How does one best learn a language if they are deaf? How does language acquisition of a deaf person differ from that of a nondeaf person on a biological level?
French-speaking hearing and deaf children, ranging in age from 6 years 10 months to 14 years 7 months were required to spell words including phoneme-to-grapheme correspondences that were either statistically dominant or nondominant. Of interest was whether the nature of linguistic experience (cued speech vs. sign language) and the precocity of such experience (early vs. late exposure) determines accuracy in the use of phoneme-to-grapheme knowledge. Cued speech is a system delivering phonemically augmented speechreading through the visual modality. Hearing and deaf children exposed to cued speech early at home relied on accurate phoneme-to-grapheme correspondences, whereas children exposed to cued speech later and at school only, and children exposed to sign language, did not. A critical factor in the development of the phonological route for spelling seems to be early and intensive exposure to a system making all phonological distinctions easily perceivable.
Most of the spelling errors made by hearing children are compatible with the word phonological form. They reflect incomplete knowledge at three possible levels: word-specific orthographic information (e.g., “brane” /brein/ for brain in English, or “trin” /tr~[varepsilon]/ for train in French), contextual rules (e.g., “woz” for was in English, or “janbon” /j~[alpha]b[Latin small letter open o]~/for jambon in French), and morphological relationship (e.g., “speld” for spelled in English, or “peti” /p[Latin small letter schwa]ti/ for petit in French). By contrast, most of the spelling errors made by deaf youngsters reflect an incomplete knowledge of the word phonology (e.g., “vingear” for vinegar in English, “moule” /mul [Latin small letter schwa] / for moulin /mul~[varepsilon]/, and “escorle” /[varepsilon]sk[Latin small letter open o]rl [Latin small letter schwa]/for escalier /[varepsilon]skalje/ in French).
Why do deaf spellers make phonologically inaccurate errors? One hypothesis is that hearing is a necessary condition for the acquisition of phoneme-to-grapheme correspondences (Gates & Chase, 1926). In this view, the sensory deficit (i.e., profound hearing loss) necessarily leads to a cognitive deficiency (i.e., the lack of use of the phonological route for spelling). An alternative hypothesis is that deaf children acquire a phonological system through the speechread input (Dodd, 1976, 1987). Deaf children's speechreading skills have been identified as the best predictors of their early reading and spelling development (Dodd, McIntosh, & Woodhouse, 1998). However, the speech information seen on the lips is inherently underspecified (i.e., ambiguous), due to the similarity in appearance of speech elements sharing the same place of articulation and to the invisibility of features such as voicing and nasality (Erber, 1974; Walden, Prosek, Montgomery, Scherr, & Jones, 1977). Any given lip movement can potentially map onto more than one phoneme. Consequently, the deaf children's phonological representations tend to be inaccurate, and underspecified. This situation does not preclude deaf children's use of phoneme-to-grapheme relationships, but strongly hinders the benefit they can get from that use (Burden & Campbell, 1994; Dodd, 1980; Leybaert & Alegria, 1995).
If deaf children's spelling is limited by the inherent ambiguity of speechreading, the addition of complementary visual information that resolves this ambiguity could improve their ability to use the relationship between phonemes and graphemes. A recent study of the spelling of deaf children educated early and precociously with cued speech (CS) provides evidence for this hypothesis (Leybaert, 2000). In CS, the speaker complements speech (speechreading for the deaf receiver) with manual cues. A cue consists of two parameters: handshape and hand placement around the mouth. Handshapes (eight in the French CS system) disambiguate the consonants, and hand placement (five in French) disambiguate the vowels. Consonants (or vowels) assigned to the same handshape (or hand placement) are easy to discriminate by speechreading, whereas those difficult to discriminate belong to two different handshapes (or hand placements). In CS, all the phonemic distinctions of spoken language can be naturally perceived by sight. The production of a consonant–vowel (CV) syllable requires a single cue (a particular handshape at a specific hand placement) that carries information about both the consonant and the vowel. Syllabic structures like vowel–consonant (VC), CCV, and CVC need additional cues to reveal the supplementary phonemes (see Cornett & Daisey, 1992, and Charlier & Leybaert, 2000 for more complete descriptions of CS in English and in French, respectively, and Messing, 1999, for a discussion of the differences between CS and sign language [SL]).
