The Journal of Space Syntax The Journal of Space Syntax The Journal of Space Syntax The Journal of Space Syntax The Journal of Space Syntax The Journal of Space Syntax ISSN: 2044-7507 Year: 2010. V olume: 1, Issue: 1 Online Publication Date: 14 July 2010 http://www .journalofspacesyntax.or g/ Solutions for visibility-accessibility and signage problems via layered-graphs Nicholas Sheep Dalton 1 and Ruth Conroy Dalton 2 1 The Open University 2 Bartlett School of Graduate Studies, University College London Pages: 164-176
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Journal of Space Syntax, Volume 1, Issue 1, Pages 164-176, 2010
Solutions for visibility-accessibility and signage problems via lay-
ered-graphs
Nicholas Sheep Dalton1 and Ruth Conroy Dalton2
1 The Open University2 Bartlett School of Graduate Studies, University College London
Abstract
One of the rare representational problems encountered in space syntax analysis arises during the
construction of axial representations of architectural configurations and is the so called 'visibility-
accessibility problem'. This describes a situation where it is possible to see a space but not to be able
to directly move towards it. This condition arises in a number of cases, for example, in an office
containing half-height partitions or glass walls, in an urban street where safety barriers prevent indis-
criminate pedestrian movement or in an atrium-building that permits direct views to the second story
but does not facilitate direct access to those same spaces. This paper introduces a new spatial repre-
sentation called the multi-layered network that is intended to serve as a more generalized representa-
tion of topological spaces than previous representations. Evidence is presented to substantiate aspects
of the proposed representation based on an extension of current topological representations and asso-
ciated computational methods. A software implementation of the multi-layered network is demon-
strated along with examples from a sample of hitherto 'problematic' cases. Finally, it is argued that
this new multi-layered representation could equally be used as the underlying mechanism for a spatial
representation that might be able to accommodate signage-information, for example, and so, by ex-
tension, could establish testable conditions with the potential to measure the effect of signage 'catch-
ment areas' and signage placement in a building.
Keywords: space syntax, representation, visibility-accessiblty, graph, integration, navigation, signage
1. Introduction
Space syntax (Hillier and Hanson, 1984; Hillier, et al., 1993; Hillier, 1996, 1999) has proven to be a
powerful predictor of movement at the urban level and capable of providing insight into patterns of
usability at the building level. One problem discovered when mapping buildings is how to construct
a representation of a building where elements of that building can be seen but are not directly acces-
sible. One example of this is in a building with an atrium: a large, open, multilevel space in a building
allowing one to look up and observe spaces on subsequent floors, without affording direct access to
them (for example, by a directly visible staircase, escalator or lift/elevator; see Doxa, 2003; and
Parvin, 2008). While it could be argued that these kinds of spaces are rare and can be handled by
adhoc methodological means, it is the contention of this paper that these problems provide an impetusto reinterpret the representations used by space syntax theory.
Journal of Space Syntax, Volume 1, Issue 1, Pages 164-176, 2010
mapped twice in both ways). In each case, the analysis would be adequate but not entirely satisfac-
tory. Intuition suggests that the space is neither irrevocably complex nor a trivial single space but, in
fact, somewhere in between these two extremes.
A second, similar and connected problem is that of the effect of signage in a building or
quasi-urban environment like the Barbican Centre (a mixed-use residential and arts centre) in Lon-
don. Most signs serve to indicate the distal presence and initial heading of a room, space or group of
spaces (such as a department). The effect of a sign is to reduce the conceptual and topological dis-
tance from one space (of a situated observer) to many others, creating an effect somewhat similar to
a 'hyperlink'. Currently, there is no clear mechanism in space syntax analysis to accommodate the
effect of signage in a complex building. Like 'programme' and other factors influencing movement
patterns in a building, signage is just one of the assumed reasons why the correlation between space
and observed movement rarely exceeds a 'threshold' correlation coefficient of approximately 0.7 to
0.8 r-squared.
The axial map is a transformation of geometric space into a graph (or network) representa-
tion; there is no facility to alter this objective spatial description to permit signage to be taken into
account. Yet the effects of signage are neither random nor unbiased: signs are situated in one location
and point to, or link to, another location. Signs exist to make space more navigable or even 'intelli-
gible' in the Hillierian sense (Hillier, 1996), something that, at the urban level, is normally attributed
solely to space. Signs are deliberate attempts to alter how the space in a building is experienced but,
unlike space, they tend to eliminate irrelevant information creating pseudo-spatial asymmetries. For example a building-user may observe a sign indicating the direction to a toilet but are unlikely to find
one indicating the direction of the air-conditioning plant or the cleaners' cupboard. Signs are also uni-
directional, for example, a visitor may regularly encounter signs indicating the way to the toilets but
it is rare to have one saying for example 'this way back to, the restaurant'.
