|
THE COMMON LANGUAGE OF
SPACE:
a way of looking at the social, economic and environmental functioning
of cities on a common basis.
Professor
Bill Hillier
Bartlett School of Graduate Studies
University College London
Gower Street
London WC1E 6BT
U.K.
phone: (44) - 171-391-1739
fax: (44) - 171-813-4363
e-mail: b.hillier@ucl.ac.uk
ABSTRACT
This paper proposes that in addition to urban research which
seeks to provide answers to policy questions involving the built environment,
there is also a need for research which directly addresses the physical
and spatial complexity of the built environment itself as the main variable
of interest, and explores any effects it may in itself have on the functioning
of the urban system. This type of research reflects the questions architects
and urban designers typically ask, rather than those that preoccupy
planners. For such research to be effective, the physical complexity
variable must be controlled at the level at which real design decisions
are made. Space syntax research attempts to do this by treating built
environments as systems of space, analysing them 'configurationally',
and trying to bring to light their underlying patterns and structure.
Results from space syntax research into the structure and functioning
of cities show a consistency which suggests that space can be used in
this way as a general means of investigating the structure and function
of cities, that is, it may be the common language of the city. On the
basis of this common language, it is argued, it should be possible to
build a domain theory of built environments as structural and functional
entities in themselves, and this will lend greater precision to studies
of its interactions with other domains.
Some questions and answers about complexity
The first sentence in any discussion about the science of cities
usually contains the word 'complexity'. One of the most obvious forms
this takes is the sheer physical and spatial complexity of the city
as an object. There is, however, in most urban research, a strange silence
on this aspect. The reason is simple: no one knows how to control the
physical complexity variable. There is no formal language is which differences
between one form of complexity and another can be described with the
required rigour and consistency, and without controlling the variable
we cannot measure its effects. What we cannot measure we prefer not
to discuss.
Space syntax research about cities (1) seeks to redress this balance.
It addresses first a problem of description: how can the physical complexity
of the city be described with sufficient rigour and consistency to permit
it to be controlled as a variable in research ? It gives what to some
is a surprising answer: that it is best captured by representing it
not as as physical stuff, but as the system of space created by the
physical stuff. This is not as odd as it sounds. Buildings are physical
things, but their purpose is to create the spaces and interconnections
that we use. The effect of every physical intervention is to create
or modify these space patterns. Cities may be aggregates of physical
stuff, but space is the universal stuff which holds the physical stuff
together and gives it its overall form.
On the basis of spatial representations (it turns out that more than
one are needed) of the city, space syntax then asks one question: does
the form the spatial complexity of the city takes make a difference,
and if so, what does it make a difference to ? It seeks answers by one
strategy: it analyses spatial representations of the physical city to
try to understands their structure, and then investigates in what ways
this structure is related to observable function. To the extent that
results are consistent, theoretical explorations become possible. Movement,
land use patterns, social and economic performance, crime patterns,
and many other aspects of function have all been investigated using
this method, with results that suggest that its may be possible to think
of it as a general means for investigating the relation between the
structure and function of cities. Space may indeed be the common language
of the city.
The idea of space syntax originated not in planning but in architecture,
with its need to answer questions about what impact different design
choices about physical and spatial structure are likely to have. The
question addressed by space syntax research is always: what, if any,
is the effect of the built environment in itself on what happens
in cities. The approach can be contrasted with policy-oriented urban
research which seeks to understand what variables, including built environment
variables, are involved in seeking overall social goals (such as energy
conservation), the answers to which are are more likely to be found
in regulation or behavioural change rather than through change to the
built environment, with the protracted timescales that entails.
But although the space syntax research programme originates at the architectural
scale, the questions is addresses are increasingly relevant to cities
as we now see them: that is, not as once-for-all planned objects in
a stable end-state, but as complex global structures which emerge
from innumerable local decisions over a long time scale. In seeking
to describe and analyse space, space syntax seeks to understand the
emergent structure of the physical city, and to account for both
its constructive functional logic and its functional impacts. Although
the main applications of space syntax today (60 major projects in the
past 5 years through UCL's Space Syntax ) are in predicting
the likely effects of architectural and urban design choices, all this
is predicated on its prior ability to analyse urban spatial structure
in a way which is informative about function. Without theoretical knowledge
at this level, applications would be guesswork.
