The Journal of Space Syntax: Is Architectural Form Meaningless?

The Journal of Space Syntax
The Journal of Space Syntax

ISSN: 2044-7507
Year: 2011. Volume:2, Issue: 2
Online Publication Date: 15 December 2011
http://www.journalofspacesyntax.org/
Is Architectural Form Meaningless?
A Congurational Theory of Generic Meaning in Architecture, and its Limits
Bill Hillier
The Bartlett School of Graduate Studies
London, UK
Pages: 125-153
Journal of Space Syntax, Volume 2, Issue 2, Pages 125-153, 2011
A Congurational Theory of Generic Meaning in Architecture, and its Limits
Bill Hillier
The Bartlett School of Graduate Studies, London
Once we have re-interpreted the optical image into a conception of space en-
closed my mass, we read its purpose from the spatial form.
(Frankl, 1914, p.1)
1. Limits to the argument
This paper is about meaning. Its aim is to outline something like a confgurational` theory oI mean-
ing in architecture. It can be thought oI as a theory oI generic meaning, by analogy with the theory
of generic function in architecture, outlined in Chapter Eight of Space is the Machine (1996, p.216-
261). Generic function meant the basic acts that people carry out in buildings beIore we consider the
contents or purposes of their acts, so occupying space and moving in space are generic functions.
Generic meaning is equally basic, and reIers to certain simple social ideas that the masses and elements
that make up the physical Iorm oI a building, especially its Iaade, can convey by being confgured
in one way rather than another.
There is a caveat. A theory oI meaning, generic or otherwise, does not take us very Iar. Perhaps
the most useIul outcome oI the paper would be to set limits oI the idea oI meaning in architecture.
It aims to identiIy these limits by distinguishing the idea oI meaning from the idea of the aesthetic
125
Order, proportion, dehnite delimitation, and simple structure are usuallv
taken as the characteristics of beautiful objects; yet these characteristics are
obviously insufficient to comprehend all the elements which make up the
aestheticallv signihcant and effective. The dehnition fails to cover a whole
class of phenomena whose reality cannot be disregarded by any observa-
tion unless it is dimmed bv theoretical prefudice. The contemplation of beautv
as harmonious proportion and strict unity of form does not awaken in us
the deepest emotions of the soul or the most intense artistic experiences.
A different and stronger emotional effect appears when, instead of unity of form,
we are confronted with its disintegration, even with its complete dissolution.
(Ernst Cassirer, 1951, p.328)
in architecture or even its poetics, though, as we will see, using this term in a technical rather than
rhetorical sense. These concepts, it will be argued, have a Iar greater potential than meaning` to clariIy
what can be conveyed to human minds by the manipulation oI architectural Iorm.
2. Preamble: most theories assume that architectural form is meaningless
We can begin by distinguishing two structurally diIIerent interpretations oI the concept oI meaning`.
We can call them signihcance and signihcation. Signihcance is where we give something like a syn-
tactic meaning to a Iorm by comparing it mentally to other Iorms, as, Ior example when we compare
the Doric order to the Ionic. We can say that the confguration oI the Iorm means` its own pattern
in contrast to the other pattern, and so means itselI`, in the frst instance. Signihcation is where a
confguration is associated with something outside itselI and quite distinct Irom itselI as, Ior example,
when we say that the Doric stands Ior manliness and the Ionic Ior Iemininity. In this sense, signifcation
emulates natural language, where the sequence oI letters or sounds that make up the word tree` signi-
fes something quite distinct Irom itselI, namely a physical entity with a trunk, branches and leaves. In
The Social Logic of Space (Hillier and Hanson, 1984) it was argued though without using the terms
signifcance and signifcation that natural language was unique in giving priority to signifcation
over signifcance, because its Iundamental aim was to make signifcation precise. Signifcance, or the
syntax that arranges words into an order, was a means to that end, and so secondary. It was argued that
this in this sense natural language was the only Iully semantic language. Other language like systems,
such as space, were morphic languages, and in these signifcance took precedence over signifcation,
as these were frst and Ioremost syntactic languages in which the syntactic pattern itselI is the primary
meaning. Here we will add architectural Iorm to the class oI morphic languages, and argue that here
too signifcance takes priority over signifcation, adding that the limited degree oI signifcation that
we fnd in architectural Iorm comes to us through its signifcance, that is, through its syntax.
This distinction refects diIIerent paradigms oI architectural theory and criticism. II we take
the work oI Colin Rowe (Rowe, 1947), Ior example, we fnd that in The Mathematics oI the Ideal
Villa`, he addresses buildings in terms oI their overall composition, Iollowing a long-standing tradition
oI architectural theory in seeking to account Ior similarities and diIIerences through Iormalistic and
even mathematical comparisons. The concept oI meaning` is never mentioned, any more than it is
in such texts as John Summerson`s The Language oI Classical Architecture` (Summerson, 1966) or
Wittkover`s Architecural Principles in the Age of Humanism (Wittkover, 1949), which locate in the
same paradigm. All three are pre-occupied with the signifcance` oI Iorm rather than its signifca-
tion`, Iollowing a long-standing main-line tradition in architectural theory.
On the other hand, iI we look at such texts as Alan Colqhoun`s Form and Figure (Colqhuoun,
1978) we fnd that he addresses the building not through its Iormal structure but through its historically
derived rhetorical fgures`. Figures, in Colquoun`s analysis, are not so much architectural elements
as the way they are stylistically elaborated. Their meaning` is conventional, and so language-like,
rather than natural`. In his essay Figure and Iorm` he says: Bv form I mean a conhguration that is
held either to have a natural meaning or no meaning at all. Bv hgure I mean a conhguration whose
126
meaning is given by culture, whether or not it is assumed that this meaning ultimately has a basis on
nature. He argues that the Iundamental dialectic` in modernism was not between Iorm and Iunction,
but between Iorm and fgure. Colqhoun`s analytic paradigm is then Iocused on signifcation rather
than on signifcance.
Without too much injustice, we could think oI the Rowe (ibid.) view as the mathematical
paradigm, and Colquoun`s (ibid.) as the linguistic. Christopher Wren refects both when he associates
what he called positive beauty` in architecture with geometrical Iorm, arguing that other aspects oI
beauty were conventional and depended on custom. Both paradigms`, however, agree on one thing:
that the ways in which buildings seem to convey social and cultural ideas to observers have nothing
to do with the essential forms of those buildings. The mathematical theory de-socialises Iorm, the
linguistic theory fnds social meaning outside Iorm. Formal theories point to signifcance, and as-
sume it lacks social meaning. Linguistic theories point to signifcation, and assume that it subsumes
all social meaning. Both paradigms agree that in itself, architectural form is meaningless.
But in both cases, it is the paradigm that sets the question and generates the answer, and
commonsense refection suggests that the common answer is in Iact decidedly odd. II we consider
the range oI built Iorms with which we are conIronted at any time in history, it is hard to avoid the
conclusion that a key way in which they convey to us ideas oI cultural and social meaning are exactly
to do with the overall Iorm and composition oI the building. For example, iI we are travelling through
a landscape and see a building in the distance [Figure 1] and ask what is that`, it`s a building` would
be an absurd answer. What we expect is: it looks like an abbey`. We assume that somehow the overall
Iorm oI the building will provide clues to its Iunctional type, and this was the object oI our question.
It is not immediately clear how in the case oI Figure 1 we would arrive at such a correct - judgement
We would note perhaps the continuous line oI the rooI running the whole length oI the building, and
underneath an improbably large, more or less rectangular surIace, uninterrupted by an entrance, and
pierced by a series oI windows more or less in alignment - and perhaps one might Ieel that the align-
ment is not weakened by its unevenness - and to the leIt a curved linear protuberance, running not
quite through two foors, and having two openings aligned one above the other down the centre. It is
not clear that any or all oI these Iactors would be enough to identity the kind oI building it is, but it
does seem to be the case that this building is commonly correctly identifed as what it is, a religious
collective dwelling in Iact the Abbeye de St
127
The Abbey of St. Hilaire in the Vaucluse region of
France
Hilaire in the Vaucluse region of France. Of
course such judgements are also aided by what
kinds oI buildings we might expect to fnd in
diIIerent landscapes, as well as by the scale and
location of the building. But this does not alter the
