Chapter 3

Space Syntax Methodology Review

3.1 Introduction

Space syntax is a street network method, which was developed in the 1970s at the
University of London as a way to quantify urban morphology features and recorded
movement and interaction within cities and buildings. In the book The Social Logic
of Space (Hillier and Hanson 1984), they firstly argued that spatial layout or
structure has a great impact on human social activities. Recently, the approach was
refined by Hillier (1996), and it particularly focuses on the arrangement of spaces
and possibilities and patterns of movement through the “spatial configuration.”
Over the past two decades, space syntax theory has provided computational support
for the development of urban morphological studies, revealing the characteristics of
spaces in terms of movement and potential use. Indeed, space syntax attempts to
define the elements of urban form by geometric accessibility, measuring the rela-
tionships between street segments by a series of measurements, such as connec-
tivity, control, closeness, and betweeness (Jiang and Claramunt 2002).
Indeed, as a technique, space syntax is powerful and successful, as it can
describe a complex urban configuration system, providing direct interpretation to
help architects, urban planners, and sociologists to understand the impact of spatial
configuration on social activities both theoretically and practically. For example, for
the regeneration project at Trafalgar square, London, the space syntax research
method provided an initial insight of tourists’ activity patterns in the Trafalgar
Square.
Despite its success, there have been critics on the space syntax research method,
mainly focusing on the definition of axial lines and the unique set of axial lines. For
example, Ratti (2004) criticized that the appropriateness of “axial maps” for
describing the urban spatial structure. It has also been suggested that the definition
and theoretical meaning of the axial map is not well enough linked to other
graph-theoretic representation of street network (Steadman 2004).

© Tongji University Press and Springer Nature Singapore Pte Ltd. 2017 41
Y. Xiao, Urban Morphology and Housing Market,
Springer Geography, DOI 10.1007/978-981-10-2762-8_3
42 3 Space Syntax Methodology Review

The motivation of this chapter is to provide a deeper insight into the space syntax
method, discussing several issues. Firstly, what is the difference between space
syntax network analysis and traditional geographic network analysis assessing
accessibility? Secondly, the algorithm of the space syntax method and three types
of accessibility measurement of space syntax are discussed. Thirdly, some key
criticism on space syntax will be summarized. Fourthly, the socioeconomic fields
where the space syntax method has been applied to will be discussed.

3.2 Overview of Urban Morphology Analysis

The study of urban morphology, frequently referred to urban form, urban land-
scape, and townscape, has an extensive literature in three genres. Firstly, in the field
of human geography, especially in Britain, there is an “indigenous British geo-
graphical tradition,” which was heavily influenced by M.R.G. Conzen. Conzen
develops two of his key ideas, the burgage cycle and fringe belts. The Conzenian is
more interested in the description and classification, and they exemplified the
characteristics of present townscapes based on the survey. Recently, the typical
works contributed to this tradition are from Whitehand (1967, 1972, 1987a, b,
1992) and Slater (1988). Additionally, by contrast, European tradition (e.g.,
Caniggian School) takes the views from architect, emphasizing the components of
the urban structure, including its elements, structures of elements, organism of
structures (Sima and Zhang 2009a, b) in architecture and planning theory. Many
scholars concerned urban morphology with prescriptive and utopian for “ideal
cities.” For example, The Garden Cities of Howard (1965), Le Corbusier’s Ville
Radieuse (Corbusier and Guiton 1981), and Wright’s Broadacre (Wright 1935) are
the outstanding examples of this genre.
However, the most prominent exponents of the critique from Lynch (1960),
Jacobs (1961), and Alexander (1964, 1974) require researchers to think city plan-
ning from anthropology perspectives, which should be based on observation of
what actually works in real cities. This sowed the seeds of mathematical approach
for quantifying urban morphology.
Alexander first attempted to introduce formal mathematical concepts into the
debate. Some scholars were interested in bringing mathematical tools into the realm
of urban morphology, in particular graph theory and set theory, and attempted to
link this idea with what works in the urban design arena (March and Steadman
1971; Martin and March 1972; Steadman 1983). For example, Q-analysis (Atkin
1978) was also influential. Work focusing on possible graph representations of
urban form (Krüger 1979) exemplifies the approach. Under this background, the
method of space syntax was gestated by Hillier and Hanson (1984), which is based
on a graphical representation describing cities as systems of open space.
3.3 Accessibility Types 43

