Intra-patient variability in the movements of tardive dyskinesia was examined through the analyses of 8-min frequency counts collected over a period of 11 weeks for six chronic schizophrenic patients. Weekly observation segments of 8 min, 4 min, 2 min, 1 min and 30 s showed considerable variation both across weeks and within sessions. Variations were of sufficient magnitude to contribute to the possibility of false negative tardive dyskinesia assessments and false positive treatment outcome designations. Measurement procedures taking this variability into account are urged for the study of tardive dyskinesia. Additionally, independently obtained rating scale scores showed no association to the collected frequency count data, suggesting fundamental differences between these two assessment procedures.