Integrative computer usage monitors have become widely used in epidemiologic studies to investigate the exposure-response relationship of computer-related musculoskeletal disorders. These software programs typically estimate the exposure duration of computer use by summing precisely recorded durations of input device activities and durations of inactivity periods shorter than a predetermined activity duration cutoff value, usually 30 or 60 sec. The goal of this study was to systematically compare the validity of a wide range of cutoff values. Computer use activity of 20 office workers was observed for 4 consecutive hours using both a video camera and a usage monitor. Video recordings from the camera were analyzed using specific observational criteria to determine computer use duration. This observed duration then served as the reference and was compared with 238 estimates of computer use duration calculated from the usage monitor data using activity duration cutoffs ranging from 3 to 240 sec in 1-sec increments. Estimates calculated with cutoffs ranging from 28 to 60 sec were highly correlated with the observed duration (Spearman's correlation 0.87 to 0.92) and had nearly ideal linear relationships with the observed duration (slopes and r-squares close to one, and intercepts close to zero). For the same range of cutoff values, when the observed and estimated durations were compared for dichotomous exposure classification across participants, minimal exposure misclassification was observed. It is concluded that activity duration cutoffs ranging from 28 to 60 sec provided unbiased estimates of computer use duration.