The discriminative stimulus properties of the prototypical atypical neuroleptic clozapine (5 mg/kg, IP) were characterised in rats using a fixed ratio assay. Clozapine induced full dose-related generalization in the absence of response suppression. Amphetamine and pentylenetetrazol failed to generalise at doses known to be discriminable, showing a degree of specificity for the clozapine cue. The typical neuroleptics haloperidol and loxapine induced minimal (20%) generalization at doses with marked behavioural effects; thus clozapine discrimination dissociates clozapine from typical neuroleptics. Atypical neuroleptics which are not clozapine congeners produced weak partial generalization when tested up to the highest doses that could be studied. The maximal levels of generalization induced by these agents were: amisulpiride 28%, risperidone 40% and sertindole 50%. Clozapine congeners typically caused more generalization, the novel pyridobenzoxapine JL13 inducing 70% maximal generalization. Most generalization (83%) was seen with the clozapine congener seroquel, although in contrast to clozapine, it only generalised at doses with marked effects on responding, so that no drug mimicked clozapine fully. Surprisingly, the clozapine congener olanzapine only induced a maximal level of 38% generalization. This apparently anomalous finding is attributed to an inability to test high doses of the drug due to its rate-suppressant actions. The clozapine cue can be used to rank atypical neuroleptics in terms of their similarity to clozapine in vivo. The clozapine cue is probably a compound cue, since only agents showing "polyvalent" receptor pharmacology induced substantial generalization.