Cognition Meets Le Corbusier-Cognitive Principles of Architectural Design

Research on human spatial memory and navigational ability has recently shown the strong influence of reference systems in spatial memory on the ways spatial information is accessed in navigation and other spatially oriented tasks. One of the main findings can be characterized as a large cognitive cost, both in terms of speed and accuracy that occurs whenever the reference system used to encode spatial information in memory is not aligned with the reference system required by a particular task. In this paper, the role of aligned and misaligned reference systems is discussed in the context of the built environment and modern architecture. The role of architectural design on the perception and mental representation of space by humans is investigated. The navigability and usability of built space is systematically analysed in the light of cognitive theories of spatial and navigational abilities of humans. It is concluded that a building’s navigability and related wayfinding issues can benefit from architectural design that takes into account basic results of spatial cognition research. 1 Wayfinding and Architecture Life takes place in space and humans, like other organisms, have developed adaptive strategies to find their way around their environment. Tasks such as identifying a place or direction, retracing one’s path, or navigating a large-scale space, are essential elements to mobile organisms. Most of these spatial abilities have evolved in natural environments over a very long time, using properties present in nature as cues for spatial orientation and wayfinding. With the rise of complex social structure and culture, humans began to modify their natural environment to better fit their needs. The emergence of primitive dwellings mainly provided shelter, but at the same time allowed builders to create environments whose spatial structure “regulated” the chaotic natural environment. They did this by using basic measurements and geometric relations, such as straight lines, right angles, etc., as the basic elements of design (Le Corbusier, 1931, p. 69ff.) In modern society, most of our lives take place in similar regulated, human-made spatial environments, with paths, tracks, streets, and hallways as the main arteries of human locomotion. Architecture and landscape architecture embody the human effort to structure space in meaningful and useful ways. Architectural design of space has multiple functions. Architecture is designed to satisfy the different representational, functional, aesthetic, and emotional needs of organizations and the people who live or work in these structures. In this chapter, emphasis lies on a specific functional aspect of architectural design: human wayfinding. Many approaches to improving architecture focus on functional issues, like improved ecological design, the creation of improved workplaces, better climate control, lighting conditions, or social meeting areas. Similarly, when focusing on the mobility of humans, the ease of wayfinding within a building can be seen as an essential function of a building’s design (Arthur & Passini, 1992; Passini, 1984). When focusing on wayfinding issues in buildings, cities, and landscapes, the designed spatial environment can be seen as an important tool in achieving a particular goal, e.g., reaching a destination or finding an exit in case of emergency. This view, if taken to a literal extreme, is summarized by Le Corbusier’s (1931) notion of the building as a “machine,” mirroring in architecture the engineering ideals of efficiency and functionality found in airplanes and cars. In the narrow sense of wayfinding, a building thus can be considered of good design if it allows easy and error-free navigation. This view is also adopted by Passini (1984), who states that “although the architecture and the spatial configuration of a building generate the wayfinding problems people have to solve, they are also a wayfinding support system in that they contain the information necessary to solve the problem” (p. 110). Like other problems of engineering, the wayfinding problem in architecture should have one or more solutions that can be evaluated. This view of architecture can be contrasted with the alternative view of architecture as “built philosophy”. According to this latter view, architecture, like art, expresses ideas and cultural progress by shaping the spatial structure of the world – a view which gives consideration to the users as part of the philosophical approach but not necessarily from a usability perspective. Viewing wayfinding within the built environment as a “man-machine-interaction” problem makes clear that good architectural design with respect to navigability needs to take two factors into account. First, the human user comes equipped with particular sensory, perceptual, motoric, and cognitive abilities. Knowledge of these abilities and the limitations of an average user or special user populations thus is a prerequisite for good design. Second, structural, functional, financial, and other design considerations restrict the degrees of freedom architects have in designing usable spaces. In the following sections, we first focus on basic research on human spatial cognition. Even though not all of it is directly applicable to architectural design and wayfinding, it lays the foundation for more specific analyses in part 3 and 4. In part 3, the emphasis is on a specific research question that recently has attracted some attention: the role of environmental structure (e.g., building and street layout) for the selection of a spatial reference frame. In part 4, implications for architectural design are discussed by means of two real-world examples. 