Strategic Differences in the Coordination of Different Views of Space

Strategic Differences in the Coordination of Different Views of Space Glenn Gunzelmann (glenng@andrew.cmu.edu) John R. Anderson (ja+@cmu.edu) Department of Psychology, Baker Hall 342-C Carnegie Mellon University Pittsburgh, PA 15213 Abstract Participants were trained to use one of two different strategies in an orientation task, which were based on verbal reports from participants in another experiment. The data provide support for the conclusion that participants in the two training conditions searched the screen differently to complete the task, but that neither group used mental transformations like image rotation. These results have implications for research in this area as well as for conceptualizing how individuals perform such tasks. A comparison of the results from the two strategy conditions is made based on an ACT-R model of one of them. Small differences in how information on the screen is scanned can produce the observed differences in performance. Introduction The coordination of different views of space is a fundamental task in human functioning. An everyday example of it involves determining which way to turn at an intersection by using a map. The visual scene presents one view of the space (egocentric), while the map presents an alternative representation (allocentric). In order to accurately decide which way to go, it is necessary to bring these two views of the space into correspondence. Of course, with a physical map it may be possible to actually rotate it to align it with your own orientation. In other situations, mental transformations may need to be done in order to coordinate these views to make accurate decisions. On a continuum of reasoning about orientation within a space, deciding whether the correct turn is left or right is a fairly straightforward task. Still, research on this issue has shown that it becomes increasingly difficult to perform as a function of the difference in orientation between the two views of space (Shepard and Hurwitz, 1984). The phenomenon bears a strong resemblance to findings in the mental rotation literature (Shepard and Metzler, 1971) where the time needed to determine that two objects are identical increases linearly as a function of the angular disparity between them. These findings have been used to support the conclusion that performance in orientation tasks involves analog mental rotation of mental images. Note, however, that the task of coordinating views of space adds a layer of complexity to the traditional mental rotation task. In a spatial orientation task, the information is presented in two different formats. Thus, deciding whether the visual scene matches the information on the map requires additional reasoning beyond the image transformation. In an important series of experiments, Hintzman, O’Dell, and Arndt (1981) had participants perform orientation tasks in a variety of ways. In the basic task, participants had to indicate the direction of a target relative to a given orientation. Figure 1 shows the orientation task used in the experiment presented here. In this figure, the left side represents the target field as viewed from a camera (on a plane above the field) and the darkened circle indicates the target. The right side represents a map-view with the target field at the center. The arrow on that side shows the camera's orientation for viewing the target field. Participants are asked to indicate in which cardinal direction the target is located relative to the center of the target field. In the sample trial in Figure 1, the correct response is South. The general finding is that decisions for targets in line with the assumed orientation are made more rapidly, and response times for other targets increase as they depart from the nearest point immediately in front of the viewpoint. Although not explicitly addressed by Hintzman, et al., this increase in response time is not strictly linear. In addition, no evidence was presented in their study about how participants claimed to be performing the task. In order to investigate what factors influence performance on this task, we asked participants to complete the task and then questioned them as to the manner in which they solved it. While we will not go into detail about this experiment, the data are presented below and bear a strong resemblance to results from similar studies, including Hintzman, et al. (1981). However, by questioning participants after they had completed the experiment, we discovered that participants were using at least two distinct strategies to do the task. Some participants claimed to be implementing a strategy that incorporated imagery and mental rotation to determine correct responses.