The big idea
Science tries to understand its subject matter at three levels, and each one builds on the one before. In behavior analysis the subject matter is behavior and its relation to the environment.
- Description: observe and record what happens.
- Prediction: find events that reliably occur together, so one can be anticipated from the other.
- Control: show that changing one event reliably changes another, and that nothing else explains the change.
Control is the highest level. Behavior analysts aim for it because it is what lets them change behavior on purpose, which is the point of applied work.
Breaking it down
Description
Description is a careful, systematic account of the behavior. It answers what happens, how often, how long, and under what conditions. Counting a child’s requests across a school day, or classifying the kinds of play seen at recess, is description.
Descriptive work produces facts that can be checked by others. It can suggest ideas worth testing, but it cannot tell you why the behavior occurs.
Prediction
Repeated observation sometimes shows that two events covary: when one changes, the other tends to change too. That relationship is a correlation, and it is enough for prediction. If tantrums are more likely when a preferred activity ends, a teacher can predict when tantrums will happen.
Prediction has a limit that the exam tests often. A correlation does not show that one event causes the other. A third variable could be responsible for both, or the relation could run in the opposite direction.
Control
Control is reached when an experimenter manipulates one variable, the independent variable, and a specific change in behavior, the dependent variable, follows reliably. The change also has to be unlikely to come from anything else, such as a confounding variable. A result that meets both conditions is a functional relation.
Single-case designs reach control through prediction, verification and replication: the analyst shows the effect, then shows it again. You will meet that logic in D.4.
Seeing it in practice
A teacher is worried about a student, Maya, who calls out during lessons.
| Level | What happens | What it tells the teacher |
|---|---|---|
| Description | An observer counts Maya’s call-outs every lesson for a week. | Maya calls out about 14 times per lesson. |
| Prediction | The counts are much higher in math than in reading, and call-outs are usually followed by the teacher saying her name. | Call-outs can be anticipated, and attention may be involved. |
| Control | For several sessions the teacher alternates between responding to call-outs and ignoring them while praising hand-raising. Call-outs rise and fall with the conditions every time. | Attention is functionally related to Maya’s call-outs. |
Only the last row justifies building a plan around attention with confidence.
Common mix-ups
- Prediction is not causation. Words like correlated, covary, associated with or tends to occur when signal prediction. They never signal control on their own.
- Control is not coercion. In science, control means a demonstrated functional relation. It does not mean forcing anyone to do anything.
- Goals are not assumptions. Description, prediction and control are what the science is trying to achieve. Determinism, empiricism, parsimony and the other assumptions in A.2 are what it takes for granted.
- Goals are not dimensions. Applied, behavioral, analytic and the other dimensions in A.5 describe what makes work applied behavior analysis. Analytic, the dimension, is the applied version of control: it asks whether the work demonstrated a functional relation.
Exam tip: Look for the verb in the scenario. Observed, recorded, counted points to description. Found to occur together points to prediction. Manipulated, introduced and withdrawn, reliably produced points to control.
Quick recap
- The three goals, from lowest to highest, are description, prediction and control.
- Description records what happens. Prediction relies on covariation. Control requires manipulating a variable.
- Control demonstrates a functional relation, and applied behavior analysis needs it to be analytic.
- A correlation, however strong, only supports prediction.
Further reading
- Cooper, J. O., Heron, T. E., & Heward, W. L. (2020). Applied behavior analysis (3rd ed.), chapter 1.
- Baer, D. M., Wolf, M. M., & Risley, T. R. (1968). Some current dimensions of applied behavior analysis. Journal of Applied Behavior Analysis, 1(1), 91-97.
Key terms
- Description
- The first level of scientific understanding: systematic observation and recording of events so they can be quantified, classified and checked by others. Description answers what happens, how much and when, but not why.
- Prediction (Correlation)
- The second level of scientific understanding. Repeated observation shows that two events reliably covary, so one can be anticipated from the other. A correlation supports prediction but does not show that one event causes the other.
- Experimental control (Control)
- The highest level of scientific understanding. An experimenter manipulates an independent variable and shows that a specific change in the dependent variable reliably follows, while ruling out other explanations.
- Functional relation
- An experimentally demonstrated relation in which a specific change in the dependent variable (behavior) can be reliably produced by specific manipulations of the independent variable, and the change is unlikely to be the result of confounding variables.
Check yourself
5 questions on A.1. Choose an answer to see whether you're right and why.
Show the answer
Answer: D. The researcher only observes and records, so the study describes the behavior. Nothing is compared or manipulated. Prediction would need two events found to vary together, and control would need a manipulated variable.
Show the answer
Answer: C. Two events that reliably occur together let you predict one from the other. That is a correlation. Nobody manipulated sleep, so the finding cannot show that sleep loss causes self-injury, and no functional relation has been demonstrated.
Show the answer
Answer: B. The analyst manipulated one variable (attention) and the behavior changed reliably each time, with the change replicated across alternations. That is control, the highest level, and it shows a functional relation. Generality to other settings would need further evidence.
Show the answer
Answer: B. Prediction rests on covariation, which needs no manipulation. It sits between description and control, and it can come from single-case or group data. Only control requires manipulating an independent variable.
Show the answer
Answer: A. Control is a technical term for an experimentally demonstrated functional relation. It has nothing to do with coercion. Forecasting from past patterns is prediction, and holding variables constant is a feature of good experimental method, not the goal itself.
Adds a step to your study record. Saved in this browser only.