Figure 136

Components of a model for the spelling of heard words. Adapted from Ellis and Young (1988).

producing appropriate spellings for non-words, succeeding less than 20% of the time. On the rare occasions he did spell a non-word accurately, he often reported that he had used a real word to assist him.

Other patients with phonological dysgraphia have been studied (see Parkin, 1996, for a review). The evidence suggests that words can be spelled in the absence of phoneme-grapheme conversion. However, as Barry (1994) pointed out, the fact that patients with phonological dysgraphia can write several non-words correctly suggests they have some residual ability to use phonological information. Evidence against this line of argument was reported by Shelton and Weinrich (1997). Their patient, EA, could not write any of 55 non-words to dictation, but was able to write 50% of regular words and 45% of irregular words correctly. As Shelton and Weinrich (1997, p. 126) concluded, "EA does not appear to be able to do any phoneme-to-grapheme conversion, yet he can write single words accurately. Thus, his data suggest that writing can be carried out completely independently of sublexical phonological generation, a conclusion contrary to previous arguments (e.g., Barry, 1994)."

Deep dysgraphia

How do we know that the semantic system is involved in word spelling? If, for some reason, only partial semantic information about a heard word was passed on from the semantic system to the graphemic output lexicon, then a word similar in meaning to the correct word would be written down. Precisely this has been observed in patients with deep dysgraphia. Bub and Kertesz (1982) studied a young woman, JC, who made numerous semantic errors, writing "sun" when the word "sky" was spoken, writing "chair" when "desk" was spoken, and so on. However, her reading aloud was very good, and did not contain semantic errors. Thus, the semantic system itself was probably not damaged, but rather the connection between the semantic system and the graphemic output lexicon.

The semantic errors made by deep dysgraphics presumably reflect damage in (or close to) the semantic system. However, Beaton, Guest, and Ved (1997) argued that processes closer to output can also be involved. They studied a female patient, MGK, who suffered from deep dysgraphia and deep dyslexia. She said and wrote down the names of pictures in rapid succession. On 12% of trials, there was a discrepancy between the written and the spoken responses. For example, when a brush was shown, MGK said "comb" and wrote "hair". Beaton et al. (1997, p. 459) concluded: "These discrepancies are interpreted as providing support for the view that semantic errors can arise at the level of selection of items from the.. .orthographic output lexicon (as well as at the level of semantics)."

Surface dysgraphia

If a patient relied largely on phoneme-grapheme conversion in spelling, what pattern of performance would we expect to see? Apart from producing misspellings sounding like the relevant word, such a patient would have some success in generating spellings of non-words, and would be more accurate at spelling regular words (i.e., those words where the spelling can be worked out from the sound) than irregular words. Patients with these symptoms suffer from surface dysgraphia.

All those features characterised the spelling of patient TP, who was studied by Hatfield and Patterson (1983). For example, she wrote "flud" instead of "flood" and "neffue" instead of "nephew". However, she could spell some irregular words correctly (e.g., "sign" and "cough"), suggesting there was some use of the graphemic output lexicon.

Graham, Patterson, and Hodges (1997) studied two patients (SC and FM) with surface dysgraphia. Both patients had a degenerative brain disease, and so other writing abilities deteriorated over time. For example, they became very poor at transcribing upper-case letters into lower-case letters (e.g., seeing "A" and writing "a"), and vice versa. As Graham et al. (1997, p. 996) pointed out, these findings cannot be explained satisfactorily by a model such as the one shown in Figure 13.6: "By this account the association between the deficits arises from the joint disruption of functionally separate subsystems, and is merely coincidental." They argued that interactive connectionist models (e.g., Olson & Caramazza, 1994) provide a better account of their findings.

Phoneme-grapheme conversion is sometimes used in the writing of normals. Unfamiliar and non-words do not have entries in either the graphemic output lexicon or the speech output lexicon. Thus, we must guess at their spelling by using some other strategy, such as phoneme-grapheme conversion. Children often seem to use phoneme-grapheme conversion, producing misspellings sounding like the word in question (e.g., "skool" instead of "school").

