Journal of Deaf Studies and Deaf Education Advance Access originally published online on May 30, 2007
The Journal of Deaf Studies and Deaf Education 2007 12(3):283-302; doi:10.1093/deafed/enm017
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Reading Comprehension of Deaf Children With Cochlear Implants
Radboud University Nijmegen
Diagnostic Centre Viataal
Radboud University Nijmegen
Radboud University Nijmegen
Correspondence should be sent to Anneke M. Vermeulen, ENT Department, University Medical Hospital St. Radboud, Radboud University, KNO huispost 377, P.O. Box 9101, 6500 HB Nijmegen, The Netherlands (e-mail: am.vermeulen{at}kno.umcn.nl).
Received October 14, 2006; revised March 25, 2007; accepted March 26, 2007
The reading comprehension and visual word recognition in 50 deaf children and adolescents with at least 3 years of cochlear implant (CI) use were evaluated. Their skills were contrasted with reference data of 500 deaf children without CIs. The reading comprehension level in children with CIs was expected to surpass that in deaf children without implants, partly via improved visual word recognition. Reading comprehension scores of children with implants were significantly better than those of deaf children without implants, although the performance in implant users was substantially lagging behind that in hearing children. Visual word recognition was better in children with CIs than in children without implants, in secondary education only. No difference in visual word recognition was found between the children with CIs and the hearing children, whereas the deaf children without implants showed a slightly poorer performance. The difference in reading comprehension performance of the deaf children with and without CIs remained present when visual word recognition was controlled for. This indicates that other reading-related skills were also contributing to the improved reading comprehension skills of deaf children with CIs.
 |
Introduction |
|---|
 TOP Introduction Materials and MethodsResultsDiscussionReferences |
|---|
The benefit that many profoundly deaf children can derive from the use of cochlear implants (CI),1 poses new challenges to the upbringing and education of deaf children. Since the 1990s, growing evidence was reported of substantial increases in auditory speech perception abilities after cochlear implantation in profoundly deaf children (e.g., Meyer, Svirsky, Kirk, & Miyamoto, 1998
; Robbins, Koch, Osberger, Zimmerman-Phillips, & Kishon-Rabin, 2004; Svirsky, Teoh, & Neuburger, 2004). The auditory speech perception skills of profoundly deaf children often even reached a level comparable to that of hard-of-hearing children over the course of time (Snik, Vermeulen, et al., 1997). Moreover, CIs proved to provide deaf children with auditory access to spoken language, reflected in increasing receptive vocabulary (Geers, Nicholas, & Sedey, 2003; Miyamoto, Houston, Kirk, Perdew, & Svirsky 2003; Svirsky, Robbins, Kirk, Pisoni, & Miyamoto 2000; Vermeulen, Hoekstra, Brokx, & Van den Broek, 1999). A subsequent focus of research was the effect of cochlear implantation on written language. For enhancement of educational attainment and for full participation in the society, reading is of course extremely important. Because the process of reading depends on the spoken language that provides the basis of the writing system (Perfetti & Sandak, 2000), one expectation of CIs is that they might enhance reading comprehension of deaf children. It is on the reading skills of deaf children with CIs that our study focuses.
Knowledge of spoken language and spoken language skills contribute to reading comprehension and to visual word recognition,2 two major components of reading competence (see below). It is clear that the development of reading in deaf children without age-appropriate spoken language skills will be difficult and slow (Musselman, 2000). Long-term research into the reading skills of deaf children and adolescents without CIs showed a limited level of reading comprehension (e.g., Holt, 1993; Holt, Traxler, & Allen, 1996; Traxler, 2000). In a recent large-scale study, on the reading skills of deaf Dutch children, Wauters, van Bon, and Tellings (2006), reported that the average reading comprehension scores of deaf children and adolescents were "shockingly low" (Wauters, 2005, p. 145). The average reading comprehension for children in the ages of 7–20 years was at a level of first grade of primary education.
There have been divergent results regarding the level of visual word recognition in deaf children without CIs. Harris and Beech (1998) and Merrills, Underwood, and Wood (1994) found that deaf children had lower word identification skills than their hearing peers. Contrastingly, Burden and Campbell (1994) and Fischler (1985) did not observe any significant difference in word recognition skills between older deaf children and children with normal hearing. The different ages of the children in the studied samples may have caused this difference. Wauters et al. (2006), however, found that the mean visual word recognition scores of deaf children in primary and in secondary education were significantly lower than those of their hearing peers, though to a lesser extent than were reading comprehension scores. Furthermore, Wauters et al. found that the visual word recognition skills did not explain the poor reading comprehension scores of children without CIs. We do not know, however, whether this is the case in children with CIs because the population in the study of Wauter et al. did not include CI users.
At present very little is known about the reading skills of deaf children with CIs. An early study on reading comprehension (Spencer, Tomblin, & Gantz, 1997) reported the results of a heterogeneous group of 40 prelingually profoundly deaf implant recipients. Their skills were compared with results of children without CIs that have been reported in other studies on reading comprehension. Almost half the children with CIs perform within the 8-month range of grade-level performance indications that were obtained from other studies. However, there was considerable variability in the scores of the CI users. Moreover, the reference data of children without CIs in this study did not permit proper comparison with the children with CIs because different tests had been used and the inclusion criteria of the implant group were not described. Spencer, Barker, and Tomblin (2003) conducted a study on reading comprehension in two groups of 16 children. Prelingually deaf children with CIs were compared to a group with normal hearing. They found significant differences between the mean standard scores for the two groups, but the ranges of the standard scores were similar. Ten of the 16 children (63%) with implants performed within 1 SD of the children with normal hearing. As part of a large-scale research study on data obtained from CI recipients, Geers (2003) reported on the performance and predictors of reading skill development. She found that over 50% of the deaf children with CIs performed within 1 SD of the mean, whereas 80% of the scores were within 2 SD of the mean. Connor and Zwolan (2004) examined multiple sources that might influence the reading comprehension skills of deaf children with implants. They reported a mean standard score of 69.8. For children implanted in their preschool ages, the mean standard score was 76.4. The mean score was more than 1 SD below that of hearing children, even in the preschool years.