Children who have been exposed to CS early and intensively at home (i.e., CS-home children) develop a spelling production system strongly guided by phonology. Their misspellings are most of the time phonologically accurate. By contrast, those who have been exposed to CS later and less intensively at school only (i.e., the CS-school children) make more phonologically inaccurate errors (Leybaert, 2000). The spelling performance of the CS-home children may reflect the effect of early and intensive exposure to CS. However, intensive CS exposure was confounded with the total amount of language exposure in Leybaert's (2000) study. Early exposure to a fully accessible language may be the critical factor, rather than exposure to CS per se. The present study aimed at comparing the spelling of the CS-home children to that of deaf children exposed early in life to a visual language, albeit of a different nature (i.e., SL).
Method
Participants
We tested 67 deaf children and 32 hearing children. Deaf children were recruited from five different specialized schools or centers for rehabilitation of language in France and Belgium and in deaf camps. All deaf children met the following criteria: (a) bilateral profound sensorineural hearing loss > 90 dB in the better ear across three frequencies of the speech range (0.5, 1, and 2 kHz); (b) no other significant handicapping conditions; and (c) hearing loss onset prior to 18 months of age. It was not the policy of the schools to allow us to test children's IQ. However, none of the children had any known associated disabilities, and each was judged by the school's psychologist to be within the normal range of intelligence. Sixty-three children were equipped with two acoustical hearing aids worn during the experiment, and the other 4 had a cochlear implant.
Younger Readers
The number of correct responses were entered in a 5 (group) × 2 (dominance) × 3 (frequency) analysis of variance (ANOVA). This analysis revealed a significant effect of dominance, F(1, 73) = 21.99, p < .0001: Dominant graphemes (M = 7.6) were better spelled than nondominant graphemes (M = 6.5). There was also a significant effect of frequency, F(2, 146) = 143.09, p < .0001. Helmert a posteriori contrasts revealed that spelling accuracy was significantly higher for graphemes included in high frequency words (M = 8.3) than in medium frequency words (M = 7.2) and low frequency words (M = 5.7), and higher for medium than for low frequency words (both ps < .001). There was no significant effect of group, F(4, 73) < 1. A significant two-way Dominance × Frequency interaction also appeared, F(2, 146) = 10.66, p < .0001. An examination of Table 2 reveals that the effect of dominance was stronger for low frequency words (M = 1.8 items) than for medium (M = 0.6 items) or high frequency words (M = 0.9 items; see Table 2). There was no significant Frequency × Group interaction, F(8, 146) = 1.94, p = .06, but there was a significant Dominance × Frequency × Group interaction, F(8, 146) = 2.12, p < .05.
To explore the Dominance × Frequency × Group interaction, a 2 × 3 ANOVA with repeated measures on dominance (dominant vs. nondominant) and frequency (high, medium, and low frequency) was performed in each group. In hearing children, all effects proved to be significant: dominance, F(1, 15) = 8.53, p < .01; frequency, F(2, 30) = 33.94, p < .0001; and the Dominance × Frequency interaction, F(2, 30) = 9.13, p < .001. The CS-home group exhibited similar significant effects: dominance, F(1, 19) = 14.03, p < .005; frequency, F(2, 38) = 49.44, p < .0001; and the Dominance × Frequency interaction, F(2, 38) = 7.69, p < .0001. In the CS-school group, the effect of dominance was marginally significant, F(1, 17) = 4.26, p = .055. The effect of frequency was significant, F(2, 34) = 20.89, p < .001, whereas the Dominance × Frequency interaction was not, F(2, 34) = 1.67. The two groups of SL children exhibited the following pattern of results: no effect of dominance, F(1, 13) = 2.09 for SL-home and F(1, 9) = 1.34 for SL-school (both ps > .10), a significant effect of frequency, F(2, 26) = 40.32, p < .001, and F(2, 18) = 23.98, p < .001, respectively, and no interaction between dominance and frequency, F(2, 26) = 2.04, and F(2, 18) = 2.92, respectively.