3.3 Accessibility Types

“Accessibility is a concept that has become central to physical planning during the
last 50 years” (Batty 2009).
In the field of population geography, Stewart (1947) firstly used the concept of
graph theory to define the gravitational potential by the weighted sum of forces.
Hansen (1959) and Wilson (1970) started with a graph theory view to explore the
spatial system, identifying accessibility as a key element in spatial interaction, while
Haggett and Chorley (1969) established the central analysis approach to analyze
spatial relationships.
Batty (2009) summarized three types of accessibility. Firstly, type one acces-
sibility defines how proximate or “near” an individual is to “opportunities,” mea-
suring the size of the opportunity at some other places or location and inversely
moderating the cost of accessing the opportunities by the distance or time. The
other two types of accessibility are based on a network. Type two accessibility
focuses more on the Euclidean distance between one location and another, mea-
suring this as the sum of the shortest routes in a planar graph. This measurement is
popular in traffic models, and high accessibility normally is associated with mini-
mum distance, travel time, and travel cost. In the metric geographical network
analysis, nodes or vertices are defined as the intersection or junction of streets,
while the links or edges are street segments linking two intersections. This is called
the prime approach. In the last 20 years, a third type of accessibility measurement
emerged which seems more abstract. The type three accessibility is based on the
dual approach, which is defined on the links of the original graph, rather than the
nodes, applying degree (cD) (control), closeness (cC) (“integration” in space syn-
tax), and betweenness (cB) (“Choice” in space syntax) as measurement of cen-
tralities (Bavelas 1948, 1950; Leavitt 1951; Shimbel 1953; Shaw 1954, 1964).
The accessibility measured by space syntax method (Hillier and Hanson 1984)
belongs to the third type, which is also called dual approach. It is more abstract,
focusing on links (street) connection in a planar graph. Similar dual approach
methods have also explored gradual directional change (Dalton et al. 2003),
characteristic point (Jiang and Claramunt 2002), and street name change (Jiang and
Claramunt 2004).
Many researchers have discussed the substantial differences between the prime
and dual approach, as shown in Fig. 3.1, and argued which approach is more
effective. Batty (2004) stated the dual approach breaks the clear links between the
Euclidean and topological space and also making the visual analysis more difficult,
compared with the prime approach, he agrees that both methods are powerful to
explain the same problem. Porta et al. (2006) introduced a new geographical net-
work analysis methodology, known as multiple centrality assessment, based on the
prime approach. Comparing results used by the two methods, they point out that the
prime approach is not efficient in measuring the cC closeness and is much more
fragmented than the dual method, which is more generalized. However, he
44 3 Space Syntax Methodology Review

Fig. 3.1 Conventional graph-theoretic representation of the street network. Source Batty (2004a)

emphasized that the prime approach is more compatible in other fields and is more
subjective, intuitive, and practical. Hillier and Penn (2004) emphasized that the dual
approach uses topological distance to prevent border effects.

3.4 Space Syntax Algorithm

What is original to space syntax is the important insight that the pattern of
movement in a city or urban area is likely to be shaped to an extent by the topology
of its route network alone. The basic analytical procedure of the original space
syntax method is divided into two steps. The first step is to create an axial map, of a
certain area at any given scale. At the early stage, the axial map is based on
experiences, which starts with identifying the fattest convex spaces. A “convex
space,” as shown in Fig. 3.2, can be defined as polygons where no lines can be
drawn between any two points (x, y) in the space which goes outside the area. The
longest line indicates the most optimal convex space with the maximum perimeter
ratio. After identifying the fattest convex space, it is possible to draw the longest
axial line within the convex space and then draw the second longest axial and to
continue until all the convex spaces are covered. The whole process should result in
the fewest axial lines representing a one-dimensional system of the spatial layout
(Hillier and Hanson 1984). Space syntax makes a default assumption that people
will move where they can see, and “convex space” could indicate people’s
movement in a visual area. Urban space could be decomposed into numbers of the
“convex space” by people’s visual zone contained in urban spatial structure or
morphology; thus, the urban space could be transformed into line–line graph with
the “longest and fewest” axial lines.
3.4 Space Syntax Algorithm 45

1 2

3 4

Fig. 3.2 The process of converting the “convex space” to axial line map. Source: Hillier and
Hanson (1984)

The results in the complex urban configuration system being represented as a

one-dimensional graph G (N, K), a mathematical entity defined by two sets, N and
K. The first set, N, denotes a non-empty set of N elements called nodes, vertices, or
points, and K is a set of K elements containing unordered pairs of different nodes
called links or edges. Once the representation of morphology has been created,
measures of spatial characteristics can be analyzed. Each axial line represents a
node N in the graph, and each intersection between lines represents a link K. An
indicator for measuring non-metric topology is depth, which is defined by the
number of turns from one axial line to another line, or the links from one node to
another node. Depth can be used to measure patterns of line connectivity called
integration (Steadman 1983; Hillier and Hanson 1984). This line graph approach
has proven to be unexpectedly successful in predicting urban movement and
measuring urban structure change (Hillier et al. 1993a; Hillier 1996).
Figure 3.3 shows how space syntax integration value is calculated for each axial
line based on accounting the numbers of depth. According to Hillier and Hanson’s
work (1984), the computation for integration of space syntax axial lines has several
steps as follows:

X
K 1
1: Sum Depth ¼ dij ð3:1Þ
j¼1;j6¼i

where K is the total number of nodes (axial line) in a graph, dij is the numbers of
depth from node i to node j.
46 3 Space Syntax Methodology Review

Fig. 3.3 Calculation of depth value of each street

3.4 Space Syntax Algorithm 47

2: Mean depth ðMD) ¼ Sum Depth ðSD)=ðK  1Þ ð3:2Þ

3: Relative Asymmetry ðRA) ¼ 2ðMD  1Þ=ðK  2Þ ð3:3Þ

4: Real Relative Asymmetry ðRRA) ¼ RA/Dk ð3:4Þ

 2
where Dk ¼ N log2 N þ
3  ðN  1Þ denotes the total depth of the root in a
diamond-shaped graph.

5: Integration ¼ 1=RRA ð3:5Þ

Relations of depth are derived from the notion of asymmetry, as spaces can only
be deeper than other spaces when passing through intervening spaces to arrive to
them. Mean depth provides the basis for RA, which provides a normalization of the
mean depth measure between the deepest a node could possibly be (at the end of a
sequence) and the shallowest it could be (when all other nodes are directly con-
nected to it). This will give a value between 0 and 1, with low values indicating a
space that tends to integrate the system, and high values a space, which tends to be
segregated from the system.
RRA provides a relativization to allow for comparisons of depth between dif-
ferent sized spatial systems. RRA eliminates the effect caused when real spatial
systems get larger (are composed of more nodes). With a larger system (area), a
node becomes relatively shallower when considered relative to how deep they could
possibly be given that number of nodes. Normally, it allows us to compare the
integration value in different urban systems, such as London and Beijing; which
requires eliminating the effects of the size on the RA of any space (Fig. 3.4).
Using the space syntax methodology, researches utilize the so-called global
integration (RN) method to measure the relative accessibility of a space within a
spatial system, measuring how many locations in the city are relatively strongly
integrated (connected) and how many are less integrated and probably belong to the
city fringe. Higher integration values are assumed to correlate with higher rates of
movement and activity. Spaces with few connections that lie “deep” within a
system have lower integration values and often experience lower levels of activity.
In the space syntax graphic, red lines represent the highest integration value; dark
blue lines represent the lowest integration value, in other words, areas with the
highest level of segregation. Integration proves to be an important spatial variable
that correlates well with social activities such as movement and interaction (Hillier
1996; Penn et al. 1998). Local integration (R3) measures accessibility up to three
steps away and in terms of axial line, means a topological distance of three turns.
There is evidence suggesting that integration at a local level correlates strongly with
local pedestrian movement, meaning short trips to local destinations. It has also
been shown correlating with the movement of “locals” or “inhabitants,” as com-
pared to non-residents and visitors who enter the spatial system from the outside
(Hillier and Hanson 1984; Hillier 1996).
48 3 Space Syntax Methodology Review

Fig. 3.4 Integration map of London. Source www.spacesyntax.com

There are other notions in the space syntax method theory, such as the notion of
“connectivity,” which measures the degree of intersection or one-step possibilities
of each axial line (Hillier 1996). The notion of “intelligibility” is defined as the
degree to what can be seen and experienced locally in the system (Hillier 1996).
The “intelligibility” value is calculated by the degree of linear correlation between
connectivity and global integration values (Hillier and Hanson 1984). “Synergy,”
denotes the degree of linear correlation between R3 and Rn, which is used to
mitigate the influence of system size (Hillier et al. 1993a).
This broadly describes the original formulation of space syntax. There have been
significant developments beyond this, which I will consider later in this chapter.
First, I will summarize some of the basic criticism of the method.