fact that initial judgements from a distance about
what building is seems to come Irom the Iorm oI
the building itself.
Setting out Irom the Iact that such judgements are normal and commonly correct, it will be
argued in this paper that it is where both theoretical paradigms agree is where they are most obvi-
ously wrong, that is in the assumption that built Iorm is in itselI meaningless, with the implication
that the study oI signifcance and signifcation (in our terms) can be kept in separate boxes. In this
paper, it will be argued that common sense experience - plus a new confgurational` way oI analys-
ing some aspects oI building Iorms - attests not only to signifcance being the most interesting thing
about architectural, but also that signifcance is the primarv source of signihcation, that is, that social
meaning outside itselI` that is conveyed to us by the building.
One reason this plausible idea has not been considered in the past is probably methodological:
there has not been a means oI capturing the confgurational nature oI architectural signifcance in such
a way as to show how it is also the basis oI signifcation. Here we argue that the notions oI confgura-
tion syntax uses as the basis oI the analysis oI space, which seem able to capture key aspects oI the
social appropriation and use oI space, can to some extent be adapted to the analysis oI architectural
Iorms in such a way as to express aspects oI signifcance and signifcation.
3. Exploring the problem
Let me begin with a story. A very long time ago I had a crowded mantelpiece above my fre, and an
enthusiastic cleaning lady. Each week she would take my disorderly array oI objects on the mantelpiece,
and re-arrange them Ior me with the largest in the centre, the next largest two at either end, and other
objects graded Ior size in between, thus maximising bilateral symmetry. On arriving home, embar-
rassed by the symmetry, I would immediately go to the mantelpiece, and muss up` the arrangement,
restoring some approximation oI the original disorder. Now this story has two points. First, you, the
reader, understand it. That is, you understand it not as a simple description oI events but as a story
about the taste oI diIIerent kinds oI people. You have in your heads it seems, some hidden prejudices,
which show not only that you know about symmetry and non-symmetry, but that you are able to give
these ideas in this case a social interpretation.
The second point is more intrinsic to our ideas oI order and non-order. In Iact, when I mussed
up` the objects to eliminate the symmetry, I don`t think I was placing them randomly. I think I was
placing the objects rather careIully, seemingly trying to make the arrangement have as little symmetry
as possible. What might this mean? Are all arrangements that lack perIect symmetry equally non-
symmetrical? Or are some arrangements more non-symmetrical than others? Is there perhaps such
a thing as the least symmetrical arrangement, or even a counter-symmetric` arrangement? Is there
perhaps some mathematical value I was trying to minimise or maximise with my mussing up`?
A key theme oI this paper will be that this is the case. It will be argued that we read objects
and arrays oI objects essentially as confgurations, and that there are generic ways oI confguring
the masses and elements that make up an architectural Iorm both to construct its signifcance` oI
an architectural Iorm, but also the limited signifcation` that buildings possess. In the mantelpiece
128
story, both signifcance and signifcation are clearly present, and we will see that something like this
is usually the case with architectural Iorms.
4. Looking for systematic intent of the architectural kind`
How then should we begin to analyse buildings confgurationally to identiIy their signifcance and
signifcation? There is a Iundamental diIfculty at the outset: that Iormal order in a building can result
Irom the laws oI construction as well as architectural intent. How can constructional order then be
distinguished Irom order intended to carry meaning? We suggest that the answer lies in the defnition
oI architecture oIIered in the frst Chapter oI Space is the Machine (Hillier, 1996, p.10-38). Here it
is suggested that architecture emerges Irom the act oI building when the abstract and non-discursive
confgurational` properties oI Iorm and space are made object oI conscious comparative thought,
leading to systematic intent oI the architectural kind`. Let us proceed by example.
In the Vaucluse region oI the South oI France there is a tradition oI dry stone circular bories`
constructed by corbelling, that is slightly oII-setting a circle oI stones inwardly at each level to defne
smaller and smaller circles until it is no more than a small hole to allow smoke to escape. |Figure 2|
Such buildings in their constructional nature will tend to acquire simple symmetrical Iorms. We also
fnd cases where the basic Iorm has been developed into a external spiral to allow one to walk up the
outside to carry out repairs, a clearly Iunctional adaptation, with visual eIIects that emerge from func-
tionality rather than being assigned by intent |Figure 3|. However, in Figure 4 there is no discernible
Iunctional reason either Ior the height oI the spire` or the asymmetry and scaling oI the entrances in
relation to the spire. The symmetry oI the spire is constructional, its height is not. It, thereIore, looks
as though it may be primitive architecture in the sense oI systematic intent oI the architectural kind`.
129
Figures 2-4. Dry stone circular 'bories', Vaucluse, France.
In contrast, consider Figure 5, an image oI a village in the Tarn area oI France. Here we see a
selection oI houses looking at their gable ends`. Because rooves tend to symmetry, each has an initially
symmetric Iorm. But iI we consider the largely asymmetrical placing and shaping oI windows, and
the ways in which rooves are extended to Iorm asymmetric shapes, we see no evidence oI a mental
intent to conserve or elaborate symmetry. So we do not read systematic intent oI the architectural
kind` into these Iorms.
Now consider Figure 7, a shrine to someone who died on the road beween Athens and Sounion,
and Figure 8, a church in Xoximilco, Mexico. Each can be described in more or less the same abstract
terms as the Greek chapel, in spite of the differences in function and content, it is this abstract scheme
that seems to give the Iorm its signifcance, and in each case leads us to anticipate its signifcation as
expressing some notion oI the sacred. The confgurational Iorm seems to carry the meaning` and the
130
Figure 6. Greek chapel.
Figure 5. Village in Tarn, France
Now consider Figure 6, a simple chapel in
a Greek island. Although the symmetry oI the Iorm
is Iar Irom perIect, its impact is heightened by two
things: frst the alignment oI elements the door, the
light, the cross along the axis oI symmetry, rein-
Iorced perhaps by the absence of such elements on the
two lateral masses that construct the bi-lateral sym-
metry: and the elaboration oI the two rooI diagonals
so that each part complements a similar part on the
other side. We could say that the pair oI rooI lines, by
being elaborated in the same distinctive way, call at-
tention to each other (as Roger Scruton says: provide
reason Ior each other` Scruton, 1979), and so draw
attention to the albeit imperIect symmetry oI the
composition, as perhaps also do the twin fgureless
masses in either side oI the axis oI symmetry.
fgural content is simply the means to this end. We have, it seems, something like a formal genotype
Ior the expression oI some idea oI the sacred (perhaps not the only one, but a common one), and it is
this that gives the Iorm both its signifcance and its signifcation.
Figure 9 is then a view a comparatively unIamiliar one - oI the main entrance to Le
Corbusier`s Chapelle de Notre Dame de Haute at Ronchamp. At frst sight it bears little obvious
resemblance to any ecclesiastical precedent or cultural tradition. But Ior the entrance confguration,
the compositional propositions that we explored Ior the Greek island church and the other two cases
also seem to hold in large part. The twin uninterrupted masses either side oI the entrance, and the
sequence oI diIIerentiated element on the central axis, including the empty space at the top, re-express
in a novel but completely clear way the genotype oI the sacred`. The repetition oI the same confgu-
rational genotype in the tower, again with diIIerent contents and this time accentuated by a deliberate
asymmetry, leave us in no doubt that this is a religious building.
131
Figures 7, 8. On the left, a shrine in Greece;
on the right, a church in Xoximilco, Mexico.
Figure 9. Le Corbusier's Chapelle de Notre Dame
de Haute, Ronchamp; the main entrance.
In all these cases, the means by which the passage through signifcance to signifcation is
achieved is by controlling the shape oI the building, and elaborating it with fgures in such a way as to
give clear evidence oI systematic mental intent oI a certain kind. It could even be said to be a matter
oI emphasis: certain aspects oI the Iorm and certain relations between the elements are picked out
for emphasis but not others. We can illustrate the principle in a simple diagram [Figure 10]. Reading
Irom the leIt, we see a square shape that is, among other things, bilaterally symmetric, but we scarcely
notice it. But iI we add the inverted V-shape oI the gable, we make the bi-lateral symmetry much more
