Abstract
We previously showed that, when asked to compare mean brightness of heterogeneous luminance ensembles, observers use a flexible weighted-averaging strategy to rely more on a few items in the ensembles such as the highest or lowest luminance patches depending on task requirement (Kimura et al., ECVP2018). The present study extended this finding and investigated whether, when asked to discriminate variability of luminance ensembles, observers also use some smart strategy of relying on proxies such as the luminance range, the highest luminance, or the lowest luminance of the arrays. We used a method of constant stimuli to measure discrimination performance. The standard and comparison stimuli were heterogeneous luminance arrays composed of 24 patches (each disk subtended 1.5°). The two arrays were presented side by side simultaneously for 47 msec and followed by a dynamic pattern mask. Observers’ task was to indicate the more variable array of the two. Mean luminance of the standard and comparison stimuli was fixed at 35 cd/m2. The standard deviation (SD) of luminance distribution of the standard stimulus was set to one of 4 levels (4.0, 8.0, 12.0, or 16.0) and that of the comparison stimulus was varied from 0 to 20.5. Results showed that observers could accurately and reliably discriminate luminance variability of the ensembles, even though the stimulus was very brief and composed of many patches. However, unlike in mean comparison, no clear evidence for smart subsampling strategy was found. The percentage of correct responses could be described fairly well as a function of the difference in SD between the standard and comparison stimuli regardless of the level of the standard SD. Thus, Weber’s law did not hold for the discrimination of luminance variability. Taken together, these results suggested that the variability of luminance ensembles is coded in a qualitatively different fashion from the mean.
Acknowledgement: Supported by JSPS KAKENHI (26285162 & 25285197)