Partial interval, whole interval and momentary time sampling

Measurement questions look simple and cost a lot of marks. The trap is that each interval method gives a different picture of the same behavior, and you need to know which way each one leans.

The BACB's 6th edition outline has a task on distinguishing continuous from discontinuous measurement (C.5) and a task on designing and applying discontinuous procedures such as interval recording and time sampling (C.6).

Continuous vs discontinuous

• Continuous measurement can detect every instance of the behavior across the whole observation. Count, rate, duration, latency and interresponse time are continuous.
• Discontinuous measurement samples the behavior. Some occurrences can be missed, so the result is an estimate, reported as a percentage of intervals, not a count or a duration.

Discontinuous methods exist because continuous recording isn't always practical, for example when one teacher is also running a class.

The three interval methods

Divide the observation into equal intervals, say 10 seconds each. Then:

Powell, Martindale and Kulp (1975) compared all three against continuous recording of the same behavior. Whole-interval consistently underestimated, partial-interval consistently overestimated, and MTS erred in both directions about equally often. Shorter intervals, or more frequent samples, brought each method closer to the continuous measure.

Why choose the method that leans against you

A useful rule: pick the method whose bias works against the change you hope to see.

• When reducing a problem behavior, partial-interval overstates it. If the data still show a drop, you can trust it.
• When increasing on-task behavior, whole-interval understates it. If the data still show a rise, the improvement is real.

Worked examples

• A teacher wants to reduce brief calling out. She marks each 15-second interval in which any call-out happens. Partial-interval.
• A therapist wants to build sustained play. He scores a 10-second interval only if the child played for all 10 seconds. Whole-interval.
• A classroom aide sets a timer for every 5 minutes and, when it beeps, notes whether each student is working. Momentary time sampling. When this is done for a whole group at once, it's often called a planned activity check (PLA-check).
• An aide tallies every hand raise across a 30-minute lesson. Not an interval method. That's continuous event recording.

Which method matches which dimension?

Interval data are often compared with the continuous measure they stand in for. Meany-Daboul and colleagues (2007) compared MTS and partial-interval recording on treatment data. Conclusions drawn from both often matched those drawn from continuous data, but MTS was slightly more likely to match duration-based conclusions, and partial-interval slightly more likely to match frequency-based ones.

A practical takeaway: for behaviors where duration matters, MTS is often a reasonable estimate. For brief, discrete behaviors, partial-interval is more sensitive, but it still doesn't give a true count.

Common traps

• Converting percentages into counts. "40% of intervals" isn't 40 occurrences, and it isn't 40% of the session's minutes.
• Mixing up MTS and partial-interval. MTS looks only at the moment the interval ends. Partial-interval needs you to watch the whole interval.
• Using MTS for brief or rare behavior. It can miss short behaviors entirely if they don't happen at the moment of observation.
• Forgetting interval length matters. Long intervals make every method less accurate.
• Calling permanent product recording discontinuous. It measures outcomes left behind by behavior. It's treated as its own method.

Practise it

For each measurement scenario, ask: "Do they watch the whole interval or just a moment? And does any occurrence count, or must it fill the interval?" Those two questions name the method every time.

The lessons on C.5: continuous and discontinuous measurement and C.6: interval recording and time sampling go through each method with practice items. Measurement is about 12% of the exam, and the free diagnostic shows how you're doing there.

FAQ

Does partial-interval recording overestimate or underestimate? It tends to overestimate how much of the time a behavior occurred, because any occurrence scores the whole interval.

Does whole-interval recording overestimate or underestimate? It tends to underestimate, because the behavior must last the entire interval to be scored.

What is the difference between momentary time sampling and partial-interval recording? MTS records only whether the behavior is happening at the end of each interval. Partial-interval recording scores an interval if the behavior happens at any point in it.

Is interval recording continuous or discontinuous? Discontinuous. It samples behavior and gives an estimate, reported as a percentage of intervals.

Sources

• BACB, BCBA Test Content Outline (6th ed.): https://www.bacb.com/wp-content/bcba-outline-6thEd (tasks C.5 and C.6)
• Powell, J., Martindale, A., & Kulp, S. (1975). An evaluation of time-sample measures of behavior. Journal of Applied Behavior Analysis, 8(4), 463–469: https://pmc.ncbi.nlm.nih.gov/articles/PMC1311883/ (whole-interval underestimates, partial-interval overestimates, MTS errs both ways)
• Meany-Daboul, M. G., Roscoe, E. M., Bourret, J. C., & Ahearn, W. H. (2007). A comparison of momentary time sampling and partial-interval recording for evaluating functional relations. Journal of Applied Behavior Analysis, 40(3), 501–514: https://pmc.ncbi.nlm.nih.gov/articles/PMC1986695/ (agreement of MTS with duration and PIR with frequency)