Analysing spatial complexity
Research which seeks to investigate the impacts of built environments
in themselves requires the built environment variables to be controlled
with much greater precision than would be normal in policy-oriented
research. The level of precision is easy to specify. It is the level
at which design decisions are made in real world projects. Taking a
spatial approach allows us to do this. Space syntax models work by taking
some pattern of real space - in cities usually the full street network
- and analysing it using simple mathematical tools that typically relate
all elements to all others up to some limit. We call this approach 'configurational',
defining this as the study of relations which take into account other
relations in a complex (Hillier 1996a, Chapters 1 and 3). This simple
strategy turns out to be quite unexpectedly powerful in detecting patterns
in what might otherwise appear as inchoate complexity. For example,
Figure 1 is a representation of part of the street network of London
as the 'fewest and longest' (i.e. street names are irrelevant) lines
that cover the system. As so often with spatial structures in grown
cities, there appears to be no obvious geometric order, or indeed any
other kind.
Figure1 Axial Map of Central
London
In fact it has a powerful interior logic, which is brought to light
by analysing the system not on the normal basis of treating the nodes
as spatial elements and the street sections between them as (weighted)
links, but by taking the lines as (unweighed) elements, and asking simple
topological questions about their interrelationships, using graph theory.
Figure 2, for example, is a simple 'syntactic' analysis in which each
line is picked up in turn and the 'complexity distance' calculated (that
is the minimum number of intervening lines that must be used, in whole
or in part, to go from one line to another) to all other lines in the
system (that is, the limit of the measure, or 'complexity radius', is
in this case no limit, or radius-n). The map is then shaded dark to
light (red to blue in the colour version) according to the total sums:
the lower the total the darker the line (2).
Figure 2 Global Integration
of Central London
The smallest total in this case is for Oxford Street, which happens
to be the main shopping street (the busiest in Europe) and the next
smallest totals are for the lines that link Oxford Street to the City
of London (the historic core and current financial centre), with the
next few linking these towards the edges of the system in several directions.
We say that this measure indexes the 'global integration' value (its
integration with respect to all others) of each line, and that the structure
brought to light by this measure is the 'integration core' of the city
for radius-n. The pattern shown in Figure 2 seems to correlate with
the intuitive idea we have of the large scale structure of London, with
the West End as the most integrated area, followed by the City, then
a series of areas to the west, south and west. Little of this 'integration
core' is south of the river. Overall, the analysis seems to make some
kind of limited sense in terms of both how London is structured and
how it functions.
The picture is suggestive, but evidently partial. It omits far more
that it includes. However, Figure 3 is the same analysis but carried
out with a restriction of three lines on the radius of the measure.
This measure is called 'local' or 'radius-3' integration, because it
picks out a much more localised structure, though still with Oxford
Street and the West End area dominant. Rather than give a picture of
the gross area structure, the measure seems to give a more detailed
picture of the movement structure, even the sinuous wandering routes
that every Londoner knows are often the best ways from a to b.
Figure 3 Local Integration
of Central London
These are interesting structures, but they are only pictures. Do they
actually make any functional sense ? It turns out that they do, and
in many different ways. To understand this, we must explain a key result
of space syntax research, one that seems to be implicated in many other
results: that the pattern of spatial integration is in the urban grid
is a prime determinant of movement patterns if the system. This sounds
improbable, so the type of study that leads to the conclusion needs
to be explained.
A key result
The technique is simple. An axial map of an urban area and its context
(which must be large enough to account for the pattern of movement into
and out of the area) is constructed and analysed. Figure 4 shows a case
study in the City of London in connection with a major architectural
project. First, the axial map is analysed and a range of numerical values,
including integration at different radii, are assigned to each line
element, indexing by colour (or shading in black and white) for graphical
clarity in the usual way. Observations are then made at different times
of day of movement flows along each street segment by counting people
passing through imaginary 'gates', and indexing them in flows per hour
through that gate. Figure 5. The various spatial values for the lines
are then compared to the movement flows by simple and multiple regression.
Figure 6 shows the degree of agreement between global integration and
rush hour movement rates, with an r-squared of .77. Midday rates correlate
similarly with radius-3. This is a fairly typical result. In most studies
the best performing spatial variable is radius-3 integration (in the
City case study the good correlation with radius-n during the rush hour
is due the to location of the main transport interchanges on globally
important lines), with R-squared values usually between .65 and .8,
depending on the smoothness of the built forms surface, that is the
degree to which the built forms and infrastructure which attract and
generate the movement are uniformly distributed throughout the grid.
Where they are not - for example there is a main shopping street - then
normally this extra attraction will have occurred on a key movement
line, and the multiplier effect that the extra attraction has on movement
is captured by taking the square root or logging the movement variable.