In the above-mentioned studies different tasks have been used to assess reading comprehension. Spencer et al. (2003) and Connor and Zwolan (2004) assessed reading comprehension with a cloze task. Performance levels reported by Geers (2003) were comparable with the levels reported by Spencer et al. Although they used different tasks and had children of different ages in their samples, the two studies indicate that children with CIs obtain higher reading comprehension levels than deaf children without CIs, when compared to hearing norms. They report that more than half the children perform within the 1-SD range of hearing children. Their results are better than those reported by Connor and Zwolan. Different child and environmental characteristics, however, are known to influence reading and reading-related skills. The participants in these two studies differed, for example, regarding age at onset of deafness, age at implantation, duration of deafness until implantation, duration of implant use, age, IQ, and educational setting. From the information provided, however, no conclusions can be drawn in this respect. The heterogeneity of the samples makes it different to compare the results of these studies.
For our investigation of the reading competence of deaf children with CIs, we examine two major skills: text comprehension and visual word recognition. Although these skills are highly correlated in the hearing population (Aarnoutse & van Leeuwe, 1988), they are regarded as relatively independent (e.g., Oakhill & Cain, 2000; Perfetti, 1985). Despite the fact that these skills have been found to be associated, word recognition and reading comprehension can be differentiated. Stothard and Hulme (1996), for instance, found that the causes of reading comprehension difficulties were different from ones that caused visual word recognition deficits. In addition, Oakhill, Cain, and Bryant (2003) described dissociation of word reading and text comprehension and the underlying abilities that account for their variance.
In their Simple View of Reading model, Hoover and Gough (1990) stated that reading comprehension is the product of decoding and language comprehension. According to this model, the enhancement of auditory speech perception skills after cochlear implantation can influence reading comprehension via three routes: via decoding, via spoken language, and via the contribution of visual word recognition to reading comprehension. First, decoding is an important underlying subskill for visual word recognition and it is likely to be facilitated by better auditory speech perception. In hearing children, decoding is a process dependent on phonological abilities that pertain to the ability to detect, store, and retrieve the basics sound elements of the spoken language (De Jong & van der Leij, 2002). Access to auditory information will lead to the use of letter to sound correspondences and thus provide a basis for phonological decoding. For deaf children who do not perceive spoken sounds via audition, decoding will be difficult. Although decoding does not uniquely rely on phonological knowledge alone (Hanson, Goodell, & Perfetti, 1991; Leybaert, 1993), auditory access to phonological information provides the most efficient way. A higher degree of hearing capacity and better speech intelligibility were found to enhance the ability of deaf children to use phonological coding (Hanson & Fowler, 1987). Therefore, cochlear implantation may lead to enhance decoding skills because it improves the hearing level and speech intelligibility.
However, as discussed by Marschark and Harris (1996), some portion of reading difficulties can be attributed to the inability to hear the sounds of spoken–written language, but other language-related components such as vocabulary and syntax influence reading skills of deaf children as well. Musselman (2000) also discussed the important role of language-specific knowledge (vocabulary and syntax) in reading comprehension of deaf children. Hence, second, as mentioned above, auditory access to spoken language has been demonstrated to positively influence the development of receptive vocabulary. Language comprehension is one of the components of reading comprehension according to the Simple View of Reading. Tunmer and Hoover (1992) argued that, after decoding, comprehension of the discourse in principle relies on the same underlying skills, regardless of whether the discourse concerns written or spoken text. They described reading as a derived skill that builds upon spoken language, and Perfetti and Sandak (2000) found that to be the case for deaf readers. Receptive vocabulary knowledge is reported to be an important factor in reading for hearing children (De Jong & van der Leij, 2002) as well as for deaf children (Marschark & Harris, 1996; Paul, 1996). Oakhill and Cain (2000) further described the relation between oral language and reading comprehension. They reported that difficulties in storytelling and the inability to detect the structure and main point of an event are likely to be the cause of reading comprehension difficulties. There are many reports about improving receptive vocabulary,3 morpho-syntactic and narrative skills after implantation (see Discussion section). Third, we expect visual word recognition to improve, partly based on the enhanced decoding skills based on auditory access to phonology. Visual word recognition, that is, recognition of the printed word, requires two types of knowledge, or skill: decoding skills and word-specific lexical knowledge (Gough & Wren, 1998). In the context of decoding, Gough, Hoover, and Peterson (1996) explained that in order to recognize a written word, the reader has to translate a meaningless set of letters into a recognizable object and locate or activate precisely the right word in the mental lexicon. Harris and Beech (1998) reported significant differences in the word recognition in deaf and hearing children, and phonological awareness appeared an important factor in both groups.
The objective of this article was to investigate the effect of the use of CIs on the literacy of deaf children. As argued above, we define reading skills as reading comprehension and visual word recognition. We expected the auditory information provided by CIs to be sufficient to enhance the poor reading comprehension in deaf children, partly via improvement of visual word recognition skills. In the first part of the study, we investigated to which extent a CI improved the reading comprehension in deaf children. Therefore, the reading comprehension in deaf children with at least 3 years of CI use was contrasted with that in deaf children without CIs and with that in children with normal hearing. In the second part of the study, we evaluated the visual word recognition in deaf children with and without CIs and in children with normal hearing. In the third part of the study we analyzed the relation between visual word recognition and reading comprehension in children with and without CIs. Moreover, we investigated whether the expected positive effect of cochlear implantation on reading comprehension remained when the contribution of visual word recognition was controlled for.
|
Materials and Methods |
|---|
|
TOPIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Participants
This study addressed the reading skills of three groups of children: deaf children with cochlear implants (DCI group), deaf children without cochlear implants (D group), and children with normal hearing (H groups). Subject characteristics of these three different groups of participants are described below.
DCI group.
