtype under
normal circumstances. The trouble is that some colour term phrases can represent
both type and instance modification. In (13), for example we have a case of generic
reference. Generally, I tried to estimate how frequent the type usage is; that is, if
there was only the odd generic phrase, such collocations were not included. I tried,
on the whole, to be inclusive, but in the end it only seemed to affect the figures
marginally. Because of this difficulty I chose to present both the generous and the
more restrictive figures in the diagram below.
Another methodological detail should be mentioned. Sporadically both the
singular and the plural forms of a word occurred among the 30 phrases under
consideration (e.g. white man and white men). On such occasions, only one form
was considered and the sample was supplemented with the most frequent noun
phrase outside the original 30. Despite this procedure, there are quite a few phrases
in the material which have similar meanings but contain distinct words. Thus, my
material contains the phrases black person, black population and black people.
Finally, it should be stated that compounds written as one word were not
taken into account. Again, to do this proved difficult in view of the method outlined
above. However, since compound material was included in the dictionary part of
my investigation, the exclusion of compounds here should not affect the general
impression.20
19

The phrase amber liquid is also frequently used in reference to whisky.

Another fact which should be mentioned is that gold was not included in this study; mostly because of the
difficulty of isolating the colour meaning. In the case of golden and silver, it was only when the colour
designation was clear to me that the collocations containing these terms were included.
20

118

Chapter 4

Below I present the overall frequencies. The continuous line represents the
restrictive approach to type modification, whereas the broken line represents the
more inclusive approach. As we can see, it springs few surprises: the pattern is very
similar to that of the OED and of the frequency list presented in Chapter 3: BCTs
occur most frequently, in particular the Primary BCTs. A rank correlation test of
the BCTs showed that there was significant correlation with the B&K hierarchy (rs
= 0.843, p < 0.01) and also the frequency rank in the BoE (rs = 0.817, p < 0.01). As
we can see the BCTs form an identifiable group apart from the presence of golden,
silver, and scarlet. There is, however, a great span within the group.
Furthermore, as stated earlier, we can see that an inclusive approach to what
is type modification does not alter the picture significantly. The most conspicuous
change is that the gap between golden (or blue) and purple increases, creating two
distinct categories. Another noticeable fact is that we can see that brown occupies a
much more prominent position in this study than in that concerned with the OED.
On closer examination, it appears that brown occurs frequently in the domain of
artefacts, but more seldom in that of natural objects. The fact that the colour
collocations listed in the OED favour the domain of natural objects made brown a
much less salient colour in that study than in this one. Recall also the different
aspects that the two corpora are likely to emphasise.
30

Instances of types

25
20
15
10
5
Black
Red
Green
White
Brown
Yellow
Grey/Gray
Blue
Silver
Golden
Purple
Scarlet
Pink
Orange
Maroon
Violet
Tan
Rose
Amber
Emerald
Beige
Turquoise
Olive
Indigo
Cream

Figure 4:5.

Number of type modifications among the 30 most frequent

collocations (T-score > 2.0) in BoE.

There are a few facts which deserve closer inspection. But before we do that let us
consider the categories of the type modifications. Since the instances are far fewer
than in the OED, I have restricted this categorisation to three categories only, which
correspond to the overall categories in the OED table natural objects, humans and
artefacts. When we look at Table 4:3 below we can note some differences as
compared to the OED table above.

Colour Terms as Adjectical Type Modifiers

Table 4:3.

119

Type modification in different categories in the BoE.

COLOUR
TERM
black
red
green
white
brown
yellow
silver
grey
blue
golden
purple
scarlet
pink
orange
TOTAL

CATEGORIES OF THE HEADS

nat obj
2
6
5
1
5
3
8
4
1
8
3
2
1
5
54

humans
14
3
2
11
3
2
0
4
5
0
0
2
0
0
46

artefacts
10
14
13
5
9
9
3
3
5
2
3
1
4
0
81

TOTAL
26
23
20
17
17
14
11
11
11
10
6
5
5
5
181

Table 4:3 demonstrates some interesting details. First, we can see that the category
most frequently exposed to type modification is that of artefacts. Furthermore, it
can be observed that the category of natural objects, which dominated the
lexicalised material, occurs comparatively seldom among the most salient
collocations of type modification in the BoE corpus. Given the type of texts we find
in the BoE, and the method of collection, this was to be expected. In fact, it is only
golden and silver which are frequently used to this end. Presumably, their
polysemous nature (the metal sense being prototypical) disfavours the use of these
terms in the artefactual category21. Finally, we can note that the most frequent
classification in the human category is based on black and white. Given the social
structure of the USA, Great Britain and Australia, the prominence of this
distinction was a likely pattern.
It would seem that a detailed comparison between black and white could be
particularly fruitful, since they are the only terms which demonstrate clear
antonymy (cf. Hunt Lazerson 1977). It might be expected that some kind of
symmetry could be found, although the method would favour the marked variety as
it were the one needing the extra specificity. If we compare the actual types of
words which are most frequently modified, we get an interesting picture.
Table 4:4 shows instances of type modification combinations of black and
white that were among the 30 most frequent adjective-noun phrases in the BoE. If
we consider the upper part of the table we find little symmetry. In most cases there
is not a corresponding opposite form: there are no white holes or white eyes, nor
any black blood cells (they are red, of course) or black flour.

21

It could, of course, be due to the classification of the present author. In order to avoid inclusions of the
metal sense, I may have been too harsh in my judgements.

120

Chapter 4

Table 4:4.

Type modifications in BoE containing black and white.

PHRASES RELATING TO RACE

NATURAL OBJECTS
Black
White
olives
water
holes

HUMANS
Black
White
hair
hair
eye
blood (cells)
collar
knight

man
woman
Americans
students
children
person
people
leaders
population
majority
guy
players

man
woman

ARTEFACTS
Black
White
box
wine
coffee
paper
pepper
bread
tie
flour
flag
township
music

ABSTRACT
Black
comedy
magic
market

community

people

minority
guy
males
supremacist

The most conspicuous difference is that which we can find between black and
white with reference to race. As we can see, although there is some symmetry, there
is also a considerable overrepresentation in the black field. Apparently, it is much
more common to identify black people as blacks, than white people as whites
unless, of course, we would like to claim that the BoE contains many more
references to people with black skin. This is not likely to be the case. How should
we understand this? Lee (1992: 92) analyses the use (in an English newspaper) of
black in a phrase such as black township as a marker of the perspective of the
speaker: The reference to the black township of Soweto rather than simply to the
township of Soweto is directed to an expectation on the part of the reader that
people are white, unless otherwise specified. Lee goes on to claim that in an
African newspaper directed to black readers this type modification would be
unnecessary. Even if we may quibble over some of the details in the analysis, it
would appear that Lee has identified an important aspect here. In the population of
types in Table 4:4 above no less than 15 relate to the black race and only seven to
the white race. If we compare the individual frequency of the phrases which occur
with both colour terms we get additional evidence for this pattern:
Black people (2110) vs. white people (932)
Black man/men (2147) vs. white man/men (1705)
Black woman (551) vs. white woman (322)

Colour Terms as Adjectical Type Modifiers

121

The most reasonable interpretation of these figures is that they support Lees
analysis. Black people form a minority in all the countries (Great Britain, the USA,
Australia) from which language material has been collected for the corpus. What is
more, the media are also controlled by the white majority. Thus, the corpus can be
said to reflect a white perspective. Furthermore, a plausible interpretation would
embrace a wider perspective and assume that it is not only, as Lee suggests, the
writer and the audience who are important factors here but in particular the
structure of society. It is part of our (subsumed) knowledge of American and
British society that students are usually white, thus it may be important information
to pass on that a particular group of students is black. This would be true even if the
writer and the reader were black. Consequently, I would prefer to understand the
frequent classification of people etc. in terms of black as reflecting the structure of
society in which blacks represent a peripheral group (in terms of power and
influence), than restricting its use only to the perspective of the writer or an
assumed audience. As we shall see below, this pattern is closely linked to the
overall pattern of type modification it would appear that type modification is
primarily used for the designation of subtypes which are peripheral in the overall
category. In this context, it should also be mentioned that once a phrase or
compound has been conventionalised, it may take on new attributes. This is
certainly the case with black in this context which also has strong social attributes.
4.5.3 Summary: a comparison between the two approaches
To sum up, although in a sense they measure different aspects of the speech
communities, the two corpus investigations present results that are remarkably
similar in certain respects. The same colour terms occur in the top positions, and
both sequences correlate with the overall frequency pattern of the BoE corpus. The
OED study gave us a general picture of the frequency with which a term is used for
type modification, and it was demonstrated that red, white and black predominated.
The category most often modified was that of natural objects. The BoE study, on
the other hand, gave us some idea of what kinds of type modification occur most
frequently in Present-Day English. The same group of colour terms occurs most
frequently in this study too, but there was a marked difference in terms of domain.
Natural objects occurred very seldom here. Presumably, this reflects the different
characters of the corpora as well as the different methods. In total, it seems that
colour terms which are very frequent are likely to be used for the purpose of type
modification. One way to interpret this is that the salience of the term and the
colour are decisive.

4.6 Decoding the process

As a final part of this chapter, I look more closely at the actual process of type
modification. Without recognising explicitly the process of type modification some
scholars have commented on the formation of such phrases, observing the
metonymical patterns involved. In this section I put special emphasis on the process
since it might explain the apparent restriction on which colour terms are used to

122

Chapter 4

this end. A further issue that is addressed is how type modification relates to
basicness and figurative usage.
It became apparent in both the dictionary and the corpus studies above that
there are, in fact, very few terms which are frequently used for type modification:
there are the six Primary BCTs and to a certain degree also brown, grey, golden and
silver. It would appear that the formation of subclasses with the help of colour
terms is a very good example of prototype phenomena. The colour terms are used
to classify marginal subtypes22 of a category, and frequently the colour of the
object constitutes an instance of the colour word which is fairly distant from the
focal colour. Thus we can talk about prototype phenomena on two levels, as it
were: within the colour domain and within the domain of the item.
The fact that colour terms occasionally designate unorthodox nuances has
been observed previously by some scholars. Bennett (1988) uses the term
radicalisation to account for this phenomenon. Mainly discussing it in the context
of the black-white dichotomy, he mentions examples such as black and white coffee
and black and white mint. No clear definition is offered of how we should
understand the phenomenon, and at the end of his discussion, Bennett (1988: 33)
suggests that the corollary of radicalisation is de-colouring and that
radicalisation is in fact a simplification of the communicative task.23 Verspoor
(1998: 10), discussing compounds containing colour words, points out that
at the semasiological level, color words in these compounds often refer to peripheral
members of the color category, especially when they serve to distinguish entities
(animals, flowers, babies) from more prototypical members of a particular category.

However, it is Clark (1992) who has addressed this issue most thoroughly from a
theoretical point of view. He discusses the phenomenon in terms of conceptual and
lexical possibilities and three broad constraints which apply to the conceptual
possibilities: similarity constraint, preference constraint and exhaustiveness
constraint. Clark (1992: 371) defines the three constraints as follows:
Similarity constraint. For each salient possibility in the specialized conceptual field,
apply the term for the most similar possibility in the general conceptual field.
Preference constraint. For terms in the specialized lexical field, prefer common
over uncommon terms from the general lexical field.
Exhaustiveness constraint. Partition the possibilities in the specialized conceptual
field in such a way that, for most practical purposes, they are exhaustively covered by
the chosen terms with the maximum amount of information value.

He observes that there are usually two sets of possibilities that interact in
communication: first, different conceptual fields contain different sets of
possibilities; second, the corresponding lexical field contains a number of
possibilities. In the case of skin colour, for example, Clark (1992: 370) identifies a
number of aspects, the important ones for our purpose being: skin color is highly
22

The whole process of subcategorisation appears to start with marginal subcategories.

Using Bennetts terminology, Wyler (1992) observes that other colour words apart from black and white
are used in the same fashion. However, his line of reasoning is at times very obscure. For a critical review,
see Biggam (1995b).
23

Colour Terms as Adjectical Type Modifiers

123

correlated with race, skin color can change with emotion and sickness, skin color
can change with exposure to the sun. An illustration of the function of the abovementioned constraints is given as he explains why white people are not called pink.
The answer is that white is the commoner term, a basic brightness term, and so it is
preferred over pink. That is, as English evolved, there was a preference for assigning
the commonest possibilities in the lexical field to the commonest possibilities in the
conceptual field, even if it led to distortion outside that field. (Clark 1992: 371)

If we want to translate this into the terminology of cognitive linguistics, we could

say that we are concerned here with semasiological and onomasiological salience
(cf. 2.6). The preference of common terms is connected with what term is most
frequently used for nuances in a particular area. This is probably an important
factor; as the two case studies above demonstrated there is a clear preference for a
few very frequent terms only. However, I would like to elaborate on Clarks
conceptual constraints; similarity and exhaustiveness.
How, then, should we understand cognitively the preference for certain terms
in the context of type modification? From the point of view of conceptual
organisation I think we can best conceive of it as a reference point phenomenon,
in which the prototype of the general category serves as the reference point (cf.
2.5.2). The type of reference point construction I suggest here is also intimately
linked with another construct in cognitive linguistics: the egocentric perspective or
vantage point of the conceptualiser. A vantage point can be characterised as the
position taken by the conceptualiser when construing a particular situation, context
or concept. It often entails a fixed position, usually that of the speaker in a
communicative situation, but, as emphasised by MacLaury (1997), it is possible to
shift vantages.
An important prerequisite for my theory is the presence of salient reference
points (foci) in the colour domain. As we could see in Chapter 1, the existence of
such foci is one of the basic claims of the B&K theory. Further evidence has been
collected by Rosch (1975c, Heider [= Rosch] 1972), who demonstrated that colour
category foci were salient cognitive reference points in terms of naming and
memorability. This evidence suggests that there are specific points in the colour
domain which are much more salient than the average colour nuances.
Accordingly, we might conceive of the colour domain in a three-dimensional form
as in Figure 4:6 below. The peaks here demonstrate the cognitive salience of
specific points in the colour domain.24 The checked pattern represents the 320chip25 Munsell colour array used in the WCS and the MCS (cf. Chapter 1). The
position of the reference points is that indicated by MacLaury (1997: 12) for a
speaker of American English (himself, actually). The height of the peaks is
intended to be suggestive (though exaggerated) and correlates roughly with the
position of the colour terms in the B&K hierarchy.

24

It should be acknowledged that reality is slightly more complex than this image suggests since the third
dimension of colour saturation is ignored in this model.
25
Achromatic colours are not included in the model as the normal mode of presentation would be
misleading.

124

Chapter 4

Figure 4:6.

Green
Blue
Purple
Pink

I
40

37

34

31

28

25

22

19

A
16

13

10

Red
Yellow
Brown
Orange

A tentative picture of the cognitive salience of the colour domain.

Only chromatic colours are included.

In my presentation below I discuss the process of reference point construction for

the purpose of type modification in relation to the three main classes distinguished
above: humans, natural objects and artefacts.
Let us first consider skin colour, and return to the question posed by Clark
(1992) of why white people are called white. In my opinion this puzzle cannot be
properly explained without taking into consideration the fact that black people are
called black. It would seem reasonable to assume that the classification of white
people is secondary to that of black people and that this classification reflects some
kind of reciprocal distinction or possibly a shift of vantage point. Type
modification entails paying attention to details that separate the subcategory from
the general category and in this case it is clear that the darkness of the skin is one of
the most conspicuous details. Moreover, the vantage point in the classification of
black people has been that of the relatively light-skinned European,26 and from that
perspective by far the most salient colour should have been black. Thus, one could
claim, BLACK would serve as the most useful reference point for classifying black
people. Conversely, then, since the categorisation of black presupposes a focus on
lightness it follows that WHITE was chosen for the opposite perspective whenever
needed. Incidentally, we can observe that white occurs much later than black in a
classificatory use in the domain of skin colour: in 1604 according to the OED
26

The classification in English and most other European languages may ultimately go back to the Latin term
niger, black. Cf. the term Negro, which occurs significantly later, according to the OED 1555, via Spanish.

Colour Terms as Adjectical Type Modifiers

125

(black, 890).27 This seems to confirm the above line of reasoning which is that we
first classify those objects that deviate from the norm and only later name and
subclassify the norm itself.
Let us consider the reference-point construction in more detail here. It entails
the use of one salient point in order to access another less salient entity of some
sort. With our metaphor of a mountainous landscape, we can clearly see how it
works. As the purpose in our context is to distinguish between types, the colour
terms referring to the peaks serve as reference points to indicate where we assume
we will be able to find the classified object. As our attention is directed towards
distinction, the present category i.e. that of the vantage is narrowed down and
the named category is expanded. MacLaury (1997: 93) emphasises the important
role of this type of cognitive activity, and mentions as a cognitive axiom the fact
that an individual attends simultaneously to similarity and distinctiveness and can
reciprocally shift the strength of emphasis.28 The colour term directs our attention
even to fairly vague nuances of the colour; apparently we organise our viewing so
that all other nuances serve as ground, and the named one as figure.
This becomes particularly clear in the categorisation of East Asians as yellow
people and North American Indians as redskins, which can be successfully
explained in the same manner. The colour terms refer to cognitively salient
reference points and emphasise difference. Thus, I would argue that the phrase
yellow people coined much later than the classification of white and black people,
the first quotation in the OED dates 1787 gives us no idea of the extension of
yellow in the 18th century, viewed on the Munsell array. Similarly, red (first used to
this end in 1587) in redskin gives us no idea of the actual colour of the peoples
skin. What we learn is that the group of people we are concerned with have a skin
colour that is distinctly redder than our own skin, and possibly redder than that of
Africans and Asians, but there is no clue as to how close to the focus of red we
should look. Using our previous metaphor, we can say that the peak of RED, cf.
Figure 4:6, gives us an idea of approximately where we could expect to find the
colour.
As pointed out previously in this chapter, the conventionalisation of a phrase
such as black people will lead to its acquisition of a much richer and more complex
structure than the surface of the phrase suggests. The phrase is defined, like any
other concept, against some background knowledge, domain, based on experience
and/or prejudice. We find good evidence of this in the decidedly pejorative
meaning of the phrases redskin and yellow people.29 Perhaps the best example of
the rich inventory of attributes is Toni Morrisons description of former president
Bill Clinton as the first black president of the USA. This suggests that additional
attributes have been centralised in this concept. In Cultural Studies and other
disciplines this richness of meanings is recognised, and concepts such as BLACK
and WHITE in their racial sense are frequently treated as social constructs (cf., for
instance, Ware 1992 and Stokes 2001). Thus, classifying use of a colour term may
27

The OED quotes 1591 as the date for the first instance of black man and 1695 for white man.
However, MacLaurys primary interest so far has been the colour domain itself, and he seems to have
been little interested in the kind of phenomenon discussed here.
29
Hawkins (2001) analyses racial classification in terms of colour in the context of political oppression.
28

126

Chapter 4

easily lead to a metonymical extension and figurative senses. An advanced attribute

structure, which the entrenchment/ conventionalisation of type modification can
give rise to, may lead to metonymical extension as is evidenced by a phrase such as
the yellow peril with reference to East Asians. Here, yellow alone stands for the
entity people, and the pejorative meanings are emphasised. The close relation
between type modification and figurative extensions is discussed in greater detail in
Chapter 7.
As an alternative classification from a different vantage point, we can mention
an example given by Siegel (1980). In Ngamambo,30 the term for RED is used to
classify Europeans, i.e. what we call white people. Siegel (1980: 169) observes
that
the Ngamambo speaker makes no claim that these individuals are women who are red
in any absolute sense, any more than an English speaker would claim that they are
women who are white in any absolute sense.

This classification indicates the importance of the vantage point for categorisation.
From the point of view of a dark-skinned person, the pinkish (and probably in an
African context slightly redder than in Europe) skin of a European could be called
either red or white; it would appear that these two salient reference points could
serve equally well for distinguishing this subcategory from the general category.
Apparently, the Ngamambo chose RED as the reference point. However, this is just
one piece of evidence and more is needed concerning other reference points and
vantages to strengthen this hypothesis.31
As an aside, we may observe that there used to be an alternative classification
of black people: as bloamon (literally blue man) as in (14) and (15) below
(possibly through the influence from Old Norse):
(14)
(15)

Of Ethiope he brohte a bleomen. (OED: 1205 Lay. 25380)

Blac as a bloamon. (OED: 1225 Ancr. R. 234)

This classification suggests a different reference point, but the vantage must
reasonably have been the same. One explanation of the use of blue could be that it
emphasised the lack of red in the colour of black people. Evidence of this use can
be found in Swedish where certain black people are occasionally referred to as
blue-black, Swedish blsvart.32 It would seem that this usage might reflect the
ambition of the user to exclude people with slightly lighter skin and more brownish
30

Ngamambo is also known as Moghamo-Menemo and is a language of Northern Cameroon. It was spoken
by approximately 87 000 speakers in 1982, (International Encyclopedia of Linguistics, 1982). According to
Siegel (1980: 158), Ngamambo only has three colour words: those denoting BLACK, WHITE and RED.
31
Another study which provides information of type modification from another vantage point, is Benders
(1983) report on Sudan Arabic terms for skin colour.
32
In a Swedish phrase such as the now extinct blmrker blue darkness the function of blue seems to be
that of emphasis. There is also an interesting habit in Swedish of using bl, blue for emphasis in contexts
where there is no colour connection at all, as in blneka (literally to deny bluely), blkra (to drive
bluely), bldre (blue fool = madman) and blljuga (to lie bluely = to lie excessively). This could
possibly be a parallel to the English phrase do something until one is blue in the face.

Colour Terms as Adjectical Type Modifiers

127

skin. This is mere speculation, however.33 The present-day form blae is defined as
blackish blue by the OED and Biggam (1997: 100) describes blwen as dark blue.
It is also thought that Swedish bl, blue, may have had a darker sense earlier.
However, it would seem that bloa only occurs in this construction, so it was
probably a marginal construction.
When we turn our attention to natural objects, the same kind of pattern can be
established. As we could see in the OED study, colour terms frequently serve the
purpose of type modification in the category of natural objects. Below, we will
consider one particular instance, namely that of oak, but I feel confident that the
general patterns can be translated to most other instances within this category. In
the case of OAK there are numerous subcategories, many of which have colour
names: blue oak (quercus douglasii), black oak (quercus velutina), golden oak,
gray oak (quercus grisea),34 (northern) red oak (quercus rubra), scarlet oak
(quercus coccinea) and white oak (quercus alba). This uniform naming habit,
however, disguises two possible areas of modification, as can be observed in Figure
4:7. The naming strategy of white oak may in fact refer to both leaves and bark
since the leaves have a whitish green colour underneath. However, most people
seem to think that it is the whitish grey bark that has given the tree its name.
Leaves:
Blue oak
Red oak
Scarlet oak
(White oak)
Bark of trunk:
Black oak
Grey oak
White oak
Figure 4:7.

Subtypes of oak. Naming strategies involving colour.

Following Langackers terminology, we can analyse this picture as a type of active

zone phenomenon (cf. 2.5.2), which in turn is a special kind of reference point
construction. Thus, it is communicatively more relevant to refer to the whole tree
rather than to the trunk and leaves as they are natural parts of a tree. Although there
are names of the type *blue-leaf oak or *black-trunk oak, they are very infrequent.
We could argue, however, that the picture is even more complex than the
above indication would suggest. Red oak and scarlet oak are called so because of
the outstanding coloration of their autumn leaves. This characteristic feature is, of
33

The colour semantics of Middle English appears to have been rather complicated, since there was another
term for black, swart, which could be used for classifying people (cf. the OED). This second term may
have affected the salience of the term black.
34
The OED gives some additional possibilities for the name grey oak: any of several American oaks, esp.
quercus coccinea or q. borealis; the first being identical with scarlet oak.

128

Chapter 4

course, only present during a relatively short period of time (although periodical).
In fact any reference to the colour of the leaves may represent a third type of
reference point construction a temporal one. Even though some trees may have
no leaves for a large part of the year, the leaves still serve as a useful reference
point. It is certainly the case that the Gestalt of a tree is one in which leaves are
present. For example, children in the north of Sweden invariably draw trees with
leaves even though all trees except firs and spruces lack leaves for approximately
eight months of the year. At any rate it would seem that salient features can serve
as naming attributes even if they are not of a permanent nature.
Let us return to the issue of colour naming again and consider the name blue
oak. If we take a closer look at a leaf from a blue oak we can establish that the leaf
is bluish green with a tinge of grey.35 However, my own impression is that few
people would call this colour blue in the context of an evening dress. In fact many
people would probably say that it is more green than blue. (In the context of an
evening dress, the term teal-green may possibly be used.) Accordingly, we could
say that the naming principle here constitutes a parallel to that of skin colour.
Because the normal colour of oak leaves is green we can identify the vantage point
as GREEN, and when naming the intended nuance, BLUE serves as a good reference
point, as attention is given to distinctiveness.36 Consider Figure 4:8 below.

R
Blue (reference point)

Green (vantage point)

The bluish greyish green colour of the

blue oak (target)

Figure 4:8.

35

A tentative picture of the reference point phenomenon blue oak.

The PLANTS National Database describes the colour of the foliage as grey-green in the section called
plant characteristics. However, in the section called plant guide, the colour is described as blue-green. The
colour pictures that are available on the database show leaves with a bluish-greyish-green colour in which
the green element clearly dominates. The PLANTS National Database is accessible at
http://plants.usda.gov.
36
This might also be analysed as a kind of figure-ground phenomenon. Consider how a blue oak stands out
in a collection of other green trees. However, that would not explain the use of the term blue.

Colour Terms as Adjectical Type Modifiers

129

The same kind of analysis would explain the other tree names. For instance, the
bark of the trunk in the cases of white oak and black oak is likely to be lighter and
darker than that of the most common oak. The colours of the leaves of red oak and
scarlet oak are characterised by a conspicuous reddish element in the autumn,
especially so in the case of the scarlet oak.37
Summarising the impression we get from viewing type modification in terms
of skin colour and tree names, it seems reasonable to conclude that the salience of a
particular term in relation to a particular nuance depends very much on the vantage
point of the conceptualiser. A particular nuance may for instance be called bluish
green, green, blue, teal or even turquoise, but which term is most useful (or salient)
can only be determined in a wider context. It would appear that in cases of type
modification, language users tend to prefer terms that have very salient prototypes.
This is clearly demonstrated by the characteristics of the OED material and a few
seconds of reflection can provide the reader with numerous examples: red and
white wine; red, black and yellow beans and so on. We are probably concerned
with a mixture of semasiological and onomasiological salience, i.e. the salience of
the term and of the cognitive reference point.
Before we conclude this section let us consider a kind of artefactual type
modification which demonstrates yet another type of reference point construction.
Consider the frequently occurring classification of coffee as white and black.
Although black coffee can exhibit a colour which is close to black (in the best case
scenario of strong coffee), white coffee is nowhere near a whitish colour. Bennett
(1988: 32-33) discusses this phenomenon and suggests that describing coffee
solely in terms of black and white is a simplification by means of the elimination of
colour; this is followed by radicalisation (viewable also as a further simplification)
of the corresponding tone into black or white. Bennett uses the term describe
but arguably the division of coffees into black and white makes no claim about the
colour but refers to the inclusion or not of milk thus it is classifying and not
descriptive. It would seem that with a reference point model we can explain the
naming much more economically. As mentioned above what is referred to in white
coffee is the milk that is poured into the coffee, and, thus, since the attribute [milk]
is at the very heart of this subcategory, it serves as an excellent reference point for
this kind of type modification.
To sum up, we have seen in this section that the cognitive abilities of
reference point reasoning and vantages combined with salient foci in the colour
domain can help us explain instances where the use of colour terms goes beyond
their normal area of designation, which can occasionally be found in type
modification.

37
There is an alternative motivation for the term red in red oak. It could also refer to the reddish interior
wood of the tree.

130

Chapter 4

4.7 Wider implications another look at previous investigations

It would seem that the conclusions that have been reached on the basis of the
material investigated here may be fruitfully applied to patterns established in
previous investigations. For instance, the distinction between type and token
modification (or description and classification) appears to have been ignored so far
in colour term semantics. In my background chapter, I mentioned Forbes (1979)
seminal study of brun and marron in French. In this article, Forbes observes that
brun and marron qualify different types of objects. For instance, brun is very much
preferred when it comes to hair or skin, whereas marron is preferred in the context
of food and clothes. It could be speculated that this pattern might very well be due
to the fact that brun (the long established term) has predominantly retained
classificatory functions, whereas marron (the more recent term) is primarily used
for descriptive purposes. Lyons (1999: 48) appears to have made the same
observation, although not giving it technical representation, since he claims that
[w]hen brun is used of the colour of someones hair, skin or eyes, it means dark
rather than specifically brown. This is true in the sense, as stated above, that the
colour itself is not of primary interest, it only has a classificatory function
delimiting a subtype. Thus, brun is used to contrast a type of eyes with blue and
green eyes, so it is possible to say that it only means dark. Further evidence
supporting this theory can be found in Schfers (1999: 98) claim that the relation
between brun and marron is similar to that between blond and jaune. It would seem
that the former term is mostly used for type modification, cf. also the use of
blond(e) in English. Furthermore, it would seem likely that Conklins (1964) and
Lyons (1999) observations concerning Hanuno and Greek colour terms can be
better understood if we take type modification into account.
Another pattern that would be very interesting to explore in the wake of this
study is the result of the step by step mapping used by MacLaury. As described in
MacLaury (1997: 77), having stopped once, the subject was encouraged to continue
mapping (putting grains of rice on colour arrays) until (s)he insisted that the term
applied to no other colours. It would seem that the continued mapping after the first
step may reflect extensions of a colour term due to type modification. It might also
be fruitful to compare the extension that MacLaury achieved in this way with the
Langackerian concept of dominion (cf. Section 2.5.2 in the context of reference
point construction). This construct was not treated in my discussion here but may
very well have explanatory power.

Colour Terms as Adjectical Type Modifiers

131

4.8 Summary
The aim of this chapter was to describe and analyse an apparent phenomenon in
colour semantics: the fact that some colour terms occasionally refer to areas which
may be better described by other terms. In the course of this chapter, it was
demonstrated that it is important to take into account the different functions that the
colour word may have. It was established that the above phenomenon appears
particularly in the context of what in cognitive linguistics has been called type
modification.
Two corpora were investigated, and although they were different in character
and the methods were distinct, the overall patterns proved to be conspicuously
similar. Type modification is primarily explored by a small number of colour terms
(the Primary BCTs). Very often the nominal heads represent basic level categories,
although in some natural domain their taxonomic position has been changed to the
subordinate level (e.g. red oak).
In the analytical part of the chapter, it was argued that, from a semasiological
perspective, the detected phenomenon could best be analysed in terms of reference
point constructions. It was demonstrated that subcategorisation entails taking a
vantage point from which attention is paid to distinctness. This includes a process
which would appear to favour salient reference points, such as WHITE and BLACK.
The favoured position of these concepts may explain the high frequency of the
corresponding terms found in Chapter 3. In conclusion, the restriction of type
modification to a few colour terms only is probably determined by both
semasiological and onomasiological salience. As pointed out by Clark (1992), there
is a preference for salient terms, and these terms refer to salient concepts in the
colour domain which can serve as reference points for categorisation.

Chapter 5

Elaborate Colour Terms:

Definition, Formation and Usage

5.1 Introduction
Whereas the previous chapter was mainly concerned with the most frequent terms
in the corpus, the basic colour terms, the present chapter deals with the other
extreme Elaborate Colour Terms. It was demonstrated in Chapter 3 that BCTs
comprise a clear majority 92.2 % of the total number of tokens in the BoE, even
when the highly frequent terms gold, golden and silver were included among the
ECTs. As we could see in Chapter 4, few ECTs are used for classificatory
purposes, which may explain their rareness. Furthermore, it seems reasonable to
assume that identifying use is proportionally rather small since the precision of an
ECT would entail the presence of a fairly similar colour nuance. Grices maxim of
Quantity1 predicts that the speaker is likely to use a more general term if there is no
special need for great precision, which in the case of the identifying function would
require a similar nuance. Moreover, even in a situation with fairly similar nuances,
I find it more probable that the speaker will opt for modified general terms rather
than an ECT, as in, for example, purplish-red, dark red. This might be an
additional factor contributing to the low frequency of these terms. If this
assumption is correct, then it would seem that ECTs are most often used for
descriptive purposes, ascribing a particular colour to an entity.
In Chapter 1, I pointed out that most research dealing with ECTs has been
gender oriented, testing whether women use ECTs more often than men. Although
these studies focussed on gender differences, some of the results have a wider
significance. One such result, relevant to the present study, was found by
Nowaczyk (1982). He was able to demonstrate that, on the whole, men and women
are not very accurate at matching colour nuances and colour terms. Although the
women in his study were significantly better than the men, the overall figures are
not impressive: women 42% correct, men 35%. I find these figures of great interest
because they seem to indicate that there is a certain looseness involved in the
usage of these specific terms. Looseness of specific terms is almost an oxymoron
and indicates a certain paradoxical aspect of the usage of these terms.
This looseness serves as the overarching theme of both this and the following
chapter. The first part of this chapter considers how these specific terms are defined
in first a technical colour dictionary, and then common language dictionaries. The
second part of the chapter deals with the emergence and formation of ECTs. If
language users generally have a fairly poor idea of the exact designation of these
terms, then their history may be important for a full understanding of their function.
As a bridge between this and the final part, sense relations pertaining to ECTs are
briefly discussed. The chapter ends with a study of the frequency of ECTs in a
1

The maxim of Quantity is expressed as follows (Grice 1989: 26):

1. Make your contribution as informative as is required (for the current purpose of the exchange)
2. Do not make your contribution more informative than is required.
133

134

Chapter 5

number of nominal categories. Since colour terms are acquired in the context of
things, the domain of the noun may be significant for the full characterisation of the
term. The study investigates a number of colour terms with similar areas of
designation. Thus, the issue of free variation is tested empirically, the aim being to
demonstrate that terms may have different salience in different nominal categories.
The next chapter takes ECT usage one step further and considers contexts where
attributes other than the nuance seem more important.

5.2 Defining ECTs

5.2.1 The standardisation approach
From a technical point of view, it is of course important that colour nuances be
properly and individually defined. It would create serious problems if
manufacturers notions of colours were as vague as those of ordinary people. They
may well be in fact, but there are means of standardisation to help people in their
production of coloured goods. Somewhat confusingly, there are several systems for
colour naming, for instance the ISCC-NBS system, the OSA system and the NCS
system, to name only a few.2 Since the differences between these systems are
irrelevant to the present study, I will not try to account for them. On the contrary,
below, I restrict myself to one system through a brief presentation of some aspects
of one standardisation book: Kelly and Judds (1976) Color: Universal Language
and Dictionary of Names. It represents the ISCC-NBS system and is published by
the US Department of Commerce, according it a privileged position among the
standards. The main purpose of this presentation is to provide a starting-point for a
discussion of how we can keep track of ECTs.
Kelly and Judds dictionary is interesting because they use two different
notational systems to realise their ambition to standardise colour reference. For the
more technical definitions, they standardise colour names with the help of formal
notations that refer the reader directly to colour charts. However, they also use a
system of controlled everyday language, or, in other words, a transparent
metalanguage.3 Thus, the three dimensions hue, lightness and saturation are
codified in everyday vocabulary Kelly and Judd (1976: A14-A16) introduce a
system of four classes for the coding of these dimensions: generic hue names,
intermediate hue names, lightness names and saturation names. The controlled
language terms are meticulously defined in that there are charts that clearly identify
their position. The first group consists of 13 names: the terms defined as BCTs plus
olive and yellow green. This is striking, not least because olive is frequently defined
as (deep) yellowish green. In Kelly and Judds description, olive has low
saturation and lightness and seems to include elements of brown. The second group
consists of sixteen hue names according to a system of derivative forms, e.g.
yellowish green and greenish yellow. The one exception in this group is violet.
Lightness and saturation are closely linked dimensions, so the naming system
2

There is a thorough presentation of most colour order systems at http://www.colorsystem.com/. A

historical background is also included.
3
This is not unique for the ISCC-NBS, however, it is in fact true of many systems.

Elaborate Colour Terms: Definition, Formation and Usage

135

combines these two. A nuance of great lightness and low saturation is termed very
pale, whereas a nuance of little lightness and high saturation is called very deep.
Extremely high saturation is termed vivid.
According to Kelly and Judd, on these principles it is possible to define one
nuance of maroon as deep reddish brown. One nuance of magenta, on the other
hand, is deep purplish red. Since Kelly and Judd standardise colour terms, they
quote various dictionaries dealing with a wide range of fields such as biology, soil
colours, plastics etc.,4 and as there appears to be some variation, the qualification
one nuance of is needed. Apparently terms like maroon and magenta can
designate slightly different nuances in different fields: the book defines eleven
nuances of maroon and seven of magenta. In total, 7,500 colour names are defined.
In the preface to the part called The Color Names Dictionary, Kelly and Judd
(1976: III) suggest that
[t]he dictionary will serve not only as a record of the meanings of the 7,500 individual
color names listed but it will also enable anyone to translate from one color
vocabulary to another. As an example, what is the meaning of Griseo-Viridis? This
dictionary shows that Griseo-Viridis (biology) = Serpentine (fashion) = Mint Green
(mass market), or in ordinary language, a light green.

In the book the terms crimson, plum, maroon, magenta, fuchsia and carmine
include either of the two neighbouring nuances dark purplish red and deep
purplish red. In view of this, we may ask how we distinguish between these terms.
Furthermore, the fact that the designation of a term might differ slightly depending
on the domain of the noun would seem to add to the problem. There is evidence of
this difficulty elsewhere; as was suggested by Nowazcyks (1982) study, it seems
that speakers find it quite hard to match terms and nuances. We can find additional
evidence of the difficulty of describing the meaning of ECTs when we look at less
technical definitions those found in language dictionaries.
5.2.2 Dictionary definitions
Definitions of colour terms in dictionaries largely resemble those discussed above,
with the obvious difference that there are no technical definitions referring to a
particular conception of the colour domain. There is also one other conspicuous
difference in that the definitions of colour terms in dictionaries can be divided into
two general types. Apart from definitions like those we could find in Kelly and
Judd, we have descriptions of the most basic terms. These colour words are not
defined with reference to other colour terms but with reference to natural objects.
As an example we can consider red. In all the six dictionaries I have studied, red
was defined in relation to blood, and other mostly natural objects; two examples
can serve as illustration.
First, we have Websters definition:
A color whose hue resembles that of fresh blood or the ruby or is that of the longwave extreme of the visible spectrum.
4

Kelly and Judd (1976:11-13) provide a detailed account of the different colour name systems that they deal
with.

136

Chapter 5

Second, the OED, which defines red in a very similar way:

The colour which appears at the lower or least refracted end of the visible spectrum,
and is familiar in nature as that of blood, fire, various flowers [] and ripe fruits.