The degree of transformation required to linearise the movement indicates
the degree to which extra attractors are present in the system. In general
the only other variable required to model movement is knowledge of these
special attractors. This may be as simple as average building height
in an area.
Figure 4 Baltic House Area; Global Integration
Figure 5 Baltic House Area; Rush Hour Average (Adults/ ph)
Figure 6 Correlation b'teen Global Integration and Rush Hour Movement
Significant results also
exist for vehicular movement. For example in a recent study of five
areas of London (Penn et al 1998) the r-squared for all the areas taken
together and correlated against the axial map of the whole of London
was .68 for 477 observed location, with much higher values for some
areas taken individually. Combining this with net road width (road width
minus car parking: not truly an independent variable because it will
have been adjusted in the light of flows) the r-squared rises to .84.
Figure 7. In a more recent study of 16 areas in Santiago de Chile (see
below for account of study) the r-squared for spatial configuration
alone was .54 for 212 locations without taking net road width into account,
again with much higher values within most areas taken separately (though
on the basis of the same axial analysis).
Figure 7
These results are now supported
by dozens of similar studies, mainly of pedestrian movement in different
parts of the world, showing that under normal circumstances (essentially
a reasonably homogeneous distribution of built forms) the spatial configuration
of the urban grid is in itself a consistent factor in determining movement
flows. Some key studies are reported in Hillier et al 1987, Peponis
1990, Hillier et al 1993, and Read 1997. The robustness of this relation
is in fact tested at least once a week in the work of the Space Syntax
Limited, since most design applications involve a movement study.
For example the recent World Squares for All masterplan for the Whitehall
area of London with Sir Norman Foster, including the re-engineering
of Trafalgar Square, was done on the basis of a syntactic study
of the spatial structure and pedestrian movement patterns in the area.
Theoretical developments
The purpose of this research, however, has never been to build a model
for accurately predicting pedestrian or vehicular flows, but to estimate
the degree to which the urban grid configuration in itself influences
movement. It is the independent effect of the built environment that
we seek to clarify. From the results we have, two theoretical propositions
have been developed concerning the nature and functioning of urban grids,
both of which have proved of great usefulness for design. The first
is the theory of natural movement , which proposes that to the
degree that the distribution of the built forms which generate and attract
movement in an area is homogeneous, then, other things being equal,
movement in the spatial system linking the buildings will be determined
by the grid configuration itself (Hillier et al 1993). The 'natural
movement' in a system is thus the proportion of observable movement
along lines that is produced by the structure of the grid itself rather
than special attractors. There is a problem of course. Surely movement
will itself attract attractors ? This leads is to the second theoretical
proposition: the theory of the 'movement economy' (Hillier 1996a).
This proposes that there is a 'central dynamic' to the spatial growth
of cities, which links the evolving grid structure and its natural movement
to the distribution of land uses and built form densities, and even
gives rise (though in different ways in different 'spatial cultures')
to the local area structures that are found in historically grown cities.
The mechanism is that as the accumulation of new built forms creates
new spaces in the expanding settlement, the emerging structure of the
spatial pattern gives rise to a natural movement pattern. Land uses
which seek movement, such as markets and retail, then naturally gravitate
towards higher movement locations, while others equally natural prefer
low movement locations. The extra attraction in the high movement spaces
then creates a multiplier effect on movement, which then attracts more,
and more diverse, movement-seeking uses, and vice versa. In this way,
the settlement pattern naturally evolves towards a seamless network
of busy and quiet areas, with the busiest in the spatially most integrated
areas, the whole process being initiated in the first place by the spatial
configuration of the grid. While the theory of natural movement notes
a regularity, the theory of the 'movement economy' tries to account
for the process by which the apparent affinity between grid structure,
movement, land uses and even building densities appears (as Figures
2 and 3 suggested) to to arise in naturally evolved urban grids like
London.
Sites and serves settlement in Santiago: a case study in the movement
economy
A recent study in which the theory of the movement economy proved useful
was a study of the evolution of 'sites and services' settlements in
Santiago, Chile funded by the European Union. This research began with
the suspicion amongst Chilean social scientists and architects studying
the development over time of 'sites and services' settlements in Santiago
that spatial, locational and design factors were implicated in the very
different levels to which settlements had developed since their common
foundation in the early nineteen seventies. From similar origins, some
had become well consolidated and seemed to be thriving, while others
remained much less consolidated and others seemed to have developed
social problems. The aim of the study was to ascertain how far and in
what way spatial factors might have contributed to these differences.
The research strategy was to combine the social scientific expertise
and experience of the Universidad Catolica de Chile in the investigation
of the social and physical aspects of settlement performance, with the
spatial modelling and observational techniques of the Space Syntax .