Our experimental group comprised 50 profoundly deaf children and adolescents with 22-channel Nucleus implant systems of Cochlear. Multi peak (M-PEAK) speech-coding strategies were being used by 15 children up to the time of reading testing. The other children have been using spectral peak (SPEAK) since the initial fitting of the device. The DCI group comprised 25 girls and 25 boys. All the children had parents with normal hearing. Major inclusion criteria for implant recipients in our study were the following: a minimum of 3 years of implant use, age at least 7 years, age at onset of deafness younger than 6 years, full insertion of the electrode array, absence of learning disorders, and participation of their school in the test procedure. At the time of testing, 82 children met the criterion of at least 3 years implant use. Of these 82 children, 50 children met the other criteria4 and were therefore included in the study. Table 1 shows the inclusion criteria and the number of children included and excluded accordingly.
|
It is important to note that there is a fairly heterogeneous group of implant users participating in our study, see the etiological and audiological characteristics of the DCI group, shown in Tables 2 and 3.
|
|
More than half the DCI group had an acquired deafness. In total, 45 children had a prelingual cause of deafness5 (onset of deafness before the age of 36 months). Note the long mean duration of auditory deprivation, which was a consequence of the inclusion of children with Usher's syndrome Type Ib (congenital very profound deafness and progressive retinitis pigmentosa) in our study who received a CI at a high chronological age. These older children with Usher’ syndrome were expected to have limited auditory speech perception abilities after cochlear implantation, due to the long period of degeneration of the nervous system (Snik, Makhdoum, et al., 1997
), but because of their progressive visual handicap they received a CI, in order to be able to maintain auditory contact with the environment.6
For the profoundly deaf children in the DCI group, the mean unaided pure-tone average (PTA)7 of the best ear is high and showed that these children had little to no access to auditory perception of speech. After implantation, auditory sound detection thresholds improved remarkably. For the purpose of indicating the effect of CI use on auditory speech perception, the equivalent hearing loss (EHL) values in the DCI group are reported in Table 3. The EHL value expresses the auditory speech recognition skill as the level of hearing loss (dB) of children with conventional hearing aids that obtain the same speech recognition score as the tested child (Snik, Vermeulen, et al., 1997). The EHL values show that before implantation, the speech recognition skills of the DCI group were at the same level as that of profoundly deaf children and after implantation were comparable with those of hard-of-hearing children, who use well-fitted conventional hearing aids, meaning that their CIs were functioning properly. Before implantation, 37 children had been at six schools for the deaf, 11 at five schools for the hard of hearing, and 2 in two regular schools (mainstream education). At schools for the deaf, at the time prior to the reading assessments, either a bilingual approach (mostly Sign Language of The Netherlands and sign-supported Dutch and until 1999 in many cases total communication) has been used or, in one school, an oral approach. In schools for the hard of hearing the mode of communication generally was sign-supported Dutch. When reading assessment took place, 24 children were at 24 mainstream schools, 9 children were at schools for the hard of hearing, and 17 children were at schools for the deaf.
D group.
We were kindly permitted to use data on visual word recognition of deaf children without CIs that have been collected by Wauters (2005). The group of 504 deaf participants without CIs included almost all the deaf children and adolescents in The Netherlands. The results of the D group were studied and described in detail by Wauters et al. (2006). There were 235 girls and 269 boys in this group. Both parents of the 466 children had normal hearing. Deafness was prelingual in 452 children. The average unaided PTA for the D group was 108 dB (ranging from 80 to 140 dB). Not all the children in this group were using (conventional) hearing aids, and data of the threshold with conventional hearing aids were not available. No children with additional handicaps or learning disabilities were included in this group. There were 423 children receiving special education for deaf children, 46 were receiving education in special schools for the hard of hearing, and 35 children were in mainstream settings. The comparability of the subject characteristics of the two samples of deaf children (one with and one without CIs) is discussed in the Results section in more detail.
H group.
Its important to note that two different reference groups of hearing children were included for comparison in this study. The reading comprehension and visual word recognition comparison data came from two different samples of hearing children. The "hearing" reading comprehension data set comprised the norm data on a standardized test that was used to assess reading comprehension skills (described in the Materials section). We were kindly permitted to use the visual word recognition data obtained from 1,475 children with normal hearing, in mainstream education, tested within the framework of the study by Wauters et al. (2006). Their sample had a mean age of 10.1 years and contained about equal proportions of boys and girls.
Materials
Reading comprehension assessment.
The "Reading Comprehension Tests" (Begrijpend Leestests, Aarnoutse, 1996) standardized for the use in primary schools were used to assess reading comprehension. Each single test, meant for a specific educational grade, consists of a booklet with 10 separate short paragraphs that have to be read silently, followed by 25–30 four-choice questions. The text remains available during answering, and there is no time limit. The number of correct answers is the raw test score. Raw scores are converted into "latent scores" that are applicable to all grades of primary and secondary education (Nijmegen Pupil Monitoring System; Aarnoutse et al., 2000).
As a basis for comparing the reading comprehension scores of the deaf and hearing children, we used instructional ages instead of chronological ages. Instructional age makes an adjustment for the amount of formal reading instruction a child received. This adjustment was necessary because in deaf children the amount of formal reading education cannot be derived directly from their school grade. Generally they do not commence with formal reading instruction at the same age as children with normal hearing. Furthermore, the curricula in special education differ from those in mainstream education. For example, a deaf child may enter elementary education (Grade 1) at the age of 7 years, whereas a hearing child is 6 years of age when entering school. After 4 years of education, this deaf child may be in Grade 3, whereas the hearing child will be in Grade 4. When the deaf child is 11 years old, it might be in Grade 3, a level obtained by hearing children at the age of 9 years. The use of instructional age, the amount of formal reading instruction perceived, yields a grade "equivalent" of 4 years for this deaf child, although the actual educational grade the deaf child is in is Grade 3. An instructional age was computed for each child in the DCI group in the same manner as that used on the D group8 in the study by Wauters et al. (2006). The instructional ages of the children of the DCI group were, on average, 13 months below their chronological ages. In what follows, for hearing children the notions chronological age and grade will be used, and for deaf children these are adjusted to instructional age and grade equivalent.
