Language Problems after Childhood: a Comparative Study of Narrative Abilities of Adolescents with Congenital Hearing Impairment (CHI) and Adolescents with Specific Language Impairment (SLI)
L. de Wael, research MA student Linguistics, University of Amsterdam Supervisor: Prof. Dr. J.C. Schaeffer, University of Amsterdam
Second reader: Dr. J.E. Rispens, University of Amsterdam
Key words: Narrative Abilities; Adolescents; Specific Language Impairment; Congenital Hearing Impairment Abbreviations: CELF-4-NL: Clinical Evaluation of Language Fundamentals – fourth edition – Dutch version; CHI: congenital hearing impairment;
CI: cochlear implant; MLU: mean length of utterances;
PTA: Pure Tone Average; SLI: specific language impairment;
SD: standard deviation; TL: typical language and normal hearing;
TTR: type/token ratio; VMH: Vulnerable Markers Hypothesis;
Abstract
Objectives: This study compared language skills of Dutch-speaking
adolescents with Specific Language Impairment (SLI) and adolescents with Congenital Hearing Loss (CHI). We investigated whether similarities and differences in performance between the groups could be explained by the Vulnerable Markers Hypothesis (VMH) (Bishop, 1994). In addition, we tested the sensitivity of a standardized language test in detecting language difficulties in adolescents. Comparing language skills of adolescents with CHI and with SLI would increase our insight in the role of auditory input in language acquisition, and the impact of impaired acquisition on later language outcomes.
Design: Speech samples from a narrative generation task, and results from
4 standardized language tasks were analyzed of 11 adolescents with SLI, 10 adolescents with CHI, and 5 adolescents with typical language and normal hearing (TL), all aged between 14 and 17 years old. Detailed analyses of morphosyntactic, lexical, and fluency errors were conducted on the speech samples and performance was compared across groups. Sensitivity of each task was tested by assigning a „pass‟ or „fail‟ to each participant, based on scoring within 1 SD from the TL norms.
Results: No differences in performance on the standardized language tasks
were found between any of the groups. On the narrative task, the SLI group made significantly more lexical and morphosyntactic errors than the TL group and was significantly less fluent than both TL and CHI groups. No other significant differences were found between CHI and TL groups on the one hand, and between CHI and SLI groups on the other, although large variation within impaired groups and the small number of participants may blur these results. The morphosyntactic errors made by the CHI group were very similar to those found in the SLI group, with the exception of determiner errors: substitutions of the neuter determiner het were common in both SLI and CHI groups, but half of the CHI group (5/10) also made a substantial amount of substitutions of the common determiner de. This was rare in the SLI group (2/11) and not attested in the TL group (0/5). Sensitivity was higher for the narrative tasks than for the standardized tasks: 11/11 SLI and 7/10 CHI individuals „failed‟ for morphosyntax. The most sensitive of the standardized tasks (Sentence Repetition) resulted in only identifying half of the SLI and half of the CHI as „failers‟. A negative correlation was found between morphosyntactic errors and Sentence Repetition.
Conclusion: CHI and SLI during childhood leads to persisting language
difficulties in adolescence, mainly at the level of morphosyntax and the lexicon. Our findings are generally in line with the VMH: firstly, a complex task such as a narrative leads to more errors than isolated language tasks. Secondly, difficulties in both groups arise in complex and vulnerable features of language and difficulties of adolescents with CHI are colored by the perceptual non-saliency of certain grammatical features. Nevertheless, difficulties of adolescents with CHI extend to other, more salient features and performance is largely similar to adolescents with SLI. It is possible that hearing loss, besides its direct effect on perception, also leads to a higher cognitive load and slower processing of linguistic input, which may explain difficulties with complex material, similar to what adolescents with SLI experience.
1. Introduction
Both individuals with Specific Language Impairment (SLI), and individuals with Congenital Hearing Impairment (CHI) show language problems, mainly in the area of morphosyntax (Delage & Tuller, 2007; Hammer, Coene, Rooryck, & Govaerts, 2014). As the source of these problems in children with CHI is better identified (i.e. reduced and often degraded auditory input due to hearing loss), directly comparing children with CHI and SLI may contribute to our understanding of language impairments.
