Abstract
Global motion perception is studied by showing moving dots, but the way that these dots are generated varies substantially across studies. We have recently shown that when the direction of each dot is generated from a Gaussian distribution, global direction perception is achieved by averaging local motion signals, as is assumed in standard models of motion perception. However, with a more typical stimulus in which dots moving in the same direction (signal) are intermixed with randomly moving dots (noise), participants actively segregate signal from noise, and then identify the global direction based on signal dots alone. This finding raises a critical question: If you can pull apart signal from noise under some stimulus conditions but not others, does your ability to monitor internal sensory uncertainty in motion representations also depend on whether or not signal and noise can be segregated? We measured conscious access to sensory uncertainty by having participants rate their confidence in the accuracy of motion direction judgments. The motion stimuli were manipulated across various levels of difficulty and stimulus type. For stimuli in which direction identification relied on averaging local motion, confidence tracked with a measure of local averaging accuracy. However, for stimuli in which signal and noise could be segregated, confidence tracked with a measure of whether or not segregation was successful, but had no relationship with measures of local averaging. Taken together, these results show that global motion perception relies on different mechanisms depending on the stimulus type. Moreover, conscious access to uncertainty in the underlying representations likewise depends on stimulus type. Therefore, the details of how stimuli are constructed can drastically alter not only how global motion is processed, but also the nature of our conscious access to these processes.