Note that most of these examples refer to the focal colour of the colour category.
This type of definition is used for the following colour terms: black, white, red,
green, yellow, blue and brown.5 Thus, this group consists of the Primary BCTs and
brown. It is also interesting to note that grey is frequently defined in the same way,
most often in relation to ashes. This is somewhat surprising since grey obviously
represents a colour between black and white, and could be defined in that way.
Incidentally, this way of defining colour terms is similar to Wierzbickas (1990,
1992, 1996) suggestion regarding the conceptual structure of colour terms (cf.
Section 1.3). In Wierzbickas (1996: 330) view, the meaning of the most basic
colour terms can be seen as quotations incorporating fundamental and visually
salient features of human environment: the sky, the sun, vegetation, fire, the sea,
the naked earth, the earth covered with snow. Without taking any standpoint on
the overall issue, we can agree with Wierzbicka that these features at least serve as
good reference points, and the dictionary usage confirms this.6
Apart from the terms mentioned above, the other terms are generally defined
in relation to the basic colour terms, in a system similar to that designed by Kelly
and Judd (1976). However, although dictionaries use the same type of controlled
language for this descriptive purpose, there is considerable variation in the
definitions. Consider the selection given in Table 5:1 below. (A full list of
definitions can be found in Appendix 4.) Aside from the first two (lemon and lime),
the terms represent three categories to be explored in greater detail in the case study
in Section 5.5 below. Table 5:1 reflects some conspicuous differences in the
definitions of the colour terms. Take for example magenta: Whereas CIDE states
that it is a dark purplish colour, Longman defines it as a bright pink colour.
Quite a few of the dictionaries define lavender, lilac and mauve in exactly the same
way, a pale purple colour. Intriguingly, the one dictionary that does distinguish
between these, Websters, defines maroon and magenta in the same way, dark
reddish-purple, two colour terms which are defined differently in most other
dictionaries. Another striking detail is that the OED tries to maintain a difference
between the terms aqua and aquamarine. It is quite clear that it is difficult to define
the exact nuance of an ECT. Moreover, it should be noted that no dictionary tries to
contrast one ECT with another; they are always defined in relation to BCTs.
If we turn to another area of the colour spectrum, we can see that Websters
defines lemon yellow as a brilliant greenish yellow suggesting both a green
element and high saturation,7 whereas all the other dictionaries stress the single
5

This is rather similar to what Nord (1997) found in language dictionaries for German, French and Italian.
Some researchers have criticised Wierzbickas claims for not being testable. A critical examination of
Wierzbickas theories can be found in Kemmerer (1999).
7
The unorthodox definition of lemon in Websters could be due to the lack of, or confusion with, the colour
term lime in the dictionary, but this is mere speculation. Somewhat confusingly, yellow is partly defined as
the colour of ripe lemons. See Appendix 4. It should be noted that there may be dialect differences present
here, and that American English is only represented by Websters Dictionary.
6

Elaborate Colour Terms: Definition, Formation and Usage

137

yellow element of lemon, and four of them low saturation; a pale yellow colour.
In the case of lime (green) we can observe that opinions differ as regards saturation,
as both bright and pale are used. More confusing is the fact that CIDE, which uses
the term lime green, defines the colour as a light bright greenish yellow colour,
thus suggesting a stronger yellow element than the green. This would appear to be
counter-intuitive given the explicit mentioning of green in the name lime green.
Table 5:1.

Dictionary definitions of some elaborate colour terms

DICTIONARIES AND DEFINITIONS

ECT
WEB

OED

LONG

OALD

CIDE

CCELD

Lemon

A brilliant
greenish
yellow

Pale yellow

A pale yellow
colour

A pale yellow colour

A pale yellow
colour

Bright yellow in
colour

Lime

[Lime green]
A light
yellowish
green colour

A pale green colour

A light bright
greenish
yellow colour

Pale yellowishgreen in colour

Crimson

Deep purplish
red

Of a deep red
somewhat inclining
towards purple

Having a deep
purplish red
colour

A deep red

Magenta

A deep
purplish red

A bright pink
colour

a colour between
purple and red

Maroon

A dark red

A particular kind of
brownish-crimson or
claret colour

Plum

A dark
reddish
purple

A shade of purple

A very dark
red-brown
colour
A dark red
colour, like a
plum

Puce

Lavender

A pale purple

Lilac

A moderate
purple

Mauve

A moderate
purple, violet,
or lilac color

Aqua
(marine)

A pale blue to
light greenish
blue

Turquoise

A light
greenish blue

Of a flea colour;
purple-brown or
brownish purple
A very pale blue
with a trace of red
Of the colour of lilac
blossom [= pale
pinkish violet]
The colour of
mauve. [ = A
bright but delicate
purple]

A dark brownish red

colour
A dark reddishpurple colour

A strong,
slightly
purplish, deep
red colour
A dark
purplish red
colour
A dark
reddish purple
colour
A darkish
purplish red
colour
A dark
purplish red
colour
A pale purple
colour
A pale
pinkishpurple colour

A dark, purplishred colour

A dark, reddishpurple colour
Dark reddishpurple in colour
Dark-red or
purple in colour

Dark brownish
purple

(of) A brownish
purple colour

A dark purple
colour

A pale purple
colour

A pale purple colour

A pale purple
colour

A pale purple colour

A pale purple
colour

A pale purple colour

A pale purple
colour

A pale purple
colour

Aqua: A light
greenish blue colour
Aquamarine: bluishgreen

A greenish
blue colour

Defined in terms of
the colour of the
stone: a pale
greenish-blue
precious stone

Bluish green

A greenish blue
colour

The colour of the

turquoise [of a skyblue to apple-green
colour]

A greenishblue colour

A greenish-blue
colour

Bluish green
in colour

A bright blue
colour that is
fairly light and
often greenish

A pale bluish
purple colour
Pale pinkishpurple in colour

Not defined as a colour term: Used for the name of a brilliant crimson aniline dye [...] Interestingly
enough, this definition is altered in the new draft edition of the OED available online at
http://dictionary.oed.com/ . The new definition suggests [a] purple-pink aniline dye.

138

Chapter 5

In conclusion, we can say that the difficulty does not reside in an approximate
definition of the terms; most dictionaries seem to agree about the BCTs between
which any specific ECT is positioned. Rather, the difficulty is to approximate
which BCT nuance is the more dominant, and, in particular, to describe the degree
of lightness and saturation of the nuance the term refers to. Here we find
considerable differences. In cases such as lemon and lime, we might have expected
more or less identical definitions in the dictionaries given the prevalence of the
fruits from which the colour terms are derived. If we can find this kind of disparity
in language dictionaries, can we then assume that the average language user will be
in full control of the differences between, say, magenta and maroon, or lavender,
lilac and mauve?
Let me finally mention an attempt to move away from the imprecise language
definitions of colour terms. This attempt can be found in Longman, in which a
number of terms are presented and illustrated in the context of a painting. This
provides the reader with a possibility of viewing, and thus identifying the reference
points of the terms directly in the colour domain. However, the success of this
approach can be questioned. First, it would seem that some colours may have come
out poorly in the printing process; for instance, the alleged turquoise colour appears
to be far from what is normally conceived of as turquoise. Second, the terms are
presented in the context of a painting scene, and there are three distinct areas in
which colour terms are expressed together with their alleged meaning: the painting
itself, a palette and a box containing tubes. Unfortunately, in this mode of
presentation, adjacent nuances (technically speaking) are not presented in
proximity, and, as a result, it may be difficult to appreciate the difference between,
e.g., purple, violet and mauve.
If it was the belief of the lexicographer that these colour terms contrast with
one another, then I think a simpler colour chart presentation would have been much
more helpful. In addition, I would suggest that it would have been more helpful if
the colour terms had been presented in the nominal domain where they are most
frequently used. For example, rather than defining beige in the domain of facial
colour (which is the case in the picture), where it is very rarely used, it could have
been defined in the domain of clothing.
5.2.3 Summary
In this section we have considered the way that ECTs are defined in a technical
dictionary and in typical language dictionaries. We could see that in the latter there
is considerable variation between the definitions and an examination of Appendix 4
will provide the reader with further evidence of this. In fact, it would appear that
the only ECTs that are consistently defined are those which designate nuances
within the area of a single BCT, as in the case of emerald, for example.
One reason for the difficulty could be that ECTs, like BCTs, cover areas of
the colour spectrum and do not refer to one particular nuance alone. Therefore we
get a slight overlap between different ECTs. On the basis of the definitions in Kelly
and Judd (1976), it is possible to draw a tentative picture of the relation between a
number of closely related colour terms. Figure 5:1 below represents such an

Elaborate Colour Terms: Definition, Formation and Usage

139

attempt. The greyish areas at the centre of the figure represent deep purplish red
and dark purplish red, two nuances which most of these terms seem to refer to.
Plum describes an area which is darker and slightly less saturated than the one
designated by magenta. Both these terms include nuances which are more purple.
Crimson, on the other hand, is redder and has high saturation. Maroon is darker and
less saturated and includes elements of brown. In contrast to language dictionaries,
where puce is given some brownish or purplish element much like maroon, Kelly
and Judd define the term as dark red.
Thus, as Figure 5:1 illustrates there is a slight overlap between the various
colour terms. The problem, if we are to make a guess on the basis of the language
dictionary definitions, appears to be that ECTs lack salient reference points (foci) in
the colour domain. There is an overlap between BCTs, but there is little
controversy about where the focus of the category lies, and all the dictionaries
provide more or less the same definition, describing the focus.

CRIMSON
MAGENTA
deep purplish red

dark purplish red

PUCE

PLUM
MAROON

Figure 5:1.

The semantic relation between some ECTs

In the case of ECTs this reference point is missing and the dictionaries define
different parts of the area designated by the term. The next section considers the

140

Chapter 5

emergence of ECTs in the light of these facts. Why, we may ask, are there so many
terms if their mutual relation is difficult to establish?

5.3 The Emergence of ECTs.

5.3.1 The motivation
The evolution of Elaborate Colour Terms has been studied by Casson (1994,
1997).9 In his articles, he convincingly demonstrates that English went through a
change from being a language that focussed on lightness aspects to becoming a
hue-focussed language.10 Although this was a long process, Casson identifies this
shift as taking place in the Middle English period. His conclusion with regard to
this shift deserves to be quoted in its entirety:
Culture members, responding to increases in societal complexity and diversity,
restructure their systems of color categorization by differentiating new concepts and
innovating new vocabulary []. The color shift from brightness to hue in the
evolution of English basic color terms can be seen as a response to an increasingly
complex color world in the Middle English period (1150-1500). The development of
secondary color terms, beginning in the late medieval period (1350-1500) can also be
attributed to this increasingly diverse array of culturally significant colors. The
response of culture members to these changes was a cognitive refocusing. (Casson
1997: 237-238)

Although Cassons overall conclusion is most persuasive, I do not fully agree with
his statement that the Middle English period saw a growing complexity from the
point of view of colour. To my mind, this is true only to a certain extent, namely in
relation to one particular domain that of textiles. As far as other areas of
experience are concerned, it is reasonable to claim that the colour nuances that
were named in the Middle English period (and later) were already present in the
everyday life of the common people in the Old English period. These nuances
could be found in the colours of flowers, of sunsets, of berries, etc. However, as
Casson so insightfully points out, they lacked cultural significance. In an illiterate
and predominantly agricultural society like that of the Old English period (and,
indeed, the Middle English period and even later), there was no, or very little,
functional need for elaborate descriptions of flowers and berries; at least not to the
extent that new terms were coined. Furthermore, there was always the possibility of
nonce formations. As suggested above, classifying and identifying functions tend to
opt for the most general contrasting term. However, with descriptive use, it is the
colour itself which is of particular importance, and this was certainly the case in the
context of clothes.
Casson sees the growing cultural importance of these nuances as intimately
linked with the increasing use of exclusive dyed products. This seems to be a
correct conclusion. Although native dyes such as woad and madder already existed,
it appears that the increasing trade with the Mediterranean region and the
9

Kerttula (forthcoming) examines in great detail the origin of 100 English colour terms.
See also Biggam (1997: 40-78) for an exhaustive summary of research in Old English colour terminology.

10

Elaborate Colour Terms: Definition, Formation and Usage

141

development of the woollen industry stimulated the use of other dyes. CarusWilson (1952: 371) claims that
the northern dyers vastly extended the range and variety of their colours with more
exotic products from Asia as well as from the Mediterranean. Kermes, now
commonly called grain (granum), from Asia Minor, Spain and Portugal was now used
in the North as it had been in the South in Roman days; with it was dyed the costly
scarlet worn by kings.

We can argue that textiles coloured with dyestuff from abroad had a high cultural
significance. Not only because the colours were different from those obtained from
native dyes, but in particular because they signalled exclusivity and, thus, wealth. It
therefore became important to designate these colours properly. Consider
Chaucers classic description of the wife of Bath in the general prologue to The
Canterbury Tales (l. 453-457):11
Hir coverchiefs ful fyne weren of ground;
I dorste swere they weyeden ten pound
That on a Sonday weren upon hir heed.
Hir hosen weren of fyn scarlet reed,
Ful streite yteyd, and shoes ful moyste and newe. [emphasis added]

The mention of scarlet here is hardly accidental, and I would argue that its major
significance lies outside the colour domain; it goes well with the characterisation of
the wife that she should wear exclusively dyed stockings. Thus, the difference
between scarlet and red was significant in the domain of clothes, but it was
probably highly insignificant in the domain of flowers. There is a great deal of
evidence of this state of affairs in the Middle English period. If we look more
closely at the oldest terms in Cassons (1994) list, that is, the ECTs established
before 1500, we find the following twelve terms: ash, auburn, azure, blond,
crimson, gold, ochre, russet, scarlet, silver, vermilion and violet. As Casson (1994:
15) points out, all but three (ash, auburn and blond) were relevant to the domain of
textiles. Furthermore, the colour red appears to have been particularly significant,
and especially the kermes dye since no less than three colour terms derive their
meaning directly or indirectly from this dye: crimson, scarlet and vermilion.
The tendency towards emphasis of ECTs in the domain of clothing appears to
be valid for modern English usage as well. A study of a random sample of the
material in the BoE reveals that the main use of ECTs in the BoE is confined to
artefactual domains such as textiles, clothes, interior decoration, cosmetics and
vehicles. Of natural domains, it is only in the domain of plants that we find
extensive use of ECTs, and then it is in the somewhat artificial context of
horticulture (see also section 5.5). It appears common to these domains that they
involve some kind of compositional thinking with regard to colour. Furthermore,
they constitute domains of experience which seem to be salient in particular types
of discourse where they are granted high cultural significance. Consider, for
11

A detailed study of Chaucers colour term usage can be found in Biggam (1993). Incidentally, she does
not include this particular instance of scarlet. This could be due to her confinement to adjectival use, or
possibly to an interpretation of scarlet as having the meaning of scarlet dye.

142

Chapter 5

instance, the great number of specialised magazines in these fields. As might be

expected, it is in the brmags subcorpus that we find the most frequent use of ECTs
and it is also in this subcorpus that there are more than five mentions of all the
ECTs studied here, cf. Section 3.6 (p 92-94). Moreover, these magazines are
characterised by a multitude of photographs and pictures, often accompanied by
comments concerning the composition of colours; the colours are not mentioned for
the purpose of classifying a type or identifying a particular instance. Often they are
mere descriptions to help the reader create a picture. Recall Magnusson and
Perssons (1986) claim that the aim of description is to draw attention to the
property, not the entity.
However, it would also appear that in reference to some entities the colour
terms are used to evoke other types of associations outside the colour domain.
Bergh (1997) demonstrates that the terminology used for car colours tends to be
more suggestive than precise; other areas which might present a similar pattern are
fashion, interior decoration and cosmetics. As an example of the associations
sought after in the creation of a colour term in cosmetics, consider the following
scene in Margaret Atwoods The Robber Bride (1994: 433). Roz, one of the
protagonists, invents a new colour term:
Shes thought of another river name too, another colour: Athabasca. A sort of bronzed
pink. Frostbite crossed with exposure. How you get in the North without sunblock.

The next section examines more closely the formation of ECTs from a cognitive
linguistics point of view.
5.3.2 The process of creation
It seems that almost all ECTs,12 and, incidentally BCTs too (cf. Shields 1979),
originate as names of entities. The evidence is overwhelming. Casson (1994: 13)
demonstrates that the entity senses predate the colour senses in all the cases of his
study. Furthermore, when a colour term is a loanword, it would seem that the
corresponding term in the supplying language originally referred to an entity. The
cognitive process that we are concerned with here is that of metonymy. In Section
2.5.2, I gave a general description of this process in a cognitive framework,
identifying several different types. The type of mapping we are concerned with
here can more precisely be described as + CATEGORY FOR SALIENT ATTRIBUTE +, or
to use Cassons (1994: 14) wording: entity stands for entitys color. On the whole
this appears to be a very common type of metonymy.
What makes the type of metonymy involved in the formation of colour terms
rather special is that it also leads to a change in word class, allowing the concrete
surface of the entity to be abstracted and moved to other entities without evoking
the whole of the entity. It can be argued, however, that in its reference to colour
space, the colour term represents an uncountable noun, cf. Kay (1999), and
Langacker (1991b). Twardzisz (1997) treats zero derivation of adjectives from
12

A conspicuous exception would be magenta, which is the name of an Italian town where there was a battle
in 1859, shortly before the dye was discovered. However, the formative process essentially parallels that of
the normal type.

Elaborate Colour Terms: Definition, Formation and Usage

143

nouns, and gives a brief account of colour terms. It would seem to make sense both
cognitively and communicatively that that which instantiates the best example of a
colour also represents that colour. However, having said this, we must make due
allowance for the fact that what constitutes the best example is culturally coded.

Concrete

Abstract

6
2

Figure 5:2.

Primary
domain:
original: GEMS
new: COLOUR

Concept:
TURQUOISE

10

Attribute list:
1.[x]
2.[blue-green]
3.[y]
4.
5.
6.
7.
8.
9.

Metonymic mapping. + CATEGORY FOR SALIENT ATTRIBUTE +,

TURQUOISE

Using the figure presented in Chapter 2, we can illustrate the metonymic mapping
+CATEGORY FOR SALIENT ATTRIBUTE+ as in Figure 5:2 above. As the figure
demonstrates, this metonymy can be viewed as a subtype of the more general partwhole metonymy. Figure 5:2 indicates that the focus of attention is shifted from the
primary domain of the default predication (the entity) to one, usually central,
attribute; hence, a new primary domain is evoked (COLOUR) at the expense of the
previous one, which becomes a part of the attribute structure. This is possible
through the reciprocity mentioned in Section 2.4.1. Furthermore, the metonymical
process leading to a new prediction will necessarily result in some change of the
attribute structure. Some attributes may be lost whereas others may be retained or
even centralised. Recall (Chapter 2) the complex network that we can assume
forms a conceptual structure. By means of an example, we may note that the form
lemon in its colour sense may vaguely evoke associations to attributes of the entity.
This, however, is not the same as to say that the lemon nuance evokes the same
associations. Furthermore, given the name of present-day vogue colour terms such
as orchid and cappuccino, it is tempting to suggest that these other attributes are in
fact almost as important as the actual nuance, the primary domain of COLOUR.13
13

Leech (1981:16-17) discusses a similar phenomenon in terms of reflected meaning.

144

Chapter 5

Berghs (1997) study of colour terms for cars provides ample evidence of the
tendency for colour terms to be used for their peripheral associations rather than for
precision, examples such as smaragdsvart, emerald black and diamantbl,
diamond blue demonstrate that associations are more important than the
information value.
Casson (1994) presents a very strong case for this kind of metonymy being
the source of all ECTs. However, he takes the establishment of this historical
development one step further and claims that we are concerned here with a
unidirectional process. A weak interpretation of this claim is that Casson only
suggests that all colour terms ultimately go back to an entity meaning, and the
evidence certainly points in that direction. But, if we are to take the term
unidirectional at face value, then Casson appears to maintain that the opposite
process is not possible, and this is unquestionably incorrect. There are numerous
examples of cases where a colour stands for an entity. Suffice it to indicate, at this
point, that the classifying terms of races, whites and blacks, represent such a
mapping. This will be dealt with in more detail in Chapter 7.
Let us finally consider an intermediate stage of development. It seems that the
mapping +WHOLE THING FOR A PART OF THE THING+ is, as it were, lexicalised in the
paraphrastic expression x-coloured (where x is an entity). In this phrase the whole
entity is evoked but at the same time the focus is placed on one attribute the
surface colour. Although this type of construction might be claimed to be a
borderline case as a colour term, I have included it in my study. It might be argued,
however, that the colour terms that demonstrate a significant proportion of such
constructions are poorly entrenched as colour terms. The terms that have the
highest proportions of this type of construction are rose (114/632),14 cream
(100/1384), and rust (42/174). In particular rose appears to have a weak identity as
a colour term, despite its relatively high frequency; there is also the phrase rosetinted, and in both cases the meaning is figurative rather than literal. It was also
shown in Chapter 3 that the term rosy is much more frequent than rose in its colour
sense.
To sum up, in this section the process of formation of ECTs has been
identified as one of metonymy. Typically, the name of an entity that constitutes a
salient example of the colour is metonymically extended to designate the colour
itself. In this process a number of attributes linked to the entity may be retained as
attributes to the colour term, which is also true of the previous primary domain. It
was argued that, in many cases, these other attributes may be just as important as
the reference to the colour domain. It was also shown that the evolution of ECTs is
a fairly recent development correlating with an increasing cultural importance of
colour nuances in clothes.

14
The figures represent the frequency of the phrase x-coloured and the total frequency of the term referring
to colour.

Elaborate Colour Terms: Definition, Formation and Usage

145

5.4 The sense relations of ECTs

So far in this chapter, we have addressed formal definitions of colour terms and the
formation of the terms. The insecurity in the dictionary definitions and the
formation of ECTs from names of entities seem to raise questions concerning the
semantics of ECTs. The purpose of this section is to look more closely at sense
relations obtaining within the field of ECTs. The sense relations of special interest
are introduced in Figure 5:3 below.
As an example of the various relations, let us consider the term fuchsia. The
colour term derives from a genus of ornamental shrubs (family Onagrace)
(OED). Thus, it could be claimed that there is a polysemous relation between the
entity sense and the colour sense. The colour term fuchsia is defined quite
differently in the dictionaries (cf. Appendix 4) so that it is actually possible to
suggest three different superordinates: red, purple and pink. Finally, depending on
our understanding of fuchsia, we might consider plum, mauve, rose or magenta to
be near synonyms of the term.
As pointed out in Chapter 2, these various sense relations are dependent on
the perspective we choose to apply to our study. Synonymy and hyponymy are the
main concern from the point of view of the onomasiological perspective. Polysemy,
on the other hand, reflects a semasiological perspective.
Entity

Superordinate2

potential polysemy

potential hyponymy

Co-hyponym1

Superordinate1

potential hyponymy

ECT

Co-hyponym2

Co-hyponym3

potential synonymy
or near synonymy

Figure 5:3.

Semantic relations relevant to the analysis of ECTs.

5.4.1 Polysemous relations

Let us first consider the relation between the entity sense and the colour sense.
Following Cassons convincing argumentation, we have concluded that almost all
ECTs are ultimately derived from names of entities through a process of
metonymy. It was also suggested that the construction x-coloured represents an
intermediate stage in this process. Once a metonymic mapping of the above type

146

Chapter 5

has been exploited or encountered by a language user a number of times, it will

have reached a certain degree of entrenchment, thus achieving unit status as a
colour term. At that point, it can no longer be claimed that an active metonymic
mapping is taking place, instead there will be a complex network, or radial
structure.15 Another way of describing the state of affairs is to say that there is a
polysemous sense relation between the entity sense and the colour sense. As was
stressed in Chapter 2 following Tuggy (1993), we can best conceive of the
dimensions of homonymy, polysemy and vagueness as a continuum. With respect
to ECTs we can observe that different ECTs appear to have different relations to
their senses of objects and that this may possibly be analysed in terms of this
continuum.16
One way of estimating the distance between two polysemous senses in the
case of colour terms could be to relate it to the dimensions of transparency and
opaqueness. This would suggest that terms which are regarded as opaque do not
share any attributes with the original entity, whereas transparent terms do so to a
varying degree. Typically ECTs which are loanwords have very little contact with
the original entity sense. However, the issue of transparency and opaqueness is not
as straightforward as one would wish. First, we have to acknowledge that different
speakers may have different opinions, depending on their knowledge of the world:
what some people may find to be transparent colour terms, others may find opaque.
Hence, we would expect to find some fuzziness when we consider the language
community as a whole. In Casson (1994), this issue is addressed and studied, and
he demonstrates that there is considerable agreement among speakers as far as
these parameters are concerned. No less than 73 of his 92 terms (i.e. 79%) were
given an almost unanimous vote as either transparent or opaque (only one
dissenting voice among the twenty informants). On the other hand, it still meant
that 19 terms showed some degree of fuzziness with respect to these dimensions.
Unfortunately, Casson does not tell us which terms. I myself was surprised to find
amber and lavender among the opaque terms, so we may speculate that these two
were among the 19 terms exhibiting fuzziness.17
A second aspect of this matter is how we should understand transparency. In
his investigation, Casson asked his informants two questions in order to determine
whether a colour term was transparent or not:
whether the term had a noncolour meaning or not, and if so,
what the meaning was.
To me, true transparency would also imply that ones knowledge of the entity
would be helpful for the identification of the colour nuance. Accordingly, I think
that Cassons questions only verify a limited degree of transparency. Consider the
15

This is of course an individual process that will vary among speakers, depending on age, interests etc.
What one speaker perceives as a novel term, others may find to be an established term.
16
It is, however, possible to make finer distinctions. Magnusson and Persson (1986) argue that there is a
qualitative distinction between homophonous transcategorial derivatives and polysemous items belonging to
the same category. The transcategorial relation (i.e. involving shift of word class) is called synsemy.
17
The opaqueness of these terms is likely to be culturally conditioned. Lavender is native to Europe, and the
largest and most significant source of amber is the shores of the Baltic Sea.

Elaborate Colour Terms: Definition, Formation and Usage

147

terms navy and orchid, both classified as transparent. No one could seriously
suggest that navy is transparent in a strict sense; indeed, there is only an indirect
link between the (abstract) entity sense and the colour sense. Similarly, orchids are
familiar to most people as flowers, but it would seem that few have actually seen
many orchids. Furthermore, since we are concerned here with a group of flowers, it
is more or less impossible to predict the nuance of the colour term. In linguistic
terms we could say that this type of polysemy is slightly different, since the keymediating attribute is missing. The formative process of navy, for instance, can be
viewed as a complex series of metonymies, whereby the name for an abstract entity
has come to serve as the name for the typical colour of the uniforms worn by the
sailors representing this entity. This aspect of marginal transparency appears to be
quite common. The presence of colour terms also designating relatively common
objects that show variation in colour testifies to this: examples of such terms in my
material are plum, olive, and lilac, and among Cassons terms we can mention
chocolate. Presumably, the colour term usually refers to the most prototypical
colour of the entity, but this is still one aspect which can be said to represent only a
limited degree of transparency.
At the other end of the scale, we can find terms like silver and lemon. They
designate entities with which people frequently interact, almost on a daily basis.
Thus, they are likely to be truly transparent. Moreover, the link between the entity
and colour sense is evident since there is very little or no variation in the colour of
lemons as we usually meet them.18
Having established these different aspects of transparency and opaqueness,
we should now be able to draw some linguistic conclusions. First, if a term is
conceived of as transparent, even in a relaxed sense, it is likely that the colour term
will share a number of attributes with the entity from which it was derived. No
doubt this is why we find terms such as orchid, cappuccino and cinnamon in
cosmetics and fashion the words are associated with exotic tastes, smells and
rareness. This is of course a modern trend, but in line with that described earlier for
Middle English, where exclusive dyes and gems tended to create new colour terms.
Second, it is tempting to see the transparent relation as a function of the
degree of entrenchment/conventionalisation that the varieties exhibit. If the object
sense is very entrenched, as we can assume is the case with silver, then we expect a
stronger influence from its attribute structure. It seems possible that the more
deeply entrenched term lends attributes to the formally identical colour term. Thus,
we would have a more complex attribute structure. For a less entrenched term such
as fawn, this would not be the case, since the attribute structure of the entity sense
might be poorly developed. Finally, in cases where the colour term was borrowed,
like magenta and mauve, there might be no influence at all from the object sense.
On the basis of this argumentation it is possible to make a rough and general
categorisation into four types of polysemous relations based on the different
degrees of entrenchment/conventionalisation of the colour term and the entity.
Figure 5:4 tries to capture the different types on the basis of the frequencies in the

18

This makes the definition of lemon in Websters (cf. p 137) even more surprising.

148

Chapter 5

BoE.19 A bold box illustrates a well-entrenched term (n > 500),20 a broken box
represents senses of low entrenchment. It may be somewhat surprising to see that
the majority of ECTs are derived from objects whose terms are not very salient.
The results can also be compared with Cassons (1994) classification of transparent
(T) and opaque (O) terms.
Thing

ECT

Olive (T)
Lime (T)
Peach (T)
Lavender (O)
Lemon (T)
Rust (T)
Charcoal (T)
Plum (T)
Mustard (T)

Figure 5:4.

Thing

ECT

Tan (O)
Amber (O)
Lilac (T)
Emerald (T)
Indigo (O)
Magenta (O)
Aquamarine (T)
Azure (O)
Fawn (T)
Tangerine (T)
Vermilion (O)
Fuchsia (O)
Puce (O)
Carmine
Teal (O)
Chartreuse (O)
Ultramarine (O)

Thing

ECT

Golden
Gold (T)
Silver (T)
Cream (T)
Navy (T)
Rose (T)
Violet (T)

Thing

ECT

Scarlet (O)
Crimson (O)
Beige (O)
Turquoise (T)
Maroon (O)
Mauve (O)

Four categories of polysemous relations based on entrenchment/

conventionalisation. The ECTs are ordered according to
frequency.

As can be observed, there is generally a good agreement, in that colour sense

derived from poorly entrenched entity senses are classified as opaque.21
However, on closer inspection the picture is even more complex. Another
complicating factor is the complex structure of some terms, e.g. lemon. In these
cases there are a number of senses which can be derived from the object. It could
be argued that these other senses may also affect the full semantic structure of the
colour sense. The question is what kind of relations obtain between the derived
senses, as in Figure 5:5 below.

19

For a detailed study of the relative basicness of English colour terms, see Kerttula (forthcoming). The
issue of prototypicality and frequency is very tricky. Alan Cruse (pc) pointed out to me that see in the sense
understand, know, is more frequent in computer corpora than see in its notionally prototypical sense of
using ones eyes.
20
This figure is arbitrarily chosen and may be justifiably challenged. However, for the current aim of
illustrating the difference between various term it serves a purpose. The character of the BoE is also likely
to affect whether an entity is classified as salient or not.
21
Carmine and golden were not tested for transparency and opaqueness by Casson (1994).

Elaborate Colour Terms: Definition, Formation and Usage

149

Entity
LEMON

smell

Figure 5:5.

taste

colour

Derived senses of the object term lemon.

Added to these directly derived senses, there are also metaphorical extensions from
the object or from one of the derived senses. Examples of such additional senses
are lemon and peach in the senses of bad and good. It seems unlikely that a car
manufacturer will ever call a particular yellow nuance lemon since the collocation
of lemon and car is likely to create undesirable meanings.
Polysemy may occasionally cause ambiguity. This would appear to be a
marginal problem in so far as ECTs are concerned. Nevertheless, ambiguity does
occur sporadically and it would seem that this is particularly frequent when the
entity domain of the colour term is that of metal or gemstones. Possibly, this is
because of the valence relation (cf. 4.3) that the modified noun makes possible.
Consequently, silver hair and amber light are not ambiguous, whereas silver buckle
and amber earrings potentially are. In the latter cases, the character of the noun
makes two interpretations possible (material and colour), the issue being whether it
is the surface, i.e., the colour, or the content of the entity that is being described.
For the compilation of my frequency lists this was occasionally a problem, but in
practice it would have little significance.
To sum up, we can say that polysemy and its related concept transparency are
important aspects of the full description of the meaning of an ECT. It is also argued
here that the polysemous relation and the issue of transparency can be described in
terms of degree, one important factor being the entrenchment of the two senses, the
object and the colour. Furthermore, it is suggested that associations linked to the
nominal concept from which the colour sense was derived may be of significance
for the usage pattern of the colour term.
5.4.2 Hyponymy
As far as hyponymy is concerned, it was pointed out in Chapter 2 that, in the colour
domain, this is a semantic relation of a complex nature. There are numerous colour
terms which designate areas located between the best examples of some BCTs.
Examples of such terms are turquoise (between blue and green), maroon (red,
brown and purple), amber (yellow and brown), crimson (purple and red), and there
are many more. Is it correct then to say that these ECTs are hyponyms of both these
colour terms? Although this is a tempting solution in some cases, for instance
turquoise, it is more problematic in others, for instance cream. Prima facie, it

150

Chapter 5

would seem that a corpus study should be able to offer us some guidance in this
respect. Quite frequently ECTs form what could be termed polysynthetic structures
together with BCTs. Thus, we find turquoise blue, lime green, and lemon yellow.
However, on closer inspection it seems that this type of construction is by far the
most frequent as regards ECTs that clearly designate nuances within the field of
reference of only one BCT, as in charcoal grey and azure blue. When it comes to a
term like maroon, for instance, the corpus is not very helpful at all; there is one
instance of maroon-red, one of maroon-purple and one of maroon-brown. In my
opinion, we have to acknowledge that certain domains of experience do not readily
lend themselves to the sometimes blunt tools represented by semantic formalism.
The colour domain appears to be one such.
5.4.3 Synonymy
From the dictionary definitions we could see that there are several terms which
appear to denote more or less the same area in the colour domain. Should not these
terms be considered synonymous? Kay (1999: 81) points out that ECTs have
succeeded one another in certain areas and that this occurred in North American
English in the twentieth century; indeed it is an on-going process:
The colors that I learned to call tan, chartreuse and turquoise are called by members
of my childrens generation beige, lime and aqua, respectively. A representative
corpus that spanned the last few decades would show the synonym pairs (tan, beige),
(chartreuse, lime) and (turquoise, aqua). In my speech each of these pairs now
occurs, as far as I can tell, in free variation, and there are doubtless other North
American speakers in the same boat.

We find two interesting statements here. First, Kay points to the possibility that we
may be faced with the phenomenon of change in apparent time (and indeed, real
time), among the terms which denote more or less the same area in the colour
domain. Second, he suggests that, in this shift, some speakers will use the terms
virtually synonymously (in free variation).
From the point of view of semantics assumed here, it does not matter whether
two entities may be used in free variation their conceptual structures are likely to
be different, and it would therefore be misleading to claim that true synonymy
exists. It has been pointed out by many commentators that it is impossible to find
perfect synonyms, since most words differ in some respect (cf. 2.6.2). Some
scholars have therefore argued that alleged synonyms are better viewed as cohyponyms (cf. Persson 1990). However, the fact that native speakers claim to use
terms in free variation deserves closer attention. A case study below explores three
areas of the colour domain in which we find colour terms defined in more or less
the same way in the colour dictionaries.

5.5 A case study of ECT synonymy domain preference and salience

The aim of this case study is to investigate the salience of effects of some ECTs in
relation to different nominal categories. The study focuses on terms which belong
to the same area in the colour domain and which could be regarded as near

Elaborate Colour Terms: Definition, Formation and Usage

151

synonyms. It is particularly alleged synonymy and free variation that are

considered from the point of view of the language corpus. As described by
Langacker (1999), the continual entrenchment of a unit is closely related to usage
and encounter; it can be viewed as the abstraction (or schematisation) of relevant
information from the context (cf. Section 2.2). This suggests that the character of
the noun may be of considerable importance in the conceptual structure of the ECT
since colour terms are mentioned in the context of some object and since the ECTs,
generally speaking, have low frequencies and thus low degrees of entrenchment. It
may be that some ECTs are used particularly often in some nominal domains, in the
same way as blond(e) and auburn are more or less restricted to the domain of HAIR.
Previously in this chapter I have pointed to domains such as FASHION and
HORTICULTURE where ECTs appear to be particularly frequent.
What led me to realise the importance of the nominal category was an answer
I received from an informant when I played around with different ideas of
transparency. I asked my informants if they knew what nuance almond was, and in
what context, if ever, they would use it. This particular American male informant
gave the following answer:
I think of two things, actually. The first color nuance is a very pale brown color close
to the color of a shelled almond or a peanut. The second is a mustard-yellow closer to
deep yellow than lemon yellow. [] I know I would use this term to talk about the
color of major appliances (refrigerators, ovens) because thats what my above
connotations are based on. The almond color was mustardy in the 70s and is the
lighter color now.

Here we can see two strategies for the designation of the term: to look for the
colour of the object, or to look for a salient context (object) where the term
occurred. Apparently the term could give access to a particular nominal context in
which this informant frequently encountered the term. Thus, the designation was
accessed through a mental space or possibly domain matrix evoked by the
described entity rather than directly in the colour domain.
The aim of my case study is to examine the relation between general nominal
categories and a number of ECTs. As pointed out above, the low frequency with
which an ECT is used makes the nominal category likely to be of importance.
BCTs, on the other hand, probably also show variation over nominal categories, but
their high entrenchment precludes nominal categories from having any deep impact
on their meaning.
Three areas of the colour domain were chosen for this case study. All three
areas are designated by more than one term, thus suggesting intricate semantic
sense relations. The terms that refer to these areas are
crimson, magenta, maroon, plum and puce (dark/deep purplish/reddish/
brownish red/purple)
lavender, lilac and mauve (pale purple)
aquamarine and turquoise (greenish blue)

152

Chapter 5

The areas differ to some extent in size and the degree of referential equivalence
between the terms in each group varies. However, as can be seen in the definitions
quoted in Table 5:1 above, there is considerable variation among the dictionaries
and we can guess that the same insecurity can be found in the speech community as
a whole. In common with Kays (1999: 81:fn 12) claim, it may be assumed that
some people will state adamantly that there are differences among lavender, mauve
and lilac, one being more bluish and one more pinkish (cf. the definitions of
CCELD in Table 5:1), whereas for others they are synonymous in terms of colour
designation (cf. most other dictionaries in the same table). Still others will only be
aware of one or two of them as colour terms. What I want to explore is the kind of
difference we can discern in the language community, as represented by the BoE
corpus.
The case study of ECT usage as described below is based on my material
from the BoE. The occurrences were categorised according to the nominal field to
which the colour term ascribed its nuance. This approach was inspired by Forbes
(1979) investigation of the French terms brun and marron. By asking informants to
list things that they would call brun and marron respectively, she was able to
demonstrate that the use of these terms partly appears to fall into complementary
categories. My hypothesis is that we may find a similar pattern when it comes to all
ECTs with potential referential identity.
However, in contrast to Forbes, I have preferred to place the entities in
general categories, which I call nominal fields. These can be equated with the
notion of domain in its vague sense, but given the technical use of domain that has
been employed here, I have found it practical to use a different term for this
categorisation. However, in want of a synonym, I have occasionally used domain.
The categories are notional and somewhat ill-defined, but are apparent enough for
the present purpose. The labels of these fields should be self-explanatory, although
some may need further explanation. First, as is demonstrated in numerous
publications, any categorisation will be complex in the sense that there are bound to
be borderline cases and fuzziness. Since I have opted for few fields these are
necessarily of a general nature and in borderline cases I have chosen to use an
inclusive approach rather than categorise something as other. As examples I can
mention that shoes are part of the clothing field, tiles are part of background
decoration. Second, the rationale behind distinguishing between furniture and
background decorations such as wallpaper, curtains, carpets and painted walls is
based on the fact that pieces of furniture tend to serve as figures, whereas the
entities belonging to the second category are characteristically ground elements.
Finally, it should be noted that each super-field natural objects, human and
artefacts contains a hotchpotch category named other, containing objects which
could not be classified in the other nominal fields.

Elaborate Colour Terms: Definition, Formation and Usage

Table 5:2.

153

The distribution of a selection of ECTs in different nominal fields.

Other

Face

Other part

Clothing

Cosmetics

Road
vehicles

Background
decorations

Furniture

Other

crimson

31

294

51

23

61

65

114

25

58

140

878

magenta

77

13

66

194

maroon

70

249

62

39

11

69

518

plum

21

48

20

24

134

puce

34

15

12

76

SUM

51

465

55

32

103

77

439

55

131

16

291

1800

lavender

109

10

64

24

40

275

lilac

159

82

10

28

37

340

mauve

271

10

18

74

35

11

60

501

SUM

13

539

24

30

17

220

23

87

18

137

1116

Alt.
fields
aqua
(marine)

Nat. Art.
water water

SUM

ECT

Celestial
phenomena

ARTEFACTS

Plants

HUMANS

Animal
kingdom

NATURAL THINGS

Human
mostly eyes

19

12

44

43

42

186

turquoise 16

15

122

23

14

179

12

13

80

21

139

640

16

141

18

29

19

223

15

16

123

28

181

826

SUM

17

Table 5:2 above gives the numbers for each category. On the basis of these figures
we can analyse the material from two different points of view; we can estimate the
relative semasiological and the onomasiological salience of the terms.22 In Chapter
1, I mentioned that MacLaurys improved fieldwork methods allowed him to
consider both these perspectives within the confines of the colour domain.
However, what I will do here is not only to go beyond the colour domain, but to
ignore it completely.
To my knowledge, the most systematic work employing these perspectives in
a study of lexical semantics is that of Geeraerts et al. (1994). One of the great
achievements of this meticulously conducted corpus investigation of words for
garments in Dutch is that the authors try to quantify semantic phenomena. In my
analysis below, I draw heavily on the approach taken in that work, with a few
necessary alterations.