The analysed map of Santiago, indicating the 17 settlements originally
studied (one was later eliminated as a special case) is given in Figure
8. The research procedure was:
- the construction of indices for settlement consolidation (including
innovative use of the Delphi technique, leading to a separate publication
- Greene, Iacobelli & Ortuszar, 1997) covering housing, communities
and neighbourhoods, and the combination of all three, a questionnaire
to 553 respondents, and a full survey of 17 settlements;
- the direct observation of the functioning of the settlements, covering
land uses, commercial and social activity, and pedestrian and vehicular
movement patterns;
- the construction of a 'space syntax' computer model of the whole of
Santiago, including all the settlements;
- the construction and analysis of data tables for 553 individuals and
17 settlements, each covering all aspects.
Figure 8 Santiago; Location of Site and Services Settlements
The key results were at the
settlement level. Data analysis was in two stages. First, a correlation
study was carried out to establish which variables were prima facie
involved in each aspect of settlement consolidation. Then multiple and
stepwise regression were used to seeks answers to specific questions
about the patterns of influence, and in this way to build a picture
of any possible process by which spatial, locational and land use factors
might have played a role in the pathways of development of the settlements.
Correlation matrices are given in Figure 9.
Figure 9 Santigo; Correlation Matrices
The first stage analysis
showed that social factors as varied as income, education, spending
patterns, age profiles, numbers in household, numbers of children, and
vehicle ownership were virtually uncorrelated with the consolidation
variables. Spatial factors, on the other hand, were suggestively correlated
with all four indices, and space use and movement factors even more
so. Strikingly, the strongest correlations of all for the consolidation
indices were with the degree to which informal business activity developed
on the outward facing edges of the settlement. This 'edge commercial
activity' was associated not only with greater consolidation of houses
and a higher level of community development (these were the strongest
elements in its correlation with the 'general consolidation index'),
but also with lower reported experience of crime in the settlement,
and (as might be expected) higher overall levels of informal economic
activity in the settlement as a whole. Figure 11 shows a case where
'edge commercial activity' is strong, and Figure 12 a case where it
is weak, with most informal commercial activity in the interior of the
settlement serving the needs of local streets.
Figure 10 Santigo; Scattergrams
Why the difference ? The
answer is quite simple; vehicular movement. The correlation between
edge commercial activity and vehicular movement is the most powerful
in the whole data set (R-squared .888), and is at the same time totally
uncorrelated with vehicle ownership in the settlement itself. Figures
9 & 10. What then is the determinant of vehicular movement rates
? The answer is equally clearly: the spatial structure, and in particular
how the settlement is embedded in its local area. The best measure we
have so far found (work is continuing) we call 'local spatial advantage'.
This is calculated by taking a circular area up to 1.5 kilometres from
the settlement edges, and calculating local integration on the basis
of this metrically uniform system (regardless of the degree to which
this area was developed). This is illustrated in Figure 13 for the 'weak
edge' case and Figure 14 for the 'strong edge' case. In effect, local
spatial conditions create more or less local vehicular movement, edge
commercial activity then takes advantage of this to the degree that
it is available, this increases the overall level of informal commercial
activity in the settlement, and it is this complex that is associated
with higher levels of housing and community consolidation - a clear
case, it would seem, of the movement economy process in action, as well
as of petit bourgeois virtue !
Figure 11 Santiago; Caupolican-Las Torres, Macul.
Figure 12 Santiago; Villa El Rodeo, Huechuraba
Why then is this process
not correlated with income ? Earlier studies at the larger scale of
the 'local authority areas' had shown that average incomes for areas
in Santiago were almost wholly a function of mean education levels (r-squared
.964), although spatial integration was also a strong correlate. In
these settlements too income was correlated strongly with both education
levels and 'local spatial advantage', with the two between them accounting
equally for over 70% of the variance in the mean per capita income for
households in the settlement. Edge commercial activity was positively
correlated with income, but weakly so and below the threshold of significance.
Income and education were then either uncorrelated with or negatively
correlated with the consolidation variables. At the same time, income
and education both correlated quite strongly with the purchase of consumer
durables ?
Figure 13 Santiago; Caupolican-Las Torres, Macul.
Figure 14 Santiago; Villa El Rodeo, Huechuraba
At this stage (again, work
is still continuing) our belief is that two processes are in operation
in the settlements, one spatial and the other independent of space:
on the one hand, there is a spatial process led by the edge oriented
economy, which leads to greater activity within the settlement, and
then to greater security and settlement consolidation where this is
successful; on the other, a nonspatial process led by education levels,
and associated with working outside the settlement (more 'professional'
jobs), and contributing to the settlement more as a consumer than as
a producer. Local spatial advantage plays a role in both processes.
