Architecture of Mind and World: How Urban Form Influences Spatial Cognition

We all require spatial knowledge of our environment. Many people spend the better part of their day in a built environment, and therefore, much of their thought about space is directly intertwined with the architectural and urban form of their surroundings. How does the form of people’s surroundings affect their spatial knowledge? Tversky (1981, 1992) has demonstrated that people’s spatial knowledge is systematically distorted by the use of heuristics to simplify alignment and rotation information. A set of computational techniques known as space syntax (Bafna, 2003) can be used to formally describe an environment. This study pairs such an analysis of a case study environment with results from spatial judgment and memory tasks. Findings suggest that space syntax measures significantly predict people’s performance on those tasks. ARCHITECTURE OF MIND AND WORLD 3 Architecture of Mind and World: How Urban Form Influences Spatial Cognition We all require spatial knowledge of our environment in order to travel to the grocery store and back, in order to answer such questions as “Which is further west, Reno or San Diego?”, and in order to provide route directions to visitors from out of town. How we acquire, store, and use this knowledge is of central interest to the field of spatial cognition. Many people spend the better part of their day in a built environment, i.e. the urban world (with varying density) of buildings, streets, parks, parking lots, blocks, and squares that has been designed and constructed by humans. Therefore, much of their thought about space (spatial judgment and memory) and movement through space (wayfinding) is directly intertwined with the architectural and urban form of their surroundings. How does the form of people’s urban surroundings affect their spatial knowledge of those places? In particular, how does urban form affect spatial judgment and memory? These questions, as well as the subject matter of spatial cognition more generally, are intimately connected with a number of other topics of study in the field of cognitive science. Spatial knowledge is commonly acquired through first-hand experience with an environment or through use of representations of that environment, such as maps. Taking in and perceiving this external information necessarily involves perception (usually visual or auditory). Moreover, in the case of first-hand experience, one’s body acts in and moves through the surrounding environment. Thus, issues of embodied/embedded action and cognition arise in the study of spatial cognition. Here I will focus specifically on the interplay of urban form with spatial judgment and memory. ARCHITECTURE OF MIND AND WORLD 4 Studying the effects of urban form on spatial cognition requires a formal system for describing the built environment. Space syntax, a set of techniques for analyzing the configuration of building interiors as well as urban environments, allows for such a computational characterization (Hillier and Hanson, 1984 is considered the seminal work). This study will pair these space syntax analyses with behavioral data on spatial judgment and memory tasks. Cognitive Maps, Spatial Judgment, and Spatial Memory The term cognitive map effectively refers to the mental representations that store a person’s spatial knowledge of an environment. However, the term is notoriously ambiguous – the form it takes on differs from theory to theory. Cognitive maps may be considered to be actual metric maps, map-like in form, acting like maps in practice, or convenient fictions (see Kitchin, 1994). For the purposes of this study, which focuses on performance in spatial judgment and memory tasks, I adopt the last view and simply consider a cognitive map to be one’s spatial knowledge of an environment. According to this perspective, a cognitive map is shaped by a number of different processes that do not necessarily follow the constraints of physical maps. Just as with most other processes of human thought, spatial cognition involves the use of heuristics. Tversky (1981) identified a set of basic heuristics that are used to encode and store memory of environments and maps, as well as of meaningless visual forms. “Remembering the absolute location of figures is difficult, and is facilitated by remembering locations relative to other figures and/or relative to the natural directions of the figure” (Tversky, 1981, p. 407). This simplification process is revealed in systematic errors, which Tversky attributes “to two heuristics that are derived from principles of ARCHITECTURE OF MIND AND WORLD 5 perceptual organization” (1981, p. 407). The alignment heuristic predicts that “two figures that are perceived as grouped together but are misaligned, that is, offset in one spatial dimension, are remembered as more aligned than they really are.” The rotation heuristic assumes that “figures are remembered with respect to a frame of reference [e.g. north-south, east-west], and that, when the orientation of the frame of reference and the natural orientation of the figure conflict, the figure’s orientation will be remembered as closer to that of the frame of reference” (Tversky, 1992, p. 136). Note the connection to perceptual processes: rotation “is similar to the Gestalt organizing principle of common fate” and alignment “is related to grouping by proximity” (Tversky, 1992, pp. 135-6). These heuristics have been demonstrated to apply to the spatial judgment and memory of large-scale spaces, such as continents, more local spaces such as streets, and both real and artificial spaces that are learned through graphical representations like maps (see Tversky, 1992 for a review). These alignment and rotation heuristics as well as other processes involved in spatial cognition lead to systematic errors in spatial judgment and memory (see Tversky, 1981, 1992 for reviews). For example, many incorrectly believe that San Diego is west of Reno, Nevada, due to the hierarchical grouping of cities within states and the subsequent alignment and rotation of the states. Tversky (2003) suggests a number of reasons that these errors exist. Cognitive processes schematize spatial information so that it can be represented and stored for future use. This schematization allows for integration of disparate knowledge from different sources and different perspectives as well as optimization of that knowledge so as to reduce the load on working memory. In the process of optimizing spatial knowledge, little metric information, e.g. distances, is ARCHITECTURE OF MIND AND WORLD 6 precisely retained or preserved in whole. My proposal that the form of an environment is intimately connected with the form of its mental representation suggests that the use of these heuristics depends upon the configuration of the environment. The identification of systematic errors in spatial judgment and memory under specific conditions can be used to indicate the involvement of the alignment and rotation heuristics. Space Syntax: Computational Analysis of Urban and Architectural Form As previously mentioned, space syntax provides us with a set of techniques for producing an abstract model of the configuration of a building interior or a part of an urban area. The spaces in question are formally described in terms of their topology – in other words, in terms of the spatial relationships among those spaces, such as connections and adjacencies. Research in space syntax proposes that these models can represent and allow for analysis of salient social and cognitive aspects of building interiors and urban areas (Bafna, 2003). (For more on the philosophical assumptions of space syntax approaches, see Hillier and Hanson, 1984, and Hillier, 1996.) This topological description takes the form of a graph indicating nodes and their interconnections. In one space syntax technique, axial map analysis, each of these nodes stands for a continuous line of sight in the environment. On the campus of Carleton College, one such sightline is the view down Winona Street running into Laird Hall (see the campus map in Figure 5). Such sightlines are commonly referred to as axes and a set of interconnected sightlines are called an axial map. As shown in the example of Figure 1, axes connect the various regions enclosed by obstacles, such as buildings that would block someone from seeing or walking through. ARCHITECTURE OF MIND AND WORLD 7 When creating an axial map, the researcher attempts to draw “the longest straight line” that will pass through the boundary between two adjacent regions of the environment in question (Bafna, 2003, p. 23). Thus the axes will connect with each other, forming a graph such as the one shown in Figure 2. These graphs retain the topology of the axes—that is, the connections among the axes—but discard all metric information such as the length and angle of the axes. “The underlying intuition is... based on the notions that first, the line of sight is a significant organizing and unifying device in experience and that second, the number of distinct turns on a route are more crucial to spatial experience than actual distance covered” (Bafna, 2003, p. 23). This “intuition” is in agreement with research in cognitive science. First, sightlines are a key component of J. J. Gibson’s ecological theory of perception (see especially Gibson, 1979, Chapter 11). His “vistas” are centered by a person as opposed to the environment – your vista changes as your location changes – but both vistas and axes emphasize that at a particular moment in time we are presented with a limited view of our environment. When we move, our view progressively changes to reveal new regions, at the same time as we leave behind our most recent view. Our visual experience of the world is defined by this serial sequence of limited views. Second, humans have been shown to be poor at estimating the metric distance of even the most familiar route or the angle formed by the best-known street intersections. Byrne (1979) has demonstrated that when people approximate distances, they rely on a heuristic that only considers the number of turns made on the route. When approximating the angle of street intersections, people are most likely to approximate the angle as nearly ARCHITECTURE OF MIND AND WORLD 8 orthogonal (at right angles) or at 45-degree angles, a finding that suggests they are using Tversky’s rotation heuristic. Once an axial map has been created, measures can be performed on the graph of the axes. One such measure is global integration, which quantifies how well individual nodes (axes) are interconnected with the graph as a whole (the map) by calculating the average number of steps required to reach each node from every other node in the graph. Since highly integrated nodes can be quickly reached from many other nodes in the graph, they are described as “shallow,” while less highly integrated nodes are “deeper” in the graph. In the example of Figure 1, Axis 3 has a high global integration value (8.30) while Axis 6 has a low global integration value (2.77). In Figure 2 it can be seen that Axis 3 is more highly integrated with the other nodes than Axis 6, which can only be accessed by way of Axis 7 and Axis 8. While global integration measures properties of the graph as a whole, local integration only considers the connections among nodes in