Spatial knowledge acquisition from direct experience in the environment: Individual differences in the development of metric knowledge and the integration of separately learned places

Abstract

Existing frameworks for explaining spatial knowledge acquisition in a new environment propose either stage-like or continuous development. To examine the spatial microgenesis of individuals, a longitudinal study was conducted. Twenty-four college students were individually driven along two routes in a previously unfamiliar neighborhood over 10 weekly sessions. Starting Session 4, they were also driven along a short connecting route. After each session, participants estimated spatial properties of the routes. Some participants’ knowledge improved fairly continuously over the sessions, but most participants either manifested accurate metric knowledge from the first session or never manifested accurate metric knowledge. Results are discussed in light of these large individual differences, particularly with respect to the accuracy and development of integrated configurational knowledge.

Introduction

As people act in the environment, they perceive surrounding space and acquire knowledge about it. Downs and Stea (1973) called this fundamental process cognitive mapping. Knowledge acquired during cognitive mapping includes the identities of places and landmarks, the patterns of path connections between places, distances, and directions between places, and so on. People use spatial knowledge of the environment to get to destinations such as home and work, to give and interpret navigational instructions, to interpret maps, to plan efficient trips, and more. These kinds of knowledge help guide people’s actions in adaptive ways, in other words, so that their behavior is coordinated not only to the environment as perceived but also to the environment as conceived and remembered. Many researchers from various disciplines have investigated human spatial knowledge of environments, particularly since the late 1960s (see Cox and Golledge, 1969, Evans, 1980, Liben et al., 1981, Lynch, 1960).

A major research question that has attracted much theoretical interest concerns the structure of spatial knowledge about environments and the process of spatial knowledge acquisition in new environments. As Ittelson (1973) discussed, the environment is larger than and surrounds the human body, so that a person cannot grasp the layout of the environment in its entirety from a single viewpoint. Instead, one must locomote about the environment, integrating knowledge acquired from separate viewpoints and travel experiences. Due to this unique characteristic, learning the spatial layout of environments from direct experience is not a straightforward task, especially for people with a poor “sense-of-direction” (e.g., Kozlowski & Bryant, 1977). In fact, researchers have shown that people’s knowledge about environments tends to be distorted, fragmented, and schematized (e.g., Golledge and Spector, 1978, Lynch, 1960, Stevens and Coupe, 1978).

Siegel and White (1975) proposed a theoretical framework for describing and explaining the process of knowledge development over time in new environments (called spatial cognitive microgenesis). In their framework, internal representations of spatial knowledge of a new place progress over time from an initial stage of landmark knowledge to a stage of route knowledge to an ultimate stage of survey knowledge. Landmark knowledge is knowledge about the identities of discrete objects or scenes that are salient and recognizable in the environment. Route knowledge consists of sequences of landmarks and associated decisions (e.g., “turn left at the gas station and go straight for three blocks”). According to Siegel and White, the space between landmarks is at first “empty” and receives “scaling” with accumulated experience (p. 29); in other words, route knowledge is initially nonmetric. The final and most sophisticated stage of knowledge in their framework is survey knowledge. This is a two-dimensional and “map-like,” quantitatively scaled representation of the layout of the environment. Survey knowledge represents distance and directional relationships among landmarks, including those between which direct travel has never occurred. For survey maps to emerge, routes need to be metrically scaled and interrelated into a global allocentric reference system.

