AD ABAdegree.org
ABA Careers
Last updated: November 4, 2026

Continuous vs. Discontinuous Measurement in ABA

Continuous measurement in ABA records every instance. Discontinuous measurement samples intervals. See which methods overestimate or underestimate, and why.

Taylor Rupe

Founder & Editor

Published

A woman at a wooden table writing on a clipboard data sheet, with a small timer on the table beside her

Key Takeaways

  • 01Continuous measurement records every instance using frequency, rate, duration, latency or interresponse time. Discontinuous measurement records a sample using partial interval, whole interval or momentary time sampling.
  • 02Each sampling method is biased in a known direction. Partial interval overestimates, whole interval underestimates, and momentary time sampling can go either way.
  • 03In a worked example, a behavior filling 50% of a minute scored 83% with partial interval, 17% with whole interval and 33% with momentary time sampling.
  • 04Count every instance of rare or dangerous behavior, because a sample can miss it. Use discontinuous methods for behavior that's too fast or continuous to count, or when you're running a group.

The difference in three lines

  • Continuous measurement records every instance: frequency, rate, duration, latency and interresponse time.
  • Discontinuous measurement records a sample: partial interval, whole interval and momentary time sampling.
  • The samples are biased in known directions. Partial interval overestimates, whole interval underestimates, and momentary time sampling can go either way.

Continuous measurement in ABA records every instance of a behavior for the whole observation. Discontinuous measurement in ABA records only a sample: you split the session into intervals and score each one by a rule, so some of the behavior goes unrecorded. Continuous data is complete. Discontinuous data is cheaper to collect, and it errs in predictable directions, which is the part most explanations leave out.

This guide is for students, RBTs and future BCBAs. It covers each method, shows one minute of behavior scored three ways, and explains which discontinuous measurement methods overestimate, which underestimate, and how to choose. Both exams test it, so it closes with where measurement sits on the outlines and a few practice questions.

What Is Continuous Measurement in ABA?

Continuous measurement means you're watching and recording for the entire observation period, and every response counts. There are five continuous measurement procedures you need to know.

  • Frequency, or count. You tally every instance. A client asks for a break 7 times in a session.
  • Rate. You divide the count by the time observed, so sessions of different lengths compare fairly. Seven requests in 30 minutes is a rate of about 0.23 a minute.
  • Duration. You time how long each instance lasts, then total or average it. A tantrum that runs 4 minutes matters more than its count of one.
  • Latency. You time the gap from a cue to the start of the response, such as from "put your shoes on" to the first shoe.
  • Interresponse time (IRT). You time the gap between one response and the next, which matters for pacing problems like eating too fast.

Continuous measurement is the more common choice in published research. A review of the Journal of Applied Behavior Analysis by Mudford, Taylor and Martin (2009) found continuous recording in 55% of the 168 articles from 1995 to 2005 that reported data on free-operant human behavior. The catch is practical. Counting every instance takes your full attention, and in a busy session you rarely have it.

What Is Discontinuous Measurement in ABA?

Discontinuous measurement trades completeness for something you can actually do while working. You divide the observation into equal intervals, such as 10 seconds each, and score each interval yes or no. The result is reported as the percentage of intervals scored, not as a count or a duration. That's a different unit, and confusing it with "percent of the time" is how most measurement errors start.

  • Partial interval recording. You score the interval if the behavior happened at any point during it, even for a second.
  • Whole interval recording. You score the interval only if the behavior happened for the entire interval.
  • Momentary time sampling. You score the interval only if the behavior is happening at the moment the interval ends. Everything in between is ignored.

Here's how the three rules treat the same minute of behavior. In this example the behavior happens in four bouts that add up to 30 of the 60 seconds, so the true figure is 50% of the session.

