Patrick Peruch

Homing in Virtual Environments: Effects of Field of View and Path Layout

Triangle completion (ie homing to the starting point after completing two legs of a triangle) is a widely used method for examining path-integration abilities in animals and humans. Two experiments are reported in which homing was used to examine the efficiency of purely visual mechanisms (eg optical flow) for spatial-information coding and integration. Adult observers had to complete triangles in an interactively simulated three-dimensional environment which consisted of two critical objects and a homogeneous set of white cylinders serving as background. Each participant completed twenty-seven triangles corresponding to a factorial combination of three geometrical fields of view (40°, 60°, or 80°) and nine triangle layouts (with variations of the first turning angle and the second leg). Homing performances revealed strong effects of triangle layout, but no effect of geometrical fields of view: variations in the amount of simultaneous visible spatial information did not influence the acquisition of spatial knowledge in the environments used. Applying the encoding-error model to the data revealed severe systematic errors of picking up directional information while moving through visually simulated environments. These results are discussed with respect to informational differences between situations of purely visual and nonvisual navigations in space.

Transfer of Spatial Knowledge from Virtual to Real Environments

The transfer of spatial knowledge from virtual to real environments is one important issue in spatial cognition research. Up to now, studies in this domain have revealed that the properties of spatial representations are globally the same in virtual and real environments, and in most cases transfer of spatial information from one kind of environment to the other occurs. Although these results suggest that virtual environments contain much of the spatial information used in real environments, it seems difficult or even impossible to draw any clear conclusion about the spatial information which is transferred and about the conditions of transfer. Being able to quantitatively and/or qualitatively predict and observe such a transfer would broaden the possibilities of training and our knowledge of the cognitive processes involved in spatial behavior. In a first step, arguments in this sense are developed on the basis of a review of some recent studies concerned with the transfer of spatial knowledge between virtual and real environments. In a second step, empirical data are reported, that illustrate the interest and limits of such studies.

Active versus passive learning and testing in a complex outside built environment

A review of the evidence on active and passive learning in virtual environments (VEs) suggests that both conditions have shown superiority under some conditions of learning and testing, but there is no consistent outcome pattern. Measures of transfer between virtual and real environments have also revealed a variety of outcomes. Following either active or passive learning in a VE, experiment 1 assessed measures of orientation and distance estimation in that VE and in a real-world equivalent environment. On measures of direct and relative distance, more accurate estimates were found for active than passive VE explorers. A suggestion was also noted for the orientation estimates to benefit from real-world rather than VE testing. With an improvement to the procedure, experiment 2 found similar real versus virtual orientation judgements, suggesting that an opportunity for active learning during the test procedure probably influenced orientation measures in experiment 1. We conclude that the effects of interactivity are unreliable and vary with the measures used, and that testing in virtual and real environments leads to similar outcomes.

Distance cognition by taxi drivers and the general public

Twenty-four taxi drivers and an equal number of non-professional or full-time drivers who had been residents of Paris for at least 10 years were requested to estimate straight-line and travel distances, in either distance or time units, between pairs of familiar locations in the city. The results showed that travel distances were invariably estimated as longer than straight-line distances, indicating that the subjects used the knowledge they had acquired of routes. Furthermore, taxi drivers did not make fewer systematic errors than the general public in estimating straight-line distances but they did estimate travel distances as shorter. In support of the view that taxi drivers differ from the general public primarily in procedural knowledge (e.g. of how to drive from A to B) rather than in declarative knowledge (of the straight-line distance between A and B), this finding was interpreted as showing that taxi drivers knew short-cuts to a greater extent than the general public.

Route and survey processing of topographical memory during navigation

We investigated the characteristics of route and survey processing of a unique complex virtual environment both at the behavioral and brain levels. Prior to fMRI scanning, participants were trained to follow a route and to learn the spatial relationships between several places, acquiring both route and survey knowledge from a ground-level perspective. During scanning, snapshots of the environment were presented, and participants were required to either indicate the direction to take to follow the route (route task), or to locate unseen targets (survey task). Data suggest that route and survey processing are mainly supported by a common occipito-fronto-parieto-temporal neural network. Our results are consistent with those gathered in studies concerning the neural bases of route versus survey knowledge acquired either from different perspectives or in different environments. However, rather than arguing for a clear distinction between route and survey processing, “mixed” strategies are likely to be involved when both types of encoding take place in the same environment.

