Introduction and Literature review

Although we take it for granted, language is a complex cognitive phenomenon which is

responsible for much of the information we receive and give. Computer systems that can
produce better language than a 4-year old have yet not been manufactured.
It is important to differentiate between language and communication. For example, is the
elaborate dance that bees perform to communicate the presence of food ‘language’? There are
several criteria which need to be present for a system to be called a language. Firstly, a
language is ‘regular’, that is, it is governed by a system of rules. Secondly, infinite
combinations of words can be made to produce infinite different meaningful phrases
(‘productivity’). Also, it is arbitrary and can be subdivided into discrete categories. Thus, the
dance of the bee is not language since it is neither productive nor arbitrary.
The structure of a language comprises of five basis systems. The sounds of a language are
called phonemes e.g., /k/. The smallest meaningful unit of a language is called a morpheme.
e.g., ‘talk’ and ‘ed’ (from the word ‘talked’) are both morphemes. Syntax refers to the
structure of sentences, the order and arrangement of words in a sentence. For example, Up the
tree, climbed Jason has a different syntax than Jason climbed up the tree. Semantics is the
meaningful value of the sentence. A syntactically correct sentence may be semantically
incorrect. E.g., The rain ate the iced tea is syntactically correct but semantically meaningless.
Semantics deal with anomaly, ambiguity, self-contradiction, synonymy and entailment.
Pragmatics involves the practical and social aspects of communication eg. Pausing to let the
other person speak, using polite words like ‘may’ when the person is a stranger

Language Comprehension

Language comprehension includes speech perception, our remarkable ability to understand

the spoken word despite the paucity of reliable cues, the omission of certain phonemes and
the continuous nature of speech. It also includes reading - comprehension of the written word.
To understand spoken language, there are three levels of processing at which language can be
analysed. At the molecular level, sounds must first be registered in the auditory system
(perception), then these sounds activate representations of the corresponding concepts on our
memory (pattern recognition) (Robinson-Reigler & Robinson-Reigler, 2008). At a higher
level, we recognize ‘morphemes’, the meaningful units of language. Several theories have
been proposed for how we recognize these words (e.g. Cohort model).
Constituent structure (phrase structure) is the breaking up of a sentence into its parts
(constituents) and how they relate to one another. This process worked well for analysis of
certain sentences but lacked clarity when it came to certain ambiguous sentences. Hence,
Chomsky proposed his deep structure and surface structure theory.
One of the defining characteristics of language is grammar – the system of rules that allows
one to take a finite set of symbols and combine it in an infinite number of ways. Noam
Chomsky’s approach to how we engage in language is called transformational grammar. The
main aspect of this theory is that there is an implicit, innate set of rules in language which
apply to the syntactic structure of sentences. Transformational grammar theory, as applied by
Noam Chomsky, proposes that each sentence has two levels of representation, a deep
structure, which is the ‘idea’, and a surface structure, which is the concrete representation
used to convey that idea. We apply /‘transformational’ rules to produce language (convert the
deep structure into a surface structure) and understand language (covert surface structure to
deep structure).
Sentence comprehension, one of the cornerstones of language comprehension, is a
complicated task. When we read a sentence like ‘The boy ate the candy’, we not only
comprehend the meaning of the sentence but also its structure – the subject (boy), the object
(candy) and the connecting verb ‘ate’. For this, we must first sense the letters and words; then
perceive them and encode them. The sentence is usually processed from left-to-right, with
each word being processed sequentially. While holding the sentence in our working memory,
we must then successfully recognize each word in the sentence by retrieving the relevant
word meanings. We must also discern the grammatical structure of the sentence so that each
word’s meaning can be related to the other to comprehend the sentence as a whole (sentence
parsing). For example, ‘My child is very dear to me’ and ‘vegetables are very dear today’
represent the same word ‘dear’ used in different contexts. Finally, various studies have shown
that when we process a sentence, we discard the exact words of the sentence and only retain
its ‘deep structure’ or meaning, that is, a ‘gist’ of the sentence. Therefore, sentence
comprehension involves processing at the morpheme (discussed earlier), syntactic and
semantic levels. The information processing approach suggests that the syntactic and
semantic processing take place serially whereas the interactionist view suggests that they are
parallel. Recent research has generally supported the interactionist view.

