A teacher writes in a note: "Rafael called out 14 times on Monday and 9 times on Wednesday. Better!" But Monday was a full-day schedule, and Wednesday was a short day with a substitute. Was it better?
Counting is the easy part. Making counts comparable is the work. This guide covers the main ways to count and time behavior, what each can and cannot tell you, and how to plan a baseline.
Pick the measure by the question
Project ACCESS lists the continuous measures plainly. Frequency answers how many times. Duration answers how long. Latency answers how long between a signal and the start of the behavior. A permanent product, such as a completed worksheet, is a fourth kind of record for academic or life skills.
| Measure | Question it answers | Best when | Example |
|---|---|---|---|
| Frequency | How many times? | The behavior is brief and easy to count | Calls out during a 30 minute lesson |
| Duration | How long? | The behavior has a clear start and stop | Time spent off task, crying, or out of seat |
| Latency | How long until it starts? | A clear signal and a delay you want to shorten | Time from "Get started" to first written word |
| Intensity | How severe? | Severity matters | A five point scale for screaming |
Project ACCESS adds that frequency is not very useful for behaviors like tantrums or crying, where a count hides how long each lasted.
Make counts comparable
Counts only compare if the observation lasted the same time. Rafael's 14 and 9 cannot be compared without knowing how long each observation was. Two fixes:
- Hold the period steady. Count in the same 30 minute block each day.
- Convert to a rate. Divide the count by the minutes. Suppose Monday's observation lasted 70 minutes and Wednesday's 30. Then 14 calls out in 70 minutes is 0.2 per minute, or 12 per hour. 9 calls out in 30 minutes is 0.3 per minute, or 18 per hour. The short day was not better.
Make durations comparable
Project ACCESS says to use percentages when comparing duration across days, calculated as duration divided by time observed, because the observation periods differ. Its reasoning is the same as above.
With invented numbers: on Monday a student is off task for 12 minutes of a 40 minute block, which is 30 percent. On Wednesday the student is off task for 9 minutes of a 20 minute block, which is 45 percent. The minutes fell, but the share of time rose. Report both numbers, and graph the percent.
Plan the baseline
Both NCII and Queensland tell you not to guess.
- Choose a simple method. Queensland says to select a simple method of data collection suited to the behavior: a frequency count for something that happens several times a day, and existing records for rare events.
- Define the behavior. NCII mentions operational definitions with anchored examples and nonexamples, so two observers count the same things.
- Collect enough days. NCII says at least five data points, ideally stable. Highly variable data may mean you need more data and a sharper definition. Queensland says a week or more, depending on how often the behavior occurs.
- Pick the time. Observe when the behavior is most likely. Keep the same time each day.
- Record safety issues. For behavior that may cause injury, Queensland says to complete a risk assessment. That is a reason to act first and collect as you go.
For the broader baseline method, see how to set an IEP baseline.
A daily sheet that fits on an index card
Keep it small enough that you will actually use it. One row per day is enough.
| Date | Start and end | Minutes observed | Count or minutes | Rate or percent | Note |
|---|---|---|---|---|---|
| Oct 12 | 9:00 to 9:40 | 40 | 6 calls out | 9 per hour | Normal day |
| Oct 14 | 9:00 to 9:20 | 20 | 3 calls out | 9 per hour | Substitute |
The two days look different as counts (6 and 3) and identical as rates (9 per hour). The note column saves you from explaining later why a day looks odd. For latency, swap the middle columns for "signal time" and "start time," and record the gap in seconds or minutes.
Write the goal about the skill
NCII's guide says IEP goals should focus on replacement behavior. Its example: "Miranda will increase the frequency of raising her hand in class" is a goal, while "decrease the frequency of loud outbursts" is not. The reason is that the goal should describe what the student will learn to do. The outbursts can still be counted alongside the new skill.
Utah's guide agrees, and adds a caution about percent goals when the opportunities are open-ended. A goal to raise a hand "80 percent of the time" cannot be measured, but "in 8 of 10 teacher questions" can. See percent accuracy vs trials vs duration for more.
Collect regularly
Utah's guide says it is best to collect behavior data every day, and that collecting only "as needed" or changing how often can produce serious errors in interpretation. A simple daily tally on an index card beats a detailed form filled in twice a week.
Graph it with a goal line
NCII says to always include a goal line on the graph, and that a goal should be monitored often enough to determine progress and make timely decisions. NCII also notes that behavior goals are often set around 80 to 90 percent for behaviors to increase, or 10 to 20 percent for behaviors to decrease, or at a level typical of peers, but that a team may start lower and raise it. That is NCII's guidance, not a rule. The graph logic is in aimline and trendline explained.
Counting slips
- Comparing raw counts across unequal periods.
- Changing the definition midweek. A new definition starts a new baseline.
- Using duration for a behavior with no clear start or end. Use intervals instead (see the next guide, interval recording and time sampling).
- Measuring the adult. NCII's list of challenges includes goals that focus on an adult behavior, such as fewer teacher prompts.
- Skipping the pattern. For when and why, see ABC data sheets explained.
Choices the sources leave to your team
None of the sources gives one right measure for a given behavior or a required number of days beyond the minimums above. Your team and your district's procedures decide. Terms are in the glossary.