In this study, we assumed that limited processing capacity in individuals with SLI, and difficulties perceiving non-salient features from the input in individuals with CHI will lead to similar problems with morphosyntax. We took as our starting point the Vulnerable Markers Hypothesis (Bishop, 1994). The VMH, originally proposed for SLI, assumes that problems arise in vulnerable areas of language (i.e. computationally difficult grammatical rules that require high amounts of input in order to acquire them) because individuals with SLI have limited processing capacity. The reduced capacity not only leads to difficulties with acquiring grammatical rules, but also leads to difficulties with applying those rules in complex situations. More recently, the VMH was also found to explain difficulties found in adults with CHI (Huysmans, de Jong, van Lanschot-Wery, Festen, & Goverts, 2014), who also experienced problems in more complex situations, but mainly with grammatical features that are perceptually non-salient and thus likely to be missed by individuals with CHI. They thus receive less input of those features, which will result in similar difficulties with and thus have difficulties with processing. In this study, we tested whether differences and similarities between the two groups could be explained by the VMH.
Up to date, only few comparative studies exist between CHI and SLI. Moreover, those few studies differ widely in language of investigation and the age at which performance is compared. Therefore, it remains unclear how differences in developmental rate (i.e. some language skills may develop more slowly in SLI than in CHI or vice versa) influence these outcomes. We thus need comparative studies after childhood, when language skills are assumed to have fully developed. This reduces the possible influence of age factors and provides an interesting perspective on language impairments and the role of auditory input in later language outcomes.
With the exception of a few (Briscoe, Bishop, & Norbury, 2001; Delage & Tuller, 2007; Hansson, Forsberg, Löfqvist, Mäki-Torkko, & Sahlén, 2004; Norbury, Bishop, & Briscoe, 2001; Tuller & Delage, 2014), comparative studies between SLI and CHI after the age of 10 are nearly absent. To the best of our knowledge, no such comparative study is done on Dutch-speaking adolescents, and thus important questions remain open. The current study aimed to fill this gap by comparing the narratives of adolescents with SLI and adolescents with CHI, all aged between 14-17 years old.
2. Background
2.1. Specific Language Impairment
SLI is usually defined as a neurological developmental disorder (DSM-IV, American Psychiatric Association, 2000) in which language acquisition is lacking behind without any apparent reason, while other aspects of development are normal (Bishop, 1997). The prevalence of SLI is estimated around 7% of the population (Tomblin, Records, Zhang, Smith, & O‟Brien, 1997), although numbers depend on the definition and inclusion criteria used. Core problems are situated in the area of morphosyntax, but differ widely per language (de Jong, 1999; Duinmeijer, 2016; Leonard, 2009): for English, verbal agreement and past tense inflections are identified as vulnerable areas (Leonard et al., 2003), and for French (and other Romance languages), pronominal accusative clitics such as le and la (a) are found to be troublesome for children with SLI (Jakubowicz, Nash, Rigaut, & Gérard, 1998). Furthermore, also problems with syntactic complexity have been attested in children with SLI: they show particularly problems with forming wh-questions (b and c) and relative clauses. This was found for several languages, such as Greek (Stavrakaki, 2006), and Hebrew (Friedmann & Novogrodsky, 2011).
(a) la giraffe le mord
the giraffe him bites ‘the giraffe is biting him’
(Tuller, Delage, Monjauze, Piller, & Barthez, 2011, p. 425) (b) Who hit the elephant? (subject question)
(Stavrakaki, 2006, p. 385) As the focus of the current study is Dutch, we will discuss problems of Dutch-speaking children with SLI in more details. Just as in English, verbal agreement has been found problematic: De Jong (1999) found that the children with SLI in his study (N = 35, mean age 7.8) significantly more often omitted third person singular finite verb morpheme –t (d), produced more often a third person plural subject together with a verb that is not marked for plural (e), and produced fewer finite verbs (f) as compared to age matched children with typical language (TL 1) (N = 35, mean age 7.6), as well as compared to younger TL children with similar mean length of utterances (MLU; a measure often used to express language level) (N = 20, mean age 5).