22

The onomasiological and semasiological perspectives are discussed and defined in 2.6.

154

Chapter 5

5.5.1 The semasiological perspective

As argued in Chapter 2, a semasiological perspective asks of a given form what its
semantic range is. Thus polysemy, homonymy and vagueness are sense relations
which can be addressed from this perspective. However, as Geeraerts et al. (1994)
demonstrate, prototype phenomena, such as fuzzy boundaries and salience effects,
can also be successfully studied; technically, these phenomena could be included in
the semantic notion of vagueness. If we apply this reasoning to the present study,
we may ask in what nominal field a particular ECT occurs most often. Should there
be any salience effects in relation to the preference of nominal field, we might
suggest that this gives us additional information about the meaning of the term. We
could say that a certain nominal field is particularly salient in the usage of the term
a salient attribute in the same way as one might propose that [hair] is a salient
attribute of BLOND(E).
As was discussed above, it seems reasonable to suppose that we acquire colour
names in the context of nouns, rather than in the context of the colour domain itself.
Thus it may be that some nominal fields are particularly salient in a term. We can
calculate the semasiological salience of a nominal field as the ratio of the frequency
of a term within a specific field to the overall frequency of the term. The null
hypothesis would suggest equal distribution across the fields. Table 5:3 below
presents the results of the calculation in percentages. The focus is the horizontal
axis of the table; the highlighted figures indicate which field is most salient in each
case. Note that whereas some colour terms show a high degree of preference for
one field, other terms are more evenly distributed. The distribution was tested for
chi-square goodness-of-fit, and showed significant deviation from the null
hypothesis (p < 0.05) for all terms.
In the brownish/purplish red field, we find an interesting pattern of differing
distributions. Crimson and magenta are most often used in the field of PLANTS,
whereas CLOTHING is the most salient nominal field in plum and maroon. Thus, the
figures suggest that an English speaker would be likely to use crimson in the
context of flowers, and plum in the context of clothing if asked to illustrate the
usage of the colour terms. Given the character of the BoE (cf. 0.3.5), however,
encounter might be a better term than use. The corpus more accurately reflects the
likelihood of being exposed to a term than of using it.
The colour term puce has a conspicuous distribution. In contrast to the other
terms in the purplish/brownish red area, however, puce is most frequently used in
the field of FACE. This is quite surprising. Why would a specific colour term be
used particularly often to designate facial colour? Dictionaries suggest that puce is
a dark brownish purple colour.

Elaborate Colour Terms: Definition, Formation and Usage

155

Table 5:3 Salience effects of nominal fields.

Other part

Clothing

Cosmetics

Road
vehicles

Background
decorations

Furniture

Other

SUM (%)

5.8

2.6

6.9

7.4

13.0

2.8

0.8

6.6

1.0

15.9

100

magenta 4.6 39.7

1.0

0.5

1.5

3.1

6.7

2.1

2.6

4.1

0.0

34.0

100

maroon 1.2 13.5

0.0

1.4

0.2

0.4

48.1

0.4

12.0

7.5

2.1

13.3

100

plum

3.7 15.7

1.5

0.7

3.0

0.7

35.8

14.9

0.0

17.9

3.0

3.0

100

puce

0.0

3.9

0.0

0.0

44.7 3.9

19.7

5.3

2.6

2.6

1.3

15.8

100

lavender 1.5 39.6

3.6

2.9

2.5

23.3

1.8

0.7

8.7

0.7

14.5

100

lilac

0.9 46.8

1.2

1.2

1.2

24.1

2.9

1.2

8.2

1.5

10.9

100

mauve

1.2 54.1

2.0

3.6

1.2

14.8

1.6

0.4

7.0

2.2

12.0

100

Plants

crimson 3.5 33.5

ECT

Animal
kingdom

Face/Eyes

ARTEFACTS

Other

HUMANS

Celestial
phenomena

NATURAL THINGS

Alt.
fields

Nat. Art.
water water

Eyes

aqua
0.5
(marine)

0.5

10.2

6.5

3.2

2.7

23.7

1.6

1.6

23.1

3.8

22.6

100

turquoise 2.5

2.3

19.1

0.9

3.6

2.2

28.0

1.9

2.0

12.5

3.3

21.7

100

The use of puce as a facial colour correlates very strongly with anger and rage as in
examples (1) and (2) below. However, I find it hard to believe that a person even at
the peak of anger would have an actual puce nuance to his/her face this seems to
be a clear case of hyperbole. The pattern can be interpreted as an indication that
puce could be on its way out of the colour domain and into designation of an
abstract quality great anger instead. Example (2) can be interpreted in this way.
Incidentally, the same kind of development can be found for the term livid. This
kind of usage is addressed in greater detail in Chapter 6.
(1)

She was in the drawing room when Billy burst in, his face puce and his eyes
wild. (BoE: brbooks)

(2)

Laudrup lobbed long to David Robertson who tapped past Thomson from five
yards. Jimmy Nicholl, puce no doubt by half-time, brought on both substitutes
for the restart. (BoE: indy)

156

Chapter 5

In the purplish/brownish red area most terms show a clear preference for one
particular nominal field. The one exception is magenta, which, apart from the field
of PLANTS, also has a high frequency of occurrences in the hotchpotch field of
ARTEFACTUAL OTHER. On closer inspection, it turns out that in this latter category,
we find two particularly frequent applications of the term. First there are quite a
number of references to printing and ink. This is due to the fact that magenta is one
of the subtractive primary colours, thus anyone familiar with colour photography
will be well aware of the colour of magenta. Incidentally one might feel hesitant
about whether this nuance and that of magenta flowers are really identical.
Amazingly, the second most frequent use of magenta in this field refers to a famous
stamp (a misprint). As we can see in the example below, it is unique and therefore
it can be referred to by its colour which functions almost as a proper name.
(3)

In the world of rare stamps, that is done by myth and legend. Some stamps
simply acquire an aura, a mystique. Some never do. Demand is what makes the
black and magenta so much more valuable than the others. (BoE: brbooks)

Finally it should be pointed out that a great number of the maroon instances in the
field of clothing are linked to sports activities. It appears that a number of sports
clubs have maroon as part of their official uniform. In particular this is true of
various teams representing Queensland (Australia), and if we look at the statistics
(Appendix II) we can see that maroon is indeed over-represented in the subcorpus
of Australian English (oznews). In all the other subcorpora but one (spoken British
English brspok) crimson is more frequent than maroon. In oznews, maroon is
four times as frequent as crimson (maroon 115 instances vs. crimson 28). Although
it can be shown to be linked to sports, it may be assumed that the term maroon
occupies a different position in the Australian speech community than in the
British. The fact that the colour term is frequently used should make it more salient
to people. In England, the same colour is usually termed claret in the context of
football; Aston Villas colours are for instance called claret and blue.
If we consider the other two areas we can see that aquamarine has no clear
field of preference, although it is clearly a term which is mostly used together with
artefactual things. Turquoise, on the other hand, shows a clear preference for the
CLOTHING domain. Interestingly enough, all the terms in the pale purple field are
most frequently used in the field of PLANTS. Although there are slight differences, it
is striking how similar these terms are as far as field salience is concerned;
lavender and lilac especially show great similarity, whereas mauve appears to be
slightly more anchored in the PLANT domain.
In sum, we can say that the ECTs studied here most frequently occur in two
nominal fields, those of vegetation and clothing. This suggests that, at least in some
cases, we are likely to associate the meaning of a colour term not only with a
particular nuance but also with a nominal field, or domain. Moreover, it could be
shown that terms in the purplish/brownish red area of the colour domain differed
with respect to the preference of nominal field. This might be an indication that the
speech community prefers to distinguish similar terms in this way. However, no
such pattern could be found in the pale purple area on the contrary, these terms
demonstrated remarkable similarity with respect to nominal field preference.

Elaborate Colour Terms: Definition, Formation and Usage

157

5.5.2 The onomasiological perspective

Let us now turn to the complementary perspective and consider onomasiological
salience. In Chapter 2, it was argued that onomasiological variation is closely
related to basic level effects and sense relations such as hyponymy and synonymy.
It has been argued that the most salient level in a hyponymy structure can be
equated with the basic level (Rosch et al. 1976, Lakoff 1987). Here we can also
include Langackers notion of entrenchment. The more deeply entrenched a unit
is, the higher its onomasiological salience, cf. Geeraerts et al. (1994).23 In my study
here, I would like to see whether there are any differences in salience among the
terms when we consider a number of different nominal fields. Since we have
stipulated full referential equivalence, the terms are treated as if they could be
exchanged without any change in nuance. Thus, to put it simply, the study asks
which ECT is most likely to be used if we want to describe a particular nuance of
an object belonging to a particular nominal field. To exemplify; if we saw a
purplish red car and wanted to describe it which term would we choose? Crimson,
magenta, maroon, plum or puce? Since I am only interested in the relation among
the ECTs of a particular area I have not taken into consideration the BCTs red and
purple, nor possible modifications of these, such as purplish-red.
A great achievement in Geeraerts et al.s work is that they give their notion of
onomasiological salience an operational definition thus making calculations
possible. They (1994: 138) define it as
the ratio between the frequency with which the members of a lexical category are
named with an item that is a unique name for that category, and the total frequency
with which the category occurs in the corpus.

Here we will modify this operational definition so that we can make it fit the
character of the present investigation. The onomasiological salience of a term was
calculated as the ratio of the frequency of a term within a specific field to the
overall frequency of the nominal field. The null hypothesis would suggest equal
distribution among the terms. The figures were tested for chi-square goodness-offit, and the marked figures within a nominal field indicate that the null hypothesis
could not be rejected (p > 0.05). The figure for the most salient term in each field is
highlighted in bold characters.
A closer look at Table 5:4 reveals some interesting details. First, as might
have been expected, crimson, given its high overall frequency, is quite salient in
most nominal fields. However, one conspicuous aspect of the deduced pattern is the
comparatively lower salience of crimson in artefactual nominal fields. In the fields
of CLOTHING, ROAD VEHICLES and FURNITURE maroon is distinctly more salient. It
might be that crimson is felt to have a fairly formal and old-fashioned tone (cf. the
age of the term), and is thus unsuitable in the discourse of fashion. Moreover, as
was pointed out above, the salience of maroon in the domain of CLOTHES can be
partly explained by its frequent occurrence in sports. The strong position of
23

Geeraerts et al. (1994: 146) maintain that the basic level hypothesis suggests that co-hyponyms should
have entrenchment values of the same magnitude, but this is disconfirmed [] here. However, as pointed
out by Ungerer (1996), this seems to be the authors interpretation, which cannot be found elsewhere. The
central idea of my study here is that the terms will differ in entrenchment in different nominal fields.

158

Chapter 5

crimson in the field of other human references can be attributed to its frequent use
in connection with blood.

Road
vehicles

Background
decorations

Furniture

Other

59.2 84.4 26.0

45.5

9.2

44.3

36.0

48.1

48.8

magenta 17.6 16.6

3.6

3.1

2.9

7.8

3.0

7.3

6.6

6.1

0.0

22.7

10.7

maroon 11.8 15.1

0.0

21.9

1.0

2.6

56.7

3.6

81.6

29.8

44.4

23.7

28.8

Face

Clothing

71.9

Other part

92.7

Plants

crimson 60.8 63.2

ECT

Animal
kingdom

Cosmetics

ARTEFACTS

Other

HUMANS

Celestial
phenomena

NATURAL THINGS

Proportion of all
instances (%)

Table 5:4 Salience effects of ECTs in some nominal fields an onomasiological

perspective.24

plum

9.8

4.5

3.6

3.1

3.9

1.3

10.9

36.4

0.0

18.3

16.0

1.4

7.4

puce

0.0

0.6

0.0

0.0

33.0

3.9

3.4

7.3

2.6

1.5

4.0

4.1

4.2

*** ***

***

***

***

***

***

***

***

***

***

***

Humans,
mostly eyes
lavender 30.8 20.2

41.7

26.7

41.2

29.1

21.7

25.0

27.6

11.1

29.2

24.6

23.1 29.5

16.7

13.3

23.5

37.3

43.5

50.0

32.2

27.8

27.0

30.5

mauve 46.2 50.3

41.7

60.0

35.3

33.6

34.8

25.0

40.2

61.1

43.8

44.9

**

lilac

***

Alt.
fields

Nat. Art.
water water

aqua
5.9
(marine)

13.5

66.7

20.7

26.3

19.7

20.0

18.8

35.0

25.0

23.2

18.4

turquoise 94.1 93.8 86.5

33.3

79.3

73.7

80.3

80.0

81.2

65.0

75.0

76.8

81.6

**

***

**

***

***

6.3

Humans,
mostly eyes

**

***

A second point worthy of attention is the status of puce in the nominal field of
FACE. We saw earlier that puce is used most frequently in this field and only rarely
24

*** = p < 0.001, ** = p < 0.01, * = p < 0.05

Elaborate Colour Terms: Definition, Formation and Usage

159

otherwise. However, this does not mean that puce is the most salient red-purplish
colour in the field of face the figures above demonstrate that crimson is much
more salient. This underscores the importance of describing ECTs from both
perspectives. However, the choice between puce and crimson is not entirely
irrelevant, as is further described in Chapter 6.
If we look at the green-blue area we can see that turquoise is much more
salient than aquamarine in all nominal fields. It is only in the field of ARTIFICIAL
WATER, i.e. in the context of pools and bathtubs, that the latter term is close in
salience to the former. This suggests that whenever the blue-green area is to be
designated by an ECT, turquoise is the more likely choice. It may even be
suggested that aqua(marine) is a hyponym of turquoise.
Finally we have the pale purple area. Mauve is the most salient term in the
fields of VEGETATION and FURNITURE, and two other categories. However, the
most conspicuous fact is that the null hypothesis could not be rejected in most
fields. Thus, it seems that these three terms are equally salient in most fields. Does
this mean that they are synonymous or used in free variation? Not necessarily, but
other methods have to be employed to prove this.
To sum up, this case study demonstrates that we can find additional features
which are important for the characterisation of the colour domain by studying the
use of the terms in the context of nominal domains. It was shown that there may not
only be differences between colour terms in relation to their own anchor points in
nominal domains, but also in terms of salience in a specific nominal domain. On
the whole, onomasiological salience in various nominal domains coincided with the
overall salience of a term. The distribution of crimson and maroon was a notable
exception.
Other terms that show this preference for certain nominal fields are beige,
fawn, navy, teal and cream, which all occur most often in the field of CLOTHING.
Lemon, on the other hand, is very frequently used in reference to the colour of
flowers. These would seem to be important facts that might well be taken into
account by lexicographers in trying to describe the meaning of ECTs.

5.6 Summary
The aim of the present chapter was to give a general semantic characterisation of
non-basic colour terms, here called Elaborate Colour Terms. Two types of
definitions were studied. It was shown that the technical dictionary used a
controlled language together with formal notations. Ordinary language dictionaries
use a similar kind of controlled language in their entries. However, a comparison
between several dictionaries revealed that there is considerable variation
concerning exactly where the lexicographers place ECTs in the colour spectrum. It
was argued that this is probably due to the lack of salient reference points for these
colour categories. This seems to verify what has been found in other studies: a
certain degree of insecurity as to the reference of ECTs. It would be interesting to
use the methodology of MacLaury (1997) to investigate further the relation
between various ECTs.

160

Chapter 5

ECTs are typically derived from names of entities and to the extent that this
connection is apparent to the language users, this polysemous state of affairs might
possibly affect the conceptual structure, giving access to a richer collection of
attributes. It was demonstrated that a number of sense relations could be discerned
and that the concept of transparency can be more complex than may appear at first
sight.
The second part of the chapter discussed various sense relations, and a case
study showed that terms which appear to be close to synonymous may differ with
respect to semasiological and onomasiological salience. In conclusion, conceptual
complexities and differences in salience undermine claims that there may be
synonymy between ECTs.

Chapter 6

In Search of Extra Meaning:

ECTs and Markedness

6.1 Introduction
So far in this dissertation, I have principally been concerned with reference within
the colour domain. Chapter 4 showed that a classificatory function of a term may
allow it to refer outside its normal area, and that the choice of term may be
dependent on the vantage of the conceptualiser. Chapter 5 dealt with ECTs and it
was argued that the domain of the noun may be of importance when we choose a
colour term to describe an object.
In the present chapter, I consider more closely some extra meaning effects
that appear to be created under particular circumstances, especially in the use of
ECTs. The chapter can be seen as a transition between the previous chapters and
the next one, which deals exclusively with figurative meanings of colour terms.
Here, I examine cases where the main reference is clearly within the colour
domain, but where it would nevertheless seem that it is the extra, non-colour,
meaning that is actually more important. In the case of figurative meaning, on the
other hand, the reference of the colour term has moved out of the colour domain
completely. However, it is not possible to draw a precise boundary and some of the
cases that I discuss below could possibly be viewed as instances of figurative use.
There are at least two phenomena that can be identified, and it is these two
phenomena that will be dealt with here. Consider (1) and (2) below:
(1)

Her cobalt eyes are hidden by taupe-coloured sunglasses which she takes off
only once. [ ] The sienna tint of her slightly weathered skin is evidence of a
love of outdoor activity. (BoE: oznews)

(2)

Maria was livid because it meant everyone was looking out of the window
instead of wishing her happy birthday. (BoE: brmags)

Sentence (1) contains very precise colour terms when describing the actress Jane
Fonda. However, it would appear that this precision is unwarranted from the point
of view of colour. I myself had no idea of the exact designation of cobalt, taupe
and sienna, and most of the native speakers I have consulted have confessed
insecurity in this respect. Presumably this would be true of most readers of the
magazine article where the phrase was recorded, so why this exactness? A number
of my informants have spontaneously pointed to the almost ridiculous effect of the
colour terms in (1). So, one possible answer to the above question is the extra
meanings that such precision seems to evoke, here in terms of emotive overtones.
In the first part of this chapter, I discuss the type of phenomenon illustrated by
(1) and analyse it with the help of Cruses (1977) modification and elaboration of
the Jakobsonian concept of markedness. Using a case study of saturation terms, I
argue that overspecification can create extra meanings and highlight attributes

161

162

Chapter 6

outside the colour domain. A previous attempt along the same lines of
argumentation can be found in Steinvall (2000), in the context of Victorian poetry.
Sentence (2) illustrates that an ECT may become closely related to specific
moods, and in this particular case, the connection has reached so far that the colour
term can stand for the mood. In fact, it can be argued that the term livid has almost
completely left the colour domain; this appears to be the case in (3) below, where
both livid and scarlet are mentioned without any apparent contradiction being felt.
(3)

Prue was livid now, scarlet with humiliation. There had better be a damn good
explanation. (BoE: usbooks)

It was demonstrated in Chapter 5 that puce is seldom used to denote other colours
than facial ones, and this appears to be even more true of livid. In both cases, the
colour term is used in connection with a particular mood. In the second part of this
chapter, I explore the relationship between colour terms and emotions and attempt
to establish whether there is any systematicity. However, since I am primarily
interested in the fuzzy territory between literal and figurative use, I take into
account only those occasions on which the colour refers to skin colour. Both
general patterns and the function of specific terms are considered.

6.2 Markedness and specificity

The notion of markedness goes back to the Prague school of linguistics and in
particular the Russian scholars Roman Jakobson and Nikolai Trubetzkoy.
Markedness in its prototypical form entails a binary opposition between a marked
and an unmarked feature or term, be they phonological, morphological or semantic,
where the unmarked feature represents a default value. The idea of markedness
has proved to be a very useful tool employed in many schools of linguistics, and as
the idea has spread the definitions of markedness have become less specific and
more vague than was Jakobsons original definition. Andrews (1990) provides a
critical exposure of the history and development of the concept of markedness.
In Lyons (1977), we can find a good illustration of the standard conception of
markedness theory in relation to the analysis of lexical structure. He (p 305-306)
identifies three different forms of markedness: formal (e.g. count countess),
distributional (e.g. high low)1 and semantic marking (e.g. dog bitch).
Semantic marking, which is the most important for the present study, is specified
by Lyons (1977: 307) as indicating that [a] semantically marked lexeme is one
that is more specific in sense than the correspondingly unmarked lexeme.
It is a slightly different notion of semantic markedness that I will take into
account here, one put forward by Cruse (1977).2 Cruses use of markedness theory
is original in two respects. First, he elaborates the notion of semantic markedness.
As we could see above, Lyons definition of semantic markedness is unidirectional
1

Here low can be said to be marked since its distribution is lower. For instance, it does not usually occur in
questions referring to height; consider the odd nature of How low is that bridge?
2
Other researchers, notably Rosch (1978), Clark and Murphy (1982), have observed the effect of
unwarranted specificity.

In Search of Extra Meaning: ECTs and Markedness

163

and has a straightforward relation to specificity: in a lexical relation it is always the

more specific term that is the marked one. Drawing on Browns (1958)
observations, Cruse, on the other hand, identifies an intermediate level of
specificity, which in his opinion is the unmarked one. He calls this level INS
(Inherently Neutral Specificity) (Cruse 1977: 155). It corresponds more or less
exactly to what is identified as the Basic level in cognitive linguistics (cf. Chapter
2). In Cruses view both subordinates and superordinates are marked in their
relation to the basic level. Basic level terms such as car and dog, are inherently
neutral as regards specificity and would be the normal terms used in conversation.3
In certain contexts, however, there may be a need for a higher degree of specificity.
Consider (4) below (taken from Cruse 1977: 156):
(4)

(a) Ill have to take the dog to the vet tomorrow.

(b) Ill have to take the alsatian to the vet tomorrow.

Dog is basic level and alsatian is subordinate. Nevertheless, depending on the

context, either sentence can be preferred. In a context in which only one dog is
concerned, sentence (b) would appear to be less normal. However, if there were
several dogs involved in the context, then sentence (a) would be uninformative and
strange.
This illustrates Cruses second modification of markedness theory in
comparison with the version suggested by Lyons. He stretches the idea of
markedness to include pragmatic considerations. With the help of Grices maxim of
Quantity,4 Cruse maintains that it is possible to identify contexts in which a
subordinate term would be contextually unmarked. In other words, the preferred
level of specificity and what can be interpreted as marked is very much
contextually bound. Cruse (1977: 156) formulates a rule to account for how an
utterance can be said to be unmarked:
To obtain an unmarked utterance in a given context, use INS unless this results in an
abnormal communication, in which case deviate from INS to the minimum degree
required to ensure normality.

This is, basically, an application of the maxim of Quantity. Following Cruse (1977:
156), we can refer to terms meeting the above rule as CNS, Contextually Neutral
Specificity. The analysis demonstrates that it is important to distinguish between
INS and CNS. Whereas INS is a cognitive-semantic unit identifiable in isolation, or
with a minimum context influence, CNS is only identifiable in a specific context.
Accordingly, the full meaning of a marked lexeme and, in fact, its very status as
marked can only be established in context. An interesting consequence of Cruses
theories, which is not predictable from Lyons model of semantic marking, is that
the system is asymmetric. Cruse (p 159) points out that
3

As stated by Ungerer and Schmid (1996: 63), the basic level is usually preferred when we first introduce a
subject.
4
The maxim of Quantity is expressed as follows (Grice 1989: 26):
1. Make your contribution as informative as is required (for the current purpose of the exchange).
2. Do not make your contribution more informative than is required.

164

Chapter 6

[t]he inherent nature of the inherently neutral level [i.e. the basic level] is reflected in
the fact that if an INS is not at the same time CNS, then it must lead to an abnormal
communication: it can never give rise to a marked utterance.
[---] A level of specificity other than INS, however, has no inherent character: it may
yield marked, unmarked or abnormal utterances, depending entirely on the context.
[Emphasis added].

Consequently, the use of hyponyms and superordinates normally has one of two
functions (if we may assume that people avoid abnormal communication): either it
is contextually motivated and produces an unmarked utterance, or it is not
contextually motivated and, hence, produces a marked utterance. A marked
utterance could be interpreted as a flouting of the maxim of Quantity, and the
hearer might look for other meanings. In what follows, I take a closer look at the
use of four ECTs in relation to the notion of markedness.

6.3 The marked use of saturation terms

As could be seen in Chapter 5, most ECTs occupy a position in the colour domain
which makes them difficult to classify in terms of hyponymy. It would, for
instance, be misleading to say that turquoise is a hyponym of blue but not of green.
The fact that many terms have this intermediate position makes them less likely to
exhibit markedness phenomena of the type discussed here. As noted by Clark and
Murphy (1982), referents which are atypical of a category may generate specific
vocabulary: it is more reasonable to refer to an ostrich by the term ostrich rather
than the term bird. By the same token, intermediate colour terms such as turquoise,
lime and maroon may be used more frequently in unmarked contexts due to their
intermediate position in relation to BCTs: a turquoise nuance is a poor example of
both green and blue and may therefore readily elicit the term turquoise in most
contexts.
There are, however, terms which appear to be good candidates for marked
usage. These colour terms are what I call here saturation terms. A saturation term,
as it is used here, refers to an ECT which designates an area of high saturation
clearly within the scope of a BCT. In everyday language but also in the controlled
language of dictionaries, such nuances are frequently referred to as bright or
vivid (cf. 0.4, 5.2.1 and 5.2.2). For example, the OALD defines scarlet as a
bright red colour which would make it into a saturation term according to my
definition. What makes saturation terms particularly interesting is the fact that they
refer to what most people appear to classify as the most typical example of a
particular BCT. Collier (1973) observed that the focus of the BCTs coincided with
the areas of highest saturation and suggested that Berlin and Kays (1969) result
could be an artefact of their experimental procedure (the Munsell chart used by
B&K was maximised for saturation). Although he was later (Collier et al. 1976)
able to demonstrate that BCT foci were stable at a uniform level of saturation, he
still maintained that (1976: 889) basic color words are associated with colors of
high saturation.
In view of these observations, we may enquire why there should be ECTs that
designate the best example of the general, BCT, category. It this particular

In Search of Extra Meaning: ECTs and Markedness

165

circumstance that makes saturation terms a fitting object of study. Given Cassons
(1994) research, it is tempting to suggest that there could be a cultural explanation.
The fact that it was difficult and expensive to create dyes and colours that could
produce bright colours made it socially and culturally important to designate these
colours using precise terms.
Here the identification of saturation terms is based on their codification in
dictionaries (cf. the dictionary study of ECT definitions in 5.2.2). The study is
delimited so as to include only saturation terms referring to areas within the four
chromatic Primary Basic Colour categories (i.e. BLUE, GREEN, RED and YELLOW).
A term is regarded as a saturation term if its dictionary description includes the
terms bright or vivid and it is only described with reference to one of the Primary
BCTs. Dictionary definitions can be found in Appendix 4. As mentioned in Chapter
5, there are considerable differences between the six dictionaries studied, but with
an inclusive approach (I have included a term if it is mentioned in any one
dictionary as bright or vivid) the eight colour terms listed in Table 6:1 can be
considered to be saturation terms.5
Table 6:1. Saturation terms
PRIMARY
BCT

BLUE

GREEN

RED

YELLOW

TOKENS IN
BOE

27102

28664

38912

13055

SATURATION TERM

Azure

Ultramarine

Emerald

Carmine

Scarlet

Vermilion

Golden

Lemon

TOKENS IN
BOE

136

35

326

80

1208

98

39836

231

TRANSPARENCY

Opaque

Opaque

Transparent

Opaque7

Opaque

Opaque

Transparent

Transparent

It was argued in Chapter 5 that the transparency of a term may influence the full
impact of the meaning of it and therefore this aspect (based on Cassons 1994
classification) has been included in Table 6:1. The frequencies clearly demonstrate
that there appears to be one term in each BCT category which dominates as a
salient saturation term. These four salient terms, azure, emerald, scarlet and
golden, are the focus of attention. We can also observe that the total frequency of
5
Crimson might be considered a good candidate too, as the OALD defines it as a deep red. Deep usually
refers to a combination of high saturation and little lightness. However, all other dictionaries included a
purple nuance in the term, which led me to decide against including it.
6
The figure for golden is an estimation, cf. 3.2.
7
Carmine does not occur in Cassons (1994) list, but I think it is safe to categorise this term as opaque. Cf.
the etymology given in the OED.

166

Chapter 6

the saturation terms is only a small fraction compared with that of the superordinate
basic term, except in the case of golden vs. yellow, where the saturation term has a
much stronger position. The structure of the yellow category was discussed in
3.2.1.
6.3.1 The pattern
Before we look at the data, a methodological issue should be mentioned. Since the
aim is to study marked use, the questions arise of how to find these marked
instances and how to abstract general patterns. If Cruses (1977) line of reasoning
is correct, then it could be claimed that, for a marked reading to occur of a
particular collocation, the minimum criterion should be that the basic level term
should be the normal choice; the INS should be the CNS, to use Cruses
terminology. One problem is that Cruses theory and examples deal with nouns,
whereas adjectives are of principal concern here. This suggests that we have to take
into account the objects to which the colour terms ascribe a specific nuance, and
therefore this study is based on adjective-noun collocations.
I decided on the following procedure in approaching my material. Since the
saturation terms are the centre of attention, I recorded the 25 most frequent
saturation term+noun collocations8 (although n could not be lower than 2).
Included in this operation were also the phrases of the type ECT-BCT + noun (e.g.
emerald-green ) since I felt that they should be part of a full description. Next,
the frequency of the superordinate BCT+noun collocations was recorded;9 thus in
this second step, the nouns were the same as from the first step. Finally, to obtain a
clear picture, the collocations were sorted into general categories on the basis of the
noun. After this initial work, the individual examples were examined in search of
marked meanings.
This section is organised in a similar way: First the overall pattern is
presented together with some general observations; then individual examples are
scrutinised, and at the end of the section I present an interpretation of the processes
involved.
The first term to be considered is emerald, the saturation term of green. Emerald is
a transparent term and most English speakers are probably aware of the precious
stone from which the colour term is derived. The nouns below were qualified by
emerald and/or emerald green at least twice in the Bank of English corpus. The
words appear in order of frequency; the most frequent collocations are mentioned
first. Those marked with an asterisk represent collocations appearing only with
emerald green.
Eyes, isle(s), greens, water(s), landscape, rice, lake(s), ray, fairways, doves,
turf, paddy, lawn, paradise, valley, glass, sea, fire,*dress,*suit, *tub,
*iguanas, *grove, *velvet, *ankle socks, *grass.

As a result of the software available at the BoE, singular and plural nouns are listed as two different
collocations.
9
The actual exploration of the BoE took place on August 9, 2001, which means that it was done after the
modification of the BoE corpus in September 2000. However, this should not have affected the result.

In Search of Extra Meaning: ECTs and Markedness

167

Table 6:2 presents the nouns organised in nominal fields and includes the total
score for each group. The nominal fields are only meant to be notional and are not
given any theoretical definitions (cf. Chapter 5). The scores can then be compared
with those of the BCT green. These groups seem to form a good basis for an
investigation of markedness effects involving emerald.
As the frequencies suggest, green is a much more common collocation of
these nouns in most cases although there are some individual exceptions. Notably,
isle(s) occurs more often together with emerald. It should be remembered that the
set phrase the Emerald Isle referring to Ireland has been excluded since it is
normally spelt with capitals. Even if the notion of nominal field is tentative and not
well-defined here, it is nevertheless obvious that emerald is most often used in
reference to predominantly natural objects and less often to artefacts. A
conspicuous detail is that the phrase emerald green, rather than emerald, referred to
all nouns within the nominal field of CLOTHES.
Table 6:2. Nouns qualified by emerald organised in nominal fields.

NOMINAL FIELD

NOUNS

TOTAL SCORE
EMERALD
(GREEN)

Eye

Eyes

12

422

Water

Lake(s), sea,
water(s)

14

125

Landscape features
not including water

Landscape, rice,
paddy, valley,
grove, pasture

14

91

Grass

Greens, fairways,
turf, lawn, grass

12

223

Clothes

Dress, suit, velvet,

ankle socks

12

163

Island
Animals

Isle(s)
Doves, iguanas

7
4

1
5

TOTAL SCORE
GREEN

The next term, azure, is classified by Casson as opaque. Like emerald the colour
term is derived from a precious stone, in this case lapis lazuli. The following nouns
occurred at least twice in the corpus together with azure. The pattern is quite
similar to that of emerald.
Sky(ies), water(s), sea(s), kingfisher, eyes, Caribbean, Mediterranean,
*flowers.10

10

The asterisk indicates that flowers was modified by azure-blue.

168

Chapter 6

In the case of azure kingfisher, it is the name of a species (alcedo azurea) and
hence, an example of classificatory use and not description. It has not therefore
been included in the table.
Table 6:3. Nouns qualified by azure organised in nominal fields.

NOMINAL FIELD

NOUNS

TOTAL SCORE
AZURE (BLUE)

TOTAL SCORE
BLUE

Water

Water(s), sea(s),
Caribbean,
Mediterranean

30

531

Sky

Sky(ies)

18

926

Eye

Eyes

1684

Plants

Flowers

247

As might have been expected, azure is mostly used in reference to water and the
sky. In common with emerald, azure is used primarily in reference to the colour of
natural objects.
If we then turn our attention to the much more frequent use of scarlet, a
slightly different pattern appears. In the case of scarlet, it was possible to collect 25
collocations with a score higher than two. In fact the number of collocations was
much higher but only the 25 most frequent were taken into consideration. The
collocations are listed below.
Fever, woman, lipstick, flowers, silk, sea perch, dress, letter, berries, blooms,
tunic, lips, macaw, velvet, leather, cloak, robe(s), cloth, jacket, women,
pimpernel, poppies, gown, hips
One conspicuous detail that makes scarlet different from the previous terms is the
fact that there are several instances which do not occur together with the
superordinate terms red. Thus, we have scarlet fever, scarlet woman and women,
scarlet sea perch, scarlet letter,11 scarlet macaw and scarlet pimpernel, but no
corresponding red fever etc. In all these cases we can talk about scarlet as
exercising type modification, i.e. having a classificatory function (cf. Chapter 4).
In some of these collocations, scarlet letter and scarlet woman, the colour term has
acquired figurative meaning. These particular collocations are dealt with in greater
detail in Chapter 7. Another detail that is different is the complete lack of nominal
collocations containing scarlet red. Below the collocations have been organised in
nominal fields on the basis of the noun. The above-mentioned instances of type
modification have not been included.
11

The collocation red letter does exist of course, but with a distinctly different meaning from that of scarlet
letter.

In Search of Extra Meaning: ECTs and Markedness

169

Table 6:4. Nouns qualified by scarlet organised in nominal fields.

NOMINAL FIELD

NOUNS

TOTAL SCORE
SCARLET

TOTAL SCORE
RED

Clothes

Silk, dress, tunic,

velvet, leather,
cloak, robe(s),
cloth, jacket, gown

91

561

Plants

Flowers, berries,
blooms, poppies,
hips

51

287

Cosmetics

Lipstick, lips

31

259

As Table 6:4 demonstrates, the use of scarlet is different from that of azure and
emerald. In contrast to those terms, scarlet occurs most frequently when qualifying
an item of clothing.
Finally, there is the transparent term golden. The 25 most frequent nominal
collocations can be found below.
Syrup, hair, light, retriever, glow, eagle(s), beaches, tan, perch, sand(s),
arches, plover, dome, curls, hue, caster sugar, pheasant, skin, seal, crust,
leaves, rod, beach, *flowers, *foliage, *leaves, *ray, *petals, *fruits,
*daffodils12
Here, as in the case of scarlet, there are a number of collocations which do not
occur together with the superordinate BCT, yellow, and which have the function of
type modification. Among the 25 collocations given above, the following can be
categorised as instances of type modification: golden syrup, golden retriever,
golden eagle(s), golden perch, golden plover, golden caster sugar, golden
pheasant, golden seal, and golden rod. There are no corresponding phrases
containing yellow.
There is another feature which makes the relationship between golden and
yellow decidedly different from that between the other terms we have considered:
the relatively low salience of yellow. This is apparent from the frequencies in Table
6:1, but it becomes even more conspicuous when we look at these collocations.
Among the 25 collocations of golden there are only two instances where the yellow
+ noun collocation has a higher score than that containing golden. These two
collocations are yellow light and yellow skin. Tellingly, these two collocations have
completely different meanings from those containing golden so we cannot talk
12
The star-marked words are collocations of golden-yellow. They are added to the 25 collocations of golden
because they are at least as frequent as golden beach.

170

Chapter 6

about greater precision in the latter case (yellow light typically refers to traffic
lights, and yellow skin refers to the skin of fruit or (rarely) East Asians). This
pattern becomes very clear when we look at the collocations organised in nominal
fields. As in the case of scarlet, I have not included type modifications in the
categorisation in Table 6:5.
Table 6:5. Nouns qualified by golden organised in nominal fields.

NOMINAL FIELD

NOUNS

TOTAL SCORE
GOLDEN
(YELLOW)

TOTAL SCORE
YELLOW

Light

Light, glow

157

160

Landscape features

Beach(es), sand(s)

140

28

Hair

Hair, curls

119

83

Skin

Tan, hue, skin

84

33

Plants

Leaves, flowers,
foliage

74

439

Building

Arches, dome

58

Cooking

Crust

17

The nominal field of COOKING should be mentioned in more detail. It would seem
that the use of golden as denoting objects within the domain of COOKING is actually
much more frequent than the collocation golden crust suggests. However, due to
the method used here, the high frequency of this domain could not be detected,
apart from this collocation. Typically the syntactic structure of this use is that of (5)
below.
(5)

For crunchy sauteed potatoes, use a floury variety. Roughly chop and parboil,
then saute in butter and olive oil, stirring until golden. (BoE: brmags)

This kind of reference to the colour of food is frequently in the form of goldenbrown. Accordingly, it would seem that golden could be viewed as a subordinate of
brown in this particular context.
To sum up, we have seen in this section that the saturation terms seem to be
particularly frequent in connection with certain nominal categories. It was also
apparent that there are some individual differences between the terms. Whereas
azure and emerald are quite infrequent and mostly used for descriptive purposes,
scarlet and especially golden are frequent and both terms are also regularly used for
classificatory purposes.
6.3.2 A look at some individual examples
Although the categorisation in nominal fields revealed the existence of patterns, it
gave us no guidance as to whether these patterns were indications of marked

In Search of Extra Meaning: ECTs and Markedness

171

readings. In what follows, individual examples are presented and compared with
the intention of demonstrating the extra meaning that these terms may exhibit in
certain contexts.
Let us first consider emerald. The nominal fields that were of special interest
listed in Table 6:2 included EYE, WATER, LANDSCAPE FEATURES NOT INCLUDING
WATER, GRASS, CLOTHES, ISLAND, and ANIMALS. Of these, the first four would appear
to be of special interest: they constitute fields in which we would not normally
assume great precision.13 Below are some examples relating to EYE.
(6)

the tears and forced the sparkle back into her remarkable emerald eyes. She
ran a hand through her hair; it sprang back into order, (BoE: brbooks)

(7)

And she retains her femininity, sparkling emerald eyes and lovely long curls of
dark red hair. (BoE: oznews)

(8)

In Dublin this week you cant move without seeing the face of Mary Black. Her
raven black hair and emerald eyes look out from giant posters all over town.
(BoE: today)

(9)

Oxford guy longish brown hair, 30 years old, 510 green eyes medium build,
clean shaven, boyish looks and straight acting. (BoE: ukmags)

(10)

Karen stopped, drove a finger straight at my face, and glared at me with her
fatigued green eyes. (BoE: usbooks)

It would seem that the use of emerald creates some extra meaning. This extra
dimension is quite difficult to pinpoint in the examples containing the word
emerald. In the first two examples much of the force is carried by the adjectives
remarkable and sparkling, and it could be argued that not much would be lost if
green was substituted for emerald in (6) and (7). But something is lost and in (8)
the force of emerald in comparison with green is apparent; there is more to it than
simply precision. In fact, this becomes very clear if we do it the other way;
substituting emerald for green in (9) and (10) creates strange sentences.
It seems that the precision that the use of emerald suggests evokes an
emotional perspective on the part of the describer which goes less well with the
assumed objectivity of (9). Furthermore, there appears to be an element of
positiveness in emerald which makes it clash with fatigued in (10). On the whole,
all eleven examples of emerald (green) eyes in the BoE carry to a greater or lesser
extent these dimensions, which will be labelled attributes below, in line with the
description of meaning presented in Chapter 2.
Preliminarily, we can observe that English speakers usually classify peoples
eye colours in three wide categories: blue, green or brown. Therefore, green eyes is
an unmarked collocation which can be used in any context. Emerald eyes, on the
other hand, is a specification beyond what is normal. It could be argued that it is a
13
Actually, this is true of ISLAND and ANIMALS too. They are excluded since ISLAND is quite similar to
landscape features and ANIMALS is very infrequent.