According to Matlin, 1995, three factors affecting sentence comprehension. Voice - active
voice is more easily processed than passive voice (Hornby, 1974). Secondly , people take
more time to process ambiguous sentences (e.g. Foss, 1970). Ambiguity has further
components. Lexical (semantic) ambiguity deals with words that have more than one
meaning. For example, ‘There are many bugs in the room’ – does it mean insects or recording
devices? It also includes words that have different spellings but the same sound e.g. The
son/sun was up early in the evening. This type of ambiguity affects comprehension more
when the sentence is spoken rather than read. Structural ambiguity occurs when a phrase or
sentence has more than one underlying structure. It includes two further types: surface
structure (or phrase structure) ambiguity, for example, He saw the old Indian dance. This is
surface structure ambiguity because the ambiguous phrase in the sentence can be represented
in two structurally different ways – He saw the [old Indian] dance OR He saw the old [Indian
dance]. An example of deep structure ambiguity is the sentence ‘It’s too hot to eat’. Does it
mean that the weather is too hot to eat or the food is too hot to eat? Here, the comprehension
of the sentence depends on whether or not the implicit subject of the infinitive clause 'to eat'
is tied the subject (‘it’) of the main clause. Another example of deep structure ambiguity is
‘Biting dogs can be dangerous’. Another type of ambiguity is garden path sentences like
‘Tom weighed 350 pounds of grapes’ which lead the comprehended to one interpretation at
the beginning and suddenly change context at the end. Thirdly, the presence of a negative
participle in the sentence also affects its comprehension. This is the aspect of sentence-
comprehension that we are currently interested in. Research on comprehension of negative
sentences has been extensive – Sherman (1976) showed how with three negatives ‘Few
people strongly deny that the world is not flat’, the sentence almost becomes
incomprehensible (Matlin, 1995). Taylor and Taylor, 1990 also showed that the presence of
an implied negative in a sentence always increases processing time. This showed, that our
cognitive ability deals better with something that exists (positive) rather than something that
doesn’t (Matlin, 1995). Other research on the comprehension of negative sentences has been
discussed in further detail below.
Most of these researches used a sentence picture-verification task, which is a common
technique to investigate the modalities of sentence comprehension, although there are some
criticisms to its use (Tanenhaus). The process of sentence-picture verification occurs in
various important cognitive tasks such as pattern identification, problem solving and others as
well as in psycholinguistics (Clark and Chase, 1972). One of the most important concerns in
psycholinguistics is how mental representations of any sentence read or spoken are made.
Investigations into the psychological processes involved in sentence-picture verification
began in the mid-1960’s (Keenan, 1982, 2001).
In this task, a human participant is presented with a sentence that describes the spatial
relationship between geometrical figures and signs (oral or written) and then presented with a
picture which depicts a certain relationship. The participant is asked to respond whether the
sentence and picture correspond or not. The Independent Variable is the framing of the
sentence and its Truth-value and the dependent variable is the participant’s Reaction Time
and the number of errors made. It is believed that a comparison of the Reaction Times and the
number of errors for the different types of sentences (four different combinations are formed:
True-Positive, False-Positive, True-negative, False-negative) will yield important insights
about the way sentence comprehension occurs in the human brain. The basic paradigm of
sentence-picture verification was developed by Clark and Chase in 1972.
Before mentioning the results from this and other studies, it is important to discuss a study by
Just and Carpenter (1971) where three types of negatives were specified. – a syntactic
negative (SYN-NOT e.g. The dots are not red, none of the dots are red, there are no red dots),
another syntactic negative that is less obvious (SYN e.g., Few/scarcely any/hardly any dots
are red) and a semantic negative (SEM - About 2 dots out of 16 are red; A small proportion
of the dots are red). In the categories SYN-NOT and SYN, true-affirmative sentences had a
shorter RT than false-affirmatives. The true-negative sentences had a longer RT than false-
negatives. (TA<FA, TN>FN). With the SEM sentences, both affirmative and negative
sentences had shorter RTs when they were true than when they were false. TA, TN < FA, FN.
The mean RT for negative sentences was greater than for affirmative sentences in all three
categories. Overall, the RTs increased from SYN-NOT to SYN to SEM. More errors were
made to negative than to affirmative sentences, with true negative sentences accounting for
majority of errors. (Main findings by Just and Carpenter)
The two basic conclusions drawn from the various studies on sentence-picture verification
are: (Just & Carpenter, 1971; Chase & Clark, 1972; Trabasso, Rollins, &Shaughnessy, 1971)
1) there was an interaction between the kind of sentence (affirmative or negative) and
the truth-value (true or false) of the response. Specifically, true affirmatives were
always verified faster than false affirmatives (TA<FA), while false negatives were
always verified faster than true negatives. (FN<TN)
2) The second main result was that negative sentences took longer to verify than
affirmative sentences (A<N).

