A model-driven treatment of a Cantonese-speaking dyslexic patient with impairment to the semantic and nonsemantic pathways

Abstract

This paper describes a case study evaluating the efficacy of a reading therapy on a Cantonese brain-injured patient, CSH, with hypothesised deficits to the semantic and nonsemantic reading routes. The treatment emphasised the re-establishment of phonetic radical-to-syllable correspondences in regular and partially regular phonetic compounds, and encouraged the patient to make use of the semantic information associated with the signific radical to assist her in arriving at the target pronunciation. By the end of the therapy, CSH read all the treatment items flawlessly and improved significantly on reading generalisation probes, while no observable change was found in the irregular phonetic compound control probes. Specific treatment effect was evidenced by the synchrony between the introduction of training and the marked progress seen at various treatment stages, and greater improvement on treatment than on generalisation probes. In addition, CSH demonstrated an increase in regularisation errors coupled with a decrease in “no responses.”

Acknowledgments

This study was supported by a grant from the Research Grant Council of Hong Kong (HKU 7157/02H). We are grateful to CSH for her participation and Ms I. Ng for patient referral.

Notes

¹In this paper, phonetic transcriptions of Chinese characters are given in jyutping, a romanisation system developed by the Linguistics Society of Hong Kong. The number in the transcription represents the tone.

²In the Chinese language family, there are seven dialect groups. Cantonese belongs to the Yue dialect group, which is different from the Mandarin group (C.N. Li & Thompson, 1981). Although speakers of different dialects may not understand each other through speech, they can by and large communicate through the written language. Speakers on the mainland use the simplified characters, whereas those in Hong Kong and Taiwan read traditional characters.

³In greater detail, the survey examined over 7,000 entries representing 1,000 phonetic radicals in Ni (1982) and more than 12,000 characters representing over 1,100 phonetic radicals in Z.-M. Li (1989), in terms of the phonological relationship between these characters and their phonetic radical. The results show that about 34% of phonetic compounds in Li and about 40% in Ni are identical in sound (including tone) to their phonetic radical. For the rest of the entries, 29% in Li and 33% in Ni are phonologically “similar” to their phonetic component. The similarity includes a difference only in tone (e.g., ho4 “river” and ho2), only in the aspiration of the onset (e.g., zing1 “eye” and cing1), in tone and the aspiration of the onset (e.g., kan4 “celery” and gan1), or sharing the same rime (e.g., jin6 “word” and gin3).

⁴Taft et al. (1999) state that semantic units can be linked to all levels of orthographic units except for the stroke level because radicals and words can all be associated with a meaning, although the connections between signific radicals and semantic features are not explicitly shown in their model of the lexicon (Figure 5.1, p. 95).

The 300 poems of the Tang dynasty [] are highly recommended by Chinese language teachers for establishing a solid foundation of written Chinese.

Sixty normal subjects were tested on the two nonverbal semantic tests. They scored a mean of 22.25 out of 23 items (SD = 1.17) on the BORB and 31.85 out of 37 items (SD = 5.79) on the PPT.

Form class is not controlled for across conditions—19 nouns, 14 verbs, 6 adjectives, and 1 functor among the regular phonetic compounds; 18 nouns, 19 verbs, and 3 adjectives among the partially regular phonetic compounds; 16 nouns, 16 verbs, 7 adjectives, and 1 functor in the irregular phonetic compounds.

Although form class is not controlled for in the “phonetic compound” list, it is rather unlikely that the better performance on regular than irregular items was attributable to grammatical class, since it is not the case that the irregular condition is dominated by verbs, or that nouns are over-represented (compared to verbs) in the regular phonetic compound condition.

In terms of the number of stimuli that are not clearly semantically related to their signific radicals, there were 5/47 among the regular phonetic compounds in Phase 2, 10/81 among the partially regular phonetic compounds presented in Phase 3, and 6/28 in the irregular phonetic compound control probes

Two phonetic compounds among the treatment items, “to determine/to judge” and “harsh/severe,” are not semantically related to the meaning of their phonetic radical, “knife/to cut” for and “plant” for . For these characters, word contexts were employed in training.

All subsequent reports of treatment efficacy are based on contrasts between the best performing sessions during the baseline (sessions 1 to 3) and the maintenance phase (sessions 17 to 19), respectively. One may suggest that the baseline for Phase 2 probes actually covered the first six sessions and the maintenance phase included the last nine sessions. Similarly, the first 10 sessions could be considered the baseline for Phase 3. We preferred to limit our analysis to the first and last three sessions because (1) once entering therapy, reading performance might have been changed as a result of any treatment, and (2) contrasting the pre-treatment performance with the maintenance phase as opposed to the last treatment session of a phase seems to be a more stringent assessment of treatment effectiveness.

It is arguable that the absence of change in reading irregular control probes might to some extent be attributable to a greater proportion of irregular phonetic compounds not semantically related to their signific radical (6/28 or 21%) compared to the sets of stimuli in Phase 2 (5/47 or 11%) and Phase 3 (10/81 or 12%). In other words, the irregular characters might have been at a disadvantage as relatively fewer items contain relevant semantic information that could lead to greater chances of retrieving the correct phonological form. However, when one contrasts the difference in performance before and after treatment on these control items with that of partially regular unaided generalisation probes (which have a comparable proportion of characters semantically unrelated to their signific radical, 4/21 or 19%), a significant improvement was still observed for the latter but not for the former. This is in line with predictions of Chinese lexical models that the presence of phonetically similar information, compared with a lack of such information, may raise the probability of selecting the target pronunciation.

This account was suggested by one of the anonymous reviewers.