Francis Gingras^1,2 (), Alex Cousineau¹, Daniel Fiset¹, Caroline Blais¹, Frédéric Gosselin³; ¹Université du Québec en Outaouais, ²Université du Québec à Montréal, ³Université de Montréal

Online data collection has multiple advantages, including access to larger, more diverse samples as well as fast data collection. However, a particular challenge for vision science studies is that they require visual stimuli to be standardized. We propose measuring perceived brightness at different luminance values to estimate screen Gamma, thus creating a luminance lookup table so researchers can correct their stimuli when running online experiments. Participants adjust the luminance of a uniform square to match the perceived brightness of flanker stimuli, lines alternating between two luminance levels on every pixel across stimulus width. We validated this task, based on an adaptation of the code included in the PsyCalibrator package (Lin et al., 2023), in an online sample of 19 participants recruited through Prolific and tested using VPixx Pack & Go (VPixx Technologies, 2021). Participants completed five 1-parameter Gamma curve measurements (nPoints= 1, 3, 7, 15 and 31 equally spaced luminance levels between 0 and 1, each measured once), twice to assess test/retest reliability. As 0 and 1 necessarily correspond to minimal and maximal luminance respectively, the estimated Gamma curve is fit using nPoints + 2 data points. The best speed/accuracy tradeoff is found using 31 luminance measurements, allowing for a precise estimate in around 3 minutes . Comparing luminance lookup tables, extrapolated to 256 luminance levels, across measurements reveals the absolute value error between measurements is on average .008 (sd = .012). Corrections applied on the 31 measured luminance levels fit a linear model with an average R2 of .999. We find this short and simple task to be a very robust measure of screen Gamma for online participants. Using it will allow vision scientists to account for varying gray level rendering of participant display configurations in their online studies, increasing their control on the presented visual stimuli.

Acknowledgements: This work was funded by a Canada Research Chair in Cognitive and Social Vision (CRC-2023-00019) as well as a grant from the Natural Sciences and Engineering Research Council of Canada (RGPIN-2019-06201), both held by Caroline Blais.