Abstract

In the single-interval paradigm, which is commonly used in separation discrimination and Vernier acuity studies, a single stimulus is presented on each trial, and the observer reports how that stimulus compares to a remembered referent. In separation discrimination, the referent is the average separation. In Vernier acuity, it is collinearity. The problem is that, if uncertainty about the remembered referent is large, the computed threshold can be an overestimate of the true threshold, whereas if it is small, it can be an underestimate. Our experiments were designed to determine the effect on threshold of using remembered referents in these two tasks. We compared thresholds obtained with a remembered mean (single-interval paradigm) and with a presented mean [two-interval forced choice (2IFC)]. We found that for separation discrimination, the single-interval paradigm yielded a slightly higher threshold than did the 2IFC paradigm, whereas for Vernier acuity, the reverse was true. In terms of signal detection theory, this implies that the variance of the remembered referent is smaller (relative to the variance of the test stimuli) for Vernier acuity than for separation discrimination. Given the naturalness of the Vernier acuity referent (collinearity) and the arbitrariness of the separation discrimination referent, this explanation seems plausible.

© 1988 Optical Society of America