Behavioral contrast and response independent reinforcement.

Blue et al. (1971) worked with pigeons in a two-key box.

One key always paid off for pecking.

The other key first paid for pecks, then either stopped paying (extinction) or kept paying no matter what the bird did (response-independent).

The team watched whether pecks on the first key sped up—classic behavioral contrast.

Contrast appeared only when the second key went to extinction.

When the key still paid but no longer required pecks, contrast vanished.

The result says the contingency change, not the drop in pecks, drives contrast.

Striefel et al. (1974) seems to disagree: they got contrast even when extinction barely cut responding.

Look closer—both studies still need the extinction contingency; the 1974 birds just reached it with fewer errors, so the papers actually match.

Kodera et al. (1976) stretched the idea further: after errorless training that almost removed mistakes, contrast still showed up, proving extinction rules even with gentle procedures.

Rodewald (1974) zoomed in on the pecks themselves, showing some are automatic stimulus-elicited reactions rather than chosen operants—helping you read data charts more carefully.

When you thin reinforcement or introduce extinction, expect bursts or drops in other behaviors—true contrast needs a real contingency shift, not just fewer responses.

If you use non-contingent reinforcement to reduce problem behavior, watch other responses; without a clean extinction component you may not see contrast, but with it you might see sudden spikes elsewhere.

Plan extra measurement and caregiver warnings any time you remove a reinforcer completely.