Figure 6.

Explanation of some of Newlin et al. ’s (1978, Experiment 1) data. (A) A condition in which participants were first pre-exposed to bright stimuli, and then presented a target at brightness level 4 (top, 1 = dimmest, 11 = brightest). We suppose that the subjects encodes a boundary separating two categories, with the target coded as being in the dim category. The categorical code provides additional information to the encoding of the metric value of the target brightness. On tests (bottom), the range of stimuli is more or less centered on the category boundary, so that the boundary does not shift during tests, and a “correct” generalization gradient peaking at 4 is found (Newlin et al.’s Figure 1). (B) A condition in which participants were first exposed to dim stimuli, and then presented a target at brightness 4 (top). We suppose that the participant would encode a boundary somewhere between 3 and 4. On tests (bottom), the range serves to shift the encoded category boundary toward the brighter end, thus causing large range effects (Newlin et al.’s Figure 1). The shift occurs because the boundary was somewhere in the middle of the range before tests (in pre-exposure and target training), and the subject is relying on the “middle of the range” description (a categorical code) in part in determining the boundary during tests.