Signage is one of the ways in which occupants of or visitors to large complex buildings can
overcome the deficiencies of poor configuration or simply accommodate the complexities of large
buildings. Given that signage is frequently used in a remedial way (small building such as domestic
dwellings tend not to need signs indicating the 'Kitchen', or 'Living Room') to overcome what are
essentially spatial problems, we have, as yet, no representational mechanisms to understand the spa-
tial context of signs. Crucial to the integration of the use of signage into a representation such as space
syntax is an interpretation of the effect that a sign has on conceptual complexity. Imagine, for ex-
ample, standing in a typical hospital such as described by Peponis (Peponis, et al., 1990). One might
ask directions to a specific ward/room and be given the instructions, 'go down here turn left then
simply follow the yellow line down main corridor'. In space syntax analysis the concept of a change
of direction being synonymous with a step change in depth is common but following 'the main corri-
dor' might actually involve several changes of direction, especially if these did not include decision
Journal of Space Syntax, Volume 1, Issue 1, Pages 164-176, 2010
Figure 5 shows an axial map representation of a courtyard-building containing a second
storey (to the right) which forms an atrium at the lower floors. At the top is a 'superlink' representing
a staircase between the ground and first floors. The two cyan-coloured links represent the visibility-
relationship between the ground-floor entrance-space and an upper level passageway. In this case, the
integration values have been computed as if the visibility link were a full spatial link: as if a staircase,
for example, linked the courtyard to the upper floors. It can be observed that such visibility links
cause the integration core to shift towards this conceptual 'staircase' (the dotted, super link). This
causes the upper floors to be more integrated than might be expected. All calculations and subsequent
visualisations were performed using Webmap@home version 0.91.1 (Dalton 2005).
The primary benefit of the layered-graph representation is that we can consider simulta-
neously two graphs/networks. First, the accessibility graph (N, EA), which must define the lower
bound of the integration structure of the graph. Clearly, it is unreasonable to assume that the integra-
tion values of the upper floor will decrease if it is visible (linked in (N, EV)). We can naturally expect
that if the upper floor is directly visible, the degree of its relative segregation might decrease or, at
worst, remain the same. Second, the visibility graph (N, EV) defines the upper limit of the integration
values. Namely, if it were treated as an accessibility graph, (one could almost imagine a visitor float-
ing magically upwards, along the 'visibility axis' towards the upper floor) then the upper floor could
never be more integrated than if it were directly connected in some physical manner.
Restating these intuitions in space syntax terms: if the step depth from the central, lower
atrium space to one of the upper levels is calculated, then the depth to the upper level space should lie between the shortest, 'accessibility' geodetic depth (9 steps in the accessibility network, in this ex-
ample) and the shortest, 'visibility' geodetic depth (3 steps in the visibility network). Thus it should be
possible to formulate a function, f, which will take into account the effect of the impact of spatial
inter-visibility on standard accessibility graphs:
f(Dea
( i,j ), Dev
(i,j))
Where Dea( i,j ) is the geodetic distance between node 'I' and node 'J' and Dev
(i,j) is the
shortest geodetic distance for the accessibility layer/graph and Dev
(i,j)) is the equivalent distance in
the visibility-layer. It might be surmised that the simplest function to measure the reduction in segre-
gation caused by the effect of spatial inter-visibility is to interpolate the new depth between the first
and second cases (accessibility and visibility). The simplest means to do this is to use a linear interpo-
Solutions for visibility-accessibility and signage problems via layered-graphs
Where ! is a constant factor taking a value between 1 and 0. In the above example, given a
! -factor of 0.5, it could be established that the layered-graph depth is ( (9*0.5)+ (6*0.5)) = 7.5. This
! =0.5 example demonstrates a situation where accessibility and visibility contribute equally to the
overall depth of spaces. Clearly this factor can be used only for positively (i.e. not negatively) weighted
graphs and can be generalised, if necessary, for a greater number of layers/edge sets (one layer for
accessibility, one for visibility and one for signage, for example).
Given the definition of layered-graph, geodetic distance provided above, a familiar J- graph
representation (all shortest paths to node 'A') can be constructed and, from this, the total and average
depths can be computed in the same manner as for non-layered J-graphs.
Figure 6 provides a visual explanation of this process: each part of the figure shows a J-graph with a origin-node, A, of the graph presented in figure 3. In this figure, the accessibility connec-
tions are shown as black lines and the visibility connections are shown as dotted, red lines. The
leftmost part (a) of figure 6 shows the typical J-graph as defined in the Social Logic of Space (Hillier
and Hanson, 1984). The rightmost part of figure 6 (c) shows the visibility J-graph, where every
visibility-link is considered to be a single step in the J-graph. It can be seen that, in this case, space H
is now at the same depth as spaces B and D due to its visible connection to space G.
