Data Collection and Graphing is content area A of the RBT exam. It accounts for 13 of the 75 scored questions and has 8 tasks. In the previous outline this area was called Measurement.
Your supervisor chooses the measurement system for each target. Your job is to use it exactly as written, every session, and to say so when something gets in the way.
Continuous measurement
Task A.1 covers continuous measurement, which means you record every instance of the behavior during the observation. Nothing is sampled, so these methods give the most complete picture.
| Measure | What you record | Session example |
|---|---|---|
| Frequency (count) | How many times the behavior occurred | Marco raised his hand 6 times |
| Rate | Count divided by observation time | 6 hand raises in 30 minutes, or 0.2 per minute |
| Duration | How long the behavior lasted, from start to end | A tantrum lasted 4 minutes 10 seconds |
| Latency | Time from the instruction (or other cue) to the start of the response | “Put on your shoes” was given, and he started 12 seconds later |
| Interresponse time (IRT) | Time from the end of one response to the start of the next | 40 seconds passed between two bites of food |
Frequency works for behaviors with a clear start and end that do not happen too fast to count. Duration suits behaviors that last a while or vary in length.
The common mistake is mixing up latency and IRT. Latency is measured from a cue to a response. IRT is measured between two responses of the same behavior. A second mistake is comparing raw counts from sessions of different lengths. Ten responses in a 20-minute session and ten in a 60-minute session are not the same level of behavior, which is why rate is used.
Discontinuous measurement
Task A.2 covers discontinuous measurement. Here the observation is divided into intervals and you record a sample of the behavior, not every instance. These methods are easier to run while teaching, but they only estimate the behavior.
| Method | You mark the interval when | Tends to | Often used for |
|---|---|---|---|
| Partial interval recording | The behavior occurred at any point in the interval | Overestimate how much of the time the behavior occurred | Behaviors to reduce |
| Whole interval recording | The behavior occurred for the entire interval | Underestimate how much of the time the behavior occurred | Behaviors to increase, such as staying on task |
| Momentary time sampling | The behavior is occurring at the moment the interval ends | Give an estimate, with more error as intervals get longer | Situations where you cannot watch continuously |
Example: with 10-second partial intervals, a single one-second scream is enough to score that interval. With whole interval recording, a learner who works for 9 of the 10 seconds does not get the interval scored.
Exam questions on this task usually describe what the observer did and ask you to name the method. Look for the key phrase: “at any time during” points to partial interval, “throughout the entire” points to whole interval, and “at the end of” or “when the timer sounded” points to momentary time sampling. Interval data are reported as the percentage of intervals with the behavior. They do not tell you how many times the behavior occurred.
Permanent product recording
Task A.3 covers permanent product recording. You measure the lasting result of a behavior after it has happened, so you do not have to watch the behavior itself.
Examples are the number of math problems completed on a worksheet, the number of dishes put away, or the number of items left on the floor after clean-up time.
This method only fits behaviors that leave a result you can count or inspect. The usual limitation is that you may not know who produced the result or how much help was given. If a caregiver finished part of the worksheet, the product no longer shows what the client did alone. Note that kind of thing and tell your supervisor.
Entering data and graphing
Task A.4 covers getting session data into the record and onto the graph. Record data during the session, as the behavior happens or immediately after each trial. Do not reconstruct it from memory at the end of the day.
Most ABA data are shown on a line graph. Know its parts:
- The horizontal axis (x-axis) shows time, usually sessions or dates.
- The vertical axis (y-axis) shows the measure of behavior, such as rate or percentage correct.
- Each data point is one session’s value. Points within a condition are joined by a line.
- A vertical phase change line marks a change in conditions, for example from baseline to intervention. Data points are not connected across it.
- Condition labels name each phase.
A common exam item asks what to do after a session with missing or doubtful data. The answer is never to estimate or fill in a number. Record what you actually collected, note the problem and tell your supervisor.
Observable and measurable terms
Task A.5 covers describing behavior and the environment in terms that someone else could see and count. A good description lets two observers agree on whether the behavior happened.
| Not observable | Observable and measurable |
|---|---|
| “He was angry and defiant” | “He threw the pencil and said no when given the worksheet” |
| “She was anxious in the store” | “She covered her ears and left the aisle twice in 10 minutes” |
| “The room was chaotic” | “Four other children were in the room and the television was on” |
Avoid words that describe inner states or guess at intent, such as frustrated, lazy, manipulative or “wanted to.” On the exam, the best description is the one that states what the person did or said, with no interpretation. The same standard applies to your session notes, which are covered in documentation and reporting.
Calculating and summarizing data
Task A.6 covers the basic calculations you use to summarize a session. Expect a few questions with simple arithmetic.
- Rate: count divided by time. 24 responses in 8 minutes is 3 responses per minute.
- Mean duration: total duration divided by the number of occurrences. Three tantrums of 2, 4 and 6 minutes total 12 minutes, so the mean is 4 minutes.
- Percentage of trials: correct responses divided by total trials, times 100. 8 correct out of 10 trials is 80%.
- Percentage of intervals: intervals scored with the behavior divided by total intervals, times 100. 15 of 60 intervals is 25%.
Read the question for the unit it asks for. People lose points by giving total duration when mean duration is asked, or by dividing time by count when calculating rate. Check that the time unit matches too (per minute or per hour).
Trends in graphed data
Task A.7 covers reading a graph. Three features describe the data path:
- Trend is the direction of the data: increasing, decreasing or flat (no trend).
- Level is how high or low the data sit on the vertical axis.
- Variability is how much the points bounce around. Stable data fall within a narrow range.
Whether a trend is good news depends on the target. An increasing trend is the aim for a skill and a concern for a behavior targeted for reduction.
As an RBT you can describe what the graph shows, for example “correct responding has gone down over the last four sessions.” Deciding what that means for the program and whether to change it is your supervisor’s job. If the answer options include the RBT changing the procedure because of a trend, that option is wrong.
Unreliable data, poor fidelity
Task A.8 covers why accuracy matters. Reliable data means the record matches what actually happened and would match another trained observer’s record. Procedural fidelity (also called treatment integrity) means the plan is carried out as written.
Supervisors make treatment decisions from your data. If the data are wrong, the risks include:
- An effective intervention is stopped or changed because it looks like it is failing.
- An ineffective intervention continues because it looks like it is working.
- A client stays in services longer than needed, or loses services too early.
- Records submitted to funders are inaccurate.
Poor fidelity creates a similar problem. If the plan was not followed, the data describe some other procedure, and nobody can tell whether the plan itself works.
Common causes are unclear behavior definitions, recording from memory, observer drift (your own use of the definition slowly changing over time) and expecting a certain result. If you notice any of these, or you were unable to collect data as planned, report it. The exam consistently favors honesty about a data problem over a tidy data sheet.
What to remember
- Continuous measures record every instance: frequency, rate, duration, latency and IRT.
- Latency runs from cue to response. IRT runs between two responses.
- Partial interval tends to overestimate, whole interval tends to underestimate, and momentary time sampling checks only the instant the interval ends.
- Permanent product recording measures the result of the behavior, not the behavior as it happens.
- Rate is count divided by time. Mean duration is total duration divided by the number of occurrences.
- On a line graph, time is on the x-axis, behavior is on the y-axis, and data points are not connected across phase change lines.
- Describe trend, level and variability. Leave treatment decisions to your supervisor.
- Never estimate or fill in missing data. Record what happened and report the problem.
Test yourself with the data collection and graphing quiz, then continue with behavior assessment.
Last checked against the BACB's documents in November 2026.