Disruptive vocalization (DV) is both a prevalent and disturbing problem in nursing homes. We developed a computer-assisted data collection system for real-time observation and recording of DV and various environmental contextual factors. Both frequency and duration of DV were recorded for 11 residents along with their location in the nursing home, their activity, environmental sound, the social environment, and whether or not the resident was physically restrained. The actual time of all events was also recorded. Measures of cognitive and ADL status were administered. The average occurrence of DV was 22 per hour and the average duration per occurrence was 26 seconds. The results show a significant upward linear trend in the occurrence of DV across the day. This is consistent with the "sundowning" hypothesis. A Cox Proportional Hazards Regression model indicates that another person present in the setting (p = .004) and resident presence at the nursing home hairdresser (p = .07) were associated with shorter duration episodes of DV. Correlational analyses indicate that both higher frequency and longer duration DV are related to greater cognitive impairment, and higher frequency DV is related to greater ADL impairment. We conclude that this computer-assisted real-time observational system is a useful and promising tool for studying disruptive behavior in its environmental context.