2 The human user in wayfinding 2.1 Navigational strategies Finding one’s way in the environment, reaching a destination, or remembering the location of relevant objects are some of the elementary tasks of human activity. Fortunately, human navigators are well equipped with an array of flexible navigational strategies, which usually enable them to master their spatial environment (Allen, 1999). In addition, human navigation can rely on tools that extend human sensory and mnemonic abilities. Most spatial or navigational strategies are so common that they do not occur to us when we perform them. Walking down a hallway we hardly realize that the optical and acoustical flows give us rich information about where we are headed and whether we will collide with other objects (Gibson, 1979). Our perception of other objects already includes physical and social models on how they will move and where they will be once we reach the point where paths might cross. Following a path can consist of following a particular visual texture (e.g., asphalt) or feeling a handrail in the dark by touch. At places where multiple continuing paths are possible, we might have learned to associate the scene with a particular action (e.g., turn left; Schölkopf & Mallot, 1995), or we might try to approximate a heading direction by choosing the path that most closely resembles this direction. When in doubt about our path we might ask another person or consult a map. As is evident from this brief (and not exhaustive) description, navigational strategies and activities are rich in diversity and adaptability (for an overview see Golledge, 1999; Werner, Krieg-Brückner, & Herrmann, 2000), some of which are aided by architectural design and signage (see Arthur & Passini, 1992; Passini, 1984). Despite the large number of different navigational strategies, people still experience problems finding their way or even feel lost momentarily. This feeling of being lost might reflect the lack of a key component of human wayfinding: knowledge about where one is located in an environment – with respect to one’s goal, one’s starting location, or with respect to the global environment one is in. As Lynch put it, “the terror of being lost comes from the necessity that a mobile organism be oriented in its surroundings” (1960, p. 125.) Some wayfinding strategies, like vector navigation, rely heavily on this information. Other strategies, e.g. piloting or path-following, which are based on purely local information can benefit from even vague locational knowledge as a redundant source of information to validate or question navigational decisions (see Werner et al., 2000, for examples.) Proficient signage in buildings, on the other hand, relies on a different strategy. It relieves a user from keeping track of his or her position in space by indicating the correct navigational choice whenever the choice becomes relevant. Keeping track of one’s position during navigation can be done quite easily if access to global landmarks, reference directions, or coordinates is possible. Unfortunately, the built environment often does not allow for simple navigational strategies based on these types of information. Instead, spatial information has to be integrated across multiple places, paths, turns, and extended periods of time (see Poucet, 1993, for an interesting model of how this can be achieved). In the next section we will describe an essential ingredient of this integration – the mental representation of spatial information in memory. 2.2 Alignment effects in spatial memory When observing tourists in an unfamiliar environment, one often notices people frantically turning maps to align the noticeable landmarks depicted in the map with the visible landmarks as seen from the viewpoint of the tourist. This type of behavior indicates a well-established cognitive principle (Levine, Jankovic, & Palij, 1982). Observers more easily comprehend and use information depicted in “You-are-here” (YAH) maps if the up-down direction of the map coincides with the front-back direction of the observer. In this situation, the natural preference of directional mapping of top to front and bottom to back is used, and left and right in the map stay left and right in the depicted world. While this alignment effect is based on the alignment between the map representation of the environment and the environment itself, alignments of other types of spatial representations have been the focus of considerable work in cognitive psychology. When viewing a path with multiple segments from one viewpoint, as shown in Figure 1, human observers have an easier time retrieving from memory the spatial relations between locations as seen from this viewpoint than from other, misaligned views or headings (Presson & Hazelrigg, 1984). In these types of studies, the orientation of the observer with respect to his or her orientation during the acquisition of spatial information, either imagined or real, seems to be the main factor. Questions like “Imagine you are standing at 4, looking at 3, where is 2?” are easier to answer correctly than “Imagine you are standing at 2, looking at 4, where is 3?”