A Comparison Study on Space Syntax as a Computer Model of Space B. Jiang† and C. Claramunt‡ †
Centre for Advanced Spatial Analysis University College London 1-19 Torrington Place, London WC1E 6BT, UK Tel: +44 171 391 1255, Fax: +44 171 813 2843 Email:
[email protected] ‡
Department of Computing The Nottingham Trent University Nottingham, NG1 4BU, UK Tel: +44 115 848 4289, Fax: +44 115 948 6518 Email:
[email protected]
Abstract Over the past two decades, space syntax has been extensively researched with a considerable amount of empirical case studies (Hillier and Hanson 1984, Hillier 1996). It has been found that space syntax can be a valuable tool for the prediction of people’s movement in urban environments. Basic to this finding is the relationship between human spatial behaviour and urban morphological structure, i.e. how people’s movement is affected by spatial perception. This paper intends to provide an analysis, based a comparison study, of space syntax as a computer model of space, i.e. a unique vision of space syntax in spatial modelling from the point of views of spatial perception and cognition. In this context, space syntax is examined in details in terms of representation and analysis capabilities, in comparison to the human and computer models of space in Geographic Information Systems (GIS). We believe that such a discussion can benefit to the space syntax research community by providing more evidence to support space syntax as a valuable tool for many disciplines and application areas. Keywords: Space syntax, spatial behaviour, urban configuration, GIS.
1. Introduction Conceptualisations of geographical spaces are based on the application of different paradigms, for example, from cognitive to symbolic spaces, and from physical to experiential spaces. Over the past two decades, space syntax has appeared as a new language to describe space in the study of the complexity and spatial patterns of modern cities. From its origin in urban research, space syntax proposes a language of space that could be of interest for many research and application areas involved in the description and analysis of spatial patterns. We believe that space syntax could provide a new vision of space for the representation of urban systems within GIS, and more generally for systems dealing with spatial configurations. However, to achieve such an objective some further theoretical developments are still required. As stated in (Hillier 1998), “Space syntax analysis turned attention away from geometrical notions of spatial order and pointed to spatial-functional patterns which are formally speaking
closer to topology than to geometry”. Therefore an analysis of the geometry and topology of space syntax is still an open issue that is especially of interest for the integration of space syntax as a theoretical support for GIS. In related previous work we have discussed and introduced some new potential of space syntax for the development of spatial structures and analysis functions (Jiang 1998, Claramunt and Jiang 1998, Jiang and Batty 1998). However, the contribution of space syntax, as a specific model of space, has not been studied in comparison to current models of space used in GIS. In this context, we propose to introduce a comparative study of the potential contribution of space syntax with respect to current spatial data models and principles used in GIS. This comparison study focuses on the resemblance of spatial models to the way people represent and perceive space, i.e., from the point of view of mental and visual perceptions. The remainder of the paper is organised as follows. Section 2 examines the computer models of space currently used in GIS, and their adaptation to the way human model space. Section 3 focuses on a unique vision of space syntax to spatial representation. In section 4, we address the contribution of space syntax to spatial analysis. Finally section 5 summarises the conclusions and proposes some future work.