One minute of behavior scored three ways A 60-second observation split into six 10-second intervals. The behavior occurs for 30 of 60 seconds, 50 percent. Partial interval recording scores 5 of 6 intervals, 83 percent. Whole interval recording scores 1 of 6, 17 percent. Momentary time sampling scores 2 of 6 checks, 33 percent. One minute, scored three ways What happened: 30 of 60 seconds, 50% Behavior occurring from 3 to 7 s (4 s) Behavior occurring from 9 to 27 s (18 s) Behavior occurring from 42 to 45 s (3 s) Behavior occurring from 53 to 58 s (5 s) 0s 10s 20s 30s 40s 50s 60s Partial interval: 5 of 6 intervals, 83% (over) Partial interval, interval 1 (0 to 10 s): scored ✓ Partial interval, interval 2 (10 to 20 s): scored ✓ Partial interval, interval 3 (20 to 30 s): scored ✓ Partial interval, interval 4 (30 to 40 s): not scored – Partial interval, interval 5 (40 to 50 s): scored ✓ Partial interval, interval 6 (50 to 60 s): scored ✓ Whole interval: 1 of 6 intervals, 17% (under) Whole interval, interval 1 (0 to 10 s): not scored – Whole interval, interval 2 (10 to 20 s): scored ✓ Whole interval, interval 3 (20 to 30 s): not scored – Whole interval, interval 4 (30 to 40 s): not scored – Whole interval, interval 5 (40 to 50 s): not scored – Whole interval, interval 6 (50 to 60 s): not scored – Momentary time sampling: 2 of 6 checks, 33% Momentary time sampling, check at 10 s: occurring, scored Momentary time sampling, check at 20 s: occurring, scored Momentary time sampling, check at 30 s: not occurring, not scored Momentary time sampling, check at 40 s: not occurring, not scored Momentary time sampling, check at 50 s: not occurring, not scored Momentary time sampling, check at 60 s: not occurring, not scored
A worked example, not study data. The same 30 seconds of behavior in a 60-second session, divided into six 10-second intervals. Filled marks were scored; dots on the bottom row are the moment at the end of each interval when the observer looks up.

Partial interval recording reports 83% because one brief bout is enough to claim a full interval. Whole interval recording reports 17% because only one interval was filled from start to finish. Momentary time sampling reports 33% here, because two of the bouts happened to end just before a check. Move those bouts a few seconds and the momentary time sampling figure moves with them, up or down.

Which Discontinuous Measurement Methods Overestimate or Underestimate?

The pattern in the example isn't a quirk of one minute. It's the classic finding. Powell, Martindale and Kulp (1975) recorded a secretary's in-seat behavior continuously and with each time-sampling method, then compared them. Whole interval recording consistently underestimated the continuous measure. Partial interval recording consistently overestimated it. Momentary time sampling overestimated and underestimated about equally often. And the more often the samples were taken, the closer every method came to the continuous record.

The reason follows from the rules themselves.

  • Partial interval recording overestimates because any part of an interval counts as all of it. A 2-second instance in a 10-second interval gets scored as 10 seconds' worth.
  • Whole interval recording underestimates because any gap disqualifies the interval. A behavior that runs 9 of 10 seconds gets scored as zero.
  • Momentary time sampling has no built-in direction. It's a snapshot, so whether it runs high or low depends on where the behavior happens to fall relative to the checks.

Later work added two refinements that matter in practice. In a computer simulation, Harrop and Daniels (1986) found momentary time sampling estimated the overall level of behavior more accurately than partial interval recording, while partial interval recording was better at detecting that behavior had changed. They also noted that at high rates partial interval recording tends to understate how much behavior changed. And when Meany-Daboul and colleagues (2007) regraphed five treatment analyses with each method, the conclusions from both discontinuous methods often matched the continuous data. Momentary time sampling was slightly more likely to agree with duration, and partial interval recording slightly more likely to agree with frequency.

So the error isn't a reason to avoid discontinuous measurement. It's a reason to pick the method whose error you can live with, and to keep the intervals short.

Continuous vs. Discontinuous Measurement: The Comparison Table

Every method on one page. The last column is the bias each discontinuous method brings to your data.

