The meeting ends with a plan to start Monday. Someone asks, "What is the number now?" Nobody knows. Without a starting number, "it is better" and "it is worse" are both opinions.
This page is about the starting number. It covers what to measure, how many days to measure, and what to do when you cannot afford to wait.
Why a baseline
A baseline is a dated record of the behavior before the plan changes anything. Without it, a team cannot tell whether a plan helped, did nothing, or made things worse. Colorado's guide says a behavioral goal should include baseline data so that progress toward the goal can be monitored, and it gives goals that begin from a measured level, such as an average of 15 times per week.
What to count
Queensland's guide says it is important to know the frequency, duration, or intensity of the behavior before interventions begin, and to select a simple method suited to it.
| Behavior | A fitting measure | Example |
|---|---|---|
| Happens several times a day | A frequency count | Tally of calling out per morning block |
| Happens less often, such as fighting | Existing records may be enough | Count of referrals over the term |
| Matters because of how long it lasts | Duration | Minutes off task, or time lying on the floor |
| Severity matters, such as screaming | A simple five-point scale | Rating each episode with written level descriptions |
The measure must match the definition, so write the definition first. See writing an operational definition.
How many days
The sources give a few anchors rather than one number.
| Source | Guidance |
|---|---|
| NCII behavior guide | At least five data points, ideally stable, unless there is an ethical reason to start intervention at once |
| Queensland | A period of a week or more, depending on how often the behavior occurs |
| IES practice guide | A few days if the behavior occurs several times a day, one to two weeks if it occurs a few times a week |
So a behavior that occurs every day needs about five to seven school days. One that occurs twice a week needs about two weeks. Rarer behavior needs a longer window or other records.
Stable or not
NCII says baseline data should ideally be stable, and that highly variable data may suggest collecting additional baseline data and revisiting the operationally defined behavior and its anchors. So if the counts are 2, 14, 1, 9, and 0, do not average them and move on. Look first at the definition: are two adults counting the same thing? Then look at the days: does the spread follow a pattern, such as a particular day of the week? The variability may be the finding.
Keep the conditions the same
Colorado's guide describes a fixed data window, the part of the day when the behavior will be counted, kept the same throughout the plan. Days outside that window are not added to the graph. Set the window before baseline begins, and keep the same sheet. If baseline is taken for 9 to 12 and the plan is monitored for 8 to 3, the two numbers cannot be compared.
Do not change how adults respond during baseline. The point is to see the behavior as it is. That can feel odd when something is going wrong in the room, which is why the next section matters.
When safety cannot wait
NCII's guide says to collect at least five data points unless there is an ethical reason to begin immediate intervention. Queensland says that for behaviors that are unsafe or may cause injury, a risk assessment should be completed to identify the level of risk and the actions to be taken.
So if someone could be hurt, do not hold back a response to get a clean baseline. Put the safety steps in place, keep counting, and mark the date on the graph where the response changed. The data you have are a baseline with a caveat, and the team can say so in the plan. A rough baseline is better than none, and safety is the reason to depart from the usual order.
Graph it
Plot the baseline days on the left, with the date and the count for each, and leave room on the right for the plan. When the plan begins, draw a vertical line. NCII's guide says to always include a goal line on a progress graph. For a behavior decrease, the goal line slopes down from the baseline level to the target. The data sheets guide, data sheets for BIP progress, shows the layouts.
A baseline for the skill you will teach
Colorado's guide says best practice is to track both the behavior to decrease and the desired behavior to increase. So count the replacement skill during baseline, even if you expect zero. A student who asks for a break appropriately zero times in a week has a baseline of zero, and the first time the number is one, you will see it. If the skill is something the student can already do sometimes, count those times, so the plan builds on what exists rather than starting from scratch.
Telling the family the starting number
A baseline is a good thing to share with a parent early, with the definition. Say what you counted, over how many days, and what the number was. Ask whether it matches what they see at home. A family that sees "about ten call-outs a morning" on paper, with the definition beside it, is in a better position to join the plan than one who hears only "he is disruptive." See parent communication about behavior.
A worked baseline
An invented example. A student calls out during whole-group lessons. Counts over six school days in the 9:00 to 11:30 window are 9, 11, 8, 12, 10, and 9. The median is 9.5 and the range is 8 to 12, which is fairly stable. The team writes: baseline, 6 school days, median 9.5 call-outs per window, definition attached, taken by the teacher with a clicker, two observers agreed on a sample day. Now the goal can say "from a baseline of about 10 to no more than 3."
Common slips
- Using memory instead of a counter.
- Baseline of two days. Too few to see a pattern.
- Changing the definition after day three. Start over.
- Mixing windows. Keep one.
- Forgetting the skill. Count the replacement behavior too, even if it is zero.
For how the baseline feeds the assessment, see FBA steps in order. Terms are in the glossary.
Rules that vary
Some districts set a minimum baseline period or a specific form. Follow yours, and keep the logic above. This guide is general information and not a clinical protocol.