(d) Die gooi* „m in de lucht
That.one throw (unmarked verb form) him in the air ‘that one throw* him in the air’
(e) dat doet* altijd mijn vader en moeder
that does (singular verb form) always my father and mother ‘my father and mother always does* that’
(f) en dan mama papa wakker maken*
and then mother father awake make (infinitive) ‘then, mother wake* up father’
(modified from de Jong, 1999, p. 70-72) Similar findings were reported by a study from Weerman, Duinmeijer, & Orgassa (2011), who compared young (N = 25, mean age 7;3) and older (N = 9, mean age 12;7) Dutch acquiring children with SLI with a younger TL group of language-matched children (N = 20, mean age 4;10). Whereas the TL group correctly assigned verbal agreement in 91.2% of the cases, the young SLI children, who were more than two years older, did so only in 80% of the cases. Interestingly, the older
1
As the focus of the current study is on language outcomes in adolescents, we specifically chose the term „typical language‟ over the more commonly used term „typically developing‟ to refer to individuals (children, adolescents, and adults) that acquired language in a typical way. Typically developing may be considered confusing when referring to adolescents, as we assume the core language skills are already fully developed. To be consistent in the terminology throughout this paper, we also used TL for referring to children, who are of course still fully developing their language skills. Additionally, the term TL also implies normal hearing, as we will compare TL groups with CHI groups.
SLI group scored significantly better (95.2%) than the younger one (and no longer different from the TL group), suggesting that children with SLI do improve their production of verbal agreement, but more slowly than children with TL.
In addition to verbal agreement errors, De Jong (1999) reported problems with past tense marking in Dutch acquiring children with SLI: the children in his study (participants reported before) used significantly fewer past tense forms (both regular and irregular) than age-matched and language-matched controls, and substituted them by present tense forms more often in obligatory contexts (g). He also found a significantly higher use of „dummy auxiliaries 2‟ (h) in the SLI group. The author interpreted this as a strategy to avoid verbal inflection, as the main verb no longer needs inflection.
(g) toen val* alles d‟r uit then fall.PRS.1SG* everything out
‘then everything fall* out’ (De Jong, 1999 p.74)
(h) die meneer die ging brood maken
that man that AUX.PST.SG bread make.INF
‘that man made bread’ (De Jong, 1999 p.127)
Another problematic grammatical feature for Dutch-speaking children with SLI is the assignment of nominal gender. This was extensively studied by Weerman et al. (2011). The Dutch gender system (see Table 1) distinguishes between common and neuter nouns, for which gender is expressed on the definite determiner and the attributive adjective combined with that noun (Weerman et al., 2011). As the Dutch nouns provide no clear grammatical cues in the attribution of gender, this system is very opaque and thus hard to acquire for both typically and atypically developing children. According to Blom, Polisenska, & Weerman (2008), typically developing children only master this around the age of 7 (in contrast with e.g. verbal agreement, which is, according to Weerman et al. (2011) already mastered around 3 years of age).
2 In Dutch, the verb gaan „to go‟ can be used as an auxiliary to express inchoative aspect. In the past tense (gingen „went‟) it is often used redundantly, as it has no semantic meaning. In this case, the auxiliary carries the tense marker and is combined with the infinitive of the main verb. Thus the use of
Common Neuter
Singular Indefinite een zwarte man a black man
een zwart[ø] paard a black horse Definite de zwarte man
the black man
het zwarte paard the black horse Plural Indefinite zwarte mannen
black men
zwarte paarden black horses Definite de zwarte mannen
the black men
de zwarte paarden the black horses
Table 1: Dutch gender assignment on determiners and attributive adjectives (adjusted from Weerman et al., 2011, p. 35).
As for determiner attribution, Weerman et al. (2011) found that children with SLI follow the same acquisition path as children with TL, as they make similar errors: both make mostly overgeneralizations of the common article de where neuter het is required, whereas the reverse pattern (het when de is required) is much less attested in both TL3 and SLI children. This is due to the fact that common nouns are relatively more frequent in the input (Blom et al., 2008). However, the children with SLI in the study from Weerman et al. (2011) were clearly behind in their acquisition, as the young SLI children showed significantly fewer correct instances of het than the younger controls. The older SLI group did not significantly differ from the young children with TL, nor did they from the younger children with SLI. This suggests that children with SLI do not improve their gender assignment skills much, in contrast with what the authors found for verbal agreement (see above).
Weerman et al. (2011) found similar results for adjectives: both younger and older children with SLI made errors similar to the younger control group (i.e. overgeneralizations of the relatively more frequent schwa, but no overgeneralizations of the bare adjective), and older and younger SLI groups did not significantly differ in performance (Weerman et al., 2011). Furthermore, the correct use of the neuter article het was similar to the percentage of correctly used bare adjective (see Table 2), which suggests that children with SLI do have some knowledge of grammatical gender as both article and adjective depend on this knowledge: if a child correctly produces the definite article het with a certain neuter noun, and correctly produces the bare adjective with the same noun, we can assume that the child correctly labeled that noun as „neuter‟, and, more importantly, understands the grammatical rules for gender assignment.
