Some concepts in behavior analysis seem deceptively easy until learners are asked to demonstrate their understanding of that information. Time-sampling procedures provide an example of this situation in which the descriptions of when to record behavior appear straightforward until a learner attempts to use these methods to record behavior.
Methods for Time Sampling
It isn’t always possible to record every instance of behavior, but the target behavior (or behavior of interest) should be monitored in every phase of the experiment. Instead, practitioners can segment the observation period into smaller intervals of equal duration (e.g., 10 seconds) and record when behavior occurs (or does not occur) in each of those intervals. In this way, practitioners can sample from the universe of behavior, as Tony Cuvo would often say in class. We assume that information collected from these observation periods are representative of other periods in which the behavior could have occurred but wasn’t recorded.
While it can be difficult to determine exactly how long the intervals within an observation period should be for a given response class, Sanson-Fisher et al. (1980) found that observers can reliably record the behavior (i.e., positive and negative self-concept, independent altruism, on task, talk, bizarre) of clients in a psychiatric unit in intervals as short as 1 second and that the frequency of target behavior decreases as the average duration increases when using longer intervals (i.e., 2, 3, 4, 5, 6, and 10 seconds). For instance, there were 219 instances of on-task behavior recorded with the 2-second intervals (88.1% of behavior recorded with 1-second intervals), but there were only 108 instances of on-task behavior recorded with the 10-second intervals (49.3% of behavior recorded with 1-second intervals). Similarly, the average duration of on-task behavior recorded with 2-second intervals was 38.87 seconds (13.6% reduction from 1-second intervals), but the average duration of on-task behavior recorded with 10-second intervals was 80.19 seconds (106.3% reduction from 1-second intervals). Infrequent and brief behavior had the lowest error rate for event recording across the different interval lengths (i.e., independent altruism), and infrequent but long bouts of behavior had the lowest error rate for duration recording (i.e., bizarre movements, aggression, or emotions). Zakszeski et al. (2017) also found that shorter intervals were better for accurately recording classroom engagement. Accurately estimating the duration of events from time sampling involves using intervals that are less than or equal to the event duration for that target behavior (Wirth et al., 2014).
There are four time-sampling techniques that can be used to record samples of behavior in an observation period. Which time-sampling method you choose depends upon the characteristics of the target behavior you’re observing and what is the anticipated effect of the intervention (i.e., increase or decrease the future likelihood of behavior).
- Whole-interval recording involves recording the presence of the target behavior if it occurs throughout the entire interval. If the intervals are 5 seconds long, then the behavior is recorded if and only if it occurs for all 5 seconds. This method is typically used if the target behavior occurs continuously or if one instance of behavior is difficult to distinguish from another (Cooper et al., 2019). For example, Graham-Day et al. (2010) used whole-interval recording to assess self-monitoring alone versus self-monitoring with edibles as reinforcement for on-task behavior (e.g., attending to and complying with the teacher’s instructions and attending to the academic task) with three high school students with ADHD. As another example, Longano and Greer (2006) used whole-interval recording to assess the conditioned reinforcement effects on appropriate play (i.e., rolling a car along the floor) for two young children with autism spectrum disorders.
- Partial-interval recording involves recording the presence of the target behavior if it occurs at any time during the interval. If the intervals are 5 seconds long, then the behavior is recorded if and only if it occurs at any point during the 5 seconds. Multiple topographies of the target behavior can be recorded simultaneously using this method. For example, Longano and Greer (2006) used partial-interval recording to assess the conditioned reinforcement effects on stereotypy (i.e., hand flapping) for their two participants.
- Momentary time sampling involves recording the presence of the target behavior if it occurs at the end of the interval. If the intervals are 5 seconds long, then the behavior is recorded if and only if it occurs at the 5-second mark. For example, Saudargas and Zanolli (1990) assessed completion of workbook pages, ditto sheets, and boardwork for 16 elementary school children with momentary time sampling in 15-second intervals.
- Planned activity check involves recording the number of learners engaged in the target behavior at the end of the interval. If the intervals are 5 seconds long, then a tally of the total number of learners engaged in the target behavior is recorded at the 5-second mark. The total number of learners who could engage in the target behavior during the interval should remain consistent for the observation period. For example, Radley et al. (2021) assessed observer accuracy when using planned activity check to record simulations of whole-class behavior (e.g., on-task behavior).