obvious. We can then emphasise the axis oI symmetry much more by placing fgures along it which
are themselves bilaterally symmetric. The result is a Iorm with clear systematic intent`. On the Iar
right we achieve a similar eIIect by diIIerent means.
Figure 11 then takes the basic Iorm with its rooI and adds a series oI fgures, which are
all symmetric, but which have diIIerent relations to the axis oI symmetry. The cases on the top line
all obstinately look like houses, because the fgures are symmetric without pointing to` the axis oI
symmetry. In the bottom line, reading Irom leIt to right, we have frst a suspicion and then complete
clarity that the building has a sacred intent as the genotype oI the sacred` becomes clear as the shape
132
A square shape is bilaterally
symmetric, but we scarcely
notice it.
But if we add the inverted V
shape, we make the bi-lateral
symmetry much more obvious
We can then emphasis the
axis of symmetry much more
by placing figure along it
which are themselves
bilaterally symmetric. The
result is a form with clear
'systematic intent' . On the
right we achieve a similar
effect by different means.
In information theoretic terms, the shape and the figures are all saying the same thing. The more we develop this, the more redundant we are
making the message, and the more obvious we are making the configurational 'meaning' . This is what we tend to do when we want to give a
form signification. We make everything say the same thing, so that each element is predictable from the others. We add nothing new by
addiing more figures once we have made it clear. This is why pure and simple forms can be more powerful than figured forms.
and the fgures, and the blank space either side
oI the axis, come to say the same thing`, so, in
information theoretic terms, adding structural
redundancy to the message. In this way, we have
moved through signifcance to signifcation. We
note that it is not the symmetry oI the Iorm that
does this, but the relations between the Iorm and
fgures and the axis oI symmetry.
Figure 10. Emphasising symmetry.
Figure 11. Figures within the basic form with differ-
ent relations to the axis of symmetry.
Now consider another building oI Le Corbusier: the monastery oI Sainte Marie de la To-
urette. [Figure 12] The abstract description we were able to give oI the Abbeye St Hilaire in Figure
1 seems to capture the confguration oI the Iorm oI la Tourette in the same minimalist way that the
Greek chapel captured that oI the entrance Iaade oI Ronchamp. For the main bulk oI the building,
a powerIul horizontal alignment dominates a large entrance-Iree rectangle oI aligned windows, this
time at several levels, separated by diIIerent repetitions oI the horizontal element; while to the leIt we
see a vertical protuberance centrally aligned in an otherwise pure mass, a clear case oI the genotype
oI the sacred`, but this time Iorming only part oI the whole complex. As with St Hilaire, these broad
and simple compositional points seem suIfcient to ensure that the building reads as a religious dwell-
ing. The main structure has a bilateral symmetry, but we do not notice it because nothing points to it.
At the same time, the unprecedented vertical fgure on the leIt building, not in itselI but in its axial
relation to otherwise blank masses, is suIfcient to establish what this part oI the overall building is:
its chapel.
These two simple themes, either singly or together, both based on the line, but one vertical
and orthogonal to the earth and the other horizontal and parallel, seem to be involved to a substantial
degree in the ways in which we seek to impose order on buildings in order to make the passage through
signifcance to signifcation. At the most primitive level, the vertical and orthogonal central axis and
the horizontal parallel axis seem to be the two simplest ways oI taking a collection oI elements and
making them read as a single object. In Figure 13 we take a square elevation and join another to it
as its neighbour. It is unclear iI it is one object or two. We then add a vertical line along the axis oI
symmetry. Now it looks like a single object because pointing up the axis oI symmetry makes the two
lateral elevations look like similar objects in a similar relation to each other, and to the axis oI sym-
metry. Then, instead oI adding a vertical line orthogonal to the earth we add a horizontal line parallel
to the earth. Again, the two initial squares come to look like similar elements in similar relations, so
again we have an overall order, and again this makes the composite object read as a single object.
133
Figure 11. Le Corbusier's monastery of Sante Marie de la Tourette.
As simple as they are, these two linear notions seem to be central to the passage through
signifcance to signifcation in many architectural traditions. For example, the European classical tra-
dition uses both themes more or less equally, and both separately and in combination. To take almost
arbitrary example, Figures 14 and 15 show the horizontal theme in the town hall, representing the
political collectivity, and vertical theme in the church, representing the ideology, in the main square
in San Gimignano, while Figure 16 shows both themes in the horizontal residential side buildings
and the vertical building representing the university at the end oI the campus at Charlottesville. These
examples could be endlessly duplicated, all showing the passage Irom signifcance to signifcation
through Iormal organisation based on verticality or horizontality.
134
Join another to it as
its neighbour. Is it one
object or two?
It is ambiguous.
Take an elevation
Join another to it as its
neighbour . Is
it one obj ect or two ? It is
ambi guous.
Then add a vertical line along the axis of
symmetry. Now it looks like a singl eobject
because pointingup the axis of symmetry makes
the two elevations look l ike similar objects in a
similar relation. In other worde we have made it
look li ke an 'order'. It i s this that m,akes it read as
onewhole object.
This ti me instead of adding a vertical line
orthogonal to the 'earth-l ine' we add a
hori zontal line parallel to theearth-line. Again,
the two initial squares come to look l ike
similarelements in similar relations so again we
have an overal l 'order, and again this makes
the composite object read as a single object.
This time instead of adding a vertical line
orthogonal to the 'earth line' we add a horizontal
line parallel to the earth-line. Again, the two initial
squares come to look like similar elements in
similar relations, so again we have an overall
'order', and again this makes the composite
object read as a single object.
Take an elevation Take an elevation
Join another to it as its
neighbour . Is
it one obj ect or two ? It is
ambi guous.
symmetry. Now it looks like a singl eobject
because pointingup the axis of symmetry makes
the two elevations look l ike similar objects in a
similar relation. In other worde we have made it
look li ke an 'order'. It i s this that m,akes it read as
onewhole object.
This ti me instead of adding a vertical line
orthogonal to the 'earth-l ine' we add a
hori zontal line parallel to theearth-line. Again,
the two initial squares come to look l ike
similarelements in similar relations so again we
have an overal l 'order, and again this makes
the composite object read as a single object.
symmetry. Now it looks like a single object
because pointing up the axis of symmetry makes
the two elevations look like similar objects in a
similar relation. In other words we have made it
look like an 'order'. It is this that makes it read
as one whole object.
Figure 13. Collection of elements read as a single object.
Figures 14, 15 and 16. Horizontal and vertical themes in architecture;
staring from the left, the Town Hall and church in San Gimignano, and on the right
the University campus in Charlotesville.
OI course, we have pointed out nothing that is not Iamiliar and clear, iI not always explicit,
to any student oI architecture. As we said at the beginning, an analysis oI signifcation in architecture
does not take us very Iar. But two points are interesting. First, the passage Irom building to signifcation
seems clearly to be by way oI signifcance, that is, through the syntax oI the Iorm. In the context oI
twentieth century architectural theory, this is a non-trivial proposition. Second, the two Iundamental
themes we have identifed, vertical and horizontal lines, seem, when they appear as dominant themes
in an architectural composition, to point in diIIerent social directions. The vertical line seems always
to point to an idea rather than to persons, to an absence perhaps, while the horizontal line seems to
point to collectivities, and to their presence. In one sense we could say that the distinction refects the
philosopher`s distinction between intension (the idea that denies a category) and extension (the enti-
ties that belong to that category). In another, we could say that it refects Marx`s distinction between
the ideological superstructure oI society (the vertical line) and the juridico-political superstructure
(the horizontal line). A chapel is a building representing an ideology. A monastery is a building rep-
resenting a political collectivity. We will see how these conjectures Iare as we try below to give more
architectural and numerical precision to these ideas.
5. Regular shapes as congurations
Can these ideas be taken Iarther by, as with space, capturing the logic oI confgurations in a numeri-