3
For Dutch-speaking TL children, Blom et al. (2008) found that at age 3, children made 0/33
substitutions of common de by neuter het, compared to 37/42 substitutions of neuter het by common
Bare adjective (%) Neuter article het (%)
Control group 47,1 48,8
Young SLI children 30,1 24,4
Older SLI children 33,7 33,8
Table 2: Percentage of correctly used bare adjectives as compared to the percentage of correctly used neuter articles for each group of participants in the study from Weerman et al. (2011). Correct use of both phenomena depends on the correct attribution of neuter gender to a specific noun.
Additionally, problems with complex syntax have been attested in Dutch-speaking children with SLI: Duinmeijer, de Jong, & Scheper (2012) found that children with SLI (N = 34, mean age 7.4) produced fewer embedded clauses than TL age-peers (N = 38, mean age 7.9) during a semi-spontaneous language production task (narrative task).
2.2. Congenital Hearing Impairment
Interestingly, children with CHI have been found to experience similar problems as children with SLI in the area of morphosyntax (Tuller et al., 2011). The prevalence of permanent bilateral hearing loss in newborns is estimated to 1-1,4 per 1000 (Thompson et al., 2001), and several studies have shown that about 50% of the children with permanent hearing loss shows some degree of language problems (Boons, De Raeve, et al., 2013a; Briscoe et al., 2001; Tuller & Delage, 2014). Tuller & Delage (2014) point out that the prevalence of language problems in children with CHI is much higher than the prevalence of SLI (7%, as mentioned above), and thus does not simply reflect comorbidity. It seems clear that CHI is a risk factor for language problems, as was pointed out by many other researchers as well (Thompson et al., 2001; Tomblin et al., 2015; Verhaert, Willems, Van Kerschaver, & Desloovere, 2008).
In contrast with SLI, however, the source of these problems is better identified, and lies in the reduced and often degraded auditory input children with CHI receive due to their hearing loss (Delage & Tuller, 2007; Hammer et al., 2014). As grammar building depends on the input children receive (Huysmans et al., 2014), CHI can lead to problems with morphosyntactic features. As the literature on children with CHI that focus on Dutch are rather limited, we will give an overview of the problems attested in several other languages as well.
Before elaborating on those problems in more details, we need to make some important remarks. Children with CHI involve a various group of children that do not only differ in terms of degree and etiology of their hearing loss, but also in the type of hearing device received (hearing aid or cochlear implant 4, henceforth CI), and whether this is unilaterally or bilaterally fitted. Lastly, also the age at which the child receives hearing devices (in children with severe or profound hearing loss receiving a CI, this is usually referred to as „hearing age‟, as they had no or very little hearing experience prior to fitting) needs to be considered, as several studies found that this highly correlates with language performance (Boons, Brokx, et al., 2013; Spencer, 2004; Verhaert et al., 2008). As the impact of such factors is beyond the scope of the current investigation, we will report studies conducted on some or all of those subgroups of CHI.
One of the difficulties that Dutch-speaking children with CHI experience is verbal agreement. Hammer & Coene (2016) investigated spontaneous speech samples of 48 children with CHI between 4 and 7 years old (all CI-users) and found that they made significantly more errors to verbal agreement (including errors such as (d) and (e) above) than TL age-matches (scores were compared with available speech samples from 240 Dutch-speaking in the same age range). The authors made a distinction between children with a CI and children with conventional hearing aids, but no differences between the two groups of hearing impaired children were found. Besides verbal agreement, the researchers also analyzed the number of finite verbs produced, and found that both groups of children with CHI performed within the age-appropriate range for this measure, although the authors noted that the children with conventional hearing aids performed rather at the lower end of the range (Hammer & Coene, 2016).
Subsequently, past tense errors (for both regular and irregular verbs) are often attested in children with CHI: results from an elicitation task performed on a small number of Dutch-speaking children with CHI 5 (all CI-users) between 5 and 7 years
4 A Cochlear Implant (CI) is a device which “electrically stimulates the auditory nerve through
electrodes placed in the cochlea” (Hammer et al., 2014, p. 69). It is usually only used for hearing loss >85 dB HL, when conventional hearing aids are no longer sufficient.
