The equal-variance signal detection (EVSD) model predicts superiority of two-alternative forced-choice (2AFC) detection over yes-no (YN) detection by a factor of √2. To make balanced comparisons between detection in these tasks, the equation for calculating 2AFC detection involves a division by √2. While the detection literature confirms this prediction (Wickelgren, 1968), the prediction sometimes fails when the model is extended to discrimination tasks (Creelman & Macmillan, 1979). Nevertheless, this model has been widely used in recent years to contrast discrimination in YN and 2AFC tasks. Three experiments tested the √2 prediction under conditions that previous research suggests are theoretically ideal for the use of EVSD in discrimination measurement; the √2 prediction failed across all three experiments. The present results challenge previous assertions that the EVSD model may be appropriate for discrimination under the present circumstances. The implications of these findings for the study of discrimination are discussed.