As a reminder, partial-interval recording tends to overestimate the duration of behavior, whole-interval recording tends to underestimate the duration of behavior, and momentary time sampling can both overestimate (at short intervals) and underestimate (at long intervals) the duration of behavior when compared to event recording (see Wirth et al., 2014). Additionally, Green et al. (1982) found that students learning to use time sampling were more accurate in approximating the event occurrence of hair twisting from a videotape with momentary time sampling than they were with whole- and partial-interval recording. Students also recorded more reliable data with momentary time sampling and whole-interval recording methods than partial-interval recording. Consistent with these results and when using more typical interval lengths between 2-5 minutes, LeBlanc et al. (2020) found that Catalyst users (e.g., Board Certified Behavior Analysts) were most accurate in recording patient data when using momentary time sampling but preferred instead to use the less accurate partial-interval recording system.
Fiske and Delmolino (2012) nicely summarize the issues related to the use of time sampling for recording behavior as well as provide a flowchart with question prompts to determine which method is best for the current situation.
- If a practitioner believes that the target behavior currently occurs at a low rate and should increase to around 100% with the intervention (e.g., on-task behavior), then whole-interval recording should be used.
- If a practitioner believes that the target behavior currently occurs at a low rate and should increase but not necessarily to 100% with the intervention (e.g., bids for attention), then momentary time sampling should be used.
- If a practitioner believes that the target behavior currently occurs at a high rate and should decrease to around 0% with the intervention (e.g., aggression), then momentary time sampling should be used.
- If a practitioner believes that the target behavior currently occurs at a high rate and should decrease but not necessarily to 0% with the intervention (e.g., echolalia), then partial-interval recording should be used.
Class Activity to Practice Using Time-Sampling Techniques
Once learners understand the basic definitions and conditions under which they would use each time-sampling procedure, they can practice using those methods with short video clips of a target behavior. Any clip with an example of appropriate or inappropriate behavior should be fine. However, I find that it’s best to use the same video clip to illustrate the difference between the time-sampling procedures. I do this especially because the time-sampling methods won’t all work equally well to adequately capture all instances of the target behavior. Then, we can use that information to discuss in class the differences between the time-sampling procedures (again) and why we need to be careful about selecting the most appropriate method for our target behavior.
We will want to choose an example video clip with a target behavior that has a clear beginning and end to the movement cycle and for which we can provide a decent operational definition. We can also have our learners help us to create an operational definition for the target behavior after they watch the video clip for the first time. The video clip that I use is of Veruca singing about all the items that she wants (Movieclips, 2016, December 29). The target behavior that I use is Veruca saying the phrase, “I want…” It’s possible to find other video clips if we want to find an example of a nonvocal target behavior that occurs at higher rates and for which we can use the planned activity check time-sampling procedure (cf. food fighting in Scene City, 2022, December 27). We can ask learners to answer the following question prompts about the video clip (individually or with choral responding).
We will want to give learners a datasheet to use for recording the target behavior and a key for how to record the occurrences and nonoccurrences of the target behavior. We will mark an “X” in an interval if we’re recording an occurrence of “I want” and an “O” if we’re recording the nonoccurrence of the target behavior.
Prompt 1: What is the frequency (or count) for how many times Veruca says, “I want?”
Answer: Veruca says “I want” 12 times between 0:00 and 1:10 in the video. Each occurrence and their timestamps appear in the table below.
| “I want” Occurrence | Video Timestamp |
| 1 | 0:01 |
| 2 | 0:08 |
| 3 | 0:18 |
| 4 | 0:19-0:20 |
| 5 | 0:24-0:25 |
| 6 | 0:32-0:33 |
| 7 | 0:36 |
| 8 | 0:47 |
| 9 | 0:49 |
| 10 | 0:59 |
| 11 | 1:01 |
| 12 | 1:04 |
Prompt 2: In how many intervals does Veruca say, “I want” if we use whole-interval recording?
