The role of plasticity-related functional reorganization in the explanation of central dyslexias

Abstract

This investigation explored the hypothesis that patterns of acquired dyslexia may reflect, in part, plasticity-driven relearning that dynamically alters the division of labour (DOL) between the direct, orthography → phonology (O → P) pathway and the semantically mediated, orthography → semantics → phonology (O → S → P) pathway. Three simulations were conducted using a variant of the triangle model of reading. The model demonstrated core characteristics of normal reading behaviour in its undamaged state. When damage was followed by reoptimization (mimicking spontaneous recovery), the model reproduced the deficits observed in the central dyslexias—acute phonological damage combined with recovery matched data taken from a series of 12 phonological dyslexic patients—whilst progressive semantic damage interspersed with recovery reproduced data taken from 100 observations of semantic dementia patients. The severely phonologically damaged model also produced symptoms of deep dyslexia (imageability effects, production of semantic and mixed semantic/visual errors). In all cases, the DOL changed significantly in the recovery period, suggesting that postmorbid functional reorganization is important in understanding behaviour in chronic-stage patients.

Keywords:

• Reading
• Connectionist
• Plasticity
• Dyslexia
• Parallel distributed processing

Acknowledgments

The research reported here was supported by grants from the Biotechnology and Biological Sciences Research Council (BBSRC; S20390); Engineering and Physical Sciences Research Council (EPSRC), Medical Research Council (MRC), and BBSRC (EP/F03430X/1); and the Gatsby Charitable Foundation (GAT2831). We are grateful for helpful contributions from Mark Seidenberg and David Plaut.

Notes

¹ This interaction is sometimes also referred to as frequency–regularity or frequency–typicality. Regularity and consistency are heavily confounded in English but they are theoretically distinct: Regular words follow the standard grapheme-to-phoneme conversion rules for pronunciation, whereas consistent words have word bodies whose pronunciation is consistent with their neighbours. In our stimuli, these properties are indistinguishable, as all the regular stimuli are also consistent. However, in the light of this and the considerable amount of theoretical effort that has been put into distinguishing regularity and consistency effects (Andrews, 1982; Cortese & Simpson, 2000; Jared, 2002; Jared, McRae, & Seidenberg, 1990; Jefferies, Ralph, Jones, Bateman, & Patterson, 2004; Taraban & McClelland, 1987), we have elected to use the term “consistency” to describe the phenomenon, as this reflects our belief that that these phenomena result from graded variation in the orthography-to-phonology mappings found in the language at multiple subword levels, rather than in the application of dichotomous rules at the grapheme–phoneme level. It is worth noting that nonword letter strings can also vary in their consistency (Zevin & Seidenberg, 2006). Similarly, we shall use the term “legitimate alternative reading of the components” (LARC; Patterson et al., 1995) rather than the more traditional term “regularization” to refer to errors where the phonological output when reading inconsistent words is what would be expected from analogy with neighbours.

² These are described in the original paper as visual errors, but visual and phonological errors are very hard to distinguish in English due to the largely regular nature of the relationship between orthography and phonology. We refer to them throughout as visual/phonological errors.

³ Although the vast majority of inconsistencies occur in the vowel portion of the word, there are some words in English with inconsistencies in the onset or coda (e.g., gaol vs. game, or cough vs. though). Our model is not large enough to allow us to include these types of rare inconsistencies. However, it seems very unlikely that this omission would be critical to our general findings.

⁴ This result depends upon a very stringent criterion whereby only one possible pronunciation of each nonword is allowed; if we had accepted variations where the vowel had been pronounced to match an inconsistent word than the accuracy rate would have been very close to 100%.

⁵ This may not be the only kind of damage that could theoretically lead to phonological/deep dyslexia, but it is by far the most common.

⁶ This is not traditionally associated with phonological dyslexia. However, although it is not often reported, phonological dyslexics do often exhibit consistency effects. A reanalysis of data from Berndt et al. (1996) reveals that 9 out of 10 of the patients in the series showed more accurate reading of consistent than of inconsistent words with the performance difference ranging from 2% to 20%. When data from all of the patients are submitted to statistical analysis, these differences are shown to be significant, t(9) = 2.32, p = .023, one-tailed.

⁷ We have elected to model surface dyslexia arising from progressive damage because that represents by far the largest number of reported cases. However, pilot simulations suggest that the model is also capable of reproducing the general pattern of surface dyslexic symptoms from acute damage followed by recovery.

⁸ Based on the data reported by Plaut, McClelland, Seidenberg, and Patterson (1996) K.T.'s performance on reading was 26% for low-frequency inconsistent words and 100% for regular nonwords. A total of 85% of K.T.'s errors on reading exception words were regularizations (we would expect the LARC error rate to be similar). The closest matching point from the patient case series read low-frequency exception words with an accuracy of 29%, and 67% of errors were LARC errors. Unfortunately, no data were collected for nonword reading, but accuracy for regular words was 100%. The closest matching point from the model was 30% accuracy for low-frequency exceptions, 100% accuracy for nonwords, and a LARC error rate of 80%.