Abstract
Recognition memory relies on sensitivity measures to identify memory effects (i.e., differences in sensitivity). Alas, all commonly used sensitivity measures (d′, A′, Percent correct, and Pr) have been shown to be confounded with response bias, raising serious validity concerns for the literature. da, a signal-detection measure grounded in the Unequal-Variance Signal-Detection model, has been proposed as a bias-independent alternative. Yet, establishing da as the gold standard for measuring memory effects requires demonstrating not only that it produces low rates of false discoveries, but also that it has sufficient statistical power to detect genuine differences in sensitivity. The low false discovery rate has been established with both simulated and large-scale experiments (Levi, Danino, et al., 2026; Levi, Rotello & Goshen-Gottstein, 2026). Here, we examined the estimated power of da using empirical data from 960 online participants who completed a recognition task under two sensitivity conditions (repetition manipulation). We used resampling techniques to test power levels for multiple sample sizes and found that da achieved significance rates comparable to those of the commonly-used measures, regardless of the statistical test or effect size. The typical 80% power threshold was achieved with approximately 48 participants. Together with our previous findings, these results establish da as a valid and statistically powerful measure, thus confirming the recommendation to adopt da as the measure of choice in recognition memory research.