assignmentAssessment and Measurement

Baseline

Before trying to change anything, you have to record where you are starting from.

Translated from the Portuguese original. Read the original (Português)

boltIn short

The baseline is the measure of behavior taken before the intervention begins — and repeated across several sessions, always the same way. It is not a waiting period: it is the point of comparison that makes it possible to say, later, whether anything really changed. Without it, “it got better” becomes an impression, subject to selective memory and to the expectations of whoever is observing. With it, the question “is it working?” gets an answer that anyone can check by looking at the same graph.

Technical definition

In Applied Behavior Analysis and in single-case research, the baseline is the condition in which the independent variable — the intervention — is absent, and during which repeated measures of the target behavior are taken. It functions as a control condition: it provides the objective basis against which the effect of the intervention will be evaluated. In Cooper, Heron & Heward, the logic of the baseline rests on three elements — prediction, verification, and replication — and it is this logic that allows the observed change to be attributed to the intervention, rather than to chance, the passage of time, or other factors in the environment.

Why measure before changing anything

Picture two families facing the same question: “is the new program helping?”. The first answers from memory — “I think so, this week was calmer”. The second opens a record that began well before anything changed. Only the second can really answer, because only it has something to compare against.

Memory is a terrible spreadsheet. It is easy to remember intense episodes more strongly, to see what one already expected to see, and, after investing time and affection in an intervention, to want it to have worked — none of that is bad faith, it is simply human. The baseline does not eliminate these biases, but it takes away their power to decide: in place of an impression comes a record made before there is any expectation about the outcome.

Recording beforehand also produces information that would not appear any other way. How many times does the behavior actually happen per day? At what times? Is it stable, increasing, or decreasing on its own? The baseline may reveal that the problem is smaller than it seemed, or that most episodes cluster around a single time of day — and a finding like that changes the plan before it even begins.

BASELINE INTERVENTION prediction: would continue like this effect of the intervention occurrences per session sessions →
The phase on the left — the baseline — describes the behavior without the intervention. Extending that level forward, as the horizontal dashed line, is the prediction: what would be expected if nothing were done. The effect of the intervention is precisely the distance between that prediction and what actually happened. Without the phase on the left, the phase on the right would be nothing more than a line going down, without meaning.

When is the baseline ready?

There is no number of sessions that works for every case. What matters is not a quantity, but a clear story in the data. To read it, three things are examined: the level (how high the points sit), the trend (whether they rise, fall, or stay flat), and the variability (how much they swing from one session to the next). The level alone does not distinguish one pattern from another; it is the combination of trend and variability that produces the four classic trajectory patterns. Flip the cards:

One extra caution is worth noting: the mere fact that someone starts recording can already change what happens — this is called reactivity of observation. That is why the way of recording needs to be the same in both phases: the same operational definition, the same observers, the same times of day. Changing the measurement method partway through ruins the comparison just as surely as not measuring at all.

lightbulb

“But how many sessions, exactly?”A common rule of thumb is to record for at least three sessions before considering the baseline stable. Talking about a trend requires its own criterion: at least three consecutive points moving in the same direction. Quality standards for single-case research, such as those of the What Works Clearinghouse, require a minimum number of points per phase and ask for more than that minimum for the evidence to receive the highest rating. Even so, the true criterion is not the count: it is stability. A short, clear baseline is worth more than a long, chaotic one.

The logic behind it: prediction, verification, and replication

The baseline may look like just “the before”, but it underpins a three-step reasoning that is the foundation of every single-case design. Explore each one:

Prediction — the bet that the baseline allows you to make

With a stable baseline, it is explicitly assumed that, if nothing were changed, the data would keep behaving the same way. That is the horizontal dashed line in the graph above. This bet is what gives the comparison its meaning: the more stable the baseline, the more reliable the prediction — and the more convincing any departure from it.