Clark and Chase (1972) presented one of the first models to explain these results.
a) They proposed that sentences are represented in terms of elementary
propositions; pictures are encoded in same interpretive format; both are
compared in algorithmic series of mental operations, each of which contributes
additively to the response latency; sentence encoding, picture encoding,
comparing and responding are four serially ordered stages which contribute
additively to response latency
b) Subjects compare proposition and picture for sameness. If not same, change
True  False, and compare again. If still not same, change False  True again.
Times for these processes are additive. (Two assumptions: principle of
congruence, initial value of truth index is true)
c) They explained their results (TP<FP<FN<TN) by three concepts:
d) They explained the interaction between affirmative-negative and true-false using the
congruity principle (Gough, 1965). Principle of congruence (Clark, 1969) states that
two representations can be compared only for identity and there must be extra
operations in case there is a mismatch. Clark and Chase assumed that the lack of
 congruence between the picture and the sentence required extra processing time and
this extra processing time was referred to as falsification time.
e) Negative sentences were said to take longer to verify than affirmative sentences
because they contained an extra element, th e negative marker. Clark and Chase
(1972) referred to the extra time consumed by processing the negative element in the
sentence as negation time. This involves extra time to code the underlying
proposition as well as decide mismatch.
f) Deep structure assumption – participant represents the semantic (“its deep
structure”, meaningful, interpretational) content of the sentence in a propositional
form. Eg. A isn’t above B = It is false that A is above B = False (A above B). ‘Above’
is less complex linguistically and is coded faster than below; positive is coded faster
than negative. These effects are additive. So, a sentence with below and negative
takes most time to form propositions.
g) There are thus four times which contribute to response latency: Below Time +
Negation Time + Falsification Time + Base Time
In their study, they conducted two versions of the sentence-picture verification task.
In Model A the participant read the sentence before picture. In Model B, the picture
was first – also has Subject Mismatch Time. Participants noted that picture-first
condition harder which confirms the presence of this extra operation)
 Argued against visual imagery model, conversion model, picture-negation model,
reading time, frequency of occurrence because their empirical results did not
support.

Beech (1980) provided an alternative model to Clark and Chase where he contended that that
the imagery model can be used to explain Clark and Chase’ experimental results. He
proposed that subjects produce an image of the sentence and compare it with the picture on
the card. In the case of the negative sentence, subjects produce an image as well as a
proposition which states that the image is false. Thus, he integrated propositional and
imaginal representation to account for the results. He claimed that since his model had fewer
operations involved in processing, it was superior.