The broad instructional age range in the DCI group and hence the limited number of participants per grade equivalent complicated further analyses. Therefore, the scores were grouped by combining them into four "(equivalent) grade levels": Grade level A, children in primary Grades 1 to and including 3; Grade level B, children in primary Grades 4 to and including 6; Grade level C, children and adolescents of secondary Grades 7 to and including 9; Grade level D, adolescents in secondary Grades 10 or higher.
Visual word recognition assessment.
Visual word recognition skills were assessed with the two lexical decision tasks described in the study by Wauters et al. (2006). One lexical decision task (the silent-reading task) used a list of 160 monosyllabic letter strings (CVC, CV, or VC) that were either highly frequently occurring words or orthographically legal pseudowords (ratio of 1:3). The words were 30 nouns, 5 adjectives, and 4 homonyms that could be a noun, an adjective, or an adverb. The words were presented in columns at the test form. Respondents have to cross out, columnwise, as many pseudowords as possible in 1 min. The number of items judged correctly within 1 min is the raw score. The other lexical decision task was the two-choices task that consisted of 80 word pairs containing a word and a pseudoword. Respondents have to cross out as many of the nonexisting items of a word pair as possible in 1 min. The words were 62 nouns, 3 adjectives, 1 verb, and 14 homonyms that could be either a noun or an adjective. Children who tend to underestimate their word knowledge or are fast instead of accurate might accept pseudowords. The second task excludes this possibility because for every pair they must mark one item as nonexisting. The raw score was the number of times the word was identified correctly. In the analyses of visual word recognition the grade levels described above were also used.
Procedure
The reading comprehension tests and the visual word recognition tasks were administered to the DCI group similar to the method used to assess the children in the D group. Teachers or peripatetic teachers administered the tests at school, where possible in groups, after receiving detailed written instructions about the assessment protocol. There were regular schools, schools for the hard of hearing, and schools for the deaf involved in the assessment.
|
Results |
|---|
|
TOPIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
To investigate whether the use of a CI improved the poor reading skills of deaf Dutch children, first we describe the findings regarding reading comprehension and second regarding visual word recognition. Next, we discuss the relation between visual word recognition and reading comprehension in deaf children with and without CIs.
Reading Comprehension
To evaluate the effect of CIs on the reading comprehension in deaf children, the scores of the DCI group are compared to those of D group. Moreover, the scores the two deaf groups, children with and without CIs, are then compared to those of hearing children. In conclusion, we put forward considerations regarding the comparability of the characteristics of the DCI group and the D group.
Reading comprehension of deaf children with and without CIs.
Figure 1 shows the individual reading comprehension scores of children in the DCI group per grade equivalent, with the mean scores for the D and the H group depicted as reference curves. The data of the H group are the norm data, as reported in the test manual (Aarnoutse, 1996).
|
The DCI scores show a large variability and are mostly distributed between the mean scores for the H and D group in all grade equivalents. Table 4 shows the mean reading comprehension scores for the DCI group and D group per grade level. The mean scores for both the DCI group and D group increased over the grade levels, but at all grade levels, the DCI group obtained higher mean scores than the D group. The Shapiro–Wilk (SW) test for normality showed that the reading comprehension scores in the D group did not have a normal distribution (SW statistic = .943, df = 504, p = .000). The reading comprehension scores of the DCI group had a normal distribution. For this reason the Kolmogorov–Smirnov two-sample test (KS test) was used to determine whether there were differences in the reading comprehension score distributions of the DCI group and D group. The nonparametric KS test has the advantage of making no assumptions about the (normality of the) distribution of the data because no parameters, such as mean or median, are tested. This test concerns the agreement between the two cumulative frequency distributions that are compared. The difference between the two frequencies is determined for fixed score intervals. The KS test focuses at the largest of these deviations (Siegel & Castellan, 1988
). Significant differences were found between the distributions of the DCI group and D group at all grade levels (see Table 4). We also computed the effect size of the difference between the group means, Cohen's d (Table 4). This parameter expresses the standardized difference between the means of two groups, d = (µ1 – µ2)/SD(
). The effect size of the differences between the DCI group and the D group at Grade levels A, C, and D are large,9 whereas at Grade level B it was medium. Table 4 shows that the reading comprehension of the children with CIs was better than that of their deaf peers without CIs. Mostly the differences were large.
|
Reading comprehension of the deaf children with and without CIs compared to the hearing norm.
After having found that the reading comprehension in the DCI group surpasses that in the D group, the reading comprehension skills of the two deaf groups were compared to the hearing standard. The z scores were computed per grade, based on the means and SDs of the hearing norm group. Figure 2 shows the mean z scores. By definition, the mean reading comprehension z scores of hearing children were equal to zero. The relative differences between the DCI group and the D group compared to the mean z scores for hearing children are expressed in the number of hearing SD units of deviation from the hearing grade-level mean z scores. The scores of the DCI group leveled off at between –3 and –4 SD.10 Relative performance in the D group, however, decreased with increasing grade level, up to an SD of –8.6 at Grade level D.
|
Next the difference between the two groups of deaf children and the hearing norm was also quantified in another way. Table 5 shows the percentage of children in the DCI and the D group per grade level who performed "below" or "in-or-above" the 95% confidence interval of the children with normal hearing. The 95% confidence interval lies between –1.96 and 1.96 hearing SD (see Figure 2) of the hearing mean z score (0), computed per grade level. Whether these percentages differ was determined with chi-square tests. The percentage of children in the DCI group who performed in-or-above the 95% confidence interval of the children with normal hearing was significantly higher than that for the D group at all grade levels. These data showed that the performance of the deaf children with CIs deviated less significantly from the children with normal hearing than that of the deaf children without CIs.
|
2 values are shown of the differences between these percentages and one-tailed significance, per grade levelGroup characteristics of the deaf children in the DCI group and D group.
