Reeve, K. F., & Fields, L. (2001).
Perceptual classes established with forced-choice primary generalization tests and transfer of function.
Journal of the Experimental Analysis of Behavior,
76, 95-114.
In Experiment 1, 20 college students learned two identity
conditional discriminations using squares that differed in
interior-fill percentage (called Fill23 and Fill77). A two-choice
generalization test was then presented with number of test trials
varied across groups of subjects. The test samples were 19
squares that ranged in fill value from 23% to 77%; the
comparisons were squares with Fill23 and Fill77. The resulting
gradients did not vary as a function of number of test trials.
When the generalization test was repeated with a third
comparison, "neither," the ranges of fill values that
occasioned the exclusive selection of Fill23 or Fill77 were
direct functions of the number of prior two-choice generalization
trials. Finally, a discriminability test revealed that Fill23 and
Fill77 were discriminable from the intermediate fill values. In
Experiment 2, perceptual classes were established with 5 new
students using 760 forced-choice generalization test trials. The
students were then trained to select a different glyph in the
presence of Fill23 and Fill77, followed by a three-choice
generalization test in which the 19 fill stimuli served as
samples and the two glyphs served as comparisons. The gradients
overlapped with those previously obtained during the three-choice
generalization test that used Fill23 and Fill77 as comparisons.
Finally, a discriminability test showed that many adjacent
stimuli along the fill dimension were discriminable from each
other. Together, the results of both experiments suggest that
ranges of fill-based stimuli functioned as members of perceptual
classes, and each class also functioned as a transfer network for
a new selection-based response.
Key words: perceptual classes, generalization, discriminability,
natural categories, keyboard press, adult humans