cal way? In principle, this seems possible. It was shown in Space is the Machine that the elementary
defnitions oI confguration Ior space could be exactly reproduced Ior the arrangement oI 3-dimen-
sional physical Iorms |Figure 17| because above` and below` are asymmetrical with respect to the
earth, just as next to` is symmetrical. It Iollows that measure oI the syntactic integration` oI a Iorm
should be straightIorward. It was also shown that notions oI symmetry could be precisely reproduced
in confgurational analysis. For example, by treating a square as an arbitrarily fne tessellation, there
would be as many identical j-graphs as there were symmetries in the Iorm |Figure 18|, so arriving at
an internal description oI symmetry, in contrast to the more Iamiliar external defnitions. Since we have
shown that the passage Irom signifcance to signifcation was dependent not simply on symmetries,
but on relations between symmetries, would it be possible to develop measures oI these.
135
Figures 17, 18. Symmetries in a form.
We can begin by looking at some simple shapes and considering the relation between their
measurable properties, and how we read or recognise them. It turns out that we can use this to dem-
onstrate in a very simple way why the diIIerence between signifcance and signifcation is important
to how we recognise objects. Figure 19 shows three fgures, which are constructed by arranging
thirty square elements in diIIerent ways. What does it means to say that we recognise these objects?
It seems to happen in two stages. In the frst stage, we identiIy a shape, in the second we assign that
shape to a category, that is, we give it a name. In the frst two shapes, we see just two shapes. We eas-
ily recognise the diIIerence between the two shapes, that is, we readily make a purely confgurational
distinction between the two objects. But we have no category to which we can assign either object.
The process oI object recognition is thereIore ended at the frst stage. In the third shape we also see a
shape, but this time we conjecture a category: the shape looks like an over-regularised humanoid, so
we conjecture it is meant to be either a robot, a caricature human, or perhaps a toy. That is we have a
second, semantic stage oI recognition built on the frst.
The frst is oI course the syntactic, or signihcance, stage of object recognition, and the second
the semantic stage, or signihcation stage. In both stages, Iorms oI knowledge must be used, which are
both non-discursive and abstract. The frst stage must use something like knowledge oI confgurational
possibility in order Ior shapes to be distinguishable. II it were not so we would not be able to tell
the diIIerence between semantic-Iree patterns. The second uses established knowledge oI categories
codifed by language. What is it then that we recognise at the syntactic` stage oI the process, that is,
what does it mean to recognise a confguration.
One approach to this, which has proved its useIulness in the understanding oI space, is to
reverse the question and ask what properties confgurations have that might allow them to be rec-
ognised? One way that immediately suggests itselI, since we have already used it extensively, is to
analyse the confgurations as distributions oI total depth, or integration, values. This gives us several
kinds oI useIul inIormation about the confguration. First, there is the distribution oI integration in
136
md = 4.609 si = .7 md = 5.604 si = .867 md = 4.073 si = .367
Figure 19. Different arrangements of thirty square elements.
md=4.609si=0.7 md=5.604si=0.867 md=4.073si=0.367
each Iorm, as shown by the red to blue pattern. Following the lessons learned Irom the distributions
oI these values in the analysis oI spatial patterns, this can be thought oI a structure within the shape.
Second, there are the integration characteristics oI the Iorm, as indexed by the mean depth values, as
shown beneath each Iorm. We see that the third shape is more integrated than the second, which is
more integrated than the frst. These depth values seem to correspond to certain intuitions we have
about the forms.
However, there is another intuition, which is not expressed in these measures. It is obvious
that the third shape is more symmetric` than either oI the other two, since it has the property oI bilat-
eral symmetry. However, while the frst and second both lack Iormal symmetries, they do not seem
to be entirely equivalent Irom this point oI view. The second fgure seems to have a greater degree oI
irregularity than the frst, in some sense that does not seem too Iar Irom the distinction between both
and the third shape. In eIIect, the frst shape seems to be closer to symmetric organisation than the
second. Can we quantiIy this property?
6. A symmetry index
Let us look more closely at our internal` defnition oI symmetry based on j-graph isomorphism.
Suppose, Ior example, we did not require j-graph isomorphism, but the same total depth. This would
seem to oIIer a weaker Iorm oI a symmetry-like property. For example, iI we load a simple linear
shape with two sets oI Iour by two cells, one horizontal, the other vertical, but each joined to exactly
We can then apply this analysis to the three shapes shown in Figure 19 in Figure 21. Each
time a shape has cells with identical total depth values we mark it with the same number, Irom the
most to the least integrating. We see that the frst shape has Iar more equal total depth values than the
second, and in the frst the equal values reach well into the integration core oI the shape, whereas in
the second they are distinctly peripheral. Both oI these properties, as well as the degree oI integration,
can be represented through a simple statistical device: the line chart in which each shape is repre-
sented by a series oI i-values, plotted Irom most to least integrated (shown as least to most depth),
together with a series representing the six by fve rectangle (shown as circles) to provide a baseline
Ior comparison. The frst shape is represented as diamonds, the second by triangles, and the third by
137
two cells in the basic Iorm, the two end shapes
created will have diIIerent distributions oI total
depth values, but all the values in the basic shape
are paired in that each cell has exactly one other
cell which is symmetrically` located and has the
same total depth. This total depth equality seems
to give a precise meaning to the idea oI balanced
asymmetry`. |Figure 20| (see Ior example, Tabor,
1982).
Figure 20. A simple linear shape showing 'balanced
asymmetry'.
squares, with the 6x5 rectangle as circles. The overall degree oI integration is indexed by the location
oI the series on the vertical axis. Thus, the rectangle is the most integrated, the third shape next, then
the frst and fnally the second. Also, the shapes diverge as they move Irom integrated to segregated
elements, so that the most integrated elements in each shape are much closer together than the least.
The line charts also show graphically the degree to which j-graphs in the shape have the same
i-value (or total depth Irom others), since identical values will be plotted at the same level and thus,
Iorm a distinct step` in the distribution. The ratio oI the total number oI elements to the number oI
elements that Iorm part oI such lines will index the degree to which values are the same. We can think
oI this as indexing something like the amount oI weak symmetry` (i.e. with the same total depths but
not necessarily isomorphic j-graphs) in the shape. Identical i-values will include both those resulting
Irom perIect symmetry as shown by isomorphic j-graphs, and those that only share the same total
depth. This may be thought oI as a kind oI symmetry index, in which a low value indicates similarity
in how the parts relate to the whole (Iew diIIerent values compared to the number oI objects), and high
values indicates many diIIerences in how the parts relate to the whole (many diIIerences compared
to the number of objects).
7. Buildings as oriented shapes
How useIul are these properties in the analysis oI architectural Iorms. First let us think about shape.
II there is a confgurational logic to shapes, can this be useIully applied to the analysis oI the arranged
masses that make up architectural shapes? This question has oI course always much exercised archi-
tectural theory. One element in the classical theory` has been the idea that pure geometric Iorms lie
behind the capability oI built Iorms to communicate to our intelligence and emotions. The starting
point oI this belieI is probably a passage in Plato`s Philebus in which Socrates is discussing the nature
oI pleasure. Pleasure, he argues, is mixed` where it is interdependent with something unpleasant,
such as relieving an itch through scratching, and unmixed where it comes through the senses Irom the
beauty oI colour and Iorm... smells... and sound`. He goes on: I do not mean by beauty oI Iorm such
138
2
3
4
5
6
7
8
9
6
x
5
r
e
c
t
a
n
g
l
e
6x5rectangle
Fig1acircs
Fig1ccircs Fig1bcircs
si = . 7 si = .867 si = .367
6 x 5 rectangle
si=0.7 si=0.867 si=0.7
Fig. 1 a circs
Fig. 1c circs Fig. 1b circs
Figure 21. Graph of Total Depth Values for shapes in Figure 19.
beauty as that oI animals and pictures, but ..... straight lines and circles, and the plane and solid fgures
that are Iormed out oI them by turning lathes and rulers and measurers oI angles, Ior these I aIfrm