3.5 Critics of the Space Syntax Method

Space syntax methodology has been developed for more than 20 years since its first
publication in 1984 and has been widely adopted for the analysis of urban con-
figuration structure and socioeconomic prediction. However, controversies here
also come with its success. The debates mainly focus on three aspects as follows:
3.5 Critics of the Space Syntax Method 49

Firstly, its prediction ability is weak. Space syntax in its original topological
formulation is criticized for only considering topology distance and ignoring other
distance metrics or 3D information about the urban system or building system, such
as building height, land use, and transportation infrastructure. For example, in terms
of early building pattern analysis, architect Lawrence and Low (1990) and soci-
ologist Leach (1978) pointed out that in a cultural context, the syntactic arguments
were insufficient to infer the social dynamics only with a building’s floor plan.
Similar sentiments were expressed by Osman and Suliman (1996), who stated that
the boundaries between spaces are normally set by some kind of furniture rather
than by physical barriers, and noted that space syntax only models open floor plans.
Recently, Ratti (2004) criticized space syntax systematically, “How is it possible
to tell so many things about the urban environment with such a limited amount of
information, that is, after having dismissed data such as the height of buildings and
the size of streets.” He also argued that topological distance is unacceptable, as the
topological distance facing “New Yorker living on Fifth Avenue between 111th and
112th Streets going to Central Park North round the corner (two changes of
direction) is the same for him going to Columbus Circle…” (Ratti 2004). In fact, the
Columbus Circle is much further.
Steadman (2004) pointed out that the space syntax method seems to be prob-
lematic at the large urban scale. Movement along a straight but congested urban
street is slow and requires the expenditure of energy. The urban traveler might be
expected to choose the shortest metric distance for lower energetic cost rather than
the fewest changes in direction. He also noted that the same value of integration
cannot predict radically different flow volumes at two dates. Take Venice as an
example. Its configuration has changed little from the eighteenth century to the
twenty-first century, and the integration value of all axial lines will have changed
little over the period. It is doubtful, however, that pedestrian flows are correlated
with the same integration over the period.
However, Hillier and Penn (2004) disagreed with Ratti over the use of topo-
logical distance in spatial modeling instead of metric distance, insisting that when
employing topology, the problem of boundary presents no border effects. For
instance, when city centers are not in the center of geometry of the study area, high
integration still occurs in the street segments near the functional center rather than at
the center of study area.
Porta et al. (2006) compared the performance of the prime approach with the
dual approach in Ahmedabad, Vennice, Richmond, CA and Walnut Greek, CA.
The first approach is a metric geographic network analysis method, and the latter
the space syntax topology method. The conclusion they made was that the prime
approach is in fact, vulnerable to the Border effect.
Secondly, space syntax is also criticized for the procedure when it creates an
axial map which has been alleged to be not objective. Although, the whole process
should be based on the principle of the “longest and fewest” lines in the street
network, several authors (Batty 2001; Batty and Rana 2002; Jiang and Claramunt
2002) showed that the process of creation axial map is arbitrary, as it seems there is
no formal evidence showing the unique set of axial lines for a space. This
50 3 Space Syntax Methodology Review

uncertainty would lead to inaccurate creation of the axial map, influencing results of
analysis. Hence, it seems different users could generate different sets of axial maps
for the same application.
Hillier and Penn (2004) countered that the process of creating an axial map is not
arbitrary, and Carvalho and Penn (2004) used a statistical method for axial line
lengths of 36 cities. They found that axial lines are statistically significant elements,
while the errors are not as significant as expected. Hillier argued that the reason for
misunderstanding the issue of arbitrary lines is that critics thought axial maps
depend on the prior construction of a unique convex map, which they do not
depend on it in fact, either practically or theoretically. He also presented an argu-
ment to show that theoretically, there is only one correct graph of a road network.
Thirdly, Ratti (2004) argued that several inconsistencies when applying the axial
map to illustrate similar the urban layouts, even there is a cross-error for space
syntax topology measurement.
In Fig. 3.5, (a) is an orthogonal axial map and (b) is a “broken” or deformed one.
It is seen that such similar configurations have such distinct integration value.
Diagram (a) scores a uniform integration value 3.134, while the range of integration
value in diagram (b) is between 0.919 (central segments) and 1.930 (peripheral
segments).
Hillier and Penn (2004) noted that Ratti’s results were correct, but that these two
configurations are actually “syntactically” different, and thus, different values of
integration are reasonable. Diagram (b) is comprised of broken segments, and
Hillier argued that Ratti’s inference is conflicted with the topological issue.
Ratti (2004) also inferred that the procedure of axial map creation represented a
discontinuity in the transformation from geometry to topology. Because at a certain
angle and certain street width, axial lines of a deformed grid would pass through all
blocks and create the same configuration as an orthogonal grid. As seen in Fig. 3.6,
there are seven axial lines for the layout in the left diagram, but when the urban grid
was deformed slightly in the right diagram, due to the “fewest and longest” prin-
ciple, three axial lines are sufficient to represent the layout—the same results for an
orthogonal version of the grid.
Another inconsistency, or logical flow, identified by Ratti is the crossed-error
implicit in space syntax topological methodology (Ratti 2004). As Fig. 3.7 shows,
when two separated axial maps are linked with an extra axial line, the integration
pattern changes dramatically. Ratti therefore concluded that the method is sensitive
to the choice of study boundary, as it influences the value of integration.
Hillier and Penn (2004) admitted that the phenomenon could happen in areal
urban study. The most integrated part would shift from its central region to the
region connecting the outside world. However, the notion of metric search radii can
relieve the problem of edge effect by setting certain radii for the mean depth of the
system though street segment. However, they shared Ratti’s view that axial analysis
does not model the real movement rate.
3.5 Critics of the Space Syntax Method 51