Siegel and White’s framework was so influential in the scientific literature (including psychology, geography, and robotics) that Montello (1998) called it the dominant framework. This was certainly true for the time period up until the early 1990s, about which Montello was writing (see, e.g., Chase and Chi, 1981, Clayton and Woodyard, 1981, Golledge et al., 1985, Hirtle and Hudson, 1991, Kuipers and Levitt, 1988, Lloyd, 1989, McDermott and Davis, 1984, Moeser, 1988, Thorndyke and Hayes-Roth, 1982). Even though writers no longer cite Siegel and White’s framework so commonly, for the most part they have not replaced it with another coherent theoretical framework, so we continue to refer to Siegel and White’s proposal as the dominant framework. And even though Siegel and White may not be cited by name so often anymore, their ideas are still influential. For instance, many researchers continue to refer to the construct of route knowledge with terms such as procedural knowledge (Golledge, 1991) or topological knowledge (Allen, Kirasic, Dobson, Long, & Beck, 1996), terms that emphasize the sequential and nonmetric nature of the knowledge. There are computational process models of spatial knowledge that are based on discrete landmarks and sequential routes as conceived by the dominant framework (Yeap & Jefferies, 2000, described such models as “object-based” in contrast to “space-based” approaches). Aguirre and D’Esposito (1997) provide a recent example from cognitive neuroscience, a field in its childhood when Siegel and White first published their framework.

A comment is in order about landmark recognition. People recognize a landmark for its salience in the surroundings in terms of size, shape, or color, and for its functionality as a navigation clue. So at least approximate metric information may exist at or in the vicinity of landmarks that allows people to attend to the size and shape of objects. But the important property of landmark and route knowledge as conceived by the dominant framework is that individual landmarks are encoded as individual entities, and related to other landmarks in terms of only connectivity and order. This “skeletal” nature of people’s spatial knowledge after early exposure to environments is similarly reflected in Golledge’s (1978) conceptual framework, called the anchor point theory.

Its popularity aside, however, Siegel and White’s (1975) dominant framework has not received convincing empirical support as a model of the structure and microgenetic course of spatial knowledge. Particularly troubling, as pointed out by Montello (1998), are the ideas that metric knowledge takes so long to begin developing, and that landmark knowledge is a necessary prerequisite for route knowledge, which in turn is a necessary prerequisite for survey knowledge. Research has shown, on the contrary, that with minimal exposure to a new environment (on the order of seconds or minutes), people can perform tasks that require some metric configurational knowledge² at a level at least better than chance—tasks such as taking shortcuts, returning directly back to starting locations, and estimating distances and directions directly between places (e.g., Klatzky et al., 1990, Landau et al., 1984, Loomis et al., 1993). Such tasks require that people understand quantitative distances and directions between places (perhaps somewhat vaguely), which the dominant framework claims should not be possible to do by people without accumulated experience.

Besides being apparently contradicted by empirical evidence, the dominant framework provides only a general description of the developmental course of spatial microgenesis (hence we call it a framework rather than a theory). It does not tell us how much time or effort is necessary to proceed from one stage to another. Blades (1991) made a similar criticism of the descriptive, nonexplicit nature of the dominant framework.

In response to these problems with the dominant framework, Montello (1998) proposed an alternative framework for explaining people’s spatial microgenesis. This framework posits continuous (or quantitative) development of metric knowledge, rather than discrete (or qualitative) development, as the dominant framework posits. This echoed suggestions by writers such as Evans, 1980, Hirtle and Hudson, 1991, Thorndyke, 1981. The idea that spatial knowledge, including metric knowledge, is acquired in new environments relatively continuously is at the core of Montello’s framework, so we refer to it as the continuous framework. Other aspects of his framework are presented below.

Most of the empirical studies of the microgenetic development of spatial knowledge have used hand-drawn sketch maps as a measure of knowledge (e.g., Appleyard, 1970, Beck and Wood, 1976, Devlin, 1976) or have used cross-sectional designs that implicitly equate length of residence with amount of exposure to the environment (e.g., Herman et al., 1979, Stern and Leiser, 1988, Thorndyke and Hayes-Roth, 1982). These studies have failed to yield consistent results about the structure and development of spatial knowledge, due mainly to (a) differences in the structure and complexity of the study area, (b) difficulties in objectively classifying the style and degree of sophistication of sketch maps, and (c) lack of control over respondents’ amount of experience and sources of knowledge (such as direct travel or maps).

