Discontinuous measurement

Sampling trades detail for feasibility

Continuous measurement aims to capture every detectable response during observation. Discontinuous procedures divide time into intervals or check behavior at selected moments. The lighter observation burden can make data collection feasible when staff must teach, support safety, or watch several activities.

Sampling also creates uncertainty. A scored interval cannot reveal every event inside it. The resulting percentage should be named for its procedure, observation schedule, and denominator.

The main procedures use different rules

• Partial-interval recording: score the interval when the response occurs at any time during it.
• Whole-interval recording: score the interval only when the response lasts throughout the entire interval under the written rule.
• Momentary time sampling: score whether the response is occurring at the designated instant, often the interval endpoint.

These scores are not interchangeable. “Occurred in 60% of partial intervals” describes a different observation from “occurred at 60% of sampled moments.”

Interval length changes what the sample can show

Longer intervals contain more time in which a response can occur. That tends to make partial-interval positives more likely. A whole-interval positive becomes harder when the response must persist across a longer span. Momentary samples can miss brief behavior between checkpoints.

Keep interval length stable within a data series unless the change is clearly marked. If the schedule changes, preserve the procedure version and effective date. A percentage from 10-second intervals should not be silently combined with one from five-minute intervals.

A fictional sampling example

Inez is a fictional adult who wants to participate in a community art group while retaining easy access to breaks. With Inez, the team defines active participation as using materials, looking at the current work area, or communicating about the shared project. Breaks and unrelated conversation are recorded separately and remain available.

An observer plans 20 one-minute intervals. For momentary time sampling, the observer checks at each interval endpoint. Inez is participating at 12 of the 20 scheduled moments, for 12 of 20, or 60%.

The observer misses two additional planned checks while helping secure a spilled paint container. Those checks stay outside the 20 completed-sample denominator and are reported as 20 of 22 planned checks completed, or 90.9%. They do not become zeros for Inez. The 60% describes sampled moments and gives no exact count or duration of participation.

Choose a procedure for the decision

Ask which behavioral dimension matters. If the team needs the exact number of requests, event recording may fit. If it needs the total time a health-related episode lasts, duration may fit. Sampling may be reasonable when continuous observation is impractical and the limitations remain acceptable for the decision.

Consider response duration and frequency, observer workload, setting visibility, safety, expected changes, and how much error the decision can tolerate. A high-stakes treatment or safety decision may require stronger measurement and observer-agreement evidence.

Preserve denominators and missingness

Before collection, define:

• observation period and interval length
• exact scoring rule and checkpoint timing
• eligible and ineligible intervals
• treatment of interruptions, late checks, and blocked views
• required communication, health, sensory, and mobility supports
• who records and how observer agreement will be sampled

Report positive intervals divided by valid observed intervals. Also report valid observations divided by planned observations. These separate measures keep missing data visible.

Preserve raw interval records so a reviewer can trace every reported percentage.

An observer who looks several seconds late may see a different event. Define a tolerance window or mark the sample missed. Quietly moving the checkpoint creates a new, undocumented rule.

Interpret the percentage with care

A sampled percentage can support comparison when procedure, interval, setting, supports, and observation coverage remain similar. It does not estimate exact count or duration unless a validated relation has been established for that use.

A change across phases can also reflect interval changes, observer availability, schedule, access, health, partners, or activities. Keep raw counts of intervals, missed samples, and procedural changes beside the percentage. Ask the person whether the measured outcome reflects their experience and goals.

Monitor measurement quality

Useful checks include valid samples divided by planned samples, samples taken inside the timing window, observer-agreement samples, device failures, excluded intervals by reason, and records tied to the current operational definition.

The BACB BCBA Test Content Outline, Sixth Edition covers continuous and discontinuous measurement, interval recording, time sampling, validity, reliability, and selection of representative procedures. It states examination content and does not endorse one sampling method for every case.