Fig. 3.5 Value changes when deforming the configuration (Ratti 2004)

Fig. 3.6 Inconsistency of axial line (Ratti 2004)

52 3 Space Syntax Methodology Review

Fig. 3.7 Cross-error for two axial line maps (Ratti 2004)

3.6 Developments of Space Syntax Theory

Criticism has always been the driving force for development. Hillier and his col-
leges have improved the space syntax methodology by enhancing the specification
of axial map theory to create unique axial line maps, as well as developing metric
segment analysis and geometric segment analysis to deal with the inconsistencies
stated by Ratti (2004) and others. Hillier and his colleagues assigned space syntax
measurements by different definitions of weighted distance, including shortest path
(metric), least angle change (geometric), fewest turns (topological) weightings to
relations between each segment and all others (Hillier and Iida 2005). Since then,
space syntax has became a two-dimensional measurement for spatial accessibility,
which brings it closer to conventional traffic models. The detailed developments are
discussed below.

3.6.1 Unique Axial Line Map

Responding to the criticisms regarding the exact definition of axial lines, and its
non-uniqueness, Hillier and Penn (2004) argued that theoretically there is only one
correct axial line map. Furthermore, Turner et al. (2005) proposed an algorithmic
solution of the axial map, which delivers a sound definition and methodology in
order to obtain a unique map. The new approach is derived from the research done
by Penn et al. (1997), and his solution involves two steps. The first step is identical
to the method Penn published in 1997 of generating all axial line maps, which
includes the reduction of the lines starting with the longest line, see Fig. 3.8a. The
second step concerns a retrieval of the complete axial line system by preserving the
topological rings, as shown in Fig. 3.8b. A detailed account of the idea of topo-
logical rings can be found in Peponis et al. (1998).
Jiang and Liu (2009) introduced an alternative algorithm for identifying unique
axial lines, named AxialGen. The function allows automatic generation of a unique
axial map by selecting the least numbers in the longest visible lines (or axial lines)
representing individual open space (convex space) in an urban environment.
3.6 Developments of Space Syntax Theory 53

Fig. 3.8 An algorithmic definition of the axial map (Hillier and Penn 2004)

The substantial difference between AxialGen and Space syntax lies in their dis-
similar first step solutions. AxialGen used isovist analysis retrieval, developed by
Batty and Rana (2004), with an algorithm similar to Peponis’s (1998). Figure 3.8a
is a fictional spatial layout, and diagram (b) shows the spaces divided into Voronoi
regions of closed spaces. Taking any point on the edge of the Voronoi region for
isovist visibility analysis creates a set of isovist ridges, and the longest isovist ridges
in the isovist area can represent the whole isovist area, as can be seen in diagram
54 3 Space Syntax Methodology Review

(c). Diagram (d) shows the final selected axial lines which contain the least number
of longest visibility lines (Fig. 3.9).
Critics also pointed out the discussion of necessity of axial lines in space syntax.
For example, Turner (2007) first raised the question “why not other representa-
tion?” and applied an angular segment analysis to compare road-center line with
axial line models of an urban road network. The results showed that the road-center
line with angular analysis had a stronger correlation with observed vehicular flows,
implying that it is feasible to combine the representations of traditional trans-
portation network analysis with the space syntax method, as it will be convenient to
get a coherent cognition on the movement in the city.

Fig. 3.9 Definition of axial line by AxialGen (Jiang and Liu 2009)
3.6 Developments of Space Syntax Theory 55