172

Chapter 6

violation of the maxim of Quantity and therefore attributes outside the colour
domain are given weight.
A look at some of the other nominal fields mentioned in Table 6:2 tells a
similar story. Consider the examples below in the field of LANDSCAPE FEATURES.
(11)

Driving the final sector down through the foothills to the coast, our road threads
its way through an emerald valley of great scenic beauty to reach the old Black
Sea port of Trabzon. (BoE: brephem)

(12)

Now, everywhere she looked she saw her beloved, emerald landscape, and no
way out. (BoE: brbooks)

(13)

But in the storms wake, a brown gash between 20 and 30 miles wide now cuts
across the emerald landscape. Andrew tore across the Everglades from east to
west like a runaway chainsaw. (BoE: npr)

(14)

Many centuries ago the remote desert oasis of San Pedro de Atacama had been
the centre of a flourishing Paleolithic civilization based within impressive rock
fortresses built upon the steep mountains that almost encircle the green valley.
Thousands of ancient graves scattered throughout the hills were beginning to
yield their secrets. (BoE: brbooks)

(15)

In ancient times, a Paez Indian chieftain would climb a peak and look out over
the surrounding rivers and green valleys. Whatever his gaze reached
determined what belonged to his clan. (BoE: npr)

(16)

High winds and rain clouds were sweeping across the flat green landscape as
Alan fought with the controls and tried to land. (BoE: brbooks)

As in the case of eye colour, green appears to have a neutral tone here. In some
cases, its function is close to classificatory, categorising these landscape features as
fertile in contrast to, say, a parched valley. It could be argued tentatively that green
here functions as a metonym for verdure. Consequently, the use of green goes well
with an objective description like that in (14).
The function of emerald is essentially the same as that in the domain of eyes.
Since we do not normally contrast shades of green in nature, at least not in cases
such as valleys, we are led to consider the use of emerald in this context as a
flouting of the maxim of Quantity. Therefore we look for external significant
features. Again there is an element of positive emotion present. The marked term
can tentatively be said to produce an association such as sunlight, both literally and
figuratively. The strong element of subjectivity is also apparent. Example (11) and
(12) can be said to illustrate this point. The marked character of emerald can also
be demonstrated by the fact that it does not function well in sentences characterised
by an objective tone, such as (14) and (15). The implication of a temporal feature
like sunshine in emerald (cf. (16) in which emerald would be strange) suggests in
turn less time stability than the use of green does. This may be a further reason why
emerald appears to be unsuitable in the sentences above. On the other hand, as
could be predicted, there is no problem in substituting green for emerald in (11),

In Search of Extra Meaning: ECTs and Markedness

173

(12) and (13). The same pattern can also be found in the other nominal fields
containing natural objects: WATER, GRASS and ISLANDS.
Let us finally turn our attention to the nominal field of CLOTHES, another field
where emerald occurs. Interestingly, it is virtually impossible to detect the same
kind of effects as those we could discern above. A substitution of emerald for green
does not seem to create the same markedness. It is significant that in the collocation
belonging to this field, emerald is often accompanied by green. Apparently, what
makes the nominal field of clothes different from those examined above is that
clothes may occur in different nuances of green, and it is therefore quite natural to
make a fine distinction. Such precision would not imply subjectivity but, on the
contrary, a focussed objectivity. This would appear to be the difference and
therefore we do not find a markedness effect; it is not perceived as a violation of
the maxim of Quantity, but rather as CNS, contextually neutral specificity.
To sum up, this examination of some instances of emerald demonstrates that
the use of this term can lead to marked meanings. It appears that these are likely to
arise in contexts where the degree of specificity that the use of emerald suggests is
unexpected. Thus, it could be claimed that emerald is always marked when used
with nouns that refer to categories such as LANDSCAPE FEATURES or EYES. As
predicted by Cruses theory, emerald does not give rise to marked readings in
nominal fields where we might expect this kind of precision. Accordingly, in the
field of CLOTHES it is not possible to detect any such effects, cf. (17) below.
(17)

There were elegant ladies in dark velvet suits and pearls, a redhead girl in an
emerald green suit, a tall blonde in corduroy shorts and a skintight animalprint top. (BoE: brmags)

If we consider azure, a similar pattern emerges. Azure is most often used in

reference to water or to the sky, and with it there follow elements of positiveness
and subjectivity; consider the examples below.
(18)

Cozumel, just a short sail from Playa del Carmen, is a divers paradise, with
clear blue waters unrivaled in the Caribbean. (BoE: usbooks)

(19)

A strong contender for the title of Most Beautiful of all Club Med villages is the
village at Punta Cana on the remote southeastern tip of the Dominican
Republic. Here, the beach is breathtakingly lovely, banked on one side by a
grove of coconut palms and on the other by the azure waters of the Caribbean
Sea. (BoE: usbooks)

(20)

I cant believe it. January 6 and Ive got miles and miles of sand to myself.
You can stick your shimmering azure seas. The North Sea here is a beautiful
autumnal shade of brown. (BoE: times)

(21)

The view was breathtaking. The cloudless azure sky contrasted vividly with the
deep green of the trees which rustled gently in the afternoon breeze, and on the
horizon he could make out the hazy outline of one of the countrys many extinct
volcanoes, which shimmered like a flickering mirage in the distance. (BoE:
brbooks)

174

Chapter 6

(22)

And we can only marvel at the golden domes and great bell-tower of Zagorsk,
silhouetted so defiantly beautiful against the crisp blue sky. (BoE: ukephem/02)

(23)

I was unconscious for three days; but I returned into consciousness for several
seconds, and I saw that I was lying on the ground, and I saw the blue sky; but
then I saw that my legs were turned something like ninety degrees to another
side; (BoE: bbc)

As these examples demonstrate, azure very often comes with a distinct tone of
exotica; azure skies and azure waters most often refer to places like the Caribbean
or the Mediterranean or somewhere near the equator. Again, the effect of azure is
best captured if we try to put it in a more neutral context: whereas azure goes well
with the subjective and emotional tone of (21), it becomes decidedly more
awkward in the more neutral narrative of (23). One may also wonder whether the
sea outside Torquay could ever be called azure even if it should happen to have that
particular nuance on one day this is of course part of the humour in (20).
Although it is true that the water and the sky have a deeper nuance of blue close to
the equator, it is quite clear that it is not the nuance that is of primary interest.
The collocation azure eyes occurs only four times, which is surprising given
the high frequency of blue eyes. One conjecture concerning the low frequency is
the strong connection to the exotic that seems to characterise other uses of azure. It
may be that this strong connection creates a notion of exaggeration which blocks
the usage. It could also be that azure is so strongly associated with WATER and SKY
(cf. Table 6:3) that other collocates appear a little strange.
If we look at the collocation azure-blue flowers, we find a similar pattern to
that of emerald in the nominal field of CLOTHING. In a sentence like (24) below,
there is no marked reading which creates extra meanings. We find only precision.
Again this is probably due to the character of the field, and possibly the type of
discourse that we find here a horticultural context.
(24)

Not just any member of the versatile Linum family, but a herbaceous queen
called Linum narbonense. A flamboyant native of Southern Europe we started
growing it in 1759 its famed for its three month display of silk, saucershaped, five-petaled, azure-blue flowers. (BoE: brmags)

As mentioned earlier, the collocations of scarlet reveal a different story. Table 6:4
shows that, ignoring the classificatory use of the term, we find scarlet in three
lexical fields: clothes, plants and cosmetics. In all of these fields, precision is quite
natural. Thus, it is very difficult to detect any markedness effects in the case of
scarlet. Furthermore, it would seem that scarlet has a slightly different meaning
structure than azure and emerald. There are a number of cultural and historical
attributes which are attached to scarlet and which may be centralised in certain
contexts. The OED mentions that scarlet may designate
[o]fficial or ceremonial costume of scarlet, as the uniform of a soldier, the gown or
robe of a doctor of divinity or law, a judge, a cardinal, etc.; also, the scarlet coat worn
in the hunting field. [The OED: scarlet A.3.]

In Search of Extra Meaning: ECTs and Markedness

175

It would seem that this kind of use is becoming rare and the few times such a
connection is evoked, it is usually in the context of some historical discussion or
referring to the military parade uniform. Another attribute that has been associated
with scarlet is prostitution. In (25) scarlet could actually produce both links, given
the status of Princess Diana, but the explicit mention of scarlet woman is likely to
eliminate the royal association altogether.
(25)

Diana swept into Paris for a gala dinner wearing a plunging gown in vivid
scarlet. A scarlet dress for a scarlet woman? The newspaper headlines in
Britain wrote themselves. (BoE: oznews)

In scarlet woman we have classificatory use only and it is not an example of a

marked reading. Although the process involved here could be viewed as analogous
to that observed for azure and emerald, there are apparent differences. First, the
describer is not emotionally affected. Second, the connection is less direct in (25),
and would probably be lost unless it was spelt out. One might assume, however,
that a collocation like scarlet lips could evoke the same kind of subjectivity as
emerald eyes.
As observed above, golden is different from the other saturation terms in that
it does not seem to contrast with its superordinate BCT (yellow) in the same way as
the other terms. It emerges that golden contrasts with yellow on very few occasions.
However, this does not appear to cancel out marked readings. Instead of a contrast
with yellow we find other types of contrasts. In the context of hair, the unmarked
choice would be blonde, of course. Consider the examples below. Here we can
detect more or less exactly the same effects as with azure and emerald.
(26)

Of sister Jessie, Howes wrote: The best looking of the girls, with long golden
hair and the most beautiful skin. She was much more sociable than the rest of
us, and from the age of six was hardly ever home to tea. (BoE: times)

(27)

Imagine that you are sitting on a beautiful island beach of golden sand with
palm trees behind you and wonderfully, brightly-coloured birds flying back and
forth between the trees. (BoE: ukbooks)

(28)

The Atlantic shores offer a choice of holiday environment. Sunny Isles, at the
northern end of Miami Beach, has a pleasant family atmosphere, golden
beaches and is within easy reach of the glamorous Bal Harbour shopping mall.
(BoE: ukephem)

It may be argued that the extra meaning of golden does not draw as much on the
contrast in precision as on its transparent character. However, one factor which
counters this is that, as in the case of the previous terms, the marked subjective
reading of golden seems to disappear in certain nominal fields such as PLANTS and
CLOTHING. Here, apparently, the specific term is CNS, cf. in (29) below.
(29)

There are several different forms of this shrub. One is of compact habit, another
has golden foliage and two varieties have golden and yellow berries. (BoE:
ukmags)

176

Chapter 6

The mention of both golden and yellow makes it quite clear that golden has CNS
status. Furthermore, in this phrase we can also see that golden and yellow can
function as co-hyponyms.
To sum up, a closer look at the individual examples reveals that we do indeed
find effects that go beyond the colour domain in the use of these terms. This is
especially clear in the case of azure, emerald and golden. It could be observed that
the effect correlated clearly with the character of the nominal field. Typically the
elements of extra meaning arose in fields where this kind of specificity is not
expected. This is also probably the main reason why few markedness effects could
be found in the case of scarlet. It appears that this term is primarily used together
with nouns from domains where high precision concerning colour is only to be
expected.
6.3.3 Theorising the effects
In this final section on saturation terms, I try to analyse the origin of the effects that
seem to follow a marked use of these terms. It was argued above that two features
are closely linked with marked use of saturation terms: subjectivity on the part of
the describer and positiveness. Thus, the use emerald eyes and golden hair in the
description of a person tells us more about the describers attitude than about the
appearance of the person. It would seem that unwarranted precision creates this
kind of effect. This has been pointed out previously: Rosch (1978: 45) observes
that substitution of subordinate terms for basic-level object names in scripts gives
the effect of satire or snobbery.
But where does the effect come from? Cruse (1977) draws on Mehrabians
(1981 [1971]) notion of verbal immediacy. In Wiener and Mehrabian (1968: 36),
we find a good example of how immediacy and specificity are connected according
to this theory. They argue that
[t]he greater the number of possible additional referents (other than the specific object
referred to in the communication) denoted by the particular symbol, the greater is the
ambiguity and the less denotative specificity in the communication. As an example,
parents referring to their sons fiance might say, our-daughter-to-be, our sons
fiance, his fiance, his lady friend, his friend, she, or the person. These
examples are ordered in decreasing degrees of denotative specificity and are
interpreted in this framework as expressing decreasing degrees of preference,
affect, or evaluation. [emphasis added]

Conversely, an increasing degree of specificity entails greater preference and

affection. However, Caffi and Janney (1994) evaluating different approaches to
emotive communication point out that Mehrabian and Wieners study is exnegativo, i.e. what can really be demonstrated is mainly non-immediate
communication.
Cruse (1977) proposes a modification of this approach to specificity. Instead
of claiming that specificity necessarily entails liking, he draws attention to the fact
that unpleasant things can also be highly specified by precise terms. Thus, he
(1977: 163) suggests that underspecification de-emphasizes the feature which is
omitted, while overspecification emphasizes or intensifies the added feature.
Essentially I follow this idea, although I adjust it slightly to be able to fit it into the

In Search of Extra Meaning: ECTs and Markedness

177

theory of cognitive linguistics adopted here. Accordingly, I talk about attributes,

and with an encyclopaedic view of meaning, attributes outside the colour domain
may be taken into account.
Broadly speaking, we can talk about two factors that are important for the
creation and interpretation of a marked utterance:
The context in which the utterance appears.
The attribute structure of the colour concept.
Context appears to be the more important factor. It is the context that will decide
whether or not the specificity in the utterance is contextually motivated (CNS). If
not contextually motivated, the specific utterance may generate extra meanings.
These pieces of extra meanings may very well be fairly peripheral attributes in an
unmarked context, which are then centralised in the marked context. As an example
of the effects consider (30) below.
(30)

For our interview, Brady chooses a rather different environment, an airy Italian
restaurant in Knightsbridge. To get there she drives me through the West End in
her Porsche Carrera. (BoE: ukmags)

The unmarked choice here would be car14 unless there has been a previous
discussion of the different cars that are available. A closer examination of the
context reveals that this is not the case here. Thus, the mention here of the make of
car could possibly highlight attributes such as [expensive] and [exclusive], and,
interestingly enough, these attributes seem to affect the owner indirectly in the
sense that we project onto her related attributes such as [wealthy] and [vain]. In a
different context the mention of the same make could highlight an attribute like
[fast], or be unmarked as in (31) below.
(31)

Lucky then recruited me for the Cuban GP in Havana, on 28 February. This was
now Castros Cuba, and the Communists certainly ran things more efficiently; I
learned the new part-aerodrome circuit in a borrowed Porsche Carrera and an
RSK 1600 before the Maserati was ready, and when I finally took it out it felt
very nice, but not exceptionally fast. (BoE: brbooks)

Consider also the effect if we substitute Trabant for Porsche in (30) our view of
Brady changes considerably.
It appears that it is a parallel type of effect that we find in the readings of
saturation terms where we would not normally expect the precision that the terms
provide. In the case of golden, it may be that the transparent character of the colour
term influences the reading suggesting that an attribute such as [valuable] is central
when marked. In fact, it may be that such an attribute is available most of the time.
Or to put it differently, golden is always marked for [positive].
In the cases of emerald, azure and scarlet, the activation of attributes could
tentatively be described in terms of metonymy. If we accept Cruses suggestion
14

If anything at all since drive suggests car.

178

Chapter 6

above that what is highlighted is the attribute added, we may enquire what this
attribute actually is. Here there are two possible interpretations. One is to restrict
our search to the colour domain. What we find then is the emphasis of saturation, or
in everyday language, the brightness of these terms, or possibly intensity. From
this analysis we can let a number of metonymies lead us to the effect. A tentative
process is outlined below.
I suggested earlier that the extra meaning could be described as positive
emotions. Through a process involving a series of (established, conceptual)
metaphors and metonymies the addressee can arrive at such an interpretation. The
characterisation below is based on eyes.
Brightness and intensity can metonymically stand for warmth (sun, fire).
[metonymy]
So if a persons eyes are bright or intense, then (s)he must be warm inside.
[metonymy]
Warmth inside a person derives from strong feelings. [metonymy]
It should be noted that the locus of these feelings need not necessarily be the
possessor of the bright eyes. On the contrary, although we are likely to construe a
person described as having emerald eyes, we may well contend that it is actually
the narrator who creates this image under the influence of her/his own feelings.
This would also seem to be in line with the argumentation of Wiener and
Mehrabians emphasis of immediacy. Recall how the highlighted attributes of the
Porsche Carrera in (30) indirectly affected the readers view of the owner.
In the case of natural objects (azure waters, emerald valleys, golden beaches)
we get an essentially parallel process:
Objects in nature which are bright and intense are exposed to sunlight.
Sunshine correlates with good mood.
If a person describes an object as being bright, then (s)he feels positive
towards it.
However, I would like to suggest that it is not only the difference in denotative
specificity that is important. In my view, peripheral attributes may also be of great
importance. Rather than focussing on brightness, we could also argue that most
saturation terms have some peripheral attribute of positiveness, in the case of
golden and emerald, such attributes can certainly be mediated by the transparent
character of the colour term. In the case of azure, these attributes may be
historically motivated. Remember the importance of the dyes, thus the use of azure
may have a history through which certain positive attributes may have been
acquired.

In Search of Extra Meaning: ECTs and Markedness

179

In connection with this second stand, the status of these attributes can be
discussed further. In a usage-based model such as cognitive linguistics, attributes
are abstracted from usage. This suggests that the frequent association of [positive]
in connection with azure, emerald and golden may lead to these attributes being
part of a conventionalised meaning. It might be felt that these terms belong to a
particular genre indeed, this can be vaguely verified: most examples occur in
written subcorpora. In this view, these three saturation terms can be construed as
forming a class of colour terms used for emphasising positive emotions rather than
specifying colour designation. The use of azure, emerald and golden for descriptive
purposes entails, in most nominal categories, no commitment to colour nuance as
much as to mood. This would also explain why these terms usually form phrases of
the type emerald-green and golden-yellow when they are used for descriptive
purposes only. Other colour terms that appear to have a similar function in certain
categories are turquoise and amber (e.g. EYES) and olive (SKIN).
Relating to what was observed above, it should also be mentioned that this
use of ECTs appears to be tied to particular genres characterised by emotional
texts. Such genres are for instance travel brochures (emerald waters, azure skies,
golden beaches) and Mills & Boon novels (golden hair, emerald eyes, olive skin).
As an example of the latter, consider (32) below.
(32)

There was a sort of evil witchery in her glittering green eyes, her pale magnolia
skin, her sensual, crimson mouth. (BoE: brmags)

In this section I have tried to demonstrate that colour terms can be used for the
purpose of creating extra meaning even though the term still refers to the colour
domain. I have looked at one particular type of colour terms saturation terms
and I have shown that three of these are frequently used for this purpose. The extra
meaning that is usually created is one of emotional affect. An important factor for
this usage has proven to be the nominal field and it is probably for this reason we
interpret the skin colour of Jane Fonda (sienna skin) favourably in (1).

6.4 Facial colour and mood

In this part I consider briefly how the description of facial colour collocates with
mood. In English and in many other European languages, it is quite common to use
colour terms as an indication of different emotions, and very often the colour term
refers to the facial colour of the person experiencing the emotion. Consider for
example such English expressions as to be white with fear, green with envy, red
with anger, to be yellow. Although the motivation behind some of these
expressions merits a detailed study of its own, my main concern here is to establish
an overall pattern and in that way shed some light on the function that an ECT may
have in this context. Generally speaking, it would seem strange that people make
the effort of specifying the exact nuance of the facial colour when people are upset.
Thus, it would seem that the use of an ECT in this context to a certain degree
parallels what was discussed in the first part of the chapter the effect of
unwarranted precision. However, the effect here appears to be slightly different.

180

Chapter 6

Below, I first describe briefly the collocations between emotions and facial colours
that could be found in the BoE and then consider the systematicity in some detail
and offer an explanation.
6.4.1 Colour terms and emotions in the BoE.
There are several difficulties involved in a study of colour terms and emotions in a
corpus. In the first part of this section, I describe some of the problems I have
encountered and how I have tried to cope with them.
One important issue that becomes immediately urgent is what emotional
words should be included. Although Ekman and his associates (Ekman et al. 1972)
claim to have identified basic emotions that correlate with facial gestures, it is clear
that many more lexemes have to be included in a corpus study than those
designating the basic emotions. Fortunately, there are a number of studies
concerned with emotional lexemes, for example Storm and Storm (1987), Shaver et
al. (1987) and Johnson-Laird and Oatley (1989). In this study, I have taken into
account the 80 highest ranked emotion words in Shaver et al. (1987). The authors
asked 112 subjects to rate 213 emotion words on a 4-point scale, so what I have
included are emotion terms that appear to be firmly established within this lexical
field. All the emotion terms considered by Shaver et al. are nouns (e.g. anger, rage,
guilt and happiness), a fact which may affect the result since emotions can also be
expressed by adjectives and verbs.
The emotion words were searched for in the BoE in combination with the
fifty colour terms that are considered in this dissertation. The amount of context
taken into account for the search was a span of 5:5, which means that to be
registered the colour word had to appear within five words on either side of the
emotion word. This span is more generous than that used by most corpus
researchers use: Sinclair (1991) claims that collocations outside a span of 4:4 are of
less interest, and Stubbs (1995) restricts his study of semantic profiles to a span of
3:3. However, even a span of 5:5 does not completely exclude the probability that
interesting patterns may occur outside this scope.
Having completed the search, I examined each example manually since quite
a large part of the assembled material contained colour words which had no relation
at all to the emotion word, as in (33) below.
(33)

Senior played excellent golf from tee to green, but had no joy with his putting.
(BoE: oznews)

Thus, a simple statistical study would be quite misleading. Since I was interested in
the fuzzy boundary between literal and figurative meaning, I only included colour
terms which referred to skin colour. Accordingly, a phrase such as black despair
was not included although there is a clear connection between the colour and the
emotion.

In Search of Extra Meaning: ECTs and Markedness

181

6.4.2 Results
Of the eighty emotion terms considered, no fewer than fifty were linked to some
colour term on at least one occasion. However, quite a few of these revealed the
black despair structure and were excluded. The number of emotion terms cooccurring with a mention of skin colour was 28, a considerably lower, but still quite
high, number. The number of colour terms used for reference to skin colour in
connection with emotions, on the other hand, was very low. Only 13 colour terms
were recorded in this context. Notably, blue was missing despite its close
connection with melancholic feelings. Niemeier (1998) and Verspoor (1998)
speculate that the connection between blue and melancholy might have been
motivated by facial colour, but there is no evidence of this from usage.
The analysis proceeded in the following way. In order to structure the
material, I used a taxonomy of the words presented Shaver et al. (1987). They
asked 100 subjects to sort the 135 emotion words that received the highest scores in
the procedure described above. This meant that all emotion words used in this
study were included in the taxonomies. After the subjects had completed their task,
Shaver et al. performed a cluster analysis of the result and were able to identify six
major clusters with a number of subclusters. The six clusters were named in
accordance with the Ekmanian tradition LOVE, JOY, SURPRISE, ANGER, SADNESS and
FEAR.
Table 6:6 summarises the collocation between facial colours and emotions,
and the results are discussed in greater detail below. The names of the
subcategories are those suggested by Shaver et al. (1987: 1067). The table is
divided into three parts which represent three different foci in the colour domain.
First we have one centred around red, and the colour terms are listed from the
lightest ones to the darkest, white and black included. The second category centres
in the yellow and green area of the spectrum, and finally there is a category of
achromatic terms. Black and white occur in two places, but this is only for
expository reasons and a term can only be scored in one place. As might be
expected, the most common structure among these collocations was colour term
with emotion term, as in green with envy. There were, however, also some other
phrases.
Table 6:6 illustrates quite clearly that there is a general tendency for positive
feelings belonging to the emotional domains such as JOY and LOVE to be only
vaguely associated with facial colours. Furthermore, the few occasions on which
there is such a colour reference, the colour term is one of lightness, pink in
particular. This correlates nicely with the fixed expression in the pink, which,
incidentally, had nothing to do with the colour initially, at least if we are to believe
the OED:
I 2. fig. a. The flower, or finest example of excellence; the embodied
perfection (of some good quality).
b. The most perfect condition or degree of something; the height, extreme. Also
freq. with ellipse of of condition, of health, etc. colloq. [The OED: pink, n 4]

182

Chapter 6

Table 6:6. Emotional terms co-occurring with colour terms in BoE.

White
Pink
Red

1
1
1

7
3

1
1

Scarlet

4
1

29
5
55
1

Purple
Puce
Black

1
22
4
3

Yellow
Green
Emerald

6
2
1

6
22
2

9
1
2

NERVOUSNESS

HORROR

NEGLECT

FEAR

Magenta
Crimson

SHAME

MENT

DISAPPOINT

SADNESS

SUFFERING

SADNESS

ENVY

RAGE

EXASPERATION

IRRITATION

ANGER

PRIDE

CONTENTMENT

ZEST

JOY
CHEERFULNESS

HAPPINESS

Colour
term

LOVE

EMOTION

1
1

3
3
2

1
1
94
2

White

Grey
Black

Here we have a few examples of pink and positive feelings.

(34)

Surprise was a splendid seaman, but in formal gathering he was usually so shy,
ill-at-ease and constrained that it was no kindness to address him; but this
afternoon he was bright pink with pleasure (BoE: brbooks)

(35)

We planned on and on, calculating railroad cars and super tankers scattering
secret seasoning to the ends of the Earth. Exhausted and pink with excitement,
he presented me a four-ounce bottle of his miraculous concoction. (BoE: npr)

The negative basic emotional concepts of ANGER and SADNESS show rather similar
patterns in colour term usage. In particular, this is true of the subcategories of
15
RAGE, SHAME and NEGLECT, which all draw on the red category. The colour term
red functions as a centre of gravity and dominates strongly. This is quite natural, of
15

The centrality of red in connection with emotions was also found by Hupka et al. (1997).

In Search of Extra Meaning: ECTs and Markedness

183

course, since most of us have, at some time, experienced and seen blushing in the
context of these emotions. Kvecses (2000) demonstrates that such disparate
languages as English, Chinese, Japanese and Hungarian share this way of
expressing anger. Possibly analogous with these underlying models, there is a
noticeable tendency for colour terms referring to darker nuances to indicate greater
amplitude of the emotion.16 The fact that purple occurred far more often with rage
than with the more modest term anger would appear to corroborate this
observation. It would seem to make sense that if the anger increases and, thus, the
pressure of the fluid or the heat according to the metonymic models, then the colour
should also be stronger. Thus, it could be claimed that these examples of hyperbole
are motivated by underlying metonymic/metaphoric models. Compare (36) and
(37) below.
(36)
(37)

Thomasina glared at Kathryn, who nodded, and the maid, her face red with
anger, backed out, throwing one last cautionary look at her mistress. (BoE:
brbooks)
He practically threw himself out of his chair at me, his face purple with rage.
How the hell did you know about that? he hissed. (BoE: brbooks)

The same kind of logic would also appear to be true of shame and embarrassment,
but it is not explored as often as in the case of anger. This is probably linked to the
fact that we do not experience these feelings as being as variable in strength as
anger.
One emotional concept that has a completely different colour association
pattern is ENVY. In fact, we have here by far the strongest collocation in green with
envy. It may be justifiably questioned whether this actually refers to facial colour,
but since the structure is identical with other phrases referring to skin colour there
is no a priori reason to exclude the phrase. The motivation for this phrase is
obscure, but Niemeier (1998) speculates that it might be linked to sickness. There
is, however, another possibility. As observed by Geeraerts and Grondelaers (1995),
the ancient and medieval physiological theory of the four humours and the four
temperaments has left many traces in English vocabulary. In this system yellow
was associated with the liver and a choleric temperament, which seems to have
included facets such as envy and jealousy; cf. the association of jaundiced. Since
there was also a green variety of jaundice, an association with green may have been
created this way. In this model, the connection between the green colour of
sickness and envy is stronger than in that suggested above, but unfortunately there
is no clear evidence of this connection.
FEAR, finally, is more strongly connected with whiteness, although the overall
number of phrases was surprisingly small. This connection would seem to be based
on experience people may turn pale when they encounter something frightening
and is conventionalised in many phrases.

16

A similar pattern was found by DAndrade and Egan (1974).

184

Chapter 6

6.4.3 ECTs and mood

Only five ECTs appear together with emotional words in the whole BoE corpus,
which is somewhat surprising. They are scarlet, crimson, puce, magenta and
emerald. Of these the first three are fairly conventionalised and I was in fact
surprised by their low frequency, six, four and seven tokens, respectively. Emerald,
on the other hand, came as a great surprise. However, it would seem that crimson,
scarlet and puce occur much more often without the emotion being spelt out, or
with it mentioned at a considerable distance. Consider the examples below.
(38)

Am too embarrassed to explain that I was talking to my son and mercifully it is

too dark for cab driver to notice that I have turned crimson. (BoE: today)

(39)

Charles Montagu Doughty (1843-1926) would have turned puce at such an

impertinent question. (BoE: indy)

The main function of these terms is to emphasise the strength of emotion. It seems
legitimate to ask why one would describe facial colours with such precision, and
the answer appears to be that to do so suggests intensity. In fact, it seems that these
terms fit the general pattern observed above by which stronger and darker colour
imply stronger emotion. It is interesting to note that puce, which realistically
denotes a nuance that could never occur in a face no matter how much rage was
felt, is on the verge of leaving the colour domain to denote instead the abstract
emotion of anger. In this, it is going through a parallel process to livid, cf. example
(2) at the beginning of this chapter.
The best example of the logic that we can see in this close connection
between colour terms and emotions is the use of emerald in connection with envy.
In the first part of this chapter, we could see that a marked use of emerald tends to
evoke positive feelings, but here we can see it used emphasise envy as in (40).
(40)

Mars is strongly placed, imbuing you with the kind of energy, courage and
sheer sex appeal that make lesser mortals emerald with envy.

Although a jocular phrase, it nicely illustrates that ECTs may adopt this function of
intensifiers even in cases where the main sense of the colour term no longer
concerns colour but something else. This may explain why terms like puce and
magenta are somewhat paradoxically used when their precision has nothing to do
with the actual colour of the object (i.e. the face).

In Search of Extra Meaning: ECTs and Markedness

185

6.5 Summary
The aim of this chapter was to study the effects that sometimes occur when specific
colour terms are used unwarrantedly. In the first part of the chapter, it was
demonstrated that the use of saturation terms such as azure, emerald and golden in
nominal domains where one does not normally make fine distinctions creates extra
meanings. The process was described in terms of markedness, which implied that
peripheral attributes were centralised, or could be viewed as a metonymy involving
the dimension emphasised by the colour term brightness. It was also suggested
that the frequent use of these saturation terms for this purpose may in fact change
their default value so that their main function becomes to create extra meaning.
However, more research is needed in this area.
In the second part of the chapter, a brief study of colour terms and emotion
words demonstrated that specific terms such as puce, crimson and scarlet can be
used to emphasise the strength of emotions. The study also demonstrated that there
appears to exist a model in English by which darker colours imply stronger
emotions of anger and sadness. Having said that, I should acknowledge that this
model is based on one part of the English-speaking population; whether or not the
same patterns of usage occur among speakers of African origin is uncertain. As
demonstrated in Chapter 4, colour and race are closely associated in English
speaking communities and the pattern identified here may also be part of that
picture.

Chapter 7

Figurative Use of Colour Terms

7.1 Introduction
In this final chapter of my dissertation, we have reached a stage where colour terms
no longer refer to colour at all, but to other concepts. In short, this chapter deals
with some aspects of the figurative use of colour terms. Since the area of figurative
use is so vast a comprehensive account would merit a dissertation in its own right
the chapter is restricted to a few themes that have been mentioned earlier in this
dissertation. These themes are the conceptual process of metonymy, type
modification and ECT use. Some attention is also given to the minor networks of
related meanings that seem to be generated once a figurative sense becomes
established.
This early stage of the chapter is a suitable place to define exactly what I
mean by figurative use. It is not always easy to draw the line between figurative
and non-figurative language, and, as we could see in Chapter 6, colour terms
frequently allude to other meanings. However, in the previous chapters, the primary
domain of the predication was the colour domain, and the other meanings could be
viewed as additional associations or attributes. In my definition, a figurative
expression of a colour term is one in which the primary domain of the predication
to which the colour term refers is not the colour domain.
In such a definition hyperbolic use like that dealt with in Section 6.3 cannot
be considered figurative, even though it evokes other meanings and even though
some metonymic models participate in the formation of these meanings.
Furthermore, this definition gives us the tool with which to determine whether
expressions like those given below should be included in an account of figurative
usage of colour terms.
(1)
(2)
(3)

Red herring
Pink elephant
Black sheep

In the account of previous studies below, it can be seen that such expressions have
very often been included, presumably under the assumption that the function of
red/pink/black serves some further purpose other than merely ascribing a colour to
the noun. However, on closer examination, we can note that these three phrases are
quite different from the point of view of figurative analysis. In the case of red
herring, this refers a particular type of smoked herring and the figurative meaning
of this expression goes back to some aspects of how this fish was used. Thus,
technically speaking, red cannot be regarded as having a figurative function in this
expression.1 Rather, it is the whole phrase that produces the figurative meaning; the
1

Cf. the following quotation in the OED: The trailing or dragging of a dead Cat, or Fox, (and in case of
necessity a Red-Herring) three or four miles and then laying the Dogs on the scent. (1686 N. COX Gentl.
187

188

Chapter 7

compositionality is very low indeed. On similar grounds, phrases such as redtops,

blue-collars, white collars and blackshirts are not dealt with here as they have the
same process of formation.
Pink elephant, on the other hand, illustrates a case in which the colour term
functions as a trigger. In this process, pink does not carry any figurative meaning in
itself its meaning has to be construed as literal but this very act forces us to
make a figurative reading of the entire noun phrase. In contrast to red herring, there
is no possible literal interpretation of the phrase, at least under normal
circumstances.2 If someone claims to see pink elephants we know that there is
something wrong with him/her; (s)he might be suffering from delirium or may have
been drinking too heavily or otherwise have upset his/her mind or vision. However,
the phrase is most often used jocularly as in (4) below.
(4)

OZRIC Tentacles got together around the time of the 1983 Stonehenge Festival
driven by a mutual liking for widdly-twiddly guitar sounds, jams that go on
for as long as it takes, and an unshakeable belief that-if you smoke enough
marijuana to floor a herd of pink elephants everything will make perfect
sense. (BoE: brmags)

Black sheep as in (5) below is a complex construction analytically.

(5)

Malena, the black sheep of a noble family, takes many years to prove her own
worth, having grown up in the shadow of her twin sister Reina, who has
qualities everyone loves. (BoE: oznews)

On the one hand, there are black sheep, which are typically less frequent than white
sheep. On the other hand, the meaning of the expression does not follow from any
aspect of the behaviour of these black sheep; apart from their colour, there is
nothing special about them. Instead, the figurative meaning draws on black in the
sense of bad, evil, a sense which occurs in many other phrases containing black.
Consequently, black sheep illustrates a proper instance of figurative usage of a
colour term.3
The present chapter differs methodologically from the previous chapters
where numbers played a part in my argumentation. Since I am mainly interested in
the figurative process and a number of structurally related questions, my
discussions are primarily based on individual examples. The majority of these are
taken from the BoE corpus. There are other, more statistically oriented, studies of
figurative use in relation to corpora: Deignan (1999a) shows that collocational
aspects are helpful in analysing metaphoric meaning. Furthermore, Moon (1998)
provides a useful discussion of how to study fixed expressions and idioms in
language corpora, and also demonstrates that most idioms are quite infrequent, less
than 1 per million tokens.
Recreat. v. (ed. 3) 65). However, this categorisation could perhaps be viewed as overstating the case, since
the figurative use of the phrase is considerably more frequent than references to the original literal entity.
2
The phrase could, of course, refer to a toy.
3
For a full analysis of the expression, sheep has to be analysed metaphorically, of course. However, my
focus is the colour term.

Figurative Use of Colour Terms

189

The chapter is structured in the following manner. First, there is a brief

presentation of previous research in the field of figurative use of colour terms, since
this was not part of the general presentation of colour research in Chapter 1. After
this presentation, two types of metonymical mapping are presented and discussed.
In this presentation I use the figure of conceptual structure introduced in Chapter 2.
Having established these general processes, the remainder of the chapter deals with
variation within a mapping.

7.2 Previous studies

In view of the fact that colour terminology and metaphors have been two subjects
of interest in linguistics in the last few decades, surprisingly little attention has been
given to the combination of the two the figurative use of colour terms. In English,
fewer than ten studies have been devoted to this particular field since 1978. These
articles and books can be seen as forming groups from the point of view of the
methodology used and the type of analysis carried out, and I discuss them as such.
The first three articles to be published in this field were all concerned with
colour metaphors in diverse languages. Derrig (1978), Wescott (1983) and Kikuchi
and Lichtenberk (1983) are similar in the sense that they look for universal
patterns, based on studies of unrelated languages. Consequently, the aim is of a
general descriptive character and little effort is made to provide a coherent theory
to accommodate the results obtained. They are also methodologically similar in as
much as they analyse phrases from various dictionaries.4 There is surprisingly little
agreement among these three studies. Derrig notes that there is a correlation
between figurative usage and the evolutionary sequence suggested by Berlin and
Kay (1969), whereas Kikuchi and Lichtenberk reject this claim. All three studies
find some patterns which they argue are universal, but the only two patterns that all
three agree on is that BLACK tends to have associations linked to malevolence,
evil and WHITE to innocence, goodness.
Ks-Dienes (1983) conducted a similar study, but she was primarily
concerned with the related languages Swedish and English and the typologically
different, but culturally related language, Hungarian. She demonstrated that there is
a great deal of correspondence between the languages. Most of it seemingly based
on shared cultural models. Moreover, Ks-Dienes investigated associations linked
to the colours as such by means of a questionnaire and found that there was some
correlation between the associations and the figurative usage. Furthermore, on the
basis of the questionnaires, she could demonstrate that, in some cases, there was
strong reciprocity between the symbolic use of a colour and the concept it referred
to. For instance, most informants said that black could symbolise death, and the
mention of death made people think of black. The relation between red and love, on
the other hand, did not show this strong reciprocity, to mention one example (p 63).
4

The great risk of confining to the material in this way can be illustrated by the fact that Wescott claims that
green has the metaphorical meaning of pretentious in Swedish. However, the sole phrase that he refers to
att gra sig grn to make oneself green, is at best a very marginal expression in Swedish: none of my
informants had any idea of the meaning of the phrase. Needless to say, there are many more figurative
expressions in Swedish that contain the term green, but with different meanings.

190

Chapter 7

Bennett (1988) and Wyler (1992), mentioned previously (Chapters 1, 3 and

4), also discuss figurative use of colour terms. Concerning colour collocations,
Bennetts work is the most thoroughly performed so far from a descriptive point of
view. The book contains a very large appendix which functions as a dictionary of
English colour collocations. Unfortunately, the theoretical side of Bennetts work is
not as strong as the descriptive and he offers very little guidance concerning the
patterns that can be discerned.
Wyler (1992) discusses figurative usage in only one of the chapters of his
book, and in contrast to the above-mentioned studies he rejects the idea that colour
terms can be used metaphorically. He argues that (p 147):
we must [] ask ourselves whether it is correct to speak of metaphor where colour
names occur in structures in which they do not retain any semantic content [+
COLOUR], or whether it would not be more appropriate to consider colour terms in
structures with figurative meaning as symbols in accordance with [the] definition a
symbol is something which stands for something.

