Elsevier

Cognition

Volume 77, Issue 3, 15 December 2000, Pages 215-250
Cognition

Updating egocentric representations in human navigation

https://doi.org/10.1016/S0010-0277(00)00105-0Get rights and content

Abstract

Seven experiments tested whether human navigation depends on enduring representations, or on momentary egocentric representations that are updated as one moves. Human subjects pointed to unseen targets, either while remaining oriented or after they had been disoriented by self-rotation. Disorientation reduced not only the absolute accuracy of pointing to all objects (‘heading error’) but also the relative accuracy of pointing to different objects (‘configuration error’). A single light providing a directional cue reduced both heading and configuration errors if it was present throughout the experiment. If the light was present during learning and test but absent during the disorientation procedure, however, subjects showed low heading errors (indicating that they reoriented by the light) but high configuration errors (indicating that they failed to retrieve an accurate cognitive map of their surroundings). These findings provide evidence that object locations are represented egocentrically. Nevertheless, disorientation had little effect on the coherence of pointing to different room corners, suggesting both (a) that the disorientation effect on representations of object locations is not due to the experimental paradigm and (b) that room geometry is captured by an enduring representation. These findings cast doubt on the view that accurate navigation depends primarily on an enduring, observer-free cognitive map, for humans construct such a representation of extended surfaces but not of objects. Like insects, humans represent the egocentric distances and directions of objects and continuously update these representations as they move. The principal evolutionary advance in animal navigation may concern the number of unseen targets whose egocentric directions and distances can be represented and updated simultaneously, rather than a qualitative shift in navigation toward reliance on an allocentric map.

Introduction

How do people and animals represent the spatial properties of their environment so as to locate objects and navigate effectively to significant places? Research on insects suggests that one form of navigation – homing – can depend on a continuous updating process over self-motion, i.e. path integration (Srinivasan et al., 1996, Wehner and Srinivasan, 1981; see also Collett, 1996, Dyer, 1996). Rodents also have a path integration system that allows them to move to and from significant locations such as the nest and the site of an enduring food source (Etienne et al., 1996, Mittelstaedt and Mittelstaedt, 1980; for review see Gallistel, 1990). Unlike many ants and bees, however, rodents also are able to navigate to familiar objects along novel paths from novel, arbitrary points, suggesting that their spatial learning involves the construction of a qualitatively different type of spatial representation: an enduring, observer-free ‘cognitive map’ of the environment (e.g. O'Keefe and Nadel, 1978, Sutherland and Dyck, 1984, Tolman, 1948; for discussion see Bennett, 1996, Gallistel, 1990). Evidence for such cognitive maps gains much intuitive appeal from studies of humans. Although humans seem to have a path integration mechanism that resembles that of insects and rodents (e.g. Berthoz et al., 1995, Fukusima et al., 1997, Loomis et al., 1993), they also can perform diverse spatial tasks, such as imagining and drawing the furniture in a room, navigating through unfamiliar territory by means of real maps, and even charting new territory during explorations. These latter abilities suggest that real maps have a mental counterpart, and that humans and other mammals navigate by constructing and using enduring mental representations of the allocentric distances and directions of the objects and places in their environment.

In addition to behavioral evidence such as that cited above, evidence from neurophysiological experiments has been interpreted as supporting the existence of one or more cognitive maps of the environment. In particular, a variety of studies have shown that individual neurons in the hippocampus of freely moving rats are active when a rat moves through a particular region of the environment (McNaughton et al., 1995, O'Keefe and Nadel, 1978). Although vestibular, somatosensory and visual cues are effective information for the establishment and modification of this firing pattern, the firing does not appear to rely exclusively on one or another cue, for it persists when a rat is carried passively through the environment, when visual cues are removed, and when the visual field is altered by changing the rat's facing direction (e.g. Gothard et al., 1996, O'Keefe and Speakman, 1987, Quirk et al., 1990; for review see McNaughton et al., 1995). Perhaps most important, the receptive fields of different place cells and head direction cells in the same animal show internal coherence during cue manipulations (Knierim et al., 1995, Muller and Kubie, 1987). These findings suggest that ensembles of hippocampal neurons serve as a cognitive map of the environment (O'Keefe and Nadel, 1978, Wilson and McNaughton, 1993).

Nevertheless, two quite different characterizations of mammalian navigation are compatible with the behavioral and neurophysiological evidence cited above. According to one class of accounts (Gallistel, 1990, O'Keefe and Burgess, 1996, O'Keefe and Nadel, 1978), mammals form a representation of the allocentric locations of the significant objects and places in the environment. As they move through the environment, moreover, they maintain and update a representation of their own allocentric position and bearing using internal and external cues. Mammals then navigate to an unseen goal by combining the enduring allocentric representation of the goal position with their current assessment of their own position and orientation. From these quantities, animals compute the current egocentric distance and direction of the goal, which they then approach by dead-reckoning or piloting. On this account, mammalian navigation resembles the processes that humans use when we navigate by means of real allocentric maps.