3.6.2 Segment Metric Radius Measurement

As Ratti already showed that there is a cross-error in space syntax topology anal-
ysis, when two axial maps commutate each other, with the integration value
changing dramatically. The radial segment analysis introduced by Hillier and Penn
(2004) can avoid this error, as the metric radii can control the integration value,
which depends on the size of the analyzed area and avoiding edge effects. To do
this, firstly, the axial line map is transformed into a segment line map, breaking
down all axial lines by intersections and removing 25 % of the overall length of
line. This procedure can be carried out automatically through the application of the
SSX “Depthmap” software. Rather than topological accessibility, which calculates
the depth from a segment x to all other segments, the radii segment analysis for
accessibility calculates the depth but within a fix metric radius (Turner 2007). So, if
we know the size of our study area, we can define the exact radius we need to take
so as to avoid edge effects.
Beside space syntax integration, space syntax brings a new notion of “choice,”
also known as “betweenness.” Betweenness is another centrality measurement
popular used in the field of social network analysis. In space syntax theory,
“choice” measures of the flow through a space (Hillier et al. 1987). It captures how
often, on average, a location may be used in journeys from all places to all others in
the city. Locations that occur on many of the shortest paths between others (i.e.,
provide a strong choice) have higher betweenness than those that do not. Normally,
a main street has a high value, whereas the choice value is a side street off a main
road has a lower volume. Thus, a side street may have high integration but low
choice.

3.6.3 Angular Segment Measurement

Angular segment analysis deals with discontinuities, when geometric configuration

changes slightly. Dalton et al. (2003) argued with her British library hypothesis that
people prefer to minimize the angular distance to their destination.
Figure 3.10 shows a simplified segmented axial map and its associated “j-graph”
of segments. Firstly, to calculate the depth to any location by lowest angular cost,
we calculate the total angular turn from one segment to another segment via the
shortest angular route,dh (x, y), where x is the starting segment and y is the end
segment of the path. Hillier and Iida’s method (2005) assigned a value in the range
of 0 (no turn) to 2 (180° turn) for each turn. So in Fig. 3.10, the depth from segment
P to segment Q is 0.5 (a turn of 45°) and the depth to segment R is 1.333 (a turn of
45° followed by a turn of 60°, noting that the direction of turn is immaterial, the
turn angle is always positive).
By introducing alternative measures of distance such as shortest path (metric),
least angle change (geometric), and fewest turns (topological), to segments within
56 3 Space Syntax Methodology Review

Fig. 3.10 Notion of angular cost (Turner 2007)

any fix radius, with radius measured on metric, geometric, and topological domains,
space syntax methodology has become a more accurate and flexible tool for
exploring a broad range of behavior in the built environment.

3.6.4 How Urban Morphology Interacts with Socioeconomic

Phenomenon

Although it emerged from architectural roots, space syntax can be classified as a

location-based social network analysis method. It abstracts the spatial accessibility
from a complex urban configuration, describing how one street segment connects to
others and produces indices that can be taken as a surrogate for the distribution of
people’s movement within the urban layout. It is known that people have an innate
ability to read or comprehend the meanings of different arrangements or layout of
space and react to physical infrastructure change (Hillier and Hanson 1984).
However, traditional socioeconomic and environmental studies ignore the spatial
information contained in urban morphology, mainly considering people’s tendency
to seek accessibility for maximum benefits with minimum cost (e.g., distance, travel
time, composite travel cost), as in traditional traffic models. Recently, some studies
have connected space syntax to other academic fields, examining how the spatial
3.6 Developments of Space Syntax Theory 57

accessibility contained in urban morphology impacts economic and social activities.

For example, Vaughan and Penn (2006) used original census data to examine the
relationship between immigrant clustering and street-level settlement patterns for
the nineteenth-century Jewish immigrant quarters of Manchester and Leeds. They
found there is a strong negative relationship between the density of Jewish
inhabitants and space syntax integration, indicating that Jewish immigrants chose
the higher density areas, but lower spatial integration streets. They sought oppor-
tunity but also sought refuge in back streets.
Hillier (2004) addressed the controversy about the relationship between crime
and spatial design, finding a strong correlation between layout type and all kinds of
crime. Traditional street patterns were found to be the best and the most “modern”
hierarchical layouts the worst, which indicates that the rich and poor benefit alike
from living in traditional streets. He also offered some simple design guidance:
Enhancing the accessibility and visibility of public space can positively influence
crime rates. Nevertheless, the research draws a critical lesson that if security is to be
enhanced, both global and local factors must be appropriate. Otherwise, the security
could not be increased. This illustrates one of the great advantages of space syntax
and related methods: the ability to examine what urban morphology effects at
multiple scales (using multiple radii).
Nes and Rueb (2009) examined how the spatial layout of neighborhoods affects
dwellers’ behavior. They compared four different types of dwelling areas, one with
large social problems (Ondiep in Utrecht) and one without (Hof van Delft in Delft).
Other two post-war urban areas were chosen—one with serious social problems
(Oosterwei in Gouda) and one without (Ommoord in Rotterdam). They found that
accessibility is highly related to antisocial behavior problems with a negative sign,
indicating streets with high degree of intervisibility and adjacent to main routes can
contribute to create safe dwelling areas.
Croxford et al. (1996) examined the relationship between street grid configu-
ration and vehicular pollution, including carbon monoxide (CO), temperature,
relative humidity, light level, and wind speed at 6-min intervals. Finally, their
research suggested that the relationship is strong, and urban pollution level could
vary cross-streets, even spatially close streets can have very different levels of
pollution.
Barros et al. (2009) discussed the potential of space syntax as a tool for esti-
mating traffic routes. In route analysis, space syntax achieved refined results when
compared to simulation and assignment of traffic to urban road networks
(SATURN). The results showed that the dual approach measuring the movement
has better fitness than the regular geometrical approach. Hence, it would be better
suited in the transport field.
Other researchers have connected space syntax to the land value field, exploring
the relationship between urban configuration and land use or housing price. For
example, Brown (1999) noted that real estate analysts found local urban configu-
ration land design to be important for the field of real estate but that they have not
had the tools to build a strong theory. He compared two different types of shopping
mall configuration using space syntax and found the reason why shopping mall
58 3 Space Syntax Methodology Review