5
Participants were divided in three age groups (N = 6, mean age 67 months; N = 5, mean age 77 months; N = 3, mean age = 91.6 months). They were compared with a Dutch-speaking group of age-peers in the Netherlands (N = 10, mean age 66.8 months; N = 6, mean age 80 months; N = 12, mean age = 88.4 months), as well as with a Dutch-speaking group of age-peers in Flanders, Belgium (N = 14, mean age 66.6 months; N = 18, mean age 79.2 months; N = 11, mean age = 87.3 months) (Hammer, 2010).
old, showed that they produced significantly fewer target-like regular past tense forms than peers with TL at all ages (Hammer, 2010). Whereas the performance of the children with TL increased over the three age groups (50% > 80% > near ceiling), all three age groups with CHI performed below 20% correct. For irregular past tense forms, both TL and CHI groups performed below chance level, with a significant group difference in favor of the TL group at age 5 years old. Interestingly, Hammer (2010) found that children in all groups made overgeneralizations such as (i), indicating that they all have some knowledge about the rules for regular past tense marking.
(i) hij val-te* he fall-ed*
‘he fell’ (Hammer, 2010, p. 153)
Just as was found for Dutch-speaking children with SLI, gender assignment in determiners seems also problematic for children with CHI. Boons, Van Wieringen, De Raeve, Peeraer, & Wouters (2011) analyzed the errors made during a sentence production task (Formulated Sentences, a subtest of the CELF-4-NL) from 10 children with CHI (median age 7;2), and found that they made significantly more errors to the gender assignment of definite determiners than normal hearing children from the same age (N = 10, median age 7;1), as well as compared with normal hearing children matched on hearing age (N = 10, median age 5;5). Whereas the median percentage of correctly assigned definite determiners was 100% for both control groups, the CHI group had a median of 82% correct determiners. Similarly, Boons, De Raeve, et al. (2013a) found Dutch-speaking children with a CI (N = 70, median age 8;2) to make significantly more errors to demonstrative pronouns (die, dat, dit, deze) than children with TL, individually matched for age (N = 70), which are also dependent on the gender of the noun it refers to. Unfortunately, the authors did not report the type of errors that were found (omissions or substitutions, either with common or neuter nouns), but data from Dutch-speaking adults with CHI suggests that errors are made in both directions (de>het and het>de) (Huysmans et al., 2014). This is clearly different from the error pattern that was found for individuals with SLI (Weerman et al., 2011) and for younger children with TL (Blom et al., 2008), who mainly make overgeneralizations of common de, but not show the reverse pattern.
Furthermore, also personal pronouns seem difficult to acquire for children with CHI across languages: this has been found for English (Elfenbein, Hardin-Jones, & Davis, 1994; Spencer, 2004) and French (Le Normand, Ouellet, & Cohen, 2003). More recently, this was also found for Dutch (Verbist, 2010): a longitudinal study of spontaneous speech samples of children with CHI (all with CI) (N = 9, aged between 1;5-3;6 years old) showed delay in the acquisition of personal pronouns compared to TL age-peers (N = 14). Whereas first emergence of a personal pronoun was found at the age of 2;4 (median for the whole group) for TL children, it took the children with CHI until 3;0 years old before this appeared in spontaneous speech. This difference was statistically significant. The full range of pronouns was acquired by the children with TL at 6 years old (median age for the group), and one year later (at age 7) for the CI children. The order in which the different types of pronouns appeared, did not differ between TL and CI groups, which was interpreted by the author as evidence for a normal, but delayed acquisition of pronouns. Besides this study, Boons et al. (2011) (as mentioned above) also found that Dutch-speaking children with CHI (N = 10, median age 7;2) made more pronoun errors than their peers matched on hearing age (N = 10, median age 5;5).
Although less frequently reported in the literature, prepositions have also been found to be difficult for children with CHI: a study conducted in French found that children with CHI (all CI-users) (N = 17) produced fewer prepositions than MLU-controls with normal hearing (N = 233, aged 2 to 4 years old) in spontaneous speech (Le Normand et al., 2003). For English, children with CHI (N = 21, 5-12 years old) were also reported to make preposition errors during spontaneous speech (Elfenbein et al., 1994), although not many details on the number or type of errors were given (the authors included them in a more general measure of morphosyntactic errors). Similarly for Dutch, Boons, De Raeve, et al. (2013) reported significantly more preposition errors for children with a CI (N = 70, median age 8;2) as compared to children with normal hearing (N = 70; individually matched on age) during an elicitation task (subtest Word Structures of the CELF-4-NL). However, this was not the case for all participants with a CHI, but only for a „low performing‟ subgroup with language quotients <2 SD than the norm.