Contrary to such accounts, humans and other mammals may navigate most accurately by means of processes that form and transform egocentric representations. On this view, mammals represent the current egocentric directions and distances of significant environmental locations. As they move or turn, they update these representations by a process of vector summation: the new egocentric positions of objects are computed by adding the objects' displacement vector relative to the animal to their previous egocentric position vectors.1 Thus, no stable, enduring allocentric cognitive map is explicitly represented.2 Instead, the representation of environmental locations is dynamic and transient. According to this egocentric updating account, basic processes of human navigation are quite different from the symbolic navigation processes made possible by real maps, and they are quite similar to the homing processes found in insects.

Recent studies suggest that humans do have such dynamic representations. Humans use egocentric representations of objects and scenes both in localizing and in recognizing objects (e.g. Diwadkar and McNamara, 1997, Roskos-Ewoldsen et al., 1998, Shelton and McNamara, 1997, Sholl, 1987, Tarr, 1995, Tarr et al., 1997, Tarr and Pinker, 1989). Moreover, humans update these representations as they move in order to recognize the scene or localize an object from a different viewpoint effortlessly (Fukusima et al., 1997, Simons and Wang, 1998, Wang and Simons, 1999). Neurophysiological studies have also shown that representations of visual space in parietal cortex are updated over intended eye movements (e.g. Duhamel, Colby, & Goldberg, 1992). The existence of these abilities nevertheless does not reveal whether human navigation depends primarily on updating an egocentric map or on updating one's position in an allocentric map.

The present research attempts to address this question through studies of human navigation. Our experiments depend on a central difference between an enduring representation and an observer-centered egocentric map that is updated continuously. Whereas both kinds of maps would allow accurate object localization from various novel points as an oriented animal moves around, the two kinds of maps should be affected differently by disorientation. If humans and animals navigate by a stable, enduring allocentric map, which itself remains the same regardless of the animal's movements, then the animal can always ‘look up’ the egocentric directions of each target from the same map according to its estimated self-position on the map.3 Because disorientation disrupts an animal's estimation of its own position and orientation, fully disoriented subjects will not be able to aim with accuracy to any locations on the map. Although subjects may know, for example, that a tower is 50 strides northeast of their home, they must guess their own position and orientation at random and so will fail to compute the tower's correct egocentric direction. Once disoriented subjects have guessed their own position and heading, however, the directions in which they localize a set of objects should depend on the same cognitive map that they use when oriented. Because the relative location of multiple targets is determined by the map, not the guessed self-position, target localization by oriented and disoriented subjects therefore should show equal internal consistency when tested under otherwise equivalent conditions. A disoriented subject's egocentric localization of all targets should deviate from that of an oriented subject by the same vector, equal to the difference between the subject's true versus guessed allocentric orientation and position.

Contrasting predictions come from the hypothesis that humans navigate most accurately by means of an egocentric representation, updated as they move. In order to compute the new egocentric coordinates of the target locations when one moves, one has to add a common vector to each individual target vector. Unlike using an enduring map, one has to make multiple additions instead of just updating the single vector of one's own position. Therefore, the new egocentric map preserves the relative locations among different targets if and only if all target locations are updated coherently over time. When the updating process is disrupted so that such coherence is reduced (e.g. through procedures that induce a state of disorientation), the internal consistency of pointing to different targets should decrease. Without a permanent map to refer to, a disoriented subject's pointing response should show not only an overall shift but also inconsistency among different targets.

Seven experiments compared subjects' pointing in the directions of a set of unseen targets when they were disoriented versus oriented. To simplify the analysis of these experiments, the disorientation procedure focused on subjects' representations of their heading directions. Subjects turned inertially while maintaining a constant position within the chamber. We assumed that this procedure would lead subjects to make random assessments of their heading, which in turn would produce a global shift in pointing to all targets. If subjects pointed to the targets by means of an enduring cognitive map, then disorientation should not perturb the internal angular relationships among targets. In contrast, if subjects pointed to the targets using dynamic egocentric representations maintained by continuous updating, then disorientation should impair the updating process and produce an increase in configuration error.

