5.4 Extinction, Extinction Bursts & Differential Reinforcement

5.4 Extinction & Differential Reinforcement

Quick Answer: Extinction occurs when a previously reinforcing consequence no longer follows a behavior, causing responding to decrease over time under relevant conditions. It is specific to the maintaining consequence and can produce variability or temporary escalation. Pairing safe extinction components with management and reinforcement of an alternative is often more practical than “just ignore it.”

Identify the Reinforcer

If jumping has produced attention, withholding attention may be an extinction component. If barking has made a delivery person leave, the natural departure can continue reinforcement even when the owner ignores the dog. If pulling has produced forward access, stopping access changes that contingency.

Therefore extinction is not synonymous with looking away. It requires a credible functional hypothesis and control over the relevant consequence. Multiple functions may operate, and a trainer cannot withhold relief from severe fear, access to water, toileting, medical care, or another welfare need in the name of extinction.

What May Happen

At the start of extinction, behavior may increase in rate, duration, intensity, or variability; novel responses can appear. This is called an extinction burst when it occurs, not a guaranteed stage. Dangerous escalation makes pure extinction inappropriate.

Previously reinforced behavior can reappear through spontaneous recovery after time, renewal in a different context, reinstatement after the outcome occurs again, or resurgence of older behavior when a newer response no longer works. Extinction is new learning and context dependent, not erasure.

Intermittent reinforcement can increase persistence. If a client sometimes gives attention after prolonged barking, the longer performance may be reinforced. This does not justify allowing unsafe behavior to intensify while waiting it out.

Differential Reinforcement

Differential reinforcement combines reinforcement for a defined class with reduced reinforcement for another:

• DRA: reinforce an alternative, such as four paws on the floor instead of jumping.
• DRI: reinforce behavior physically incompatible at that moment, such as stationing away from the doorway.
• DRO: reinforce absence of the target during a defined interval; monitor for other problem behavior.
• DRL: reinforce a lower rate when zero is unnecessary or unrealistic.
• DRH: reinforce a higher rate when increased responding is the goal.

Choose an alternative that accesses the same or an acceptable outcome and is easier than the problem behavior in the training context. A sit may not be functional when the dog needs distance; an emergency U-turn may be.

Antecedent Management

Prevent high-risk rehearsal while teaching: use gates, distance, leashes, visual barriers, planned visitor entry, or another safe arrangement. Management does not reinforce “bad behavior”; it changes opportunity and protects participants. Combine it with practice at a workable level.

Welfare and Safety

Do not use extinction alone for aggression, panic, severe escape behavior, self-injury, or behavior maintained by relief from pain or fear. Withholding escape during intense exposure can become flooding, which current CCPDT policy prohibits. Refer for medical evaluation when health may contribute.

Monitor frustration, new behavior, recovery, and client ability. If risk or distress increases, stop and revise rather than insisting an extinction burst must be completed. Reinforcement of a replacement does not guarantee absence of frustration, but it gives the learner an available route to a useful outcome.

Building a Plan

Define target and alternative, hypothesized consequence, management, reinforcement schedule, measurement, generalization, lapse response, and stop criteria. Teach the alternative before the difficult context when possible. Coordinate everyone who controls the consequence so implementation is consistent.

Track target rate, alternative rate, intensity, latency, setting events, and relapse. When improvement occurs, continue reinforcement and practice across contexts rather than declaring the behavior erased.

Example: Attention-Maintained Jumping

Before visitors arrive, teach four paws on the floor and stationing with frequent reinforcement. Use a gate and leash transition so jumping cannot knock someone down. Ask visitors to provide attention only while the alternative occurs and to step out of reach if jumping begins, without pushing or shouting. Reinforce the alternative promptly. If jumping escalates, increase distance and return to an easier rehearsal rather than waiting through dangerous contact. Across sessions, measure jumping and stationing separately and practice with different willing visitors. This combines antecedent management, a possible extinction component for attention-maintained jumping, and DRA; it does not assume attention is the only function without data.

If the visitor’s arrival is frightening, use distance and respondent-focused work rather than treating all barking or jumping as attention seeking.