Acquisition of Spatial Knowledge Through Visual Exploration of Simulated Environments

Abstract

The capacity of observers to learn spatial layouts of objects located in a graphically displayed wall-limited space was tested in a target localization task after active or passive exploration of the simulated environment. The spatial layouts were composed of four target cubes of different colors and obstructing inner walls. In the active exploration condition, the observers freely explored the space, smoothly changing the view point by means of a joystick and attempting to locate each target cube. In two passive exploration conditions, the observers explored the environments by looking either at a continuous changing scene as viewed along a smoothly travelled path or at a series of static scenes as viewed from successive points of view along a path. Immediately after exploration, the observer's task was to reach, using the shortest path, a specified target not visible from the starting point situated near the center of the environment. We hypothesized that spatial acquisition resulting after active exploration would be more accurate than after passive exploration, and that dynamic visual information used to display the environment would be more efficient than static visual information. As expected, the performance was better for active exploration than for passive exploration, but dynamic and static passive conditions yielded equivalent performance. Further spatial analyses of the trajectories produced to reach the target and subsequent analyses between trajectories and corresponding target reaching score and target reaching time allowed us to classify the participants' strategies according to different levels of learning of the environment layouts. Some participants were able to build up an absolute representation of the environment, including the objects' locations and orientations, and others learned the spatial layouts of objects by relating spatial clues to each other. A few participants were unable to memorize the environment layouts and exhibited random-like search paths. These results show the link between the level of performance and the level of spatial knowledge, and confirm the importance of active motor behavior combined with active perception to extract invariants in the environment.