An often used animal model based on the effects of neuroleptics on spontaneous behaviour is the catalepsy test. However, this test seems to be particularly insensitive to the atypical neuroleptics thioridazine and, especially, clozapine. We have therefore developed an alternative test, the paw test, which measures the ability of drugs to prevent the spontaneous withdrawal of fore- and hindlimbs in rats, and have compared this with the classical catalepsy test. The results show that: 1) the classical neuroleptic drugs haloperidol and chlorpromazine, the atypical neuroleptic drugs clozapine and thioridazine, the potential atypical neuroleptic drugs molindone and SCH 23390, and the potential classical neuroleptic drug metoclopramide are potent in increasing the hindlimb retraction time; 2) the paw test discriminates between classical neuroleptics which are equipotent in prolonging both the forelimb (FRT) and hindlimb retraction time (HRT) an atypical neuroleptics which are much more potent in prolonging HRT than in prolonging FRT; 3) the non-neuroleptic drugs desipramine, diazepam and morphine do not influence the variables measured in the paw test, although morphine does produce catalepsy; 4) Molindone as well as SCH 23390 behave like atypical neuroleptic drugs in the paw test. In comparison with the classical wood block catalepsy test, the paw test is shown to be superior for predicting the profile of the neuroleptics tested. Although more neuroleptics and non-neuroleptics have to be tested to determine whether false positives and false negatives do occur, we feel that the paw test might be an interesting animal model, because the increase in hindlimb retraction time was associated with the antipsychotic potential, whereas the increase in forelimb retraction time was associated with the potential to induce so-called extrapyramidal side effects.