Abstract
Two experiments were performed to investigate whether two measures of spatial memory, working memory (WM) and reference memory (RM), can be generalized over learning situations. In the first experiment, outbred WU rats were used; in the second four inbred strains. In both experiments rats were given 50 training trials in both a holeboard and a high-sided radial maze: 4 out of 16 holes and 4 out of 8 arms were baited. The number of WM (revisits to baited holes/arms) and RM (visits and revisits to never baited holes/arms) errors were calculated for 5 blocks of 10 trials; odd-even reliabilities were determined for the last 20 trials and were shown to be satisfactory for both measures in both types of apparatus. Error-decrements were analyzed by trend components. For both WM and RM, simple and canonical correlations were determined between means and significant trends in maze and holeboard as a test for construct validity; in both experiments r's and R c's were very low. It was concluded that the concepts WM and RM are not validly generalizable over tests.