Control as a scientific goal

Description, prediction, and control answer different questions

The BACB BCBA Test Content Outline, 6th edition identifies description, prediction, and control as goals of behavior analysis as a science.

• Description: What behavior and relevant events occurred, where, when, and with what measurable properties?
• Prediction: Under which conditions is the behavior more or less likely, based on repeated covariation?
• Control: Does deliberate and credible manipulation of a variable produce systematic change in the measured behavior?

Description supports prediction, and prediction helps identify a possible relation to test. A reliable correlation can guide action without establishing control. Experimental evidence becomes stronger when the outcome repeatedly changes with the independent variable while plausible alternatives are addressed.

The outline is certification-examination content rather than a research protocol or authority to intervene. Actual design and service decisions remain subject to competence, consent, ethics, law, setting rules, and the person's priorities.

A functional relation is the core claim

The independent variable is the condition deliberately manipulated. The dependent variable is the measured outcome. A functional relation is supported when a credible design shows that changes in the independent variable reliably produce changes in the dependent variable for the case and conditions studied.

The WWC Single-Case Design Technical Documentation organizes this logic around prediction, repeated manipulation, and visual analysis. In common single-case terminology:

1. Prediction: Baseline supports an estimate of what responding would look like without the new condition.
2. Verification: Withdrawing or withholding the independent variable, when safe and feasible, tests whether the baseline prediction remains credible.
3. Replication: Reintroducing the condition or demonstrating the effect at another time, behavior, setting, or participant tests whether the relation occurs again.

The exact logic varies by reversal, multiple-baseline, multielement, changing-criterion, or another design. No single sequence fits every question. A design must match the expected onset, reversibility, carryover, available observations, and ethical limits.

Control belongs to the relation, not to ownership of a person

Scientific language can sound dehumanizing when it is detached from context. The analyst manipulates an environmental variable or arranges a comparison. The person retains legal rights, preferences, communication, agency, relationships, and the ability to accept, decline, pause, or withdraw when applicable.

A change can be experimentally controlled and still be unacceptable. An intrusive event may influence behavior while creating pain, fear, exclusion, dependence, or loss of access. Ethical evaluation asks whether the goal matters to the person, the procedure is proportionate, alternatives were considered, consent and assent are valid, and benefits and adverse effects are monitored.

The classic Baer, Wolf, and Risley paper placed analytic evidence alongside applied importance, directly measured behavior, replicable procedures, conceptual grounding, meaningful effect, and generalized outcomes. Experimental control supplies one dimension. That evidence cannot repair a trivial goal, inaccessible plan, or harmful procedure.

Experimental control and procedural control differ

Experimental control concerns whether the design isolates a functional relation. Procedural control concerns whether the assigned conditions were implemented as defined. Weak procedural integrity can destroy the intended comparison, while perfect integrity cannot eliminate every rival explanation.

Measure the independent variable and dependent variable separately. Record implementer, setting, materials, timing, dose, prompts, opportunities, exclusions, health and medication changes, and concurrent services. Calibrate observers and sample agreement when interpretation depends on their scores.

History, maturation, measurement drift, sequence, carryover, unequal exposure, and phase-aligned staffing changes can compete with a causal account. Inspect raw data for level, trend, variability, immediacy, overlap, and consistency. Report uncertainty when the pattern is ambiguous.

A fictional demonstration of control

Mina is a fictional adult employee who chooses to test a reminder package for beginning a preferred inventory task. An eligible opportunity is a scheduled inventory block with materials available. Starting means opening the current inventory record and scanning the first item within two minutes, without a person prompting.

Across ten baseline opportunities, Mina starts within two minutes in 3 of 10. A chosen phone alert plus a protected two-minute transition is introduced, and she starts in 8 of 10. During an agreed brief return to the ordinary calendar notice, she starts in 4 of 10. The package returns, and she starts in 9 of 10. Staff and systems implement the assigned condition in 40 of 40 opportunities.

The outcome repeatedly moves with three condition changes, supporting a functional relation for the reminder package, task, and measured window. Because the alert and transition changed together, the evidence does not identify either component as independently necessary. It also does not establish long-term maintenance, generalization, work quality, or benefit for another person. Mina's reported usefulness and burden remain separate outcomes.

Ethical feasibility sets the boundary

Some variables cannot be manipulated safely or acceptably. Never withdraw food, water, bathroom access, AAC, mobility, prescribed care, pain care, emergency help, or effective safety supports to demonstrate control. Do not provoke dangerous behavior, recreate trauma, or prolong deterioration for a clearer graph.

The current BACB Ethics Code applies to BCBA and BCaBA certificants and people who completed an application. It addresses competence, client involvement, informed consent and assent when applicable, medical needs, risk, data, and continual evaluation. Immediate safety, assent withdrawal, new health information, or adverse effects can require a pause or stop even when another replication was planned.

When experimental control is infeasible, use the strongest ethical design available and describe the inference accurately. Useful clinical improvement can still matter. The report should separate observed change, predicted relation, experimentally demonstrated control, generality, and the person's experience.