Translated from the Portuguese original. Read the original (Português)
Measuring behavior means choosing which property of it will become a number. Behavior Analysis organizes that choice into three fundamental quantities: the behavior repeats (so it can be counted, and at what rate), takes up time (so you can measure how long each episode lasts), and happens at some point in relation to other events (so you can measure how long it takes to start, or how much time passes between one response and the next). The choice is not a minor detail: the same scene turns into a very different graph depending on which dimension is measured — and a forty-minute crying episode counts as exactly the same "1" as a two-minute one.
The dimensions of behavior measurement are the quantifiable properties by which a response can be recorded. According to Johnston and Pennypacker, and as systematized in Cooper, Heron & Heward, behavior has three fundamental dimensional quantities: repeatability (occurrences can be counted — count, rate, and celeration), temporal extent (every occurrence lasts some amount of time — duration), and temporal locus (every occurrence happens at a point in time in relation to other events — latency and interresponse time). To these are added measures of the response's form and force (topography and magnitude) and derived measures, obtained by combining the previous ones — such as percentage and trials to criterion.
Why the dimension you choose changes what you see
After agreeing on what counts as the behavior — the operational definition — comes the second decision: exactly what is going to be recorded. And here there is no neutral number. Counting occurrences, timing episodes, or measuring how long it takes to start are different operations on the same scene, and each one reveals an aspect that the others hide.
A common example: a child has episodes of intense crying. If the record is by count, a two-minute episode and a forty-minute episode both go in as "1" each — and an intervention that reduced episodes from forty minutes to five minutes will show up on the graph as no improvement at all. If the record is by duration, the same intervention shows a sizable drop. Neither number is wrong; they simply answer different questions.
That is why the practical rule is to choose the dimension that corresponds to the socially relevant aspect of the complaint — what actually bothers, limits, or matters to the person and to those who live with them — and that is sensitive to the changes the intervention is expected to produce.
The three fundamental quantities
Every direct measure of behavior rests on one of these three properties. Explore each one:
Repeatability — the behavior repeats, so it can be counted
Responses occur repeatedly over time, and each occurrence can be added up. From this come three nested measures:
- Count — the simple number of occurrences. On its own, it says little: "got out of the chair 12 times" only makes sense once you know whether that happened in 10 minutes or over a whole day.
- Rate — the count divided by the observation time (responses per minute, per hour). It is what makes it possible to compare sessions of different lengths.
- Celeration — how much the rate itself is changing across days, expressed as a factor of multiplication or division per unit of time. It is the measure of learning itself, not of performance on a single day.
A note on vocabulary: in ABA manuals, rate and frequency often appear as synonyms for "count per unit of time," while part of the literature reserves "frequency" for the simple count. What resolves the ambiguity is always recording, along with the number, how long the observation lasted.
Temporal extent — every occurrence takes up a chunk of time
No behavior is instantaneous: it begins, extends, and ends. The corresponding measure is duration, recorded either as total duration (how much time, adding everything up, the person spent in that behavior during the session) or as duration per occurrence (how long each episode lasted).
It is the dimension to use when the problem — or the goal — is about how much time the behavior takes up: long crying episodes, staying engaged in a task, playtime with a peer, time to calm down after a frustration. It is also the natural choice for continuous behaviors, whose individual occurrences are hard to separate from one another — in those cases, counting "how many times" is almost arbitrary.
Temporal locus — when the occurrence happens
Beyond lasting some amount of time, each occurrence happens at a specific moment in relation to other events. Two measures explore this aspect:
- Latency — the time between the onset of an antecedent stimulus (the instruction, the request, the cue) and the start of the response. It is the measure behind the complaint "he does do it, but takes forever to start" — and also of responses that are too fast, such as answering before the question is finished.
- Interresponse time, or IRT — the interval between the end of one response and the start of the next. Short IRTs correspond to high rates; long IRTs, to low rates. It is the dimension that some differential reinforcement procedures manipulate directly, by reinforcing responses separated by a minimum interval.
Form, force, and derived measures
Not everything that matters fits into the three quantities above. Flip the cards to learn about four measures common in practice:
Percentage: the most popular measure, and the easiest to misreadIt is convenient, but it hides three traps. First, it hides the denominator: "80% correct" can mean 8 out of 10 opportunities or 4 out of 5 — and with few opportunities, each response shifts the result enormously (with 20 opportunities, one response is worth 5 percentage points; with 4, it is worth 25). That is why manuals recommend caution when interpreting percentages based on few opportunities. Second, it has a ceiling: once it reaches 100%, the graph freezes, even though the person keeps getting faster and more fluent. Third, it erases time: scoring 100% in ten seconds and in ten minutes yields the same percentage. When fluency matters, use rate.
How to choose the right dimension
In practice, the choice usually comes from the complaint itself, once it is carefully translated. Some frequent correspondences:
- "It happens too much" or "it almost never happens" → count and, whenever sessions vary in length, rate.
- "Once it starts, it doesn't stop" → duration (per occurrence, to see episodes getting shorter).
- "It takes forever to start" or "answers before I finish talking" → latency.
- "Fires off one after another, with no pause" → IRT.
- "Gets a lot wrong in activities" → percentage of opportunities, as long as the number of opportunities is sufficient and consistent.
- "Talks too loudly" or "does it the wrong way" → magnitude and topography, with an observable criterion agreed upon in advance.
Three precautions round out the decision. The first is to always record the observation time, even when the chosen measure is count — without it, nothing is comparable later. The second is to keep the same dimension from start to finish: switching the measure partway through tracking makes comparison with the baseline impossible. The third is to remember that measuring more than one dimension at the same time is legitimate and sometimes necessary — episodes can become rarer and, at the same time, longer, and only two parallel measures tell that whole story. How these numbers get collected day to day is the subject of the entry on recording and data collection.
Two different things with the same name"Dimensions" shows up in Behavior Analysis in two senses that should not be confused. Here we are dealing with the dimensions of behavior measurement — repeatability, temporal extent, temporal locus. The seven dimensions of Applied Behavior Analysis, proposed by Baer, Wolf, and Risley in 1968, are a different thing: they describe what characterizes the field itself — being applied, behavioral, analytic, technological, conceptually systematic, effective, and generalizable. One list is about how to measure a response; the other, about how a discipline defines itself.
Test your understanding
Quick quiz
Three questions to lock in the concept. You see the explanation right away.
Further reading
- Cooper, J. O.; Heron, T. E.; Heward, W. L. Applied Behavior Analysis. 3rd ed., 2020 — chapter on behavior measurement.
- Johnston, J. M.; Pennypacker, H. S. Strategies and Tactics of Behavioral Research. 3rd ed., 2009 — dimensional quantities of behavior.
- Baer, D. M.; Wolf, M. M.; Risley, T. R. "Some current dimensions of applied behavior analysis". Journal of Applied Behavior Analysis, 1968 — the seven dimensions of ABA (a different sense from the one covered in this entry).
- Bailey, J. S.; Burch, M. R. Research Methods in Applied Behavior Analysis. 2nd ed., 2018 — choosing and recording behavioral measures.
This entry is educational and does not replace assessment, diagnosis or treatment by a qualified health professional. Every person is unique — for a specific case, seek individual guidance.