Children with CHI have also been attested with difficulties with complex syntax, such as relative clauses (examples j and k) and wh-questions: In a study from Friedmann & Szterman (2006) Hebrew-speaking children with CHI (N = 20, mean
age 8;9) were found to have significantly more difficulties with the comprehension of object relative clauses than children with TL (N = 10, mean age 6;2). The children with CHI did not differ from the TL controls in comprehension of subject relatives and simple sentences with canonical SVO word order (subject, verb, object), but had significantly more difficulties than another TL control group (N = 20, mean age 6;4) with understanding topicalized sentences with a non-canonical OVS-word order (which is optional in Hebrew) (Friedmann & Szterman, 2006).
The same authors also performed an elicitation task and found that children with CHI (N = 14, mean age 9;7) produced significantly fewer correct object relative clauses than a group of age-matched children with TL (N = 28, mean age 9;0). Instead, they often produced an ungrammatical sentence, or used a resumptive pronoun (i.e. a personal pronoun used to refer to the head of the relative clause, which is grammatical in Hebrew but typically seen in younger TL children) (Friedmann & Szterman, 2006).
(j) This is the woman that draws the girl. (subject relative clause)
(k) This is the girl that the woman draws.
(object relative clause) (Friedmann & Szterman, 2006, p. 60)
Similar results were found for Italian-speaking children with a CHI, all implanted with a CI (Volpato & Vernice, 2014): as a group, the CI subjects (N = 13, mean age 9;2) produced significantly less correct object relative clauses than a control group matched by hearing age (N = 13, mean age 6;10), as well as compared to chronologically age-matched controls (N = 13, mean age 9;1). The CHI group did not significantly differ with a younger TL group, matched on performance on a morphosyntactic test (N = 13, mean age 6;7). Among the incorrect productions many avoidance strategies were used, such as passive constructions (l). Within each group, subjects had much less difficulties with subject relative clauses. The authors concluded that performance of the children with CHI was comparable to that of younger TL children (Volpato & Vernice, 2014).
(l) The child that is combed by the father.
2.3. Similarities and differences between SLI and CHI
Although it becomes clear from section 2.1 and 2.2 that children with SLI and children with CHI show similar language difficulties, some differences have been found in the proportion of those difficulties, in which, overall, children with CHI outperform peers with SLI on several linguistic tasks. From a theoretical point of view, comparing those two groups is interesting, as it may provide cues of what underlies language difficulties. On the one hand, the exact etiology of SLI is unknown. On the other hand, the only difference between children with CHI and children with TL (leaving comorbidity aside) is their reduced access to spoken language. Thus, differences and similarities between children with SLI and children with CHI provide us with information on the exact role auditory information plays in language acquisition. Hammer et al. (2014) who investigated Dutch-speaking children with CHI (all CI-users, N = 48) and children with SLI (N = 38), all 4 to 7 years old 6, found that both groups scored below age-expectations for verbal agreement errors/omissions and for finite verb production. Within each group, youngest (4-year-olds) and oldest (7-year-olds) groups did not differ in performance on verbal agreement, indicating that children with CHI and children with SLI did not improve their verbal agreement skills over time. However, for finite verb production SLI and CHI groups showed different patterns: the children with SLI in the different age groups showed similar production rates (30,4%>36,7%>35,5%>42,9% finite verbs produced), whereas production rates for the CHI group increased over the different age groups (41,3%>45,3%>50,8%>53,6% finite verbs produced). This resulted in significant group differences between CHI and SLI groups at 6 and 7 years old. These results suggest that children with CHI acquire verb finiteness more slowly than peers with TL (as they scored lower than TL group but improved over time), whereas children with SLI seem to stagnate in their performance on this feature, at least in this age range.
