Sometimes you can't record every instance of a behavior. The behavior may have no clear start or end, it may happen too often to count, or the person collecting data may also be teaching a group. Interval recording and time sampling let you estimate behavior in those situations, but only if you design them well.
Why it matters
These methods don't measure count or duration directly. Each one estimates behavior, and each one has its own predictable error. The interval length, the scoring rule, and the observation schedule you choose decide whether the data are useful or misleading. As a BCBA, you design the data system, train staff to use it, and make treatment decisions from it, so you need to know which way each method's error runs.
Be ready to match a method to a behavior and goal, predict whether a method will overestimate or underestimate, and spot design flaws such as intervals that are too long.
Key ideas
- All three methods divide the observation period into intervals, usually of equal length. You score each interval as yes or no, and you report the percentage of intervals scored. That percentage is not a count, a rate, or a duration.
- Whole-interval recording scores an interval only if the behavior lasts through the entire interval. It tends to underestimate how much of the time the behavior occurs. It fits continuous behaviors you want to increase, such as on-task work or cooperative play, because the underestimate makes progress look smaller, not larger.
- Partial-interval recording scores an interval if the behavior happens at any point in it. It tends to overestimate the proportion of time the behavior occurs, and it underestimates count when several responses fall in one interval. It fits behaviors you want to decrease, because the overestimate keeps you from declaring success too early.
- Momentary time sampling (MTS) scores an interval only if the behavior is happening at the instant the interval ends. Its error can go in either direction. It works well for ongoing behaviors and for watching several learners, but it misses brief or infrequent behaviors.
- Shorter intervals give better estimates. Interval recording usually uses brief intervals, often about 10 seconds. With MTS, shorter intervals mean more samples, so the estimate gets closer to the true proportion of time.
- Keep the design constant across phases. If you change the interval length or the method between baseline and intervention, the percentages aren't comparable.
Definitions
- Discontinuous measurement: any method that samples behavior instead of recording every instance, so some occurrences may be missed.
- Whole-interval recording: an interval is scored only if the behavior occurs throughout the whole interval.
- Partial-interval recording: an interval is scored if the behavior occurs at any time during the interval. Once you score an interval, you can stop watching until the next one starts.
- Momentary time sampling: an interval is scored only if the behavior is occurring at the moment the interval ends.
- PLACHECK (planned activity check): a group version of MTS. At each scheduled moment, the observer counts how many people in the group are doing the target activity and reports that number as a percentage of the people present.
- Observe-and-record intervals: a format that alternates observation intervals with short recording intervals, such as 10 seconds to observe and 5 seconds to record.
Examples from a session
- Whole-interval in a classroom. Priya, age 8, is learning to stay engaged during independent reading. Her BCBA sets 10-second intervals and scores an interval only if Priya keeps her eyes on the book for all 10 seconds. If she reaches 70% of intervals, you can be fairly confident she was engaged at least that much of the time.
- Partial-interval at home. Marcus, age 4, engages in loud vocal stereotypy during dinner. An RBT uses 15-second intervals and scores each interval where any stereotypy occurs. As the percentage drops, you know the true level is probably lower still.
- MTS in a day program. At an adult day program, a staff member checks every 2 minutes whether Ms. Okafor, 52, is taking part in the group activity at that instant. The staff member leads the activity between checks, so continuous observation wouldn't be practical.
- PLACHECK in a group. A teacher scans a table of six students at the end of each 5-minute block and writes down how many are working on the assigned task.
- Non-example: timing every episode. In a clinic, an RBT starts a stopwatch each time Tomás, age 11, leaves his seat and stops it when he sits back down. That is continuous duration recording, not interval recording or time sampling: nothing is sampled.
Common mistakes / exam traps
- Reporting interval data as a count or duration. "Scored in 40% of intervals" does not mean the behavior happened 40% of the time, and it doesn't tell you how many times it happened.
- Using intervals that are too long. Long partial intervals inflate the estimate, and long whole intervals rarely get scored unless the behavior is very persistent.
- Changing interval length between phases. A change in the percentage might come from the design change rather than a change in behavior.
Exam trap
Whole-interval vs. partial-interval recording: match the error to the goal. When the goal is to decrease a behavior, partial-interval recording is the usual pick, because it overestimates. When the goal is to increase a behavior, whole-interval recording is the usual pick, because it underestimates. Either way, the error works against you, so any improvement you see is probably real.
Exam trap
Momentary time sampling vs. partial-interval recording: the difference is when you look. Partial-interval scores behavior at any point in the interval; MTS scores only what is happening at the instant the interval ends. So MTS misses brief, low-rate behavior: a slap that happens three times an hour will almost never land on a sampling moment. Count responses like that with event recording.
Quick check
A BCBA watches a 5-minute video of Ava, age 6, in 10-second intervals and scores an interval only if Ava is rocking at the exact moment it ends. Which procedure is this?
Related topics
- Continuous vs. discontinuous measurement (C.5)
- Measuring occurrence: count, rate, and percentage (C.3)
- Measuring time-based dimensions: duration, latency, and IRT (C.4)
- Judging the validity and reliability of measurement (C.8)
- Choosing a measure that captures the critical dimension within real-world limits (C.9)