Cocaine was administered to gravid C57BL/6J mice on embryonic days E8-18 at doses of either 17.5 or 20 mg/kg x 2 per day; controls received equal volumes of vehicle. The two cocaine dose groups were indistinguishable in their effects on maternal weight gain, offspring survival or body weight; therefore, the two groups were combined. Offspring were assessed as adults in straight channel swimming, cued and spatial reference-memory and working memory versions of the Morris water maze (MWM), and in the Barnes spatial maze to escape from a light, tone and fan. Cocaine offspring had shorter latencies in the straight channel and increased cumulative distance from the platform and path length in the spatial version of the Morris maze, but only when the platform size was reduced, not under standard platform conditions. In the working memory test, cocaine offspring showed deficits in acquisition and, following random trials, on relearning during a final test phase. In the Barnes maze, cocaine offspring were delayed in utilizing more efficient search strategies and took longer to find the goal. Taken together, the data suggest that prenatal cocaine induces modest but significant long-term alterations in both reference and working memory-based spatial learning and memory.