Probabilistic, positional averaging predicts object-level crowding effects with letter-like stimuli

• DOI: 10.1167/10.10.14
• Publication Date: 2010-08-18T06:19:09Z
• Authors: S. C. Dakin; J. Cass; J. A. Greenwood; P. J. Bex
• Keywords: Crowding; Feature (linguistics)

We investigated how crowding-a breakdown in object recognition that occurs the presence of nearby distracting clutter-works for complex letter-like stimuli. Subjects reported orientation (up/down/left/right) a T target, abutted by single flanker composed randomly positioned horizontal and vertical bars. In addition to familiar retinotopic anisotropies (e.g., more crowding from eccentric flankers), we report three object-centered anisotropies. First, inversions target element were rare: errors included twice as many ±90° 180° rotations. Second, flankers intrusive when they lay above or below (end-flanking) compared left right (side-flanking) an upright (an effect holds under global rotation target-flanker pair). Third, end induce subjects make erroneous reports resemble (producing structured pattern errors), but induced side do not (instead producing random errors). A model based on probabilistic weighted averaging feature positions within contours can account these effects. Thus, demonstrate set seemingly "high-level" effects arise "low-level" interactions between features elements.