An attempt to account for controversial estimates of working memory persistence in the radial maze.

Factors affecting performance in the radial maze were examined to explain widely differing estimates of the persistence of spatial working memory. In Experiment 1, two groups of 10 rats each, trained in the 8-arm elevated maze (EM) or two-level maze (TLM), showed after 30 trials 0.66 and 0.34 errors per eight choices, respectively. In Experiment 2 insertion of a delay between choices 4 and 5 increased error incidence in choices 5 to 8 from 0.8 and 1.0 at 2-min delay to 1.8 and 1.9 at the 1-hr delay in the EM and TLM, respectively. (Chance level is 2.0 errors in choices 5 to 8). In Experiment 3, omission of the masking food odor sources improved performance in the EM at the 1-hr (1.1 error) but not at the 4-hr delay (1.85 error). In Experiment 4, elimination of trail marking by replacement of maze A, used in choices 1 to 4, by maze B in choices 5 to 8 caused a significant deterioration of delayed performance in the TLM but not in the EM. In Experiment 5, a continuous series of 15 trials with the 40-min delay improved performance in the EM from 1.60 to 1.35 errors in choices 5 to 8. High error incidence (0.41) on choice 5 was due to rat's preference for channels visited on choices 1 and 2. Significant decrease of errors on choice 6 (to 0.25) indicates that the rats responded to incorrect choice 5 by activation of an error-correcting strategy. Examples of such strategies are given which increase delayed performance well above chance level even in the absence of working memory. Error-induced change of response pattern is indicated by the observation that the interchoice distances 1--2 and 5--6 are the same after correct, but differ by 0.83 channel after incorrect choice 5. Possible reasons for the failure to confirm the long persistence (more than 4 hr) of spatial working memory reported by Beatty and Shavalia (Behavioral and Neural Biology, 28, 454--462, 1980) are discussed. It is argued that longevity of the working memory record gives the rat no advantage in foraging and must be, therefore, carefully checked against possible alternative explanations.