Abstract
It takes a skilled reader approximatively the same time to read aloud words made of three to six letters. With pseudowords, however, reaction times increase linearly with the number of letters (Lavidor et al., 2002; Weekes, 1997). The cognitive and perceptual processing underlying this difference of performance is still unclear. We utilized a dynamic version of the Bubbles technique (Gosselin & Schyns, 2001; Vinette & Gosselin, 2002) to reveal the effective use of visual information in time in a lexical decision task. Five participants had to decide whether 1,000 stimuli each presented for a duration of 200 ms were words or not as fast as possible without making too many mistakes. The stimuli were 2.75 × 0.6 deg five-letter words and pseudoword randomly drawn from a database of 500 words and 500 pseudowords. The stimuli were sub-sampled in space-time by multiplicative filtered Gaussian white noise (the noise was convolved with a 3D Gaussian bubble — space std = 0.17 deg and time std = 20.41 ms). A linear regression was performed on the logarithm of participants' response times and the corresponding multiplicative noise volumes. Preliminary results indicate a clear difference between the effective space-time spread of attention for words and pseudowords. For words, but not for pseudowords, more than one letter is processed at a time suggesting partial parallel letter processing; in all cases no simple left-to-right reading strategy is observed.