. These results have been taken as an indication of alignment effects between the orientation of an observer during learning and the imagined orientation during test. Fig. 1. Sample layout of objects in Presson & Hazelrigg (1984) study. The observer learns the locations of objects from position 1 and is later tested in different conditions. Later studies have linked the existence of alignment effects to the first view a person has of a spatial layout (Shelton & McNamara, 1997). If an observer learns the location of a number of objects from two different viewpoints he will be fastest and most correct in his response when imagining himself in the same heading as the first view. Imagined headings corresponding to the second view are no better than other, not experienced headings. According to the proposed theory, a person mentally represents the first view of a configuration and integrates new information from other viewpoints into this representation, leaving the original orientation intact. Similar to modern view-based theories of object recognition (Tarr, 1995), this theory proposes that spatial information should be easier accessible if the imagined or actual heading of a person coincides with this “remembered” viewing direction, producing an alignment effect. In the theories described above, the spatial relation between the observer and the spatial configuration determines the accessibility of spatial knowledge without any reference to the spatial structure of the environment itself. Indeed, most studies conducted in a laboratory environment try to minimize the potential effects of the external environment, for example by displaying a configuration of simple objects within a round space, lacking in any salient spatial structure. This is in stark contrast to the physical environments a person encounters in real life. Here, salient axes and landmarks are often abundant and are used to remember important spatial information. Recently, studies of human spatial memory have started to explore the potential effect of spatial structure on human spatial memory and human navigation (Werner, Saade, & Lüer, 1998; Werner & Schmidt, 1999). If an observer has to learn a configuration of eight objects within a square room, for example, she will have a much easier time retrieving the spatial knowledge about the configuration when imagining herself aligned with the room’s two main axes parallel to the walls than when imagining herself aligned with the two diagonals of the room. This holds true even when all potential heading directions within the room have been experienced by the observer (Werner, Saade, & Lüer, 1998). Similarly, people seem to be sensitive to the spatial structure of the large-scale environment they live in. When asked to point in the direction of important landmarks of the city they live in, participants have a much easier time imagining themselves aligned with the street grid than misaligned with the street grid (Werner & Schmidt, 1999; see also Montello, 1991). In this case, the environment has been learned over a long period of time and from a large number of different viewpoints. Additional research strongly suggests that the perceived structure of an environment influences the way a space is mentally represented even in cases where the acquisition phase is well-controlled and the observer is limited to only a few views of the space (Shelton & McNamara, 2001; McNamara, Rump, & Werner, in press). In sum, the perceived spatial structure of an environment seems to play a crucial role in how spatial information is remembered and how easy it is to retrieve. In the following section we will review which features of the environment might serve as the building blocks of perceived spatial structure. 3 The perceived structure of the environment Natural and man-made environments offer a large number of features that can influence the perception of “environmental structure.” Visual features, such as textures, edges, contours, can serve as the basis for structure as can other modalities, such as sound or smell. Depending on the scale of the environment, the sensory equipment of the user, and the general navigational goal, environments might be perceived very differently. However, in many cases a consensus seems to exist among observers as to the general structure of natural environments. Following are a few examples. When navigating in the mountains, rivers, valleys, and mountain ranges constitute the dominant physical feature that naturally restrict movement and determine what can be perceived in certain directions. Paths within this type of terrain will usually follow the natural shape of the environment. Directional information will often be given in environmental terms, for example “leaving or entering a valley,” “crossing a mountain range,” or “uphill” and “downhill” (see Pederson, 1993), reflecting the importance of these physical features. A recent study confirmed that observers use environmental slant not only to communicate spatial relations verbally, but also to structure