Punishment and dynamic choice: Assessment of the direct‐suppression model
Punishment guides choice toward richer or safer options instead of wiping out the punished response.
01Research in Context
What this study did
Fontes and team tested pigeons in a two-key chamber.
Each key paid food on its own changing schedule.
Sometimes pecks also produced a quick shock.
The shocks came at the same rate on both keys, or at different rates.
The birds could switch keys at any time.
Researchers watched which key the birds chose over many sessions.
What they found
When shocks were equal on both keys, birds still pecked more on the side that paid food faster.
They did not stop pecking just because shocks were present.
When one key had more shocks than the other, birds moved away from the harsher side.
The old direct-suppression model said shocks should cut all responding.
That did not happen.
Choice followed relative shock rate, not absolute shock rate.
How this fits with other research
Deluty (1976) saw the same shift in rats.
When one lever had more shocks, rats moved to the other lever.
Fontes et al. (2025) shows this still holds when the food schedules keep changing.
DARDANO et al. (1964) used harder and harder food ratios instead of changing food timing.
They also found punishment pushed animals to the safer option.
All three studies agree: punishment steers choice, it does not simply stop it.
Why it matters
If you use mild punishment in a token system or classroom, do not expect the behavior to vanish.
The learner may just switch to another task that pays better or hurts less.
Check what other choices are available before adding a punisher.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →List every task or response your client can do in the session, then note which ones pay better or feel safer before adding any punisher.
02At a glance
03Original abstract
The effects of punishment rate on response allocation were investigated using a choice procedure where relative reinforcement rates changed rapidly within session. Predictions from a modified version of the direct-suppression model were tested in two separate experiments. In both experiments, sessions were composed of seven unsignaled components, each programming a different reinforcement ratio. In Experiment 1, equal punishment rates were superimposed on the schedule of reinforcement for both responses and absolute punishment rates increased across blocks of sessions. Punishment increased preference for the richer schedule of reinforcement, but the degree of the preference shift was not a function of absolute punishment rates. In Experiment 2, unequal punishment rates were superimposed on the schedules of reinforcement for both responses and relative punishment rates changed across blocks of sessions. Response allocation shifted away from the richer punishment schedule creating a bias toward the option associated with less frequent punishment. The results from both experiments challenged the predictions of the direct-suppression model. Furthermore, fits of the generalized matching law to the data indicated that superimposition of equal or unequal punishment schedules on responses maintained by unequal reinforcement schedules differentially affect the values of sensitivity and bias.
Journal of the Experimental Analysis of Behavior, 2025 · doi:10.1002/jeab.4237