Persistence and relapse of reinforced behavioral variability

Galizio et al. (2018) used a lag schedule to pay pigeons only when they produced new four-peck sequences.

After the birds learned the rule, the team stopped paying for variety and later removed pay for any alternative behavior.

They tracked whether the birds kept varying or fell back into old patterns.

The birds kept making new sequences even when variety no longer paid.

When the only other paid option ended, the new sequences bounced back strong.

The authors say this shows variability itself can act like an operant — it obeys the same extinction and resurgence rules as pressing a lever.

Doughty et al. (2015) saw the same pattern earlier: pigeons varied more when the rule said "be different" than when it just said "switch."

Dugdale et al. (2000) and Hopkinson et al. (2003) stretched the idea to kids — teens with autism and depressed college students both produced wider response sets once variability earned tokens.

Yet Nergaard et al. (2020) push back. Their review claims the extra variety is not a new operant — it is just the old responses dying out under extinction.

The two views clash on paper, but the data sets differ: Galizio used full extinction and resurgence tests; the review pooled studies that rarely ran those checks.

If variability is an operant, you can reinforce it directly. Program a lag schedule when you want a learner to try new ways to solve a task — whether that is a new play sequence, a new sentence, or a new route through a worksheet.

Watch for resurgence: when you stop reinforcing the new variation, it may dip, then spike again once other sources of reinforcement end. Plan extra practice or signals during those dips to keep the skill alive.