Behavior analysis is a natural science, and like any science it aims to understand the events it studies. That understanding comes in three levels: description, prediction, and control. Each level depends on the one before it, and each lets you say something stronger about behavior.
Why it matters
The level of understanding you have reached decides what you can honestly claim. A careful record of what a client does is valuable, but it does not tell you why the behavior happens. Two events that reliably occur together let you anticipate behavior, but they still do not show cause. Only when you change one variable and watch behavior change with it, reliably and with other explanations ruled out, can you say you know what is responsible.
This matters every day in practice. Treatment decisions, conversations with caregivers, and claims about what "worked" all rest on which level your evidence supports. Questions on this task typically give a short vignette and ask you to name the level a study or a clinician has reached, or to spot an overstated claim.
Key ideas
- Description is the first level. You make systematic observations, then quantify and classify them. The result is a collection of facts about behavior and its context. Descriptive facts often suggest research questions and hypotheses for later study.
- Prediction is the second level. Repeated observation shows that two events consistently covary: when one occurs (or changes), the other tends to occur (or change) too. This covariation is a correlation, and it lets you predict one event from the other.
- Correlation is not causation. Two events can covary because a third variable affects both, or by coincidence. Prediction is useful, but it cannot tell you which event, if either, is responsible for the other.
- Control is the highest level. You show that a specific change in behavior is reliably produced by a specific manipulation of an environmental variable, and that other factors are unlikely to explain the change. That demonstration is a functional relation.
- Control comes from experimentation. Getting there requires deliberately manipulating an independent variable, measuring the dependent variable repeatedly, and replicating the effect. This is the job of experimental analysis.
- Each level builds on the last. You cannot predict without good descriptive data, and a demonstration of control depends on the same careful measurement.
Definitions
- Description: systematic observation of behavior and related events that is quantified and classified, producing facts that can be examined and compared.
- Prediction: the level of understanding reached when repeated observations show that two events reliably covary, so the occurrence of one can be anticipated from the other.
- Correlation: a consistent relation between two events in which changes in one go along with changes in the other, without showing that either causes the other.
- Control: in a scientific sense, demonstrating that manipulating one variable reliably changes another, and that extraneous variables are unlikely to explain the change.
- Functional relation: an experimentally demonstrated relation in which a specific change in the dependent variable is reliably produced by manipulating the independent variable and is unlikely to be the result of confounding variables.
Examples from a session
One question, three levels
- Description: Jae-won, a BCBA in an outpatient clinic, records each time Mia, age 4, leaves the table during snack across two weeks. She notes how often it happens, how long Mia is away, and what was happening at the time.
- Prediction: Looking across those records, Jae-won sees that table-leaving happens far more often when snack runs longer than ten minutes. She can now anticipate the behavior, but she does not yet know why it occurs.
- Control: Jae-won shortens snack time for several sessions, returns to long snacks, and then shortens it again. Table-leaving drops each time snack is short and rises each time it is long. Because the effect was produced and replicated by her manipulation, she has evidence of a functional relation.
The same progression works in other settings. A school team might describe how often a 15-year-old student starts assignments, notice that starting is more likely after a check-in with a teacher, and then test a planned check-in in some class periods and not others. An adult day program might go from describing a client's requests for help to showing that a picture-based request card changes how often he asks.
Common mistakes / exam traps
- Treating a correlation as proof of cause. Records showing that behavior is higher on certain days support prediction only. Without a manipulation, a third variable could explain both events.
- Reading "control" as coercion. In science, control means that the analyst has shown which variables a behavior depends on. It does not mean forcing a person to behave a certain way. Showing control is what makes an intervention reliable and accountable.
- Skipping description. A study can only be as good as its measurement. Poorly defined or inconsistently recorded behavior undermines both prediction and control.
- Mixing up the levels with design logic. Single-case designs also use the word "prediction" (anticipating what baseline would look like). On a question about the goals of science, prediction means the level based on covariation.
Exam trap
Prediction vs. control. Ask one question to place a vignette: did anyone deliberately change a variable and see behavior change reliably with it? If not, the highest level possible is prediction, no matter how strong or consistent the relationship looks.
Exam trap
One manipulation vs. a demonstrated functional relation. The trap: a clinician introduces an intervention once and behavior improves, and an option calls that "control." A single before-and-after change could also come from history or maturation. Control needs the effect replicated (withdraw and reintroduce, or repeat across settings or people) so other explanations become unlikely.
Quick check
A researcher has repeatedly found that a student's on-task behavior is higher on days with a substitute teacher. Which next step would allow her to reach the level of control?
Related topics
- Philosophical assumptions: selectionism, determinism, empiricism, parsimony, and pragmatism (A.2)
- Explaining behavior through radical behaviorism (A.3)
- Behaviorism vs. EAB vs. ABA vs. science-guided professional practice (A.4)
- The dimensions of applied behavior analysis (A.5)
- Independent vs. dependent variables (D.1)
- What makes a design single-case: repeated measures, prediction, verification, replication (D.4)