Older Readers
The data of the older hearing and CS-home participants were entered in a 2 (group) × 2 (dominance) × 3 (frequency) ANOVA. This analysis yielded significant effects of dominance, F(1, 19) = 10.16, p < .005; of frequency, F(2, 38) = 22.53, p < .0001; and of the Dominance × Frequency interaction, F(2, 38) = 19.66, p < .0001. Neither the effect of group, nor the interactions involving the group factor, were significant.[/QUOTE]
The misspellings at the word level are summarized in the bottom of Table 3. Separate one-way ANOVAs were carried out on number of word substitutions and number of misspellings that were orthographically illegal. Post hoc tests (Newman-Keuls, at the .05 level) were used to determine which groups differed from each other when the ANOVA revealed a significant group effect. The groups of younger readers significantly differed on percentage of word errors, F(4, 73) = 11.70, p < .0001: SL-home and SL-school groups made a higher proportion of such errors than the hearing group, the CS-home group, and the CS-school group. The five groups did not differ regarding the percentage of errors that were orthographically illegal (F < 1). The two groups of older readers did not differ regarding the number of word errors, nor regarding the number of orthographically illegal errors.
Discussion
Our experiment was designed to investigate the possible influences of deaf children's nature and precocity of linguistic experience on the use of phoneme-to-grapheme relationships. We hypothesized that exposure to CS entails a use of accurate phoneme-to-grapheme relationships, all the more so because exposure is early and intensive. The effect of grapheme dominance and the interaction between the effect of grapheme dominance and word frequency should be more marked in groups of CS-home children than in groups of CS-school, SL-home, and SL-school children. The majority of spelling errors should be phonologically accurate in CS-home groups as in hearing children.
In line with our predictions, only the hearing children and the CS-home children exhibited a significant effect of dominance and a significant interaction between dominance and word frequency. Dominant graphemes were better spelled than nondominant graphemes, and the effect of dominance was larger for low frequency words than for medium or high frequency words. As predicted also, the CS-home children and the hearing children made a high and similar proportion of phonologically accurate errors. When required to spell a word for which they do not have a fully detailed orthographic representation, hearing and CS-home children start from accurate phonological representations (Charlier & Leybaert, 2000), and apply dominant correspondences between phonemes and graphemes.
Early acquisition of a natural language and language having the phonological structure of the spoken language are thus necessary conditions for the development of accurate use of phoneme-to-grapheme relationships. The fact of learning a language early in life cannot explain our results on its own, because SL-home children did not achieve a larger effect of grapheme dominance than SL-school children. Exposure to CS is also in itself insufficient to explain our results, as the CS-home children differed from the CS-school children. CS-home children also have a fairly sophisticated knowledge of the relations between phonemes and graphemes in French, as hearing children do. They knew that the phonemes have more than one possible spelling, and that one spelling is more common than others. The development of a spelling system sensitive to the statistical relationship between phonemes and graphemes is not precluded in the case of prelingual and profound deafness. It can be promoted by exposure to visual speech information, provided that all phonological distinctions can be easily perceived.
The CS-school, SL-home, and SL-school children performed differently in spelling than did the CS-home and hearing children. First, they displayed a similar effect of frequency on dominant and nondominant graphemes. Second, they made fewer phonologically accurate misspellings, indicating a lower ability to use accurate phoneme-to-grapheme mappings. Their performance may result from inaccuracy at the level of phonological representations, deficiency in segmentation of these representations, or a difficulty in attributing graphemes to phonemes. CS-school children and children educated with SL have underspecified representation of phonology in previous work (Charlier & Leybaert, 2000). Inaccuracy of their phonological representations hinders these children in applying phoneme-to-grapheme correspondences in the present study. Possibly, they need more exposure to print in order to acquire correct spelling for dominant as well as for nondominant graphemes.
