Antipsychotic drugs can be of great benefit in a range of psychiatric disorders, including schizophrenia and bipolar disorder, but all are associated with a wide range of potential adverse effects. These can impair quality of life, cause stigma, lead to poor adherence with medication, cause physical morbidity and, in extreme cases, be fatal. A comprehensive overview of tolerability requires a review of all available data, including randomised controlled trials (RCTs), observational studies and postmarketing surveillance studies. Assessing the relative tolerability of atypical antipsychotics is hampered by the paucity of RCTs that compare these drugs head-to-head, and limited and inconsistent reporting of adverse effect data that makes cross-study comparisons difficult. Despite methodological problems in assessment and interpretation of tolerability data, important differences exist between the atypical antipsychotics in the relative risk of acute extrapyramidal symptoms (highest risk: higher doses of risperidone), hyperglycaemia and dyslipidaemia (highest risk: clozapine and olanzapine), hyperprolactinaemia (highest risk: amisulpride and risperidone), prolongation of heart rate-corrected QT interval (QTc) [highest risk: ziprasidone and sertindole] and weight gain (highest risk: clozapine and olanzapine). Sedation, antimuscarinic symptoms, postural hypotension, agranulocytosis and seizures are more common with clozapine than with other atypical antipsychotics. The variation in their tolerability suggests that it is misleading to regard the atypical antipsychotics as a uniform drug class, and also means that the term 'atypical antipsychotic' has only limited usefulness. Differences between the atypical agents in terms of efficacy and pharmacodynamic profiles also support this view. As tolerability differs between specific conventional and atypical drugs, we conclude that broad statements comparing the relative risk of specific adverse effects between 'atypical' and 'conventional' antipsychotics are largely meaningless; rather, comparisons should be made between specific atypical and specific conventional drugs. Adverse effects are usually dose dependent and can be influenced by patient characteristics, including age and gender. These confounding factors should be considered in clinical practice and in the interpretation of research data. Selection of an antipsychotic should be on an individual patient basis. Patients should be involved in prescribing decisions and this should involve discussion about adverse effects.