Route knowledge in different spatial frames of reference

Orienting oneself in space requires establishing a correspondence between various spatial frames of reference (SFR) in which the same information about the environment can be encoded in different ways and formats. In this encoding process, one key point is the alignment of the SFRs, which may require additional operations such as a mental or real rotation. Three experiments were conducted to investigate the process of spatial orientation under aligned and misaligned conditions. Subjects were shown animated sequences of decision points perceived along a route (Experiments 1 and 2) or verbal route instructions (Experiment 3) to which they had to attribute a path on a map. The results showed that when the orientations of the map and the route were different (misalignment) both total time and errors increased. The route length (Experiment 1), and the need to reverse the direction of the path (reverse response condition in Experiments 2 and 3) also led to a decline in performance. In Experiment 3, the map-rotation strategy was found to be pertinent for solving misaligned spatial problems.

Mental Representation and the Spatial Structure of Virtual Environments

This study aimed to investigate the influence of the spatial structure of virtual environments on the mental representation that people form. Three groups of participants visited the locations of several objects in simulated environments that differed in aspects of their structure. After learning, the objects were removed, and participants were required to estimate their former positions in terms of angle of orientation and straight-line distances. Results indicated that the structure of the environment influences performance and suggested the use of different underlying mental representations: If the spatial structure does not obscure direct perception between locations, a procedural type of spatial representation may be sufficient; however, if the spatial structure prevents direct perception between locations, it is likely that people need to elaborate a configurational type of spatial representation. These results stress the importance of spatial structure in the acquisition and use of mental representations.

How do we locate ourselves on a map: a method for analyzing self-location processes

Abstract The activity of self-location can be described as the making of a connection between the spatial relationships perceived from the subjects position (egocentric frame of reference) and all the external spatial relationships which contain this position (geocentric frame of reference: a map for example). We propose a method for analyzing the cognitive processes involved in a self-location task where the subjects had to match 2 spatial frames of reference. The processing methods observed by measuring various parameters used to evaluate the spatial relationships showed that the subjects (adults) could use a wide range of spatial tools - which were often equivalent and thus redundant - to locate themselves on a map, and that they drew upon this spatial repertory according to their requirements.

Spatial Orientation in Virtual Environments: Background Considerations and Experiments

Spatial orientation strongly relies on visual and whole-body information available while moving through space. As virtual environments allow to isolate the contribution of visual information from the contribution of whole-body information, they are an attractive methodological means to investigate the role of visual information for spatial orientation. Using an elementary spatial orientation task (triangle completion) in a simple virtual environment we studied the effect of amount of simultaneously available visual information (geometric field of view) and triangle layout on the integration and uptake of directional (turn) and distance information under visual simulation conditions. While the amount of simultaneously available visual information had no effect on homing errors, triangle layout substantially affected homing errors. Further analysis of the observed homing errors by means of an Encoding Error Model revealed that subjects navigating under visual simulation conditions had problems in accurately taking up and representing directional (turn) information, an effect which was not observed in experiments reported in the literature from similar whole-body conditions. Implications and prospects for investigating spatial orientation by means of virtual environments are discussed considering the present experiments as well as other work on spatial cognition using virtual environments.

Mental Scanning of Images Constructed from Visual Experience or Verbal Descriptions: The Impact of Survey versus Route Perspective

Mental scanning was used to assess the metric properties of mental spatial representations derived from visual experience or the processing of a verbal description, and either from survey or route perspective. Participants were asked to mentally scan their images of a spatial environment they had learned in one of the following four conditions: Visual-Survey, Visual-Route, Verbal-Survey, and Verbal-Route. No difference was found between the scanning times of the visual and verbal conditions, but scanning times were shorter after survey than route acquisition, and they consistently increased as a function of the Euclidean distances between locations in the environment, with steeper slopes in the route conditions. These results demonstrate that mental spatial representations derived from different sources and perspectives are endowed with similar properties and preserve the Euclidean characteristics of the original environment, but that they are easier to access when they have been constructed from a survey perspective.