For educated people, local spatial advantage is associated with greater
opportunities working outside the settlement, while for those with less
education local spatial advantage provides informal economic opportunities.
The former process is associated with greater income but less settlement
development, and vice versa. The chief factor in consolidation seems
to be the degree to which economic activity within the settlement gives
people a stake in the settlement, and leads them to invest in its future,
and this is governed by a process which is initiated by space. More
educated people work outside the settlement, invest in it less, and
buy more consumer goods, presumable because they intend to leave the
settlement and invest in movable rather than fixed goods.
Can spatial design be implicated in the social decline of new housing
areas ?
Movement has also been shown to be critical in studies of the decline
of new housing areas, though in a very different way. Figure 15 is an
axial map of the area of central London around the Kings Cross Railway
Lands site (3). The housing estate to be discussed here is ringed above
the site. It has a certain notoriety because it went from being a much-praised
award-winning design to being described by the police as a 'ticking
time-bomb' in less than four years. The axial map of the estate shows
a number of properties that are very common in social housing: the axial
scale of space is dramatically reduced; the spatial pattern is very
much more complex; and the estate overall displays structural segregation,
meaning that there is no pattern of integrated lines linking the interior
of the estate to the surrounding area. Are there consequence for people
from this type of spatial pattern, and can it in any way be implicated
in the rapid decline of the estate ?
Figure 15 Maiden Lane Estate
To suggest an answer, we
must first be clear how this estate differs structurally from normal
street based urban areas. In inner London, for example, most named areas
have a structure similar to that found for London as a whole: an integration
core composed of a partial grid at or near the centre linked by strong
lines to the edge in different directions, with the quieter, more exclusively
residential areas lying in the interstices formed by this pattern. Figure
16. This turns out to be expressible statistically. Figure 16a. The
lines that connect edge to centre, and the centre itself, are stronger
local integrators than other lines in the area, while the interstitial
areas are less so. The means that if we plot local against global integration
for those lines against a scattergram for the whole of London (Figure
16a) we find that well-formed areas tend to have a linear scatter crossing
the main regression line at a steeper angle. The linearity of this scatter,
its slope, and its location in the overall scatter are powerful numerical
indicators of the characteristics of an area. However, if we take the
ringed housing estate we find a scatter (Figure 15a) which is layered
rather than linear, with each layer more or less vertical (each line
corresponding to one step of axial depth into the estate), and is bottom
left in the scatter meaning strongly segregated, and so on.
Figure 16 Barnsbury
This type of spatial structure
creates a local situation in which there is no relation between internal
and external movement, the average background encounter rate is reduced
by an order of magnitude (from about 2.7 people per minute to .27 people
per minute, or one person every four minutes), and the lack of natural
space occupancy becomes associated with the social misuse of these 'structurally
abandoned' spaces. The mechanism by which the social misuse of space
is facilitated by spatial design is that in most street-based areas,
the space structure is such that natural movement leads to all social
groups - men and women, adults and children, inhabitants and strangers,
and so on - using the space pattern in a similar way, so that all spaces
are used by all categories. This yields a background pattern of natural
co-presence and mutual surveillance between categories. This can be
merely pleasurable, but is can also become important when it is between
inhabitants and strangers or adults and children. In overly complex
housing estates there are virtually no strangers (the scale and complexity
of the space excludes it from local natural movement networks) and far
fewer inhabitants. In such circumstances strangers, who seem normal
in streets, become a source of fear, because their presence is unexpected.
However, because the space structure is much more complex in housing
estates, and because relatively little of it is used, the way is open
to exploitation of the empty spaces by groups who take a different view
of space. For example, children and teenagers tend to form larger groups,
and occupy not the spaces that are used for adult movement, but those
that are not. This has a critical effect: the natural surveillance from
adults to children and teenagers is broken. Compared to streets, children
spend more time in larger groups and out of contact with adults in complex
housing estate. Greater vandalism tends to be found in and around these
spaces. This process increases fear on the estate, which is already
increased by the very low background encounter rate, and this is in
turn exacerbated by the social misuse of the spaces which are not ordinarily
used. This creates the sense that the estate has problems, and this
decreases its desirability for incoming tenants. From then on the estate
is on the way to stigmatisation. Thus a social process can be initiated
by a spatial one. In a sense, the symptoms bring about the disease.
