Purchase this article with an account.
Bosco S. Tjan, Chen He, Susana T. L. Chung, Noah Z. Schwartz; Letter crowding in the periphery is best modeled by an increase in additive equivalent noise. Journal of Vision 2004;4(8):509. doi: 10.1167/4.8.509.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Letter crowding typically refers to the decrease in letter identification performance when a target letter is flanked by other letters. Crowding is most pronounced in the periphery, where it impedes form vision. Previous studies have shown that crowding is unlike classical contrast masking, but the functional mechanism of crowding remains unknown. We used an ideal observer model to uncover the functional components of crowding. Subjects identified lowercase letters of Times-Roman font at 5 deg eccentricity inferior to fixation. The target letter was presented alone, flanked by a letter on each side, or flanked by a noise patch on each side. The noise patch had the identical power spectral density and resided within the same bounding box as a randomly selected letter. Letters were of size 0.4 log units above a subject's acuity at 5 deg. Letter spacing was 1 x-height center-to-center. Flanking letters were at 20% contrast. Threshold contrast energy (E) of the target letter at 79% correct was determined for the 3 crowding conditions in 4 levels of white noise (N), with spectral densities of 2.4, 5.9, 15, 38 micro-deg^2. Two subjects with normal vision participated. For both subjects, E vs. N was well approximated by a straight line, and no significant change of slope was observed between the three crowding conditions (only vertical shifts). Statistical efficiency was therefore unaffected by crowding. Equivalent noise, as determined by the x-intercept of the E vs. N lines, averaged 4.4 (unflanked), 24 (letter flankers), and 15 (noise flankers) micro-deg^2. Equivalent noise increased by a factor of 5.6 when flanked by letters and 3.4 when flanked by noise patches. Our results suggest that crowding impedes performance by inducing noise into the visual system. The quality of the observer's “letter template” and the associated computations are otherwise unaffected. The spectral density of the induced noise is yet unknown and may depend on the flankers.
This PDF is available to Subscribers Only