their spatial memories (Werner, 2001; Werner, Schmidt, & Jainek, in prep.). In this study, participants had to learn the location of eight objects on a steep hill. Their spatial knowledge of the environment was later tested in the laboratory. Accessing spatial knowledge about this sloped environment was fastest and most accurate when imagining oneself facing uphill or downhill, thus aligning oneself with the steepest gradient of the space. In many instances, natural boundaries defined through changes in texture or color give rise to the perception of a shaped environment. Looking at a small island from the top of a mountain lets one clearly see the coastal outline of the land. Changes in vegetation similarly present natural boundaries between different regions. Both, humans and other animals seem to be sensitive to the geometrical shape of their environment. Rats, for example, rely heavily on geometrical structure when trying to retrieve food in an ambiguous situation (Cheng & Gallistel, 1984; Gallistel, 1990). Young children and other primates also seem to favor basic geometrical properties of an environment when trying to locate a hidden toy or buried food (Hermer & Spelke, 1994; Gouteux, Thinus-Blanc, & Vauclair, 2001). The importance of geometric relations might be due to the stability of this information over time, compared to other visual features whose appearance can change dramatically throughout the seasons (bloom, changing and falling of leaves, snow cover; see Hermer & Spelke, 1996). Different species have developed many highly specialized strategies to structure their environment consistently. For migrating birds, local features of the environment are as important as geo-magnetic and celestial reference points. Pigeons often rely on acoustical or olfactory gradients to find their home (Wiltschkow & Wiltschkow, 1999). The desert ant Cataglyphis uses a compass of polarized sunlight to sense an absolute reference direction in its environment (Wehner, Michel, & Antonsen, 1996). Similarly, humans can use statistically stable sources of information to create structure. When navigating in the desert, the wind direction or position of celestial bodies at night might be the main reference, whereas currents might signal a reference direction to the polynesian navigator (see Lynch, 1960, pp. 123ff, for anecdotal references). In the built environment, structure is achieved in different ways. At the level of the city, main streets and paths give a clear sense of direction and determine the ease with which spatial relations between different places or regions can be understood (Lynch, 1960). In his analysis of the “image of the city,” Lynch points out the difficulty to relate different parts of Boston because the main paths do not follow straight lines and are not parallel. The case of Boston also nicely illustrates the interplay between the built and natural environment. In Boston, the main paths for traffic run parallel to the Charles river – resulting in an alignment of built and natural environment. As mentioned above, the perceived structure of the city plays a large role in how accessible spatial knowledge is for different imagined or real headings within the space (Werner & Schmidt, 1999). At a smaller scale, individual buildings or structures impose their own structure. As Le Corbusier notes, “architecture is based on axes” which need to be arranged and made salient by the architect (p. 187). Through these axes, defined by walls, corridors, lighting, and the arrangement of other architectural design elements, the architect communicates a spatial structure to the users of a building. Good architectural design thus enables the observer to extract relevant spatial information. This feature has been termed architectural legibility and is the key concept in research on wayfinding within the built environment (Passini, 1984, p. 110). In the last section we will focus on the issue of architectural legibility and how the design of a floor plan can aide or disrupt successful wayfinding. 4 Designing for Navigation 4.1 Architectural legibility and floor plan complexity Research linking architectural design and ease of navigation has mainly focused on two separate dimensions: the complexity of the architectural space, especially the floor plan layout, and the use of signage and other differentiation of places within a building as navigational aids. As many different research projects have shown both from an architectural and environmental psychology point of view, the complexity of the floor plan has a significant influence on the ease with which users can navigate within a building (O’Neill, 1991, Weisman, 1981, Passini, 1984). The concept of complexity, however, is only vaguely defined and comprises a number of different components. Most often, users’ ratings of the figural complexity of a floor plan, often interpreted as a geometric entity, has been used to quantify floor plan complexity for later use in regression models to predict navigability. Different authors have mentioned different underlying factors that influence an observer’s judgment of complexity; most notably, the symmetry of a plan and the number of possible connections between different parts of the figure. An attempt to quantify the complexity of a floor plan analytically, by