The effects of intradentate colchicine injections on the performance of tasks requiring spatial working and reference memory are controversial. Multiple-site colchicine injections (7 microg/microl; via a drawn micropipette) throughout the dentate gyrus (DG) of rats (nine sites in each hemisphere, 0.06 microl at each site) selectively destroy about 90% of the DG granule cells, as revealed by quantitative stereological estimates; stereology also revealed minor neuronal losses in the CA4 (33%) and CA1 (23%) subfields, but lack of damage to the CA3 hippocampal subfield. Spatial reference and working memory were assessed in Morris' water maze; in the reference memory task, the rats were required to learn a single, fixed location for the platform over several days of training; in the working memory task, animals were required to learn a new platform location every day, in a matching-to-place procedure. Compared to sham-operated controls, lesioned rats showed significant disruption in acquisition of the reference memory water maze task; however, the data reveal that these rats did acquire relevant information about the task, probably based on guidance and orientation strategies. In a subsequent probe test, with the platform removed, lesioned rats showed disruption in precise indexes of spatial memory (e.g., driving search towards the surroundings of the former platform location), but not in less precise indexes of spatial location. Finally, the lesioned rats showed no improvement in the match-to-place procedure, suggesting that their working memory for places was disrupted. Thus, although capable of acquiring relevant information about the task, possibly through guidance and/or orientation strategies, DG-lesioned rats exhibit a marked difficulty in place strategies. This is particularly evident when these rats are required to deal with one-trial place learning in a familiar environment, such as in the working memory version of the water maze task, which requires flexibility in the use of previously acquired information.