Predicting generalization in the training of irregular-word spelling: Treating lexical spelling deficits in a child

Abstract

This paper presents a single case study investigating the mechanisms underlying generalization of treatment benefits to untrained words in spelling. Brunsdon, Coltheart, and Nickels (2005) observed that untreated words that improved tended to be those whose errors were closest to being correct prior to treatment. These words also tended to be high in written frequency. The present study employed the same treatment techniques as those used by Brunsdon et al. with K.M., a developmental surface dysgraphic. During a first treatment the characteristics of words whose spelling improved without specific training were identified. These characteristics were then used in a second treatment to test whether it was possible to predict generalization. The results showed that treatment generalization to untreated irregular words was best predicted by neighbourhood size and frequency. We suggest that the processes underlying treatment generalization are based on the interaction between the orthographic lexicon and the graphemic buffer. Clinical implications are discussed.

Keywords:

• Surface dysgraphia
• Treatment
• Generalization
• Spelling

We would like to thank K.M. and her family for their cheerful and enthusiastic participation. Thank you also to K.M.'s teachers and the control children. We are grateful to Alan Taylor for statistical support. Helpful comments on an earlier draft were provided by M.-J. Tainturier and an anonymous reviewer. This paper was prepared while Saskia Kohnen was supported by an International Macquarie University Research Scholarship (iMURS). Lyndsey Nickels was funded by an National Health and Medical Research Council Senior Research Fellowship and Max Coltheart by Australian Research Council Special Research Centre and Federation Fellowship grants.

Notes

¹ A very convincing case was presented by McCloskey et al. (2006). Their patient's lexical errors (e.g., < SOLD > written as < COLD > ) were best explained by activation from the grapheme level feeding back to the lexical level. First, it was shown that the number of lexical errors that the patient made was greater than that predicted by chance—that is, the lexical errors were “true” lexical errors, which resulted from the selection of a lexical entry (e.g., < COMPUTE > was activated when the patient was asked to spell < CARPET > ). Secondly, the lexical errors systematically included letter intrusions from previous spelling responses. Persisting activation at the grapheme level from previous responses was fed back into the lexical level and resulted in the selection of incorrect lexemes.

² The exact nature of this improvement is not specified and would depend on the particular architecture of the theory used. However, one possibility, in a connectionist model, might be that there is a general increase in connection strength across all the links to the lexicon (see Dell, Schwartz, Martin, Saffran, & Gagnon, 1997, for an example of a connectionist model where this could apply, albeit in a different domain).

³ Note that data used to determine frequency and neighbourhood are based on adult language. There are few norms available for children.

⁴ This is a different measure from estimating whether untrained words included the same subregularities as trained words. For the subregularities we analysed which parts of the training words were not the most frequent phoneme–grapheme correspondence according to the Perry list—for example, spelling the sound /∫/ as < CH > in “chef”. We then checked which of these subregularities were present in the control words.

⁵ Note that phonological neighbours of training words may receive some activation, which then cascades to the orthographic lexicon, possibly leading to improved access for phonological neighbours of training words as well. The authors would like to thank M. J. Tainturier for this suggestion.

⁶ Note that some of the clinical decisions and implications regarding both sublexical and lexical treatment are discussed elsewhere (Kohnen & Nickels, 2008).