Graphemic and phonological output lexicons

How do we know there are separate graphemic output and phonological output lexicons? If information about the written forms (graphemic output lexicon) and the spoken forms (speech output lexicon) of words were stored in the same lexicon, then presumably patients who had problems with speaking would have problems with writing. We have seen already that this is not always the case. EB was able to write fairly well despite apparently having no inner speech, and RD could write many words he could not say properly. Presumably patients such as EB and RD have a relatively intact graphemic output lexicon, but a severely impaired speech output lexicon (or connections to or from it). The more common pattern among brain damaged patients is to have a greater problem with writing than with speech (see Parkin, 1996). Such patients presumably have more severe impairment of the graphemic output lexicon than of the speech output lexicon.

One or two orthographic lexicons?

Knowledge of word spellings (orthography) is important in reading and in writing. The simplest theoretical assumption is that a single orthographic lexicon is used in both reading and writing. An alternative position (e.g., Weekes & Coltheart, 1996) is that there are two orthographic lexicons, one for reading (visual input lexicon) and one for spelling (graphemic output lexicon; see Figure 13.6).

Evidence in favour of a two-lexicon model has been obtained from patients showing large discrepancies between reading and spelling. Hanley and Kay (1992) studied a patient whose spelling was much worse than his reading. In addition, most of his spelling errors involved making use of phoneme-grapheme correspondence rules (e.g., he wrote "serkel" instead of "circle"), whereas most of his reading errors involved visual confusions (e.g., "feather" was read as "further"). In contrast, Patterson (1986) studied a patient whose spelling was better than his reading. He could spell irregular words (e.g., yacht) as accurately as regular ones (e.g., capsule), but could not read printed words aloud. Thus, his knowledge of spelling did not assist his reading performance.

Weekes and Coltheart (1996) studied a patient, NW, whose main difficulty lay in reading irregularly spelled words. A reading programme produced a significant improvement in his ability to read irregular words, but did not enhance his spelling performance. Weekes and Coltheart (1996, p. 302) concluded: "We interpret the results we have obtained with distinctly favouring a two-orthographic-lexicon model." However, as Holmes and Carruthers (1998, pp. 266-267) pointed out, "brain-damaged patients often have deficits that extend beyond their difficulties in reading and spelling. Thus, their reading and spelling behaviour may be affected in multiple ways, with compensatory strategies having little to do with normal reading and spelling determining performance."

Evidence favouring the one orthographic lexicon model was reported by Funnell (1992). A 10-year-old boy was given the task of deciding whether printed words were correctly or incorrectly spelled. He performed this task well only for words he could spell accurately.

Holmes and Carruthers (1998) used a modified version of Funnell's procedure. The participants were presented with five versions of each word they could not spell: the correct version; their own misspelling; the most popular misspelling (if that differed from their own misspelling); and two or three other misspellings. The participants showed no ability to select the correct spelling over their own misspelling (see Figure 13.7). As Holmes and Carruthers (1998, p. 284) concluded, "Normal adult readers access the same orthographic representation for both reading and spelling."

In sum, the evidence remains inconclusive. There are no strong reasons at present for rejecting the simple assumption that there is a single orthographic lexicon.

Phonological output lexicon

There is evidence from normals that the phonological output lexicon is sometimes involved in writing. All of us sometimes write down a word sounding the same as the one we intended to write down. Several examples are given by Hotopf (1980), including "sought" instead of "sort", "their" instead of "there", and "scene" instead of "seen". The fact that actual words are nearly always written down suggests the involvement of the speech output lexicon. If the sounds of words were being used to produce spellings by phoneme-grapheme correspondence, then we would expect numerous non-words to be produced (e.g., "sawt" instead of "sort").

Correct spelling

Correct spelling

Very Qutte Unconfident confident confident

Very Qutte Unconfident confident confident

Helping Your Child Learn To Read

Helping Your Child Learn To Read

When parents help their children learn to read, they help open the door to a new world. As a parent, you can begin an endless learning chain: You read to your children, they develop a love of stories and poems, they want to read on their own, they practice reading, and finally they read for their own information or pleasure. They become readers, and their world is forever expanded and enriched.

Get My Free Ebook

Post a comment