Prelexical representations and processes in reading: Evidence from acquired dyslexia

Abstract

We report a detailed and extensive single-case study of an acquired dyslexic patient, L.H.D., who suffered a left-hemisphere lesion as a result of a ruptured aneurysm. We present evidence that L.H.D.’s reading errors stem from a deficit in visual letter identification, and we use her deficit as a basis for exploring a variety of issues concerning prelexical representations and processes in reading. First, building on the work of other researchers, we present evidence that the prelexical reading system includes an allograph level of representation that represents each distinct visual shape of a letter (e.g., a, A, etc., for the letter A). We extend a theory proposed by Caramazza and Hillis [Caramazza, A., & Hillis, A. (1990a). Spatial representation of words in the brain implied by studies of a unilateral neglect patient. Nature, 346, 267–269] to include an allograph level, and we probe the nature of the allograph representations in some detail. Next, we explore the implications of visual similarity effects and letter perseverations in L.H.D.’s reading performance, arguing that these effects shed light on activation dynamics in the prelexical reading system and on the genesis of L.H.D.’s errors. We also probe the processing of letter case in the visual letter identification process, proposing that separate abstract letter identity and case representations are computed. Finally, we present evidence that the allograph level as well as the abstract letter identity level implement a word-based frame of reference.

• Acquired dyslexia
• Pure alexia
• Reference frames
• Letter identification
• Perseverations
• Reading.

We are grateful to L.H.D. for her cooperation and patience over many sessions of testing, and we thank Jeremy Purcell for creating the figure of L.H.D.'s lesion.The first author was supported by an National Science Foundation (NSF) Integrative Graduate Education and Research Traineeship (IGERT) grant to the Johns Hopkins Department of Cognitive Science.

Notes

1 Caramazza and Hillis (1990a, 1990b) refer to this level as the letter shape level. We use the more neutral character shape instead of letter shape to emphasize that this level represents the shape of any typographic character, including unfamiliar characters such as pseudoletters.

2 The few misspellings that were not phonologically plausible (4 for word stimuli and 2 for nonword stimuli) appeared to reflect minor phoneme–grapheme conversion errors (e.g., the nonword /flok/ orally spelled as FLOACE), which are not uncommon even in neurologically intact individuals.

3 Some researchers (e.g., Cipolotti & Warrington, 1996; Warrington & Langdon, 2002) have suggested that naming from oral spelling recruits spelling rather than reading processes. However, no specific account has been offered for how the task would be accomplished via spelling mechanisms. Furthermore, given that L.H.D. has a central spelling deficit, her intact naming from oral spelling performance provides evidence that naming from oral spelling does not implicate spelling processes.

4 One might have supposed that when naming letters in random letter strings, L.H.D. would have processed the letters one at a time, as if the letters had been presented individually. However, the finding of lower accuracy for letters in strings than for letters presented individually and the serial position effects (lower accuracy for final than initial letters) indicate that she typically processed the strings as wholes, or at least in multiletter chunks (see McCloskey & Schubert, 2014, for additional results and discussion).

5 The conclusion that L.H.D.'s reading improves when letter representations have not recently been activated leads to a prediction: L.H.D. should show higher reading accuracy on the first trial of a list than on subsequent trials. Tabulating results across the 116 lists in the reading corpus, we found that L.H.D.'s accuracy was 80% on Trial 1, but dropped to 62% on Trial 2 and fluctuated around a mean of 60% (range 54%–67%) on Trials 3–10. The accuracy difference between Trials 1 and 2 was significant, χ²(1, N = 232) = 9.25, p < .01, as was the difference between Trial 1 and Trials 2–10, χ²(1, N = 1160) = 17.18, p < .001.

6 All analyses of perseveration positions reported in this article are based upon the both-edges scheme that McCloskey et al. (2014) refer to as the narrowly graded both-edges scheme.