Abstract
The visual system has well adapted to process complex information efficiently to compensate for its limited capacity. We form a gist of a scene at a glance (Alvarez, 2011) by representing statistical properties of the scene. Due to the summary aspect of the scene, forming statistical representations is sometimes considered to be a coarsely occurring processing, during which fine-details of individual stimuli are not accessed. Thus, we tested if statistical properties were computed based on low spatial frequency (LSF) information. It has been known that the magnocellular (MC) adaptation reduces the impact of LSF information, suggesting that the precision of mean discrimination should decrease after the adaptation if mean computation is based on LSF information. We tested this idea in three experiments. Twelve young adults participated in each experiment including adaptation and no adaptation conditions. During adaptation trials, participants were presented with flickers prior to mean discrimination tasks. This fast flicker adaptation paradigm (Arnold, Williams, Phipps, & Goodale, 2016) was designed to adapt the MC pathway and showed participants dynamic noise patterns rapidly updated at the refresh rate of 85 Hz. Across the experiments, participants discriminated means of orientations (Experiment 1), of circle sizes (Experiment 2), and of emotions on facial expressions (Experiment 3). We fitted psychometric functions to individual data and averaged them in the two adaptation conditions, yielding the widths of the functions in the two conditions. We found that the widths of the functions in the adaptation condition were significantly narrower than or at least comparable to the width of the function in the no-adaptation condition, indicative of higher or at least similar precision after the adaptation. These results suggest that averaging is not merely a coarse processing relying on LSF information. It is still precise only with high spatial frequency information after the MC adaptation.
Acknowledgement: This work was supported by National Research Foundation of Korea (NRF) grant funded by the Korean government (MSIT) (NRF-2016R1A2B4016171).