computing the mean number of potential paths from any decision point within the floor plan, was devised by O’Neill (1991). Fig. 2. Different schematic floor plans and their ICD index after O’Neill (1991). Five basic floor plan layouts used in his study are shown in Figure 2 and the corresponding inter-connection density index (ICD) is listed underneath each plan. The basic idea in this approach consists of an increase in floor plan complexity with increasing number of navigational options or different paths. The correlation of the ICD measure and empirical ratings of complexity for the plans used in his study were fairly high. One theoretical problem with this index, however, is demonstrated in Figure 3. Here 4 different figures depict three different floor plans with exactly the same ICD index. Their perceived complexity, however, rises from left to right, by making the figures less symmetric, changing the orientation, or making the figure less regular. Fig. 3. Four different floor plans with identical ICD but different perceived complexity. A serious problem with all approaches using figural complexity as a measure, is to treat the geometrical complexity of a floor plan as indicative of the navigational complexity of the spatial environment depicted by the plan. As Le Corbusier pointed out almost 80 years ago, the easily perceivable and pleasant geometrical two-dimensional depiction of a spatial environment can differ dramatically from the perceived structure of a spatial environment (1931, p. 187). In it, space is experienced piecemeal, from multiple different viewpoints, in which only small portions of the space are visible at one time, and in which spatial relations have to be inferred by integrating spatial knowledge across multiple viewpoints and over long periods of time. The basic city layout of Karlsruhe, for example, includes as its main design characteristic a radial (star) arrangement of streets emanating from the castle in the center of the environment. While providing a very salient structure when looking at the city map, the global structure is hidden from each individual view. What is perceived is often a single, isolated street. In a similar fashion, when judging the complexity of the two fictitious floor plans at the top of Figure 4, The left floor plan might be judged as less complex than the right floor plan. This is due to the meaningfulness of the left geometrical figure. If a person has to navigate this floor plan without prior knowledge of this structure, however, the meaningfulness will not be apparent, and the two floor plans will be perceived as similar in their navigational complexity (see the two views from viewpoints within the two floor plans in the lower half of Figure 4). These examples strongly suggest that the two-dimensional, figural complexity of a depiction of a floor plan should not uncritically be taken as a valid representation of the navigational complexity of the represented spatial environment. Fig. 4. Top: Two similar floor plans with different perceived complexity; Below: Views from similar viewpoints within the two floor plans (viewpoints and viewing angles indicated above). 4.2 Global and local reference frames in perceiving spatial layout When viewing a visual figure, such as a depiction of a floor plan, on a piece of paper or a monitor, the figure can usually be seen in its entirety. This allows an observer of the floor plan to see the spatial relations between different parts of the plan, which cannot be perceived simultaneously in the real environment. One of the first steps in the interpretation of the visual form consists of the assignment of a common frame of reference to relate different parts of the figure to the whole (Rock, 1979). There are multiple, sometimes competing solutions to the problem of which reference frame to assign to a figure. For example, the axis of symmetry might provide a strong basis to select and anchor a reference frame in some symmetric figures, whereas the viewpoint of the observer might be chosen for a less symmetric figure. In general, the distinction between intrinsic and extrinsic reference frames has proven useful to distinguish two different classes of reference systems. Fig. 5. Determining the “top” of a geometrical figure. Figures A & B exemplify the role of intrinsic reference systems and C & D the role of extrinsic reference systems. The perceived orientation of each figure is marked with a black circle. See text for details. Intrinsic reference systems. An intrinsic reference system is based on a salient feature of the figure itself. In Figure 5 a number of examples illustrate this point. The axis of symmetry in a isosceles triangle determines the perceived direction the triangle is pointing at (example A). It also determines how spatial information within the triangle and surrounding space is organized (e.g., left half and right half, see Schmidt & Werner, 2000). Example B shows a situation in which the meaning of the object determines a system of reference directions (e.g., above and below the chair, see Carlson, 1999). An isolated experience of a particular part of a building will most likely result in the dominance of the intrinsic reference system of the particular space. Extrinisc reference system. Besides intrinsic features of a figure, the spatial