Momentary Time Sampling: The Most Practical Measurement System in ABA

Momentary time sampling (MTS) is the measurement system most likely to give you usable data from a real session, and it is chronically underused because it feels too simple to be rigorous.

The rule: divide the observation period into intervals. At the moment each interval ends, look at the learner and record whether the target behavior is occurring right then. Nothing that happened during the interval counts. Only the instant matters.

Why the instant is enough

The objection is obvious: you are ignoring almost all of the observation. How can that be accurate?

The answer is that MTS is estimating a proportion — what share of the time the behavior occupies — and repeated instantaneous samples estimate a proportion well. It is the same logic as a poll: you do not need to ask everyone to estimate the split accurately, you need enough samples taken without systematic bias.

That is exactly what MTS has going for it over its siblings. Partial interval overestimates because any occurrence scores the whole block. Whole interval underestimates because any pause disqualifies the block. MTS has no structural reason to lean either way, so across enough intervals it tends to land near the true proportion.

The caveat is honest: with few intervals or a behavior that happens to be periodic in sync with the interval, MTS can miss badly. More intervals mean a better estimate, as with any sampling method.

What MTS is good for

Behaviors defined by how much time they occupy — on-task, engagement, stereotypy, in-seat, appropriate play, sustained attention. If the clinical question is what proportion of the session, MTS answers it directly.

Group observation. MTS scales to multiple learners in a way no other method does. At the end of each interval, scan the group and score each learner once. This variant is often run as planned activity check, and it is the only realistic way to measure engagement across a classroom.

When the observer is also the instructor. This is the decisive practical advantage. A single scoring decision on a timer cue is something an RBT can do while running a session. Continuous duration recording is not, and partial interval — which requires attending to the behavior throughout every interval — is only marginally easier.

What MTS is bad for

Low-frequency behavior. If aggression occurs twice a session for ten seconds each, MTS will almost always score zero. The behavior is real, clinically urgent, and invisible to the method. Never use MTS for behaviors that are dangerous or rare — count those.

Brief, discrete behaviors. Hand raises, requests, vocalizations. These have essentially no duration, so the chance of one coinciding with an interval boundary is near zero. Count them.

Anything where you need the actual number of episodes. MTS estimates proportion of time. It cannot tell you a behavior occurred four times rather than once. If the treatment decision depends on episode count, MTS is the wrong tool.

Choosing the interval

The trade-off is precision against workload, and it is less painful than with the other interval methods because MTS asks so little per interval.

• Fifteen to thirty seconds for a single learner in a focused observation. This gives 60 to 120 samples in a thirty-minute block, which is plenty.
• One to two minutes for group observation or a full-day sample, where you are trading precision for coverage you could not otherwise get.
• Five minutes or longer only for behaviors that persist for long stretches, and with the understanding that the estimate is coarse.

A useful rule: the interval should be meaningfully shorter than the typical duration of the behavior. If on-task episodes last two minutes, a thirty-second interval samples them well. A five-minute interval will miss most of them entirely.

Whole-interval, partial-interval and MTS on the same session

It is worth seeing the three disagree. Take a learner on task in this pattern across ten thirty-second intervals — on task for the first twenty seconds of each interval, off task for the last ten.

• True proportion: about 67% on task.
• Partial interval: 100% of intervals scored — behavior occurred in every one.
• Whole interval: 0% of intervals scored — it never persisted through a full interval.
• Momentary time sampling: 0%, because at every interval boundary the learner happens to be off task.

That last line is the honest warning about MTS. When the behavior is periodic and the period aligns with your interval, MTS fails badly. The defense is to vary interval length or use a variable-interval schedule of observation, which removes the possibility of systematic alignment. In routine clinical practice this is rarely necessary — but it is the reason MTS is described as generally less biased rather than unbiased.

Running it without a second person

MTS lives or dies on the interval cue. If the RBT has to watch a clock, they will look up early or late and the sample stops being momentary.

What works: an unobtrusive prompt — a vibration or a soft tone — at the end of each interval, with the scoring control immediately available. What does not work: a timer on a phone in a bag, a paper sheet on a clipboard across the room, or an interval the staff member is asked to estimate.

This is one of the clearest cases where measurement fidelity is a tooling question rather than a training question. When the cue and the scoring control are in the same place, MTS runs at high fidelity indefinitely. When they are not, it degrades quietly into a rough guess that still gets graphed as data.

Reporting it properly

MTS data is reported as percentage of intervals scored, never as a duration or a count. Writing on task 60% of the session when you ran MTS is a misstatement — you measured 60% of intervals, which estimates but does not equal 60% of time.

The program document should record the method, the interval length, whether the schedule was fixed or variable, and the observation period. Those four facts let anyone reading the graph later know what the y-axis means.

The summary

Momentary time sampling asks one question per interval and gets a good estimate of how much time a behavior occupies. Use it for duration-like behaviors, especially with groups or when the observer is also teaching. Do not use it for rare, dangerous or brief discrete behaviors. Keep the interval well below the typical behavior duration, make the cue impossible to miss, and report it as percentage of intervals.