Carpenter and Just’s Constituent Comparison Model (1975) remains the most widely-
accepted and detailed model of sentence picture verification. They discovered that the
estimated value for negation time was always either two or four times greater than the value
for falsification time. The fact that negation time was always an integer multiple of
falsification time suggested to them that both parameters were probably being determined by
a single mental operation and that the difference in verification times for different sentence
types could be explained by the number of times this mental operation was executed
(Keenan, 1982, 2001).
a) People form Mental Representations to represent sentences and pictures. These
mental representations are in the form of an abstract conceptual structure called a
proposition. E.g., The star is above the plus would be (ABOVE, STAR, PLUS) =
(ABOVE, STAR). Propositions can be affirmed or negated. E.g., ‘The star isn’t above
the plus’ would be NEG (ABOVE, STAR). Even complicated sentences can be
represented in this way.
b) Carpenter and Just assume that pictures too are represented in this same propositional
format and they are always represented affirmatively. E.g., A picture of…
c) Each proposition is referred to as a constituent, hence the term ‘The Constituent
Comparison Model’ as the next step in processing is that the corresponding
constituents from the sentence and picture representations are retrieved and compared,
 pair by pair, beginning from the innermost constituent. Every time there is a
mismatch, the response index, which is initially always set at ‘True’ will be changed
and the comparison process is reinitialized. Thus, since the verification times are
determined by the number of comparison operations, they will increase with the
number of mismatches. Since the True Positive condition encounters no mismatches
(= 2 comparisons) and the True Negative condition encounters two mismatches (= 5
comparisons), the trend is predicted to be TP<FP<FN<TN. Strong evidence for this
model was provided by further experiments.

Carpenter and Just’s Recoding Model is a supplemental model to the Constituent Comparison
Model which accounts for results where TN are verified faster than FN.
Sentences that contain a negative element can often be rephrased so as to eliminate the
negative while retaining the meaning. This is called recoding. The advantage of recoding a
negative sentence is that the sentence representation becomes affirmative and thus matches
the picture representation which is always affirmative. Recoding typically occurs only after
the subject has had considerable exposure to the sentence-picture pairs because the subject
needs to learn which sentence forms are equivalent e.g. The dots are not red = The dots are
black is not immediately obvious. Carpenter and Just also mention that recoding is
encouraged by "a delay between the presentation of the sentence and the e second source of
information"
According to this model, a TN sentence requires a total of 4 operations whereas a FN requires
5. Thus, TP<FP<TN<FN

MacLeod et al. (1978)

One difference in MacLeod’s sentence-picture verification study was that they studied
Comprehension RT (time taken to comprehend the sentence, as indicated by participant
pressing a buzzer) and Verification RT (sentence-picture verification time) separately. They
also correlated verbal and spatial ability with these times.
They found that the results of the subjects could be divided into two groups. The larger group
fit well into the Constituent Comparison Model but for some of their subjects, the only factor
affecting their verification times was the true/false distinction. Both Verbal and Spatial ability
were good predictors of Verification RT and not Comprehension RT. The predictive ability
of the spatial ability test was unexpected. Thus, they proposed that what these subjects were
doing was reading the sentence and converting it into a pictorial representation; then when
the picture was presented, instead of representing it propositionally, they directly compared
the visual representation of the picture against the pictorial representation they had
formulated from reading the sentence. Thus, the linguistic structure of the sentence (e.g.,
presence of a negative) could not affect their reactions times and time to convert the picture
into a propositional format is reduced from the Verification RT. The only thing that could
affect their reaction times was whether their pictorial representation of the sentence did or did
not fit the presented picture. As predicted, picture-model subjects were thus faster in
Verification than the linguistic model subjects. Thus, TA=TN<FA=FN. Thus, they
demonstrated that two different comprehension strategies could be used by subjects in
sentence-picture verification, and which strategy the subject will use can be predicted by
psychometric measures of spatial ability (measured, in their experiment, by the Washington
Pre-College (WPC) test). Those with a higher spatial ability are more likely to use the visual
imagery strategy, and those with a higher verbal ability are more likely to use the linguistic
strategy.
The Experiential Simulations Account: In their experiment, Zwaan et al (2004) used only
negatively-framed sentences, along with filler affirmative sentences.
 Responses were found to be faster following negative sentences when the depicted entity
matched rather than mismatched negative situation. FN<TN For example, when the
sentence ‘There was no eagle in the sky’ was matched with the picture of an eagle with
folded wings, the RT was greater than when the sentence ‘There was no eagle in the sky’
was matched with the picture of an eagle with its wings outstretched. Thus, it was
suggested that participants simulate the negated condition when processing a negative
sentence.
 This supported their theory that people create mental simulations on hearing a sentence
that are experiential in nature, that is, grounded in perception and action. In fact, they
theorize that creating experiential simulations is integral and even necessary to sentence
comprehension and cite several studies as evidence (Boroditsky, 2000; Glenberg and
Kaschak, 2002). For example, in one study, the probe recognition of a sentence was
greater when the earlier sentence had the words ‘half an hour’ as compared to ‘five hours’
because the subject imagines time on an actual timeline.