As discussed in the introduction, many factors are associated with reading comprehension in deaf children. Geers (2003)
reported that performance intelligence quotient (PIQ), family socioeconomic status, gender, later onset of deafness, educational placement, and ethnic background are important factors. In order to exclude the possibility that such child and environmental variables contributed to the difference in reading comprehension between the DCI group and D group, the comparability of the groups was examined. The variables the influence of which we could investigate, based on the available data, were gender, pre- or postlingual onset of deafness, (preimplant) educational placement, and ethnicity (Dutch or other nation of origin [no Dutch mother tongue] of the parents). The PIQ's were not available for all the children in the DCI group, and their effect could therefore not be investigated. However, Oakhill et al. (2003) found that PIQ did not have any effect on reading comprehension or on visual word recognition in children with normal hearing. As already mentioned in the Participants section, no children with learning disabilities have been included in the D and DCI group. Unfortunately, data were lacking on the socioeconomic status of the families of the children from either group. The data concerning the factors we could study were classified for the DCI group in the same way as those for the D group in the Wauters’ (2005) study. Onset of deafness was classified into prelingual (before the age of 36 months) or postlingual. The factor preimplant educational setting had the following levels: schools for the deaf, schools for the hard of hearing, and mainstream education. The factor nation of origin of the child's parents was Dutch: two Dutch parents; Dutch and other: one Dutch parent and one parent not of Dutch origin; or other: both parents not Dutch. We carried out chi-square tests to determine the comparability between the DCI and D groups with respect to these factors. Table 6 shows these results. Gender can be considered to be equally distributed in the two groups. The same applied to the variable pre- or postlingual onset of deafness. Significant differences were found in the percentages of participants classified according to parents' native origin and preimplant educational setting. The DCI group contained more children with "Dutch-only" parents and the D group more "Dutch and other." In the DCI group more children were attending schools for the hard of hearing, whereas relatively fewer were attending schools for the deaf than in the D group. In the preimplant situation in the D group, most of the children had been receiving mainstream education than in the DCI group. Hence, analyses of variance were carried out to determine the effect on reading comprehension within the DCI group of the variables that differed between the DCI and D group: parents nation of origin and preimplant educational setting. We included instructional age as the covariate because the correlation between age and reading comprehension, rs = .68, p < .01, was strong. Preimplant educational setting showed no significant effect on reading comprehension. The nation of origin of the parents, however, showed a marginally significant effect (F(2, 46) = 3.36, p = .044). From the outcomes of the chi-square tests we can conclude that gender and age at onset of deafness were not causing the differences in the reading comprehension of the DCI group and D group. The analysis of covariance (ANCOVA) showed that preimplant educational setting did not affect reading comprehension outcomes. Based on the last ANCOVA, the only factor that could affect the reading comprehension difference between the DCI group and D group, apart from the use of a CI, was the nation of origin of the parents.
|
We considered it important to further analyze the data without the influence of the nation of origin of the parents. Although we found only a marginal significant effect of the nation of origin of the parents on reading comprehension in our sample, Wauters et al. (2006) reported that in her group of deaf children with conventional hearing aids 20% of the poor readers were children with parents of another nation of origin compared to 5% of the better readers. Therefore, we reanalyzed the reading comprehension scores of the two groups of deaf children, restricting the analyses to native Dutch children (children with parents of Dutch origin) only. The limited number of implanted children of other ethnic backgrounds, n = 4, in the DCI group did not permit further statistical analysis of their data. Table 7 shows the number of DCI and D group native Dutch children per grade level, the mean visual word recognition scores and SDs. The KS test showed significant differences between the DCI and D distributions at all grade levels. In order to express the effect size of the standardized difference between the group means, Cohen's d values were computed for the native Dutch group. The KS p levels and Cohen's d values are included in Table 7.
|
The native Dutch children in the DCI group and the D group performed better on average at every grade level than the total groups (see also Table 4). The mean scores of the native Dutch children of the DCI group are higher than those of the D group. The distributions of the DCI group and the D group of native Dutch children showed significant differences at all grade levels. Effects were large at all grade levels.
Inspection of the latent scores revealed that four of the five DCI children with the lowest scores per grade equivalent were children of two not-Dutch parents. The performance of these nonnative Dutch children at higher grade equivalents was better than that of nonnative Dutch children at the lower grade equivalents, though.
Conclusions regarding reading comprehension.
In conclusion, we found that at all grade levels, the deaf children with CIs obtained higher mean reading comprehension scores than the deaf children without CIs. There were significant differences in favor of the CI group, between the score distributions of the two groups of deaf children, and the effect size was substantial. These findings were not due to group differences of the variables gender or pre- or postlingual onset of deafness. Furthermore, we have shown that preimplant educational setting did not have any significant effect on reading comprehension within the group of children with CIs. Although there was a marginal effect of the nation of origin of the parents within the children with CIs, a comparison of the reading comprehension scores of the native Dutch children showed that the performance of the CI users was better than that of the children without CIs at all grade levels. Whether the use of another language than Dutch in the home environment was the only cause of this effect cannot be concluded from our data. Other factors that are known to negatively influence reading cannot be excluded from being present in these children's environments. Such factors are, for instance, a lower educational level of the parents and the limited use of sign language at home. However, no structural differences between the school environments of these and other children are known.
Unfortunately, the positive effect of cochlear implantation did not prevent the reading comprehension of the children with CIs from lagging far behind that of the children with normal hearing. On average, the performance of the children with CIs was still more than 3 SDs below the hearing norm, whereas for the children without implants the average was even as low as –7 SDs.
Visual Word Recognition
The second part of our study assessed the visual word recognition skills in deaf children with CIs, using the two lexical decision tasks described in the Materials and Methods section. The visual word recognition in deaf children with CIs is described, and their skills are contrasted with those in deaf children without CIs and those in children with normal hearing. Furthermore, the relative performance of the two groups of deaf children compared to the hearing norm was evaluated.
In order to derive a more robust variable, we pooled the data from the two lexical decision tasks. The strong Spearman rank correlations between the z scores on these two tasks, in the DCI group (rs = .80, p < .000) and the D group (rs = .72, p < .000) justified this. In our analyses we therefore used the mean of the z scores from Task A and Task B, further referred to as visual word recognition.