to be not only relatively beautiIul, like other things, but ... eternally and absolutely beautiIul` (ibid.).
This passage has caused endless conIusion in architectural theory, partly through the conIusion
oI these Phileban` Iorms with the regular, or Platonic` solids, but more through the sheer diIfculty oI
accounting Ior more than a small handIul oI built Iorms through these simple geometrical ideas. Even
so, it is hard to rid the mind altogether of the suspicion that simple geometric forms do sometimes
play a signifcant role in architecture. In design terms, also, it is not easy to dispense with the idea
that simple geometric Iorms are implicated in the felds oI possibility that the designer manipulates
in the search for form.
One reason Ior optimism about confgurational descriptions oI shapes is that they take into
account all the space oI the Iorm, that is the periphery, the centre, and all the space in between the
two, rather than just the outline. Confgurational description captures topological properties oI, Ior
example, circular or square Iorms which are not explicit in their geometrical description, but which
can be critical to the way shapes can be useIul to us Ior example, the way we shape and use tables.
Perhaps then, the confgurational analysis oI shapes can lead us to useIul propositions about archi-
tectural forms.
However, we must begin by acknowledging that buildings are not simply shapes, in the geo-
metric sense oI Iree-standing Iorms in a uniIorm context, but oriented shapes, in the sense that they
are oriented towards - and away Irom - the ground on which they stand. For example, experientially
a square Iaade standing on the surIace oI the earth does not read as a shape with 8 symmetries, but
more as a bi-lateral symmetry, which we relate to the bi-lateral symmetry oI our own bodies, and this
is perhaps a principle reason why we cannot use pure geometry to inIorm architectural intuition oI
the feld oI possibilities. This eIIect can be shown clearly by confgurational analysis. On the leIt oI
Figure 22 is a depth analysis oI an 8x8 square shape, showing the pattern oI mean depth Irom red
Ior low through to blue Ior high. It shows the expected centre to edge pattern, with the lowest values
in the centre and the highest in the corners. To its right is the same Iorm with a line representing the
surIace oI the earth and earth-line - added. The pattern oI depth now reads clearly as a bi-lateral
symmetry, a remarkable but liIe-like eIIect.
139
Figure 22. On the left the analysis of a 8x8 square shape;
on the right the analysis with the earth-line added.
Figure 23. The same analysis (as in
g.22) for a 12x4 shape.
II we now take a 12x4 shape Figure 23, and add an earth line, frst to the short side, creating
a vertical Iorm and then to the long side, creating a horizontal Iorm, we fnd equally liIe like eIIects.
In both cases the bi-lateral symmetry oI the square shape has all but disappeared, and in its place we
have, in the vertical Iorm, strong diIIerentiation Irom bottom to top, and in the horizontal Iorm, strong
similarities Irom bottom to top. These visual eIIects are confrmed by the numbers. Figure 24 shows
mean depth and the symmetry index Ior the square and rectangular Iorms without the earth-line. The
square has oI course lower mean depth (more integration) than the rectangle, and also more symmetry.
When the earth line is added to create three forms, [Figure 25] the square becomes less integrated
and less symmetrical (more diIIerentiated), the vertical Iorm becomes a little more segregated, and
markedly more diIIerentiated (less symmetrical), while the horizontal Iorm has become markedly the
most integrated and least diIIerentiated (more symmetrical) oI the three. So verticality segregates and
diIIerentiates, while horizontality integrates and makes symmetric.
1
We believed these numbers refect the intuitive semantics oI these Iorms. In the horizontal
Iorm, insoIar as elements are horizontally related, they will tend to become more similar to each other,
by virtue oI their closeness to the earth-line. This corresponds to the intuition that the more shapes are
aligned along a surIace, the more equal they become. Formally, the relation oI elements approaches
that oI neighbour`, which is a symmetrical relation in that iI a` is b` `s neighbour, then b` is a` `s
neighbour. In contrast, the vertical dimension stresses diIIerence, in that the relations oI above and
below are asymmetrical. Horizontality, we may say, equalises, while verticality diIIerentiates. Both
eIIects will in due course be shown to be critical to the sense that buildings can convey signifcant
information to us through their forms.
140
5.25 .141
6.56 .234
4.91 .462 3.98 .308 6.65 .492
Figure 24. Mean depth and symmetry index for the 8x8 and the 12x4
shapes of Figures 22-23.
Figure 25. Mean depth and symmetry index for the 8x8 and the 12x4
shapes with the line-earth added.
See Note 1
Relevant examples: San
Sebastian Design by Alberti
and Mies van der Rohes
Barcelona Pavilion.
8. Complicating forms and adding gures
Even the simplest aspects oI shape, then, have the potentiality to pass Irom signifcance to signifca-
tion. But so Iar our argument has been confned to orderly Iorms. What about less orderly and more
complex Iorms? And what about the relation between Iorms and fgures? In what Iollows we will
explore some ways in which we can exploit the two confgurational measures oI integration and sym-
metry index to take the argument a little Iarther. Suppose, Ior example we add sub-Iorms, such as
vertical or horizontal sub-Iorms, to the original Iorm. In what Iollows, various ways oI elaborating
regular Iorms will be examined, though always within the envelope oI a 64-cell Iorm with an earth-
line, to preserve comparability with the Iorms so Iar analysed. For example, in Figure 26 horizontal
8x2 sub-Iorms are added to a 16x3 rectangle in various locations Irom edge to centre, and in Figure
27 the 8x2 Iorms are added vertically, again in various locations Irom edge to centre.
The eIIects oI these elaborations on integration are, oI course already known Irom the parti-
tioning` theory Ior space in Chapter Eight oI Space is the Machine (Hillier, 1996, p.216-261). Adding
sub-Iorms centrally will be more integrating than adding them at the edge, and horizontal additions
will be more integrated than vertical additions. However, the eIIects on the symmetry index are less
simple. DiIIerentiation (low symmetry index) is minimised when sub-Iorms are added centrally, but
maximised towards the edge, but not at the edge. The eIIect is most clearly shown in Figure 28 in
which symmetry index is plotted Ior both sequences against the distance Irom the edge counted as the
number oI cells. Both sequences show a curve that rises Irom centre towards the edge, but then Ialls
again as the edge is approached. The curve Iormed by the vertical additions shows that the horizontal
141
.0960 .431
.0963 .600
.0965 .692
.0969 .723
.0960
.431
.0963 .600
.0965 .692
.0969 .723
.098 .677
rra .131 si .369 rra .131 si .413 rra .132 si .446
rra .132 si ..523 rra .134 si .615 rra .136 si .569
Figure 26. 16x3 rectangles: horizon-
tal addition of 8x2 sub-forms.
Figure 27. 16x3 rectangle: vertical
addition of 8x2 sub-forms.
sequence has a higher peak oI diIIerentiation, a higher overall degree oI diIIerentiation, and a more
rapid descent towards more symmetry (lower symmetry indices) at the centre. The vertical sequence
has a lower peak, and lower overall diIIerentiation, but Iollows a similar, though gentler, curve. In
both cases, diIIerentiation is maximised towards the edge but not at the edge. Again, this seems to
agree with intuition.
9. Adding gures to forms: is there an opposite to symmetry?
What about adding fgures to Iorms? Suppose we frst consider the eIIect oI adding a single archi-
tectural fgure`, in the Iorm oI a circular window, to the square Iorm without an earth-line. |Figure
29] It is clear that symmetry will be greatest (the lowest symmetry index) when it is centrally placed.
But where will this value be at its highest, that is, where can we place the fgure to achieve the least
symmetry? The results are Iairly commonsense. Symmetry is minimised when the fgure is placed at
the edge midway between centre and corner. But iI we add an earth-line, as Figures 30 and 31 show
that symmetry is not longer minimised in this location, but when the fgure is close to the edge just
below the horizontal centre oI the Iorm. This does seem to refect intuitions about least ordered Iorms.
The same is the case Ior horizontal shapes with an earth-line. |Figure 32| Symmetry is minimised
when the fgure is placed a little way in Irom the edge rather than at the edge. In contrast, symmetry is
maximised in the vertical Iorm when the fgure is put at the top oI the Iorm, and minimised at the bot-
tom. Figure 33 then summarises the locations where a single circular fgure maximises and minimises
the symmetry index on vertical, square and horizontal Iorms. These results are very preliminary, but
do suggest there is an opposite to symmetry counter-symmetry perhaps though how stable these
results would be under diIIerent scales oI tessellation is uncertain.
142
.35
.4
.45
.5
.55
.6
.65
.7
.75
-1 0 1 2 3 4 5 6 7 8
cellsfromedge
8
x
2
h
o
r
S
I
8x2horSI 2x8vertSI
Scattergram for columns: X
1
Y
1
X
1
Y
2
Figure 28. The effect on the symmetry index by the elaborations
in Figures 26 and 27.
143
md 4.58 si .169 4.65 .492
4.77 .738
md 4.61 si .431
4.74 .415
4.87 .523
4.40 .485