In the current experiment, the experimenter conducted the sentence-picture verification

task with an individual subject and then integrated group data from 20 subjects to check
for a specific trend.
AIM:

To study the effect of negatively framed sentences on sentence picture verification time and
errors in verification

HYPOTHESIS:
Sentence picture verification is hindered in the case of negatively framed sentences as
compared to affirmatively framed sentences. That is
1) The Mean Reaction Time (RT) is higher for negatively framed sentences as compared
to affirmatively framed sentences.
2) The number of errors is higher for negatively framed sentences as compared to
affirmatively framed sentences

Null hypothesis: Sentence picture verification is not hindered in the case of negatively framed
sentences as compared to affirmatively framed sentences. That is
1) The Mean Reaction Time (RT) is equal to or lower for negatively framed sentences as
compared to affirmatively framed sentences.
2) The number of errors is equal to or lower for negatively framed sentences as
compared to affirmatively framed sentences.

OPERATIONAL DEFINITION OF VARIABLES

Independent Variable:
Negative vs Affirmative (Positive) framing of sentences. Negatively framed sentences
used the word, ‘not’ to describe the relations between representations of geometrical
figures/ signs/ symbols/ letters/ numbers.

Dependent Variable
Sentence verification measured in terms of:
1) Time taken to respond correctly (RT in seconds with decimals not rounded off to
an integer) that the sentence does or does not describe the relationship between the
geometrical figures/ signs/ symbols/ letters/ numbers in the pictures, and
2) Number of errors

CONTROLS

1) Half the sentences of each type- negative and positive- were so framed that ‘True’
was the correct response, while for the other half, ‘False’ was the correct response.
The four types of cards (T+, F+, T-, F-) were presented in a random order. This was
done to prevent the development of a set to respond in a certain way.
 2) Four counterbalanced sets of cards were prepared in such a way that each
representation of geometrical figures was paired with each type of sentence (Positive-
True, Positive- False, Negative- True, Negative- False) across the four sets. Each set
was used equally often across the group of participants.
3) In each set, 32 sentences were used, out of which 8 each were T+, F+, T-, F-.
4) Each set consisted of 32 cards, with each of the 4 kinds of relationship between the
geometrical figures- Left-Right, Above-Below, Bigger-smaller and Inside-Outside
presented on 8 cards.
5) Two cards depicting the same kind of relationship were not presented successively.
6) The E gave a ‘ready’ signal, started the stopwatch and simultaneously started reading
the sentence aloud during stimulus presentation. Immediately after E finished reading
the sentence, E showed the card with geometric figures on it to the participant over
the screen. E stopped the stopwatch exactly when the participant responded ‘True’ or
‘False’. E did all three tasks with the same precision across all the sentences. (The
sentence was written on the reverse side of the card to be shown to the participant, so
that E could read it with ease, just before showing the card on the screen).
7) The reaction time for each participant was accurately recorded.
8) All four kinds of relationship were demonstrated with the help of 4 sample cards.