Verification — ruling out the hypothesis that it would have changed on its own

The intervention is introduced and the data change. There is still room for doubt: “what if it would have changed anyway?”. Verification answers that by demonstrating that the previous level would have held steady had the intervention not been introduced — for example, by temporarily withdrawing the intervention and watching the data return to the old level, or by keeping a second context in baseline that has not yet received the plan.

Replication — the coincidence that does not repeat itself

Finally, the manipulation is repeated and the same result is obtained again. A coincidence happens once; the same result reappearing time after time is no longer luck. It is replication within the same case that turns “the data changed” into “it was the intervention that changed them” — what behavior analysts call experimental control.

When there is no time to wait

There is a real and well-recognized tension in the field: a longer baseline produces a more convincing graph, but it also means delaying the start of the intervention for someone who is suffering right now. When the behavior in question involves risk — intense self-injury, aggression, running into the street — prolonging data collection to improve the demonstration reverses the priorities: it puts the interest of the study ahead of the person's well-being. The field's codes of ethics are explicit about placing the client's well-being first.

This does not mean the measure has to be abandoned. There are options that preserve both things: shortening the baseline to the minimum that still gives a reliable reading, or choosing a design that does not require taking away from anyone what is already working — such as the multiple baseline, in which the intervention is introduced in a staggered way across different behaviors, settings, or people, and the comparison comes from those whose turn has not yet arrived.

What does not solve the problem is replacing the baseline with records that already existed — reports from school, notes kept by the family, logged incidents. That material is valuable for understanding the picture and raising hypotheses, but it was not produced under the same recording conditions as the data from the next phase. It describes the past; it does not replace the baseline as the point of comparison for what comes afterward.

warning

Baseline is not synonymous with “doing nothing”During this phase, no one goes without care, safety, or support. What has not yet entered the picture is the specific intervention being evaluated — not the care, not the protection, not the relationship. If someone presents a baseline data collection as a reason to stop acting in the face of a risk, the problem is not the baseline — it is in how it is being used.

From recording to decision

The baseline only fulfills its role if someone looks at it afterward. The full cycle is simple to describe and takes discipline to maintain: define the behavior in observable terms, record before intervening, introduce the intervention without changing how it is measured, compare the phases, and — this is the part that tends to be missing — act on what the data show, including when they say the plan is not working.

That is why the baseline shows up so early in any serious work: it is the difference between following someone's progress and simply hoping for the best. See also the entries on operational definition, functional assessment, and evidence-based practice.

baseline logic of the baseline level, trend, and variability control condition multiple baseline reactivity of observation

Test your understanding

quiz

Quick quiz

Three questions to reinforce the concept. You see the explanation right away.

Further reading

  1. Cooper, J. O.; Heron, T. E.; Heward, W. L. Applied Behavior Analysis. 3rd ed., 2020 — chapters on data analysis and on single-case designs (logic of the baseline).
  2. Baer, D. M.; Wolf, M. M.; Risley, T. R. “Some current dimensions of applied behavior analysis”. Journal of Applied Behavior Analysis, 1968.
  3. Kazdin, A. E. Single-Case Research Designs: Methods for Clinical and Applied Settings. Oxford University Press.
  4. What Works Clearinghouse (Institute of Education Sciences, U.S. Department of Education). Procedures and Standards Handbook — standards for single-case designs.
  5. Behavior Analyst Certification Board (BACB). Ethics Code for Behavior Analysts — priority of client well-being in service-context research.
info

This entry is educational and does not replace assessment, diagnosis or treatment by a qualified health professional. Every person is unique — for a specific case, seek individual guidance.

Related entries

Looking for professional guidance?

Instituto Walden4 is a psychology and education center in Brasília, Brazil, offering evidence-based assessment and care for autism, psychotherapy and development — in person and online. Our team serves in Portuguese; get in touch and we will find the best way to help.

chat Message us on WhatsApp
chat

Conta Walden4

Entrar

Use o e-mail e a senha da sua conta.