Visual word recognition of children with and without CIs and hearing children.
For the purpose of investigating differences in visual word recognition skills between deaf children with and without CIs and children with normal hearing, we carried out pairwise KS tests. (The scores of the children in the D group did not follow a normal distribution [SW statistic = .977, df = 504, p = .000], but those of the DCI group did.) But first the descriptives of visual word recognition in the three groups of participants are summarized in Table 8. The mean scores for the DCI group were higher than those for the D group, at all grade levels. In the two deaf groups an increase over grade levels was observed, however, as in the hearing group.
|
Table 9 shows the p levels of the differences between the distributions of scores of the three pairwise combinations of groups per grade level. The effect size, Cohen's d values, was computed and is also included. No significant differences were found in the distributions of the visual word recognition scores between the DCI group and the D group at primary education levels (Grade levels A and B). At the secondary level (Grade levels C and D), the visual word recognition skills of the DCI group were better than those of the D group. There were no significant differences in the distribution of scores between the DCI and the H group. In the D group, however, there were differences in favor of the H group at Grade levels B, C, and D. However, at Grade level A, the D group performed better than the H group. In the cases where the differences between the pairwise comparisons of visual word recognition score distributions were significant, the effect sizes of differences between means were medium: DCI versus D group at Grade levels C and D; D versus H group at Grade levels B, C, and D; and in one case the effect size was small: D versus H group at Grade level A, in favor of the D group.
|
However, similar to the reading comprehension results, the visual word recognition in the two deaf groups relative to that of a reference group of hearing children was expressed in two more commonly used and comprehensible measures, which are parametric. The raw scores on lexical decision Tasks A and B were transformed into z scores, based on the hearing grade-level mean scores and SDs. (The hearing group is not the same group as the normative group of hearing children for the reading comprehension scores.) Figure 3 shows the z scores of visual word recognition for the DCI group and the D group, relative to the hearing mean, expressed in hearing SD units. By definition, the mean z scores for hearing children were equal to zero. The mean for the DCI group was .07 and that for the D group was –.60.11
|
As with the raw scores, children in Grade levels A and D in the DCI group obtained higher mean z scores than the H group mean scores, although the range is very large. For the D group the mean scores were lower than the H group mean scores. Table 10 shows the percentages of children performing below or in-or-above the 95% confidence interval of the children with normal hearing. These data showed that the DCI group had significantly better results than the D group at all grade levels except for grade level A, in which there was no difference between the two deaf groups. In this case the performance of both the DCI group and the D group was better than that of the H group.
|
Conclusions regarding visual word recognition.
An important finding was that, although the visual word recognition skills of deaf children without CIs were below those of hearing children, after the same amount of reading instruction, the visual word recognition skills of the children with CIs did not differ significantly from those of the hearing children. Furthermore, at secondary education level, the visual word recognition skills of deaf children with CIs were better than those of children without CIs. No differences between children with and without implants, however, were found at primary level. From the z scores we can conclude that in comparison with the visual word recognition skills of the children with normal hearing, the abilities of the two groups of deaf children were relatively good, that is, mostly within 2 SD of the hearing grade-level mean.
The Relation Between Visual Word Recognition and Reading Comprehension
In this section we examine the relation between visual word recognition and reading comprehension in deaf children with and without CIs. First we studied the association between these two skills in the DCI group and the D group. Moreover, we evaluated whether the reading comprehension in deaf children with CIs remained better than that in children without CIs when the influence of visual word recognition was controlled. As shown above, the reading comprehension in deaf children with CIs was significantly better than that in deaf children without CIs, whereas no significant differences were found in their visual word recognition in elementary education, but only in secondary education. Therefore, we analyzed the contribution of visual word recognition skills to the reading comprehension in the DCI and D group at the four grade levels.
Spearman rank correlations were calculated between the visual word recognition scores and the reading comprehension scores. This value was rs = .73, p < .01, in the DCI group and rs = .61, p < .01, in the D group. Figures 4a and b show the scatter plots of the visual word recognition and the reading comprehension scores. The distributions of the scores of visual word recognition and of reading comprehension of the deaf children without and with CIs are fairly similar, suggesting that the visual word recognition might not be responsible for the difference in reading comprehension between the two groups of children.
|
Further examination of the data aimed to determine the explanatory power of visual word recognition on reading comprehension. Separate linear regression analyses were carried out on the DCI and the D group. In the DCI group, visual word recognition explained 52% of the variance of reading comprehension, whereas in the D group only 32% of reading comprehension variance was explained by visual word recognition. We found large differences in the results of the D group between primary education (34% explained variance) and secondary education (only 21%).
We found that in the DCI group visual word recognition accounted for 13% of the variance in reading comprehension, after the variance due to instructional age was removed (beta = .49, t = 3.85, p = .000).
An ANCOVA was carried out to evaluate the effect of the factors group (with two factor levels: CI vs. no CI) and grade level on reading comprehension, with visual word recognition as the covariate. No interaction of the factors group (CI vs. no CI) and grade level with the covariate visual word recognition was found, which implies parallel slopes and homogeneous regression and justifies straightforward ANCOVAs. Eliminating the effect of visual word recognition, we found significant effects of group (F(1, 513) = 26.63, p < .001) and grade level (F(3, 513) = 4.67, p < .005) on reading comprehension. Thus, the effect of group on reading comprehension was significant and applies to all grade levels. The effect of group was still present after the contribution of visual word recognition skills on reading comprehension had been ruled out. This suggests that the difference in reading comprehension skills between the deaf children with and without CIs was not explained by differences in word recognition skills alone.
Table 11 shows that the differences in reading comprehension between the two deaf groups at each grade level are still present when the reading comprehension scores were adjusted for the effect of visual word recognition. The predicted means for reading comprehension after eliminating the influence of visual word recognition are depicted. Comparison with the results summarized in Table 4 shows that the differences between the scores of children with and without implants were of a similar magnitude after adjusting for the effect of visual word recognition.
|
We demonstrated that there was high correlation between visual word recognition and reading comprehension in the two groups of deaf children but that the contrasts between visual word recognition skills of the deaf children with and without CIs did not explain the differences in reading comprehension. This is in accordance with the Simple View of Reading model (Hoover & Gough, 1990
) because language competence is a second factor that on its own or in combination with visual word recognition might explain the variance in reading comprehension.
