The big idea
Continuous measurement can detect every instance of the target behavior during an observation period. Discontinuous measurement samples the behavior, so some instances, or parts of instances, may never reach the data sheet.
“Continuous” describes what happens within the observation period, not how long you observe. Counting every hand raise during a 20-minute lesson is continuous measurement, even though nobody watched the other six hours of the school day.
The trade-off is completeness against effort. Continuous data give the truest picture of the behavior, but they can be hard to collect when behavior is very frequent, has no clear start and stop, or must be recorded by someone who is also teaching. Discontinuous procedures are easier but produce estimates, and each one is biased in a known direction (C.6).
Breaking it down
Continuous procedures
- Event recording: tallying each occurrence as it happens, with tally marks, a counter or an app. It yields count, rate and percentage (C.3).
- Timing: using a stopwatch or app to record duration, latency or interresponse time (C.4).
Continuous procedures work best when the behavior has a clear beginning and end, occurs at a rate the observer can keep up with, and the observer can attend to it for the whole period.
Discontinuous procedures
These divide the observation period into intervals and record whether the behavior met a scoring rule, not how many times it occurred or how long it lasted:
- Whole-interval recording: score the interval only if the behavior lasted the entire interval.
- Partial-interval recording: score it if the behavior occurred at any point.
- Momentary time sampling: score it if the behavior is occurring at the moment the interval ends.
- Planned activity check (PLACHECK): momentary time sampling applied to a group.
The result is the percentage of intervals scored, an estimate of how much the behavior occurred. Terms vary. Cooper, Heron and Heward use time sampling for all four, while other sources call the first two interval recording and keep time sampling for the momentary methods. Either way, all four are discontinuous.
Comparing the two
| Continuous | Discontinuous | |
|---|---|---|
| What is recorded | Every instance during the period | Whether behavior met the rule in each interval |
| Typical output | Count, rate, duration, latency, IRT | Percentage of intervals |
| Accuracy | True values, given a good definition and trained observers | Estimates that can run high or low |
| Demand on the observer | High | Lower; one observer can track several behaviors or people |
| Best fit | Discrete behaviors at manageable rates | Frequent or continuous behaviors without clear starts and stops |
Seeing it in practice
| Situation | Procedure | Type |
|---|---|---|
| An RBT tallies each full-sentence request Kenji makes during a 30-minute play session. | Event recording | Continuous |
| A teacher running a lesson looks up every 2 minutes and notes whether Amara is working. | Momentary time sampling | Discontinuous |
| Staff time each of Mr. Lewis’s showers from water on to water off. | Duration recording | Continuous |
| Observers score a toddler’s vocal stereotypy in 10-second partial intervals because it starts and stops too fast to count. | Partial-interval recording | Discontinuous |
| A supervisor tallies every specific praise statement an RBT gives during a 15-minute observation. | Event recording | Continuous |
Common mix-ups
- Continuous does not mean all day. It means every instance within the observation period.
- Discontinuous data are not counts or durations. “Scored in 40% of intervals” does not mean 40 occurrences, and it does not mean the behavior filled 40% of the session.
- Observing on some days is not discontinuous measurement. Watching three of five days is a scheduling decision. If you record every instance during the sessions you observe, the measurement is continuous. Poorly scheduled observation is its own problem (C.8).
- Discontinuous is not always worse. For behavior that runs without clear breaks, trying to count can be less reliable than interval scoring.
- Rare, brief behavior needs continuous recording. Momentary samples tend to miss it, and long partial intervals inflate it.
Exam tip: Ask whether every occurrence during the observation could end up in the data. If yes, it is continuous. If the procedure only records whether behavior met a rule in or at each interval, it is discontinuous, however short the intervals.
Quick recap
- Continuous measurement can capture every instance during the observation period: event recording and timing.
- Discontinuous measurement samples behavior: whole-interval, partial-interval, momentary time sampling and PLACHECK.
- Discontinuous data are reported as a percentage of intervals and are estimates.
- Continuous fits discrete behavior at manageable rates; discontinuous fits frequent or continuous behavior and busy observers.
Further reading
- Cooper, J. O., Heron, T. E., & Heward, W. L. (2020). Applied behavior analysis (3rd ed.), chapter 4.
Key terms
- Continuous measurement
- Measurement that can detect every instance of the target behavior during the observation period, such as event recording and timing duration, latency or interresponse time. Continuous refers to every instance within the period, not to observing all day.
- Discontinuous measurement
- Measurement that samples behavior, so some instances or parts of instances may not be recorded. Whole-interval recording, partial-interval recording, momentary time sampling and planned activity checks are discontinuous, and they produce estimates reported as a percentage of intervals.
- Event recording (Frequency recording, Tally recording)
- Recording each occurrence of a behavior as it happens, with tally marks, a counter or an app, to produce a count. It is a continuous measure and suits behaviors with a clear beginning and end that occur at a rate an observer can keep up with.
- Interval recording
- A discontinuous procedure that divides an observation period into equal intervals and scores each interval for whether the behavior occurred, under a whole-interval or partial-interval rule. Data are reported as the percentage of intervals scored. Some texts group it with momentary methods under the broader label time sampling.
Check yourself
4 questions on C.5. Choose an answer to see whether you're right and why.
Show the answer
Answer: D. Partial-interval recording notes only whether call-outs happened in each interval, so three call-outs in one interval are recorded as one. That makes it discontinuous no matter how closely the teacher watches or how short the intervals are. Observing one lesson a day is a scheduling choice, not what makes a procedure discontinuous.
Show the answer
Answer: A. Timing every latency in the session captures every instance, so it is continuous. The other three are momentary time sampling, partial-interval recording and a planned activity check, all of which sample the behavior and can miss or merge instances.
Show the answer
Answer: C. Momentary time sampling scores only what is happening at the end of each interval, so a few seconds of aggression twice a week will almost never be caught. Event recording of every episode fits brief, rare behavior. Momentary time sampling can err in either direction, not only upward, and it is widely used in practice.
Show the answer
Answer: A. Five minutes holds 30 ten-second intervals, and 18 ÷ 30 × 100 = 60% of intervals. Partial-interval data say only that flapping happened at some point in those intervals, so they are not a count (there may have been several flaps per interval), a rate (18 ÷ 5 = 3.6 treats intervals as responses) or a duration (partial-interval recording overstates time).
Adds a step to your study record. Saved in this browser only.