Thus, Wyler comes close to recognising the metonymic character of the use, but
instead advocates a symbolic analysis without clarifying the difference. He claims
that a symbolic analysis would allow us to focus our attention on the structure as a
whole, and the empirical experience represented in, and expressed by, this
structure (Wyler 1992: 147). The problem that Wyler encounters is a result of his
view of semantics, which is based on the notion of semantic features. With an
encyclopaedic view of language, this problem does not arise. Some of Wylers
observations are clever but his theoretical basis prevents him from creating a
comprehensive picture, which is why he has to resort to the symbolic theory,
although he notices that (p 148) colour terms must be associated with human
experience.
In order to be able to explain the symbolic force of a colour term, say, red, he
has to isolate semantic features which are common to this term, and which can be
deduced through associations. Wyler (1992: 155) states that
[c]ommon semantic features which constitute the figurative usage of a specific colour
term have little to do with the underlying hue: they are part of a meaning expressed
with the help of an arbitrary sign or lexeme in a given context.

Thus, he completely overlooks the possibility of a cultural motivation for figurative

uses of a colour term, despite his observation of the link to experience. For red
Wyler identifies the features [+DYNAMIC], [+DIRECTED TOWARDS
/AGAINST] and [POSITIVE]. To my mind, the vagueness of these features
illustrates that this approach represents a kind of cul-de-sac.
In three recent articles the subject of the figurative usage of colour terms has
been addressed on the basis of the assumption that this use could best be
understood in terms of metonymy. This could possibly be viewed as a sign of the
development of the field of conceptual metaphors and metonymies during the 80s
and 90s.
Verosub (1994) restricts her study to the colour red, and she sets out to
demonstrate that (p 27) RED forms the basis of a highly abstract, greatly extended

Figurative Use of Colour Terms

191

metonymic system. Thus she is the first to identify the metonymic character of the
acquisition of figurative meanings. According to Verosub the unifying idea at the
centre of RED is IMPORTANT THINGS ARE RED5. From this general idea she
identifies two universal and highly important substances that are red: blood and
fire. These are meant to form the bases for metonymic systems of signs. It is
interesting to note that these substances are also part of Wierzbickas (1990)
definition of RED. However, although Verosub states explicitly that she is mostly
concerned with language evidence, it would seem that she looks at red primarily
from the point of view of semiotics and only rarely takes linguistic evidence into
consideration.
A more relevant study is Niemeiers (1998). In common with Verosub, she
assumes a metonymic basis for figurative expressions of colour terms and,
furthermore, a possible prototypical centre for the category which may be
universal. The latter point is, however, much more vaguely suggested. Niemeier is
a linguist, and focuses exclusively on the linguistic side of colour. Furthermore,
using a theory of cognitive semantics, she can easily handle the difficulties that
forced Wyler and Verosub into ad hoc solutions. Niemeier (1998: 123) declares
that the conceptual contiguity of metonymy is based on extralinguistic experiences
and connotations and is therefore culture-dependent. Another commendable aspect
of Niemeiers study is that she is not content with using dictionaries in addition,
she uses two corpora, the BNC and the Collins Cobuild CD-Rom on Collocations.
The extensions of the colour terms are presented as radial networks emanating from
a prototypical centre of universal character. Figure 7:1 below illustrates the
network of the term red. It should be noted that Niemeier agrees with Verosub and
Wierzbicka in her identification of the universal features of blood and fire.
Nevertheless, there are a few details for which Niemeier can be criticised.
First, she does not define exactly what she means by metonymies. The term
suggests that there are several different metonymic processes involved at work, but
they are never identified. Instead collections of individual expressions are analysed,
and groups are formed which seem to have similar motivations. However, whether
or not these correspond to different processes is not made clear. The fact that a
word attracts different associations is by no means a sign of difference in process.
A second criticism that can be levelled at Niemeier is that she does not make
explicit what she means by the term radial network,6 and how exactly it refers to
her figures (as in Figure 7:1 below). The term itself is vague and suggests a blend
of Lakoffs (1987: 91-114) radial category and Langackers (1987: 162-166)
network model. However, it is unclear whether Niemeier actually claims that her
model has conceptual status or whether it is to be taken as a typology. If it is
conceptual, one is tempted to see the centre as more basic. However, although
universal tendencies may suggest cognitive basicness, it is difficult to see why this
should be the case here. On the whole, the merits of Niemeiers study clearly
outweigh the question marks.

Verosub writes these phrases in italics, but since she seems to imply that they have some sort of conceptual
status, here I follow the linguistic tradition of writing conceptual metonymies in small caps.
6
This term is also used by Dirven and Verspoor (1998).

192

Chapter 7

WESTERN WORLD
6. colour for ideology
socialist revolution
communism
Red Army
red flag
red peril
Red Square

7b. colour of urgency 8. colour for heated emotions

Red Cross
love, anger, joy
red telephone
to see red
red-hot anger
red-letter day

3a. skin colour/people

Red Indian
red face
red skin

UNIVERSAL

5. colour of
danger
red battle
to catch s.o. red-handed

red
1. colour of naturally red entities
blood
fire
setting sun

2. Attention colour
traffic signs
prostitution
red light
red alert
red-letter day
in the red
red card

ANGLO-BRITISH WORLD
7a. colour of immediacy
red-hot anger
red-hot news

4. Colour of opposition/contrast
red cabbage
red herring
to red-pencil
to be in/out of the red
red-letter day
redneck

colour of bureaucracy
red tape

3b. clothes col/people

red hat
red cap
redcoat

Figure 7:1. Radial network of metonymic extensions of red. (Adapted from

Niemeier 1998: 131)

In another recent article, Verspoor (1998) discusses some aspects of figurative use
of colour terms. She too adheres to the idea of metonymy and she points out that
metonymic use of colour terms is closely linked to our interaction with things and
what we find important in our environment. Thus, she is more interested in the
individual instances of metonymic motivation than the overall patterns that Wyler,
Verosub and Niemeier strive to illustrate.7
The latest and hitherto most ambitious contribution to the study of figurative
use of colour terms is Ohtsukis (2000) A Cognitive Linguistic Study of Colour
Symbolism. In this book the author presents a new model based on symbolic
mappings which are said to be different from metaphoric and metonymic
mappings: symbolic operations reflect an independent, autonomous relationship
between different symbolic meanings, even if they may contain metaphoric and
metonymic elements (p 79). It is Ohtsukis ambition to give a unifying account of
colour symbolism as economically as possible. In his model he tries to merge
cognitive linguistic features with traditional logic. Thus, for each colour term he
identifies a number of prototypical meanings, Initial meanings (p 81), from
which the various symbolic expressions of the colour term can be derived in
accordance with a limited number of logico-conceptual operations (pp 91).
Moreover, Ohtsuki suggests a number of overarching Principles which are
necessary to explain the nature of the operations and variables that are part of the
7

Dobrovolskij and Piirainen (1996, 1998) also treat figurative use of language and colour symbols.
However, since they do this in a larger theoretical context, I have not included their study in the overall
presentation.

Figurative Use of Colour Terms

193

description. Finally, his analysis results in a suggestion for a new categorial

structure, an arbor symbolica (p 108), which is said to incorporate both the
classical category and Lakoffs (1987) radial category.
This short summary cannot do justice to Ohtsukis work; the theoretical
model is quite complex. However, there are a few aspects of his model that I find
less convincing. Let me here briefly point to one problem. To me it is not clear
what the model is supposed to capture. A cognitive model would seem to claim that
it is concerned with the organisation of a speakers mind. However, in his
application of the model to English, Ohtsuki (2000: 122) points out that the
derivations
go beyond particular languages, the following derivations of symbolic meanings of
colour terms should not be regarded as being confined to English alone. Some of the
senses, even though they may not be realized in English, are theoretically necessary
as intermediate steps in deriving other senses.

Thus, Ohtsuki (p 124-126) suggests that black as in black market, i.e. in the sense
of illegal, should be viewed as conceptually derived via a very advanced chain
from the initial meaning darkness:
DARKNESS -> NIGHT - > SLEEP - > DEATH - > HELL - > DEVILS/DEMONS/SATAN - >
EVIL/BADNESS/IMPURITY - > ILLEGALITY

I do not find this particularly psychologically plausible. From the point of view of
my native language Swedish, in which a corresponding phrase, svart marknad
black market, illegal market exists, I can truthfully say that I see no reason at all
why this meaning of svart should be derived via HELL. I can appreciate that black
in the sense of illegal may be motivated in relation to DARKNESS, but I can see
many more intuitive paths to this connection. In short, occasionally Ohtsukis
ambitious system of derivations does not seem psychologically reasonable.
To sum up, to date the study of the figurative use of colour terms has
developed from sketchy descriptive and dictionary-based studies in search of
universals, to methodologically and theoretically well-grounded approaches in
which cultural significance and metonymic mapping have been observed as
important features. However, although metonymy has been established as the
formative process, the details of the process have not been closely scrutinised.
Moreover, no one has given any attention to the close relation between type
modification and figurative use,8 and ECTs have generally been ignored altogether.
These are themes that are treated below. First, however, a specification of my view
of the formative process.

8
Ks-Dienes (1983: 35), however, observes in passing that the colour term in a phrase like white lie has a
classifying function.

194

Chapter 7

7.3 Basic assumption

As the presentation above illustrates, the figurative use of colour terms has been
explained in various ways: metaphor, symbol and metonymy. In my analysis, I
follow Verosubs, Niemeiers and Verspoors insight that metonymy is the basis of
the figurative use of colour terms. This assumption has in practise an axiomatic
character and thus my analysis does not take any other option into consideration.
The axiom can be formulated as follows:
Ultimately, all cases of figurative use of colour terms go back to metonymic
processes in which some experience is linked to a contiguous colour.

This is not to exclude metaphorical mappings from the entire framework, but such
a process can only be of a secondary nature, i.e. there must first have been a
metonymic process. Research in the area of synesthesia may in the end demonstrate
that direct metaphorical mappings in the basic domain are possible, but, in my
opinion, the evidence presented so far is not convincing, and the phenomenon is far
too rare to set its mark on whole speech communities.9
The emphasis on cultural and experiential grounds may be felt to be a case of
stating the obvious, but there is lexicographic evidence of an opposing view. The
quotes below from the OED appear to indicate a belief that the inherent character
of the colour forms the basis for the figurative meaning of the term.
Rhetorically. With reference to the qualities of this colour: bright-hued, brilliant,
splendid, gaudy, gay; (of sin) deep-dyed, grave, heinous.
purple patch, passage, piece, a brilliant or ornate passage in a literary composition.
(The OED: purple, sense A.3.a [emphasis added])
In allusions to the glaring effect of the colour.
1820 Hazlitt Lect. Dram. Lit. 16. The deathblow which had been struck at scarlet vice
and bloated hypocrisy. (The OED: scarlet, sense B.2.a [emphasis added])

I find it very questionable indeed to claim that there are associations of this kind. If
this were so, people would have no trouble interpreting phrases like purple prose,
since they would be self-explanatory. However, my own experience is that native
speakers unaware of this phrase find it almost impossible to figure out the meaning
unless they are presented with some context. To my mind, it seems much more
likely that these meanings arise from intricate patterns of metonymy based on
culturally defined meanings of a colour. I hope to be able to show that this is so as
we proceed. If these motivations are no longer apparent, it may be argued that these
senses have become independent and should be treated as homonyms. I return to
this issue in 7.6.1.

A thorough treatment of synesthesia is presented by Cytowic (1989).

Figurative Use of Colour Terms

195

7.4 Figurative colour expressions and classificatory use

Before we take a look at the metonymical processes involved in the creation of
figurative colour expressions, I would like to point to the resemblances that exist
between figurative use and type modification. The phenomenon of type
modification or as it was also designated, classificatory use, was dealt with in
Chapter 4. It was demonstrated that a few colour terms, notably the Primary Basic
Colour Terms, are frequently used to categorise purposes with respect to the noun
they modify. Thus, the function of the term is not to describe an entity but to
classify it as belonging to a category. A property used for this purpose should thus
be both salient and general, and I argued in Chapter 4 that this might explain the
preference for certain colour terms. Conceptually, this process was described as a
kind of reference point construction in which the term for a salient colour was used
as the vehicle for accessing the category.
It seems that figurative use of colour terms to a certain extent parallels this
description. It is classificatory and thus related to type and not token. What is
sought after is generality and not precision and it is therefore hardly surprising that
the same terms that are frequently used for type modification are also used
figuratively. This becomes particularly apparent if we take into account the most
frequent figurative use of black in the BoE, in reference to people of African origin.
Consider the phrases below.
Black people
Black community
Black pride
Black music
Black English
Except for the first phrase, black has here a figurative meaning suggesting of black
people, thus the primary domain of BLACK is no longer the colour domain. It
should also be noticed that these examples are not really on an equal footing. The
first two figurative expressions, black community and black pride, primarily evoke
a sense of African-American people we cannot talk of community and pride being
modified. However, in the last two examples, this is precisely the case. Black music
and black English are a type of music and a type of English10 and not just anything
performed or spoken by this particular racial and social group. Thus, the phrases
give rise to other meanings than just race. Experience and knowledge (which to
varying degrees may be based on prejudice) are likely to generate more attributes
than just race. Admittedly, the link to African-American people is strong, but in
principle these entities have become independent and, as was argued in Chapter 4,
it can be claimed that they are less compositional than the first types mentioned
above. A white artist can play black music (cf. (6) below), but it might be
questionable whether a white person would ever be categorised as belonging to the
black community.
10

In the recognition of Black English as a dialect in its own right, or even a language, other names for this
dialect have been preferred, e.g. African American Vernacular English (AAVE) (Trudgill 2000: 42) and
Ebonics.

196

Chapter 7

(6)

I think the fact that Elvis first national TV appearance and Brown vs. Board of
Education happened just about the same time is no coincidence. I mean, he was
the first white man who sang black music and made it acceptable for white kids
to listen to it. (BoE:npr)

The whole issue of black and white in reference to race is a very delicate matter,
and the most important reason why this is so is precisely the additional meanings
that arise in type modification. The close relation between social, political,
economic, cultural and racial issues creates a situation in which white and black
have absorbed a great many attributes which may be important in certain contexts.
This example serves to illustrate that the figurative use of colour terms appears to
presuppose a process similar to that of type modification. In the case of figurative
usage, a colour has to co-occur with some object or some other quality to the extent
that a reference to the colour can suggest the object or the other quality.
It seems that there are other similarities between figurative use and classifying
use of colour terms. Quite often, it appears, the function of the figurative
expression is to classify a subtype. Consider examples such as blue movie, purple
prose, yellow journalism, black magic, and white lie. They all refer to subtypes: a
type of movies, a type of prose, a type of journalism etc. However, in contrast to
the examples of type modification discussed in Chapter 4, the colour terms above
refer to abstract notions and not to concrete colours. Below the process of
connecting the colour term and the abstract quality is discussed in some detail.

7.5 Metonymical processes

In this dissertation I have earlier shown that metonymy plays a central role in
colour semantics. One such important role is its presence in the derivation of colour
terms, which appears to be based on a +THING FOR SALIENT PROPERTY+ metonymy.
Below I survey the two types of metonymical processes that are involved in the
creation of the figurative use of colour terms. One of these is the exact opposite of
that mentioned above; a property stands for some entity. The patterns correspond to
two general configurations of metonymy-producing relationships identified by
Radden and Kvecses (1999). They can be described as part-whole and part-part,
respectively. The first type is very productive in the use of colour terms and it is
this process that is usually referred to when discussing figurative colour
expressions, cf. Niemeier (1998). The second type, however, entails an attribute
standing for another attribute and in Radden and Kvecses analysis this usually
takes place at a higher level of organisation that of an event.11 However, I think it
can be used for lower levels and I suggest that the formation of some colour
metonyms can be modelled successfully in this way.

11

This level of organisation would appear to correspond to what Ungerer and Schmid (1996) treat as a
frame.

Figurative Use of Colour Terms

197

7.5.1 Part-whole metonymy

Part-whole metonymy is a simple and very productive type of metonymy. We find
it in an example such as (7) below.
(7)

The tests that the military administers to all recruits or to all potential recruits
before they come in is rather similar to other standardized tests in that blacks
tend to score lower than whites. (BoE: npr)

The mapping relies on a simple part-whole relationship in which a salient attribute

may stand for the entire category. In the context of race, Radden and Kvecses
(1999) suggest the metonymy +DEFINING PROPERTY FOR THE CATEGORY+, but this
is far too strong a suggestion in my opinion. As mentioned above in Section 7.4,
blackness in this sense can be problematised largely in social contexts. As
additional evidence the racial classifications in apartheid South Africa can be
mentioned. Instead I prefer the slightly less absolute +SALIENT PROPERTY FOR THE
CATEGORY+. This kind of metonymy is very close to the classificatory phrases
black people and white people. Using the figure that was introduced in Chapter 2,
we can illustrate the difference as below.

Concrete

Concept:
AFRO-

Abstract

AMERICANS

1
2

Primary
domain:

PEOPLE

10

Figure 7:2. Conceptual structure of the phrase black people.

Attribute list:
1.
2. Black
3.
4.
5.
6.
7.
8.
9.

198

Chapter 7

Concrete

Abstract

6
2

Primary
domain:
PEOPLE

Concept:
AFROAMERICANS

10

Attribute list:
1.
2. Black
3.
4.
5.
6.
7.
8.
9.

Figure 7:3. Conceptual structure of the phrase blacks.

As the figures illustrate, the difference can be said to be that in the nominal phrase
the primary domain is explicitly mentioned together with the typifying attribute,
whereas in the second case the primary domain is maintained although
unmentioned and the salient attribute stands for the whole category. We can
contrast this metonymy with that forming new ECTs, described in Chapter 5.
Whereas that type of metonymy introduces a new primary domain (COLOUR), here
it is retained although indirectly accessed via the mentioning of the salient colour
attribute.
Warren (1999: 123) discusses referential metonyms and noun-noun
compounds, but her observations are also valid here. She notes that what they have
in common is that they involve two referents which are connected by an implicit
link. The main difference is that in the case of compounds, the two referents are
mentioned, whereas in the case of metonymy only the vehicle12 is mentioned.
Moreover, Warren notes that the referring item in the compound also functions as
the head (her example is chair in armchair). She suggests that the implicit referent
in a metonymy serves the same function of head and referring item.
The part-whole type of metonymy is not at all restricted to skin colour but is
also applicable in the context of other domains. As observed by Norrick (1981), it
is conspicuously productive in sports, in which the colour of the jerseys can stand
for the club or the players, especially in football and rugby. There is a notable
pattern in this usage: a newspaper summary of a game could use a phrase like the
Sky Blues even if Coventry City played in their yellow away colours. However, it
12

For a definition, see footnote 22, page 51.

Figurative Use of Colour Terms

199

would not be possible to use the phrase the Yellows in the newspaper.13 In a report
on TV, on the other hand, the Yellows could be an option since the reference would
be obvious. As mentioned in 2.5.2, it seems that a metonym has to reach a certain
degree of conventionalisation for it to be an option outside the immediate situation
(cf. the ham sandwich). In this it is similar to what we observed for type
modification in Chapter 4.
A similar kind of metonymic pattern can be identified in the context of other
types of uniforms. However, very often when it comes to uniforms and other such
clothing, the attribute is not a single colour term but a nominal containing a colour
term and a noun. As was pointed out at the beginning of this chapter, this is
technically not a proper colour metonym, but there are obvious similarities to what
was discussed above and therefore it is briefly discussed. Consider the examples
below.
(8)

In Rwanda the Australians have won widespread praise from other members of
the 5000-member multinational force of blue berets for the pivotal role they
have played, especially in establishing medical services that have helped an
estimated 30,000 people. (BoE: oznews)

(9)

The anti-fascist, seven-strong outfit signed the deal last Friday (February 12), in
Cable Street, East London, site of pitched battles between blackshirts and antifascists in the 1930s. (BoE: ukmags)

The difference between these and the aforementioned pure types is really only
one of degree and definitely not type. These latter examples use noun phrases as
their point of reference, and hence coincide with the overall typical pattern of
metonymy. These compounds are known as exocentric compounds (Bauer 1983)
or sometimes bahuvrihi compounds. Their metonymic character has previously
been observed by Warren (1995). One important reason we find one type in
football and rugby, and the other type of relation when it comes to other uniforms,
would seem to lie in the very nature of football and rugby shirts. The characteristics
of the game friend and foe mixing, the passing of the ball etc suggest that the
players should wear shirts of very distinctive colours. And since all teams wear
shirts which are distinctive it does not make sense to mention this specifically;
hence we get generalisations of the type Sky Blues and Claret and Blues and not
*sky-blue shirts, *claret and blue shirts.
The +SALIENT ATTRIBUTE FOR THE CATEGORY+ also occurs in references to
objects in nature. The most common ones are blue in reference to the sky, as in out
of the blue, which has taken on additional meanings and green in reference to
grassed ground as in village green and a green in golf.
In principle the same kind of metonymy, +SALIENT ATTRIBUTE FOR THE
CATEGORY+, can be found in an expression such as the green peril, which used to
be a name for absinthe, the strong greenish alcoholic liquor popular at the turn of
the century. The quote below, taken from the OED, is an instance of such use.

13

An additional factor is that football teams tend to change their away colours frequently.

200

Chapter 7

(10)

This taxing of the green peril will no doubt be popular. (The OED)

As with black in the sense of Afro-American, we are concerned here with a

metonymic function that is referential. The difference is that the target is reached
through two vehicles (in fact), one of which is concrete, and the other abstract.14
The domain for such a formation is given in Figure 7:4 below.
Concept:
Concrete
1

6
2

Primary
domain:
LIQOUR

ABSINTHE

Abstract

10

Attribute list:
1
2.
3.
4.
5. green
6.
7.
8.
9. dangerous
10.

Figure 7:4. ABSINTHE as the green peril.

Let us now consider another subtype of the part-whole metonymic structure. This
subgroup can be illustrated by the expression green fingers as in (11).
(11)

If youre not blessed with green fingers or the urge to cultivate, you can still
enjoy cooking with herbs. (BoE: oznews)

This frozen phrase represents, in its figurative sense, the meaning

to be unusually successful in making plants grow. (The OED)

What makes this structure more difficult to analyse is the fact that we are
concerned here with two mapping processes which together form one coherent

14

It is interesting to note that the head designates the abstract entity. One would be tempted to think that the
opposite formation would be the one preferred. There is evidence of such phrases as in the potent green in
the following example: I have not felt myself since my generous allowance of the potent green on
Thursday (the OED). Incidentally, this formation with peril was quite productive; consider the red peril
and the yellow peril. These two phrases are still being used but then in reference to sport, and in particular
football, where yellow and red cards are given for offences.

Figurative Use of Colour Terms

201

figurative expression. However, it is possible to identify the vehicles and targets

that operate within the expression.

Green fingers - successful plant growing

Again the colour term represents a simple part-whole metonymy of the type
+SALIENT ATTRIBUTE FOR THE CATEGORY+, where green, the colour of the leaves
of the plant, stands for the whole plant. The case is slightly more complex when we
look at growing and fingers. It seems clear that we are concerned with metonymy
in this case too. However, fingers can hardly be seen as standing for GROWING,
rather it seems to be the case that the whole domain of PLANT GROWING has to be
evoked. One of the more important attributes of plant growing is manipulation by
the human cultivator; we have to water the plants, give them fertiliser, and change
the soil. All this manipulation is done with the help of the hands and therefore it
seems reasonable that the hand, or the fingers (or as the case may be, the thumb
green thumbs) represent this manipulation. It should be noticed that it is indeed the
fingers that are most useful when it comes to plant growing a prototypical scene
is someone putting his/her finger into the soil to check the moisture.15 Thus fingers
stands for manipulation through a metonymic process in the domain of plant
growing. Plant growing represents in fact an event domain in which we find
participants and processes. Fingers for manipulation represents a part-part
metonymy; tentatively, the relationship can be described as +INSTRUMENT FOR
ACTION+. To summarise, we can say that green fingers represent a double
metonymy but not in the same sense as the green peril, as the two vehicles do not
activate the same target.
To sum up this section, we can note that part-whole metonymy is a common
process which may occur in patterns of varying complexity. Colour term
metonymy of this kind can typically be described as +SALIENT ATTRIBUTE FOR THE
CATEGORY+.
7.5.2 Part part metonymy
As was mentioned earlier, Radden and Kvecses (1999) identify another type of
metonymy which seems to be more complex. It may be that this type, part part
metonymy, can explain examples of the kind below.
(12)

15

The British public think green issues are important, but they do not think they
are sufficiently important to warrant the return of single subject campaigners as
Members of Parliament. (BoE: brbooks)

What may be discussed is whether the growing of plants can be seen as anything more than just
manipulation of the circumstances under which the plant is being cultivated.

202

Chapter 7

(13)

This month is a busy one for the travel industry, which will bombard us with
120m brochures. But do we need so many? This is about purple prose, right?
Wrong. Its a question of waste paper. (BoE: times)

(14)

Most Russians avoided overt protests but took risks nevertheless by turning to
the black market for such things as jeans. (BoE: indy)

(15)

Concurrently, Kanes ideals for his newspaper as set out in his Declaration of
Principles are contrasted with the actuality that his newspapers engaged in
yellow journalism under his direction. (BoE: usbooks)

These metonyms are clearly different from the ones we have discussed so far. It is
not the case that the colour term stands for an object whose colour it describes.
Rather it seems that the colour term represents an abstract quality.
Let us consider purple in purple prose more closely, and have a look at its
motivation. As mentioned above (Section 7.3), the OED suggests that this is a
metaphor, where the character of the colour motivates the construction. Another
alternative, closer to my own but still different, is Ohtsukis (2000: 178) suggestion
which is presented below. He is only aware of purple prose in the sense of unduly
pungent and profane words (2000: 179) and this might explain the derivation. As
Figure 7:5 illustrates, he identifies BRUISE/WOUND/SWELLING as the Initial
meaning. This is based on the experience that the human body turns a purplish
colour when bruised. The next step is based on causality, purple may represent
the cause of the bruise hence, an assault. The final step is based on involvement,
where profanations (purple language) typically contain assaults against God
(Ohtsuki 2000: 177)
BRUISE/WOUND/SWELLING

ASSAULT/VIOLENCE

PROFANATION

DEATH/MARTYR

MOURNING/LOSS OF BODILY DESIRES/

SPIRITUALIZATION

FAST/TEMPERANCE/ABSTINENCE

Figure 7:5. Ohtsukis (2000) derivation for a few senses of PURPLE.

Figurative Use of Colour Terms

203

Although elegant, I find this derivation of the phrase less convincing, in particular
in view of phrases such as purple patch and purple passage, which are obviously
related to purple prose. Furthermore, the intermediate step of VIOLENCE/ASSAULT
is not lexicalised in English. It is important to distinguish between the historical
motivation and the synchronic state of affairs, but there may very well be some
connection. In Section 7.6.1, the synchronic picture of purple in its figurative sense
is given, whereas the immediate focus is that of diachrony.
The OED points to the phrase purpureus pannus in Horaces De Arte Poetica,
and it is possible that the phrase purple patch was originally a calque. Nevertheless,
I think we can assume some motivation for its adoption. Purple, in its original
meaning Lat. purpureus used to be a typical attribute of a kingly court. For
instance, the OED lists the first meaning of purple thus:
1. a. Of the distinguishing colour of the dress of emperors, kings, etc.; = L.
purpureus, [] in early use meaning crimson; hence, imperial, royal.

The frozen phrase born in the purple is another piece of evidence for this
association. I suggest therefore that the formative domain of purple as splendid
was that of ROYALTY. In this domain, there are both [purple] as a typical concrete
attribute and [splendour] as an abstract attribute, i.e. a typical association.

Concrete

Abstract

6
2

Primary
domain:
POWER

Concept:
ROYALTY

10

Attribute list:
1.
2.
3.
4.
5.purple
6.
7.
8. splendour
9.
10.

Figure 7:6. Purple as splendour, brilliance in the formative domain of ROYALTY.

The arrow indicates that the attribute [splendour] becomes linked to and associated
directly with purple and can be referred to through the term purple. On the basis of
a phrase like born in the purple, we can also conclude that purple was a salient
feature in the domain of royalty. Splendour was, of course, also linked to the

204

Chapter 7

highest representative on earth. Thus, the relation between the attributes [purple]
and [splendour] is motivated partly because of their relative salience in the domain
as argued above. The metonymy can be described as +SALIENT CONCRETE
ATTRIBUTE STANDS FOR SALIENT ABSTRACT ATTRIBUTE+. Figure 7:6 illustrates this
connection. However, the analysis demands yet another step; to be able to account
for the use of purple with reference to texts, we have to assume a conceptual
metaphor, which can be identified tentatively as +A TEXT IS AN OBJECT FOR
DECORATION+. Then we can dress up a text in purple and create a splendid piece.
This formation is interesting for another reason. It illustrates the process
through which a metonym may lose its original motivation and become a frozen
phrase, a fossil. In this case, two things have happened which have created a
certain amount of opaqueness round a phrase like purple prose. First, the meaning
of purple changed from the original crimson tone, to what we today conceive of as
purple but which is not the colour that kingly robes used to have. Second, and more
importantly, in modern monarchies, it is no longer the habit of the royalty to wear
such purple clothing as a symbol of their status. Thus, [purple] (or, more correctly,
[purpureus]) is no longer an attribute of ROYALTY. Moreover, although, there is
still a fascination with royalty in Great Britain, it seems likely that there are a good
many other attributes which are much more salient than [splendour]. So in fact both
attributes have lost their salience and therefore the connection is lost. In Figure 7:6,
this is illustrated by the arrows directed out of the domain. This development is
possible since neither [purple] nor [splendid] is in any sense a defining attribute of
ROYALTY. This is likely to be a common pattern of the part-part metonymy. The
colour represents a salient and fairly general but non-defining attribute and the
abstract concept is also a general association.
The development of purple illustrates the vulnerability that exists because
of the indirect link between the vehicle and the target. Should the attributes, for
some reason, lose their status in the formative domain, it may become difficult
even impossible to decode the metonym, and it has to be learnt as a lexeme in its
own right. It may even perhaps be valid to talk of homonymy being created. This
seems to be a process which has happened time and again, and it is likely that
phrases like yellow press and blue joke are other examples. Here the original
motivations are no longer present for most language users, and, if unfamiliar with
their meaning, a speaker will have great difficulty deciphering their senses,
although the context tends to provide sufficient help. Moreover, at least in some
cases, it seems that once the motivation has been lost, the metonymic meaning is
only marginally productive. That is, it appears to be preferred only in frozen
phrases, often with a classificatory function. Consider for instance the alliterative
use of purple: purple patch, purple piece and purple prose.
Let me also add a note on the identification of the formative domain. It should
be stressed that the formative domain of ROYALTY is only meant to be suggestive. It
may very well be the case that this domain can be more accurately identified.
However, this should not undermine the force of my argumentation the main point
of which is the presence of a part-part mapping and the lingering of phrases after
the original motivation is lost due to changes in the formative domain.

Figurative Use of Colour Terms

205

There are other types of figurative phrases which may be analysed in the same
way. One such meaning is that of evil, malevolence etc. in connection with
black. Kikuchi and Lichtenberk (1983) could demonstrate in their cross-linguistic
comparison that this connection is one association which seems to be almost
universal in character. Even if this is not the case, the cross-linguistic evidence
suggests that we are concerned with a connection that may be based on some
universal experience. One such experience, which is common, is night. Kikuchi and
Lichtenberk (1983: 31) give the following account of the connection, which I
believe is essentially correct:
Blackness is associated with darkness/night. And darkness/night typically evokes fear
in people: one cannot see what goes on, and one feels threatened. In other words,
darkness/night is perceived as malevolent, and its outstanding characteristic,
blackness, is metaphorically transferred to malevolent events, deeds and characters.

The only reservation I have against the suggested connection is that I insist that the
association is based on metonymy not metaphor. This model seems to be a much
more convincing explanation than Ohtsukis (2000) attempt, mentioned above, to
derive the meaning evil from blackness in no less than seven steps, via SLEEP and
HELL. It is cognitively more economical to view the mapping as taking place in one
formative domain to which the two concepts are contiguous. Accordingly, we have,
in essence, a formative process similar to that for purple in its meaning of
splendid, the difference being that this connection is still open to interpretation
since we all experience night. Graphically, we would have a similar picture to that
of Figure 7:6, without the arrows pointing out of the domain. Similarly, the
connection between green and ecology as in (12) can be modelled as a part-part
metonymy in the formative domain of NATURE.
Let us finally consider the example yellow journalism. According to the OED,
this is
[a] use derived from the appearance in 1895 of a number of the New York World in
which a child in a yellow dress (The Yellow Kid) was the central figure of the
cartoon, this being an experiment in colour-printing designed to attract purchasers.

Apparently, we are here dealing with a very salient individual example, a paragon
(Lakoff 1987: 87), serving the purpose of connecting one concrete attribute
[yellow] with an abstract attribute which can be tentatively identified as
[unscrupulous] or [sensational]. Although the original motivation is lost, the phrase
has been retained in the domain of journalism in a phrase such as the yellow press.
It seems plausible that this sense should remain since it fits the general derogatory
senses that the figurative use of yellow tends to have. Consider, for instance, the
sense cowardice in yellow streak as in (16) below.
(16)

The majority of Members of Parliament are said to be in favour as, I am sure, is

a majority of the public. Yet suddenly a yellow streak has appeared, running
from right-wing MPs through the whips office to spread as a stain down the
Cabinets back. Someone is running scared of something. (BoE: times)

206

Chapter 7

Thus, we may find cases of remotivation, where the original motivation for a sense
development has been lost.
To sum up, in this section we have identified two types of metonymical
processes which are possible explanations for the formation of figurative
expressions with colour terms. Of these types the first, based on a part-whole
mapping appears to be very productive. The second mapping, part-part, can be used
to explain phrases such as black market. However, it was also demonstrated that if
the motivation for such a mapping is culturally determined it may be lost, as in the
cases of purple patch and yellow press, and then the figurative meaning of the
colour term has to be learnt as a separate word, and the phrase as an idiom.

7.6 Post-mapping development: local networks and domain logic

As mentioned above, it is unrealistic to view these metonymical processes as being
activated all the time. On the contrary, we must consider these processes initial
stages in the formation of figurative expressions. Like any phrase, the colour phrase
will become conventionalised if the motivation is viewed as being good enough. At
this stage, the original mapping may be retained as a connection in a schematic
network of senses (cf. Chapter 2, and Langacker 1987); the colour term has become
polysemous. This new element of meaning may, however, develop further, and
motivate new, slightly different, meanings which can be viewed as forming local
networks of meanings or local radial structures. Another type of development that
can be noticed in some domains is that once a metonymic pattern has become
conventionalised, more colour terms can be mapped onto that domain. It is as if the
logic of the colour domain itself had been mapped. Below, I discuss these two
developments briefly on the basis of a few examples.
7.6.1 Local networks and valence structure
When I talk about a local network of meanings, I suggest that within the overall
schematic network which can characterise a concept, there might be nodes that are
intimately linked and from which one may extract a subschema. This process
would parallel the kind of process described by Langacker (1987: 383) in the
context of the category TREE. Another possibility is to analyse the development in
terms of Lakoffs (1987: 204) radial structure.
It was mentioned earlier that Niemeiers (1998) analysis of colour metonyms
appears to be influenced by the idea of radial structure. However, the term she
prefers is radial network. On the whole Niemeiers analysis is quite convincing, but
it is necessary to understand the details of the network or radial structure. In the
present section these details are precisely the focus; I restrict my discussion here to
established substructures of the larger network that the colour term may comprise.
Furthermore, I discuss how we should view these substructures internally, but also
their relation to the entire colour structure.
In the first part I focus my attention on one particular extension of a colour
term, namely that of green in the sense which may be loosely described as

Figurative Use of Colour Terms

207

environmentally friendly or ecological, for a detailed analysis.16 The OED

defines the meaning as
[o]f, pertaining to, or supporting environmentalism (esp. as a political issue); that
belongs to or supports an ecological party; loosely, environmentalist, ecological.

The CIDE suggests a similar definition: (someone who cares a lot about matters)
relating to the protection of the environment.
It is clear that green in connection with the environment originally had a strong
political overtone. The association dates from the early 1970s in West Germany
where political groups emphasising the environment made use of this symbol.17 I
argued above that the connection between ecological and green could possibly be
modelled as a part-part metonymy in the formative domain of NATURE. An
additional important prerequisite for the success of this mapping was no doubt the
tradition in Europe of symbolising political movements in colours. However, my
focus here is the development since then. Although there is still a strong element of
political meaning, it appears that green has taken on the more general meaning of
ecological which does not necessarily have political undertones. Consider the
collection of examples below from the BoE.

16

(17)

He added that as part of the BMW Group they were investing in green cars
such as hydrogen-powered ones. (BoE: times)

(18)

Currently, the EU is also trying to force them to abandon their green policy on
recycling plastics because countries like England want to sell beer in cans.
(BoE: guard)

(19)

The Green party at its annual conference has been discussing how a Green
government would run the economy on environmental principles. The partys
Environment spokesman, Mr David Fitzpatrick, told delegates meeting in the
Midlands city of Wolverhampton that a green economy would be based on
energy efficiency, recycling and sustainable growth. (BoE: bbc)

(20)

It would be very useful to be able to extend the measurement of economic

activity to changes in the environment and the resource base, he said. This
would enable broader assessments to be made of implications of economic
growth, the contribution of sectors particular environmental problems and the
sectorial and national implications of environmental policy measures. Mr
Willis said one way to measure the contribution of environment to the economy
would be an environmentally adjusted or green GDP. (BoE:oznews)

(21)

In the US, the oil industry still has its head in the sand. But other corporate
players see profits in green industry. (BoE: newsci)

Blatn (1996) briefly compares the use of green and Czech zelen in the sense of ecological.
Two such groups were Grne Aktion Zukunft, Green Campaign for the Future, and the grne Listen,
green lists (of ecological election candidates).
17

208

Chapter 7

(22)

But opting for an extra 5-8 per cent on your electricity bill still does not
guarantee that you personally will receive green electricity the companies
simply promise to use the money for research and to buy some green energy.
(BoE: guard)

(23)

The second myth which annoys me is the claim by the Gas Industry that gas is a
green fuel. Gas is a fossil fuel the same as oil and coal and helps towards
global warming. (BoE:bbc)

(24)

because you know consumers increasingly and you know do look to green
products and green companies you know. (BoE: brspok)

(25)

Ted Halstead, of the New America Foundation, a think-tank with backing from
Silicon Valley, points out that green ideas, such as transferring taxes from
payroll to carbon emissions, might actually be supported by many sunrise
knowledge industries to the detriment of what he calls sunset extractive
industries. (BoE: econ)

(26)

Mr Hague will tell a business environment awards ceremony: There is a

fundamental problem with green taxes. They are not what businessmen call
win-win. For if you just impose more and more green taxes, you put up costs,
make Britain less competitive and destroy peoples jobs and livelihoods.
(BoE:indy)

(27)

Indeed, the tax has been promoted not as an environmental measure, but simply
as a way to cut the $300 billion budget deficit. But the effects of the tax would
certainly have been green. (BoE:econ)

As these examples amply demonstrate, the use of green is quite flexible, and in
some cases it may be difficult to decide whether the term refers to ideas linked to
the Green Party or to a more general concept of ecological. It seems reasonable to
interpret this as a diachronic process through which green becomes more and more
flexible. Its great flexibility is also demonstrated by its ability to qualify far more
nouns than the potential synonym ecological. Using Langackerian terminology, it
could be claimed that green (in this sense) can be involved in a great number of
valence relations (cf. Chapter 4). Originally restricted mostly to the domain of
POLITICS, green can now be used to modify distinct objects in other domains
concrete objects, as in green car. This, of course, opens up the possibility of
ambiguity as in the aforementioned phrase. In these nominal constructions green
may modify different conceptual structures of the noun: the surface of the car (the
default reading the car has green colour) or the exhaust aspect of a car (a green
car emits very little exhaust fumes).
Using the concept of local network introduced above, we can see the
expanding use in various new domains as forming nodes in a minute local network
as in Figure 7:7. This network expands as new domains are explored by the
adjectival concept. The difference between these uses of green is very small and is
to a large extent based on the valence relation that exists between the noun and the
adjective due to the nature of the noun or the particular domain to which the noun

Figurative Use of Colour Terms

209

belongs. Strictly speaking, it is doubtful whether we could say that green in green
GDP and green car has the exactly same meaning; whereas environmentally
friendly could be used for green in green car, it works less well in green GDP.

