Do hippocampal pyramidal cells signal non-spatial as well as spatial information?

It is generally agreed that the rat hippocampus is involved in spatial memory. Whether this is its sole or primary function, or merely one component of a broader function, is still debated. It has been suggested, for example, that the hippocampus stores information about flexible relations between stimuli, both spatial and non-spatial. In this paper, I reiterate the basic tenet of the cognitive map theory that the processing and storage of spatial information is the primary and perhaps the exclusive role of the hippocampus in the rat, and that data that appear to contradict this have been misinterpreted. These data are found in reports of non-spatial correlates of unit activity recorded in the awake animals and reports of deficits on non-spatial tasks following hippocampal lesions. In this paper, I examine both claims and suggest alternative explanations of the data. The first part of the paper contains a review of some of the properties of hippocampal place cells, which might be misinterpreted as non-spatial in "non-spatial" tasks. For example, if an animal is trained to carry out a sequence of stereotyped actions in different parts of an environment, there will be a strong correlation between the performance of each behaviour and the animal's location, and it is necessary to rule out the locational correlate as the cause of the firing pattern. The second part of the paper looks at the results of experiments on conditioning and non-spatial discrimination tasks and concludes that the results are less supportive of a more general relational theory of hippocampal function than has been suggested. Furthermore, there is often a discrepancy between the correlates of unit firing in non-spatial tasks and the absence of an effect of hippocampal damage on these same or similar tasks. It is concluded that, contrary to the claims of its detractors, the cognitive map theory is still the theory of hippocampal function that is most clearly specified, makes the most testable predictions, and for which there is the strongest experimental support.