and visual context of a figure can also serve as the source for a reference system. In example C, the equilateral triangle is seen as pointing towards the right because the rectangular frame around it strongly suggests an orthogonal reference system and only one of the three axes of symmetry of the triangle is parallel to these axes. Similarly, example D shows how the perceived vertical in the visual field or the borders of the page are used to select the reference direction up-down as the most salient axis within the rightmost equilateral triangle. When viewing a floorplan, all the parts of the building can be viewed in unison and the plan itself can be used as a consistent extrinsic reference system for all the parts. Based on the distinction between extrinsic and intrinsic reference systems we can now re-examine one of the main differences between a small-scale figural depiction of a floor plan and the large-scale space for navigation which is depicted by it. In the case of the small figure, each part of the figure is perceived within the same, common reference system. This reference system can be based on an extrinsic reference system (e.g., the page the plan is drawn on), or a global intrinsic reference system of the plan (e.g., the axis of symmetry of the plan). The common reference system then determines how each part of the plan is perceived. 4.3 Misalignment of local reference systems as a wayfinding problem: Two examples In section 2 we discussed navigational strategies and how misalignment with the perceived structure of an environment increases the difficulty for a navigator to keep track of the spatial relations between parts of the environment or objects therein. This concept of misalignment with salient axes of an environment fits very well with the concept of a reference system as discussed above. If an environment’s structure is defined by a salient axis, this axis will serve as a reference direction in spatial memory. The reference system used to express spatial relations within this environment will most likely be fixed with respect to this reference direction (see Shelton & McNamara, 2001; Werner & Schmidt, 1999). As discussed in section 2.2, the task of keeping track of one’s location in the built environment often requires the integration of spatial information across multiple places. An efficient way to integrate spatial information consists of the expression of spatial relations within the same reference system (Poucet, 1993). A common reference system enables a navigator to relate spatial information that was acquired separately (e.g., by travelling along a number of path segments). Architectural design can aide this integration process by assuring that the perceived spatial structure in each location of a building suggests the same spatial reference system and is thus consistent with a global structure or frame of reference. This does not imply, however, that buildings have to be organized around a simple orthogonal grid with only right angles. Other, more irregular designs are unproblematic as long as the architect can achieve a common reference system by making common axes salient. The following two examples are illustrating the effects of a common reference system and alignment effects at the scale of an individual building (example 1) and the layout of a city (example 2). Example 1: The town hall in Göttingen, Germany. Figure 6 depicts a schematic floor plan of the town hall of Göttingen, Germany. Informal interviews with people working in or visiting this building revealed that it is difficult to understand and navigate. The architectural legibility is very low. With respect to the aim of this paper, we will mainly focus on the layout of the floor plan in order to discern how it might impact people’s ability to find their way around in the building. When looking at the floor plan, the building appears to consist of three separate areas. To the left and the right, two large areas stand out. They are almost mirror images of each other and slightly offset against each other. At the top of the floor plan, centered horizontally between these two areas is a smaller, third area which includes the main elevator vertically connecting the floors. This area appears to have a diamond shape in the floor plan. To the left, bottom, and right, this area is connected with the hallways serving the other two main areas. The overall shape of the building appears to consist of two offset octagons merged touching on one side with the diamond shaped elevator area connecting them. Fig. 6. Floor plan of the city hall of Göttingen, Germany (hallways are depicted in white). The area around the elevator at the top is rotated 45o with respect to the rest of the building. The naïve description of the visual appearance of the floor plan listed above nicely illustrates the point made above in the context of Figure 4. Especially the description of the elevator area as a “diamond shaped area” needs to be re-evaluated. Unlike a viewer of the floor plan, a user of the physical space will not perceive the area around the elevator as a diamond. Instead, the area will be perceived as a square, thus choosing a different reference system as in the description above. Figure 7 summarizes this situation. Not knowing the global reference system that was used