METHOD

Participant
Individual data

Name of the Participant Age Gender Class Emotional state

Apparatus and Materials:

1) 32 stimulus cards, each having a representation of geometrical figures/ signs on it and
a list of 32 sentences describing the figures on the cards. 16 sentences were negatively
framed, and the other 16 were affirmatively framed.
2) Four sample cards- the figures on them being , ,? , +
X
and the sentences on the reverse side being, ‘The circle is not to the right of the
triangle, ‘The triangle is bigger than the square’, ‘The question mark is above the
cross’ and ‘The plus sign is not inside the oval’.

3) Digital stopwatch calibrated in milliseconds or showing seconds in fractions (decimal

points)
4) Screen
5) Record sheet
6) Stationery
Design

Repeated measures design with one independent variable having 2 levels. Each participant
was exposed to both levels – negatively and affirmatively framed sentences.
(In actuality, the design is a here is a 2x2 factorial one with Type of Sentence- Positive,
Negative and Truth Value of the sentence- True, False as the two independent variables, for
which a two-way ANOVA would be the appropriate statistical technique which is not a part
of Sem V TYBA syllabus. Hence, the design here is restricted to a single independent
variable and the analysis to t tests.

PROCEDURE:

The E arranged all materials and called the Participants inside the laboratory. She/ he/ they
was/were made to sit comfortable facing the screen, rapport was established and the
following instructions were given:

“This is a simple experiment about understanding sentences. I will read aloud some
sentences, one sentence at a time. Each sentence will describe a relationship between some
geometrical figures or signs. As soon as I finish reading the sentence, I will immediately
show you over this screen, a card with some geometrical figures or signs on it. You have to
say whether the sentence I read out described or did not describe the relationship shown
between the geometrical figures on the card and respond “True” or “False” accordingly. That
is if the sentence described the relationship shown on the card, say “True”, if the sentence
did not describe the relationship shown on the card, say “False”. I will give you a ready
signal before reading each sentence. If on a card, one figure is on the left and the other is on
the right, you have to judge left or right from your perspective, that is left or right as you see
it, not how I see it, and make the response “True” or “False”.

(The E demonstrated with the help of 4 sample cards and observed if the Participant gave the
correct responses – False, True, True and False respectively, and explained further if it was
felt necessary).

Please listen to each sentence very carefully as I read it, because I will not repeat any
sentence again. Have you understood? Shall we start?

After making sure that the participant had understood the instructions, the experiment started.
The E presented 32 cards, 1 card at a time, doing 3 tasks: i) E read aloud the sentences
describing the figures on the card, simultaneous started the stopwatch the moment she/he/
they started reading the sentence, ii) presented the card with the geometrical relationship over
the screen and iii) stopped the stopwatch when the participant gave the response. The
participant’s RT were carefully recorded.

If the participant was not able to hear a sentence clearly because of distraction or a loud noise
or a disturbance from another E or a participant entering or exiting, and there was a large RT,
the sentence was presented again later at the end of 32 cards. This was done to ensure that the
RT in such cases (outliers) would not influence the mean. If the participant was not able to
respond at all because of the same reasons, the sentence was presented again later at the end
of 32 cards to ensure that either “False” or “True” response was obtained on each of the 32
cards.

The E examined the Record sheet for the following (i) the number of times the participant
said ‘True’ and ‘False’, (ii) the number of correct responses and the number of errors, (iii) the
type of sentence(s) and the geometrical pattern(s) for which errors were recorded, and (iv) the
range in RT. E entered the data in Table 1, calculated the four Mean RTs and noted – what
were the shortest and longest RTs; for what kind of sentences were they observed; whether
there was any consistent trend regarding True versus False sentences within a single (- or +)
frame- whether the Mean RT was less for True sentences or less for ‘False’ sentences; and for
what kind of sentences there were most/ least errors.

After entering the data in Table 2, E noted the mean RT for [Positively and Negatively
framed sentences; whether the difference was in the expected direction; whether the number
of errors were more for Negatively framed sentences; whether a single, extreme RT had
influenced the Mean RT; whether there was a speed-accuracy tradeoff observed. Depending
on these observations, relevant Post task questions were asked.