The Relation Between (Spoken) Language Comprehension and Reading Comprehension
In search of causal factors that might offer an explanation for the observed difference in the reading comprehension of deaf children with and without CIs, we investigated the other major contributor to reading comprehension, spoken language comprehension (Vermeulen, in press). The results of evaluations of postimplant (spoken) receptive vocabulary tests are reported briefly here, as they were not the main focus of this article. Receptive vocabulary was assessed with standardized tests for children with normal hearing in the age range of 2–12 years. The DCI group was evaluated preimplant and at yearly intervals until 36 months after implant. Based on the language age equivalents, the ratio of language development was computed (months of language development over actual number of months). Prior to implantation only 24 of the 50 children were able to perform the test. For the other 26 children the spoken receptive vocabulary level was too low to score. The ratio of the 24 best performing children before implant was 0.63. In the first year after implantation, the ratio for children in scorable range (n = 41) was 0.24, in the second year the ratio for children in scorable range (n = 50) was 0.76, and in the third year after implant (n = 50) it was 1.19, a remarkable rate of improvement. By comparison, the ratio of language development of deaf children who use conventional hearing aids is in general considered to be only 0.5, half the rate of children with normal hearing (Connor, Graig, Raudenbush, Heavner, & Zwolan, 2006; Svirsky et al., 2000). The ratio of the children with CIs in the third year after implantation is higher than that of children with normal hearing. The interpretation of this finding is difficult though because the age of the children is much higher than that of the norm group (see Discussion section).
The association of receptive vocabulary with reading comprehension was strong. We found that the receptive vocabulary age equivalent accounted for 29% of the variance in reading comprehension, after the variance due to instructional age was removed (beta = .54, t = 2.88, p = .015). Thus, receptive vocabulary explains a substantial part of the variance of reading comprehension, whereas visual word recognition explains less.
There is an important finding regarding the effect of CI on reading comprehension and the role of receptive vocabulary in this effect that we consider worthwhile to report here. After cochlear implantation the auditory speech perception skills in the DCI group improved significantly. Linear regression analyses showed that postimplant receptive vocabulary was strongly associated with earlier postimplant auditory speech perception skills. Instructional age and auditory speech perception at 12 months together explained 41% (adjusted R2 = .386) of the variance in receptive vocabulary age equivalent at 24 months after implant. Receptive vocabulary at 36 months was best predicted by auditory speech perception at 24 months after implantation. Instructional age and EHL at 24 months after implantation together explained 40% (adjusted R2 = .37) of the variance in receptive vocabulary age equivalent at 36 months after implantation. These findings strongly suggest that there is a causal relation of the auditory speech perception skills with later receptive vocabulary.
|
Discussion |
|---|
|
TOPIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The objective of our study was to investigate whether the use of a CI was associated with improvements in the poor reading skills of deaf Dutch children. To evaluate this, the reading comprehension in children who had been using their CI for at least 3 years was compared to that in deaf children without CIs. In addition, the scores obtained from these two deaf groups were evaluated in relation to those of two groups of children with normal hearing. To gain further insight into the causes of differences in reading comprehension between the deaf children with CIs and those without, we investigated the relation between visual word recognition and reading comprehension.
In the introduction, we mentioned that reading comprehension in deaf children without CIs seemed to be lagging behind that of hearing children (Holt, 1993; Holt et al., 1996; Traxler, 2000). A recent large-scale study by Wauters et al. (2006) showed that deaf children without CIs in The Netherlands also had very low reading comprehension levels compared to children with normal hearing. As expected, we found that the reading comprehension performance of deaf children with at least 3 years of CI use was better than that of deaf children without CIs. Our results show that the use of a CI was associated with improvement of the reading skills of deaf children. Within the available data, there were no secondary group characteristics that could account for the differences between the scores of the deaf children with and without CIs. Nevertheless, the reading comprehension of deaf children with CIs was still lagging behind that of the children with normal hearing. As a group, the average performance for the deaf children with CIs was –3.6 SD below the norm, but it was much better than the mean for the deaf children without CIs (–7.2 SD). Over the four grade levels, the percentage of children with CIs whose performance was within or above the 95% confidence interval of the children with normal hearing varied between 25% and 50%. At each grade level this percentage was significantly higher than that of the deaf children without CIs.
Until now only a few studies have assessed reading comprehension in deaf children with CIs. In agreement with these studies, we found that the reading comprehension skills of the children with CIs were better than those of the deaf children without implants, but was poorer than that of the children with normal hearing. The reading comprehension level of the children with CIs in our study was slightly lower than the levels reported in the studies by Geers (2003) and by Spencer et al. (2003). About half their children with CIs performed within the 1-SD range of the norms of hearing children. In our sample, half the youngest children performed within 1.96 SD of the hearing mean, whereas at the three highest grade levels, this applied to only about 25% of the children. The differences between our findings and those reported by Spencer et al. (2003) and by Geers may have been caused by subject and environmental characteristics or by the tasks used to measure reading comprehension. Group characteristics that deviated included the speech-coding strategies of the speech processors. The children who participated in our study were using M-PEAK speech processors. Recent coding strategies, such as SPEAK and Advanced Combined Encoder, with wide dynamic ranges, are known to provide auditory input with speech characteristics that facilitate the use of phonological coding. Furthermore, our participants had a relatively long duration of deafness and an older age at implantation, which is known to limit auditory speech perception after cochlear implantation. This is shown by the lower auditory speech perception outcomes of the children with Usher's syndrome, with a long duration of deafness. Thus, if the positive effects of CIs on literacy take place via enhanced auditory perception, possible explanations for differences between the findings reported in the recent literature on literacy in children with CIs and the results of our study are the indication criteria for implantation applied by the implant center and the available coding strategies.
