Reinforcement

Reinforcement is a relation, not an object

People often call a sticker, snack, compliment, or break “reinforcement.” Technically, reinforcement is the measured relation; a reinforcer is the consequence shown to produce the strengthening effect.

A defensible claim names the response class and measure, consequence and contingency, conditions and supports, and a baseline or comparison showing that later responding increased or was sustained.

For an opportunity percentage, the numerator is eligible opportunities with the response at the defined level, and the denominator is all predeclared eligible opportunities. Report exclusions by reason. If none occur, report 0 eligible opportunities; percentage not calculated, not 0%. For rate, divide response count by observation time. For a duration percentage, divide time engaged by observation time.

If praise follows a response but later responding does not improve relative to an appropriate baseline or comparison, the data do not establish praise as a reinforcer. A flat trend alone cannot establish maintenance unless a comparison shows responding would otherwise have declined.

Positive and negative reinforcement both strengthen behavior

• Positive reinforcement: An event is added after a response, and later responding increases. Asking a question produces useful information, and asking increases in similar situations.
• Negative reinforcement: An event is removed, reduced, postponed, or prevented, and later responding increases. Closing a door reduces hallway noise, and door closing increases when that noise occurs.

“Positive” does not mean kind, and “negative” does not mean bad. Goals, consent, access, risks, and results still matter. Baron and Galizio describe ambiguity in the addition-versus-removal distinction, so records should name the actual change and measured effect.

Reinforcement, reward, preference, and bribery are different ideas

Lill, Shriver, and Allen synthesized 65 peer-reviewed articles into SPADS, a model for selecting context-specific stimulus preference assessment procedures. Their model considers client, stimulus, setting, administration-time, and motivating-operation variables. A preference assessment predicts stimuli that may function as reinforcers; it does not demonstrate that contingent access will increase a particular response. A motivating operation can temporarily alter a consequence’s effectiveness and related responding, so preference results should not be treated as fixed.

A fictional worked example

Sofia is a fictional seven-year-old who uses gestures and a speech-generating device. She chooses a goal of joining her sibling’s block play. Before collecting data, the team defines an eligible opening as one in which Sofia selected blocks from an activity menu, her sibling is available, and both the device and invitation card are within reach. Refusal, sibling unavailability, and missing communication access are logged by reason before invitation responding is scored.

The target response is independently selecting “play?” or handing over the card. The planned partner response is acknowledgment within ten seconds and, when both children still choose it, access to shared play. A clear “later” counts as acknowledgment, not as shared-play delivery.

Sofia independently invites in 1 of 6 eligible baseline openings and 9 of 12 after the sibling learns the response plan. Of nine emitted invitations, eight receive timely acknowledgment and six lead to shared play. Report the learner measure as 9/12 eligible openings; partner measures are 8/9 invitations acknowledged and 6/9 followed by shared play. The learner denominator is eligible openings; each partner denominator is emitted invitations.

This one baseline-to-intervention series is preliminary evidence, not proof that the consequence caused the increase: partner training, time, and other variables also changed. It does not establish which consequence component mattered or whether the relation occurs elsewhere. Stronger claims require repeated measurement or a safe, acceptable experimental comparison.

Timing and schedules shape what happens

Immediate, reliable consequences often make a contingency easier to contact, but no universal delay applies. Record actual timing; signals, tokens, rules, and learning history can sometimes bridge delays.

Reinforcement can follow every response or a schedule. Use data and the person’s experience when thinning; abrupt changes can bring disengagement or recurrence.

When they fit the goal, natural consequences reduce reliance on contrived outcomes. Tokens or prizes require choice, clear exchange rules, monitoring, and an exit plan.

Reinforcement can be unplanned

Reinforcement need not be intentional. A consequence may be socially mediated by another person or produced directly by the response, called automatic reinforcement. A successful second click or peer laughter may strengthen repeated clicking or joking.

When an unplanned pattern is unwanted or harmful, measure partner and environmental behavior alongside client behavior. Check pain, accessibility, communication failures, task fit, and safety before changing consequences.

Ethical reinforcement starts with the person's goals

The linked Association for Behavior Analysis International Basic Principles page is a recommended prerequisite course outline within its culturo-behavior science higher-education materials, not a clinical practice standard. The current BACB BCBA Test Content Outline (6th ed.) covers positive and negative reinforcement, automatic and socially mediated contingencies, preference assessment, reinforcement procedures, motivating operations, and possible unwanted effects. It describes entry-level examination content, not permission for a case-specific plan.

The current BACB Ethics Code for Behavior Analysts applies to BCBA and BCaBA certificants and people who have completed an application for either certification. The BACB states that it has no separate jurisdiction over organizations or corporations. For behavior analysts within its jurisdiction, the Code addresses client and stakeholder involvement, required informed consent and assent when applicable, assessment- and evidence-based intervention, preference, risks and side effects, minimizing harm, data use, and continual evaluation.

Do not make ordinary meals, water, toileting, prescribed care, pain communication, emergency protection, AAC, mobility aids, or respectful interaction contingent on performance. Keep the person’s communication modes available before, during, and after teaching. ASHA’s AAC guidance emphasizes access across settings, vocabulary for refusal and basic needs, multimodal communication, and partner training. Treat refusal, distress, or changing preferences as reasons to pause and reassess whenever safety permits. Reinforcement data cannot make an unwanted or harmful goal acceptable.

Questions to ask about reinforcement

• What response class and dimension are being measured?
• Which consequence is contingent on it, and what comparison shows an effect?
• Are raw counts, denominators, observation time, exclusions, prompts, and supports reported?
• What is reported when no eligible opportunity occurs?
• Does the person want the goal, outcome, and teaching arrangement?
• Are informed consent, applicable assent, refusal, and changing preferences addressed?
• Could teaching, health, access, partner behavior, time, or another intervention explain the change?
• Are communication, basic needs, and other access supports protected outside the contingency?
• What data will guide schedule changes or stopping the procedure?