2. Models of space in GIS Within GIS the relative and absolute visions are the two main concepts used for the representation of environmental and urban systems, i.e., space as an attribute of objects or space as a container, respectively (Peuquet 1988, Couclelis 1992). These visions support two different perceptions of space: the relative view considers space as a collection of objects since the objects themselves are the space (Nunes 1991), the absolute vision considers space as the content of things. From these two paradigms, which are analogous to the philosophical and physical visions of space, different spatial models are currently developed and used within GIS. The relative view is oriented toward the description of objects and their relationships. The absolute view is more concerned with the representation of the properties of space. As a primitive form of spatial language, these data models reflect the way human beings translate their cognitive representations in mapping concepts. A human model of space uses many concepts and abstraction mechanisms such as perception, metaphors, cognition combined with cultural and experiential influences. A human model of space provides a conceptualisation of space that acts as a support for the development of a computational model of space (Figure 1). Real world
Human models
Computer models
Figure 1: Spatial modelling in GIS From a spatial modelling perspective, different spatial structures are currently used for the representation of space: the object model of space which leads to a vector representation of space with the application of geometrical and topological principles, and the continuous representation (i.e., field model) of space which implies the definition of a tessellation with the application of some basic geometrical principles1. Continuous representations of space are either based on a regular tessellation, e.g., raster or grid spatial structures, or on a irregular tessellation, e.g. triangular or Voronoi diagrams. A vector representation of space (figure 2) is based on some well-known Euclidean and topological properties of space that together allow the geometrical description of a set of objects in a geographical space. Spatial relationships are described (e.g. adjacency, connections), application of simple or complex metrical and geometrical operations are the most common used operations. Moreover, the concept of object fits very well with the development of spatial databases that describe the thematic attributes and properties of these objects. The raster view of space (figure 2), on the other hand, favours a field representation of space that describes the distribution of a continuous phenomena, e.g., regular tessellation. An irregular tessellation of space covers a geographical space by a set of non overlapping polygons. With the development of GIS, many irregular tessellation structures have been used (e.g. triangular irregular networks, Thiessen-Voronoi diagrams). Particularly, Voronoi diagrams are based on a geometrical analysis of space: informally from a set of points in a planar region, each point is surrounded by an area that contains the points of closest proximity.
Real world
Human models
Map
Computer models Raster
Vector
Figure 2: Human and computer models of space in GIS
1
As stated in (Nunes 1991) both vector and raster structures of space are only geometric representations.
“People manipulate objects (but cultivate fields)” as nicely stated in (Couclelis 1992). Spatial data structures provide some useful computational representations of space. However, how do these models encompass the way people elaborate a spatial knowledge ? As previously stated the relative view of space, that describes objects as the space, provides an interesting similarity with the way we identify “things” in reality. Continuous representations could also provide a correct support for the cognitive description of natural landscapes or built environments. However the geometrical organisation of space, whatever the kind of spatial data structure used, is quite distant from a human spatial cognition. A similar observation can be made for irregular representations of space (i.e., raster and Voronoi models) that are mainly based on a geometrical description of space that cannot be considered as a cognitive and mental vision of space. Thus, if some of the characteristics of these spatial data structures are relevant for spatial cognition, these geometrical descriptions of space do not provide a complete relevant support for the description of human perceptions and behaviours in space. There is still a need for some form of data concept and structures in space that better represent the way people interpret and behave in spatial environments. Within GIS, new initiatives are in progress to create new spatial data representation and reasoning models, which are more suited to human perception and cognition. These initiatives are mainly based on the fact that spatial perception and cognition integrate many concepts and mechanisms which are far from a strict geometrical representation of space. Particularly the emerging domain of naive geography, extended from naive physics, which attempts to formalise and identify models which are closer of a common-sense geographic world and the way people behave in real world (Mark and Egenhofer 1995). Human behaviours in space constitute some recurrent patterns that people learn through physical and repetitive experience (Lakoff 1987). They are composed of parts and relationships that organise objects in space (i.e. possible relationships between an object and a container, a part and a whole). However, it is still recognised that the development of naive geography is limited by a lack of formalization and application validations. Space syntax can also be considered as a new language of space. Some challenging questions are: how could space syntax be formalised to provide a closer description of human perception and behaviour in space? How does space syntax compare with current spatial data models within GIS? Is space syntax closer to a classic geometrical representation of space or to a common-sense representation of an urban system?
3. Space syntax representation of space Traditional maps provide a support for visual representations of space as a continuous surface or as a collection of objects. Such representations are assumed from the point of view of an observer in the air without distortion at edges. These two observations set two basic distinguished perspective of space syntax in the representation of space. Firstly space syntax is only concerned with spaces with free movement; secondly space syntax provides a sort of egocentric point of view of space as it reflects the point of view of an observer moving in an urban system.
