Abstract
Noise in sensory signals can vary over both space and time. Moving random dot stimuli are commonly used to quantify how the visual system accounts for spatial noise. In these stimuli, a fixed proportion of “signal” dots move in the same direction and the remaining “noise” dots are randomly replotted. The spatial coherence, or proportion of signal versus noise dots, is fixed across time; however, this means that little is known about how temporally-noisy signals are integrated. Here we use a stimulus with low temporal coherence; the signal direction is only presented on a fraction of frames. Human observers are able to reliably detect and discriminate the direction of a 200 ms motion pulse, even when just 25% of frames within the pulse move in the signal direction. Using psychophysical reverse-correlation analyses, we show that observers are strongly influenced by the number of near-target directions spread throughout the pulse, and that consecutive signal frames have only a small additional influence on perception. Finally, we develop a model inspired by the leaky integration of the responses of direction-selective neurons, which reliably represents motion direction, and which can account for observers’ sub-optimal detection of motion pulses by incorporating a noisy decision threshold.
Highlights
Noise in sensory signals can vary over both space and time
A school of fish has strong local motion cues associated with individual animals, but to herd the population, a predator may need to determine its average velocity over time, despite continual fluctuations in the velocity from moment to moment
We show that human observers can reliably extract direction information from stimuli with low temporal coherence, even when as few as 5 frames, or 12.5% of individual dots within a 20 frame, 200 ms motion pulse, move in a target direction
Summary
Noise in sensory signals can vary over both space and time. Moving random dot stimuli are commonly used to quantify how the visual system accounts for spatial noise. In the domain of visual motion processing, perceptual and neuronal sensitivity has been characterised in detail by manipulating the coherence of random dot stimuli[1,2] In these stimuli, a set of signal dots moves in a single direction, while the remaining noise dots move randomly. When viewing a random sequence of motion that changes every 20 ms, human observers are not consciously aware of the direction at all points in time, but are more likely to report that upwards motion occurred immediately after near-upwards motion is presented[10,11] In this scenario, most individual upwards periods go undetected. In order to examine the perceptual integration of direction information over time, we designed a sequential detection-discrimination task requiring judgments about the presence and direction of brief, noisy motion pulses www.nature.com/scientificreports/
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call