The breakdown of parallel letter processing in letter-by-letter dyslexia

Abstract

Two critical issues were examined regarding letter-by-letter (LBL) dyslexia: (1) What is the nature of the functional impairment responsible for the incapacity of LBL patients to overtly recognise words on the sole basis of parallel letter processing? (2) What is the purpose of sequential letter processing? Four experiments focusing on these issues were conducted in LH, an LBL dyslexic. Expt 1 showed facilitatory effects of increased phonographic neighbourhood size, lexical frequency, and imageability on the word naming performance of LH. These high-order effects reflect a modulation of parallel letter processing in LH and demonstrate that he is able to rapidly access phonological, lexical, and semantic knowledge during reading. Congruently, Expt 2 demonstrated that all three high-order effects are eliminated when words are presented one letter at a time, from left to right. Expt 3 showed that these high-level effects are also abolished if target words are made of letters that are highly confusable (i.e., visually similar) to other letters of the alphabet. These observations suggest that LBL dyslexia may rest on an impairment at the letter encoding level that causes an excessive level of background noise in the activation of higher-order representations (i.e., letter combinations) when letters are processed in parallel. An additional experiment (Expt 4) shows that the letter confusability effect is also eliminated when words are presented one letter at a time, from left to right. This latter finding suggests that compensatory sequential processing invoked by LBL dyslexics serves to avoid the confusion between visually similar letters, which is present with parallel letter processing, and to amplify the signal-to-noise ratio required to achieve overt word identification.

Acknowledgments

We are grateful for the kind collaboration offered by LH. We also thank Frédéric Gosselin and Caroline Blais for their comments on an earlier draft and their invaluable help in the writing process. This research was supported by a grant from the Canadian Institute for Health Research to Martin Arguin. Martin Arguin is chercheur-boursier of the Fonds de la Recherche en Santé du Québec.

Notes

The fact that the word length effect in normal subjects is modulated by high-level effects such as lexical frequency and number of orthographic neighbours (Weekes, 1997) suggests that, without high-level feedback, the capacity of normal readers for parallel letter processing is limited. The rapid access to high-level knowledge thus appears necessary for the parallel processing of letters. However, difficulties in the rapid access to this knowledge could strongly alter the probability of efficient parallel processing.

This metric is based on published letter confusion matrices, which were averaged (Gilmore, Hersh, Caramazza, & Griffin, 1979; Loomis, 1982; Townsend, 1971; Van Der Heijden, Malhas, & Van Den Roovaart, 1984). Letter confusion matrices are only available for upper-case letters and not for lower-case letters. Consequently, all the experiments in this article were conducted with stimuli printed in upper-case letters.

Although very high, this correlation was not significant given the small number of conditions (and thus the low number of degrees of freedom) involved.