4.1 Continuous vs. Intermittent Schedules of Reinforcement

4.1 Continuous and Intermittent Reinforcement

Quick Answer: Reinforce densely while establishing a behavior, then thin only when performance remains useful and the learner stays engaged. Continuous reinforcement is one reinforcer for each response meeting criterion. Intermittent reinforcement delivers consequences for only some eligible responses. There is no universal 80–90 percent fluency threshold or 20–30 percent thinning formula.

Continuous Reinforcement

On FR1, every defined response produces reinforcement. This can clarify a new contingency and provide many successful repetitions. It is not literally required for every acquisition. Captured behavior may contact reinforcement irregularly; natural opportunities may be sparse; shaping changes criteria; and practical teaching can combine food, access, play, and social outcomes.

The important features are contingency, timing, attainable criteria, and adequate reinforcement. “Every response” means every response that meets the stated criterion, not every approximation forever.

Intermittent Schedules

Ratio schedules depend on number of responses; interval schedules make the first qualifying response after time eligible. Fixed schedules use predictable requirements; variable schedules vary around a programmed value. In applied training, real schedules often combine features and environmental reinforcers.

Intermittent reinforcement can support maintenance when behavior is already established and can make responding more persistent when reinforcement later stops. The partial-reinforcement extinction effect is an experimental pattern, not a promise that an FR1-trained dog stops in five trials or a variable-ratio dog persists for thousands.

Thinning a Schedule

Define the desired real-world schedule first. A recall may need reinforcement often because it competes with valuable environmental activity; settling may be maintained partly by comfort and ongoing context. Do not thin merely to eliminate food.

Use data such as latency, success under defined conditions, rate, duration, error pattern, engagement, and recovery. Change one dimension at a time when practical: repetitions per reinforcer, delay, duration, distance, distraction, or reinforcer magnitude. If performance declines, return to an easier condition and reassess.

Fixed numeric rules can mislead. Eight correct trials out of ten may reflect chance, prompting, easy conditions, or a criterion that is too broad. A five-to-seven-step increase is not universally safe just because it is less than 30 percent. Select increments from the individual trend.

Ratio Strain and Lean Schedules

When response requirements become too costly relative to reinforcement, behavior may slow, become variable, or stop. The learner may disengage, seek other outcomes, or show frustration. This is commonly called ratio strain. It is a performance description, not a medical diagnosis.

Increase reinforcer value, reduce effort, lower the ratio, improve timing, simplify distractions, or check health. A return to FR1 may help but is not the only remedy. Avoid punishing disengagement created by an unrealistic schedule.

Duration and Differential Schedules

For duration, reinforce achievable intervals and vary them around a level the learner can meet rather than extending predictably until failure. Differential reinforcement schedules reinforce specified alternatives, incompatible behavior, lower rates, higher rates, or omission during a period. Each requires a clear definition and safety analysis.

Natural and Life Reinforcers

Transfer does not mean removing reinforcement. Everyday outcomes can maintain behavior: a loose leash gains access to sniffing, stationing gains a chew, or waiting opens a door. Continue occasional deliberate reinforcement and monitor whether behavior remains reliable in the required contexts.

Schedule Examples and Misclassification

An FR3 requires three responses for reinforcement; a VR3 varies around an average requirement. An FI30 makes the first qualifying response after thirty seconds eligible, while a VI30 varies the interval around an average. These laboratory labels are useful only when the actual arrangement matches them. Reinforcing “about every few good responses” without programmed values is intermittent but may not be a true variable-ratio schedule. A jackpot changes reinforcer magnitude, not necessarily the schedule. Duration, delay, and response effort can change at the same time, so record them separately. In daily life, a behavior may contact multiple schedules from different people and the environment.

Avoid gambling analogies that imply variable schedules automatically create compulsion or are always strongest. Schedule effects depend on the response, reinforcer, competing opportunities, and history. The professional goal is reliable useful behavior with adequate welfare, not maximal persistence after reinforcement disappears.

Designing the Plan

Record what response qualified, which consequence occurred, schedule, context, and outcome. Plan what happens after an error and how the client will deliver the schedule consistently. Thin gradually enough to preserve success, but let evidence—not folklore—determine the exact step.