Another comparative study on finite verb production, performed in English, found no significant differences between children with CHI (N = 19, mean age 8.6) and age-matched TL children (N = 20, mean age 8.45) on verbal agreement of third
6
In order to study the development of finite verb production, CHI and SLI groups were subdivided in 4 different age ranges. For CHI: 15 4-year-olds (M = 50,9 months); 14 5-year-olds (M = 63,2 months); 10 6-year-olds (M = 73,5 months); and 9 7-year-olds (M = 85,8 months). For SLI: 5 4-year-olds (M = 54,3 months); 9 5-year-olds (M = 65,2 months); 15 6-year-olds (M = 76,5 months); and 9 7-year-olds (M = 87,6 months) (Hammer et al., 2014).
person singular and past tense during an elicitation task, whereas an SLI group (N = 14, mean age 9) performed significantly worse than the age-matched children with TL on both features (Norbury et al., 2001). The authors also compared individual performance within the groups for the third person singular elicitation task and found that none of the TL subjects scored <66% correct, whereas 7/14 SLI and 3/19 CHI subjects did. On the past tense task, Norbury et al. (2001) found an effect of regularity and frequency of the verbs elicited, but this was equal in all groups (i.e. TL, CHI, and SLI groups had slightly better scores for regular past tense forms as compared to irregular past tense forms, and for high frequent verbs over low frequent verbs).
Subsequently, comprehension of morphosyntax was tested by several authors by means of the Test for Reception of Grammar (TROG; Bishop, 1983). In this test, participants need to select one out of four pictures based on a given word or sentence with increasing grammatical complexity. English-speaking children with SLI scored significantly worse than children with TL, whereas children with CHI did not differ from the TL group (Norbury et al., 2001; same participants as mentioned before). Similar findings were reported for Swedish-speaking children who were slightly older (CHI: N = 18, median age 10;10; SLI: N = 27, median age 10;0; TL: N = 38, mean age 10;6), using the same test. Delage & Tuller (2007) assessed older participants (11-15 years old) using a grammatical judgment task (receptive) and a clitic pronoun (example a) elicitation task (productive). The authors found that both SLI and CHI groups scored significantly worse than a younger TL group (N = 12, mean age 11;4) on both the receptive and productive task, but in addition the SLI group (N = 12, mean age 14;8) was significantly worse than the CHI group (N = 19, mean age 13;8) on both measures as well.
In addition, both SLI and CHI subjects experience difficulties with complex syntax such as object relative clauses, as we discussed before. Tuller & Delage (2014) found French-speaking children with SLI (N = 8, mean age 9;6) and children with CHI (N = 14, mean age 9;8) to perform similarly on the production of object relative clauses, and both groups scored significantly more poorly than 6-year-old TL controls (N = 12).
Furthermore, the study from Hammer et al. (2014) (mentioned above) compared general language performance (by means of MLU) of their SLI and CHI
groups and found that 75% of the children with SLI scored below age-expectations, in contrast with only 35,4% of the children with CHI.
Next, in an elicitation task French-speaking adolescents with CHI (N = 15, mean age 12;6) scored significantly better on the production of relative clauses as compared to the younger SLI group (no direct comparison with the younger CHI subjects was reported). Interestingly, this older group of participants used more avoiding strategies such as subject relative clauses and passive constructions (examples j and l, respectively), whereas the younger CHI group and the SLI group rather used simple constructions without relative clauses at all (Tuller & Delage, 2014). According to the authors, finding better performance and different avoiding strategies for the older subjects with CHI suggests that they still furthermore mature their language skills. On a par, recall that Hammer et al. (2014) found differences in performance for verbal agreement between young children with CHI and age-matches with TL, whereas no such difference was found in the study of Norbury et al. (2001), who studied older children with CHI. This as well suggests that children with CHI mature their language skills at a slower pace than TL peers. For SLI, recall the study from Weerman et al. (2011). These authors found significant differences on a verbal agreement production task between a younger (mean age 7;3) and an older (mean age 12;7) SLI group, also suggesting a form of maturation.
As becomes apparent from the above, different outcomes and developmental rates are found depending on the language area that is investigated, and the age group studied: in some studies children with CHI and children with SLI are found to perform similar, and both different from TL peers (Hammer et al., 2014), whereas in others only children with SLI perform worse than children with TL (Norbury et al., 2001). Some studies report improvement over time (Tuller & Delage, 2014), whereas others find some sort of stagnation (Hammer et al., 2014). It is possible that age effects and different developmental trajectories between and within groups blur those outcomes, as was also pointed out by several researchers (Norbury et al., 2001; Tuller & Delage, 2014). Overall, it seems that the difficulties of children with CHI are often less severe and less persistent than the problems of children with SLI, but that the areas of their problems are the same.