Debriefing: (not more than 5 minutes)

Should include:
- The purpose of the experiment. Aim and hypothesis
- Explaining the experiment with the help of demo cards (explain relationships PN, TF)
- Showing and explaining results to the P; whether results in line/ not in line with the
hypothesis.
- Explain 2 theories of sentence picture verification relevant to the hypotheses and also
relevant to Participant’s results obtained, along with responses to PTQs.
(e.g., Just and Carpenter constituent comparison theory, Clark and Chases model,
Beech’s model, Macleod model’s)
- If results contradictory to the hypothesis, mention some reasons why, also taking help
from PTQs.
- Mention one application of this study
- Ask participants for questions if any? And if time permits, ask them what do they
think an application be for this study.
- Request to maintain confidentiality about this experiment.
 INDIVIDUAL DATA

Name of the Participant Age Gender Class Emotional state

Vidhi joshi 20 F TYBA Stable

Individual data

Table 1: Comparison of Participant’s RT in seconds and the errors made in the

verification of True and False, Positively and Negatively framed sentences.

RT in seconds and errors made

+ Frame - Frame
Trial True Trial False Trial True Trial False
No. Sent No. Sent No. Sent No. Sent
4 4.41 2 7.40 7 ERROR 1 ERROR
5 4.40 8 5.47 10 5.50 3 4.53
9 3.46 11 3.97 12 6.07 6 4.25
13 5.16 16 7.78 14 11 15 5.09
19 4.81 22 4.97 18 4.93 17 5.91
21 3.32 26 3.53 24 6.28 20 6.97
25 4.22 28 3.50 29 5.37 23 5.68
31 3.06 30 4.44 32 5.54 27 6.75
Total
RT 32.84 41.06 44.69 39.18
Mean
RT 4.10 5.13 6.38 5.60
Range 2.1 4.28 6.07 2.72
Errors 0 0 1 1

Note: Only the correct responses are taken into account while calculating these Means

Graphs:
- Line graph drawn to depict the data in table 1
- Bar graph drawn to depict the data in Table 2
Table 2: Comparison of Mean RT in seconds and No. of errors for positively and
negatively framed sentences.
 Mean RT in seconds No of errors
Negatively framed 4.61 0
Sentences
Positively framed 5.99 2
Sentences
Difference = -1.38 -2

Table 3: Comparison of Participant’s RT in seconds and Errors for 4 types of

relationships between geometrical patterns

Types of relationships
Trial Left- Trial Above- Trial Bigger- Trial Inside-
No. Right No. Below No. Smaller No. Outside
1 Error 3 4.53 2 7.40 6 4.25
5 4.40 8 5.47 4 4.44 10 5.50
9 3.46 13 5.16 7 Error 14 11.00
11 3.97 16 7.78 12 6.07 21 3.32
15 5.09 20 6.97 17 5.91 23 5.68
18 4.93 25 4.42 19 4.81 26 3.53
22 4.97 29 5.37 27 6.75 28 3.50
24 6.28 32 5.54 30 4.44 31 3.06
Total 33.1 45.24 39.82 39.84

Mean 4.73 5.66 5.69 4.98

Range 2.82 3.36 2.96 7.94

No. of / 0 1 0
errors
Pointers for Individual data discussion:

1. With reference to Table 2, briefly state the difference in the Mean RT and the Number
of Errors that can be observed between negatively and positively framed sentences.
Was the data in the expected direction? Was the mean RT higher for negatively framed
sentences? Were the errors greater for negatively framed sentences? Which dependent
variable (RT/ No. of errors) showed greater influence of positive vs negative framing?
Were the two DVs necessary? Why/ Why not? (To check for speed-accuracy trade off)

