Abstract
Identifying a crowded object in close proximity to neighboring flankers often leads to errors. One common type of error is mislocation — incorrectly identifying a flanker as the target. Mislocation errors occur on approximately one-third of all error trials, but little is known about their properties. To better understand mislocation errors, we measured letter identification accuracy for strings of three random letters (trigrams), as a function of letter contrast. In separate blocks of trials, observers reported all three letters (full report), or only one of the three letters (partial report). Responses were scored according to two criteria: (1) the reported letter matching the stimulus letter presented at a given letter position; (2) the reported letter matching any of the three letters presented, regardless of letter position. The difference in performance between these two scoring criteria represents the rate of mislocation errors. Five strabismic amblyopes were tested with their amblyopic eyes, and four normally-sighted observers were tested monocularly at 10° in their lower and nasal visual fields. Results from amblyopic observers and the normal periphery were qualitatively similar. Mislocation error rates were higher for partial report than for full report, and were highest for reporting the middle letter of trigrams. Averaged across all observers and contrast levels, mislocation error rates for the left, middle and right letter position of trigrams were 4.6±3.5%[SD], 9.6±4.8%, 6.6±4.4%, respectively, for full-report responses; and 7.3±4.3%, 18.3±13.6%, 7.3±5.1% for partial-report responses. There was a weak effect of contrast such that mislocation error rates were higher for lower-contrast letters, regardless of letter position. Among all mislocation error trials, only ~8% of the trials were due to switching of two adjacent letters. Our results suggest that while mislocation errors contribute to crowding, they are not the full story. Our findings also provide constraints on current models of visual crowding.
Meeting abstract presented at VSS 2012