The Problem of Common False-Alarm Rates and the Informational Asymmetry When Comparing Pure and Mixed Lists in Recognition Memory: Reconsidering the List-Strength Effect

The list-strength effect (LSE)—the interaction between item strength and list composition (pure versus mixed)—is theoretically diagnostic in models of memory. In contrast to free recall, in old-new recognition, the LSE was long considered negligible. This dissociation prompted the development of new recognition models. Recent work, however, suggests that the magnitude and direction of the LSE depend on the strength manipulation. Here, we argue that conclusions about the LSE are undermined by two methodological limitations in pure-mixed comparisons: a design flaw and an informational asymmetry. Sensitivity measures in recognition memory depend jointly on hit and false-alarm rates (HRs and FARs), which co-vary when sensitivity, response bias, or both change. When strength conditions differ only for studied items, sensitivity estimates from pure and mixed lists are not constructed from equivalent components. This is because traditional mixed-list designs, which impose a common FAR across strength conditions, are compared to pure lists wherein FARs can vary freely. Critically, there is no contingency in which given differences in HRs, a common FAR of any magnitude would lead researchers to attribute those differences to bias rather than to sensitivity—regardless of which sensitivity measure is used. We consider two possible arguments in favor of common FARs and refute both. Using a hypothetical reanalysis of recent LSE data, we show that empirically plausible changes in FAR—without altering observed HRs—can produce LSEs of any direction. We conclude that pure-mixed comparisons cannot adjudicate between theories of recognition memory. Alternative experimental designs are then discussed.