The extinction burst: Effects of reinforcement magnitude
Rich reinforcement histories create bigger extinction bursts, so scale your prompts and alternative rewards accordingly.
01Research in Context
What this study did
Shahan et al. (2025) worked with rats to see how the size of past rewards changes the extinction burst. One group got one food pellet for each lever press. The other group got six pellets. After the rats learned the routine, the team stopped all pellets and watched what happened.
They used a single-case design and tracked every press minute-by-minute. The goal was to test if bigger past rewards create a bigger spike when rewards disappear.
What they found
The six-pellet group showed a much larger extinction burst. Their lever pressing jumped higher and lasted longer once food stopped. The one-pellet group had a smaller, shorter spike.
A matching-law model fit the data well. It predicted that richer reinforcement histories would produce bigger bursts, and the numbers matched the prediction.
How this fits with other research
Shahan et al. (2025) extends its own lab work. A companion paper by the same authors showed that giving six pellets as an alternative reward during extinction shrinks the burst. Together the two studies tell a clear story: big rewards before extinction worsen the burst, but big rewards during extinction can soften it.
Lambert et al. (2024) found the same pattern with adults who have developmental disabilities. Higher reinforcer value led to larger bursts, backing up the rat data across species and settings.
KELLEHER et al. (1963) first showed that more continuous reinforcement creates bigger bursts in both rats and children. The new study adds the twist that pellet size, not just number of trials, matters.
Muething et al. (2024) looked at 108 clinical cases and saw bursts only 24 % of the time. The lab studies remind us that when bursts do happen, their size depends on the reinforcement history we bring into treatment.
Why it matters
If your client has been getting large or frequent reinforcers, plan for a stronger extinction burst. Add extra prompts, offer dense reinforcement for alternative behaviors, or stretch the schedule before you stop rewards. The matching-law model gives you a ruler: the richer the history, the bigger the planned buffer you need.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Before you start extinction, list the last three reinforcer sizes or schedules the client received; if they were large or dense, pre-load extra alternative reinforcement for the first five minutes.
02At a glance
03Original abstract
The extinction burst is an increase in an operant behavior early in the transition to extinction. A matching-law-based quantitative theory suggests that it results from the elimination of competition from reinforcement-related behavior that accompanies the transition to extinction. This experiment examined the effects of reinforcement magnitude on the extinction burst with rats. Responding produced either 1 pellet or 6 pellets prior to a within-session transition to extinction. Baseline response rates were lower with 6 pellets than with 1 pellet but increased more in the first minute of extinction following 6 pellets. Furthermore, an order effect was observed where rats having first experienced extinction following 1 pellet showed a smaller extinction burst when subsequently trained with 6 pellets—an outcome resulting from a faster deceleration of responding during extinction. The model described the data well, including the order effect, when augmented to include the potential asymmetrical influence of learning to discriminate the continued absence of different magnitudes of reinforcement. We conclude that the approach holds promise by formally incorporating the extinction burst into the corpus of matching theory and serving as an example of the utility of better aligning basic research methods and theorizing with areas of applied concern.
Journal of the Experimental Analysis of Behavior, 2025 · doi:10.1002/jeab.4238