An evaluation of time-sample measures of behavior.

Researchers compared three common time-sampling methods. They used a stopwatch and paper data sheets in a lab setting. Adults performed a simple repetitive task while observers recorded behavior using whole-interval, partial-interval, and momentary-time sampling. Each method was tested with 5-second, 10-second, and 20-second observation windows.

Whole-interval sampling always underestimated how long the behavior really lasted. Partial-interval sampling always overestimated. Momentary sampling gave mixed errors—sometimes high, sometimes low, but balanced overall. Shorter intervals (5 seconds) cut the size of the errors in half compared with 20-second intervals.

The findings echo Iwata et al. (1990), who also used single-case lab methods to study schedule effects. Both papers show that small procedural choices—interval length or schedule type—change the data you see. McLennan et al. (2008) later showed that signaling reinforcer ratios within a session gives cleaner data; the 1975 paper tells us that picking the right sampling method is just as critical for clean data. No direct contradictions appear; the later studies simply extend the theme that measurement details matter.

When you track tantrums or on-task behavior, your sampling rule becomes part of the treatment decision. If you use whole-interval recording, you will miss brief episodes and might withdraw reinforcement too soon. If you use partial-interval, you will over-count and might keep an intervention longer than needed. Momentary sampling gives the least biased picture, especially with 5- or 10-second windows. Pick one method, stick with it, and report it so future teams know the limits of your data.