Section snippets

Experiment 1

In this experiment, subjects pointed to six targets first with their eyes open, then blindfolded after a small rotation, and finally blindfolded after disorientation. The initial eyes-open condition provided a measure of the represented target locations by each subject. The first eyes-closed condition served as a measure of pointing accuracy without vision in a state of orientation, after updating for a small rotation. The final disorientation condition, in which the subjects pointed to the

Experiment 2

In Experiment 2, subjects participated in the same three conditions as in Experiment 1, with two changes in procedure. First, subjects were seated throughout the experiment in a swivel chair at a fixed position but variable orientation, in order to minimize the possibility that they would perceive their position to change over the course of the study. Second, subjects were given a 30 s recovery period after the disorientation procedure and before the final pointing test, so as to allow their

Experiment 3

In Experiment 3, subjects pointed to body-centered orientations (e.g. ‘to the right’) rather than to the target objects in the environment (e.g. ‘to the table’) under conditions that closely paralleled those of the preceding experiments, and the consistency of different pointing responses was compared when subjects were oriented versus disoriented. If the disorientation procedure directly affected subjects' motor control but not their spatial memory, then the subjects in Experiment 3 should

Experiment 4

In Experiment 4, subjects were given the same pointing task for the same three conditions as in Experiment 2. They pointed to external targets with eyes open, with eyes closed after a small rotation, and with eyes closed after the extensive rotation that was used in Experiments 1–3 to induce a state of disorientation. In the present experiment, however, subjects were tested with a blindfold that was translucent rather than opaque, through which a single asymmetrically placed light produced a

Experiment 5

In Experiment 5, subjects were tested under identical conditions to those of Experiment 4, with one exception. Although they wore a translucent blindfold and were trained and tested with a single, asymmetrically placed light, the light was extinguished during the rotation procedure that preceded the final pointing condition and then re-illuminated for that condition. Therefore, subjects lost their sense of orientation during the disorientation procedure, but they reoriented themselves before

Experiment 6

In this experiment, subjects were presented with a rectangular room furnished with four distinct objects arranged in a similar, but smaller, rectangular configuration. Each subject participated in an objects task and a corners task, in which they pointed to targets (objects or corners) without vision, both before and after disorientation. For each task, we obtained three error measures from subjects' pointing responses as before: the heading error, the pointing error, and the configuration error

Experiment 7

Experiment 7 compared subjects' pointing to room corners with their pointing to objects both before and after disorientation. In this experiment, however, we used an irregularly shaped room instead of the rectangular room, and four identical objects instead of distinctive ones. In both conditions, subjects were given the same verbal commands, to point clockwise or counterclockwise to all targets (objects or corners), starting from a single verbally specified target. If the differing effects of

General discussion

The present experiments provide evidence that the representation of the relative directions of objects is impaired when subjects lose their sense of orientation. Subjects made significantly larger errors in their assessment of the spatial relationship among target objects, as measured by pointing to individual targets, after they lost track of their own orientation. This effect was not due to a decrease in pointing accuracy per se, was not a direct result of physical stimulation of the

Acknowledgements

We thank James Cutting, Thomas Gilovich, C.R. Gallistel, and Nancy Franklin for comments on earlier versions of this paper, and James Knierim, Linda Hermer-Vazquez, Richard Darlington, Steven Pinker, Nancy Kanwisher, Mary Potter and William Warren for valuable discussions. This study was supported in part by a grant to E.S.S. from NIH (R37-HD 23103). Experiments 2, 4 and 5 were presented at the 19th Meeting of the Cognitive Science Society, Stanford, CA, 1997. Experiment 6 was presented in the

References (46)

  • V.A Diwadkar et al.

    Viewpoint dependence in scene recognition

    Psychological Science

    (1997)
  • J.R Duhamel et al.

    The updating of the representation of visual space in parietal cortex by intended eye movements

    Science

    (1992)
  • F.C Dyer

    Spatial memory and navigation by honeybees on the scale of the foraging range

    Journal of Experimental Biology

    (1996)
  • R Epstein et al.

    A cortical representation of the local visual environment

    Nature

    (1998)
  • A.S Etienne et al.

    Path integration in mammals and its interaction with visual landmarks

    Journal of Experimental Biology

    (1996)
  • S.S Fukusima et al.

    Visual perception of egocentric distance as assessed by triangulation

    Journal of Experimental Psychology: Human Perception and Performance

    (1997)
  • C.R Gallistel

    The organization of learning

    (1990)
  • K.M Gothard et al.

    Dynamics of mismatch correction in the hippocampal ensemble code for space: interaction between path integration and environmental cues

    Journal of Neuroscience

    (1996)
  • L Hermer et al.

    A geometric process for spatial reorientation in young children

    Nature

    (1994)
  • J Huttenlocher et al.

    Categories and particulars: prototype effects in estimating spatial location

    Psychological Review

    (1991)
  • J.J Knierim et al.

    Place cells, head direction cells, and the learning of landmark stability

    Journal of Neuroscience

    (1995)
  • Levinson, S. C. (in press). The role of language in everyday human...
  • J.M Loomis et al.

    Nonvisual navigation by blind and sighted: assessment of path integration ability

    Journal of Experimental Psychology: General

    (1993)
  • Cited by (319)

    View all citing articles on Scopus
    View full text