Space syntax uses axial maps and convex maps as representations of an urban system for further analytical purpose. The two maps are referred to as syntax maps in the context of this paper. In this section, we present an analysis of these representations, for instance, how useful they are in understanding and reasoning in space configuration? How do they differ or resemble to the human visual perception of spatial processes?
3. 1 Open space perspective (traditional maps and syntax maps) Space syntax models an urban system by concentrating on open spaces. An urban system involves two parts: one termed as closed spaces and another termed as open space. The open versus closed spaces distinction is generated by the existence of boundaries between the streets and the built environment, i.e. open and closed spaces are interdependent as they share a common physical boundary. We can remark that such a property is similarly defined within existing boundary theories (Smith and Varzi 1997). These theories provide a theoretical support for modelling the spatial configuration of an urban system. Space syntax is based on human or cognitive models that manipulate open and close spaces. Within urban systems, an open spaces is explicitly represented as the centre of interest. Closed-spaces are described implicitly as they are derived from open space boundaries. To date the computer representation of space syntax in urban systems is generally concerned with the computational representations of the open space. Moreover and as traditional maps do, space syntax may provide a multi-dimensional representation of an urban space, i.e., the representation of an urban system at complementary abstraction levels (Claramunt and Jiang 1998). This property encompasses the accepted vision of a city as a hierarchical structure (Alexander 1982). The distinction between open and closed spaces relates to the way movements are possible within these structures: an open space allows (potentially) a human being to have free movement from any point to any other point in the open space, while a displacement within a close space is, by definition, restricted to the boundaries of this closed space. In the context of a city, urban blocks or plots are considered as closed spaces, while streets and squares as parts of the open space. The human perception and understanding of the city or the formation of mental representation is through navigating, exploring landmarks and places, alternatively by reading city maps. Traditional maps, whatever the spatial data structure used (i.e., object or continuous based), provide a relevant representation of an urban system, with closed and open spaces represented graphically (figure 3). These maps can be helpful in our daily life as well as for urban studies and research. As previously stated, GIS provides some powerful functionality for spatial analysis and reasoning in urban spaces.
Real world
Human models
Axial map
Map
Computer models
Vector
Raster GIS vision
graph Space syntax vision
Figure 3: Human and computer models of space An open space provides a unique vision in understanding the configuration of an urban system. Open space allows the detection of the skeleton / structure of an urban system and the derivation of the so called configuration knowledge. Syntax maps, based on the open spaces provide a precise skeleton representation of an urban configuration and a reference for analytical measures. We believe that syntax maps bear a resemblance to human visual perception. Let us take a look at a human body as illustrated in figure 4, where the body shape is represented as the form of axial lines. This is much similar to the basic idea of medical axis transform (Blum 1967) concerning the shape classification of biological forms with the aim of creating mathematical tools for shape description. This work has been taken up in the image analysis, and from this much of current activity in pattern recognition and mathematical morphology has been developed (Serra 1982). In this case the space syntax is applied to the closed space, i.e., the human body. Such an example shows the flexibility of space syntax and its relativity in terms of the role of open and closed spaces.
