Partial Interval Recording Data Sheet

When to use it

Partial interval recording divides an observation into equal intervals. You score an interval if the behavior occurs at any point during it, even for a moment. You do not count how many times it happened, and you do not time how long it lasted.

Use it when:

• The behavior happens too often, or too quickly, to count accurately (for example, vocal stereotypy or rapid hand movements).
• The behavior has unclear boundaries between instances.
• You want to track a reduction target. Partial interval recording tends to overestimate how much a behavior occurs, especially for brief, frequent behavior. For a reduction goal, that means the data errs on the side of caution.

Avoid it for increase targets, such as on-task behavior. Because partial interval scores an interval for even a brief occurrence, it can make a skill look stronger than it is. Whole interval recording, which requires the behavior to last the entire interval and tends to underestimate, is usually a better fit for behavior you want to increase.

Partial interval data is an estimate of occurrence. Do not report it as frequency or duration.

How to fill it in

1. Write the operational definition at the top of the sheet.
2. Pick an interval length and keep it the same across sessions. Shorter intervals (about 5 to 15 seconds) usually give a closer estimate. Longer intervals are easier to run but less accurate. The grid below uses 10-second intervals. Practice varies, so change the column headers if your program uses a different length.
3. Use a timer that signals each interval, such as a vibrating interval timer or an audio cue in an earpiece. Don’t try to watch a clock and the client at the same time.
4. During each interval, watch for the behavior. If it occurs at any point, mark + in that interval’s box. If it doesn’t occur, mark -. Once an interval is scored +, you don’t need to watch for more instances in that interval.
5. Some protocols use observe-then-record intervals (for example, observe 10 seconds, record 5 seconds). If yours does, note it on the sheet and only score behavior seen during the observe portion.
6. Fill every box. A blank box means “not observed,” which is different from “did not occur.” Mark intervals you missed with X and leave them out of the calculation.

Common mistakes:

• Scoring + only when the behavior happened “a lot” in the interval. Any occurrence counts.
• Changing interval length between sessions, which changes how much the data overestimates.
• Using partial interval data to make claims about rate (“she flapped 20 times”).
• Leaving boxes blank instead of marking - or X.

The sheet

Partial Interval Recording Data Sheet https://streamaba.com/resources/data-sheets/partial-interval-recording

Client:Date:Observer:Setting / activity:

Target behavior and definition:

Interval length: 10 secondsStart time:Stop time:

Code: + occurred at any point in the interval | - did not occur | X not observed

Total intervals scored +:Total intervals observed (excluding X):Percent of intervals:

Summarizing the data

Percent of intervals = number of intervals scored + divided by number of intervals observed, times 100.

Worked example: a 5-minute observation with 10-second intervals gives 30 intervals. The behavior was scored + in 12 intervals, and no intervals were missed.

• 12 / 30 x 100 = 40% of intervals

Report this as “vocal stereotypy occurred in 40% of 10-second intervals,” not “40% of the time.” Because partial interval overestimates, the behavior was probably present for less than 40% of the actual time.

If two intervals were missed (X), the denominator becomes 28: 12 / 28 x 100 = 42.9%.

Tips

• Keep observation length, interval length, and activity consistent across sessions. That makes your trend lines meaningful even though the absolute numbers are estimates.
• For IOA, interval-by-interval agreement is standard. For low-rate behavior, scored-interval IOA gives a more conservative agreement figure, and for high-rate behavior, unscored-interval IOA does.
• If you need to track several behaviors at once, give each behavior its own row per minute or its own sheet. Tracking too many behaviors on one grid lowers accuracy.
• If the behavior turns out to be clearly countable, switch to frequency recording. You’ll get more precise data.