Abstract
When comparing two sets of items based on one visual property like size or position, we should be able to make a decision without influence from irrelevant visual features. However, across a collection experiments that asked subjects to compare the average value of two sets, seemingly arbitrary visual features like set luminance biased which set a subject chose. Each experiment had over 800 trials. For each trial, subjects were shown two sets of 20 items side-by-side and asked which set had the higher average vertical position or which set had the higher average size (the target feature). The difference between the averages varied by trial. Each experiment had a one target feature (position or size) and one distractor feature that varied the appearance of each set as a whole. Distractor features included contrast (100% vs 50%), set size (20 vs 12), shape (square vs circle), and filled vs outlined. Item width (100% vs 66%) was another distractor feature but only when position was the target. Fitting the responses to psychometric functions, each experiment found that subjects were biased to select the more visually salient set as having a higher average position or a larger average size. Despite subjects getting feedback after every trial for hundreds of trials, the bias persisted. Moreover, another experiment modulated the distractor contrast, which resulted in similar modulation of the bias strength. To counter the possible explanation that the bias is caused by subjects conflating the two features, additional experiments asked subjects to report the lowest average position or size. In these experiments, subjects were still biased towards the set with the more salient distractor feature (e.g., higher contrast or larger set size). The consistent direction of this bias even when inverting the task suggests a decision-making bias rather than a perceptual illusion.