number one was successful while shopping mall number two had failed was con-
figuration design.
Min et al. (2007) explored the relationship between the characteristics of location
and land use, by a case study in Seoul. Their research showed that space syntax
integration value has a high correlation with land value and the location propensity
of industries. Similar findings are also found in the work of Kim and Sohn (2002).
These studies only focus on identifying positive or negative correlations between
urban morphology and socioeconomic activities, which cannot be employed for
prediction. Desyllas (2000) explored the relationship between urban street config-
uration and office rent patterns in Berlin by multiple regression analysis. A number
of spatial and non-spatial independent variables (e.g., pre-letting time, unit floor
space size, and indexes and multipliers) were used in the determination of rent in
his research. The results showed a shift in the pattern of location rent within the
short period under analysis. Whereas the Western CBD was the peak area of
location rents from 1991 to 1994, the pattern by 1995–1997 had shifted to Mitte.
This was offered as a direct evidence that prime location rents have reorganized
around the new spatial structure of the city revealed by the measure of space syntax
global integration. His finding is supported of the idea that spatial integration has an
explanatory value for office rent.
Chiaradia et al. (2009) examined the relationship between street layout and
residential property value by space syntax segment measurement (integration and
choice). Their study applied Council Tax Band as a proxy for residential property
value and considered other attributes, including non-residential land uses (mainly
retail), property size, building centered density, and age. Their findings showed that
integration is much stronger than choice value, and integration is strongly and
positively related to higher tax bands. When property size, age, and building
ambient density are added, space syntax spatial locational variables are slightly
weakened but still contribute to the property value. The likely reason is that
betweenness has a high variance within the tax band while closeness tends to be
more homogenous across locally continuous roads.
Matthews and Turnbull (2007) examined how street layout affects property value
by using space syntax network analysis for both eastern and western lake
Washington. They used two methods to measure street layout indices: space syntax
integration and ratio of segments/intersections. They found integration significant in
both Western and Eastern samples using a 1400-feet walking distance radius, but the
coefficient signs are opposite in the two samples. The ratio of segments/intersections
had the same result. They concluded that the portion of house value contributed by
street layout critically depends upon the context of the surrounding developmental
pattern. Enström and Netzell (2008) also used space syntax methods in a hedonic
price approach to test the urban street layout impact on commercial office rents in
Stockholm. They approached space syntax integration and found that it showed a
positive impact on office rent. More generally, street layout added an additional
explanatory power to hedonic models of office rent levels in a major city.
Several points need to be noted about space syntax use in causal modeling.
Firstly, most studies connecting space syntax to other academic fields apply the
3.6 Developments of Space Syntax Theory 59

space syntax topology measurement, which has been criticized by Ratti (2004) for
its inconsistencies. However, it seems that Ratti exaggerated the critics of space
syntax. The main point of Ratti is that two quite similar axial maps could have two
distinct results. In that case, the inappropriate axial map creation would bias the
analysis results. Indeed, the axial map is unique theoretically, which is proved by
the works of Peponis et al. (1998), Turner et al. (2005), and Jiang and Liu (2009).
Therefore, it is expected that two axial maps should have two different results even
they are similar, as the space interpreted by axial map is actually “syntactically”
different. Furthermore, regarding to the axial map edge effect pointed out by Ratti
(2004), the phenomenon that “cross-error” would happen in topology measurement,
because it is only one dimension. Indeed, considering the notion of metric, the
problem has been addressed (Hillier and Penn 2004), which also implies that there
is no intrinsical inconsistency of topology measurement for axial map, but it seems
difficult to interpret in different urban systems with certain social activities of
people. Therefore, metric segment and geometric analysis seems to be more
advanced, as different metric radii could interpret different kind of interactions
between built environment and social processes.
Secondly, correlation is not prediction, so space syntax should be supported by
statistical data, accounting for other attributes in urban system, such as land use, the
composition of the population, the location of houses and hobs and services and so
on (Ratti 2004). For example, location theory suggests that accessibility could be the
most important determinant of land value or property value. Although there is a
strong correlation of urban form and property value, few studies address the question
of whether urban from can be treated as a determinant of property value statistically.
Thirdly, the theory has not outlined clearly whether the relationship between
built environment and human societies could vary cross-space or different cultural
contexts. However, there is broadly a cognition of spatial heterogeneity of human
activities (Collinge 1996; Monn 2001).

