Cue competition between shapes in human spatial learning

In many species, including humans the basic ability to move to a goal is essential to survival. Central to understanding how this ability operates in the cognitive systems of humans and other animals is whether learning about spatial relationships follows the same principles as learning about other kinds of contingent relationships between events. In non-spatial contingent relationships, learning about one stimulus can influence learning about other stimuli. For example, in blocking, learning that cue-A predicts an outcome can reduce learning about a subsequently added cue-B that is paired with cue-A when both cues predict the same outcome (Kamin, 1969). To the extent that spatial learning operates according to similar principles to other forms of contingency learning, spatial cues that can be used to locate a goal should also compete with each other. Failure to find blocking between spatial cues that can be used to locate a goal would be consistent with an alternative account of how spatial knowledge is acquired and used: one that assumes a quite different learning mechanism. For example, the hypothesis of locale learning assumes that a cognitive map of the environmental layout is automatically updated when cues are added or removed from the environment (O'Keefe and Nadel, 1978). Automatic updating implies that added or removed cues will be processed irrespective of what is learned about other cues, rather than competing with or otherwise interacting with those other cues. A second, related, hypothesis is that the geometric properties of the environment are processed in an independent module that is impervious to cue competition from non-geometric features (Cheng, 1986; Gallistel, 1990). This hypothesis implies that geometric cues within the module are also immune to competition from each other. In the current experiments, evidence for blocking of goal location learning was investigated in virtual environments (VEs) in which the presence or absence of large-scale structures can be manipulated. Experiment 1 found that an irregular-shaped flat-walled enclosure blocked learning about a landmark subsequently placed within its boundaries, providing preliminary evidence that shape may not be processed in a specialised module. However, many participants appeared not to be using shape to locate the goal. In the remaining experiments, spatial cues were large-scale 2D shapes presented on the ground which ensured that participants perceived overall shape. Experiments 2 and 3 found no evidence of blocking between shapes when these stimuli were presented in the context of minimal "auxiliary" cues. When additional auxiliary stimuli were presented throughout learning in Experiment 4, a direction consistent with blocking was found, but the effect was not statistically significant. In Experiments 5 and 6 a clear blocking effect was found under circumstances that suggested that the critical variable to finding blocking was the number of irrelevant shapes present either during training or at test. Experiment 7 confirmed that, rather than the test conditions, the presence or absence of stimuli during one or both training phases was the crucial variable in promoting blocking. Experiment 8 investigated the hypothesis that an initial process of learning to ignore irrelevant shapes in phase 1 is a requirement for blocking of learning. In the absence of auxiliary cues in phase 1, blocking was not found. The implications of these outcomes are discussed in relation to the hypothesis of specialised geometric processing, changes in attention, and the conditions of discrimination learning