Abstract
Microsaccades are thought to be a measure of covert attention. However, the relationship between attention and microsaccades is often obscured because studies probe the microsaccade-attention relationship by presenting a cue on each trial that is highly predictive of target location (~100% cue validity). This approach tests an extreme version of cueing at a single location. If microsaccades truly measure covert attention, then the microsaccade-attention relationship should generalize to more naturalistic settings where cue validity is dynamic and cues are stored in memory. We analyzed microsaccades from sessions with blocks of trials with remembered, less predictive cues. Monkeys performed an orientation change-detection task using two oriented Gabors presented simultaneously. Monkeys fixated and reported an orientation change by saccading to the changed stimulus. Attention was cued to a stimulus location using a visual cue on instruction trials at the start of each block of trials. During cued blocks, the change occurred with 80/20% probability at the cued/uncued locations. During neutral blocks, visual cues were presented at both stimulus locations during instruction trials, and the change occurred with 50% probability. We investigated how the directions of microsaccades were modulated during cued and neutral blocks. Unlike previous results, we found that monkeys did not direct microsaccades towards the target locations. Instead, monkeys made microsaccades towards the midpoint between the two targets. Nevertheless, in sessions consisting of only cued blocks, attentional cueing biased microsaccade direction toward the cued locations. The bias in direction was weaker when sessions consisted of both cued and neutral blocks. We also saw no difference in microsaccade direction in detected versus missed trials, despite robust changes in behavioral performance and neuronal activity. Our results suggest that microsaccades may not correlate with the locus of attention in more naturalistic settings that involve multiple remembered cued locations with variable cue validity.