It must be borne in mind that deaf children's phonologically accurate responses likely do not constitute the only attempts to represent the word's phonemes. As discussed by Treiman (1983) in the case of hearing first grade spellers, many phonologically inaccurate errors are almost surely attempts to represent the word's phonology. A similar argument was developed about deaf children: Errors that are illegal in relation to the conventional phonological representation of hearing adults, may not be illegal when deaf children's unconventional representations of phonology are considered (Burden & Campbell, 1994; Leybaert & Alegria, 1995). Attempts to generate spellings from unconventional phonological forms likely explained misspellings like “tirop” for sirop and “douvercle” for couvercle observed here. The children symbolized one phoneme (for example /s/ or /k/) with a letter (t or d) that is used for a phoneme articulated at the same, or at a near, place. These errors could reflect phonological processes rather than visual memorization.
The groups also differentiate in the number of word substitution errors. SL-home and SL-school groups made more such word errors than the hearing, the CS-home, and the CS-school children. The words they gave as responses were visually similar to the targets, in nearly all the cases (e.g., “villa,” “ville,” or “village” for vilain; “moule” for moulin). That is, there was a difficulty in the selection among the visually similar orthographic representations. The spelling production system of SL-children seems more governed by orthographic knowledge (Perfetti & Sandak, 2000). Interestingly, such errors were also found in hearing phonological dyslexics whose spelling is little constrained by phonology (Broom & Doctor, 1995). Accurate phonological representations thus constrain the possible patterns used to spell a word, and, consequently, make easier the selection of the conventional spelling.
This could explain why the CS-home children reached a reading level appropriate for their chronological age, whereas the other deaf children were reading-delayed. However, differences in parents' attitudes toward reading cannot explain that CS-home children use more phoneme-to-grapheme relationships than other deaf children matched for word recognition level. Finally, could the results of the hearing and CS-home younger groups be explained by a lesser exposure to print because of their younger chronological age?
.
It thus seems reasonable to conclude that the spelling by deaf children is phonologically guided, to various extents, depending on their language experience. A critical condition ensuring adequate spelling development seems to be early and intensive exposure to a system that makes all phonological distinctions of spoken language visually accessible. A late and less intensive exposure to systems like CS does not have the same effect on the use of phoneme-to-grapheme correspondences. Finally, the early advantage in language development displayed by deaf children born to deaf parents over deaf children from hearing parents, does not induce a larger use of phonology-to-orthography mappings. Possibly, there is a “missing link” between this early language experience and reading acquisition (Padden & Hanson, 1999).
Providing deaf children with inputs that could serve as a principled way of remembering the word spelling, such as CS or fingerspelling, could improve their spelling abilities (Hanson et al., 1983; Padden & Ramsey, 2000). This is not contradictory to the use of sign language as a primary language. Early experience with fingerspelling or CS may allow the development of metaphonological awareness before, or in interaction with, reading. We plan to test this hypothesis by studying deaf children who have a SL as primary language and who benefited from acquiring French, English, or any other traditionally spoken language via CS before they started learning to read. We would expect that the spelling by such children could be phonologically guided as is the spelling by hearing or CS-home children. The study of the development of metaphonological skills, reading and spelling, of profoundly, prelingually deaf children fitted with a cochlear implant is also on the agenda. The information provided by the implant could interact with that perceived through lip-reading and CS to set the stage for the development of phonological awareness.
In sum, the reading and spelling difficulties of deaf children are long-lasting problems. Our research demonstrates that access to an input providing information about phonological contrasts is the critical question rather than deafness per se.
Variability in Deaf Children's Spelling : The Effect of Language Experience__
Journal of Educational Psychology, Vol. 93(3), September 2001. pp. 554-562
ummm maybe a couple of questions on the thread would be suffice, eh?
