Abstract
We often need to make judgments about one feature of a stimulus amid variation across other features. For example, in vision, the shape of objects must be estimated despite changes in orientation or lighting. In hearing listeners face a similar challenge when discriminating changes in pitch amid variations in timbre, as when hearing melodies played across different instruments. Previous results have suggested that listeners are influenced by irrelevant changes in timbre when asked to discriminate the pitch of two tones. However, timbre does not appear to be retained in memory as robustly as pitch, so adding a memory delay decreases this bias. This leads to a counter-intuitive performance improvement across a delay (McPherson & McDermott, 2022). Here we tested whether similar effects are observed for vision. Participants (N=91) completed a shape discrimination task. On each trial, participants were presented with two shapes from a continuous shape space (Li et al. 2020) for 200ms each and were asked whether the two shapes were the same or different. The color of the shapes could either be the same or different, and shapes were either presented in series or with a 2000ms delay between onsets. Our hypothesis was that interference from task-irrelevant color changes would decrease across a delay. However, we found that changes in color were similarly likely to bias participants to report changes in shape regardless of delay (no interaction between delay and color: F(1,90)=1.59, p=0.21). In the case of shape/color stimuli with a shape discrimination task, it appears that task-irrelevant color information is retained across a delay, and continues to interfere with shape discrimination. Overall, this raises questions about the similarity of memory representations across vision and audition. It also suggests limits on the invariance of visual object representations to changes in task-irrelevant features, which commonly occur in real life.