Generalization gradients of inhibition after different amounts of training.

Pigeons pecked a key for food on a VI schedule. A black vertical line was always present.

The birds then learned a new rule: peck in the presence of a white key, but stop when the vertical line returns.

After 2, 4, or 8 days of this go/no-go training, each bird saw test lines tilted 5° to 45° from vertical. The team counted how often the birds pecked at each tilt.

Birds that received more training days showed sharper drops in pecking as the line moved away from vertical.

In other words, extra discrimination sessions made the "do not peck" rule stick to nearby tilts. The inhibitory gradient grew steeper.

Snapper et al. (1969) ran a near-copy study the next year. They added a group that got extinction without the S-. That group produced flat gradients, proving that differential reinforcement (S+ vs S-) is the key ingredient.

Okouchi (2003) later showed the same gradient logic works with college students. After learning which line length paid money, their response rates also peaked near the trained length.

HOFFMAN et al. (1964) looked at conditioned suppression instead of key pecking. They found that discrimination training skewed the gradient but did not produce peak shift. Together these papers show the effect holds across species and response types.

If a client responds correctly to the trained S- but also stops responding to similar stimuli you want to keep, add more discrimination trials before you probe. Extra practice tightens control and reduces over-generalization.