Once again we see that space structure and its impact on movement are
critical to the link between the built environment and its social functioning.
This process is described in greater detail in (Hillier 1996a, Chapter
5).
Does urban layout affect the pattern of urban
crime ?
In other urban issues the
mechanisms of influence from space to behaviour are less complex, but
no less critical. In trying to detect any effect of urban layout on
the spatial distribution of crime, for example, the mechanisms would
be expected to be simply what design factors would cause a criminal
to think one location was more vulnerable than another. Even so, there
is much confusion about the matter. There is a widespread tendency for
people to think that the mechanism is that the more people pass your
front door the more likely one of them is to be a thief. Those who have
read about 'defensible space' (Newman, 1972) will think that there is
a theory that backs this. In fact, something like the opposite seems
to be the case. Other people, including strangers, keep you safe.
The technique here is in two stages. Figure 17. We take an area and
first create a map in which dots indicate the exact locations of different
kinds of crime - burglary in this case. We then create a space syntax
model of the area in its context (subject to the same strictures as
before), and carry out the various analyses. We then superimpose the
crime map on the syntax map, and proceed in two ways: first, look for
visual patterns, and their relation to the visual syntactic patterns;
then correlate syntactic values with crime rates. It is important in
these cases to consider the lack of crime as well as concentrations.
More significantly we take take analysis beyond the simple analysis
of 'clusters' and show that although spatially dispersed, certain crimes
tend to occur in specific types of syntactic location on the layout.
For example, in the map shown, a designed area of New Town X, crime
seems to be highest where the urban grid is most broken up (in effect
creating most local segregation), and lowest where the lines are longest,
and in fact most integrated. The linear routes through the estate have
least burglary, and the most broken up, locally enclosed spaces the
most.
Figure 17 New Town X (Burglary in dwellings shown by red dots)
This (and other results)
runs against the general belief that spatial segregation decreases crime.
On the contrary, recent results at an area level show that other things
being equal (for example one not being in the city centre and the other
on the periphery) a more integrated area is likely to have less crime
that a more segregated one. We also find that crime is higher where
access is through spaces which are unrelated to building entrances.
In other words it is not the surveillance of the space you are in that
is critical, but surveillance by neighbouring groups of houses on the
way to your space. This interdependency is critical. For example, the
use of a general cul de sac layout tends to create vulnerability by
increasing entrance-free space, and greater delinearisation of spaces.
In general, safety seems to lie in linearity (including occasional short
linear cul de sacs attached to streets, but not cul de sac complexes)
and in the continuity of building entrances through all spaces (rather
than focusing entrances on a selected space to create a supposed 'sense
of community') coupled to the minimisation of rear and side access.
This research is being presented as a keynote paper at the forthcoming
UK Home Office Crime Prevention College Conference on What Really Works
on Environmental Crime Prevention in October 1998.
What effect does the street network have on urban pollution ?
Once we have the technique of regressing values representing the 'configurational'
properties of locations with functional values (such as movement rates)
for that location, then we can apply it to any functional phenomena
which can be located exactly and expressed as a number. For example,
recent work has used space syntax to look in great detail at the spatial
diffusion of pollution emanating from vehicles, and how it is affected
by the configuration of the street grid, thus potentially affecting
pedestrians. In order to monitor pollution in enough locations, a simple
'streetbox' capable of measuring carbon monoxide, temperature, humidity,
daylight and windspeed with an accuracy within 5%, and which could be
fixed to a lamp post was developed by Dr Ben Croxford and data downloaded
periodically to a Psion Organiser. A syntactic model of a local street
system was made and tested against a sample of real flow points. Two
early findings were that there were huge variations in pollution both
according to time and space. For example, within tens of metres of very
high pollution streets, rates on local streets were no more than background.
The effects of wind were also examined, and again this showed great
variation according to wind direction and side of street. Taking all
these factors into account, pollution levels were related to traffic
flows, and good predictions were also obtained from the syntax model,
which was also experimentally modified to take account of the prevailing
wind.(Figure 18) Work is continuing on the modelling aspects of this
in order to predict pollution levels in the fine scale structure of
the urban environment, with a view to better understanding the exposure
of pedestrians to traffic pollution.
Figure
18 Urban Pollution
Can there be a spatial
theory of the city ?
The consistency of these result across a whole range of urban phenomena
suggest that space may indeed offer something like a common language
of the city. Many if not most of the relations between the form of the
city and the way it functions seem to pass through space in some sense,
and many also involve the space-movement relation. The fact that the
most important of our results are about the urban structure itself are
also suggestive. Whatever functional phenomenon we pursue, the use of
syntactic techniques seem to make some kind of sense out of apparently
disorderly urban patterns, and shows they have functionally sensitive
structures. It is, it seems, these spatial structures that characterise
cities, and it is these that relate the form of the city to its function.