in describing the floor plan, a user upon entering the space will find four hallways surrounding the elevator connected at right angles, leading to the perception of a square. Fig.7. Schematic display of the spatial situation in the town hall. When viewing image A, the center figure will be labelled diamond. In B, the relation between the figure inside and the outer figure is unknown to the observer and the smaller figure will be seen as a square. As is evident from this analysis, an important part of the navigational difficulties in this environment stem from two conflicting spatial reference systems when perceiving different parts of the environment. This misalignment between the parts makes integration of spatial knowledge very difficult. Example 2: Downtown Minneapolis. The second example deals with a city-scale environment. Figure 8 shows two maps of different parts of downtown Minneapolis. Due to its vicinity to the Mississippi river, the street grid of downtown Minneapolis does not follow the North-South, East-West orientation of the streets and main traffic arteries found in the surrounding areas. As can be seen in the left map of the warehouse-district, the streets run south-west to north-east or orthogonal to this direction. The map to the right gives an overview of the street grid found downtown and how it connects into the surrounding street pattern (e.g., the streets to the south of downtown). Fig. 8. Maps of downtown Minneapolis. Left: A blown-up map of the Warehouse district. North is up. Note the lack of horizontal and vertical lines. Right: A larger scale depicting all of downtown. In this map, the main street grid consists of vertical and horizontal lines. North is rotated approximately 40 ̊ counterclockwise. It is interesting to note that the map designers for the two maps chose different strategies to convey the spatial layout of the depicted area. On the left, a North-up orientation of the map was chosen, which has the effect that all the depicted streets and buildings are misaligned with the vertical and horizontal. On the right, the map designer chose to align the street grid with the perceived horizontal and vertical on the page, in effect rotating the North orientation by approximately 40 ̊ counterclockwise. In a small experiment we tested these types of map arrangements against each other and found that observers had an easier time interpreting and using spatial information gathered from a map in which the depicted information was aligned with the visual vertical and horizontal, whereas a misalignment with these axes led to more errors in judgements about spatial relations made from memory (Werner & Jaeger, 2002). It seems evident, from these results and from the theoretical analysis presented in the context of the town hall, that the information in the map should be presented in the same orientation as it is perceived in the real environment, namely as an orthogonal street grid running up-down, and left-right. The map example on the right also points towards another problem discussed above. When displaying spatial information only about downtown Minneapolis, a rotation of the grid into an upright orientation on the map makes a lot of sense from a usability point of view. However, when this information has to be integrated with spatial information about areas outside the downtown area, the incompatibility of the two reference systems becomes a problem. If information about downtown and the surrounding areas has to be depicted in the same map, only one alignment can be selected (which usually follows the North-up orientation which aligns the streets outside of downtown with the main visual axes). 4.4 Design recommendations for wayfinding As the examples and the discussion of empirical results show, misalignment of reference systems impairs the users ability to integrate spatial information across multiple places. There are a number of design considerations that can be derived from this finding. When designing a building in which wayfinding issues might be relevant, the consistent alignment of reference axes throughout the building, all other things being equal, will greatly reduce the cognitive load while keeping track of once position. The architectural structure as perceived from different locations thus has direct implications for the navigability of the building and determines the buildings overall legibility. Providing navigators access to a global frame of reference within a building will greatly support wayfinding tasks. This can be achieved by providing visual access to distant landmarks or a common link, such as a courtyard or atrium. If the preexisting architectural environment does not allow for a consistent spatial frame of reference, as in the case of downtown Minneapolis, the navigational demands on the user should take this into consideration. If integration across different reference systems is not required, the problem of misaligned reference systems becomes a moot point. In the case of Minneapolis, for example, the activities in downtown are mainly confined to the regular street grid. Only when leaving the downtown area and trying to connect to the outside street system does the misaligned reference system become an issue. In this case, allowing for simple transitions between the two systems is essential.