Only a few studies have experimentally examined changes in the accuracy of people’s spatial knowledge over time. Gärling, Böök, Lindberg, and Nilsson (1981) and Herman, Blomquist, and Klein (1987) had participants travel a route, and asked them to do spatial tasks such as distance and direction estimation. Participants repeated this set of route travel and estimation tasks a couple of times. Adult participants in these studies acquired knowledge of landmarks immediately, and their direction and distance estimates got better during early sessions and then leveled off, never attaining perfect accuracy.

Overall, the results from these empirical studies, which used a variety of different methods, seem to indicate that route knowledge and at least an approximate knowledge of distances and directions are acquired early on in a new environment. However, the results are not consistent enough to allow one to draw a solid conclusion about spatial microgenesis. Missing are studies that examine microgenetic development over an extended period of time with appropriate experimental control.

Both the dominant framework (Siegel & White, 1975) and Montello’s (1998) continuous framework argue that the acquisition of coordinated and comprehensive survey knowledge is an important process in microgenetic development. For survey maps to emerge, routes need to be metrically scaled and interrelated into a global allocentric reference system. In other words, places and routes learned during separate travel experiences are integrated³ and interrelated with each other in a common frame of reference. This is a fairly sophisticated step in spatial microgenesis. For example, researchers (e.g., Lynch, 1960, Rand, 1969) have often observed that people who are familiar with two separate regions may not understand the spatial relationship between them.

Some researchers have examined how accurately people can integrate separately learned places into a common frame of reference (e.g., Hanley and Levine, 1983, Holding and Holding, 1988, Golledge et al., 1993, Moar and Carleton, 1982, Montello and Pick, 1993). These researchers presented two separate routes to participants, and then assessed their spatial knowledge about each route (within-route tasks) and about the spatial relation between the two routes (between-route, or integration, tasks).

The results of these studies on integration are somewhat inconsistent. Participants in some studies (Holding and Holding, 1988, Moar and Carleton, 1982) successfully integrated two separate routes, showing evidence of spatial knowledge that was just as accurate for the integrated routes as for the separate routes. Participants in other studies (Golledge et al., 1993, Hanley and Levine, 1983, Montello and Pick, 1993) failed to integrate the separate routes, or were less accurate for the integrated than for the separate routes. This inconsistency may have been due to differences in (a) the source of information (direct navigation vs. slides), (b) the spatial scale (environmental routes vs. tabletop paths), (c) the complexity of the routes, or (d) the information given to participants that allowed integration of the two routes (the two routes shared a common segment vs. the two routes were completely separate and information about the relationship between the two was given later as verbal descriptions or direct experience).

Montello’s (1998) continuous framework states that individuals will vary in the extent and accuracy of the spatial knowledge that they acquire from direct experience. He posited that this variation, especially in the ability to integrate separately learned places, was so large that it deserved special note because it could confound general statements about the form of microgenesis. For example, individual differences could confound questions of how much metric information is acquired early, how precise is the metric knowledge, whether integration occurs, and if so, when. Such individual differences are ignored (relegated to the error term) in an aggregate analysis. There are studies of individual differences in spatial cognition, for example, with respect to underlying spatial abilities (Allen et al., 1996), sense-of-direction (Hegarty, Richardson, Montello, Lovelace, & Subbiah, 2002), and neural correlates of spatial thinking (Hartley, Maguire, Spiers, & Burgess, 2003). The study we report here specifically addresses in detail individual differences in the developmental pattern of people’s spatial knowledge, which is quite unusual in the literature on spatial microgenesis.

Except for the few attempts discussed above, microgenetic development of spatial knowledge has not been investigated much in controlled longitudinal studies. As we discussed above, the dominant framework, explicitly proposed 30 years ago, has strongly influenced spatial cognition research in several disciplines; even if not cited by name, its concepts of nonspatial landmark knowledge and, especially, nonmetric and sequential route knowledge still echo through the writings of various cognitive-science fields (as we cited above). This is true even though it has never received clear empirical support. Likewise, critiques of the dominant framework and the ideas suggested as alternatives have not had much empirical research specifically directed at them. In our research, therefore, we aim to investigate how people’s spatial knowledge of the environment, particularly the accuracy and precision of their knowledge, develops over time.

