Dana, a teacher with a class of 22, wants to know how often a student stays on task during a 10 minute independent activity. Counting each moment of attention is not possible. Timing each stretch with a stopwatch is not possible while teaching. What is possible is looking up once a minute.
That is time sampling, and it is built for exactly this. This guide shows how it works, compares the three common methods with one set of numbers, and explains what to write on the graph.
Continuous and non-continuous
Project ACCESS divides data collection into two groups. In continuous methods, the observer watches the student the entire time and marks every occurrence. Frequency, duration, and latency are continuous. In non-continuous methods, the observer watches for a short time and collects data in intervals. Interval recording and time sampling are non-continuous.
The trade is simple. You give up an exact count in return for being able to collect data while you do other things.
The steps
South Dakota's Department of Education form gives the steps for interval recording:
- Select the length of the observation session.
- Break it into smaller, equal intervals.
- Mark whether the behavior occurred in each interval, either at any time during the interval (partial interval) or throughout the entire interval (whole interval).
- Calculate the percent of intervals in which the behavior occurred.
It also notes that a timing device is necessary: a clock, a watch, or a timer. The form's sheet lists observation blocks in 10 minute steps and asks for the number of marked intervals out of the total, as a percent.
Three methods, one example
Project ACCESS lists momentary time sampling, whole interval recording, and partial interval recording together as time sampling methods. Definitions for the first are not on the South Dakota form, so the version here is the common one: look at the student at the instant each interval ends.
Take a 10 minute observation with ten one-minute intervals. In the invented record, the behavior, say being out of seat, happens like this.
| Interval | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 |
|---|---|---|---|---|---|---|---|---|---|---|
| Seconds | 20 | 0 | 60 | 60 | 10 | 0 | 45 | 0 | 60 | 5 |
| Out of seat at the end? | No | No | Yes | Yes | No | No | Yes | No | Yes | Yes |
Now score it three ways.
| Method | Rule | Intervals marked | Result |
|---|---|---|---|
| Whole interval | Marked only if the behavior fills the interval | 3, 4, 9 | 3 of 10 = 30% |
| Partial interval | Marked if it occurs at any time in the interval | 1, 3, 4, 5, 7, 9, 10 | 7 of 10 = 70% |
| Momentary | Marked if it is happening when the interval ends | 3, 4, 7, 9, 10 | 5 of 10 = 50% |
The actual total, from the seconds, is 260 out of 600, about 43 percent. Whole interval came in under it and partial interval came in over it. That pattern is commonly described for these methods, but the sources here do not rank their accuracy, so treat it as a caution and not a rule.
What to write on the graph
Because the three methods give 30, 70, and 50 for the same behavior, the method is part of the number. Write on every sheet and graph:
- the method (whole, partial, or momentary),
- the interval length and the observation length,
- the behavior definition,
- who observed.
If the baseline was taken with partial interval at one minute, every later point must be too. See how to set an IEP baseline.
Choosing a method
A rough guide, flagged as practice and not from a rule:
| Behavior | Method that often fits | Why |
|---|---|---|
| Should be sustained: on task, working with a partner | Whole interval | Credit only when it holds the whole time |
| Brief and disruptive: calling out, tapping | Partial interval | Catches short events |
| Long or frequent, and you cannot watch constantly | Momentary | One quick look per interval |
The South Dakota form leaves the choice to the observer. NCII's behavior guide lists percentage of time as one scale a team can choose for a goal. If you use intervals, the goal should say so, for example "on task in at least 80 percent of one-minute intervals using whole interval recording." See behavior frequency and duration data for the continuous measures.
Check that two observers agree
A percent of intervals is only as good as the observers behind it. A simple check, standard in behavior practice but not set by these sources: have two people score the same ten-minute session independently, then compare interval by interval. Count the intervals where both marked the same thing, divide by all ten intervals, and multiply by 100.
If you agree on 9 of 10, that is 90 percent. If you agree on 6 of 10, look at where you differ. The cause is almost always the definition: one observer counted leaning back as out of seat and the other did not. Fix the wording, then try again. Do this once at the start of a goal, and again whenever a new person takes over the sheet, such as a paraprofessional or a substitute.
Practical tips
- Use a timer with a quiet signal. A vibrating watch or a silent visual timer keeps the class from noticing.
- Pre-print the sheet. Boxes for each interval, with a plus or minus.
- Keep the observation short. Ten to twenty minutes in the same activity is enough for one data point.
- Practice together. Two observers using the same sheet should land within a point or two. If not, tighten the definition.
- Sample across the group. For a small group, see small group data tracking.
Interval recording errors
- Changing method mid-goal. A switch from partial to whole interval can look like improvement or decline.
- Changing interval length. Longer intervals inflate partial interval results.
- Vague definitions. "Off task" means different things to different observers.
- Treating the percent as a count. 70 percent of intervals is not 70 percent of the time.
Interval length and accuracy checks
The South Dakota form and the Project ACCESS fact sheet describe the methods but do not set interval lengths, minimum intervals, or accuracy checks. Your district or behavior specialist may. Terms are in the glossary.