The middle diagram of figure 6 (b) presents the new, hybrid J-graph showing the effect of
applying the depth formula as outlined above. In this case, the ! -factor is 0.5. By examining space-
G, for example, it can be seen that it is now situated between depth 4 (its 'accessibility' level) and
depth 1 (its 'visibility' level). The formula calculates a new depth, of 3, for space-G in the hybrid J-
graph. In the pure visibility case (c), space-H has a depth equal to spaces D and B which is counter-
intuitive. In contrast, in the hybrid J-graph, space-H still occupies the most distant location from A,
which is aligned with our intuition.
If the axial map first presented in figure 2 is processed in the hybrid-manner described above
(merging both accessibility and visibility links), the result is that the total depth values for the upper
floors are lowered, as originally hypothesized. The presence of the visibility links reduce the degree
Figure 6. a J-graph from space-A from figure 3 for a) accessibility b) hybrid accessibility/visibility c) visibility
Journal of Space Syntax, Volume 1, Issue 1, Pages 164-176, 2010
of segregation of the upper level. The ! -factor of 0.5 is just a example intended to demonstrate the
influence of the visibility links on the resultant patterns of integration. By examining figure 7, it can
be seen that the integration core of the building moves away from the courtyard entrance, yet it
remains more integrated than in the visibility-only case. The upper floor becomes more integrated but
not as integrated as in the pure visibility case in figure 5. Overall, the integration pattern shifts so that
the upper and lower spaces become equally accessible (in this example, it is assumed that the lower
floors do not have windows overlooking the ground floor) and the back plane becomes the most
integrated part of the building.
Table 1 shows the values of total depth computed for each node in the visibility (EV), the
accessibility (EA) graphs. The third column contains the values which have been computed using thenew values of depth for the layered-graph method with a !-factor of 0.5. As to be expected, the
visibility depths are always lower than the accessibility ones (making a space visible lowers its con-
ceptual distance from everywhere else). From this case it can be see that it is possible to interpolate
between the visibility and accessibility total depths and, by implication, it should also be possible to
interpolate between the global integration (radius infinity) values. While this could be used as an
approximation, in many cases, this could mask any benefits of this approach. First, the interpolation
between integration values would break down in the case of using some non-global depths (for ex-
ample radius 3) as the value would shift due to the change in connectivity. Second, the interpolation
of global values also fails when asymmetric values are used, these would occur in the case of signage
(A points to B which does not point back to A) and may also appear in the case of visible upper levels.
For example, it could be thought that looking down on a lower level might be more likely than
looking up; in this case a different !"factor could be applied depending upon the direction of gaze. As
such the layered-graph computation has certain advantages and hence was the method programmed
Journal of Space Syntax, Volume 1, Issue 1, Pages 164-176, 2010
pirical and observational work will need to be completed and, to this end, the software which has
been developed to perform the accessibility-visibility calculations, for axial line analysis, is to be
made freely available.
Extensions
Given the layered-graph mechanism, it might be fruitful to explore the utility of this representation to
other problems. Consider, once again, figure 7 and, this time, imagine that the axial line graph repre-
sents a non-atrium building: the cyan-coloured links, instead of representing an inter-visibility condi-
tion between two spaces, represent two different signs indicating the presence and the initial direction
of travel of two spatially remote and physically unconnected spaces. Again, the core hypothesis of
this paper is that the mere existence of a sign certainly cannot make the target location appear to be
any further away compared to its actual 'accessibility-distance' and certainly can not cause it to appear
any closer than if it were little more than a single change of direction away, must still hold true.
Observe also that, in this case, the sign-as-link is one way (it is a directed graph) insomuch as a sign
may indicate the presence of a destination from an origin, but once that destination has been reached
the same sign cannot be re-utilised to return to the origin (although, in rare situations, a second sign
might serve this purpose). The effect of signage is as if a building has been artfully constructed using
numerous one-way mirrors, providing brief glimpses to spatially remote locations. The layered-graph,
as a description of a building's signage system, could also be used to form simple experiments. If
pedestrian, mean flow-rates at regular 'gates' within a building could be observed both with and
without the effect of signage, then it might be possible to calculate the !-factor value which is neces-
sary to calibrate the signage-link model. Clearly it is simpler to alter a buildings' signage (to someextent) than it would be to modify the inter-visibility of spaces within a building and so form a viable
empirical test of the layered-graph representation.