3.7 Summary

In this chapter, I have briefly reviewed space syntax methodology, which is dual
approach to network analysis, measuring the accessibility information contained in
a model of urban morphology. Although some researchers have pointed out the
inconsistencies in its topological distance measurement, a series of improvement
had been made for the previous critics, such as metric segment analysis and
geometric angular cost segment analysis.
As such, these researches will approach the geometric angular segment analysis
measurement of space syntax. The reasons for choosing this measurement are as
follows: There are some inconsistencies for typology distance measurement, and
metric segment analysis does not consider depth “cost” between two axial lines
according to their angle of incidence. Hillier and Iida (2005) found that angular
segment analysis is efficient and robust for measuring pedestrian movement in a
60 3 Space Syntax Methodology Review

large area of London. In a city-wide level, both vehicular and pedestrian, users are
assumed to be sensitive of the cost to angular of street layout, and making 135° turn
seems to cost more than making 90° turn. Furthermore, in these studies, I adopt
Turner’s suggestion (2007) on creating the axial map based on road central line.
The space syntax axial map was created as follows:
1. A vector urban road network map was prepared and imported into Map info 9.5.
2. An axial line map was created based on the principle of road central lines.
3. The map was modified to take into account of connections in 2D that do not
hold in 3D (such as “intersections” involving viaducts and tunnels).
4. Using confeego 1.0, space syntax accessibility values were calculated for each
segment.
Two essential indices of space syntax will be applied for capturing the acces-
sibility contained in urban configuration, namely integration and choice. Their
formulas (Turner 2007) are provided as follows:

3.7.1 Radius Integration

Space syntax integration measures to what extent a segment is close to all others
segments along the shortest depths (paths) of the network. Space syntax integration
for a segment i is defined as:
PN
ðN  1Þ ‘i
Integrationi ¼ PN i
ð3:6Þ
j¼1;j6¼i di;j ‘i

where
N is the total number of segments in the network,
dij is the shortest depth between segment i and j, and
‘i is the search length through segment i.
The integration value commonly measures how well connected or how close one
node is to all other nodes within the search radius. If the integration value is high, it
means it has a high spatial accessibility, as it is easier to move from one segment to
others. Integration has been widely been used in urban and regional analysis as a proxy
for the cost of overcoming spatial separation within a network of road segments.

3.7.2 Radius Choice

Space syntax choice is based on the idea that a segment is more central when it is
traversed by a larger number of the shortest paths connecting all couples of seg-
ments in the network, which is defined as (Turner 2007):
3.7 Summary 61

N X
X N
choicex ¼ r‘ ði; x; jÞ i 6¼ j ð3:7Þ
i¼1 j¼1

where
rði; x; jÞ is the number of shortest paths from segment j to k through x,
r‘ is the weighted sigma function.
Using a social network analog, choice is like the kind of prominence of a person
who acts as an intermediary among a large number of other persons. The choice can
demonstrate how movements go through the network within a specific radius, and
many studies have used different scales of choice to simulate movements associated
with different travel modes. Generally, a high value of choice implies high fre-
quencies of movement passing through the segment, which is expected to have a
negative impact on residential property value.
I believe the new development of space syntax could improve its measurement
in explaining the interaction between urban spatial structure and socioeconomic
activities.
In addition, I have also reviewed the studies that connect space syntax to other
social science fields, especially land value theory. Location is considered as an
essential determinant of housing price distribution, but currently most researchers
ignore the spatial information hidden in an urban street grid. Given that a number of
researchers in the space syntax community have confirmed that there is a high
correlation between urban morphology and housing price pattern, there would seem
to be a fruitful research agenda. Hence, in the next chapters, I explore both the static
and dynamic relationships between urban morphology and housing price, supported
by statistical data that accounts for other attributes in urban system. Specifically, I
will look at (a) the basic relationship between urban morphological measures of
accessibility and house price, holding other price determinants statistically constant;
(b) the use of space syntax in delineating housing market areas; and (c) the dynamic
relationship between changes in systemic accessibility (through incremental net-
work improvements) and house price.