Nine VA Medical Centers are participating in a 2-year double-blind placebo controlled study of antioxidant treatment for tardive dyskinesia (TD) conducted by the Department of Veteran Affairs Cooperative Studies Program. One of the principal outcome measures of this study is the score derived from the instrumental assessment of upper extremity dyskinesia. Dyskinetic hand movements are quantified by assessing the variability associated with steady-state isometric force generated by the patient. In the present report, we describe the training procedures and results of a multi-center reliability assessment of this procedure. Data from nine study centers comprising 45 individual patients with six trials each (three from left hand and three from right hand) were reanalyzed by an independent investigator and the results were subjected to reliability assessment. For the statistic of interest (average coefficient of variation over trials 2 and 3 for each hand, then take the larger of these two values), we found very high intraclass correlation coefficients for reliability over all patients across sites (ICC = 0.995). We also calculated the reliability of the measures across trials within patient for each combination of hand (right, left, dominant), rater group (site, control), and trials set (all three, trials 2 and 3). For a given hand and trial set, the reliability of the site raters was similar to that of the control. This study demonstrates that instrumental measures for the assessment of dyskinesia are reliable and can be implemented in multi-center studies with minimal training.