STIMULUS GENERALIZATION AND THE RESPONSE-REINFORCEMENT CONTINGENCY.

HEARSKELLEHER et al. (1964) tested how different reinforcement schedules shape stimulus generalization. They trained pigeons to peck a key under three schedules: DRL, short VI, and long VI. After training, they presented colors that gradually shifted away from the trained hue and recorded how responding spread across those new colors.

The goal was to see whether the schedule itself changes how sharply the birds discriminate between similar stimuli.

DRL and longer VI schedules flattened the generalization gradient. That means the birds pecked almost as often at colors that looked different from the training color. The schedule had become a stronger cue than the color itself.

Shorter VI kept the peak sharp; most pecks stayed near the original hue.

Powell et al. (1968) later showed that simply adding more discrimination sessions steepens the gradient again. Their work acts as a successor: once you know schedule matters, you can override the flattening by giving extra training.

Okouchi (2003) extended the same gradient idea to college students using line lengths and got the same asymmetric curve, proving the effect is not just for pigeons.

Wesp et al. (1981) pulled these threads together in a big review. They list schedule type as one of the key levers you can turn to make stimulus control tighter or looser.

HOFFMAN et al. (1964), from the same year and lab, swapped the schedule for an aversive contingency and still saw gradient changes, showing the effect holds across reinforcement and punishment setups.

When you want tight stimulus control—say, a child to respond only to the exact word “cat” and not “bat” or “cap”—avoid DRL or very lean VI schedules during teaching. Instead, use richer VI or add extra discrimination trials like Powell et al. (1968) to steepen the gradient. When you need broader generalization, a DRL or stretched VI can help responses spread naturally across similar stimuli.