RBT Measurement Cheat Sheet: Every Data Type Compared

Measurement is roughly an eighth of the RBT exam, and it is the most reliably scoreable domain on the entire blueprint. There is no judgment involved and no ambiguity. Either you know that partial interval overestimates duration, or you do not.

This is the reference page. Learn the tables and this domain stops costing you points.

Continuous versus discontinuous

Continuous measurement records every instance of the behavior during the observation period. It is accurate and labor-intensive.

Discontinuous measurement samples the behavior — you record whether it occurred during defined intervals rather than counting every occurrence. It is efficient and introduces known bias.

The trade-off is the whole story: continuous when you need accuracy and can spare the attention, discontinuous when you are teaching and observing at the same time.

The continuous methods

Two discriminations the exam tests constantly:

• Latency versus duration. Latency is the gap between the antecedent and the start of the behavior. Duration is how long it runs once it starts. "How long after the instruction did she begin?" is latency.
• Rate versus frequency. If sessions differ in length, raw counts are not comparable, so rate is the correct choice. This is a very common stem.

Also worth knowing:

• Percentage — a proportion of opportunities (8 correct out of 10 trials = 80%). Useful when the number of opportunities varies.
• Trials to criterion — how many trials until the learner hits the mastery standard. A measure of learning efficiency, not of the behavior itself.

The discontinuous methods

This is the table to memorize.

The logic is worth understanding, not just memorizing:

• Whole interval requires the behavior to be present the entire time, so any brief pause makes the whole interval score as zero. The result reads lower than reality — hence underestimate.
• Partial interval credits a full interval for one second of behavior. The result reads higher than reality — hence overestimate.
• Momentary time sampling checks a single instant, so errors run in both directions and roughly cancel out.

A memory hook that survives exam pressure: Whole = Way less. Partial = Plenty more.

Permanent product

Sometimes you do not observe the behavior at all — you measure what it left behind. Worksheets completed, dishes washed, toys put away.

Permanent product is efficient and objective, but it cannot tell you who produced the outcome or how. If a scenario mentions a completed task the technician did not witness, permanent product is likely the answer, and the limitation is likely the follow-up.

Choosing a method: a decision path

1. Does the behavior have a clear beginning and end? If not, duration or an interval method rather than count.
2. Is the concern how often, how long, or how quickly? Count/rate, duration, or latency respectively.
3. Do session lengths vary? Use rate rather than raw frequency.
4. Can you attend to the behavior continuously? If not, use a discontinuous method.
5. If discontinuous, which direction is the program going? Increasing a behavior favors whole interval; decreasing it favors partial interval; a general estimate favors momentary time sampling.

Behavior definitions come first

No measurement system rescues a bad definition. A usable operational definition is:

• Observable — someone else could see it
• Measurable — it can be counted or timed
• Objective — no inference about internal states

"Aggression" is not a definition. "Any instance of the client's open hand or closed fist contacting another person's body with audible force" is.

Fast test: could two people watching the same session independently record the same number? If not, rewrite it.

Graphing basics the exam expects

• Line graphs for behavior over time; the standard in applied practice
• X axis = time or sessions; Y axis = the measure
• Phase change lines are vertical and separate conditions; do not connect data points across them
• Data points are plotted as collected — you do not smooth, average, or omit inconvenient sessions

And the professional boundary that gets tested: an RBT collects data and updates graphs. Interpreting trend and making programmatic decisions from that graph belongs to the supervisor.

Five-minute self-test

Cover the tables and answer these:

1. Which interval method underestimates duration?
2. Sessions are 30, 45, and 60 minutes. Frequency or rate?
3. A learner takes 20 seconds to respond after an instruction. Which measure?
4. Which method credits an entire interval for a two-second occurrence?
5. Which measure describes the gap between two consecutive responses?

Answers: whole interval; rate; latency; partial interval; interresponse time. Five for five means this domain is done. Anything less, reread the table and retest tomorrow.