Method Type What it records When to use it Over or under?
Frequency (count) Continuous Every instance, as a tally Discrete behavior with a clear start and end, at a rate you can count None from sampling
Rate Continuous Count divided by observation time Comparing sessions of different lengths None from sampling
Duration Continuous How long each instance lasts Behavior where time matters more than count, like tantrums or on-task work None from sampling
Latency Continuous Time from an instruction or cue to the start of the response Slow starts, like following directions None from sampling
Interresponse time (IRT) Continuous Time between the end of one response and the start of the next Pacing, like eating too fast None from sampling
Partial interval Discontinuous Whether the behavior happened at any point in each interval Behavior you want to decrease, especially brief or high-rate behavior Overestimates
Whole interval Discontinuous Whether the behavior lasted the entire interval Behavior you want to increase that should be sustained, like engagement Underestimates
Momentary time sampling Discontinuous Whether the behavior is happening at the instant the interval ends Ongoing behavior when you can't watch the whole time, or groups Either way

How to Choose Between Continuous and Discontinuous Measurement

Start with the behavior, then look at what you can realistically do during the session. Four questions settle most cases.

  1. Does it have a clear start and end, at a rate you can count? Use frequency or rate. Hitting, requests and correct responses are all countable.
  2. Does how long it lasts matter more than how often it happens? Use duration. A single 20-minute tantrum and twenty 1-minute tantrums aren't the same problem.
  3. Is it too fast, too continuous or too hard to count, or are you running a group? Use a discontinuous method. Momentary time sampling suits ongoing behavior like engagement. Partial interval recording suits behavior you're trying to reduce, because its overestimate keeps you conservative about claiming progress. Whole interval recording suits sustained behavior you're trying to increase, for the same reason in the other direction.
  4. Is it rare or dangerous? Count every instance. A sample can miss a behavior that happens twice a day, and that's exactly the behavior you can't afford to miss.

The choice also feeds everything downstream. A functional assessment built on inflated partial interval data can make a behavior look more frequent than it is, which is one reason the measurement plan belongs alongside the work on the functions of behavior. If you need a quick definition of any term here, the ABA glossary has short entries for each interval method.

Discontinuous Measurement on the RBT and BCBA Exams

Both exams test this directly, and the outlines name the methods.

On the RBT exam, the RBT Test Content Outline (3rd ed.) puts measurement in Domain A, Data Collection and Graphing, which carries 13 of the 75 scored questions. Task A.1 is "Implement continuous measurement procedures (e.g., frequency, duration, latency, interresponse time)." Task A.2 is "Implement discontinuous measurement procedures (e.g., partial & whole interval, momentary time sampling)." For an RBT, the questions are about doing it correctly: which rule scores which interval.

On the BCBA exam, the BCBA Test Content Outline (6th ed.) puts it in Domain C, Measurement, Data Display, and Interpretation, which carries 21 of the 175 scored questions. Four tasks cover this page:

  • C.4: measure temporal dimensions of behavior (duration, latency, interresponse time).
  • C.5: distinguish between continuous and discontinuous measurement procedures.
  • C.6: design and apply discontinuous measurement procedures.
  • C.9: select a measurement procedure that accounts for the critical dimension of the behavior and environmental constraints.

The BCBA questions lean on choice and consequence: why a method suits a behavior, and which way its data will be wrong. The full domain-by-domain breakdown is in our guide to the BCBA task list 6th edition.

Practice Questions on Continuous and Discontinuous Measurement

These four are original questions written for this page. They aren't taken from any BACB exam. For a full-length run at RBT-level items, try our RBT practice exam.

1. An RBT splits a 15-minute session into 30-second intervals and marks each interval in which the client hit the table at least once. Which procedure is this?

  1. A.Whole interval recording
  2. B.Partial interval recording
  3. C.Momentary time sampling
  4. D.Frequency recording
Show the answer

B. Partial interval recording. One instance anywhere in the interval is enough to score it, which is the definition of partial interval recording. Frequency recording would count every hit.

2. A teacher wants to know how much of a work period a student spends writing. She can't watch him the whole time, so she glances up when a vibrating timer goes off every 2 minutes. What is she using?