Figure 4: Human body and its axial line representation
3. 2 Small-scale space perspective Real case studies and applications provide a useful validation for the acceptance and diffusion of space syntax. However, the definition of a formal theoretical support and the demonstration of its computational relevance, still need some additional developments. One of the potential problems we have identified, come from the axial line representation. For example, ring roads are very integrated streets, so why are they represented as many axial lines, each one represented as a very segregated segment ? The distinction between large- and small-scale spaces is a fundamental assumption for the application of the space syntax. From the point of view of spatial perception, we make a distinction between large- and small-scale spaces. A large-scale space is beyond human body perception, and cannot be perceived from a single vantage point; while a small-scale space is presumably larger than human body, but can be perceived from a single vantage point. It is our belief that to perceive large-scale environments, a human being updates his/her perception of small-scale space whilst moving along in the large-scale space. In other word, a perception of a large-scale space is generated from perception in small-scale spaces. Small-scale space is empty or at least it can be thought of as empty. For instance, a room may be occupied by some furniture like a table and chair, but one can perceive the room’s structure without any difficulty. Small-scale spaces are continuous (not discrete) and interconnected. When we are walking along a street, we perceive our surrounding environment as a small-scale space. Small-scale spaces are interconnected in the sense that a large-scale space consists of an infinite, but countable, number of small-scale spaces. The popular saying that all roads lead to Rome indicates that all streets are interconnected, in which one can go from anywhere to anywhere else. Small-scale spaces are interconnected to form the so called open space. In order to derive an analytical tool based on a small-scale space perspective, there is a need for, first, a representation of a large-scale space as a number of small-scale spaces and, secondly, a definition of a link between these individual small-scale spaces to form a graph. Figure 5 shows an example a closed building plan, and its related graph with each room or corridor represented as a small-scale space. Some points are important in the graph representation, e.g. how each node is connected to its immediate neighbours, and how each node is connected to every other node. Answers to these questions lead to some morphological measures given by space syntax parameters. 3
2 3
7 4
5
1
6
4
2
7
6
1
5
Figure 5: Closed building plan and its graph representation 4. Space syntax analysis of space
Within GIS, the study of a spatial configuration is based on the application of topological, metrical and neighbourhood properties. Qualitative reasoning provides a set of important primitives for the identification of topological and cardinal relationships between objects in space (e.g. Egenhofer and AI-Taha 1992, Frank 1996). Some basic functions like adjacency, proximity or buffers are some examples of functions that can provide a computational representation of these relationships (Gold 1992). In terms of computational analysis, in the raster structure of space, von Neumann and Moore neighbourhoods are often used, while in the vector structure of space, the neighbourhood is defined from the distances surrounding geographic objects and the application of topological relationships. However, it is difficult for these models to have a sense of context meaning, i.e., in addition to immediate neighbours, we need various degrees of neighbour, e.g., cardinal relationships (Frank 1996), relative distances (Worboys 1996). Moreover, many spatial operations are concerned with the context meaning. For instance, cartographic generalisation is not just an isolated graphic operation, it should be put in context by considering surrounding situation in order to produce a reasonable generalisation. In this connection, a network is a sort of context concept, i.e., each node has certain statuses in a network. Delaunay triangulation, based on the Voronoi model of space also provides such networks for context generalisation (Hangouet and Djadri 1998). In this respect, space syntax as well as the Voronoi model of space show some flexibility about the relationships of part and whole, because they have the same sort of dual graph representation.
4. 1 Analysis based on dual graph (spatial adjacency) Mathematically, syntax maps bear a significant similarity to the Voronoi model of space. Spatial adjacency is the central notion to the Voronoi model of space as elaborated by Gold (1993). Two polygons are considered to be adjacent if there is a common boundary between them, this adjacency is represented as edge in the dual graph. In syntax maps, spatial adjacency is also centre to the model. Two convex polygons are considered to be adjacent if they have a common boundary, and two axial lines are considered to be neighbours if they are intersected each other (figure 6). Connectivity is one of measures concerning spatial adjacency. It is defined by the number of edges of a vertex. Local integration shows another local property of space, not concerning immediate neighbours rather than those neighbours located a few steps away. In addition to the similarity one, a second difference does exist between Voronoi representations and axial maps. Within space syntax, an axial map is based on a visual perception, i.e., each axial line can be approximated as an individual space (see section 3.2), and adjacency is defined at the level of visual perception rather than physical. Each space (represented by either an axial line or a convex space) can be regarded as a homogenous space at the level of visual perception, while each space in the Voronoi model is rather homogenous at the level of geographic entities. Both models bear some resemblance to human visual perception processes. As stated by Edwards (1993) that “the Voronoi model of space concords closely with perceptual and linguistic spaces of humans and hence Voronoi zones around objects are meaningful. The Voronoi model of space is also a closer fit to qualitative date
representation and analysis than other models.” Fruitful application studies of space syntax has proved that it is a good model for human spatial behaviour prediction, say movement. However, space syntax has an important advantage over the Voronoi model in some sense, closed and open spaces correspond to some physical concepts we can perceive in space, on the other hand Voronoi zones are a result of a computation (i.e., derivation of the Thiessen-Voronoi polygons) and not directly present in the real world.
