Drug discrimination: stimulus control during repeated testing in extinction.

Rose et al. (2000) tested pigeons in a drug-discrimination task. Birds first learned to peck one key when given a drug and another key when given saline.

After training, the team ran daily extinction sessions. No food was given for correct or wrong pecks. They repeated these drug-free tests many times.

Response rate dropped across extinction days, but stimulus control stayed strong. The birds still chose the drug-appropriate key most of the time.

Counting the percentage of FR10 runs showed clearer control than simple rate. Low peck numbers had hidden the true accuracy until they used the ratio measure.

Podlesnik et al. (2016) later showed the same logic works with alternative reinforcement. They gave rich food in a separate context before mixing it with the target task. Extinction went faster, proving the 2000 finding extends to reinforcement-rich histories.

Fisher et al. (2020) applied the idea to kids with destructive behavior. Adding an S-delta during FCT extinction cut resurgence. The birds' stable stimulus control under repeated extinction told the team that a clear 'no-reinforcement' signal would hold even when response rate fell.

Shahan et al. (2020) measured resurgence as alternative reinforcement thinned. They also moved past simple rate, modeling relapse size. Their fine-grained tracking echoes the 2000 paper's switch to percentage of FR10s: both warn that raw counts can mislead.

When you run extinction, watch how you score stimulus control. If responses drop but accuracy looks shaky, switch to a consecutive-response ratio instead of total count. Add an S-delta card or tone when you probe, just like Fisher did, to keep control tight. And plan for resurgence any time you thin reinforcement—Shahan's curve tells us the relapse jumps fast. One practical step: record ten correct responses in a row before you say the learner 'has it' during extinction probes.