Abstract
We examined how practice alters response bias in a yes-no visual detection task. Observers detected band-limited textures in two levels of static white Gaussian noise over two consecutive days. Half of the observers saw the same textures on day 2 and half saw novel textures. The textures on signal-present trials were presented at several contrasts using the method of constant stimuli: signal-present and signal-absent trials were equally likely. The data were used to estimate global and local response criteria. The global criterion (Jones et al, 2015) was defined using hit rates at all contrasts, whereas local criteria were measured for the two contrasts corresponding to hit rates of 70% and 80% on Day 1 (Wenger and Rasche, 2006). Sensitivity in both groups of observers increased across days. We also found that practice had small effects on response bias: on average, observers made fewer false alarms on Day 2 than Day 1. For the global measure of criterion, this shift in response bias meant that observers were, on average, closer to the criterion that maximized the percentage of correct responses. The effects on the local measures of criterion were more variable across groups and noise levels. In summary, practice shifted the response criterion in a direction that reduced false alarms and increased response accuracy.