BACKGROUND: The long-term evolution of neuroleptic-induced extrapyramidal syndromes (EPS) of Parkinsonism, akathisia and tardive dyskinesia (TD) is still a controversial issue worth exploring. METHOD: A total of 200 inpatients on regular typical neuroleptics (NL) and/or clozapine were assessed in 1995 with regard to the prevalence of EPS. Altogether, 83 patients could be reassessed in 2003/04 (63 had died) using the same methods. Strict definitions of EPS were used. The complete account of NL therapy the patients were prescribed between 1995 and 2003/04 (including atypical NL other than clozapine) was considered. RESULTS: The prevalences found in 1995 and 2003/04 were 17% and 29% for Parkinsonism, 14% and 14% for akathisia, and 24% and 13% for TD. There were considerable intra-individual fluctuations in EPS occurrence even when the overall prevalence rate remained the same. In intra-individual comparisons of EPS ratings on both assessments, there was a tendency for worsening of Parkinsonism to be associated with a current (2003/04) therapy with typical NL; worsening of akathisia was associated with a current therapy with atypical NL other than clozapine, amelioration of akathisia with a current therapy with clozapine; and, basically, there were no significant associations found between the changes in TD ratings and the long-term therapy with typical NL, clozapine, and other atypical NL, considering cumulative doses of all these drugs. In a multivariate analysis, there was a tendency for the long-term evolution of TD to depend on illness duration as the only variable. CONCLUSIONS: There are intra-individual fluctuations in all EPS over longer time periods. The choice of current NL therapy has an impact on Parkinsonism and akathisia. The long-term evolution of TD appears independent of NL prescriptions.