  1. A.Momentary time sampling
  2. B.Latency recording
  3. C.Partial interval recording
  4. D.Interresponse time
Show the answer

A. Momentary time sampling. She scores only what is happening at the moment each interval ends. That is momentary time sampling, and it suits an ongoing behavior she can't watch continuously.

3. A supervisor wants a measure of on-task behavior that is cautious, so a reported increase can't be an artifact of generous scoring. Which discontinuous method tends to make that error least likely?

  1. A.Partial interval recording
  2. B.Momentary time sampling
  3. C.Whole interval recording
  4. D.Permanent product recording
Show the answer

C. Whole interval recording. Whole interval recording only scores an interval if the behavior lasted all of it, so it tends to underestimate. Any increase it shows is unlikely to come from overcounting.

4. A client engages in head-hitting about twice a day. Which measurement choice fits best?

  1. A.Momentary time sampling every 10 minutes
  2. B.Partial interval recording in 1-minute intervals
  3. C.Frequency recording of every instance
  4. D.Whole interval recording in 30-second intervals
Show the answer

C. Frequency recording of every instance. A low-rate, high-risk behavior has a clear start and end and is easy to miss with a sample. Count every instance.

Frequently Asked Questions

What is the difference between continuous and discontinuous measurement in ABA? +
Continuous measurement records every instance of a behavior across the whole observation, using frequency, rate, duration, latency or interresponse time. Discontinuous measurement records only a sample. The session is split into intervals and each interval is scored with a rule, using partial interval recording, whole interval recording or momentary time sampling. Continuous data is more complete. Discontinuous data is easier to collect when you can't watch nonstop.
What are the three types of discontinuous measurement? +
Partial interval recording scores an interval if the behavior happened at any point in it. Whole interval recording scores an interval only if the behavior lasted the whole interval. Momentary time sampling scores an interval if the behavior is happening at the moment the interval ends.
Does partial interval recording overestimate or underestimate behavior? +
It overestimates. A single brief instance scores the whole interval, so the percentage of intervals scored runs higher than the real share of time the behavior took up. Whole interval recording does the opposite and underestimates. In Powell, Martindale and Kulp's 1975 comparison in the Journal of Applied Behavior Analysis, momentary time sampling overestimated and underestimated about equally often.
When should you use momentary time sampling? +
Use it for ongoing behavior, like engagement or stereotypy, when you can't watch one person continuously, for example while running a group or teaching. Don't use it for brief or rare behavior, because a short instance between checks is never recorded.
Is frequency continuous or discontinuous measurement? +
Frequency is continuous measurement. You count every instance during the observation period. Rate is also continuous: it is the count divided by the time you observed.

Sources

Last checked November 4, 2026.

  1. Powell, Martindale and Kulp, An evaluation of time-sample measures of behavior (Journal of Applied Behavior Analysis 8(4):463-469, 1975)
  2. Harrop and Daniels, Methods of time sampling: a reappraisal of momentary time sampling and partial interval recording (Journal of Applied Behavior Analysis 19(1):73-77, 1986)
  3. Meany-Daboul, Roscoe, Bourret and Ahearn, A comparison of momentary time sampling and partial-interval recording for evaluating functional relations (Journal of Applied Behavior Analysis 40(3):501-514, 2007)
  4. Mudford, Taylor and Martin, Continuous recording and interobserver agreement algorithms reported in the Journal of Applied Behavior Analysis, 1995-2005 (Journal of Applied Behavior Analysis 42(1):165-169, 2009)
  5. BACB, RBT Test Content Outline (3rd ed.), tasks A.1 and A.2 and the Data Collection and Graphing domain weight
  6. BACB, BCBA Test Content Outline (6th ed.), tasks C.4, C.5, C.6 and C.9 and the Domain C weight

Keep going

Explore ABA programs

Ready to take the next step? Browse our expert-ranked program listings.

Explore careers Browse programs

Sponsored

ABA Degree Programs

Schools pay to appear in these listings. They are separate from our rankings, which no school can pay to change.