An empirical examination of the utility of partial interval and momentary time sampling as measurements of behavior

This study was detailed examination of the ability and accuracy of two observational time sampling systems in the estimation of real time measures of behaviors. Nine mother-child pairs were videotaped three or four times. Twenty-two behaviors were recorded, and rate, response durations, temporal arrangement, and the percent of an observation during which they occurred were measured. The same behaviors were also measured using two time sampling procedures, momentary and partial interval. In momentary time sampling (MTS) the presence or absence of a behavior is judged at the end of sequential, identical length periods across an observation. In partial interval sampling (PI) a behavior is judged to either occur or not occur during sequential identical length periods across an observation. Seven period lengths were investigated for each type sampling, 6, 12, 18, 24, 30, 45, and 60 seconds. Sample scores were compared to actual percent total times, and the effects of rate, response duration, and percent total time on the accuracy of sample estimates was assessed. Results showed that accuracy of estimation for both sample types was effected by rate, response duration, and sample length. Shorter sample lengths were more accurate for both sampling procedures. MTS was the most accurate procedure for all sample lengths and types of behavior. PI scores introduced large errors of estimation at every sample length and error was not consistent across behaviors. Changes in rate, response durations, and interresponse times severely altered the accuracy of PI scores as estimators of actual percent total time. MTS is recommended as the preferable sampling procedure as long as sample length are 30 seconds or shorter.