Introduction: An animal with an unimpaired reference memory can distinguish between alternatives that belong to a rewarded set and those that are always unrewarded. An animal with an unimpaired working memory can distinguish beween alternatives where it has been rewarded (e.g. food has been eaten, but not replaced) and those where it will still be rewarded. Olton et al., 1979 proposed that fimbria-fornix or hippocampal lesions impairs working rather than reference memory in a radial maze. This hypothesis was tested for rats with damage to the hippocampus, limbic ventral tegmentum (VTA A10 ) and neocortex, intact and operated controls on a 16-hole-board search task. Methods: Food-deprived animals searched for food pellets placed consistently in 4 holes of a 16-hole-board (figure 1). They were presented with 11 sessions of 10 trials/session. There were five groups of animals, - one with aspiration lesions of the hippocampus and overlying neocortex, one with damage only to the overlying neocortex and sham-controls that went through the procedure but the brain was left intact (Oades and Isaacson, 1978) - VTA damage resulted from coagulation with a stylet in a sterotaxically implanted cannula and their controls received the cannula alone. Working memory error = a visit to a correct hole that has just been visited, and thus no longer contains a food pellet. Reference memory error = visit to a hole that is never baited. Results: 1/ A referenceand a working memory impairment (in terms of errors made) was recorded for animals with hippocampal or with VTA damage. 2/ The impairments were significant by session 3 and the differences amounted to more than 50% by the end of testing. Conclusions: .There was a striking similarity between the performances of animals with damage to the hippocampus and those with damage to the VTA (that projects to the lateral septum, entorhinal cortex and dentate gyrus. The different results obtained by Olton in the radial maze may be explained by the discrete trial testing conducted in the radial maze that contrasts with the multiple choices that an animal makes on a hole-board. Further in the current study training occurred exclusively post-operatively, while in the radial maze animals had received some preoperative training. Both lesioned and control animals expressed preferred sequences of hole-visits. The preference was weaker in the lesioned animals but the number of changes of preference between sessions did not differ between groups. Thus it is argued that limbic and mesolimbic DA substrates are crucially involved in attentive mechanisms important to adaptive learning and the impairment is not merely one of forming and using memory.