| “I want” target behavior | |||||||
| Interval length | 0:01 – 0:10 | 0:11 – 0:20 | 0:21 – 0:30 | 0:31 – 0:40 | 0:41 – 0:50 | 0:51 – 1:00 | 1:01 – 1:10 |
| Interval | 1 | 2 | 3 | 4 | 5 | 6 | 7 |
Answer: I chose 9-second intervals, but it is possible to use shorter intervals (e.g., 1 or 2 seconds) and have this target behavior occur throughout the entire interval. We would record seven O’s on our datasheet because none of the “I want” instances occurred for a duration of 9 seconds. The target behavior occurred in 0% of the seven intervals.
| “I want” target behavior | O | O | O | O | O | O | O |
| Interval length | 0:01 – 0:10 | 0:11 – 0:20 | 0:21 – 0:30 | 0:31 – 0:40 | 0:41 – 0:50 | 0:51 – 1:00 | 1:01 – 1:10 |
| Interval | 1 | 2 | 3 | 4 | 5 | 6 | 7 |
Prompt 3: In how many intervals does Veruca say, “I want” if we use partial-interval recording?
| “I want” target behavior | |||||||
| Interval length | 0:01 – 0:10 | 0:11 – 0:20 | 0:21 – 0:30 | 0:31 – 0:40 | 0:41 – 0:50 | 0:51 – 1:00 | 1:01 – 1:10 |
| Interval | 1 | 2 | 3 | 4 | 5 | 6 | 7 |
Answer: We would record seven X’s on our datasheet because there was an “I want” instance that occurred at least once in each of the seven intervals. The target behavior occurred in 100% of the seven intervals.
| “I want” target behavior | X | X | X | X | X | X | X |
| Interval length | 0:01 – 0:10 | 0:11 – 0:20 | 0:21 – 0:30 | 0:31 – 0:40 | 0:41 – 0:50 | 0:51 – 1:00 | 1:01 – 1:10 |
| Interval | 1 | 2 | 3 | 4 | 5 | 6 | 7 |
Prompt 4: In how many intervals does Veruca say, “I want” if we use momentary time sampling?
| “I want” target behavior | |||||||
| Interval length | 0:01 – 0:10 | 0:11 – 0:20 | 0:21 – 0:30 | 0:31 – 0:40 | 0:41 – 0:50 | 0:51 – 1:00 | 1:01 – 1:10 |
| Interval | 1 | 2 | 3 | 4 | 5 | 6 | 7 |
Answer: I counted any “I want” instance that occurred at 9 or 10 seconds into the interval just in case my timing was off by a second. We would record three X’s on our datasheet for when Veruca said, “I want” as the interval timed out. Three out of seven intervals with “I want” means that Veruca emitted the target behavior on 43% of the seven intervals.
| “I want” target behavior | O | X | O | O | X | X | O |
| Interval length | 0:01 – 0:10 | 0:11 – 0:20 | 0:21 – 0:30 | 0:31 – 0:40 | 0:41 – 0:50 | 0:51 – 1:00 | 1:01 – 1:10 |
| Interval | 1 | 2 | 3 | 4 | 5 | 6 | 7 |
Prompt 5: Which time-sampling method is the best one for recording Veruca’s behavior that doesn’t overestimate or underestimate her behavior?
Answer: If we had 24 2-second intervals and used partial-interval recording, then we would have 11 intervals in which we recorded the occurrence of the target behavior. This is close to our frequency of 12 instances of the target behavior. That means that the target behavior occurred in 46% percent of the intervals. The closest time-sampling method that we used for this video to 46% was momentary time sampling with the target behavior occurring in 43% of the intervals.