In an attempt to explain the better reading comprehension of the deaf children with CIs, we investigated the relation between visual word recognition and reading comprehension. According to the Simple View of Reading model devised by Hoover and Gough (1990), reading comprehension is the product of word recognition and language comprehension, and visual word recognition explains a large part of the variance in reading comprehension. Lower visual word recognition skill levels were reported by Harris and Beech (1998). In other studies, no significant differences were reported between deaf children without CIs and children with normal hearing (Burden & Campbell, 1994; Fischler, 1985). However, the latter studies were performed on older children. The visual word recognition skills of deaf Dutch children without CIs were only slightly poorer than those of children with normal hearing (Wauters et al., 2006). In this article we found that the visual word recognition skills of the children with CIs were better than those of children without CIs only at secondary education level, not at primary education level.12 Our data might reflect a trend similar to that described by Burden and Campbell (1994), who did not find any differences in visual word recognition skills between relatively old deaf children without CIs and their peers with normal hearing, and by Fischler (1985).
When Merrills et al. (1994) investigated the contribution of visual word recognition to reading comprehension, they found that poor visual word recognition skills only explained part of the reading difficulties in deaf children without CIs. Marschark and Harris (1996) argued that, apart from phonological recoding, vocabulary and syntax were expected to account for reading difficulties. Indeed, Wauters et al. (2006) reported similar results in deaf Dutch children without CIs. They stated that even if the visual word recognition scores of the children without CIs would have been age appropriate, their reading comprehension would have still lagged behind that of children with normal hearing. We found that although the reading comprehension in deaf children with CIs was poorer than that in children with normal hearing, there were no large differences in visual word recognition. Nevertheless, these two skills were correlated in the children with CIs (note that the reference data for reading comprehension and visual word recognition were from two different groups of hearing children), which is also the case in the hearing population. Differences in visual word recognition between the children with CIs and the deaf children without CIs could not explain the contrast in reading comprehension skills between these two groups. The better reading skills of the children with CIs can be attributed to their implants and must depend on other factors than visual word recognition alone.
In the Simple View of Reading model (Hoover & Gough, 1990), reading comprehension is defined as the product of single-word reading and listening skill (language competence). Part of the difference in reading comprehension between the deaf children with and those without CIs should therefore lie in language comprehension according to the model. Wauters et al. (2006) argued that the reading comprehension problems of deaf children without CIs might also be the result of linguistic comprehension factors. We argue that it is highly plausible that language skills make a prominent contribution to the better reading comprehension of children with CIs. The results regarding the increase of postimplant receptive vocabulary of our group of children with implants do suggest that this is the case (Vermeulen, in press). The language development ratio of this group of children prior to implantation showed that the children developed according the expectations of deaf children with conventional hearing aids up to the time of implantation, implying that no prior advantages in vocabulary knowledge are likely to have caused the better reading comprehension of the children with CIs.
Frequent reports have been published on enhanced language skills after cochlear implantation and many authors described increases in the development of receptive vocabulary after cochlear implantation13 (e.g., Connor et al., 2006; Svirsky, Stallings, Lento, Ying, & Leonard, 2002; Tomblin, Spencer, Flock, Tyler, & Gantz, 1999; Vermeulen et al., 1999).
Fewer data are available on the development of morpho-syntactic skills. Spencer, Tye-Murray, and Tomblin (1998) reported that CI users demonstrated better perception and comprehension of bound morphemes than hearing aid users. Svirsky et al. (2002) and Szagun (2004) compared the morphological development of children with CIs to that of children with normal hearing and found different patterns. However, in the latter studies, the tasks involved speech production. Nevertheless, all the studies indicate that auditory perceptibility strongly influenced the development of morphological skills in children with CIs. Markers (morphemes) that are best perceived through audition are the first to be produced. Another language component that is associated with reading comprehension is narrative skill. Several studies reported improvements in narrative abilities after implantation (Nikolopoulos, Dyar, Archbold, & O'Donoghue, 2004). Furthermore Crosson and Geers (2001) found not only that narrative skills improved after cochlear implantation but also that they had predictive value for reading comprehension in children with CIs.
In conclusion, cochlear implantation was positively associated with reading comprehension of deaf children. This is a very important finding, improved reading comprehension may increase their academic achievements and promote their participation in society. In our opinion, the achievement of auditory access to spoken language caused the swift progress in reading comprehension. Although the visual word recognition skills of the children with CIs were better than those of the deaf children without CIs, this does not explain the superior reading comprehension skills of the CI users. Our preliminary findings regarding enhanced postimplant receptive vocabulary, that it has a strong association with postimplant auditory speech perception, show that language comprehension skills after cochlear implantation indeed explain a significant part of the variance of reading comprehension scores.
The results of this article did not provide a clear answer to the question of how (i.e., through which specific aspects of language competence) the improvements in reading comprehension were actually brought about, but our data on receptive vocabulary suggest that that component is an important factor in the causal chain. Further analyses of the relation between language development after cochlear implantation and the relation with the improved reading comprehension in children with CIs will provide information that will be very useful for educational practice.
|
Acknowledgments |
|---|
This study was supported by University Medical Centre St. Radboud, Nijmegen, and Viataal, Sint-Michielsgestel, The Netherlands. No conflicts of interest were reported.
|
Notes |
|---|
1. A CI enables sound perception in profoundly deaf children who were unable to perceive speech with conventional hearing aids before. In profoundly deaf persons, CIs generally provide much more auditory input, in a broader frequency range, than conventional hearing aids do. A CI works in a totally different manner than conventional acoustic hearing aids, which just amplify auditory signals that still have to be processed by a damaged cochlea. The internal part of a CI transfers signals that are already coded by a speech processor into electrical stimuli directly to the auditory nerve endings in the cochlea.
2. The notion "visual" is generally not employed in this context. We use it to distinguish the notion from auditory word recognition.
3. When the term "vocabulary" is used without any further specification, it refers to spoken language only.
4. We excluded children who had severe medical problems or had moved abroad and adolescents who were not willing to cooperate.
5. As described in the Results section, there was no significant difference between the percentage of children