2a As an ancillary observation what does Table 1 show with respect to RT (Mean,

Range)? For ‘True’ and ‘False’ sentences? Was the trend in line with previous
research? (Carpenter and Just’s constituent comparison model TP<TF<FN<TN,
Recoding model TP<TF<TN<FN, Macleod’s Pictorial representation model
TP=TN<FP=FN)
2b Did negatively framed sentences take longer to verify than positively framed sentences
irrespective of whether they were True or False?
2c Were errors more for negatively framed True or False sentences? For what kind of
sentences were most/ least errors observed?
2d Was there a speed accuracy trade-off observed? If yes, justify it with references to
trials on record sheets or tables.
2e As another ancillary observation, What does Table 3 indicate about the RT (Mean,
Range) and the Number of Errors for the four types of relationships between
geometrical patterns.

3. Any specific aspects or details that need attention, linking these to observations during
conduct as well as the post task responses.
3a On how many trials were very long RTs observed? Was the Mean RT influenced by a
single, extreme RT? Were the cards presented again if a very long RT was observed?
Was there an error or long RT when the card was presented again?
3b Did participant’s performance correspond with their post-task responses? Discuss.
3c. Were controls in place? Were there any sources of confounding? (Manual
measurements of RT, sentence length, reading speed etc.)

4. Explain results obtained with reference to

4a relevant theory (models) and research (refer introduction- Principle of congruence,
Deep structured assumption, the constituent comparison and recoding models, the
experiential-simulations account and Just and Carpenter and Macleod, Hunt and
Mathews studies. Which model would best explain P’s results?)
4b Participant’s strategy- representation and comparison process used (linguistic or
visual?)
Pointers for Group data discussion:

1a) With reference to tables 4 and 6, describe the results obtained (Means and SD) for RT.
Interpret among similar lines as the Individual data Table 2. Which dependent
variable
(RT/ No of errors) showed greater influence of positive vs negative framing?

1b) Which is the appropriate inferential statistic to use here? Why?

(Design: repeated measures design with one IV having 2 levels, DV- Ratio scale,
Parametric statistic used to check whether the difference between the two conditions
was significant or not)

1c) Was the research hypothesis supported? Were the differences observed statistically
significant?

(Hint for reporting: The significance of the obtained t value was found out by referring to the
appropriate statistical table (one tailed test, df= N-1). For Mean RT, t (19) =______,
p<____, one-tailed. Or, t (19) = ____, p >____, n.s. . Accordingly the null hypothesis was
rejected/ retained. For errors, t (19)= ___, p < ____, one tailed Or, t (19) = ___ p > _____,
n.s. Accordingly, the null hypothesis was rejected/ retained)

Note: If the research hypothesis was not supported, discuss with reference to theory (models)
research in the area.

1d) Report ancillary observations referring to tables 5 and 7. What do the tables show with
respect to Mean RTs and Errors for True and False, positively vs negatively framed
sentences? Was the trend in line with previous research (Constituent comparison model/
Recoding model/ Macleod’s picture representation model)

2) Briefly explain the results obtained from a theoretical perspective, comparing the results
obtained with previous findings. In what way is the design and results of this experiment
different from that reported in previous research? (e.g., Clark and Chase, Just and
Carpenter)

Present some methodological issues in the SPV task, for example presenting sentence first or
picture first or simultaneously (Clark and Chase, 1972), the need for separate measures of
negation time, comprehension time, verification time (Just and Carpenter, 1972; Macleod et
al., 1978).
Are there any modifications, improvements and new lines of inquiry that you would like to
suggest? (use of apparatus/ computer to measure comprehension time and verification time,
number of trials, feedback on errors; materials: fewer figures: 16 sentences can be generated
from one figure by interchanging the figures in the sentence, for example, eight sentences (i)
Star is above plus, (ii) Star is below plus, (iii) star is not above plus (iv) star is not below plus
(v) Plus is above star, (vi) Plus is below star, (vii) Plus is not above star, (viii) Plus is not
below star, and pairing each sentence with one of two pictures: either a star directly above a
plus or a plus above star))
Discuss the practical relevance of research in this area (writing or speaking in positively
framed sentences rather than in negatively framed sentences).