The big idea
There is no single best measure of behavior. The right one depends on two questions: what about this behavior matters, and what can realistically be measured in this setting?
The first question is about the critical dimension, the property of the behavior that the concern or the goal is really about. The second is about environmental constraints: who is available to observe, when and where the behavior happens, how many behaviors or people must be tracked, and what tools are available.
A measure that is perfect on paper but never gets collected produces no data. A measure that is easy but misses the critical dimension produces misleading data. The goal is representative data: a fair picture of the behavior as it actually occurs.
Breaking it down
Step 1: Find the critical dimension
| If the concern is… | Dimension | Measures |
|---|---|---|
| How often | Repeatability | Count, rate, percentage (C.3) |
| How long | Temporal extent | Total duration, duration per occurrence (C.4) |
| How soon after a cue, or how fast the pace | Temporal locus | Latency, interresponse time (C.4) |
| How hard or how loud | Magnitude | Force, volume in decibels, size of the effect |
| What it looks like | Topography | Checklist of correct form, steps of a task analysis |
| What it leaves behind | Outcome | Permanent products (C.2) |
Magnitude is the force or intensity of a response. Topography is its physical form. Both are measured less often than count or duration, but sometimes they are the whole concern: a worker who speaks too loudly, a learner whose handwriting cannot be read.
Some behaviors need two measures. Tantrums often need count and duration per occurrence. Reading often needs both rate and accuracy.
Step 2: Account for the setting
- Observer availability. A teacher with 25 students cannot keep a continuous tally on one student all lesson, so momentary time sampling or a permanent product may be more realistic. An aide assigned to one learner can handle continuous recording.
- When and where the behavior happens. Schedule observation for the times and places the behavior occurs. A scatterplot (C.10) or an interview can show when that is.
- The behavior’s form and rate. Discrete behavior at a manageable rate suits event recording. Frequent behavior without clear starts and stops suits interval methods (C.6). Rare, brief behavior needs continuous recording across the whole period, because sampling will miss it.
- Number of targets. Several behaviors or many people may call for interval methods, PLACHECK or rotating observation.
- Intrusiveness. Observers and cameras can change behavior (C.2).
Step 3: Check that the data will be representative
- Observe often enough, and long enough, to capture the behavior’s usual pattern.
- Make sure the measure can show change. A probe with a 100% ceiling cannot show growth in fluency, and observing only when behavior is rare creates a floor.
- Pick a discontinuous method whose bias suits the goal (C.6).
- Pilot the system and check interobserver agreement before baseline (C.8).
A result that reflects how the behavior was measured rather than the behavior itself is a measurement artifact, and it threatens the validity of everything built on the data.
Seeing it in practice
| Case | Critical dimension | Constraint | Measure |
|---|---|---|---|
| Kai, 6, hits peers two or three times a day at school. | Repeatability | Brief and rare; an aide is with him all day. | Event recording, reported as hits per school hour |
| Grace, 15, waits a long time before starting assigned work. | Temporal locus | The teacher gives instructions to the whole class. | Latency from instruction to first written word, on three instructions per class |
| Ms. Delgado, 50, speaks so loudly that coworkers complain. | Magnitude | Busy, noisy workplace | Sound-meter readings during set conversations |
| Noor, 3, hums almost constantly with no clear breaks. | Temporal extent | Clinic session with several programs running | Momentary time sampling at 30-second intervals |
| Liam, 13, does homework while his parent is at work. | Outcome | No adult present | Problems completed and correct, scored from the worksheet |
| A class of 22 during independent work | Group engagement | One teacher | PLACHECK every 5 minutes |
Common mix-ups
- Choosing the data sheet before the dimension. Start with the concern, not the form you already have.
- Counting everything. Count misses problems of duration, latency and intensity.
- Measuring at convenient times. Data from the wrong time of day can misrepresent the behavior completely.
- Ignoring feasibility. The most precise procedure is worthless if staff cannot carry it out. Choose the most accurate procedure the setting can sustain.
- Insensitive measures. If the measure cannot move, it cannot show improvement.
Exam tip: In a scenario, mark the words that describe the concern (how often, how long, how quickly, how hard) and the words that describe the setting (one teacher, many students, brief, constant, nobody present). The best answer fits both.
Quick recap
- Choose a measure by the critical dimension first, then adjust for constraints.
- Repeatability, temporal extent, temporal locus, magnitude, topography and outcome each point to different measures.
- Constraints include observer time, the timing and location of the behavior, its form and rate, and the number of targets.
- Representative data sample the behavior fairly and can show change.
- Pilot the system and check agreement before baseline.
Further reading
- Cooper, J. O., Heron, T. E., & Heward, W. L. (2020). Applied behavior analysis (3rd ed.), chapters 4 and 5.
- LeBlanc, L. A., Raetz, P. B., Sellers, T. P., & Carr, J. E. (2016). A proposed model for selecting measurement procedures for the assessment and treatment of problem behavior. Behavior Analysis in Practice, 9(1), 77-83.
Key terms
- Repeatability (Countability)
- The property that instances of a behavior can occur again and again over time, which makes the behavior countable. Count, rate and celeration are measures of repeatability.
- Temporal extent
- The property that every instance of behavior takes up some amount of time, from its start to its end. Duration is the measure of temporal extent.
- Temporal locus
- The property that every instance of behavior occurs at a particular point in time relative to other events. Latency and interresponse time are measures of temporal locus.
- Magnitude (Intensity, Force)
- The force or intensity with which a response is emitted, such as how loudly a person speaks or how hard they press. It is measured when strength, rather than count or duration, is the concern.
- Topography
- The physical form or shape of a response: what it looks like, including the body parts involved, the movement and where it occurs. The force of a response is usually treated as a separate dimension, magnitude. Responses with different topographies can serve the same function, and the same topography can serve different functions.
- Measurement artifact
- A result that appears in the data only because of how the behavior was measured, not because of the behavior itself. Common sources are discontinuous methods that distort the behavior, poorly scheduled observation and measurement scales too limited to show change.
Check yourself
4 questions on C.9. Choose an answer to see whether you're right and why.
Show the answer
Answer: A. 'Go on forever' is about how long each episode lasts, so duration per occurrence is the critical dimension. A weekly count would show one or two and miss the problem entirely. Partial-interval recording all day would be costly and would overstate the time involved.
Show the answer
Answer: B. Ms. Torres can only look up now and then, and momentary time sampling needs just one glance at set moments while giving a fair estimate of a long-lasting behavior like engagement. Continuous duration and event recording require watching Ivy the whole time. Whole-interval recording with 10-second intervals also demands constant attention.
Show the answer
Answer: A. The finished boxes are a permanent product, and dividing the count by the hours worked gives a rate that reflects speed. The other three options all require the coach to watch during the shift, which the setting does not allow. The product is valid as long as only Hamid packs at that station.
Show the answer
Answer: D. The data reflect when staff were available, not when the behavior happens, so they will badly underestimate self-injury. That is a measurement artifact and a threat to validity. Drift, agreement and definitions concern how behavior is scored, not whether observation happens when the behavior occurs.
Adds a step to your study record. Saved in this browser only.