4.44 .758
4.53 .667

4.41 .497
4.5 .409
4.61 .636
4.582 .878 4.52 ..833 4.43 .818
4.449 .758 4.63 .833 4.68 .636
4.40 .485

4.44 .758
4.53 .667

4.41 .497
4.5 .409
4.61 .636
3.88. .591 3.86 .591
3.86 .682 3.86 .636
3.86 .621 3.86 .576
3.86 .424
Figures 29, 30. Adding gures to forms.
6.35 .303 3.86 .424 4.50 .409
6.21 .500
3.86 .682
4.582 .878
.116 .425 .089 .438 .072 .315 .144 .521 .093 .452 .078 .384
Finally we may return to one oI our frst Iorms, the Greek chapel. Figure 37 shows that the
serrated rooI line oI the chapel increases symmetry more than the Iorm shown on the right, and Fig-
ure 38 shows that a circular fgure placed high on the central axis creates more symmetry than two
circular fgures either side oI the axis. A high fgure is also more symmetric than an identical fgure
low on the axis. It seems that the creators oI the Iorms with sacred intent` that we saw were making
correct judgements to increase the symmetry oI the Iorm.
144
If horizontal strips are added to a vertical form, differentiation increases, but if to a horizontal
form, differentiation decreases. NB in all these forms other forms were added to the earth-line
SI .465 SI .496 SI .503 SI .517
SI .380
SI .338
SI .321
SI .316
It is also striking and intuitive when we examine
the eIIects oI multiple uniIorm fgures on the basic Iorms.
In Figure 34 and 35 we add vertical and horizontal fgures to
vertical and horizontal Iorms. For the already diIIerentiated
vertical Iorm, vertical fgures reduce diIIerentiation, while
horizontal fgures increase it. For the already symmetri-
cal - horizontal Iorm, horizontal fgures decrease symmetry
while vertical fgures increase it.
2