We are specifically motivated by three major research questions. Our first question concerns the development of metric knowledge (distances and directions) in a new environment learned directly via locomotion. In particular, we look at the level of people’s configurational understanding of routes (i.e., beyond the sequential, nonmetric route knowledge of the dominant framework’s terminology). Our second question concerns people’s ability to integrate separately learned places into a common frame of reference (i.e., the acquisition of integrated survey knowledge). Related to this question, we also look at whether people acquire spatial knowledge differently when they travel a route in two opposite directions, compared to when they travel a route in only one direction. Our third major question concerns individual differences in people’s spatial knowledge; specifically, how different individuals’ developmental curves look with respect to accuracy and a developmental pattern, and whether the difference is especially large for integration.

To investigate these questions, we conducted a longitudinal experiment in a naturalistic setting, with as much experimental control as possible concerning the amount and method of exposure to the environment. We examined participants’ performance on tasks that assessed their spatial knowledge, both metric and nonmetric, once a week for 10 consecutive weeks.

If spatial microgenesis progresses through qualitatively distinct stages as conceived by the dominant framework, participants’ performance on metric tasks should be at or near chance level in early sessions. Only after repeated sessions, participants’ performance should become better than chance and begin to improve. On the other hand, if spatial microgenesis is continuous, as the continuous framework argues, participants’ performance should be at least better than chance from the first session and gradually improve. Both the dominant and continuous frameworks agree that the acquisition of integrated configurational knowledge (or survey knowledge) is a large, sophisticated step in microgenetic development. If so, people’s performance should be poor on between-route, as opposed to within-route, tasks, at least in early sessions. Also, their performance should be affected by the complexity of routes. Fig. 1 schematically illustrates the predicted patterns of development by the two frameworks.

As well as comparing participants’ performance to chance performance, which assumed complete lack of metric knowledge, we considered possible performance by “hypothetical” participants that we gave varying degrees of accuracy and precision of metric knowledge. We examined the performance of these hypothetical participants by conducting Monte Carlo simulations (Table 1). First, we simulated participants who would know only the identities of the landmarks. These participants can be thought of as being in the stage of landmark knowledge as conceived by the Siegel and White framework. When asked to draw a sketch map, such participants would simply locate points randomly on paper. Other such participants might locate points based on some heuristic rule, such as aligning the points in a straight line, or locating landmarks from one route on the right side of paper and landmarks from the other route on the left side. Researchers have reported that people tend to use these kinds of simplifying heuristics to judge or recall angles of turns (Byrne, 1979, Lynch, 1960, Moar and Bower, 1983).

Second, we simulated participants who would know the identities and the sequence of landmarks. These participants can be thought of as being in the stage of route knowledge as conceived by the Siegel and White framework. When asked to estimate the direction from one landmark to another, they would randomly choose a forward or backward direction, depending on whether the target landmark came after or before the current landmark in the sequence. Montello and Frank (1996) conducted similar simulations of qualitative direction judgments.

Third, we simulated participants who would know the identities and sequence of landmarks, but also possess some degree of quantitative knowledge. These participants can be thought of as being in transition from the stage of route knowledge to that of survey knowledge as conceived by the Siegel and White framework. Such participants would notice major changes in heading during travel and encode them as right-angled turns to the right or left. Other such participants might roughly conceive of the spatial relation between the two routes as one being to the right or left of the other. Other participants might possess minimal distance knowledge, in terms of two categories of relatively short or long, or in a little more detail, in terms of a rank order. These kinds of qualitative spatial reasoning have been discussed in the literature, particularly by computer scientists (Forbus et al., 1991, Frank, 1996).