Ever since Liepmann's original descriptions at the beginning of the century, apraxia has usually been attributed to damage confined to the cerebral cortex andtex cortico-cortical connecting pathways. However, there have been suggestions that apraxia can be due to deep subcortical lesions, which raises the question as to whether damage to the basal ganglia or thalamus can cause apraxia. We therefore analysed 82 cases of such ‘deep’ apraxias reported in the literature. These reports consisted of a small number (n = 9) of cases studied neuropathologically, and a much larger group (n = 73) in which CT or MRI was used to identify the size and extent of the lesion. The reports were subdivided into (i) those with small isolated lesions which involved nuclei of the basal ganglia or thalamus only, and not extending to involve periventricular or peristriatal white matter; (ii) those with large lesions which involved two or more of the nuclei, or one or more of these deep structures plus damage to closely adjacent areas including the internal capsule, periventricular or peristriatal white matter; and (iii) lesions sparing basal ganglia and thalamus but involving adjacent white matter. The main conclusions to be drawn from this metanalysis are that lesions confined to the basal ganglia (putamen, caudate nucleus and globus pallidus) rarely, if ever, cause apraxia. Lesions affecting the lenticular nucleus or putamen nearly always intruded into the adjacent lateral white matter to involve association fibres, in particular those of the superior longitudinal fasciculus and frontostriatal connections. Apraxia occurred with deep lesions of the basal ganglia apparently sparing white matter in only eight out of the 82 cases. Apraxia was most commonly seen when there were lesions in the lenticular nucleus or putamen (58 out of 72 cases) with additional involvement of capsular, and particularly of periventricular or peristriatal, white matter. Lesions of the globus pallidus (no cases) or caudate nucleus (three cases) rarely caused apraxia. The caudate lesions also had white matter involvement. Indeed, involvement of periventricular or peristriatal white matter alone caused apraxia. The vast majority of cases described with apraxia associated with deep lesions were in the left, dominant hemisphere. Ideomotor apraxia was described in most reports (72 out of 82 cases). Orofacial apraxia was less common (37 cases), usually with ideomotor apraxia. Ideational apraxia was rare (five cases), all with ideomotor apraxia. Apraxia was either bilateral or involved the left hand if there was a right hemiparesis, in those cases where descriptions were available. Lesions of the thalamus can sometimes cause apraxia (26 cases), even if there is no apparent involvement of white matter (12 cases). Small lesions confined to the thalamus can also sometimes cause apraxia (eight cases). The role of the thalamus in higher order motor control and apraxia remains to be determined. It is suggested that the term limb-kinetic apraxia should be retained to describe motor deficits in planning ‘what to do’, ‘how to do it’ and ‘when to do it’; decisions which appear to involve activation of a complex distributed network of dorsolateral prefrontal cortex, supplementary motor areas, anterior cingulate regions and lateral premotor cortex. Such deficits need to be quantified. If they are present in patients with basal ganglia disease, over and above classical akinesia, bradykinesia and hypokinesia, then such patients could be said to exhibit limbkinetic apraxia.