Can we then go on from here and build a spatial theory of the city ?
The fact that the most important of our results are about the urban
structure itself suggest that this might be possible. It is clear from
the similarities of the forms of cities all over the world that they
grow according to a certain logic. If we could capture this 'generative
logic', then it is likely that we could also understand how the evolving
form of the city relates to how it functions. The idea proposed here
is that the 'generative logic' of the city is essentially about space:
more precisely about how the now piecemeal now orderly aggregation of
buildings creates a continuous pattern of space which links the buildings
together into a system and in doing so constitutes in itself the essential
structure of the city. By learning the language of this spatial evolution
- a matter of understanding first of all what all cities have in common
spatially, and then considering the range of differences - we can learn
to ask questions of the city and get intelligible answers. But we can
only have a common language of space to the extent that it is also a
theoretical language, and we can only have a theoretical language to
the extent that it is the language of the city itself.
Reflections on the need for a domain theory.
It is easy to see that this type of research can serve the needs of
designers. But how does this fit into the overall pattern of urban research
? In the long run, it may be best seen as answering the emerging need
for domain theories in the study of cities, that is theories which deal
with autonomous laws within domains such as space which need to be understood
if we are to understand the relations across domains - such as the relations
between space and society - out of which urban complexity is made. Space
syntax is an attempt to build a domain theory of the urban object itself.
Let us be clear why we need this ?
Cities are the largest and most complex objects that human beings make.
With a few exceptions, they come into existence not through once for
all design but through a process of growth and change spread over tens,
hundreds, or even thousands of years. Each generation examines what
it inherits, then extends, substitutes, re-arranges and adapts according
to its needs, before passing it on to the next generation. What we call
the city at any one point in time is as much process as object: an emergent
structure created by a large number of smaller scale decisions, one
of the large class of artifacts (like languages and societies) which
human beings create but which remain puzzles for us.
For much of the twentieth century we have tried to use science to improve
our understanding of the city. Even where this has been successful,
the consequences have not been unmixed. Often we have solved one problem
only to create another. For example, we reacted to the environmental
problems of traffic by turning away from streets, and as a result, down
graded public space and the activity it supported, increasing the social
isolation of some of our communities. Policy was driven by research,
and research could only deal with one issue at a time. We could not
deal with the systemic interactions between the very different kinds
of phenomena that make urban life what it is.
More recently, under the urgent influence of problems like sustainability,
some of our scientific effort has switched to a more systemic study
of cities, with the object of understanding the interaction between
the built environment aspects of the city and the social, economic and
behavioural phenomena that animate it. This research is aimed at urban
policy: how should we guide the development of our cities in the light
of broad social objectives. For example, if we are to save energy, we
need to know whether this could best be achieved by re-engineering our
pattern of settlement, or simply by regulating and managing behaviour
in existing settlements, for example by increasing fuel costs or road
pricing.
Suppose however that built environment variables did turn out to be
important. The fact is that the huge inertia of the existing built environment
is such that the timescale for bringing about change makes effective
policy implementation unrealistic. It can even lead to absurdity. We
have for example in general in Europe adopted a compact city policy,
on the grounds that this will reduce the length of journeys and thus
save energy. Although it will be decades before this has a significant
effect, we have recently heard from my new colleague Professor Steadman
that one effect of the compact city policy may be to increase the price
of land in city centres and drive out residents, thus creating longer
journeys to work, and working against the policy. Changing urban policy
is, it seems, like turning a tanker, but with an added twist: by the
time the tanker is turned, knowledge has advanced and it is time to
turn again. The form of movement that emerges from this process needs
little clarification.
It is clear from this that a policy orientation is not the only useful
framework for built environment research. We have long since abandoned
the utopian notions of 'end-state' planning through massive urban re-engineering
and replacement. Effectively we have reverted to the historic reality
of the city as a distributed process in which the whole emerges from
innumerable local actions. In western countries most urban initiative
is now in the private sector, and the role of planning is a regulatory
one, its effectiveness limited by the timescale factors we have noted,
and by its weakness in playing any initiatory role in projects. The
motor of urban change is now once again the privately initiated and
privately funded design and development project.
