Abstract
Letters are special. They form a set of shapes learned at a young age and used throughout adulthood. Therefore, the human visual system might use mechanisms specifically devoted to letter identification. Recent studies of letter identification appear to be consistent with this, such as evidence for letter channels, lack of off-frequency looking in low- or high-pass masking noise, and scale dependence. Majaj et al. (Vis Res 42:1165–1184, 2002) characterized the channels used for letter identification using critical-band masking. Observers never switched to a channel with a different peak spatial frequency to escape the effects of masking noise. We repeated this experiment, but also varied the masking noise power density. Contrary to Majaj et al., we demonstrated off-frequency looking and found that its magnitude increased with increasing noise power density. That is, Majaj et al. used insufficient noise power density to induce subjects to switch channels. Majaj et al. also found scale dependence: peak letter channel frequency (cycle/letter) varied with letter size. However, Chung et al. (Vis Res 42:2137–2153, 2002) claimed scale dependence was simply an effect of what an ideal observer would do given the equivalent input noise (from the contrast sensitivity function). We repeated the critical-band masking experiment in the presence of additional white noise (to swamp equivalent input noise), but still found scale dependence. Letter identification is scale dependent even when it no longer provides a benefit. This scale dependence is not found with, e.g., grating detection, and may indicate that letters are indeed special.
NIH EY08266 and EY16165 (to MSL)