Atypical antipsychotics are expected to be better tolerated than older antipsychotics because of their lower propensity to cause certain adverse effects. All atypical drugs have been shown to cause fewer acute extrapyramidal symptoms (EPS) than a standard typical agent (usually haloperidol) and some (clozapine, sertindole and quetiapine) appear to cause these effects no more often than placebo. In the longer term, clozapine, olanzapine and (less robustly) other atypical antipsychotics are thought to cause less tardive dyskinesia than typical antipsychotics. Problems caused by hyperprolactinaemia occur less often with some atypical antipsychotics than with typical drugs although risperidone and amisulpride appear to have no advantages in this respect. Other adverse effects may occur as frequently with some atypical antipsychotics as with some typical drugs. Clozapine, risperidone and quetiapine are known to cause postural hypotension; clozapine, olanzapine and quetiapine are clearly sedative; and anticholinergic effects are commonly seen with clozapine, and, much less frequently, with olanzapine. Some adverse effects are more frequent with atypical drugs. Idiosyncratic effects seem particularly troublesome with clozapine and, to a lesser extent, sertindole, olanzapine and zotepine. Bodyweight gain is probably more problematic with atypical antipsychotics than with typical drugs. Overall tolerability, as judged by withdrawals from therapy, is not clearly proven to be better with atypical drugs, although some individual trials do indicate an advantage with atypical agents. Differences in tolerability between individual atypical antipsychotics have not been clearly shown. The tolerability profile of atypical drugs certainly benefits from a lower incidence of acute EPS effects, along with less certain or less uniform benefits in symptomatic hyperprolactinaemia or tardive dyskinesia. Other, perhaps more trivial, adverse effects militate against their good tolerability, and effects such as bodyweight gain may severely reduce tolerability. Without clear advantages in tolerability in patient groups used in trials, drug choice in regard to adverse effects should continue to be on a patient to patient basis.