Philippe Blondé¹ (), Sabrina Hansmann-Roth¹, David Pascucci², Árni Kristjánsson¹; ¹Icelandic Vision Lab, School of Health Sciences, University of Iceland, Reykjavík, Iceland, ²Laboratory of Psychophysics, École Polytechnique Fédérale de Lausanne (EPFL), Switzerland

The visual system constantly extracts statistical regularities from the current environment to predict future visual input. Color is a prominent visual feature and a good candidate for extracting such regularities. Here we investigated whether a color ensemble can cue statistical regularities in target location. We assessed how associations between a target’s location and the average and variance of different color distributions affected visual search. In our first experiment, 26 participants (61.54% female, 29 ± 10 years old) completed an odd-one-out task, looking for an oddly colored target in a 6 x 6 set of 36 colored diamonds. The targets and distractors came from two distributions with different color averages. Each distribution was associated with different target location probabilities, with an 80% chance that the target would appear on the left side of the display for one distribution, and on the right side for the other. Participants gradually became faster at detecting the target when it appeared on the high probability side of the display, showing how observers learned the association between color distribution and location probability. In a second experiment, 18 participants (44.4% female, 21 ± 2 years old) completed the same task but now the two distributions had the same mean color and different variance. This time, there was no evidence that observers learned the association between distributions and probable location. These results show that statistical properties of color distributions are not used in the same way to extract regularities, with color average being an informative property while variance is not. It is also possible that the variance differences between the two distributions of experiment 2 were too small to distinguish them. Additional experiments will be performed, using larger variance differences to increase their discriminability.