Layered-graph and intelligibility
It has become apparent, that one effect of signage is to reduce the overall segregation (or raise the
mean integration) of a complex building. Another effect that might be anticipated of a layered-graph
representing signage-links is that it might raise the apparent intelligibility of a sufficiently complex
building (since the intelligibility of small systems will always be high, this effect must be confined to
systems in excess of approximately fifty spaces). This hypothesis could be tested via the basic obser-
vation that the intelligibility of the layered-graph of a well signed building should be reasonably
expected to be higher than that of the accessibility-only graph. This leads to one theoretical problem:
since the calculation of intelligibility depends upon a correlate of the measure of a node's connectiv-
ity, what precisely is connectivity in the context of a layered-graph? Each and every node has two, or
more, edge sets leading from it so the degree of a node (its connectivity), instead of a single integer
value, is transformed into an ordered list of degrees drawn from each edge set. That is to say, for each
layer there is a different connectivity/degree. It would be reasonable to assume that one definition of
connectivity, in the context of a layered-graph, would be to apply the !-factor to the degree/connec-
tivity value. So for a node, N, with two Edge sets, EA and EV,
Solutions for visibility-accessibility and signage problems via layered-graphs
K(N) = k(N,EA) * ! !
! ! ! + (1-!!!!!) * k(N,EV)
Where K(N) is the degree/connectivity of node, N, and k(N, E) is the degree of edge set E for
node N, given this equation for K, an approximation of intelligibility can be duly created.
3. Conclusions
This paper has introduced the concept of a layered-graph and shown how it can be used in space
syntax to represent a wide range of real-world situations including a solution to the visibility-accessi-
bly problem and, potentially, signage in buildings. This layered-graph approach appears to be a fur-
ther generalisation of the current space syntax representations. For example, when the more general
case of accessibility being directly equivalent to visibility, then the layered-graph is reduced to the
simpler graph typical of space syntax methods. While this paper has not incorporated either the effect
of angular analyses or other graph-weightings, the concept of the layered-graph is not antithetical to
these representational methods. It is evident that the layered-graph approach can be used to represent
a number of different problematic cases found in real world buildings and even urban environments.
As such, it seems reasonable to warrant further research and testing in this area.
References
Bakke, J.W. and Yttri, B. (2003), “Hybrid infrastructures for knowledge work,” in Proceedings of the Fourth International
Space Syntax Symposium, London. pp. 17-19.
Benedikt, M.L. (1979), “To take hold of space: isovists and isovist fields,” Environment and Planning B, vol. 6 no 1, pp. 47-65.
Dalton, N. (2005), WebmapAtHome, available for down load from http://tinyurl.com/34nhq5a.Davis, L.S. and Benedikt, M.L. (1979), “Computational models of space: Isovists and isovist fields,” Computer Graphics and
Image Processing, vol. 11 no 1, pp. 49-72.
Doxa, M. (2001), “Morphologies of co-presence in interior public space in places of performance: the Royal Festival Hall and
the Royal National Theatre of London,” in J. Peponis, J. Wineman, and S. Bafna, eds., Proceedings, 3rd International
Space Syntax Symposium, pp. 16.1-16.15, A. Alfred Taubman College of Architecture and Urban Planning, Michigan..
Harary, F. (1994), Graph Theory, Westview Press, Colorado.
Hillier, B. (1996), Space is the Machine, Cambridge University Press, Cambridge (MA).
Hillier, B. (1999), “The Hidden Geometry of Deformed Grids,” Environment and Planning B, vol. 26, pp. 169-191.
Hillier, B. & Hanson, J. (1984), The social logic of space, Cambridge University Press, Cambridge (UK).
Hillier, B. & Iida, S. (2005), “Network and psychological effects in urban movement,” in A.G. Cohn and D.M. Mark, eds.,
Proceedings of Spatial Information Theory: International Conference, COSIT 2005, Ellicottville, N.Y., U.S.A., September 14-18, 2005, pp. 475-490, Springer Verlag, Berlin.
Hillier, B., Penn, A., Hanson, J., Grajewski, T., and Xu, J. (1993), “Natural movement: or, configuration and attraction in urban
pedestrian movement,” Environment and Planning B: Planning and Design, vol. 20 no 1, pp. 29-66.
Kostakos, V. and O'Neill, E. (2007), ”Quantifying the effects of space on encounter,” in A.S. Kubat, Ö. Ertekin, Y.I. Güney,
and E. Eyübo_lou, eds., Proceedings, 6th International Space Syntax symposium, Istanbul, vol. II, pp. 097-1 - 9, ITU
Faculty of Architecture, Istanbul.
Major, M.D. and Penn, A. (2008), “Following the crowd: spatial layout and crowd behaviour,” in P. Carolin and T. Fisher, eds.,
Architectural Research Quarterly, vol. 4 part 3, pp. 257-264, Cambridge University Press, Cambridge (UK).
Parvin, A., Min, A.Y. and Beisi, J. (2008), “Effect of visibility on multilevel movement: a study of the high-density compact
built environment in Hong Kong,” Urban Design International, vol. 13 no 3, pp. 169-181.
Peponis, J., Zimring, C. and Choi, Y.K. (1990), “Finding the building in wayfinding,” Environment & Behavior, vol. 22 no 5, pp. 555-590.