Where is the cheese? Within-compound spatial learning in rats

Spatial cognition is crucial for our survival as it allows us to remember where to find food, water and shelter. But, with a vast amount of information surrounding us, which of the many cues do we actually use when trying to reach a goal? And how do we learn about these competing cues? This question was investigated by Joe Austen and Anthony McGregor (Durham University) in a recent paper published in Learning and Behavior.

We know that animals can use a variety of different cues to get to a goal and that, in some cases, they can use multiple types of cues. One current theory suggests that learning about the different types of cues, for example the geometry of a space vs. the landmarks within it, progresses independently. However, the results of a series of studies by Austen and McGregor have called this idea into question.

Austen and McGregor trained rats to search for a hidden platform in a wedge-shaped water maze. In initial training, the rats could use landmark information and/or geometric information to locate the platform. After learning about the position of the platform, the rats received a test trial in which they were placed in a different shaped pool. The test pool was circular and thus did not contain the geometric information that the rats could have learned about during initial training. The landmarks, by contrast, were again present in the novel pool and the rats’ search behavior was monitored. This test revealed that the rats had learned about the landmarks; they generally searched in the area surrounding the “correct” landmark (the one which had been above the platform during training). This tells us that the rats can use landmark information to find a goal even when geometric information is absent or uninformative.

In the next training phase, the rats were again placed in the original pool but no landmarks were present. The animals received additional training in which, for half of them, the platform was in the same geometric position as in initial training (the “consistent” group) whilst for the other half, the platform was moved to a new position – where the previously “incorrect” landmark had been (“inconsistent” group). Remember that the landmarks are not present at this stage. After a number of training sessions, the animals were then given another test, identical to the first, in which they were presented with the landmarks in the circular pool.

The rats in the consistent group behaved in the same way as they had in the initial test – searching around the landmark which had been above the platform during initial training. However, the inconsistent group showed no discrimination between the two landmarks. This is exciting as it suggests that learning about the different cue types is not independent, but rather, whilst learning, the animals formed a within-compound association between the cue types:  The presence of the landmark made the rats think about the geometric cue and, as the geometric cue no longer indicated the presence of the platform, the rats did not search under the landmark.

This suggests a very different mechanism controlling the learning of geometric and landmark cues in a spatial task. Austen and McGregor suggest that the animals form within-compound associations when learning the task. This changes the way we think about processing of these cues and may be able to explain apparently anomalous results in the spatial learning literature.