The effective use of spatial frequencies through time in reading

Abstract

We used the Bubbles method (Gosselin & Schyns, 2001) to examine the effective use of spatial frequencies through time in a reading task. Ten participants viewed a total of 2600 dynamic stimuli. Each stimulus sampled one of 2,000 four-to-seven-letter words (one letter spanned 0.38 × 0.61 deg of visual angle × 180 ms) by dot multiplying their Fourier spectrum with a 2D white Gaussian noise field convolved with a Gaussian function (Std's = 0.156 of the Nyquist frequency and 25 ms). The subjects' accuracy was maintained at 50% correct by adjusting, on a trial by trial basis, the surface under the sampling noise. Multiple linear regression was performed on response accuracy and sampling noise. Subjects showed an early use of a narrow band of spatial frequencies centered at 1.6 cycles per letter. Around 40 ms following the stimulus onset, another band of spatial frequencies centered at 3.0 cycles per letter became useful. These results show that two bands of spatial frequencies are important for visual word identification and suggest a two-step process: low spatial frequencies would activate a subset of word attractors and mid-spatial frequencies, known to be useful for letter recognition (Solomon & Pelli, 1994; see Fiset, Chauvin, Dupuis-Roy, Blais, Arguin & Gosselin, VSS, 2005), would add fine grain letter information to complete word recognition.