Further reappraisal of momentary time sampling and partial-interval recording.
Momentary time sampling can badly misstate duration when behavior is chopped into quick bursts.
01Research in Context
What this study did
The authors wrote a theoretical paper. They looked at momentary time sampling (MTS) when behavior is broken into many short bursts.
They warned that MTS can badly misstate how long the behavior really lasts.
What they found
If the behavior you track happens in quick scattered bursts, MTS can dramatically misstate duration.
The paper tells you to treat MTS results with caution in this case.
How this fits with other research
Neef et al. (1986) ran an earlier test and said MTS beats partial-interval recording (PIR) for true duration. The 1993 paper keeps the same topic but adds a warning flag for fragmented behavior.
Decasper et al. (1977) showed MTS was already better than old interval methods. The new paper keeps that view but says even MTS can fail if the behavior is chopped into tiny bits.
Gardenier et al. (2004) later tested kids with autism and found PIR wildly inflates stereotypy duration, while MTS gives smaller, two-way errors. This real-data study backs up the 1993 caution about PIR and supports careful use of MTS.
Why it matters
Next time you set 15-second MTS to log self-talk or hand flaps, picture the behavior split into two-second bursts. The paper says your data could hide the real duration. Run a quick check: count continuous seconds with a stopwatch for one session. If the bursts are short and scattered, either shorten the MTS interval or switch to continuous timing for that client.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Test one behavior for true burst length with a stopwatch; adjust MTS interval or record continuously if bursts are shorter than your window.
02At a glance
03Original abstract
We found the comments of Suen, Ary, and Covalt (1991) on our paper (Harrop & On some im- portant points, however, we disagree with their analysis. The comment that "for any given true total duration, MTS can be expected to record behavior in the same number of intervals whether that behavior is massed in one bout or is distributed among many bouts" (p. 804) is potentially mis- leading. Consider, for example, a behavior, perhaps a student banging a desk lid in school, that occurs for a total duration of 6 min in a 60-min lesson and that is observed using momentary time samples (MTS) undertaken every 30 s. If the 6 min of behavior occur in one bout, the MTS record would show 12 (perhaps 13) instances recorded, or 12/ 120 of the observations made, which is the correct proportion. If, on the other hand, the behavior occurred in bouts of 3 s each, then the MTS record could conceivably show any number of instances of behavior occurring from 0 to 120 (i.e., from 0/120 to 120/120). That on average, the number recorded in these circumstances would be 12/120 (as may be demonstrated by sampling theory) is, in practice, oflittle use or interest to the practitioner who relies on single periods of observation. Herein
Journal of applied behavior analysis, 1993 · doi:10.1901/jaba.1993.26-277