Time sampling is a behavioral observation method where you divide an observation period into equal intervals and record whether a specific behavior occurs during each interval, rather than tracking every single instance of the behavior continuously. It’s one of the most widely used techniques in education, psychology, and applied behavior analysis because it dramatically reduces the effort of data collection while still producing useful estimates of how often a behavior happens.
There are three main types of time sampling, each with different rules for when you mark a behavior as “occurred.” The method you choose depends on what kind of behavior you’re watching and whether you expect it to increase or decrease over time.
How Time Sampling Works
The basic setup is the same across all three types. You pick a target behavior (say, a student being on-task), decide on an interval length (commonly 10 to 30 seconds), and create a data sheet with one box per interval. Each time an interval ends, you record whether the behavior happened, using the specific rules of whichever method you’re using. At the end of the session, you divide the number of intervals where the behavior was scored as occurring by the total number of intervals. That gives you a percentage estimate of how much of the session the behavior was present.
For example, if you observe a student for 5 minutes using 15-second intervals, you’d have 20 intervals. If the behavior was scored in 14 of them, your estimate is 70%.
Whole Interval Recording
With whole interval recording, you only mark a behavior as occurring if it lasts for the entire interval. If a student is on-task for 8 out of 10 seconds in an interval, that interval gets scored as “did not occur.” This makes whole interval recording a conservative measure. It tends to underestimate how much a behavior actually happens because even brief pauses within an interval cause it to be scored as absent.
That conservative bias is actually useful in certain situations. If you’re working with a behavior that should be happening nearly all the time, like sustained attention or staying seated, whole interval recording sets a high bar. A practitioner who believes a target behavior currently occurs at a low rate and should increase toward 100% with intervention would choose this method. It’s also helpful when one instance of a behavior is hard to distinguish from the next, such as continuous engagement with a task, because you’re measuring sustained presence rather than counting individual occurrences.
Partial Interval Recording
Partial interval recording works in the opposite direction. You mark a behavior as occurring if it happens at any point during the interval, even briefly. If a student calls out just once during a 10-second window, the whole interval is scored as an occurrence. This method catches everything but at a cost: it consistently overestimates how much a behavior is really happening.
Research published in the Journal of Applied Behavior Analysis found that partial interval recording consistently overestimated the duration of behaviors like stereotypy. Because it flags every interval where any instance of the behavior appears, it inflates both the apparent rate and duration. The margin of error is larger than with other time sampling methods. This creates a risk of false positives when you’re trying to estimate how frequently something happens. At the same time, it can underestimate the magnitude of change when a high-rate behavior starts decreasing, because even residual instances keep lighting up intervals.
Despite these limitations, partial interval recording is a practical choice when you’re tracking a behavior that occurs at a high rate and you want to see it decrease but not necessarily disappear entirely. Repetitive vocalizations or fidgeting are common examples. The overestimation works as a built-in sensitivity detector: if the behavior truly drops, even this generous scoring method will reflect the change.
Momentary Time Sampling
Momentary time sampling is the simplest version to carry out. You look up at the exact moment an interval ends and record whether the behavior is happening right then. You don’t need to watch continuously during the interval itself, which makes it far less demanding on the observer. A kitchen timer, watch alarm, or audio cue can signal when to look up and record.
This method is especially useful for behaviors that are hard to count because they happen at very high rates or blend together without clear start and stop points. In classroom settings, momentary time sampling with 15-second intervals is commonly used to track academic engagement, defined to include both active participation (writing, answering questions) and passive engagement (listening, reading).
The tradeoff is that momentary time sampling can underestimate behavior, since it only captures a snapshot. A behavior could be happening throughout most of an interval but stop just before the observation moment, and it would be missed. For that reason, it works best for behaviors that are relatively frequent and last for longer stretches of time. Research comparing the two methods found that while momentary time sampling sometimes overestimated and sometimes underestimated duration, its margin of error was consistently smaller than partial interval recording’s.
How Interval Length Affects Accuracy
The length of each interval matters more than many observers realize. Shorter intervals produce more accurate data but require more attention. In one-hour observation sessions, momentary time samples taken every 10, 20, or 30 seconds yielded 90% of estimates within 10% of the true value for behaviors that occurred 25%, 50%, or 70% of the time. That’s a solid level of accuracy for most practical purposes.
But rare behaviors are a different story. Behaviors occurring for only about 2% of a session were not accurately captured even at 10-second intervals. For five-minute observation windows, acceptable accuracy was only approached when data from eight hours of observation was averaged together, and only for behaviors present at least 25% of the time. The takeaway: time sampling works well for behaviors that happen reasonably often. For rare events, you need either much longer observation periods or a different method entirely.
Time Sampling vs. Continuous Recording
Continuous recording, sometimes called all-occurrences sampling, means tracking every single instance of a behavior as it happens in real time. This produces the most accurate frequency and duration data and is the gold standard when the behaviors are few, clearly defined, and easy to count.
Time sampling exists because continuous recording often isn’t practical. It requires constant, undivided attention from the observer, which is exhausting over long periods and nearly impossible if you’re also doing something else, like teaching a class. Recording at fixed intervals reduces that labor significantly. A teacher using momentary time sampling can glance up every 15 seconds, mark a box, and return to instruction. That’s sustainable in a way that continuous observation is not.
The cost is precision. Research comparing the two approaches in animal behavior studies found that sampling at intervals of 30 seconds or longer underestimated how often subjects performed brief, frequent behaviors, because some of those events occurred between sample points and were never captured. When knowing the exact count of a behavior matters, continuous observation is necessary. When you need a reasonable estimate of how prevalent a behavior is across a session, time sampling delivers that with far less effort.
Choosing the Right Method
Your choice comes down to three questions: what kind of behavior are you tracking, what do you expect to happen to it, and what resources do you have for observation?
- Whole interval recording is best for behaviors you want to see sustained throughout each interval, like on-task behavior or independent play. It sets a high standard and underestimates, so improvements will be clearly visible when they happen.
- Partial interval recording is best for behaviors you want to see decrease, like calling out or self-stimulatory behavior. It catches any occurrence within an interval, so it’s sensitive to the presence of unwanted behavior.
- Momentary time sampling is the most practical when you can’t watch continuously. It’s the least biased overall and works well for frequent, longer-duration behaviors. It’s the most common choice in classroom settings where the observer is also the teacher.
Other factors include the duration of individual behavior episodes (brief events are harder to capture with momentary sampling), how many people or animals you’re observing at once (scan sampling, a form of momentary time sampling applied to groups, lets you survey an entire class at each interval), and whether you need to know presence/absence or precise frequency. For presence and absence questions only, any time sampling method can work. For precise counts, continuous recording remains the better tool.
Digital Tools for Time Sampling
Paper data sheets and kitchen timers still work fine for basic classroom observation, but digital tools have made time sampling faster and less error-prone. Several software applications designed for handheld devices support scan sampling and other time-based methods, including Animal Behaviour Pro, BORIS, Prim8, and ZooMonitor. These apps handle interval timing automatically, prompt the observer when it’s time to record, and export data directly for analysis. Most support multiple sampling methods (focal, scan, all-occurrence) within the same platform, so you can switch approaches without switching tools.