ECOLOGY
green ideas

ENERGY

POLITICS

green energy,
green fuel

Green party,
green issues

ECONOMY
green economy,
green GDP,
green taxes

INDUSTRY
green products,
green car,
green companies

Figure 7:7. The local network of green in the sense of ecological.

However, the meanings are similar enough for an overarching schema to be very
narrow. Furthermore, it would seem that this schema is much more entrenched
/conventionalised than the distinct meanings for each collocation. Thus it is
reasonable to say that green in this sense exhibits vagueness, rather than polysemy,
cf. Tuggy (1993) and Chapter 2, and especially Figure 2:9.
In a similar fashion we can assume that other metonyms may expand and drift
in meaning. A good example of this is the change of meaning of purple, in phrases
like purple prose and purple patch. As could be seen in the definition of these
phrases quoted from the OED, the traditional sense was a brilliant or ornate
passage in a literary composition. Although the meaning appears to have been
positive originally, it is very clear from the material in the BoE that there is quite
frequently a disapproving element involved. Purple passages and instances of
purple prose are exaggerated and kitsch. The quotations below show both senses.
(28)

These maps accompany equally gorgeous photographs and the kind of purple
prose that has made Oz Clarke such a popular contributor to the BBCs Food
And Drink programme. (BoE: today)

(29)

Reillys book has its failings; he is not a natural writer and has a tendency to
launch into purple prose that should have been curbed by his editor, but his is
an important book. (BoE: times)

As ideals change, what is regarded as eloquent by people in one generation may be

thought of as pretentious and tasteless by the next generation. However, the local

210

Chapter 7

network of purple in this sense has developed further, and there is also the
possibility of using purple in reference to writing or speaking that contains a lot of
offensive or taboo words (The CIDE), which is the sense Ohtsuki (2000) derives
from bruising. It could perhaps be argued that taboo words are used to produce
adorn language. Here we clearly have a case of category extension that is greater
than that described above in the case of green. However, I could not find any clear
example of this use in the BoE, so the construction may still be marginal. In the
OED there are many more examples of purple in the sense of splendid in the
quotes, but it has been difficult to find instances of this usage in the BoE. It would
seem that the use is mostly confined to the domain of LANGUAGE. There is one
notable exception, however; purple patch is quite often used with reference to the
domain of sport, in which it designates a moment of brilliance as in (30) below. In
fact, of all the derived senses in this local network, this is the one occurs most
frequently in the BoE.
(30)

But he is a talented swing bowler who ended last summer with his county in
splendid style. It was probably that purple patch which brought his potential to
the attention of the England selectors. (BoE: indy)

This sense of brilliance is, of course, slightly different from that which refers to
style in literature so here we would be concerned with another node in the local
network.
The overall confinement of this network to one particular domain, and the
relatively low frequency of this usage in the BoE, could probably be partially
explained by the fact that the original motivation for this sense of purple has been
lost. Thus, the phrase may be difficult to interpret it is almost like a fossil in the
language. The same can be said about phrases such as yellow in the sense of
unscrupulous and blue in the sense of indecent, as in yellow press and blue
movie. It could be argued that in so far as these local networks are distant from
other derived meanings of the colour term, they are probably more like
homonymous than polysemous senses. Again, consider Tuggys (1993) continuum
of homonymy and polysemy (cf. Figure 2:9). Where the motivation is still clear, we
can talk about polysemy, but once it has been lost, then it seems that the
overarching schema is likely to be very abstract and vague and, consequently, it is
more reasonable to treat the senses as homonyms.
7.6.2 Domain logic
Another phenomenon that may occur once a metonymic mapping has established
itself and reached a certain degree of conventionalisation is something that could be
described as the development of domain logic. This can be exemplified by the use
of pink in the sentence below.
(31)

Im a Labour Party member, a second member of the delegation said, but Id

happily support the Liberal Democrats if they had the best chance of winning in
my constituency. Should they? Should you, if you consider yourself left-wing
or even pale pink. (BoE: indy)

Figurative Use of Colour Terms

211

We have no difficulty at all of understanding this use of pink. What is interesting is

that there is no direct mapping here between the concepts of (a smaller degree of)
SOCIALISM and PINK, but instead the motivation is based on some notion of domain
logic. Intriguingly, this application could possibly be viewed as a kind of
metaphoric mapping. A perceived similarity between the two domains IDEOLOGY
and COLOUR makes it possible to map the concrete domain onto the abstract one.
The similarity appears to reside in the boundless character of the two domains.
Exactly like the colour domain, the political domain is characterised by salient
prototypes and fuzzy boundaries; just as there is a salient red colour, there are
hardcore communists, but there are also less salient nuances that belong to the same
category. However, as this reasoning demonstrates, there is no full mapping
between the two domains. On closer inspection, we can observe that what is
mapped is, in fact, the area of one basic colour; a colour which is metonymically
linked with one particular ideology. Furthermore, there is only one dimension from
the colour domain that actually participates in the mapping, namely lightness. Thus,
a light nuance of the colour suggests a low commitment to the political idea. In
English, this use appears to be restricted to the colour red whereas, in Swedish it is
possible to characterise a lukewarm supporter of the right-wing party as ljusbl,
light-blue.18 On the other hand, rosa, pink is not (or very seldom) used to
designate a supporter of moderate socialism.19
The same kind of application of a colour term is very frequent in the case of
grey, which then designates a middle area of some domain. For instance, in the
domain of economy, the phrase grey may refer to what is a borderline case between
what is legal and illegal. Black in reference to illegal is quite common as in black
market and black economy, whereas white does not occur at all. This can probably
be explained by the fact that it is the natural vantage point for categorising the other
activities. It should be noted that, strictly speaking, from the point of view of the
law, no true fuzzy border exists between what is legal and illegal. However,
morally and ethically such vagueness may exist. According to the OED, the
meaning of grey market can be defined as
[a]ny unofficial or unorthodox trading (usu. of items which are scarce or in great
demand) which, although legal, may be considered unethical; also loosely, = grey
economy s.v. GREY a. 8 a. Cf. *PARALLEL importing. (OED online: grey market,)

Thus, by using this colour term, speakers are able to create a zone of fuzziness
where non-figurative language would suggest strict divisions. What has happened
is that the structure of the domain of lightness and its continuum is mapped onto the
legal domain. In this sense the figurative expression is much more practical and
efficient than a non-figurative one. Grey market can also have another meaning of
18

Askedal (1987) discusses political colours in Norwegian. He observes that whereas mrkerd and
mrkebl (dark red and dark blue) are regularly used with ideological implication, this does not seem to
be the case with grnn (green). This is true of Swedish too, and presumably this is due to the fact that the
green ideology has not yet been developed far enough to suggest shades of commitment.
19
It would be tempting to suggest that this could be due to Swedish rosa containing more elements of purple
than English pink. However, Sivik and Taft (1994) demonstrate that Swedish rosa is centred in the unique
red hue plane whereas pink also has some elements of blue at the centre.

212

Chapter 7

[t]rading in a security prior to its official quotation on the Stock Exchange (OED
Online). This is actually the most frequent meaning in the BoE, but this is probably
due to the character of the corpus.
In this section we have seen that the conventionalisation of a colour term in
reference to some domain may open up the possibility of a metaphor-type mapping
in which domain features from the colour domain are mapped onto the abstract
concept. In doing so it is possible to talk about indefinable phenomena in a very
economical way. However, although this second stage may be viewed as
metaphorical, it is important to stress that the initial mapping is based on
metonymy.

7.7 Figurative use of ECTs

It has previously been observed by researchers (e.g. Bennett 1988) that ECTs and
some Secondary BCTs, notably orange, very seldom take on figurative meanings
and only rarely occur in figurative expressions. In the light of my analysis here,
where I have indicated the similarity between type modification and figurative use,
this is not surprising. It seems that the low salience of both the colour concept and
the term makes an ECT a less likely candidate for figurative use. There are a few
instances of figurative use of ECTs, but they are very rare. Below, I present the few
that I have found, including material outside the BoE. Apart from just mentioning
the instances, I point to possible motivations for the figurative development. In the
BoE corpus, I have been able to identify four ECTs which may be used
figuratively. They are scarlet, rose, lavender and puce. The development of puce
was mentioned in Chapters 5 and 6 and it is therefore not dealt with in any detail
here. To these ECTs, we could also add the terms silver, gold and golden, which
quite often occur figuratively. However, it seems that most of these meanings draw
on the metal meaning of these terms and for that reason they are ignored. In
addition to the above-mentioned four terms, I am aware of a figurative use of beige,
and although this use could not be found in the BoE, I take it into account in my
discussion below.
Of these ECTs, it seems that scarlet has the most conventionalised figurative
meaning, although it is mainly restricted to the phrase scarlet woman, a
notoriously immoral woman; a prostitute. (The OED) The origin of this meaning
can be traced back to the Bible, Rev. xvii. 1-5:
1. And there came one of the seven angels which had the seven vials, and talked with
me, saying unto me, Come hither; I will shew unto thee the judgment of the great
whore that sitteth upon many waters:
2.With whom the kings of the earth have committed fornication, and the inhabitants
of the earth have been made drunk with the wine of her fornication.
3.So he carried me away in the spirit into the wilderness: and I saw a woman sit upon
a scarlet coloured beast, full of names of blasphemy, having seven heads and ten
horns.
4.And the woman was arrayed in purple and scarlet colour, and decked with gold
and precious stones and pearls, having a golden cup in her hand full of abominations
and filthiness of her fornication:

Figurative Use of Colour Terms

213

5.And upon her forehead [was] a name written, MYSTERY, BABYLON THE
GREAT, THE MOTHER OF HARLOTS AND ABOMINATIONS OF THE
EARTH.
[capitals original, bold emphasis added]

As in the case of yellow press mentioned above, here we are clearly concerned with
a paragon (Lakoff 1987: 87), a very salient individual example in which the
mention of scarlet colour and the abstract quality of sin are contiguous. Moreover,
it is quite clear that the use of this colour here is not accidental but is closely linked
to the fact that clothes dyed with scarlet were very exclusive.20 Too much fondness
for worldly and exclusive things was seen as a deadly sin in earlier ages.
Nowadays, the use of scarlet in this way probably draws on another paragon,
Nathaniel Hawthornes novel The Scarlet Letter. In this novel, a woman is forced
to wear the scarlet letter A on her dress since she refuses to name the father of her
illegitimate child. The important point is that the figurative meaning has to draw on
cultural meanings (the domain of CLOTHING and the non-defining attribute of
exclusivity) and cannot be derived from the inherent qualities of the colour itself.
Below are two examples of scarlet woman from the BoE.
(32)

Id betrayed him by allowing her to leave. He called our mother a whore and
a scarlet woman. (BoE: brmags)

(33)

She seemed to accept her role as the English aristocracys scarlet woman - in
fact, she played it to the hilt, relishing her own shamelessness. (BoE: brmags)

Ironically, scarlet could previously refer to the gown or robe of a doctor of

divinity or law, a judge, a cardinal, etc.; also, the scarlet coat worn in the hunting
field (The OED). Furthermore, it could refer to the rank that followed with the
robe. So, actually the same original culturally defined property, namely that of
being the colour of an exclusive dye, led to two very different figurative meanings.
This second meaning of scarlet may, incidentally, be compared with that of purple
discussed above.
Lavender and beige have probably also acquired their figurative meanings via
the domain of CLOTHING. Lavender is associated with homosexuality and there are
a few instances of this use in the BoE, as in (34) below.
(34)

As Madsen is eager to reveal, lavender marriage, a marital contract between

a lesbian woman and a gay manwas common in Hollywood. It afforded deep
cover and decency and the studios often forced it where scandal loomed.

According to the Bloomsbury Dictionary of Contemporary Slang, lavender in the

sense of homosexual has developed as

20

The choice of scarlet is almost accidental, as the whore is also dressed in purple. In fact this illustrates the
original meaning of purple as purpureus a nuance almost identical to that of scarlet and based on a very
exclusive dye. Consider also the colour of the robe that Christ wore during his crucifixion; Matt. xxvii. 28
describes it as scarlet, whereas Mark. xv. 17 and John. xix. 2 describe it as purple.

214

Chapter 7

[a] facetious term appropriated from the heterosexual mockery for use by the gay
community itself; the colour and the scent of lavender being thought as
quintessentially feminine and old-maidish, respectively.

Thus, we would be dealing with a parallel to the development of pink, which is the
more established and much more linguistically productive colour reference to
homosexuality. This formation could tentatively be viewed as a type of part-whole
metonymy, similar to that of uniforms, which has been taken one step further. The
colour of the uniform may stand for the people. Later, some attribute associated
with these people has been associated with the colour and the colour term. An
important difference is that we are not talking about a formal uniform here, but
some (prejudiced) idea of typical clothing.
Beige may be used with the figurative meaning of uninteresting,
unimaginative or boring as in (35).
(35)

Maybe thats what was wrong with your marriagetoo beige.21

From the quotes in the Random House Dictionary of American Slang, it appears
that this use of beige developed fairly recently, in the 1980s, and the formation
could be explained in the same way as lavender. Beige occurs most often as a
textile colour and since the colour beige has little hue and saturation, clothes of this
colour may appear boring as fashion changes, and so too the people who wear the
colour. The important point is that the mapping must have taken place in some
nominal domain and not directly in the colour domain we experience and contrast
colours as attributes in nominal domains, not directly in the colour domain. This
use of beige is essentially parallel to the more established use of grey. It seems that
the figurative use of beige was originally an American phenomenon,22 which may
explain why it could not be found in the BoE. This figurative use of beige is
currently becoming frequent in Swedish. Below is one example.
(36)

Att ha ett namn som blir obsolet om mindre n tv r r en aning beige, men det
r i mitt tycke ocks det enda beigea med tidskriften.
[To have a name which will be obsolete in less than two years is somewhat
beige, but it is in my opinion the only beige thing about the magazine.]
(Sprkbanken: SVD2:0903 LED)

Incidentally, it is possible to trace a previous parallel development to that of beige

in this sense. Drab originally referred to a kind of cloth and later to the colour of
the cloth. Finally it developed a figurative meaning of boring, dull, which is the
one mentioned in dictionaries. In the BoE, both the colour sense and the figurative
sense occur. Below is an example of the latter, where presumably beige or grey
could also be used.

21

The quote is taken from the Random House Dictionary of American Slang.
The NTCs Dictionary of American Slang and Colloquial Expressions suggests that it is a Californian
phenomenon.
22

Figurative Use of Colour Terms

(37)

215

The latest Royal break-up proves that our rulers have nothing to offer their
subjects, millions of whom endure drab but happy married lives. (BoE: today)

Although it is my belief that these meanings are based originally on metonymy, it

seems that the connection between colour and certain moods has become almost
metaphorical. It would seem that it is in particular the dimensions of saturation and
lightness that play a role here: colours of very low saturation such as grey, beige
and drab are considered boring, whereas high saturation (brightness cf. Chapter
6) is considered positive. Consider the use of colourful person in (38) below.
(38)

She was an outgoing, beautiful and colourful person. She spoke her mind and
we loved her for that. (BoE: oznews)

Similarly light can stand for what is positive and dark for what is negative.
However, this applies mainly to white and black.
Finally there is rose, which may be associated with cheerful optimism and
indicate that a persons views are nave and unduly optimistic (The OED); this is
particularly the case in the phrases rose-coloured and rose-tinted glasses or
spectacles. Dobrovolskij and Piirainen (1998: 21) point out that this association
has cross-linguistic parallels which they suggest that this may be due to the colour
pink being frequently associated with pleasant things such as female babies. True
as this may appear, it seems equally possible that the meaning is derived from the
entity rose. What supports this latter interpretation is that it is not possible to use
pink in the same sense. Furthermore, in all the cases that Dobrovolskij and
Piirainen present, the term is that which can also refer to a rose. This is the case in
Swedish where skrt, pink, does not give rise to the association, but rosenrtt,
rosy-red, can. It can also be noted that rosy may carry the same associations.

7.8 Summary
In this chapter I have discussed the process that leads to the figurative use of colour
terms, and I have identified this process as one of metonymy. It was suggested that
there are two types of configurations that can explain the types of figurative
meanings that tend to develop. There is the part-whole relation in which the colour
term refers to some object and can stand for that object. This relation appears to be
very productive and particularly in sport. The other configuration can be described
as part-part and entails a colour term being associated with some object or domain
able to stand for some abstract quality which is also linked to the entity or domain.
It was shown that if these attributes lose their salience in the formative domain, the
link becomes opaque which may result in the usage being restricted to a few frozen
phrases. Furthermore, it was argued that the identification of salient colours and
colour terms for the mapping is essentially parallel to that for type modification
discussed in Chapter 4. This probably explains why the salient Primary Basic terms
are those that predominate. Very few ECTs have figurative meanings and those that
do, either seem to draw on one nominal domain in particular, or appear to acquire
the figurative meaning from the original entity rather than the derived colour, as in
the case of rose.

Concluding Remarks
Commenting on the Berlin and Kay tradition of colour semantics research, some
scholars, notably Lucy (1997) and Lyons (1999), have criticised the method of
mapping colour terms onto colour charts, a very common procedure in the study of
colour semantics. It is claimed that the method leads the researcher to restrict his
investigation, and thus to underestimate the complexity of the field of colour
semantics. Lyons (1999) points out that the above-mentioned method favours
context independent reference (i.e. reference directly in the colour domain) rather
than contextual use (i.e. description of the colour of entities). The present thesis has
been written with this remark in mind. One aim was to explore a new method of
approaching colour semantics to consider English colour terms in context. For
this purpose I used an extensive computerised text corpus, the Bank of English.
Although the corpus represents the type of language English speakers are exposed
to rather than what they produce, it is hoped that the study of this corpus will
contribute to the understanding of colour categories. The material was approached
on three levels: general corpus statistics, the categories of the nouns described by
the colour terms, and individual examples.
Because of the nature of text corpora, my focus was reference outside the
normal area of designation of a colour term, i.e. extensions from the prototype.
Using the theory of cognitive linguistics, I tried to analyse and describe the
processes involved. Some theoretical constructs proved to be particularly useful in
accounting for English colour term usage: notions such as entrenchment, plane,
reference-point, vantage, salience, markedness and metonymy.
In this closing chapter, in addition to summarising some of the results, I
would like to point to some general conclusions that may be suggested on the basis
of what has been described here. These suggestions concern the existence of a
colour domain, and the structure of the colour term category. I also mention some
possible future studies where the results obtained here could be refined and
complemented by other methods.
Some critics of Berlin and Kays (1969) idea of Basic Colour Terms reject the
existence of a colour domain on the grounds that it cannot be assumed to be
universal (cf. Section 1.2). In many languages, it is argued, one cannot reduce other
associated meanings to mere connotations, and thus mapping procedures on a
colour chart exclude important aspects of these terms. A frequently quoted example
of the lack of a colour domain is Hanuno colour terms as described by Conklin
(1964). Conklin shows that the terms malatuy (GREEN) and marara (RED) in
addition to colour evoke the meanings of succulence and desiccation,
respectively. This division (or opposition) is mainly significant in the domains of
plants, and Conklin (1964: 191) mentions that [a] shiny, wet, brown-colored
section of newly cut bamboo is malatuy. Thus, although the colour is closer to
marara, the bamboo is still described as malatuy because of the succulence.
However, it is doubtful whether this can be said to disqualify the notion of a colour
domain. Consider how a dark-skinned European can be referred to as white and a
217

218

Concluding Remarks

light-skinned African as black in the context of race. As I have tried to show in this
thesis, it is quite often the case that other concepts or domains are associated with
colour. With an encyclopaedic view of meaning, this is not a problem it is reality.
We can assume the existence of a basic colour domain against which colour terms
are primarily defined while other meanings can nevertheless be successfully
accounted for.
In this thesis, I have demonstrated that colour terms may take on a classifying
function (classifying a subtype) and in doing so they can refer to nuances which
might well lie outside their normal area of designation (Chapter 4). As an example
we can mention that white wine is usually far from white. Two powerful theoretical
constructs for my analysis of the usage of colour terms in classifying function are
vantage and reference point. The vantage represents the perspective of the
conceptualiser and the reference point is a salient landmark through which the
target, the classified entity, is accessed. The choice of reference point is largely
based on the vantage point of the conceptualiser and since it is a matter of
characterising a type, generality (and thus, salience) is striven for. It seems that the
use of malatuy in Hanuno in classifying bamboo (cf. above) could be modelled in
the same way as the use of blue in e.g. blue oak and white in e.g. white people. In
so far as attributes co-occur, they may be associated with one another, as in
Hanuno. The importance of salience is further evidenced by the fact that the
classifying function is primarily restricted to a small number of terms, those
referred to as Primary BCTs (i.e. black, white, red, green, yellow and blue).
The co-occurrence of colour and a concept or another attribute is taken one
step further in figurative usage. It has been argued in this thesis that figurative
expressions of colour terms are based on metonymy (Chapter 7). Langacker (1999)
claims that reference-point constructions and metonymy are related in character,
and that metonymy can be viewed as a special case of the former. Moreover,
figurative expressions show another similarity to classifying usage in that they
frequently have a classifying function, designating a subtype of some sort (purple
prose, yellow press). In cases such as whites white people, the closeness between
the two types of usage becomes very apparent. This may also explain why
figurative usage appears to be restricted to only a few colour terms.
Two different types of metonymy were identified as participating in the
formation of figurative usage of colour terms: a part-whole metonymy (+SALIENT
ATTRIBUTE FOR OBJECT+) and a part-part metonymy (+SALIENT CONCRETE
ATTRIBUTE FOR SALIENT ABSTRACT ATTRIBUTE+). The first type is very common
and usually transparent as in the case of skin colour whites. However, if the
second type is generated from cultural patterns and these are changed, then the
original motivation for the mapping may be lost, as in purple prose and yellow
journalism. In such cases it may be valid to talk about the creation of homonymy,
since the meaning of purple in purple prose has to be learnt quite separately from
the colour sense. Generally, figurative senses of colour seem to be culturally
motivated, which explains why previous researchers have found few universal
patterns.
The third type of usage evoking meanings outside the colour domain has been
analysed in this thesis in terms of markedness. It was argued that the precision of

Concluding Remarks

219

ECTs may in some nominal domains be interpreted as a flouting of Grices maxim

of Quantity, and hence lead the hearer to look for additional meanings (Chapter 6).
My case study of saturation terms showed that these additional meanings may
derive from the nature of the added feature (in this case brightness), or the
transparency of the term (as in golden). However, it is also clear that nontransparent terms are able to yield extra meanings (sienna skin). In conjunction
with expressions of emotions, specific terms referring to facial colour appear to
suggest an intensification of these emotions (scarlet with embarrassment).
Transparency is a common feature among non-basic colour terms, elaborate
colour terms (ECTs), and is the result of their metonymical derivation from entities.
This metonymy (+OBJECT FOR SALIENT ATTRIBUTE+) is the exact opposite to one of
the processes creating figurative meanings of colour terms. In describing the
semantics of ECTs, entrenchment (within the speaker) and conventionalisation
(within the speech community) are important dimensions. Some colour terms are
closely linked to certain nominal domains such as VEGETATION and COSMETICS and
are thus poorly conventionalised, whereas other terms are used more generally and
show a higher degree of conventionalisation. In nominal domains where ECTs
commonly occur, marked readings do not arise since the specificity is to be
expected and is not interpreted as a flouting of the maxim of Quantity.
On the basis of the observation made about usage patterns and frequencies of
occurrence, it is possible to give a usage-based characterisation of the two
categories of colour terms discussed in this thesis Basic Colour Terms and
Elaborate Colour Terms.
A BCT is a colour term which is centred on a salient point in the colour
domain. In my description of English colour term usage in the Bank of English, I
have identified a number of features that seem to correlate with basicness. English
BCTs have the following linguistic characteristics to a greater or lesser degree:
They may be frequently used for type modification (i.e. classificatory
use of subtypes), e.g. blue oak, white coffee.
They occur with the derivational suffix ish, e.g. reddish, greenish.
They occur with the inflectional forms er and est e.g. blacker.
They have high overall frequency.
They occur in colour-colour combinations with fixed positions and
without implying a hyponymic relation, e.g. blue-green.
They have figurative meanings, e.g. green products (ecological), to be
green (inexperienced).
On the whole, the terms usually defined as BCTs fit this category description, but
the usage patterns do not yield a clear-cut picture of basicness. The Primary BCTs
(black, white, red, green, yellow and blue) form a more homogeneous group than

220

Concluding Remarks

the group of BCTs as a whole if we only look at these features. The BCT orange,
for example, has few of these characteristics. There are also some ECTs which
show degrees of basicness; in particular, silver and gold(en).
Elaborate Colour Terms also appear to have some typical features. On the
basis of what was described in Chapters 3, 5 and 6, we can identify the following
characteristics as being typical of ECTs:
They are derived from some entity, and frequently the entity can be
identified, e.g. turquoise, cream.
They are used almost solely for descriptive purposes (token plane).
They may generate extra meanings through a violation of Grices maxim
of Quantity in some nominal domains, e.g. emerald eyes, azure waters.
They occur mostly in written texts, typically magazines and books.
They are frequently inconsistently defined in dictionaries.
They may occur with the derivational suffix y, e.g. rosy, rusty.
Among the BCTs, orange shares quite a few of these features with the ECTs. From
this linguistic evidence, it is tempting to suggest that the category of colour terms
should be viewed as a radial category (Lakoff 1987). The centre of the category
comprises the Primary Basic Colour Terms and the periphery the ECTs. The most
peripheral members of the colour term category would be nonce terms and colour
terms restricted to certain nominal domains. From a cognitive point of view, this
category would have to be considered dynamic, meaning that a colour term may
move towards the centre and out from the centre, depending on its entrenchment
within a speaker. On a higher level, i.e. the speech community, the dynamics of
such a change could possibly be described in a diachronic study, but also
synchronically as in Forbes (1979, 1986) studies of brun and marron in French.
As a colour term becomes increasingly entrenched or conventionalised it
moves towards the centre of the category. As described in Chapter 5, increased
entrenchment also entails greater freedom from the entity of derivation. The more
entrenched a term becomes the more likely it is to acquire the characteristics listed
for basic terms. A colour term may also move out of the colour term category if it
is consistently used in one nominal domain and in close connection with other
attributes. This appears to be the case with English puce (Chapters 5 and 6).
To sum up, this study of English colour terms in context demonstrates that
with the assumption of a basic domain of colour and important cognitive notions
such as domains, attributes, reference point constructions, metonymy, and
entrenchment described above, it is possible to account for many facets of English
colour term usage. It may be possible to use these models to explain patterns in
languages which have been regarded as being very different. It is hoped that the

Concluding Remarks

221

results obtained here will encourage other researchers to explore the corpus based
method further and include other languages.

Prospects for further research

The fact that there have been relatively few approaches to colour semantics based
on texts means that there are numerous possibilities for further research. Below I
have indicated a few of these.
Chapter 4 demonstrated that type modification might extend a term
outside its normal area of designation. A comparison between this
type of extension and MacLaurys (1997) step-by-step mapping
technique in the colour chart would probably be very fruitful.
It may be potentially rewarding to review the previous studies of
Lyons, Conklin and Forbes to determine whether some of the
observed usage patterns may be part of the phenomena that arise
from type modification (classifying function).
Closer textual examination of the identifying and descriptive
function of colour terms is necessary. I have here assumed that the
identifying function is mainly restricted to BCTs but a detailed study
might bear out this assumption.
The borderline between basic and non-basic terms should be
investigated more carefully, preferably using a number of different
tools texts as well as colour chips. Interesting terms to study would
be purple, orange, beige, turquoise, olive, and crimson and possibly
a few more.
The study of ECTs in Chapter 5 could also be followed up by other
methods. Rather than testing colour naming with the help of colour
chips one could do it with the help of a nominal context of clothes or
flowers or cars. This could more clearly establish whether some
ECTs are firmly anchored in certain domains only.
As these suggestions indicate, great advances may be made by combining
established anthropological and psychological methods with text studies.

Bibliography
AARTS, J. and J. CALBERT 1979. Metaphor and Non-Metaphor. The Semantics of Adjective-Noun
Combinations. Tbingen: Niemeyer.
ABELIN, . 1989. Frgordssammansttningar. Gteborg: Gothenburg Papers in Theoretical
Linguistics, S 14.
ADAMS, F.M. and C.E. OSGOOD 1973. A Cross-Cultural Study of the Affective Meanings of
Color. Journal of Cross-Cultural Psychology, vol. 4, no. 2, 135-156.
AITCHISON, J. 1994. Words in the Mind: An Introduction to the Mental Lexicon. 2nd ed. Oxford:
Basil Blackwell
ANDREWS, E. 1990. Markedness Theory: The Union of Asymmetry and Semiosis in Language.
London: Duke University Press.
ANDRICK, G.R. and H. TAGER-FLUSBERG 1986. The Acquisition of Colour Terms. Journal of
Child Language, vol. 13, 119-134.
ARCHIBALD, J. 1989. A Lexical Model of Color Space. In Corrigan, R., F. Eckman and M. Noonan
(eds.), Linguistic Categorization. Amsterdam: John Benjamins, 31-53.
ASKEDAL, J. O. 1987. Om Politiske Farger i Norsk. Meel og Minne, 196-207.
ASTON, G. and L. BURNARD 1998. The BNC Handbook : Exploring the British National Corpus
with SARA. Edinburgh: Edinburgh University Press.
ATWOOD, M. 1994. The Robber Bride. London: Virago Press Limited. [First published by
Bloomsbury Publishing Ltd in 1993.]
AU, T.K. and D.E. LAFRAMBOISE 1990. Acquiring Color Names via Linguistic Contrast: The
Influence of Contrasting Terms. Child Development, vol. 61, no. 6, 1808-1823.
AVERILL, E. 1980. Why Are Colour Terms Primarily Used as Adjectives? The Philosophical
Quarterly, vol. 30, no. 118, 19-33.
BALDWIN, D.A. 1989. Priorities in Childrens Expectations about Object Label Reference: Form
over Color. Child Development, vol. 60, no. 6, 1291-1306.
BALL, C.N. 1994. Automated Text Analysis: Cautionary Tales. Literary and Linguistic
Computing, vol. 9, no.4, 295-302.
BARBER, C. 1997. Early Modern English. 2nd edition. Edinburgh University Press.
BARCELONA, A. (ed.) 2000. Metaphor and Metonymy at the Crossroads: A Cognitive Perspective.
Berlin: Mouton de Gruyter.
BARLEY, N.F. 1974. Old English Colour Classification: Where Do Matters Stand? Anglo-Saxon
England, no. 3, 15-28.

223

224

Bibliography

BARRATT, L.B. and M. KONTRA 1996. Matching Hungarian and English Color Terms.
International Journal of Lexicography, vol. 9, no. 2, 102-117.
BARSALOU, L.W. 1983. Ad hoc Categories. Memory & Cognition, vol. 11, no. 3, 211-227.
1985. Ideals, Central Tendency, and Frequency of Instantiation as Determinants of
Graded Structure in Categories. Journal of Experimental Psychology: Learning, Memory,
and Cognition, vol. 11, no. 4, 629-654.
1992. Frames, Concepts, and Conceptual Fields. In Lehrer, A. and E. Kittay (eds.),
21-74.
BATTIG, W.F. and W.E. MONTAGUE 1969. Category Norms for Verbal Items in 65 Categories: a
Replication and Extension of the Connecticut Category Norms. Journal of Experimental
Psychology, vol. 80, no. 3, part 2, 1-46.
BAUER, L. 1983. English Word-formation. Cambridge: Cambridge University Press.
BENDER, M.L. 1983. Color Term Encoding in a Special Lexical Domain: Sudanese Arabic Skin
Colors. Anthropological Linguistics, vol. 25, no. 1, 19-27.
BENNETT, T.J.A. 1988. Aspects of English Colour Collocations and Idioms. Heidelberg: Carl
Winter Universittsverlag.
BERGH, G. 1997. Varfr r inte bilen bara bl? Sprkvrd, no 1, 8-15.
BERLIN , B. and E.A. BERLIN 1975. Aguaruna Color Categories. American Ethnologist, vol. 2, 6187.
BERLIN, B. and P. KAY 1969. Basic Color Terms: Their Universality and Evolution. Berkeley:
University of California Press.
BIGGAM, C.P. 1993. Aspects of Chaucers Adjectives of Hue. The Chaucer Review, vol. 28, no. 1,
41-53.
1995a. Sociolinguistic Aspects of Old English Colour Lexemes. Anglo-Saxon England,
vol. 24, 51-65.
1995b. Review of Wyler (1992). Anglia, vol. 113, no. 1, 87-90.
1997. Blue in Old English: An Interdisciplinary Semantic Study. Amsterdam: Rodopi
B.V.
1998. Grey in Old English: An Interdisciplinary Semantic Study. London:
Runetree Press.
BIRREN, F. 1978. Color and Human Response. New York: Van Nostrand Reinhold.
BLANK, A. 1999. Co-presence and Succession: A Cognitive Typology of Metonymy. In Panther,
K.-U. and G. Radden (eds.), 169-191.

Bibliography

225

BLATN, R. 1996. New Lifestyle New Word Meanings. In Weigand, E. and F. Hundsnurscher
(eds.), Lexical Structures and Language Use. Proceedings of the International Conference on
Lexicology and Lexical Semantics. Mnster, September 13-15, 1994. Volume 1: Plenary
Lectures and Session Papers. Tbingen: Max Niemeier Verlag, 197-193.
BLOOMFIELD, L. 1933. Language. New York: Holt, Rinehart and Winston.
BOLINGER, D. 1967. Adjectives in English: Attribution and Predication. Lingua,
vol. 18, no.1, 1-34.
1980. Language the Loaded Weapon: The Use and Abuse of Language Today. London:
Longman.
BORG, A. (ed.) 1999. The Language of Color in the Mediterranean. Stockholm: Almqvist &
Wiksell International.
BOSCH, P. 1985. Context Dependence and Metaphor. In Paprott, W and R. Dirven (eds.), 141176.
BOYNTON, R.M. 1997. Insights Gained from Naming the OSA Colors. In Hardin and Maffi (eds.),
135-150.
BOYNTON, R.M. and C.X. OLSON 1987. Locating Basic Colors in the OSA Space. Color Research
and Application, vol. 12, 94-105.
1990. Salience of Chromatic Basic Color Terms Confirmed by Three Measures. Vision
Research, vol. 30, no. 9, 1311-1317.
BRAISBY, N. and B. FRANKS 1997. What does Word Use Tell us about Conceptual Content?
Psychology of Language and Communication, vol. 1, no. 2, 5-16.
BRANDT, P.A. 1993. Cognition and the Semantics of Metaphor: A General Outline. Acta
Linguistica Hafniensia, vol. 26, 5-21.
BRENNER, A. 1982. Colour Terms in the Old Testament. Sheffield: JSOT Press.
1999. On Color and the Sacred in the Hebrew Bible. In Borg, A. (ed.), 200-207.
BROWN, C.H. 1990. A Survey of Category Types in Natural Language. In Tsohatzidis, S.L. (ed.),
Meanings and Prototypes: Studies in Linguistic Categorization. London: Routledge, 17-47.
BROWN, R. 1958. How Shall a Thing Be Called? Psychological Review, vol. 65, no. 1, 14-21.
BROWN, R and E. LENNEBERG 1954. A Study in Language and Cognition. Journal of Abnormal
and Social Psychology, vol. 49, 454-462.
BROWN, W.P. 1972. Studies in Word Listing: Some Norms and Their Reliability. Irish Journal of
Psychology, vol. 1, 117-159.
1978. A Cross-National Comparison of English-Language Category Norms. Language
and Speech, vol.21, part 1, 50-68.

226

Bibliography

BURNLEY, J.D. 1976. Middle English Colour Terminology and Lexical Structure. Linguistische
Berichte, vol. 41, 39-49.
1992. Lexis and Semantics. In Blake, N.F. (ed.), The Cambridge History of the English
Language, II. Cambridge: Cambridge University Press, 409-499.
BUSS, D. 1997. The Use of Colour Names in Idiomatic and Colloquial Language Colour as
Idiom. In Sivik, L. (ed), 170-174.
CAFFI, C. and R.W. JANNEY 1994. Towards a Pragmatics of Emotive Communication. Journal of
Pragmatics, vol. 22, no. 3-4, 325-373.
CARUS-WILSON, E. 1952. The Woollen Industry. Chapter in Postan, M. and E.E. Rich (eds.), The
Cambridge Economic History of Europe. Volume II: Trade and Industry in the Middle Ages.
Cambridge: Cambridge University Press, 355-428.
CASSON, R.W. 1994. Russett, Rose, and Raspberry: The Development of English Secondary Color
Terms. Journal of Linguistic Anthropology, vol. 4, no. 1, 5-22.
1997. Color Shift: Evolution of English Color Terms from Brightness to Hue. In
Hardin, C.L. and L. Maffi (eds.), 224-239.
CASSON, R.W. and P.M. GARDNER. 1992. On Brightness and Color Categories: Additional Data.
Current Anthropology, vol. 33, no. 4, 395-399.
CHAUCER, G. (1987) The Riverside Chaucer. (Gen. ed. Benson, L.D.) Oxford: Oxford University
Press.
CHURCH, K.W. and P. HANKS 1990. Word Association Norms, Mutual Information, and
Lexicography. Computational Linguistics, vol. 16, no.1, 22-29.
CHURCH, K.W., W. GALE, P. HANKS and D. HINDLE 1991. Using Statistics in Lexical Analysis. In
Zernik, U. (ed.) Lexical Acquisition: Exploiting On-line Resources to Build a Lexicon.
Hillsdale: Lawrence Erlbaum, 115-164.
CLARK, E.V. 1992. Conventinality and Contrast: Pragmatic Principles with Lexical Consequences.
In Lehrer, A. and E.F. Kittay (eds.), 171-188.
CLARK, H.H. 1992. Arenas of Language Use. Chicago: The University of Chicago Press.
1996. Communities, Commonalities, and Communication. In Gumperz, J.J. and S.C.
Levinson (eds.), 324-355.
CLARK, H.H. and G.L. MURPHY 1982. Audience Design in Meaning and Reference. In Le Ny, J.-F.
and W. Kintsch (eds.), Language and Comprehension. Amsterdam: North-Holland
Publishing Company, 287-299.
CLAUSNER, T.C. and W. CROFT 1999. Domains and Image Schemas. Cognitive Linguistics,
vol. 10, no. 1, 1-31.
CLEAR, J. 1992. Corpus Sampling. In Svartvik, J and H. Wekker (eds.), New Directions in English
Language Corpora: Methodology, Results, Software Developments. Berlin: Mouton de
Gruyter, 21-31.

Bibliography

227

COLLIER, G.A. 1973. Review of Berlin and Kay (1969), Language, vol. 49, 245-48.
COLLIER, G.A. et al. 1976. Further Evidence for Universal Color Categories. Language, vol. 52,
no. 4, 884-890.
CONKLIN, H.C. 1964 [1955]. Hanuno Color Categories. In Hymes, D. (ed.), Language in Culture
and Society: Reader in Linguistics and Anthropology. London: Harper & Row, 189-192
(Reprinted from Southwestern Journal of Anthropology, vol. 11, no. 4, 339-44.)
1973. Color Categorization (Review of Berlin and Kay 1969). American
Anthropologist, vol. 75, 931-942.
CONLEY, K. and W.E. COOPER 1981. Conjoined Ordering of Color Terms by Children and Adults.
Studies in Language, vol. 3, 305-22.
CORBETT, G. and G. MORGAN 1988. Colour Terms in Russian: Reflections of Typological
Constraints in a Single Language. Journal of Linguistics, vol. 24, no. 1, 31-64.
CORBETT, G. and I.R.L. DAVIES 1995. Linguistic and Behavioural Measures for Ranking Basic
Colour Terms. Studies in Language, vol. 19, no. 2, 301-357.
CORBETT, G. and I.R.L. DAVIES 1997. Establishing Basic Color Terms: Measures and Techniques.
In Hardin, C.L. and L. Maffi (eds.), 197-223.
CRAWFORD, T.D. 1982. Defining Basic Color Term. Anthropological Linguistics, vol. 24, no. 3,
338-43.
CROFT, W. 1993. The Role of Domain in the Interpretation of Metaphors and Metonymies.
Cognitive Linguistics, vol. 4, no. 4, 335-370.
1998. Linguistic Evidence and Mental Representations. Cognitive Linguistics, vol. 9,
no. 2, 151-173.
CRUSE, D.A. 1977. The Pragmatics of Lexical Specificity. Journal of Linguistics, vol. 13,
152-164.
1986. Lexical Semantics. Cambridge: Cambridge University Press.
1990. Prototype Theory and Lexical Semantics. In Tsohatzidis, S.L. (ed.), 382-402.
1992. Antonymy Revisited: Some Thoughts on the Relationship Between Words and
Concepts. In Lehrer, A. and E.F. Kittay (eds.), 289-306.
1995 Polysemy and Related Phenomena from a Cognitive Linguistic Viewpoint. In
Saint-Dizier, P. and E. Viegas (eds.), Computational Lexical Semantics. Cambridge:
Cambridge University Press. 33-49.
CRUSE, D.A. and P. TOGIA 1995. Towards a Cognitive Model of Antonymy. Lexicology, vol. 1,
no. 1, 113-141.
CYTOWIC, R.E. 1989. Synesthesia: A Union of the Senses. NewYork: Springer-Verlag.
DANDRADE, R. and M. EGAN 1974. The Colors of Emotion. American Ethnologist, vol. 1, 49-63.