2
v2
v6
6 3
v1
1 5
v4
4
8
7 2 5
v7 v8 v2
1 3
v3
v5
v1
4 6
v5
v3
v4 v6
Figure 6: Voronoi representation and axial map and their dual graphs
4. 2 Local-global relationship Basically there are two types of morphological measures in space syntax: first local measures like connectivity, control value and local integration, secondly global measure such as global integration. These two types of measure provide an important perspective to an urban system perceived as large-scale space both from local and global viewpoints. In addition, the relation of these two types of measure offer another property about urban systems such as how a local area relates to a global one. Intelligibility is a term used in space syntax to describe the part-whole relation. It is defined by the co-efficiency of correlation between local and global measures. A local area is said to be intelligible if its co-efficiency value is higher than the one of global area as illustrated in Figure 7. The large oval represents the cluster of all spaces of a whole area, while the small oval represents the cluster of a selected local area. The meaning of intelligibility can be compared to the notion of self-similarity in fractal geometry. Self-similarity represents the fact that parts are considered to be similar to the whole, for instance, a tree branch is similar to a tree as a whole. Intelligibility has cognitive meaning in space syntax. As a recent effort to space syntax research, Kim (1998) has investigated how residents perceive their neighbourhoods and their city. Subjects (local residents in respective areas) are chosen from two local areas in the north west of London. They are required to sketch a more global area in terms of their understanding. Surprisingly, it was found that people living in intelligible areas have a more developed sense of global integration. Thus, from a cognitive point
Local measure
of view, intelligibility is the property of space that, in some senses affects the human perception of a part-whole relationship.
Global measure
Figure 7: Illustration of intelligibility In cognitive science, it is generally admitted that human cognition is generated from parts to the whole. The same evidence has been mentioned in section 3.2, where we have illustrated how a small-scale space perspective facilitates the representation of an urban system. From the analytical point of view, intelligibility help us in analysing this local-global relationship.
5. Conclusions This paper compares space syntax to other computer models of space in GIS both in terms of their adaptation to human models of space and of their analysis capabilities of a urban configuration. The unique vision of space syntax in spatial modelling is discussed and illustrated. As the result of the comparison study, we conclude that, as a model of space, space syntax provides some important advantages over other spatial models in modelling human spatial perception and cognition. Space syntax is based on a graph representation of space. How can this representation act as a support for a theoretical development that could act as a spatial data model within GIS, at both a model and computational level? As demonstrated in this paper, space syntax can be considered as a model particularly suitable the prediction of human spatial behaviour. These concepts include the respective contributions of open and closed spaces for human and computational models, small and large scale spaces at the perception level. Space syntax offers a complementary perspective to the current continuous or object representations of space. Particularly, one suggestion we can make, through our comparison study, is that the same sort of morphological analysis can be applied to a triangulation network, which corresponds for example to the dual graph of Voronoi polygons. Accordingly each polygon in space has both local and global properties. This study shows an interesting advantage of the space syntax in terms of its adaptation to human models of space. Space syntax has its origin in the study of the built environment. We believe that more experimental research is needed in spatial studies that are based on similar spatial structures. Indeed, it is quite surprising to see that the principles of space syntax haven’t yet been applied to the description of graphs in space that could be derived from any spatial configurations (e.g., urban networks
such as gas or electricity, biological or ecological systems). Such applications could integrate multi-disciplinary views and consequently provide insights that could serve for the development of a broader use of space syntax.