| X | O | X | O | O | X | X | O | X | O | X | X |
| 0:01 – 0:03 | 0:04 – 0:06 | 0:07 – 0:09 | 0:10 – 0:12 | 0:13 – 0:15 | 0:16 – 0:18 | 0:19 – 0:21 | 0:22 – 0:24 | 0:25 – 0:26 | 0:27 – 0:29 | 0:30 – 0:32 | 0:33 – 0:34 |
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 |
| X | O | O | O | X | O | O | O | X | X | O | O |
| 0:35 – 0:37 | 0:38 – 0:40 | 0:41 – 0:43 | 0:44 – 0:46 | 0:47 – 0:49 | 0:50 – 0:52 | 0:53 – 0:55 | 0:56 – 0:58 | 0:59 – 1:01 | 1:02 – 1:04 | 1:05 – 1:07 | 1:08 – 1:10 |
| 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 |
Learners seem to enjoy this activity even if they have never seen the video clip before and don’t anticipate recording the specific target behavior displayed in the video. More realistic target behaviors in the types of settings practitioners actually encounter can be used, but those videos tend to be removed from the platform and/or set to restricted permissions if the instructor isn’t the creator of the video. Because of this, larger and more stable content creators are a better resource for videos used in class or training. Regardless of the target behavior chosen and the context in which the behavior is depicted, learners tend to also gain a greater appreciation for how important the data collection methods are when participating in a live demonstration of time sampling.
Image credits
[1] Image provided courtesy of Pixabay under Pexels CC0
[2] Image provided courtesy of Gustavo Fring under Pexels license
References
Cooper, J. O., Heron, T. E., & Heward, W. L. (2019). Applied behavior analysis (3rd ed.). Pearson.
Fiske, K., & Delmolino, L. (2012). Use of discontinuous methods of data collection in behavioral intervention: Guidelines for practitioners. Behavior Analysis in Practice, 5, 77-81. https://doi.org/10.1007/BF03391826
Graham-Day, K. J., Gardner, III, R., & Hsin, Y. W. (2010). Increasing on-task behaviors of high school students with attention deficit hyperactivity disorder: Is it enough? Education and Treatment of Children, 33, 205-221.
Green, S. B., McCoy, J. F., Burns, K. P., & Smith, A. C. (1982). Accuracy of observational data with whole interval, partial interval, and momentary time-sampling recording techniques. Journal of Behavioral Assessment, 4, 103-118. https://doi.org/10.1007/BF01321385
LeBlanc, L. A., Lund, C., Kooken, C., Lund, J. B., & Fisher, W. W. (2020). Procedures and accuracy of discontinuous measurement of problem behavior in common practice of applied behavior analysis. Behavior Analysis in Practice, 13, 411-420. https://doi.org/10.1007/s40617-019-00361-6
Longano, J. M., & Greer, R. D. (2006). The effects of a stimulus-stimulus pairing procedure on the acquisition of conditioned reinforcement on observing and manipulating stimuli by young children with autism. Journal of Early and Intensive Behavior Intervention, 3, 62-80. https://doi.org/10.1037/h0100323
Movieclips. (2016, December 29). Willy Wonka & the chocolate factory [Video]. Youtube. https://youtu.be/5wAlQf4WdiE
Radley, K. C., Dart, E. H., Schrieber, S. R., & Davis, J. L. (2021). The accuracy of peer comparison observations: A simulated analysis. Behavioral Disorders, 46, 120-129. https://doi.org/10.1177/0198742920944845
Sanson-Fisher, R. W., Poole, A. D., & Dunn, J. (1980). An empirical method for determining an appropriate interval length for recording behavior. Journal of Applied Behavior Analysis, 13, 493-500. https://doi.org/10.1901/jaba.1980.13-493
Saudargas, R. A., & Zanolli, K. (1990). Momentary time sampling as an estimate of percentage time: A field validation. Journal of Applied Behavior Analysis, 23, 533-537. https://doi.org/10.1901/jaba.1990.23-533
Scene City. (2022, December 27). Hook: Imaginary dinner and food fight scene (Robin Williams) [Video]. Youtube. https://youtu.be/7N-a72jEFh0
Wirth, O., Slaven, J., & Taylor, M. A. (2014). Interval sampling methods and measurement error: A computer simulation. Journal of Applied Behavior Analysis, 47, 83-100. https://doi.org/10.1002/jaba.93
Zakszeski, B. N., Hojnoski, R. L., & Wood, B. K. (2017). Considerations for time sampling interval durations in the measurement of young children’s classroom engagement. Topics in Early Special Education, 37, 42-53. https://doi.org/10.1177/0271121416659054
Blog post contributed by Melissa Swisher