the context of their immediate surroundings. Local integration is usually computed with radius 3 – that is, only nodes within a radius of 3 are used when computing the integration for a particular node. Local integration values are given below each global integration value in Figure 1. Integration measures have been found to accurately quantify various aspects of human behavior in the environment in question. Bafna (2003) cites studies in which the integration value of an axis in an urban setting was a significant predictor of the average number of pedestrians in that location—the higher the integration, the more pedestrians present. Haq and Zimring (2003) demonstrated such measures to be strong predictors of wayfinding behavior (where people walked) and abilities (how well they performed on ARCHITECTURE OF MIND AND WORLD 9 assessments of their knowledge) inside large buildings, such as hospitals. Kim and Penn (2004) found that the integration values of street maps sketched by people correlate with actual integration values of the streets. In particular, local integration in the sketch maps correlated highly (r = .728) with local integration in the actual map. In other words, participants’ spatial knowledge of the environment, as represented in their sketch maps, accurately represented the axial properties of the real world. The space syntax measure of integration is closely associated with the concept of legibility proposed by Lynch (1960): “the ease with which [the cityscape’s] parts can be recognized and can be organized into a coherent pattern. Just as this printed page, if it is legible, can be visually grasped as a related pattern of recognizable symbols, so a legible city would be one whose districts or landmarks or pathways are easily identifiable and are easily grouped into an over-all pattern” (p. 3). In the case of axial map analysis, such an overall pattern takes the form of a graph of axes. Integration measures how tightly each axis is knit into local groups (local integration) and the map as a whole (global integration). Connecting Spatial Cognition and Space Syntax Space syntax research suggests that places with certain configurational characteristics (i.e. higher integration) are used more frequently and by more people, are recalled more often by people, and are more accurately represented when recalled (Kim and Penn, 2004; Haq and Zimring, 2003). Spatial cognition research suggests that people perform better on spatial judgment and memory tasks for places that can be accurately schematized and integrated with other spatial knowledge (Tversky, 1981, 1992, 2003). I propose that there is an intimate connection between these features of the built world and ARCHITECTURE OF MIND AND WORLD 10 of spatial judgment and memory. Accordingly, space syntax measures, such as integration, should predict performance on spatial judgment and memory tasks. I will test this conjecture with a real-world case study using the campus of Carleton College and a set of undergraduate volunteers. I will perform an axial map analysis of the campus and ask participants to complete a set of spatial judgment and memory tasks. In particular, I hypothesize that: 1. Since the underlying form of spatial knowledge is presumably consistent across people, demographics (the sex and class year of participants) will not affect performance on spatial judgment and memory tasks. Previous research (Dara-Abrams, 2004) indicated no significant differences in performance on any of the tasks to be used in this study between underclass students (six or fewer trimesters on campus) and upperclass students (seven or more trimesters on campus). In addition, other research (German, Kail, and Siegel, 1979) has shown that undergraduate students become familiar with much of their campus within three months (approximately one trimester) if not within three weeks. However, using extreme groups will, it is assumed, conclusively demonstrate that experience is not a confounding factor. 2. Individual differences in spatial ability will affect performance on spatial judgment and memory tasks. In a pilot study (Dara-Abrams, 2004), a selfreport questionnaire, the Santa Barbara Sense of Direction Scale (Hegarty, Richardson, Montello, Lovelace, and Subbiah, 2002) was used. Significant correlations were found between scores on that instrument and performance on spatial judgment and memory tasks. However, the creators of that ARCHITECTURE OF MIND AND WORLD 11 instrument acknowledge it to be a better measure of ability relevant to realworld wayfinding than to spatial memory and judgment. This study will instead use the Mental Rotation Test (Vandenberg and Kuse, 1978), which has the advantage of being more closely associated with spatial memory and judgment tasks; performance on the test has been found to also measure wayfinding ability (Malinowski, 2001). 3. Systematic distortions, as described by Tversky (1981, 1992), will be observable in participants’ responses on spatial judgment and memory tasks. 4. Participants’ performance on one spatial judgment and memory task will agree with their performance on another task, as the tasks are designed to make use of the same spatial knowledge. 5. An axial map analysis of the case study environment will produce measures – in particular, measures of global and local integration – that can be compared with results from spatial judgment and memory tasks. 6. These space syntax measures will predict participants’ performance on spatial judgment and memory tasks. That is, the integration value associated with particular areas of the case study environment will predict participants’ performance when they are asked questions about those areas.