Abstract
In many orientation estimation experiments, participants report the orientation of a target stimulus while ignoring the orientation of a distractor stimulus. The distractor interference is visualized by plotting the estimation error (reported orientation - target orientation) against the relative distractor orientation (distractor orientation - target orientation). Depending on whether the reported orientation and the distractor orientation are on the same side or on the opposite side of the target orientation, the distractor interference is characterized as attractive or repulsive. Through a series of experiments, we show that this characterization of the distractor interference is incorrect. The experimental stimuli consisted of several small rectangles spatially arranged to create a single large rectangle (Experiment 1) or several large rectangles (Experiment 2). Participants estimated the orientation (mean orientation for Experiment 2) at the global or local level, while ignoring the orientation at the other level. Besides the interference from the distractor level, an additional bias was present. The reported orientations were also biased towards the central tendency of the distribution of target orientations. The traditional approach of measuring distractor interference indicated that the distractor interference was attractive for small relative orientations and repulsive for large relative orientations. After accounting for the central tendency bias, however, we found that the distractor interference was attractive for small relative orientations but zero for large relative orientations. Experiment 3 used target distributions with different ranges and demonstrated a bias towards the central tendency of the respective distribution. Experiment 4 showed that the bias is towards the mean and not the mode of the target distributions. The central tendency bias is robust across the experiments. These results highlight the importance of accounting for the central tendency bias when measuring distractor interference.