A parametric manipulation and meta‐analysis of target‐response punishment on resurgence

Martinez-Perez et al. (2024) asked a simple question: does adding punishment during DRA stop the behavior from coming back later? They ran three lab experiments with college students pressing buttons for points.

Each person first earned points by hitting the 'target' button. Next, the target paid nothing while a new button paid big. During this DRA phase, some people also lost points every time they hit the target. Later, both buttons paid nothing. The team watched to see if the punished target response would 'resurge.'

More punishment meant fewer target presses during DRA. Heavy fines knocked the response down to almost zero. But when reinforcement stopped later, the behavior bounced back no matter how tough the fine had been.

A meta-analysis of 18 similar studies showed the same pattern. Punishment cuts responding in the moment, yet resurgence still happens about half the time.

Fahmie et al. (2018) found that DRA alone can prevent problem behavior from starting. The new study flips the timeline: it shows punishment plus DRA cannot stop the behavior from returning once it has been reinforced before.

Hartmann et al. (1979) showed that signaled extinction keeps pigeons pecking. The human data now echo that idea: even clear fines do not break the underlying Pavlovian link between the response and its old reinforcer.

Dougherty et al. (1996) tweaked reinforcement size with rats and saw mixed effects. The 2024 team tweaks punishment size with humans and also finds mixed long-term payoff, underscoring that parametric changes often matter less than we hope.

If you use response-cost during DRA, do it to cut immediate harm, not to prevent future resurgence. Plan for resurgence anyway: program extra practice of the alternative response, thin the original reinforcer slowly, or add an extinction burst plan. Punishment helps today; it does not inoculate tomorrow.