Figure 36 then shows that in general iI horizontal
strips are added to a vertical Iorm, diIIerentiation increases,
while iI vertical strips are added to a horizontal Iorm, diI-
Ierentiation decreases. Again, these are powerIully intuitive
eIIects. In general, vertical Iorms minimise integration and
maximise diIIerentiation, while horizontal Iorms maximise
integration and minimise diIIerentiation. This is why the
horizontal Iorms we initially noted in the landscape seemed
so apt at signiIying collectivities oI equals.
3
See Note 2
Relevant example: The AEG
Turbine Factory by Behrens.
See Note 3
Relevant example: Wells
Coates Isokon Building,
Hamstead London.
Figure 36. Effects on differentiation by horizontal additions to vertical and
horizontal forms.
10. Beyond meaning
These are no more than preliminary results Irom what would be a large and complex feld oI research,
and it must be stressed again that, however intuitively clear, it is uncertain how stable these results
would be under diIIerent scale conditions. But the results we have do seem to support three key
conjectures in principle. The frst is that the pathway to social signifcation in architecture is frst and
Ioremost by way oI signifcance, or, put another way, that, unlike natural language, the pathway to
semantics is by way oI syntax. The second is that symmetry has an inverse, one that can play a key
role in architectural meaning as well as in the meaning oI arrays oI objects. The third is that the kind
of meaning that architecture transmits is not meaning in the sense of natural language, but something
else, which we might plausibly call generic meaning`.
What exactly is generic meaning`? To give a clear defnition, we need to look at the relation
between mind and world in which language intervenes. A key relation oI the human mind to the world
is captured by the notion oI attention. When we pay attention to something, we Iocus our minds on
it, and momentarily exclude everything else, at least at that level. But what we attend to is not all we
experience. At the same time as we Iocus our attention we are also aware at a much lower level oI a
vast number oI other things in the feld around us. Part oI the reason we can move our attention Irom
one thing to another so quickly is that we have this low level awareness oI much more than what we
are focusing our attention on.
Attention and awareness seem to work in diIIerent ways and it is tempting to Iollow cognitive
science in describing the diIIerence through a computer analogy. Attention is said to be serial`, in that
it happens in well-defned sequences, while awareness is parallel` in that a large number oI things
happen at the same time, iI at the lower level. The coexistence oI the two is explained in terms oI
computational necessity in the human brain: The contents of conscious experience, which are usuallv
confunctional, must hrst pass through a serial processor that focuses attention on specihc combina-
tions of features. The onlv processors in the human brain that do this are re-entrant thalamo-cortical
loops. Parallel processors outside these loops perform pre-attentive sub-conscious computations, but
145
5.11 .462 4.99 .415 4.66 .363 4.55 .403
Figures 36, 37. Symmetry of the Greek chapel.
serial processors embodied by the loops perform attentive conscious ones. (Kong H More Self than
Self: at Autisms Edge, p.243). Pinker adds: Parallel unconscious computation stops after it labels
each location with a colour, contour, depth and motion. The combinations then have to be computed,
consciously, one location at a time (Pinker 1997, p.141). The sheer combinatorics oI the parallel
world Iorbid its processing into serial Iorm other than highly selectively. Hence, the permanence oI
the attention-awareness pairing.
Now natural language is attention-like in the sense that within a huge feld oI linguistic pos-
sibility, which must in some sense exist in parallel Iorm in our minds, it outputs a serial pattern oI
elements, which represent some kind oI order in that feld. More remarkably, does so with consider-
able precision. On refection, it is surprising that language is able to convey meanings as precisely as
it does. II we bear in mind that nearly all words in language are abstract universals between` is as
much a universal as bird` or thought` and that most words (as dictionaries attest) stand Ior a list
oI possible meanings, and which is to be used only becomes clear in the context in which it is being
used at that moment, it is remarkable that language does manage to output serial patterns oI words
that make such clear, attention-like sense oI the ambient world. Against the odds, sentences in natural
language manage to be phenotypical in that they reIer to the world, as it actually seems to be at any
moment in time.
This is what we mean when we say that natural language is unique in giving priority to
signifcation over signifcance: no other alleged language works in this way. The patterns constructed
by the language` oI space, Ior example, relate to basic types oI human activity, not, other than con-
tingently, to actual events in the world, and this is what we mean by generic Iunction` in space. In
this paper we have seen that something similar is the case Ior the language` oI architectural Iorm.
Meaning is generic, not specifc, genotypical not phenotypical. It reIers to the kinds of things that are
Iound in the real world, not to actual events. The reason this is so is Iundamental. Architecture is not
serial. It is parallel. It exists in the frst instance as non-discursive order in the pre-attention parallel
world. Whatever meaning it has does not take the Iorm oI serial propositions, but that oI pre-attention
intuitions oI order in the parallel world.
This is why the pursuit oI meaning in architecture by analogy with natural language leads
us away Irom the real nature oI architecture and what is communicable by architectural means. But
acknowledging the reasons Ior this can lead us to a much more promising and precise analogy: be-
tween architecture and poetry. There are clear analogies that can be made between what poetry does
and what architecture can do. But, as we will see, the resolution between the two is not on the terms
oI poetry. It is on the terms oI architecture. Poetry, it will be argued, is language used in an architec-
tural way, or, more generally, poetry is language used as a morphic language. The argument I will
construct to support this proposition will depend, critically, on distinguishing meaning` Irom what
I will call the aesthetic. Meaning, it will be argued, is necessary to architecture at the generic level
outlined in this paper, but iI it is taken Iarther in the direction oI signifcation that is, striving to
make architecture stand Ior something outside itselI it tends to trivialise architecture, and distract it
146
Irom its richer purpose in the realm oI the aesthetic`. The defnition oI the aesthetic cannot be given
at this stage, since it is in a sense the culmination oI my argument, and many prior defnitions must
be in place beIore I can attempt it. But put crudely, we can say that meaning` is about trying to make
architecture work like a language in the everyday sense, while the aesthetic is about trying to make
architecture work, in a technical sense, in the way poetry works. To make this argument clear, we
will make comparisons between architecture and both poetry and science as linguistic techniques,
and show how both can be brought together to arrive at a defnition oI the aesthetic in architecture,
which will serve our purposes.
11. Poetry as a linguistic technique
So what is poetry as a linguistic technique? It is oIten thought that poetry is a special kind oI selI-
expression, and so should take the Iorm oI a clear representation oI the poet`s thoughts. This is oI
course a perIectly good way oI looking at natural language, with its serial and attention-like Iocus,
but iI we examine poetry careIully, it seems unlikely that any oI these three aspects selI-expression,
clarity and the pre-existence oI the thoughts expressed are necessary to a defnition. What distin-
guishes poetry Irom the ordinary use oI natural language, according to Mallarme, is that Poetry is not
written with ideas, it is written with words`. By this he means that poetry is not so much the use oI
words to express meaning but the use oI words to create meaning. It uses the rich potentials oI words
in unprecedented combinations, and so creates new and unprecedented meanings. Put another way,
while natural language extracts seriality Irom the parallel world, poetry uses words and combinations
to take the reader or listener back into the parallel world to fnd new connections and new richness.
In natural language, the ideas determine the words; in poetry the words also determine the ideas.
So while the ordinary use oI language limits our awareness momentarily, poetry tends to ex-
pand it. It does so not by pointing to more things in our ambient circumstances, but by using language
to create a more multi-dimensional awareness. It does not replicate the complexity oI our awareness,
but creates a momentary situation in which such a world is brought into existence. It does so with a
specifc technique: to maximise the dimensions oI abstraction (or meaning`) that reverberate Irom
the verbal means that we use. This implies yet another concept oI meaning: one which it is not given
prior to expression, but created out of expression and which probably cannot be transcribed into any
other Iorm. Seen this way, it is clear that the non-transcribability oI poetry is one oI it principle merits.
We cannot create by other means what we fnd in poetry, since that complex oI meanings, Iar Irom
existing in advance and being expressed through the poem, only exists by virtue oI the existence oI
the poem.
This can be Iormulated an inIormation-theoretic terms. The redundancy, or necessary struc-
ture, in speech (or writing), in the sequencing oI letters or words, reIers to those aspects which are
governed by the nature oI the language, that is the rules governing the combination oI letters into
words and words into meaningIul sentences; while those parts not so governed represent the choice the
speaker (or writer) has and so convey the information content oI what is said. The degree oI Ireedom
oI choice exercised by the speaker is thereIore related to the degree to which the hearer`s (or reader`s)
147
uncertainty is reduced by what is said. It was Ior this reason that the inIormation content oI language
could be quantifed as entropy, since it is represented by the degree oI choice that the speaker has
at every point and the relative improbability oI those choices. A successIul poet at each stage oI the
poem is able to substantially expand this feld oI possibility by selecting words outside the normal
Irame oI reIerence given by the transition probabilities in everyday language. That is why one oI the
pleasures oI poetry is the hearer`s surprise at the unexpected use oI words. By using words, the poet
expands the inIormation content oI what is said by increasing the hearer`s degree oI surprise. Put
simply, poetry expands the inIormation content oI language, and so shiIts language towards greater
entropy.
But there is a caveat. The poet does not simply use unexpected words. In two senses this is
not enough. First, the poet must not lose the thread oI meaning, the sense that the words used make
sense in relation to each other. A computer could select a series oI words in which their unexpectedness
in relation to each other is maximised, but they would be unlikely to be heard as poetry because in
the loss oI the continuous thread oI meaning the reader would Iail to see evidence oI poetic intent`.
Just as architecture engages the viewer with the cerebral processes oI the designer by showing in his
buildings, however unexpected, evidence oI systematic intent`, so the poet need to persuade the hearer
that he or she is more or less in control oI the fow oI meaning through poetic intent. So in expanding
the feld oI possible words at each stage oI a text while not losing the thread` oI meaning, the poet
in eIIect re-balances the structure-choice (or redundancy-entropy) content oI language.
12. Architecture as poetic technique
