Reinforcer control and human signal-detection performance.

Panlilio et al. (2000) ran adult volunteers through a signal-detection game. On each trial a faint tone was either present or absent. The adults pressed one key for "tone" and another for "no tone."

Half the time the researchers fixed the ratio of reinforcers at 50:50. The other half they let the ratio drift with the participant's own response pattern. They drew ROC curves to see how each setup shaped bias.

When the reinforcer ratio was locked at 50:50, the ROC curve slid left or right exactly where signal-detection theory said it should. The adults showed a clean criterion shift.

When the ratio floated, the curve wobbled. The neat bias prediction fell apart. Uncontrolled pay-offs muddied the measurement.

Reberg et al. (1979) showed the same story in pigeons: fixed reinforcer ratios shifted bias while leaving true discrimination untouched. Panlilio et al. (2000) now proves the rule holds for humans.

Alsop et al. (1995) used human matching-to-sample and also saw bias track the pay-off ratio. Their data line up with the new finding; the procedure changed but the principle stayed.

Lie et al. (2010) swapped reinforcers for punishers in the same signal-detection setup. Punishment frequency pulled bias the opposite way, yet discriminability stayed constant. Together the two papers bracket how consequence ratios steer choice without touching sensory acuity.

If you run conditional-discrimination probes or signal-detection tasks with learners, lock your reinforcer ratios across choices. Letting the child's own response rate set the pay-off will smear your bias data and hide true stimulus control. Program equal reinforcement or track every reinforcer delivery; your ROC curve—and your treatment decisions—will be clearer.