An investigation of the accuracy of momentary time sampling procedures with time series data

Momentary time sampling is a procedure in which one records whether or not responding is occurring at the end of an interval. As a sampling procedure, it presents both the advantage of ease and the disadvantage of imperfectly representing data collected through continuous recording. The present study was conducted to determine the interval sizes at which momentary time sampling accurately represents continuous recording. Observations using an event-recorder produced permanent records of five behaviours which were sampled five times, each time at intervals of 10, 20, 30, 60, 120 and 240 s, for each day's data with starting delays of 0, 12, 24, 36 and 48 s. These procedures allowed comparison of the results of this study with previous studies which had grouped their results across sessions, and provided repeated samples of each behaviour which allowed examination of the accuracy of momentary time sampling on a session-by-session basis. When the data were grouped as in previous studies, results indicated that (a) the 10, 20, 30, and 60 s intervals accurately estimated the average duration of behaviours, (b) the 120 s interval was accurate when the behaviour occurred more than 10 per cent of the time, and (c) the 240 s interval was accurate when the behaviour occurred more than 20 per cent of the time. When the data were not averaged, however, the results indicated that the 60, 120, and 240 s intervals were inaccurate.