Trimethyltin (TMT) was used as a positive control neurotoxicant to evaluate a repeated acquisition procedure for the 8-arm radial maze. Ten male Long-Evans rats were trained to collect a single food pellet at the end of each baited arm on each trial of a daily 12-trial test session. Four of the eight arms were baited on all trials of a given session. The set of four baited arms was changed each day: thus the rats were required to learn a new set of baited arms in each session. In trained rats, error frequencies (entries into unbaited arms) declined from about 4 on Trial 1 to less than 1 on Trials 4-12 in each session: this within-session error reduction thus defined an acquisition baseline which was evaluated for its sensitivity to TMT. Learning was impaired after 7 mg/kg (iv) TMT, as shown by a slower decline in within-session error frequencies in all treated rats. Errors and response times were elevated for 5 weeks after TMT but returned to control levels thereafter. Histological examination of hippocampi showed damage in all treated rats 18 weeks after treatment; however, no significant relationship between degree of damage and behavioral effect was observed. Analysis of errors showed that TMT more strongly impaired the rats' ability to avoid arms in the current unbaited set than those baited arms already entered on a given trial (i.e., working memory). These dissociations between behavior and hippocampal morphology in terms of time course, magnitude of effect across animals, and error type suggest that performance of this task does not depend upon hippocampal integrity, as do other tasks involving spatial working memory. Recovery of function in this kind of task may shed light on processes of neural plasticity after exposure to neurotoxic compounds.