References Alexander, C. (1982), A City is not a Tree, in: S. Kaplan and R. Kaplan (eds.), Humanscape Environments for People, Ulrich’s Books Inc., Ann Arbor, pp. 377-402. Blum H. (1967), A Transformation for Extracting New Descriptors of Shape, in: W. WhatenDunn (ed.), Models for the Perception of Speech and visual Form, Cambridge, Mass.: MIT Press, pp. 153 - 171. Claramunt, C. and Jiang, B. (1998), A Multi-scale Approach to the Evolution of Urban Structures, submitted for publication. Couclelis, H. (1992), People Manipulate objects (but cultivate fields): beyond the raster-vector debate in GIS, in: Frank A. U. And I. Campari (eds), From Space to Territory: Theories and Methods of Spatio-temporal Reasoning in Geographic Space, Springer-Verlag, pp. 65-77. Edwards, G. (1993), The Voronoi Model and Cultural Space: Applications of the Social Sciences and Humanities, in: Frank A. U. and Campari I. (eds.) Spatial Information Theory: A Theoretical Basis for GIS, Springer-Verlag, pp. 202 - 214. Egenhofer, M. and Mark, D. (1995), Naive Geography, in: Frank A. U. and Kuhn, W. (eds.), Spatial Information Theory: A Theoretical Basis for GIS, Springer-Verlag, pp. 1-15. Egenhofer M. J. and AI-Taha K. K. (1992), Reasoning About Gradual Changes of Topological Relationships, in: Frank A. U., Campari I. and Formentini (eds.), Theories and Methods of Spatio-Temporal Reasoning in Geographic Space, Springer-Verlag, pp. 196 - 219. Frank A. U. (1996), Qualitative Spatial Reasoning: Cardinal Directions as an Example, International Journal of Geographic Information Systems, 10 (3), pp. 269-290. Gold, C. G. (1992), The Meaning of “Neighbour”, in: Frank A. U., Campari I. and Formentini U. (eds.), Theories and Methods of Spatio-Temporal Reasoning in Geographic Space, Springer-Verlag, pp.220 - 235 Hangouet, J. F. and Djadri, R. (1998), Voronoi Diagrams on Line Segments: Measurements for Contextual Generalisation Purposes, in: Poiker T. K. and Chrisman N. (eds.), Proceedings of 8th International Symposium on Spatial Data Handling, International Geographical Union, pp. 207 - 222. Hillier B. and Hanson, J. (1984), The Social Logic of Space, Cambridge University Press. Hillier B. (1996), Space is the Machine: A Configurational Theory of Architecture, Cambridge University Press. Hillier, B. (1998), The Common Language of Space, Forthcoming, University College London. Jiang, B. (1998), Axwoman: An ArcView Extension for Urban Morphological Analysis, Proceedings of Geoinformatics’99, 15-19 June, Beijing, pp. 394 - 401.
Jiang B. and M. Batty (1998), The Incorporation of Urban Morphological Analysis into GIS, CASA working paper series, 13 pages. Kim Y. (1998), Personal Communication. Lakoff, G. (1987), Women, Fire, and Dangerous Things: What Categories Reveal About the Mind, The University of Chicago Press, Chicago, IL. Nunes, J. (1991), Geographic Space as a Set of Concrete Geographical Entities, in: D. M. Mark and A. U. Frank (eds.), Cognitive and Linguistics Aspects of Geographic Space, Kluwer Academic Publishers, pp. 9-33. Peuquet, D. (1988), Representations of Geographic Space: Toward a Conceptual Synthesis, Annals of the Association of American Geographers, 78: 375-394. Serra, J. (1982), Image Analysis and Mathematical Morphology, Academic Press: London. Smith, B. And Varzi, A. C. (1997), Fiat and Bona Fide Boundaries: Towards an Ontology of Spatially Extended Object, in: S. H. Hirtle and A. U. Franks (eds.), Spatial Information Theory, A Theoretical Basis for GIS, Springer-Verlag, pp. 103-120. Worboys M. F. (1996), Metrics and Topologies for Geographic Space, in: M. J. Kraak and Molenaar M. (eds.), Advances in GIS Research II, Taylor & Francis, pp. 7A.1 - 7A.11.