We can say then that whatever its essential nature, the linguistic nature oI poetry is to use language
to create, rather than simply to refect, meaning. Once this is said, it is clear than this is in general
what morphic languages do. They do not simply aim to express a pre-given meaning in the manner oI
natural language, but to create a meaning out of an assemblage of elements and relations. We could
perhaps qualiIy this in the case oI architecture by suggesting there are two pathways oI systematic
intent Irom building to architectural Iorm. The frst is where the possibilities oI building shape and
fguring are used in such a way as to support each other, so that each confrms the eIIect oI the other.
In such cases, the diIIerent layers oI Iorm say the same thing`. The second is where the diIIerent
layers are used in diIIerent directions, so that tensions are created between the diIIerent layers oI
the Iorm. The frst pathway, it might be suggested, is the pathway to meaning. The correspondence
between the layers oI the Iorm eliminates ambiguity, and give rise to the sense that the building gives
a strong sense oI what it is.
The second pathway might then be seen as the pathway towards what I will call the aesthetic`.
The lack oI correspondence between layers creates ambiguity at the level oI any simple meaning, but
at the same time creates more complex possibilities oI adumbrated meanings, in much the same way
as good poetry creates felds oI possible meanings, rather than simple and precise meanings in the
manner oI everyday language. An example is the protuberance in the chapel at La Tourette.
148
Can we Iurther clariIy this notion oI the aesthetic`? Let us start by looking in an unexpected
place: the relation between aesthetics and science. As is well known, science has built into it an analytic
principle, which is also a normative principle: theoretical statements must be as economical as possible
Ior the phenomena they cover. II two theories account Ior the same phenomena, one simple and one
complex, then we must preIer the simple one. As Ockham`s razor has it: entia non sunt multiplicanda
praeter necessitatem (entities are not to be multiplied beyond necessity). OIten this is expressed by
scientists as an aesthetic preIerence: theory 1 is preIerred to theory 2 aesthetically` because it is
more elegant. This is an interesting use oI aesthetic`. It implies simplicity`, or at least economy oI
means. We could hardly say that we preIerred theory I to theory 2 because it was more byzantine`.
The very nature oI science Iorbids us to say that we preIerred a theory, which, while explaining` the
same phenomena, was more complex than another.
This notion oI the aesthetic is oI relevance because there seems also to be an unstated - Ock-
ham`s razor Ior art. Just as the criteria oI the scientifc` implies the simplest possible genotypes (i.e.
theoretical entities) covering the maximum number oI phenotypes (concrete entities), so the criterion
oI the aesthetic` seems to imply the least number oI phenotypes to create the maximum genotypical
complexity that is the minimum oI means in the Iorm oI concrete elements and relations to create
the maximum ends, in terms oI complexity oI meaning. This seems clearly to be true oI poetry. By
the use oI words out oI their normal context and combination, poetry creates complexes oI meaning
or abstract entities which are outside our everyday experience but which by being inscribed in the
poem become part oI our experience. To the extent this is done with an economy oI means, we think
oI it as better technique, and iI there is a superfuity oI means it seems to be much less poetic`.
In eIIect, the poet both takes advantage oI the indeterminacy oI meaning in words, and the
richness oI their potential connotations, to put them in new combinations in which each words takes
on a new liIe by becoming part oI the confgurational context Ior others. So poetry specifcally exploits
the raw materials that language oIIers to do something quite diIIerent, namely to create rather than
to refect meaning, and Ior this to be eIIective it must conIorm to the inverse Ockham`s` principle:
149
The meaning` is simple. The placing oI the element
on the axis oI symmetry against an otherwise blank
background, make the genotype oI the sacred`
unmistakeably clear. But at the same time the Iorm
against which it is placed is strictly rectilinear, and
with a powerIul horizontal line at the top, the whole
part oI a regular box-like structure. This both creates
a tension between the Iorm and its fgure, but also
reverberates in other parallel realms, which have no
reIerence to the sacred, creating a complex semantic.
So architecture as the aesthetic is doing Ior syntax
what poetry does Ior semantics.
4
See Note 4
Relevant examples: Le
Corbusiers Villa Stein and Villa
Savoie, and Mario Bottas House
at Massagno.
maximising the meaning we extract Irom words, and minimising the number oI words. So this inverse
principle is as much in the nature oI poetry as the original is in the nature oI science.
So we might suggest that in general, and in particular Ior poetry and architecture, that the
aesthetic can be distinguished Irom meaning as Iollows: aesthetic objects (including and perhaps
especially architecture) become oriented towards meaning to the extent that they use more phenotypi-
cal means to construct genotypical simplicity and towards the aesthetic to the extent that they use the
simplest phenotypical means to create the greatest genotypical complexity. So in science the rule is
many phenomena with Iew abstractions, while in art the rule is many abstractions with Iew phenomena.
This is why a distinction between meaning and the aesthetic in architecture suggests itselI. Meaning
means using the layered potentials oI architecture in correspondence to clariIy one abstraction. The
aesthetic means using the layered potentials in non-correspondence to create abstract complexity.
13. Finally
One thing is clear: the notion oI meaning in architecture as it has been canvassed on the basis oI the
analogy with language, is largely misconceived, and Ior Iundamental architectural reasons. It is the
very generality oI meaning and its syntactic nature, which conjointly give architecture its most power-
Iul property: its syntactic allusiveness, which, because it is multi-layered, can be made to reverberate
with unexpected reIerences. II meaning in architecture were like natural language this would not be
possible or useIul. Buildings would have to mean` by making everything point one way. By using its
very imprecision architecture acquires the character, which comes to be called poetic. This is not an
imprecise word, not because in this architecture is imitating poetry but in that in using words poeti-
cally language is imitating architecture, that is not conjuring precision oI meaning out oI imprecision
by the use oI context but by conjuring richness oI layered meanings out oI simplicity. In this sense
poetry is language used architecturally, that is as a confgurational language in which the primary
sense of meaning comes from the order and structure inherent in the form itself, not from its links to
some pre-given realm oI meaning. Architecture is Iundamentally a language oI signifcance rather
than signifcation, and insoIar as signifcation is Iound it is only as an outcome oI the pathways oIIered
by the feld oI possible signifcance.
150
151
Figure c3. The AEG Turbine Factory by Behrens.
2
A characteristic example oI a horizontal Iorm
with multiple uniIorm vertical fgures (the third
Iorm in Figure 34) is the AEG Turbine Factory
by Behrens, a maximally integrated and least
differentiated form resembling a compositional
typology oI a Greek temple and intending to
elevate the Iactory type to a cultural standard
[Figure c3].
Notes: Sophia Psarra's Comments
1
Examples oI this gravitational impact oI the earth
line on built Iorm are Iound in many classical buildings
siting` on a platIorm oI steps (known as the stylobate),
which accentuates the bi-lateral symmetry oI a building,
and expresses the anchoring` oI the Iorm onto the earth
as a confgurational eIIect based on marked integration
|Figure c1, San Sebastiano, design by Alberti, 1460|.
Barcelona Pavilion designed by Mies van der Rohe, is
another example in which a raised platIorm is used to
emphasise the horizontal elongation of the form and its
gravitational attachment to the earth-line |Figure c2|.
Figure c1. San Sebastian, Italy,
Design by Alberti
Figure c2. Mies van der Rohes Barcelona Pavilion.
152
Figure c4. Wells Coates Isokon Building,
Hamstead London.
3
The horizontal extension oI Iorms is oIten
encountered in modern housing blocks (though
linear elements oI deck access) as an expression
oI collectivity, reIlecting the belieI oI many
modernist architects that the design of hous-
ing projects were capable oI shaping society
into ideal social communities [Figure c4. Wells
Coates` Isokon Building, Hamstead London|.
4
This lack oI correspondence between layers` can
also be seen at the Iront elevation oI Le Corbusier`s
Villa Stein, where the linear extension oI the ribbon
windows is in tension with the central placement oI the
terrace opening, and the symmetrical positioning oI the
balcony and the canopy either side oI the central axis
|Figure c6|. Villa Savoie is another example construct-
ing a tension between the elongated Iorm oI the piano
nobile (emphasised by the linear extension oI the ribbon
window) and the central organization oI the ground
foor volume |Figure c7|. Finally, Mario Botta`s House
at Massagno is an example where an elongated Iorm is
contrasted by a centrally placed circular opening. The
composition balances not only the tendencies oI linear
extension and central organization, but also symmetry
and differentiation, as the location of the central open-
ing maximizes symmetry but the eIIect oI the sloping
earth adds more diIIerentiation to the Iaade |Figure c5|.
Figure c7.
Le Corbusiers Villa Savoie.
Figure c5. Mario Bottas House at Massagno
Figure c6. Le Corbusiers Villa Stein.
References
Cassirer, E. (1944), Chapter on Art in An Essay on Man`, Yale, 1944; Bantam Matrix edition, 1970.
Ernst Cassirer (1951). The Philosophv of the Enlightenment. Boston: Beacon Press.
Colquhoun, A. (1978), Form and Figure`. In: Essays in Architectural Criticism: Modern Architecture and Historical Change,
Cambridge, MA: MIT Press.
Frankl, P. (1914), The principles of architectural historv, trans. O`Gorman, J. (1968), Cambridge, MA: The MIT Press.
Hillier, B. (1985) Quite Unlike the Pleasures oI Scratching: Theory and Meaning in Architectural Form`, In 9H, p.66-72.
Hillier, B. (1999), Space as a paradigm Ior describing emergent structure in strongly relational systems`. In: Major, M. and
Amorim, L. (eds.), Proceedings of the Second International Symposium on Space Syntax, Brasilia: University oI Brasilia.
Also published (2003) in Seitz F & Terrin J-J (eds.) Architecture des Systemes Urbaines, Paris: L`Harmattan.
Plato`s Philebus, HackIorth R. (ed. & trans.) (1972), Cambridge: Cambridge University Press
Rowe C (1976) The Mathematics of the Ideal Jilla MIT Press
Scruton, R. (1979), Chapter 3, Has architecture an essence` and Chapter 7 The language oI architecture`. In The Aesthetics of
Architecture, London: Taylor & Francis. Those interested could also Iollow up with the Chapter on Frege Irom Scruton`s
A Short History oI Modem Philosophy, ARK Paperbacks (RKP), 1981/5.1; pp 241 - 254.
Summerson, J. (1966), The Classical Language of Architecture, Cambridge, MA: The MIT Press.
Wittkover, R. (1949), Architecural Principles in the Age of Humanism, London: The Warburg Institute
Contact Details:
ProIessor Bill Hillier b.hillierucl.ac.uk
UCL, Bartlett School of Graduate Studies
Central House
14 Upper Woburn Place
London, WC1H 0NN
UK
153

The Journal of Space Syntax: Is Architectural Form Meaningless?

Uploaded by

Copyright:

Available Formats

The Journal of Space Syntax: Is Architectural Form Meaningless?

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Journal of Space Syntax: Is Architectural Form Meaningless?

Uploaded by

Copyright:

Available Formats

The Journal of Space Syntax

The Journal of Space Syntax

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.