This has created a knowledge emergency. New kinds of knowledge are needed
to support design and development, and to ensure that it carries out
its tasks in a socially and environmentally responsible way. For the
most part, the knowledge that comes out of policy research is in the
wrong shape to serve the knowledge needs of this sector. It is too nonspecific,
too lacking in detail, and too little oriented to the variables that
designers and developers can manipulate, namely the physical and spatial
variables of the built environment itself. Designers and developers
needs to answer immediate questions about the impact of built environments:
what would be the social, economic and environmental impact of developing
this area or this site with this design and this relation to the surrounding
urban context ? What would be the accumulative impact of similar interventions
on the urban system over time ?
We need, in effect, to move towards a science capable of supporting
evidence-based design, that is design which takes place in the knowledge
of the social, economic and environmental impact of different kinds
of intervention. What kind of science is required for this ? The answer
is simple: research which treats the built environment not as one of
a number of intervening variables in a policy question, but as the principle
variable, as it is in the real world design and development process.
Research in support of evidence based design is research which seek
to understand the impact of built environments on how people, organisations
and communities live their lives, and the accumulative effects of built
environment decisions on the larger scale pathways of our cities and
the longer term pathways of our societies.
Research with built environment as the principle variable will theoretically
come in three kinds:
- Type 1 research which treats the built environment as an autonomous
variable , in which we ask such question as: what are the possible
forms for cities to take, what are the laws of emergence from local
decisions to global patterns - how in short does the built environment
behave as a form of complexity in itself ? The current rapid spread
of work on cellular automata, some aspects of fractal theory, and also
syntactic approaches to computer generation of settlement forms (for
example, Erickson and Lloyd Jones, 1997) are all relevant to this.
- Type 2 research which (preferably in the light of type 1 research)
treats the built environment as the dependent variable , and
asks what kinds of social, economic and cultural processes modify the
autonomous processes and give rise to the different kinds of built environment
complexity we associate with different types of society and culture;
and
-Type 3 research which (preferably in the light of types 1 and
2 research) treats the built environment as the independent variable,
and asks what follows functionally from selecting one 'complexity strategy'
for the built environment rather than another.
These three forms of enquiry are cyclic in the sense that knowledge
of the relation between spatial configuration and movement - a type
3 effect - can then be fed back as a constraint into type 1 research
to see how movement affects the evolutionary patterns of space itself
(Hillier 1996a, Chapter 9). Both of these are then involved in type
2 questions about why different kinds of society adopt different spatial
forms. In this sense the theory of the city as a movement economy is
cyclic: it deals in some way with all three aspects of the built environment
process. Any domain theory of the built environment will eventually
need to be cyclic in this sense.
bh 8.9.98
Notes
1 - Space syntax also address issues of space in buildings, from the
largest quasi-urban complexes down to the cross cultural analysis of
houses.
2 - The mathematical basis for these measures was first given in Hillier
et al 1983, and then in Hillier & Hanson 1984, with further discussed
in Hillier 1996, and numerous other texts, including Steadman 1983.
3 - One of the earliest urban design projects in which space syntax
was used was the Norman Foster masterplan for the redevelopment of this
site. As a consequence, the surrounding area came to be studied intensively.
Bibliography
Croxford B, Penn A., Hillier B. (1996) Spatial distribution of urban
pollution: civilising urban traffic, Science of the Total Environment
189/190 pp3-9.
Erickson B & Lloyd-Jones T (1997) Experiments with settlement
aggregation models Environment and Planning B; Planning and Design
Vol 24, 903-928
Greene M, Iacobelli A, & Ortuzar J - Subjective valuation of
social housing attributes Proceedings Fourth Interntional Conference
on Computers in Urban PLanning and Management, Melbourne, Australia
Hillier B (1996a) - Space is the Machine Cambriudge University
Press - 470pp Paperback 1998
Hillier B et al (1987) - Creating life: or, does architecture determine
anything? (with Burdett R, Peponis J, & Penn A) in Architecture
& Behaviour Vol 3 No 3 pp 233-250.
Hillier B et al (1993) - Natural movement: or configuration and attraction
in urban pedestrian movement - Environment & Planning B: Planning
& Design Vol. 19 38pp
Newman O (1972) - Defensible space Architectural Press, London
Penn A et al (1998) Configurational modelling of urban movement networks
Environment and Planning B: Planning & Design Vol 25 pp 59-84
Peponis j et al (1989) The spatial core of urban culture Ekistics
334/335; Special Issue on Space Syntax pp 43-55
Read S (1997) Space syntax and the Dutch city Proceeding of the
First International Space Syntax Symposium, Vol 1 02.1-13, Space Syntax
Limited, University College London
Steadman P (1983) Architectural Morphology Cambridge University
Press
|