228

Bibliography

DAVIDOFF, J. and P. MITCHELL 1993. The Colour Cognition of Children. Cognition, vol. 48,
121-137.
DAVIES, I.R.L. and G.G. CORBETT. 1997. A Cross-Cultural Study of Colour Groupings: Evidence
for Weak Linguistic Relativity. British Journal of Psychology, vol. 88, no. 3, 493-517.
DAVIES, I.R.L., P.T. SOWDEN, D.T. JERRET, T. JERRET and G.G. CORBETT. 1998. A Cross-Cultural
Study of English and Setswana Speakers on a Colour Triads Task: A Test of the Sapir-Whorf
Hypothesis. British Journal of Psychology, vol. 89, no. 1, 1-15.
DE VALOIS, R., R.L.I. ABRAMOV, and G.H. JACOBS 1966. Analysis of Response of LGN Cells.
Journal of the Optical Society of America, vol. 56, 966-77.
DEDRICK, D. 1998. Naming the Rainbow. London: Kluwer Academic Publishers.
DEIGNAN, A. 1999a. Linguistic Metaphors and Collocation in Nonliterary Corpus Data. Metaphor
and Symbol, vol. 14, no. 1, 19-36.
1999b. Corpus-based Research into Metaphor. In Cameron, L. and G. Low (eds.),
Researching and Applying Metaphor. Cambridge: Cambridge University Press, 177-199.
DERRIG, S. 1978. Metaphor in the Color Lexicon. In Farkas, D., W.M. Jacobsen and K.W. Torys
(eds.), Papers from the Parasession on the Lexicon. Chicago: Chicago Linguistic Society,
85-96.
DIRVEN, R. 1993. Metonymy and Metaphor: Different Mental Strategies of Conceptualisation.
Leuvense Bijdragen, vol. 82, no. 1, 1-28.
DIRVEN, R. and J.R. TAYLOR 1988. The Conceptualisation of Vertical Space in English: The Case
of Tall. In Rudzka-Ostyn (ed.), Topics in Cognitive Linguistics, 379-402.
DIRVEN, R. and M. VERSPOOR 1998. Cognitive Exploration of Language and Linguistics.
Amsterdam: John Benjamins (Cognitive Linguistics in Practice 1).
DIXON, R.M.W. 1982. Where Have All the Adjectives Gone? and Other Essays in Semantics and
Syntax. Berlin: Mouton Publishers.
DOBROVOLSKIJ, D. and E. PIIRAINEN 1996. Symbole in Sprache und Kultur. Studien zur
Phraseologie aus kultursemiotischer Perspektive. Bochum: Brockmeyer.
1998. On Symbols: Cognitive and Cultural Aspects of Figurative Language. Lexicology,
vol. 4, no. 1, 1-34.
DOWNING, P. 1977. On the Creation and Use of English Compound Nouns. Language, vol. 53,
no. 4, 810-842.
DURBIN, M. 1972. Basic TermsOff Color? Semiotica, vol. 4, 257-288.
ECO, U. 1985. How Culture Conditions the Colours We See. In Blonsky, M. (ed.), On Signs.
Oxford: Basil Blackwell.
EKMAN, P., W.V. FRIESEN and P. ELLSWORTH. 1972. Emotion in the Human Face: Guidelines for
Research and an Integration of Findings. New York: Pergamon Press

Bibliography

229

EMBER, M. 1978. Size of Color Lexicon: Interaction of Cultural and Biological Factors. American
Anthropologist, vol. 80, 364-67.
EVANS, N. 1997. Review of Geeraerts et. al (1994). Functions of Language, vol. 4. no. 2, 299-303.
FASS, D. 1997. Processing Metonymy and Metaphor. London: Ablex Publishing Corporation.
FAUCONNIER, G. 1994. Mental Spaces: Aspects of Meaning Construction in Natural Language.
Cambridge: Cambridge University Press. (First published in 1985 by The Massachusetts
Institute of Technology.)
1997. Mappings in Thought and Language. Cambridge: Cambridge University Press.
FAUCONNIER, G. and M. TURNER 1999. Metonymy and Conceptual Integration. In Panther, K.-U.
and G. Radden (eds.), 77-90.
FERRIS, C. 1993. The Meaning of Syntax: A Study in the Adjectives of English. London: Longman.
FILLMORE, C.J. 1985. Frames and the Semantics of Understanding. Quaderni di Semantica, vol.
VI, no. 2, 222-254.
FORBES, I. 1979. The Terms Brun and Marron in Modern Standard French. Journal of
Linguistics, vol. 15, 295-305.
1986. Variation and Change in the Basic Colour Vocabulary of French. Sigma, vol. 10,
81-103.
FRANK, J. 1990. Gender Differences in Color Naming: Direct Mail Order Advertisements.
American-Speech, vol. 65, no. 2, 114-126.
FRANKOWSKA, S. 1995. Multifunctional Affixes in Word Formation: The Case of the English
Suffix -ish. Unpublished masters essay. Lublin: Katolicki Uniwersytet Lubelski.
FROMKIN, V. and R. RODMAN 1998. An Introduction to Language. (6th ed.) London: Harcourt
Brace.
FLT, E. 1995. Snvit och de sju frgerna. En studie i frgord, srskilt i de slaviska sprken.
Uppsala: Uppsala Slavic Papers 22.
GAGE, J. 1995. Colour and Culture. In Lamb, T. and J. Bourriau (eds), 175-193.
GEERAERTS, D. 1997. Diachronic Prototype Semantics: A Contribution to Historical Lexicology.
Oxford: Clarendon.
GEERAERTS, D., S. GRONDELAERS and P. BAKEMA. 1994. The Structure of Lexical Variation:
Meaning, Naming, and Context. Berlin: Mouton de Gruyter.
GEERAERTS, D and S. GRONDELAERS 1995. Looking Back at Anger: Cultural Traditions and
Metaphorical Patterns. In Taylor, J. and R. MacLaury (eds.), Language and the Cognitive
Construal of the World. New York: Mouton de Gruyter, 153-179.
GIBBS, R.W. JR. 1999. Speaking and Thinking with Metonymy. In Panther, K.-U. and G. Radden
(eds.), 61-76.

230

Bibliography

GIBBS, R.W. JR. and G.J. STEEN. (eds.) 1999. Metaphor in Cognitive Linguistics. Amsterdam: John
Benjamins.
GLEASON, H.A. 1961. An Introduction to Descriptive Linguistics. New York: Holt, Rinehart and
Winstone.
GODDARD, C. 1998. Semantic Analysis. A Practical Introduction. Oxford: Oxford University
Press.
GOOSSENS, L. 1995. Metaphtonomy: The Interaction of Metaphor and Metonymy in Figurative
Expressions for Linguistic Action. In Goossens, L., P. Pauwels, B. Rudzka-Ostyn, A-M.
Simon-Vandenbergen and J. Vanparys (eds.), By Word of Mouth: Metaphor, Metonymy and
Linguistic Action in a Cognitive Perspective. Amsterdam: John Benjamins, 175-204.
GRLACH, M. 1997. The Linguistic History of English. London: MacMillan.
GOTTFRIED, G.M. 1997. Comprehending Compounds: Evidence for Metaphoric Skill? Journal of
Child Language, vol. 24, no. 1, 163-186.
GRADY, J. 1999. A Typology of Motivation for Conceptual Metaphor: Correlation vs.
Resemblance. In Gibbs, R.W. Jr. and G.J. Steen (eds.), 79-100.
GRADY, J., T. OAKLEY and S. COULSON. 1999. Blending and Metaphor. In Gibbs, R.W. Jr. and
G.J. Steen (eds.), 101-124.
GREENE, K.S. and M.D. GYNTHER 1995. Blue versus Periwinkle: Color Identification and Gender.
Perceptual and Motor Skills, vol. 80, 27-32.
GREENFELD, P.J. 1986. What is Grey, Brown, Pink, and Sometimes Purple: The Range of the
Wild-Card Color Terms. American Anthropologist, vol.88, no. 4, 908-16.
GRICE, H.P. 1989. Studies in the Way of Words. Cambridge: Harvard University Press.
GUMPERZ, J.J. and S.C. LEVINSON (eds.) 1996. Rethinking Linguistic Relativity. Cambridge:
Cambridge University Press.
HALFF, H.M., A. ORTONY and R.C. ANDERSON 1976. A Context-Sensitive Representation of
Word Meaning. Memory & Cognition, vol. 4, no. 4, 378-383.
HALLIDAY, M.A.K. 1994. An Introduction to Functional Grammar. 2nd edition. London: Arnold.
HARDIN, C.L. 1988. Color for Philosophers: Unweaving the Rainbow. Cambridge: Hacket.
1993. Van Brakel and the Not-So-Naked Emperor. British Journal for the Philosophy of
Science, vol. 44, no. 1, 137-150.
1997. Basic Color Terms and Basic Color Categories. In Sivik, L. (ed), 146-157.
HARDIN, C.L. and L. MAFFI (eds.) 1997. Color Categories in Thought and Language. Cambridge:
Cambridge University Press.
HARRISON, B. 1973. Form and Content. Oxford: Basil Blackwell.

Bibliography

231

HAWKINS, B. 2001. Ideology, Metaphor and Iconographic Reference. In Dirven, R., R. Frank and
C. Ilie (eds.), Language and Ideology. Volume II: Descriptive Cognitive Approaches.
Amsterdam: John Benjamins Publishing Company, 27-50.
HAYS, D.G., E. MARGOLIS, R. NAROLL, and D.R. PERKINS 1972. Color Term Salience. American
Anthropologist, vol. 74, 1107-1121.
HEIDER, E.R. 1972. Universals in Color Naming and Memory. Journal of Experimental
Psychology, vol. 93, no. 1, 10-20.
HICKERSON, N. 1971. Review of Berlin and Kay (1969). International Journal of American
Linguistics, vol. 37, 257-270.
HOLLAND, D. and N. QUINN (eds.) 1987. Cultural Models in Language and Thought. Cambridge:
Cambridge University Press.
HONECK, R.P. and R.R. HOFFMAN (eds.) 1980. Cognition and Figurative Language. Hillsdale:
Lawrence Erlbaum Associates.
HUNT LAZERSON, B. 1977. Antonymy and Ten Modern English Color Names. Journal of
Psycholinguistic Research, vol. 6, no. 1, 21-36.
HUPKA, R.B., Z. ZALESKI, J. OTTO, L. REIDL and N.V. TARABRINA 1997. The Colors of Anger,
Envy, Fear, and Jealousy: A Cross-Cultural Study. Journal of Cross-Cultural Psychology,
vol. 28, no. 2, 156-171.
HUTCHINGS, J. 1997a. Colour in Folklore. In, Sivik, L. (ed.), 182-186.
1997b. Folklore and Symbolism of Green. Folklore vol. 108, 55-63.
JAKOBSON, R. 1956. Two Aspects of Language and Two Types of Aphasic Disturbances. In
Jakobson, R. and M. Halle (eds.), 1971 (2nd ed.), Fundamentals of Language. Paris: Mouton,
67-96.
JESPERSEN, O. 1924. [Reprinted 1963] The Philosophy of Grammar. London: George Allen &
Unwin.
JOHANSSON, S and K. HOFLAND 1989. Frequency analysis of English Vocabulary and Grammar:
Based on the LOB Corpus. Volume I: Tag Frequencies and Word Frequencies. Oxford:
Clarendon.
JOHNSON, A., O. JOHNSON and M. BAKSH 1986. The Color of Emotions in Machiguenga.
American Anthropologist, vol. 88, no. 3, 674-681.
JOHNSON, M. 1987. The Body in the Mind: The Bodily Basis of Meaning, Imagination, and
Reason. Chicago: The University of Chicago Press.
JOHNSON-LAIRD, P.N. and K. OATLEY 1989. The Language of Emotion: An Analysis of a
Semantic Field. Cognition and Emotion, vol. 3, 81-123.
KAY, P. 1975. Synchronic Variation and Diachronic Change in Basic Color Terms. Language in
Society, vol. 4, 257-270.

232

Bibliography

1999a. The Emergence of Basic Color Lexicons Hypothesis: A Comment on The

Vocabulary of Color with Particular Reference to Ancient Greek and Classical Latin, by
John Lyons. In Borg, A. (ed.), 76-90.
1999b. Color. Journal of Linguistic Anthropology, vol. 9, nos. 1-2, 32-35.
KAY, P., B. BERLIN, and W. MERRIFIELD 1991. Biocultural Implications of Systems of Color
Naming. Journal of Linguistic Anthropology, vol. 1, 12-25.
KAY, P., B. BERLIN, L. MAFFI and W. MERRIFIELD 1997. Color Naming across Languages. In
Hardin, C.L. and L. Maffi (eds.), 21-56.
KAY, P. and C.K. MCDANIEL 1978. The Linguistic Significance of the Meaning of Basic Color
Terms. Language, vol. 54, no. 3, 610-46.
KAY, P. and L. MAFFI 1999. Color Appearance and the Emergence and Evolution of Basic Color
Lexicons. American Anthropologist, vol. 101, no. 4, 743-760.
KELLY, K.L. and D.B. JUDD 1976. Color: Universal Language and Dictionary of Names.
Washington: U.S. Department of Commerce.
KEMMERER, D. 1999. Neuroscientific Evidence against Wierzbickas Analysis of the Meaning of
Basic Color Terms. In Hiraga, M.K., C. Sinha and S. Wilcox (eds.), Cultural, Psychological
and Typological Issues in Cognitive Linguistics. Amsterdam: John Benjamins, 249-266.
KERTTULA, S. Forthcoming. English Colour Terms: Etymology, Chronology, and Relative
Basicness. Ph.D. diss. Helsinki: Mmoires de la Socit Nophilologique de Helsinki.
KIKUCHI, A. and F. LICHTENBERK 1983. Semantic Extension in the Colour Lexicon. Studies in
Language, vol. 7, no. 1, 25-64.
KOCH, P. 1999. Frame and Contiguity: On the Cognitive Bases of Metonymy and Certain Types of
Word Formation. In Panther, K.-U. and G. Radden (eds.), 139- 167.
KORNERUP, A. and J.H. WANSCHER 1978. Methuen Handbook of Colour. 3rd ed. London: Eyre
Methuen.
KS-DIENES, D. 1983. Om Frgtermernas Symboliska Anvndning. Gteborg: Papers in
Anthropological Linguistics 10.
KVECSES, Z. 1986. Metaphors of Anger, Pride, and Love: A Lexical Approach to the Structure of
Concepts. New York: John Benjamins.
2000. Metaphor and Emotion: Language, Culture, and Body in Human Feeling.
Cambridge: Cambridge University Press
KVECSES, Z. and G. RADDEN 1998. Metonymy: Developing a Cognitive Lingistics View.
Cognitive Linguistics, vol. 9, no. 1, 37-77.
KRIEG, M.L. 1976. Semantic Fields of Color Words in Old French, Old English, and Middle
English. Dissertation. Ann Arbor: UMI Dissertation Services [1995].

Bibliography

233

KRISTOL, A.M. 1980. Color Systems in Southern Italy: A Case of Regression. Language, vol. 56,
no 1, 137-145.
KUHN, T.S. 1962. The Structure of Scientific Revolutions. Chicago: University of Chicago Press.
LAKOFF, G. 1987. Women, Fire, and Dangerous Things: What Categories Reveal about the Mind.
Chicago: University of Chicago Press.
1990. The Invariance Hypothesis: Is Abstract Reason Based on Image-Schemas?
Cognitive Linguistics, vol. 1, no. 1, 39-74.
LAKOFF, G. and M. JOHNSON 1980. Metaphors We Live By. Chicago: University of Chicago Press.
1999. Philosophy in the Flesh: The Embodied Mind and Its Challenge to Western
Thought. New York: Basic Books.
LAKOFF, G. and M. TURNER 1989. More Than Cool Reason: A Field Guide to Poetic Metaphor.
New York: Mouton de Gruyter.
LAKOFF, R. 1975. Language and Womans Place. New York: Harper & Row.
LAMB, T. and J. BOURRIAU (eds.) 1995. Colour: Art&Science. Cambridge: Cambridge University
Press.
LANGACKER, R.W. 1984. Active zones. Proceedings of the Annual Meeting of the Berkeley
Linguistic Society, vol. 10, 172-188.
1987. Foundations of Cognitive Grammar. Vol. 1, Theoretical Prerequisites. Stanford:
Stanford University Press.
1991a. Concept, Image, and Symbol. The Cognitive Basis of Grammar. New York:
Mouton de Gruyter.
1991b. Foundations of Cognitive Grammar. Vol. 2, Descriptive Application. Stanford:
Stanford University Press.
1993. Reference-Point Constructions. Cognitive Linguistics, vol. 4, no. 1, 1-38.
1997. The Contextual Basis of Cognitive Semantics. In Nuyts, J. and E. Pederson (eds.),
Language and Conceptualization. Cambridge: Cambridge University Press, 229-252.
1999. Grammar and Conceptualisation. Berlin: Mouton de Gruyter.
LEE, D. 1992. Competing Discourses: Perspective and Ideology in Language. London: Longman.
LEE, P. 1996. The Whorf Theory Complex: A Critical Reconstruction. Amsterdam: John
Benjamins.
LEECH, G. 1981. Semantics. 2nd ed. Harmondsworth: Penguin Books.
LEHRER, A. and E.F. KITTAY (eds.) 1992. Frames, Fields, and Contrasts. New Essays in Semantic
and Lexical Organization. Hillsdale: Lawrence Erlbaum Associates.

234

Bibliography

LENNEBERG, E.H. and J.M. ROBERTS 1956. The Language of Experience: A Study in Methodology.
Indiana University Publications in Anthropology and Linguistics. Memoir 13: supplement to
International Journal of American Linguistics, vol. 22, no. 2. Baltimore: Waverley Press Inc.
LEVI, J.N. 1976. The Syntax and Semantics of Non-Predicating Adjectives in English.
Bloomington: Indiana University Linguistics Club.
1978. The Syntax and Semantics of Complex Nominals. London: Academic Press.
LEVINSON, S.C. forthcoming. Yl Dnye and the Theory of Basic Color Terms.
LUCY, J.A. 1992a. Language Diversity and Thought: A Reformulation of the Linguistic Relativity
Hypothesis. Cambridge: Cambridge University Press.
1992b. Grammatical Categories and Cognition: A Case Study of the Linguistic
Relativity Hypothesis. Cambridge: Cambridge University Press.
1997. The Linguistics of Color. In Hardin, C.L. and L. Maffi (eds.), 320-346.
LUCY, J.A. and R.A. SHWEDER 1988. The Effect of Incidental Conversation on Memory for Focal
Colors. American Anthropologist, vol. 90, no. 4, 923-931.
LUMSDEN, C.J. 1985. Color Categorization: A Possible Concordance between Genes and Culture.
Proceedings of the National Academy of Sciences USA, vol. 82, 5805-5808.
LYNCH, C.M. and M. STRAUSS-NOLL 1987. Mauve Washers: Sex Differences in Freshman
Writing. English Journal, vol. 76, 90-94.
LYONS, J. 1963. Structural Semantics: An Analysis of Part of the Vocabulary of Plato. Oxford:
Blackwell.
1977. Semantics. Cambridge: Cambridge University Press.
1995. Colour in Language. In Lamb, T. and J. Bourriau (eds.), 194-224.
1999. The Vocabulary of Color with Particular Reference to Ancient Greek and
Classical Latin. In Borg, A. (ed.), 38-75.
MACLAURY, R. 1987. Color-Category Evolution and Shuswap Yellow-with-Green. American
Anthropologist, vol. 89, no.1, 107-124.
1991. Social and Cognitive Motivations of Change: Measuring Variability in Color
Semantics. Language, vol. 67, no. 1, 34-62.
1992. From Brightness to Hue: An Explanatory Model of Color-Category Evolution.
Current Anthropology, vol. 33, no. 2, 137-186.
1995. Vantage Theory. In Taylor, J.R. and R.E. MacLaury (eds.), 231-276.
1997. Color and Cognition in Mesoamerica: Constructing Categories as Vantages.
Austin: University of Texas Press.
1999. Basic Color Terms: Twenty-five Years After. In Borg, A. (ed.), 1-37.

Bibliography

235

2000. Linguistic Relativity and the Plasticity of Categorization. In Ptz, M. and M.

Verspoor (eds.), Explorations in Linguistic Relativity. Amsterdam: John Benjamins
Publishing Co, 251-293.
MACLAURY, R.E., J. ALMSI and Z. KVECSES 1997. Hungarian Piros and Vrs: Color from
Points of View. Semiotica, vol. 114, no. , 67-81.
MAFFI, L. 1990. Somali Color Term Evolution: Grammatical and Semantic Evidence.
Anthropological Linguistics, vol. 32, no. 3-4, 316-334.
1991. A Bibliography of Color Categorization Research 1970-1990. In Berlin, B. and P.
Kay, Basic Color Terms: Their Universality and Evolution, 1st paperback edn. Berkeley:
University of California Press, 173-189.
MAFFI, L. and C.L. HARDIN 1997. Closing Thoughts. In Hardin, C.L. and L. Maffi (eds.),
347-372.
MAGNUSSON, U. and G. PERSSON 1986. Facets, Phases and Foci: Studies in Lexical Relations in
English. Stockholm: Almqvist & Wiksell International.
MANNING, C.D. and H. SCHTZE 1999. Foundations of Statistical Natural Language Processing.
Cambridge, Mass: MIT Press.
MARCHAND, H. 1969. The Categories and Types of Present-Day English Word-Formation.
Wiesbaden: Otto Harrassowitz.
1966. On Attributive and Predicative Derived Adjectives and Some Problems Related to
the Distinction. Anglia, vol.82, no. 2, 131-149.
MAXWELL-STUART, P.G. 1981a. Studies in Greek Colour Terminology. Volume I: .
Leiden: E.J. Brill.
1981b. Studies in Greek Colour Terminology. Volume II: .
Leiden: E.J. Brill.
MCMANUS, I.C. 1983. Basic Colour Terms in Literature. Language and Speech, vol. 26, part 3,
247-252.
1997. Half-a-million Basic Colour Words: Berlin and Kay and the Usage of Colour
Words in Literature and Science. Perception, vol. 26, no. 3, 367-370.
MCNEILL, N.B. 1972. Colour and Colour Terminology. Journal of Linguistics,
vol. 8, 21-33.
MEDIN, D.L. and L.W. BARSALOU 1987. Categorization Processes and Categorical Perception. In
Harnad, S. (ed.), Categorical Perception: The Groundwork of Cognition. Cambridge:
Cambridge University Press, 455-490.
MEHRABIAN, A. 1981. [1971] Silent Messages: Implicit Communication of Emotions and
Attitudes. 2nd ed. Belmont, Calif.: Wadsworth.
MERVIS, C.B. and E.M. ROTH 1981. The Internal Structure of Basic and Non-Basic Color
Categories. Language, vol. 57, 384-405.

236

Bibliography

MOON, R. 1998. Fixed Expressions and Idioms in English: A Corpus Based Approach. Oxford:
Clarendon Press.
MOONWOMON, B. 1994. Color Categorization in Early Greek. Journal of Indo-European Studies,
vol. 22, no. 1-2, 37-65.
MORGAN, G. 1993. Basic Colour Terms: Comparative Results for French and Russian. Journal of
French Language Studies, vol. 3, no. 1, 1-17.
MOSS. A.E. 1989. Basic Colour Terms: Problems and Hypotheses. Lingua,
vol. 78, 313-20.
1989. Does Russian Have a Basic Colour Term for Purple? Linguistics, vol. 27, no. 1,
145-155.
MOSS, A.E., I. DAVIES, G. CORBETT, and G. LAWS 1990. Mapping Russian Basic Colour Terms
Using Behavioural Measures. Lingua, vol. 82, 313-332.
MUNSELL, A.H. 1961. A Color Notation. 11th ed. Baltimore: Munsell Color Company.
MURPHY, G.L. 1988. Comprehending Complex Concepts. Cognitive Science, vol. 12, 529-562.
NICKERSON, D. and S.M. NEWHALL 1943. A Psychological Color Solid. Journal of the Optical
Society of America, vol. 33, 419-422.
NIEMEIER, S. 1998. Colourless Green Ideas Metonymise Furiously. In Ungerer, F. (ed.), Kognitive
Lexikologie und Syntax. Rostock: Universitt Rostock (Rostocker Beitrge zur
Sprachwissenschaft: 5), 119-146.
NORD, B. 1997. Von Azurblau bis Zinnoberrot. Farbbezeichnungen und ihre Darstellung im
Wrterbuch. Lebende Sprache, vol. 42, no. 1, 27-30.
NORRICK, N.R. 1981. Semiotic Principles in Semantic Theory. Amsterdam: John Benjamins.
NOWACZYK, R.C. 1982. Sex-related Differences in the Color Lexicon. Language and Speech, vol.
25, part 3, 257-265.
NUNBERG, G. 1978. The Pragmatics of Reference. Ph.D. diss. City University of New York.
1979. The Non-Uniqueness of Semantic Solutions: Polysemy. Linguistics and
Philosophy, vol. 3, no. 2, 143-184.
OAKES, M.P. 1998. Statistics for Corpus Linguistics. Edinburgh: Edinburgh University Press.
OGDEN, C.K. and I.A. RICHARDS 1972. The Meaning of Meaning. 10th edn. [First edn. 1923]
London: Routledge & Kegan Paul.
OHTSUKI, M. 2000. A Cognitive Linguistic Study of Colour Symbolism. Tokyo: Institute for the
Research and Education of Language, Daito-Bunka University.
OJA, V. 2001. Linguistic Studies of Estonian Colour Terminology. Ph.D. diss. Tartu: Tartu
University Press.

Bibliography

237

OSGOOD, C.E., W.H. MAY and M.S. MIRON 1975. Cross-Cultural Universals of Affective
Meaning. Chicago: University of Illinois Press.
PANTALEO, N. 1994 Ase Roser When Hit Redes: Semantic Shifts and Cultural Overtones in
the Middle English Colour Lexicon. In Fernndez, F., M. Fuster and J.J. Calvo (eds.),
English Historical Linguistics 1992: Papers from the 7th International Conference on English
Historical Linguistics. Amsterdam: John Benjamins, 273-284.
PANTHER, K.-U. and G. RADDEN (eds.) 1999. Metonymy in Language and Thought. Amsterdam:
John Benjamins.
PAPROTT, W. and R. DIRVEN (eds.) 1985. The Ubiquity of Metaphor: Metaphor in Language and
Thought. Amsterdam: John Benjamins.
PAWOWSKI, A. 1999. The Quantitative Approach in Cultural Anthropology: Application of
Linguistic Corpora in the Analysis of Basic Colour Terms. Journal of Quantitative
Linguistics, vol. 6, no. 2, 222-234.
PERSSON, G. 1990. Meanings, Models and Metaphors: A Study in Lexical Semantics in English.
Stockholm: Almqvist & Wiksell International.
1994. Social Change, Prototypes and Word Meaning. In, Todtenhaupt, M and I.
Valfridsson (eds.), Sprache als lebendiger Kulturspiegel. Festschrift fr Astrid Stedje.
Stockholm: Almqvist & Wiksell International, 169-177.
1997. Frnskivling eller poison pie. Kognitiva aspekter p namngivning av svamp i
olika sprk. In, Nilsson J. (ed.) Aspekter p vxtnamn. Skrifter utgivna av Dialekt-, ortnamns
och folkminnesarkivet i Ume. Serie E, Vxtnamn, no. 1, 63-83.
2000. Vit flugsvamp or Destroying Angel: Cognitive Aspects of Names for Fungi in
Some Languages. In, Morgensen J.E., V Hjrnager Pedersen and A Zettersten (eds.),
Symposium on Lexicography IX: Proceedings of the Ninth International Symposium on
Lexicography April 23-25, 1998 at the University of Copenhagen. Tbingen: Max Niemeyer
Verlag, 305-322.
RADDEN, G. and Z. KVECSES 1999. Towards a Theory of Metonymy. In Panther, K.-U. and G.
Radden (eds.), 17-59.
REDDY, M. 1979. The Conduit Metaphor. In Ortony, A. (ed.), Metaphor and Thought. Cambridge:
Cambridge University Press, 284-324.
RICH, E. 1977. Sex-Related Differences in Colour Vocabulary. Language and Speech, vol. 20,
404-409.
ROSCH, E. 1975a. Cognitive Reference Points. Cognitive Psychology, vol. 7, 532-547.
1975b. Cognitive Representations of Semantic Categories. Journal of Experimental
Psychology: General, vol. 104, no. 3, 192-233.
1975c. The Nature of Mental Codes for Color Categories. Journal of Experimental
Psychology: Human Perception and Performance, vol. 1, no. 4, 303-322.

238

Bibliography

1978. Principles of Categorization. In Rosch, E. and B.B. Lloyd (eds.) Cognition and
Categorization. Hillsdale: Lawrence Erlbaum Associates, 27-48.
1983. Prototype Classification and Logical Classification: The Two Systems. In Kofsky
Scholnik, E. (ed.), New Trends in Conceptual Representation: Challenges to Piagets
Theory? London: Lawrence Erlbaum Associates, 73-86.
1988. Coherence and Categorization: A Historical View. In Kessel, F.S. (ed.), The
Development of Language and Language Researchers: Essays in Honor of Roger Brown.
Hillsdale: Lawrence Erlbaum Associates, 373-392.
ROSCH, E. and C.B. MERVIS 1996 [1975]. Family Resemblance: Studies in the Internal Structure
of Categories. In, Geirsson, H. and M. Losonsky, (eds.), Readings in Language and Mind.
Cambridge, MA: Blackwell Publishers Inc, 442-460. [Originally in Cognitive Psychology,
vol. 7, 573-605.]
ROSCH, E., C.B. MERVIS, W.D. GRAY, D.M. JOHNSON and P. BOYES-BRAEM 1976. Basic Objects
in Natural Categories. Cognitive Psychology, vol. 8, 382-439
ROWE, C. 1972. Conceptions of Colour and Colour Symbolism in the Ancient World. Eranos, vol.
41, 327-364.
RUSIECKI, J. 1985. Adjectives and Comparison in English: A Semantic Study. London: Longman.
SAHLINS, M. 1976. Colors and Cultures. Semiotica, vol. 16, no.1, 1-22.
SAMPSON, G. 1980. Schools of Linguistics. Stanford: Stanford University Press.
SAUNDERS, B. 1995. Disinterring Basic Colour Terms: A Study in the Mystique of Cognitivism.
History of the Human Sciences, vol. 8, no. 4, 19-38.
SAUNDERS, B. and J. VAN BRAKEL 1997. Are there Nontrivial Constraints on Colour
Categorization? Behavioral and Brain Sciences, vol. 20, no. 2, 167-228.
SCHFER, B. 1999. Les termes fondamentaux de couleur en franais. In Borg, A. (ed), 91-104.
SETO, K. 1999. Distinguishing Metonymy from Synecdoche. In Panther, K-U. and G. Radden
(eds.), 91-120.
SHAVER, P., J. SCHWARTZ, D. KIRSON, and C. OCONNOR 1987. Emotion Knowledge: Further
Exploration of a Prototype Approach. Journal of Personality and Social Psychology, vol. 52,
no. 6, 1061-1086.
SHIELDS, K. 1979. Indo-European Basic Colour Terms. Canadian Journal of Linguistics, vol. 24,
142-46.
SIEGEL, M.E.A. 1980. Capturing the Adjective. New York: Garland Publishing.
SIMPSON, J. and A.W.S. TARRANT 1991. Sex- and Age-related differences in Colour Vocabulary.
Language and Speech, vol.34, no.1, 57-62.
SINCLAIR, J. 1991. Corpus, Concordance, Collocation. Oxford University Press.

Bibliography

239

SIVIK, L. 1997. Color Systems for Cognitive Research. In Hardin, C.L. and L. Maffi (eds.),
163-193.
SIVIK, L. and C. TAFT, 1992. Frgkombinationer och associerade betydelser: Semantiska
dimensioner och olika frgackord. Stockholm: Frginstitutet. (Frgrapport F41)
SIVIK, L. and C. TAFT 1994. Color Naming: A Mapping in the NCS of Common Color Terms.
Scandinavian Journal of Psychology, vol. 35, 144-64.
SIVIK L. (ed.) 1997. Colour and Psychology. From AIC Interim Meeting 96 in Gothenburg.
Stockholm: Frginstitutet. (Colour Report F50)
SLEIGHT, C.C. and P.M. PRINZ 1982. Childrens Color Vocabulary. Language and Speech, vol. 25,
part 1, 75-79.
SMITH, E.E. and D.L. MEDIN 1981. Categories and Concepts. Cambridge: Harvard University
Press.
SMITH, J.J., L. FURBEE, K. MAYNARD, S. QUICK and L. ROSS 1995. Salience Counts: A Domain
Analysis of English Color Terms. Journal of Linguistic Anthropology, vol. 5, no. 2, 203-216.
SMITH, L.B., S.S. JONES and B. LANDAU 1992. Count Nouns, Adjectives, and Perceptual Properties
in Childrens Novel Word Interpretations. Developmental Psychology, vol. 28, no. 2, 273286.
SPENCE, N.C.W. 1989. The Linguistic Field of Colour in French. Zeitschrift fr Romanische
Philologie, vol. 105, 472-497.
SPRINGER, K. and G.L. MURPHY 1992. Feature Availability in Conceptual Combination.
Psychological Science, vol. 3, no. 2, 111-117.
STEINVALL, A. 2000. Aspects of Elaborate Colour Terms in Victorian Poetry. In Spnberg, S-J, H.
Kardela and G. Porter (eds.), The Evidence of Literature: Interrogating Texts in English
Studies. Lublin: Maria Curie-Skodowska University Press, 403-419.
STOEVA-HOLM, D. 1996. Farbbezeichnungen in deutchen Modetexten: Eine morphologischsemantische Untersuchung. Ph.D. dissertation. Uppsala: Acta Universitatis Upsaliensis.
STOKES, M. 2001. The Color of Sex: Whiteness, Heterosexuality, and the Fiction of White
Supremacy. Durham: Duke University Press.
STORM, C. and T. STORM 1987. A Taxonomic Study of the Vocabulary of Emotions. Journal of
Personality and Social Psychology, vol. 53, no. 4, 805-816.
STUBBS, M. 1995. Collocations and Semantic Profiles: On the Cause of the Trouble with
Quantitative Studies. Functions of Language, vol. 2, no. 1, 23-55.
1996. Text and Corpus Analysis: Computer-assisted Studies of Language and Culture.
Oxford: Blackwell Publishers.
SUMMERS, D. 1996. Computer Lexicography: The Importance of Representativeness in Relation to
Frequency. In Thomas, J. and M. Short (eds.), Using Corpora for Language Research:
Studies in Honour of Geoffrey Leech. London: Longman, 260-266.

240

Bibliography

SUN, R.K. 1983. Perceptual Distances and the Basic Color Term Encoding Sequence. American
Anthropologist, vol. 85, no. 2, 387-391.
SVENSSON, P. 1998. Number and Countability in English Nouns: An Embodied Model. Ph.D.
dissertation. Uppsala: Swedish Science Press.
SWARINGEN, S., S. LAYMAN, and A. WILSON 1978. Sex Differences in Color Naming. Perceptual
and Motor Skills, vol. 47, 440-442.
TAFT, C. 1997. Generality Aspects of Color Naming and Color Meaning. Ph. D. dissertation.
Gteborg: Department of Psychology, Gteborg University.
TAFT, C. and L. SIVIK 1997. Salient Color Terms in Four Languages. Scandinavian Journal of
Psychology, vol. 38, no. 1, 29-34.
TAYLOR, J.R. 1992. Old Problems: Adjectives in Cognitive Grammar. Cognitive Linguistics,
vol. 3, no. 1, 1-36.
1995. Linguistic Categorization: Prototypes in Linguistic Theory. 2nd ed. Oxford:
Clarendon.
TAYLOR, J.R. and R.E. MACLAURY (eds.) 1995. Language and the Cognitive Construal of the
World. Berlin: Mouton de Gruyter.
TEYSSIER, J. 1968. Notes on the Syntax of the Adjective in Modern English. Lingua, vol. 20,
225-249.
THOMAS, L.L., A.T. CURTIS and R. BOLTON 1978. Sex Differences in Elicited Color Lexicon Size.
Perceptual and Motor Skills, vol. 47, 77-78.
THOMPSON, E. 1995. Colour Vision. A Study in Cognitive Science and the Philosophy of
Perception. London: Routledge.
THOMPSON, S.A. 1989. A Discourse Approach to the Cross-Linguistic Category Adjective. In
Corregan, R., F. Eckman and M. Noonan (eds.), Linguistic Categorization. Amsterdam: John
Benjamins, 245-265.
TRUDGILL. P. 2000. Sociolinguistics : An Introduction to Language and Society. 4th ed. London:
Penguin.
TUGGY, D. 1993. Ambiguity, Polysemy, and Vagueness. Cognitive Linguistics, vol. 4, no. 4,
273-290.
TURNER, M. 1996. The Literary Mind. Oxford: Oxford University Press.
TURNER, M. and G. FAUCONNIER 1995. Conceptual Integration and Formal Expression. Metaphor
and Symbolic Activity, vol. 190, no. 3, 183-204.
TVERSKY, A. 1977. Features of Similarity. Psychological Review, vol. 84, no. 4, 327-352.
TWARDZISZ, P. 1997. Zero Derivation in English: A Cognitive Grammar Approach. Lublin:
Wydawnictwo UMCS.

Bibliography

241

ULLMANN, S. 1957. The Principles of Semantics. 2nd ed. Oxford: Basil Blackwell.
1962. Semantics: An Introduction to the Science of Meaning. Oxford: Basil Blackwell.
UNGERER, F. 1996. Review of Geeraerts, Grodelaers and Bakema (1994), Lexicology, vol. 2, no. 1,
78-84.
UNGERER, F. and H.-J. SCHMID 1996. An Introduction to Cognitive Linguistics. London: Longman.
VAN BRAKEL, J.

1991. Meaning, Prototypes and the Future of Cognitive Science. Minds and
Machines, vol. 1, no. 3, 233-257.

1993. The Plasticity of Categories: The Case of Colour. The British Journal for the
Philosophy of Science, vol. 44, no. 1, 103-135.
VARELA, F.J., E. THOMPSON and E. ROSCH 1991. The Embodied Mind. Cambridge, Mass.: The
MIT Press.
VEROSUB, A.L. 1994. Scarlet Letters: Metonymic Uses of the Color Red. Semiotica, vol. 102,
no. 1/2, 27-47.
VERSPOOR, M. 1998. Colourful Bits of Experience that Count. Series A: General & Theoretical
Papers, Paper No. 450. Essen: LAUD, 1- 21.
WARE, V. 1992. Beyond the Pale: White Women, Racism and History. London: Verso
VON WATTENWYL, A.

and H. ZOLLINGER 1979. Color-Term Salience and Neurophysiology of

Color Vision. American Anthropologist, vol. 81, 279-88.

WARREN, B. 1984. Classifying Adjectives. Gteborg: Gothenburg Studies in English 56.

