A further study of stimulus generalization following three-stimulus discrimination training.
Adding a third stimulus to discrimination training pulls the peak of generalization away from the unreinforced cue and makes the gradient steeper.
01Research in Context
What this study did
The team taught pigeons to peck a key when they saw one color, but not when they saw two other colors.
Each correct peck earned food on a variable-interval schedule.
After training, the birds saw a rainbow of colors to see where they would still peck.
What they found
The birds pecked most at colors that sat away from the no-food color.
The generalization curve rose higher and dropped faster than curves from birds that had only two-color training.
How this fits with other research
Thomas et al. (1960) ran the same lab setup with only two colors. Their curve peaked right on the reinforced color, not shifted away.
Reynolds (1968) later showed the shift happens because responding to the no-food color is suppressed, not because the food rate changed.
Locurto et al. (1980) added that the early “shoulders” next to the reinforced color fade with longer training, so the final peak you see is clean.
Why it matters
When you teach a learner to tell two stimuli apart, expect the strongest responding to move away from the unreinforced cue.
Plan extra probes near that edge to catch the shift. If you add a third stimulus, the control becomes sharper, so use three-stimulus sets when you need tight discrimination, such as teaching similar speech sounds or safety signs.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Run a quick generalization probe across five similar pictures after your three-picture discrimination program and note where responding peaks.
02At a glance
03Original abstract
A group of pigeons was trained in a Skinner box to peck for VI reinforcement when the key was illuminated by a monochromatic light of 540 or 580 mmu but were non-reinforced for responding to 560 mmu. Two control groups, differing in amount of training, received only the two positive stimuli. At the completion of training all Ss received generalization tests under extinction. Both control groups produced bi-modal generalization gradients with the peaks of responding at the S+ values. The post-discrimination gradient revealed peaks displaced from the S+ values in the direction away from the S-, low responding and increased steepness in the region of S- and a general elevation of the gradient.
Journal of the experimental analysis of behavior, 1963 · doi:10.1901/jeab.1963.6-171