1988. Ambiguity and Vagueness in Adjectives. Studia Linguistica, vol. 42, no. 2,
122-72.
1992. Sense Developments: A Contrastive Study of the Development of Slang Senses
and Novel Standard Senses in English. Stockholm: Almqvist and Wiksell.
1993. Nominal and Adjectival Modifiers of Nouns. Anglistik und Englischunterricht,
vol. 49, 57-68.
1995. Distinguishing between Metaphor and Metonymy. In Melchers, G. and B. Warren
(eds.), Studies in Anglistics. Stockholm: Almqvist and Wiksell International, 137-151.
1999. Aspects of Referential Metonymy. In Panther, K-U. and G. Panther (eds.),
121-135.
WEI, P. and R. MANGOLD 1997. Bunt gemeint, doch farblos gesagt: Wann wird die Farbe eines
Objektes nicht bennant? Sprache & Kognition, vol. 16, no. 1, 31-47.
WESCOTT, R.W. 1983. Color Metaphors in Three Language Phyla. In Di Pietro, R.J. W. Frawley
and A. Wedel (eds.), The First Delaware Symposium on Language Studies. London:
Associated University Presses, 87-98.

242

Bibliography

WESTPHAL, J. 1987. Colour: Some Philosophical Problems from Wittgenstein. Oxford: Basil
Blackwell.
WIENER, M. and A. MEHRABIAN 1968. Language Within Language: Immediacy, a Channel in
Verbal Communication. New York: Appleton-Century-Crofts.
WIERZBICKA, A. 1986. Whats in a Noun? (Or: How do Nouns Differ in Meaning from
Adjectives?) Studies in Language, vol. 10, no. 2, 353-389.
1990. The Meaning of Colour Terms: Semantics, Culture, and Cognition. Cognitive
Linguistics, vol. 1, 99-149.
1992. Semantic Primitives and Semantic Fields. In Lehrer, A. and E. Kittay (eds.),
209-207.
1996. Semantics: Primes and Universals. Oxford: Oxford University Press.
WILLIAMS, J.M. 1976. Synaesthetic Adjectives: A Possible Law of Semantic Change. Language,
vol. 52, no. 2, 461-478.
WITKOWSKI, S.R. and C.H. BROWN. 1983. Marking-Reversals and Cultural Importance.
Language, vol. 59, no. 3, 569-582.
WITTGENSTEIN, L. 1958. Philosophical Investigations. Translated by G.E.M. Anscombe. 2nd ed,
Oxford: Basil Blackwell.
1978. Remarks on Colour. Ed. G.E.M. Anscombe, trans. L. McAlister and M. Schttle.
Oxford: Basil Blackwell.
WOODS, A., P. FLETCHER and A. HUGHES 1986. Statistics in Language Studies. Cambridge:
Cambridge University Press.
WYLER, S. 1992. Colour and Language: Colour Terms in English. Tbingen: Gunter Narr Verlag.
XU, W. 1994. Chinese Colour Semantics. Ph.D. dissertation. Australian National University.
YANG, Y. 1996. Sex- and Level-related Differences in the Chinese Colour Lexicon. Word, vol. 47,
no. 2, 207-220.
ZADEH, L.A. 1965. Fuzzy Sets. Information and Control, vol. 8, 338-353.
ZASORINA, L.N. (ed.) 1977. astotnyj slovar russkogo jazyka: okol 40 000 slov. Moscow:
Russkij jazyk.
ZAR, J.H. 1996. Biostatistical Analysis (3rd ed.). Upper Saddle River, N.J.: Prentice Hall.
ZIMMER, A.C. 1982. What Really is Turquoise? A Note on the Evolution of Color Terms.
Psychological Research, vol. 44, 213-230.
1984. There Is More Than One Level in Color Naming A Reply to Zollinger.
Psychological Research, vol. 46, 411-416.

Bibliography

243

ZLATEV, J. 1997. Situated Embodiment: Studies in the Emergence of Spatial Meaning. PhD
dissertation. Stockholm: Gotab.
ZOLLINGER, H. 1984. Why Just Turquoise? Remarks on the Evolution of Color Terms.
Psychological Research, vol. 46, 403-409.

244

Bibliography

Dictionaries (with abbreviations)

Bloomsbury Dictionary of Contemporary Slang. 1990. THORNE, T. London : Bloomsbury
Cambridge International Dictionary of English. 1995. PROCTER, P. (Editor in Chief)
Cambridge: Cambridge University Press. [CIDE]
Collins Cobuild English Language Dictionary. 1993. SINCLAIR, J. (Editor in Chief)
London: Harper Collins Publishers. [CCELD]
Longman Dictionary of Contemporary English 1995. 3rd ed., Summers, D. (Director)
Harlow: Longman. [Longman]
NTCs Dictionary of American Slang and Colloquial Expressions. 1995. SPEARS, R.A.
Lincolnwood: National Textbook Company
Oxford Advanced Learners Dictionary. 1995. 5th ed. CROWTHER, J. (Ed.). Oxford: Oxford
University Press. [OALD]
Oxford English Dictionary, The. 1993. CD-ROM version. Oxford: Oxford University
Press. [OED]
Websters New Encyclopedic Dictionary.1995.Revised ed. Cologne: Knemann.
[Websters]

Internet Sources
Encyclopdia Britannica Online. http://search.eb.com/
OED Online. http://dictionary.oed.com/
Sprkbanken. http://spraakbanken.gu.se/
Svenska Akademiens Ordbok. http://g3.spraakdata.gu.se/saob/index.html
The PLANTS National Database. http://plants.usda.gov/
Colour Order Systems in Art and Science. http://www.colorsystem.com/
Chroma. http://pourpre.com/chroma/

Appendix 1. Description of the Bank of English Corpus

This appendix presents subcorpora in the Bank of English and their sizes at the
time of my sampling of the corpus (February June 1998). The Bank of English
contained approximately 323 million words at that time.
Subcorpus

Abbreviation

UK Independent newspaper
indy
Australian newspapers
oznews
UK ephemera
brephem
UK general interest magazines
brmags
UK spontaneous speech
brspok
US ephemera
usephem
UK BBC World Service broadcasts bbc
UK Guardian newspaper
guard
UK New Scientist magazine
newsci
US National Public Radio broadcasts npr
UK Books
brbooks
US miscellaneous books
usbooks
US regional newspapers
usnews
UK Economist magazine
econ
UK Times/Sunday Times newspapers times
UK Today newspaper
today

245

Size
(number of words)
19,452,295
33,378,314
4,721,964
30,137,896
20,181,050
1,255,655
18,522,600
24,261,095
6,087,440
22,259,602
42,127,619
32,656,385
8,578,632
12,125,208
20,950,497
26,606,537

Appendix 2. Regular Expressions Eliminated via Automatic

Commands.
Below is a list of the regular expressions that were automatically eliminated with a
grep-v command when I collected my material. Grep is a UNIX command for
selecting lines matching a regular expression, on which the lookup commands g
and v are based. The v-command selects only those lines which dont match the
entered regular expression. After this, the material was examined, and all the
remaining non-colour tokens were eliminated.
Peach
ripe peach\|peach essence\|peach halves\|peach juice\|peach kernel\|peach melba\|peach
orchard\|peach orchards\|peach palm\|peach schnapps\|peach tree\|peach trees\|peach
baskets\|peach
blossom\|peach
brandy\|peach
chutney\|peach
cobbler\|peach
flavouring\|peach flavour\|peach flesh\|peach fuzz\|a peach of a\|peach perfume\|peach
pie\|peach potato\|peach pie\|peach sauce\|peach-skin\|peach slices\|peach aroma

Rust
anti-rust\|anti rust\|emerge from rust\|damage from rust\|surface rust\|wont rust\|doesnt
rust\|will rust\|against rust\|any rust\|can rust\|cluster-cup rust\|for rust\|from rust\|get rust\|it
rust\|mean rust\|more rust\|not rust\|noticed rust\|old rust\|prevent rust\|ring-rust\|ring
rust\|structural rust\|vitually rust
rust away\|rust has\|rust inhibitor\|the rust off\|rust-proof\|rust proof\|rust-proofing\|rust
resistant\|rust-resistant\|rust wreck\|rust disease\|bit of rust\|flakes of rust\|heap of rust\|a lot
of rust\|signs of rust

Plum
ripe plum\|Victoria plum\|wild plum\|plum brandy\|plum cake\|plum draw\|plum
flavours\|plum fruit\|plum job\|plum jobs\|plum position\|plum pudding\|plum ride\|plum
role\|plum roles\|plum sauce\|plum tomato\|plum tomatoes\|plum tree\|plum trees\|plum
assignment\|plum chutney\|plum jam\|plum marmalade\|plum orchard\|plum post\|plum
posts\|plum posting\|plum prize\|plum puree\|plum stones\|plum post\|plumposts\|plum
posting\|plum prize\|plum puree\|plum stones\|plum assignment\|plum chutney\|plum
jam\|plum marmalade\|plum orchard

Rose
actually rose\|they all rose\|anxiety rose\|asstes rose\|audience rose\|autumn rose\|Autumn
rose\|average rose\|ball rose\|balloon rose\|BHP rose\|blossoms rose\|body rose\| Brian
rose\|business rose\|China rose\|Christmas rose\|climbing rose\|color rose to\|Commonwealth
rose\|companies rose\|affair rose\|aircraft rose\|air rose\|Airways rose\|Anna rose\|anger
rose\|ANZ rose\|approvals rose\|American rose\|yen rose\|GDP rose
confidence rose\|Conservatives rose\|contract rose\|New Corp rose\|costs rose\|Cross
rose\|crowd rose\|curtain rose\|debt rose\|decisions rose\|delegates rose\|deputies rose\|dollar
rose\| division rose\|dust rose\|earnings rose\|Earnings rose\|English rose\|Europe rose\|Eva
rose\|eventually rose\|executives rose\|expectations rose\|exports rose\|Exports rose\| eye
brows rose\|factories rose
figure rose\|figures rose\|flames rose\|funds rose\|George rose\|Gil rose\|Gilts rose\|Gold
rose\|goods rose\|group rose\|he rose\|He rose\|heart rose\|heat rose\|hills rose\|Holdings
rose\|homes rose\|House rose\|I rose\|immedietely rose\|imports rose\|index rose\|Index rose

247

248

Jesus rose\|Jill rose\|leaders rose\|level rose\|levels rose\|of living rose\|London rose\|market

rose\|month rose\|moon rose\|mortgages rose\|numbers rose\|mortgage rose\|offences
rose\|optimism rose\|orders rose\|pay rose\|period rose\|permits rose\|pound rose\|price
rose\|prices rose\|production rose\|products rose\|profit rose\|profits rose\|rate rose\|rates
rose\|red rose
registrations rose\|repayments rose\|revenues rose\|revenue rose\|sales rose\|services
rose\|share rose\|shares rose\|She rose\|she rose\|slopes rose\|smoke rose\|spending
rose\|spirits rose\|States rose\|steam rose\|stock rose\|stocks rose\|sun rose\|surplus rose\|tax
rose\|temperature rose\|temperatures rose\| tempo rose\|they rose\|They rose\|thing rose\|toll
rose\|trade rose\|trees rose\|Turnover rose\|turnover rose\|trusts rose\|UK rose\|unemployment
rose\|US rose
vacasion rose\|value rose\|visitors rose\|voice rose\|voices rose\|volume rose\|volumes
rose\|wages rose\|water rose\|We rose\|we rose\|week rose\|which rose\|White rose\| white
rose\|who rose\|wild rose\|wind rose\|workers rose\|world rose\|year rose\|a yellow rose\|yield
rose\|yields rose\|1994 rose\|1997 rose
a rose is\|a rose by\|rose bed\|rose beds\|rose-beds\|a rose between\|rose bowls\|rose
bowl\|rose breeders\|rose bud\|rose buds\|rose-bush\|rose bush\|rose-bushes\|rose bushes\|rose
collis\|rose
collection\|rose
country\|rose-covered\|rose
&dollar\|rose
every\|rose
fragrance\|rose fragrances\|rose family\|rose fertilizers\|rose fertilizer\|rose for\|rose arches
rose garden\|rose-garden\|rose gardens\|rose-gardens\|rose geranium\|rose given\|rose
grew\|rose grafted\|rosegrower\|rose-grower\|rose growers\|rose-growers\|rose growing\|rosegrowing\|rose had\|rose hip\|rose-hip\|rose hips\|rose-hips\|rose is\|rose, jasmine\|rose
last\|rose leaves\|rose lovers\|rose may\|rose motif\|rose named\|rose oil\|rose on\|rose
otto\|rose out\|rose pattern\|rose-patterned\|rose perfume
rose petal\|rose petals\|rose-petal\|rose-petals\|rose rust\|rose-scented\|rose show\|rose
smell\|rose so \|rose specialists\|rose stem\|rose stems\|rose symbol\|rose that\|net rose
to\|rose tree\|rose trees\|rose type\|rosetypes\|rose varieties\|rose water\|rose-water\|rose
were\|rose window\|rose-window\|rose windows\|rose-windows\|rose wine\|rose wines
rose a\|rose above\|roseabruptly\|rose after\|rose again\|rose alarmingly\|rose an\|rose
last\|rose higher\|rose less\|rose markedly\|rose modestly\|rose nearly\|rose quickly\|rose
rapidly\|rose roughly\|rose sharply\|rose slightly\|rose slowly\|rose steadily\|rose
unexpectedly\|rose up\|rose yesterday\|rose-coloured\|rose-colored\|rose colour\|rose color
Rose\|rose and fell\|rose and a\|rose and put\|rose and came\|rose and asked\|rose and
ran\|rose and said\|rose and shook\|rose and stepped\|rose and stood\|rose and streched\|rose
and walked\|rose and told\|rose and took\|rose and then\|rose and began

Cream
anticeptic cream\|avocado cream\|barrier cream\|body cream\|butter cream\|camouflage
cream\|chocolate
cream\|cleansing
cream\|clotted
cream\|coconut
cream\|cold
cream\|conditioning cream\|custard cream\|dairy cream\|double cream\|eye cream\|face
cream\|fresh cream\|full cream\|hand cream\|heavy cream\|moisturizing cream\|mustard
cream\|night cream\|potato cream\|pouring cream\|salad cream\|sharish cream\|single
cream\|skin cream\|sour cream\|soured cream\|spermicidal cream\|special cream\|sun
cream\|whipped cream\|whipping cream\|whip the cream\|waterproof cream\|zinc
cream\|triple cream\|shaving cream
cream bun\|cream buns\|cream, butter\|cream cake\|cream cakes\|cream cheese\|cream
cleanser\|cream cleansers\|cream cracker\|cream crackers\|cream eggs\|cream-filled\|cream,
jelly\|cream liqueur\|cream pie\|cream pies\|cream sauce\|cream sauces\|cream soup\|cream
tea\|cream teas\|cream until\|strawberries and cream\|cream and strawberries\|add the
cream\|add cream\|apply the cream\|apply cream\|cream and fruit\|cream and garlic\|cream
and jam\|ice cream\|ice-cream\|Ice cream\|Ice-cream\|ICE CREAM\|ICE-CREAM

249

Navy
American navy\|airforce and navy\| airforce and the navy\|army and navy\|army and the
navy\|Party and navy\|Australian navy\|British navy\|Chinese navy\|the entire navy\|federal
navy\|former navy\|French navy\|German navy\| Israeli navy\|Italian navy\|Japanese
navy\|merchant navy\|official navy\| the old navy\|the regular navy\| republican
navy\|Russian navy\|Swedish navy\|Soviet navy\|US navy\|United States navy\|the whole
navy\|Weimar navy\|Yugoslav navy\|Navy\|NAVY
navy bean\|the navy can\|the navy has\|the navy had\|the navy is\|the navy could\|the navy
would\|the navy was\|navy aircraft\|navy and airforce\|navy and army\|the navys\|head of
the navy\|navy, army\|navy, airforce\|the navy began\|navy ship\|navy ships\|navy
boats\|navy came\|navy captain\|navy commanders\|the navy continued\|navy divers\|navy
engineer\|navy found\|navy took\|navy waited\|navy gunboat\|navy gunboats\|the navy had
been\|navy vessels\|navy veteran\|navy high command\|navy leaders\|navy lieutenant\|navy
minister\|navy spokesman\|secretay of the navy\|strength of the navy\|size of the
navy\|history of the navy\|record of the navy\|join the navy

Olive
olive branch\|olive bread\|olive grove\|olive groves\|olive hatches\|olive leaf\|olive
leaves\|olive nymphs\|olive oil\|olive paste\|olive pate\|olive press\|olive presses\|olive
processing\|olive orchard\|olive orchards\|olive salad\|olive tree\|olive trees\|olive
wood\|olive virgin

Orange
Orange\|orange blossom\|orange drink\|orange fruit\|orange fruits\|orange grove\|orange
groves\|orange juice\|orange-juice\|orange marmalade\|orange peel\|orange rind\|orange
sauce\|orange slice\|orange slices\|orange sorbet\|orange tree\|orange trees\|orange
zest\|orange flavour\|orange-flavored
orange, apple\|orange brandy\|orange butter\|orange cake\|orange crop\|orange
custard\|orange
flavoured\|orange-flavoured\|ORANGE\|orange
grower\|orange
growers\|orange liqueur\|orange oil\|orange orchard\|orange orchards\|orange pekoe\|orange
plantation\|orange plantations\|orange seller\|orange squash\|orange squeezer
orange-flower water\|orange flower water\|orange, grapefruit\|orange harvest\|orange
segments\|orange box\|orange boxes\|orange crates\|orange season\|orange soda\|orange
stick\|orange tea\|orange and\|orange bomb\|orange bone\|lorange\|dorange\|orangebluhende\|orange-blossom\|orange, blackcurrant\|orange biscotti
orange and almond\|orange and blackcurrant\|orange and chcolate\|orange and clove\|orange
and date\|orange and grapefruit\|orange and grapes\|orange and honey\|orange and
lemon\|orange and lime\|orange and other\|orange and pineapple\|orange and
raspberry\|orange and strawberry\|orange and sweet\|orange and you

Violet
Violet\|VIOLET\|Ultra violet\|ultra violet\|Ultra-violet\|ultra-violet\|African violet\|shrinking
violet\|dog violet\|dog-violet\|nodding violet\|violet cream\|violet tea\|violet leaf\|violet
leaves\|violetscent\|violet-scented\|water violet\|tooth violet\|shy violet

Silver
Silver\|SILVER\|silver
dagger\|silver
medal\|silver
medallist\|silver
lining\|silver
plated\|silver-plated\|silver medalist\|silver jewellery\|silver spoon\|silver plate\|silver
chain\|silver-plate\|silver medals\|silver tray\|silver wedding\|silver jubilee\|silver horn\|silver
gilt\|silver-gilt\|silver spoons\|silver bracelet
silver platter\|silver anniversary\|silver ring\|silver cutlery\|silver wires\|silver trophy\|silver
earrings\|silver rats\|silver-framed\|silver coins\|silver bullet\|silver jewellery\|silver
mining\|sterling silver\|Sterling silver\|olympic silver\|silver and bronze

250

Golden
Golden\|GOLDEN\|golden age\|golden opportunity\|golden rule\|golden boy\|golden
girl\|golden oldies\|golden handshake\|golden rules\|golden goal\|golden chance\|golden
oldie\|golden future\|golden era\|golden handcuffs\|golden wedding\|golden years\|golden
triangle\|golden days\|golden period\|golden summer\|golden couple\|golden jubilee\|golden
statue\|golden chances\|golden opportunities\|golden touch
golden share\|golden chain\|golden parachute\|golden egg\|golden eggs\|golden
thread\|golden serve\|golden hello\|golden parachutes\|golden boot\|golden remove\|golden
moment\|golden treasure\|golden moments\|golden scenario\|golden hellos\|golden
crown\|golden cool\|golden fleece

Gold
Gold\|GOLD\|Olympic gold\|carat gold\|won gold\|solid gold\|win gold\|Games gold\|struck
gold\|copper-gold\|pure
gold\|team
gold\|African
gold\|individual
gold\|winning
gold\|Commonwealth gold
gold medal\|gold medals\|gold medallist\|gold medalist\|gold medallists\|gold medalists\|gold
plated\|gold rush\|gold leaf\|gold-leaf\|gold mining\|gold standard\|gold bars\|gold chain\|gold
stocks\|gold bullion\|gold price\|gold coins\|gold watch\|gold mines\|gold production\|gold
reserves\|gold ring\|gold assets\|gold chains\|gold jewellery\|gold prices\|gold producer\|gold
project\|gold digger\|gold-digger\|gold rings\|gold bracelet\|gold miner\|gold teeth\|gold tooth

33

4829

2446

933
0

11
3

173

28

28

12

13

46

4165

1465

500

74

46

20

10

Beige

Black

Blue

Brown

Carmine

Charcoal
(grey/gray

Chartreuse

Cream

Crimson

Emerald

Fawn

Fuchsia

BREPHEM

251
0

18

249

354

547

18

BRMAGS

23

27

55

172

533

11

44

29

2609

4365

12807

207

23

16

316

658

1946

11

83

229

409

11

70

331

3416

14

39

50

20

557

1728

5762

48

10

241

499

1387

10

12

324

1069

7853

34

12

34

83

356

244

24

27

2939

5563

10039

117

41

27

103

14

14

37

106

60

11

1616

3624

9665

93

18

25

71

13

103

426

1141

131

561

2886

37

57

70

13

552

1587

3757

35

13

11

21

16

41

102

704

2197

4286

76

25

93

124

326

878

1384

35

151

80

11927

27102

74895

748

136

186

427

Total in
BoE

2961

757
8

1583

683

Green

Grey/Gray

Indigo

146

539

50

2256

4581

251

1004

88

257

123

405

830

2033

156

750

376

930

73

3608

5612

15

1903

3177

109

317

198

740

851

1629

10

778

2146

200

13113

28664

*3983

Azure

BRSPOK

59

USEPHEM

BBC

Golden

32

Aqua(marine)

GUARD

23

NEWSCI

14

NPR

BRBOOKS

USBOOKS

USNEWS

73

ECON

14

TIMES

*3636

28

INDY

23

SUBCORPORA
TODAY

Gold

OZNEWS

Amber

COLOUR
TERM

Appendix 3. Frequency of Colour Terms in the Bank of English

28
10
42
24
10
69
14
173
13
508
6
12
145
2087
23
12
67

Magenta

Maroon

Mauve

Mustard

Navy (blue)

Olive

Orange

Peach

Pink

Plum

Puce

Purple

Red

Rose

Rust

Scarlet

INDY

Lime

OZNEWS

16

44

30

3449

265

11

696

22

231

61

108

23

44

115

12

13

BREPHEM

Lilac

12

601

70

163

10

57

23

13

16

BRMAGS

211

40

171

6110

757

58

2133

104

755

133

378

23

150

98

50

93

121

76

BRSPOK

23

11

982

67

269

76

41

15

USEPHEM

11

285

31

54

30

19

BBC

Lemon

733

17

42

31

GUARD

79

10

33

2632

245

11

511

217

13

63

24

30

19

39

25

16

20

1060

56

53

96

NEWSCI

11

11

1327

103

177

12

128

21

NPR

457

41

148

7369

975

24

24

1789

40

858

102

128

13

153

71

62

51

61

55

67

BRBOOKS

137

28

75

4423

529

969

43

417

55

89

10

31

53

14

15

20

17

56

USBOOKS

70

454

27

79

41

USNEWS

816

35

76

36

ECON

18

81

49

2580

221

604

10

212

12

55

30

28

55

20

10

TIMES

SUBCORPORA

80

46

4004

236

10

801

27

195

11

136

24

33

28

40

14

11

TODAY

Lavender

COLOUR
TERM

1208

174

632

38912

3779

76

134

8924

302

3553

449

1126

108

501

518

194

339

340

231

275

Total in
BoE

252

WORDS

PER MILLION

829.3

703.7

859.8

4060

1795.6

54117

405.2

8178

392

1650.1

2072

100

384.4

7121

177

1746

846.3

20532

726

4551

876.4

5335

206

768

813.6

18110

461

5098

1363.3

57432

3077

12514

242

37

1198.0

39121

1696

9658

91

456.2

3914

148

957

644.2

7811

121

2128

809.1

16950

772

3399

22

53

787.9

20964

922

3794

28

47

944.4

305337

13055

68037

583

98

35

640

75

111

23488

2637

383

15

16132

TOKENS IN

SUBCORPORA

190

2052

13

816

11521

11

77

12

614

928

85

Yellow

5074

112

18

3466

15

White

18

28

42

15

23

Violet

11

16

13

Vermilion

192

126

12

65

Ultramarine

31

35

24

52

26

Total in
BoE

Turquoise

INDY

OZNEWS

Teal

GUARD

17

NEWSCI

BBC

NPR

USEPHEM

15

BRBOOKS

Tangerine

BRSPOK

USBOOKS

459

BRMAGS

115

USNEWS

27

ECON

Tan

TIMES

*3631

BREPHEM

SUBCORPORA
TODAY

Silver

COLOUR
TERM

253

255

Beige

Azure

Aqua
(marine)

Amber

COLOUR
TERM

A pale
creamy
brown
colour.

A very pale
creamy
brown
colour.

Bluish
green.

A greenish
blue colour.

The bright
blue colour
of the sky on
a sunny day.

[Not defined
as a colour] :
the
yellowishorange
traffic light
which shows
between the
green and
the red

Yellowishbrown or
orange.

Bright blue,
the colour of
the sky on a
sunny day.

CIDE

CCELD

Very light
yellowish
brown.

A bright
blue colour,
like the sky.

The colour
of the stone,
which is said
to be] a pale
greenishblue
precious
stone.

[The colour
of the
substance
amber,
which is said
to be] a
yellowishbrown
colour.

OALD

A pale dull
yellowish
brown
colour.

A bright
blue colour
like the sky.

A greenish
blue colour.

A yellowish
brown
colour.

LONG

DICTIONARIES

Yellowishgrey.

Coloured like
the unclouded
sky; orig. of a
deep intense
blue, now
usually of a
soft clear
bright blue

Aqua: A light
greenish blue
colour.
Aquamarine:
bluish-green.

Of a clear
yellowish
brown.

OED

A light
grayish
yellowish
brown.

The blue
color of the
clear sky.

A pale blue
to light
greenish
blue.

A dark
orange
yellow.

WEB

Appendix 4. Definitions of Colour Terms

The following list illustrates the definition of the colour terms under investigation as they
are defined in six contemporary dictionaries.
The dictionary are: The Oxford English Dictionary (OED), Cambridge International
Dictionary of English (CIDE), Collins Cobuild English Language Dictionary (CCELD),
Websters New Encyclopedic Dictionary (Web), Longman Dictionary of Contemporary
English (Long), and Oxford Advanced Learners Dictionary (OALD).
The text is quoted from the dictionaries. Text within brackets [ ] is my own
addition or summary.

Charcoal
(grey/gray)

Carmine

Brown

Blue

Black

COLOUR
TERM

A deep bright red

colour.

Dark grey.

Deep bright red

colour.

A very dark grey

colour.

A deep red colour.

Having the colour

of earth, wood or
coffee.

Having the colour

of chocolate, or
coffee mixed with
milk.

The colour of earth

or wood.

The colour of the

clear sky or of the
sea on a fine day.

Having the colour

of a clear sky or
the sea in sunlight.

The colour of the

sky without clouds
on a bright day, or
a darker or lighter
without variety of
this.
The colour of
chocolate or earth.

A deep purplish
red colour.

The dark colour of

night or coal.

Of the very darkest

colour; the
opposite of white.

The darkest colour

there is, like night
without stars.

LONG

Of the darkest
colour there is, the
colour of the sky at
night when there is
no light at all.
The colour of the
sky on a sunny
day.

OALD

CIDE

CCELD

A dark grey colour.

Deep crimson.

A composite
colour produced by
a mixture of orange
and black []
Brown is the
colour produced by
partial charring or
carbon-ization of
starch or woody
fibre, as in toasted
bread or potatoes,
peat, lignite,
withered leaves,
etc.

Absorbing all light:

of the colour of
night (J.), of the
colour of soot or
coal.
Of the colour of
the sky and the
deep sea.

OED

A vivid red.

The color of the

clear daytime sky:
a color lying
between green and
violet in the
spectrum.
Any of a group of
dull colors between
red and yellow in
hue

The color of soot

or coal

WEB

256

Gold(en)

Fuchsia

Fawn

Emerald

Crimson

Cream

Gold: the colour of

gold [= a precious
yellow metal]
Golden: like gold
in colour.

A pale yellowish
brown colour.
A pinkish-purple
colour.
Gold: a yellowish
colour, like that of
gold.
Golden: the colour
of gold.

A pale yellowish
brown colour

Gold: a bright
yellow colour
Golden: bright
yellow in colour
and looks rather
like gold.

A light yellowish
brown.

[[Not defined, but

the stone is
described as] a
bright green
transparent
precious stone.

Bright green in
colour

Of a bright green
colour.

A deep red.

A strong, slightly
purplish, deep red
colour

Gold: the colour of

the metal: a bright
golden yellow.
Golden: of the
colour of gold;
that shines like
gold.

A red colour like that

of the fuchsia flower.

A bright pink
colour.
Gold: the colour of
this metal [ = a
yellow metal]
Golden: having a
bright, rich, yellow
colour, like gold

A light yellowish
brown.

Brilliantly green like

the emerald.

Of a deep red
somewhat inclining
towards purple.

Cream-coloured,
yellowish white.

A pale apple-green.

OED

Pale yellow brown.

Bright green.

Having a deep
purplish red colour.

Yellowish-white in
colour.

Pale yellow, almost

white.

A dark, purplishred colour

Chartreuse

A bright green
colour.

LONG

A pale yellow or
pale green colour.

The colour of
cream, [which is
defined as] thick
yellowish-white
liquid.

OALD

A yellowish-white
colour

CIDE

CCELD

COLOUR
TERM

Gold: a deep
yellow.
Golden: having
the color of gold.

A vivid reddish
purple.

A light grayish
brown.

Bright or richly
green.

Deep purplish
red.

A pale yellow.

A brilliant yellow
green.

WEB

257

A bluish purple
colour.
A pale purple
colour.
A pale yellow
colour.
A pale pinkishpurple colour.

A dark colour between

blue and violet.
A pale bluish purple
colour.
Bright yellow in
colour.
Pale pinkish-purple in
colour.

Lavender

Lilac

Lemon

Grey/Gray

Indigo

A colour between
blue and yellow;
the colour of grass.

The colour of grass or

leaves.

Is a mixture of
black and white,
the colour of
clouds on a rainy
day.

CIDE

CCELD

The colour of ashes or

of clouds on a rainy
day.

Green

COLOUR
TERM

A pale purple
colour.

A pale yellow
colour.

A pale purple
colour.

A deep blue colour.

Of the colour of
ashes or lead.

Of the colour of
growing grass and
the leaves of most
plants and trees.

OALD

A pale purple
colour.

A pale yellow
colour.

A pale purple
colour.

A dark purplish
blue colour.

Having a colour of
black mixed with
white, like the
colour of ash; the
colour of smoke
and rain clouds.

The colour of grass

or leaves.

LONG

Of the colour of
lilac blossom [=
pale pinkish
violet].

Pale yellow.

A very pale blue

with a trace of red.

A deep violet-blue
colour.

The colour intermediate between

black and white, or
composed of a
mixture of black
and white with
little or no positive
hue; ash-coloured,
lead-coloured.

The colour which

in the spectrum is
intermediate
between blue and
yellow; in nature
chiefly
conspicuous as the
colour of growing
herbage and leaves.

OED

A moderate purple.

A brilliant greenish
yellow.

A pale purple.

A dark grayish
blue.

One of the series of

shades formed by a
blending of black
and white.

A color whose hue

is somewhat less
yellow than that of
growing fresh grass
or of the emerald
or is that of the part
of the spectrum
lying between blue
and yellow.

WEB

258

A pale purple
colour.

A pale purple
colour.

An interesting discussion of the status of olive as noun or adjective can be found here in the OED.

Orange

Olive

Navy (blue)

Mustard

Mauve

A colour between
red and yellow.

A colour between
red and yellow.

Yellowish green.

The colour of
olives that are not
ripe.

Yellowish-green in
colour.

The reddish yellow

colour of an
orange.

(Of the skin):

yellowish brown.

Dark blue.

Dark blue

A fairly dark
yellow colour.

A pale purple
colour.

A dark brownish
red colour.

A colour between
purple and red

A pale green
colour.

OALD

Very dark blue in

colour.

A dark reddish
purple colour.

Dark reddishpurple in colour.

Brownish yellow.

A dark purplish red

colour.

A dark, reddishpurple colour.

Magenta

Maroon

A light bright
greenish yellow
colour.

CIDE

Pale yellowishgreen in colour.

CCELD

Lime (green)

COLOUR
TERM

A colour that is
between red and
yellow.

Olive
skin/complexion
that is yellowish
brown.

A deep yellowish
green colour

Very dark blue.

A yellow-brown
colour.

A pale purple
colour.

A very dark redbrown colour.

A bright pink
colour.

A light yellowish
green colour.

LONG

A yellow to
yellowish green
color
Olive green: a color
greener, lighter, and
stronger than average
olive color.

A color between
red and yellow.

The reddish yellow

colour of the
orange.

A dark blue.

A moderate purple,
violet, or lilac
color.

A dark red.

A deep purplish
red.

WEB

1)A dull somewhat

yellowish green.
2)Yellowish brown
or brownish yellow
in the complexion
of persons or races.
3)A dull ashy
green with a
silvery sheen.1

A dark blue

Brownish yellow.

The colour of
mauve. [ = A
bright but delicate
purple dye].

A particular kind
of brownishcrimson or claret
colour.

OED

259

A delicate red or light crimson

colour.

The colour of
blood or fire.

A pink colour.

Of the colour of
fresh blood or a
similar colour.

Rose

Red

Purple

Reddish-pink or
brownish pink in
colour.

Pink.

A pink colour.

The colour which appears at

the lower or least refracted
end of the visible spectrum,
and is familiar in nature as
that of blood, fire, various
flowers []and ripe fruits.

A dark colour that

is a mixture of red
and blue.

Having the
colour of red and
blue mixed
together.

A dark reddish
blue colour.

A reddish-blue
colour.

Puce

The colour of
fresh blood.

Dark brownish
purple.

(Of) a brownish
purple colour.

A dark purplish
red colour.

A dark purple
colour.

Plum

The colour of
blood or of a ripe
tomato.

A dark red colour,

like a plum.

A dark reddishpurple colour.

A darkish
purplish red
colour.

Mixtures of red and blue in

various proportions, usually
containing some white or
black, or both, approaching on
the one side to crimson on the
other to violet.

Of a flea colour; purple-brown

or brownish purple.

A shade of purple.

A light or pale red colour with

a slight purple tinge.

Dark-red or purple
in colour.

A pale red colour.

The colour of a ripe peach, a

soft pale red. Or, the colour of
peach-blossom, a delicate rose
or pink.

OED

Of a pale red
colour.

A pale pinkishorange colour.

LONG

A pale red
colour.

Pink

Peach

A pinkishorange colour.

OALD

A colour that is
between red and
white.

CIDE
A pale pinkish
orange colour.

CCELD
Pale pinky-orange
in colour.

COLOUR
TERM

A moderate purplish
red.

A color whose hue

resembles that of fresh
blood or the ruby or is
that of the long-wave
extreme of the visible
spectrum.

A color about midway

between red and blue.

A dark reddish purple.

A pale red.

A moderate yellowish
pink.

WEB

260

Turquoise

A bright blue colour

that is fairly light
and often greenish.

Bluish green in
colour.

Dark or reddish
orange colour.

Teal

A deep orangeyellow colour.

Pale yellowish
brown.

A light yellowishbrown colour.

A yellowish-brown
colour.
The brown colour of
the skin of normally
white people after it
has been exposed to
the sun.

The colour of
silver: [which is
described as] a
valuable shiny
white metal.

Greyish-white in
colour.

A greenish-blue
colour.

Looking like silver:

[which is described
as] a shiny white
precious metal.

A bright red colour.

Bright red.

Bright red in colour.

A reddish-brown
colour.

OALD

A reddish brown
colour like that of
rust.

CIDE

A reddish-brown
colour.

CCELD

Tangerine

Tan

Silver

Scarlet

Rust

COLOUR
TERM

A greenish-blue
colour.

A greenish blue
colour.

A light yellowish
brown colour
The brown
colour that
someone with
pale skin gets
after they have
been in the sun.

The colour of
silver [ = a shiny
whitish valuable
metal].

Bright red.

LONG

The colour of the

turquoise [=] of a skyblue to apple-green
colour.

A shade of dark
greenish blue
resembling the patches
of this colour on the
teal.

A deep orange colour.

The colour of tan or

tanned leather; of a
yellowish or reddish
brown; tawny; bronzed,
sunburnt.

A silvery lustre or
colour. Silver-coloured;
having the colour of
silver, of a greyish white
hue with a metallic
lustre.

A brilliant vivid red

colour, inclining to
orange.

The colour of rust [a

red, orange or tawny
coating].

OED

A light greenish
blue.

A light yellowish
brown.
A brown color
imparted to the
skin by exposure to
the sun or weather.

A medium gray.

A bright red.

A strong reddish
brown.

WEB

261

A bluish purple
colour.
A colour like
that of snow,
milk or bone.

A bluish purple
colour.
The lightest of
colour there is,
the colour of
snow or milk.

Yellow

White

A colour like
that of a lemon
or gold or the
sun.

A bright red
orange colour.

Bright red in
colour.

Vermilion

The colour of
lemons or egg
yolks.

Bright blue.

A very bright
blue colour.

Ultramarine

Violet

CIDE

CCELD

COLOUR
TERM

Of the colour of
e.g. a ripe
lemon, an egg
yolk or gold.

Of the very
palest colour,
like fresh snow
or milk.

The colour of
wild violets;
bluish-purple.

A bright red
colour.

A very bright
blue colour.

OALD

The colour of
butter, gold, or
the middle part
of an egg.

The colour of
milk, salt, and
snow.

A colour
between purple
and blue.

A bright
reddish-orange
colour.

A very bright
blue colour.

LONG

Of the colour of gold,

butter, the yolk of an egg,
various flowers and other
objects [] in the spectrum
between green and orange.

Of the colour of snow or

milk; having that colour
produced by reflection,
transmission, or emission of
all kinds of light in the
proportion in which they
exist in the complete visible
spectrum, without sensible
absorption, being thus fully
luminous and devoid of any
distinctive hue.

The colour of violets, a blue

or bluish-purple colour.

Bright red or scarlet.

A special deep-blue colour.

OED

A color whose hue

resembles that of
ripe lemons or
sunflowers or is
that of the part of
the spectrum lying
between green and
orange.

The color of fresh

snow.

A reddish blue.

A vivid reddish
orange color.

A vivid blue.

WEB

262

Appendix 5. English Colour Terms and their Age

Colour
Term
Amber
Aqua(marine)
Azure
Beige
Black
Blue
Brown
Carmine
Charcoal
(grey/gray)
Chartreuse
Cream
Crimson
Emerald
Fawn
Fuchsia
*Gold
*Golden
Green
Grey/Gray
Indigo
Lavender
Lemon
Lilac
Lime
Magenta

First
recored

Colour
Term

First
recorded

1500
1862
1481
1879
700
1300
1300
1799

Maroon
Mauve
Mustard
Navy (blue)
Olive
Orange
Peach
Pink

1791
1860
1848
1840
1662
1600
1848
1720

1952
1884
1872
1400
1598
1881
1923
1400
1300
700
700
1622
1882
1796
1791
1923
1860

Plum
Puce
Purple
Red
Rose
Rust
Scarlet
*Silver
Tan
Tangerine
Teal
Turquoise
Ultramarine
Vermilion
White
Violet
Yellow

1878
1787
975
700
1530
1716
1386
1481
1749
1899
1923
1853
1598
1400
950
1370
700

263

SKRIFTER FRN MODERNA SPRK

Institutionen fr moderna sprk, Ume universitet
Redaktr:
Raoul J Granqvist
1.
2.
3.
4.

Mareike Jendis, Mumins wundersame Deutchlandabenteuer. Zur

Rezeption von Tove Janssons Muminbchern. Diss. 2001.
Lena Karlsson, Multiple Affiliations: Autobiographical Narratives
of Displacement by US Women. Diss. 2001.
Anders Steinvall, English Colour Terms in Context. Diss. 2002.
Sensuality and Power in Visual Culture edited by Raoul J.
Granqvist. Monograph 2002.