2.4. Language problems beyond childhood
Although core language problems are extensively studied in children, far fewer studies have been conducted on language skills after childhood. However, for both SLI and CHI separately, it has been previously shown that language problems often persist into adolescence and even later adulthood. Moreover, with the exception of a few (Delage & Tuller, 2007; Hansson et al., 2004), comparative studies between SLI and CHI after the age of 10 are nearly absent. To the best of our knowledge, no such comparative study is done with Dutch-speaking adolescents. As we pointed out before, it is possible that age effects, and a slower pace of development between and within the different groups may obscure the findings in comparative studies. Therefore, studying performance of both populations after childhood, when language skills are assumed to have fully developed, reduces the possible influence of age factors and provides an interesting perspective on language impairments and the role of auditory input in later language outcomes.
For SLI, two independent longitudinal studies on English report that about 70% of the individuals that were diagnosed with SLI during childhood, had persisting language problems into adolescence (Johnson et al., 1999; Stothard, Snowling, Bishop, Chipchase, & Kaplan, 1998). More recently, also for Dutch-speaking adolescents (N = 31, mean age 14;5) who were diagnosed with SLI as a child, persisting problems were attested (Duinmeijer, 2016). The researcher found that the adolescents with SLI as a group still scored significantly lower than TL peers (N = 30, mean age 14;5) on those morphosyntactic aspects that were identified as core problems in SLI at a younger age: verbal agreement, grammatical gender (articles and adjectives attribution) and relative clauses (Duinmeijer, 2016). Delage & Tuller (2007) found that both adolescents with SLI and those with CHI (11-15 years old, as mentioned in section 2.3) performed significantly more poorly than TL age peers on several grammar tasks. Even in adulthood difficulties persist: Huysmans et al. (2014) showed that Dutch-speaking adults with CHI (N = 10, mean age 25) scored significantly different from adults with TL on several morphosyntactic features (determiners, bound morphemes and adverbs). Poll, Betz, & Miller (2010) reported similar results for adults with SLI (N = 13, mean age 21).
What is important when studying adolescents or adults, is that although persisting, language problems are often much more discrete (Wetherell, Botting, &
Conti-Ramsden, 2007), and appear mostly in complex or time-pressured situations (Huysmans et al., 2014). This makes the use of standardized and norm-referenced language tests less suited for the assessment of adolescents, as they usually only test isolated skills and thus cannot always accurately identify small language problems in older individuals (Wetherell et al., 2007; Wolgemuth, Kamhi, & Lee, 1998). Moreover, standardized assessment tools usually do not have age-norms available over the age of 12. An exception is the Clinical Evaluation of Language Fundamentals, fourth edition (CELF-4) (Semel, Wiig, Secord, & Kort, 2008), a test battery available in many different languages, which provides norms up to 18 years of age. However, most subtests involve isolated skills or focus on lexical knowledge, which does not appropriately match the typical language problems of adolescents.
An interesting alternative to testing language skills beyond childhood is a narrative task, which tests the ability to tell a coherent and comprehensive story (Trabasso & Rodkin, 1994). Narrative tasks are more complex as they require the storyteller to combine a range of linguistic skills such as morpho-syntactic, lexical, and pragmatic skills and does not focus on only one of these skills in isolation (Duinmeijer et al., 2012). In addition, in order to tell a comprehensible story, a multitude of cognitive skills are needed, such as working memory, organization and planning (Botting, 2002; Duinmeijer et al., 2012).
Narrative skills have been proven successful in identifying small linguistic errors, both in younger children (Miniscalco, Hagberg, Kadesjö, Westerlund, & Gillberg, 2007) as well as in adolescents (Wetherell et al., 2007), and are found to be good predictors of later language outcomes (Botting, Faragher, Simkin, Knox, & Conti-ramsden, 2001; Stothard et al., 1998).
Furthermore, narratives provide useful information about the linguistic capacities in daily life. In both French (Tuller, Henry, Sizaret, & Barthez, 2012) and English (Marinellie, 2004), adolescents with SLI seem to be able to produce complex syntactic units (e.g. relative clauses and wh-clauses), as several such structures were found in the samples. However, they produced significantly fewer of those than age-matches with TL. Tuller et al. (2012) argue that this can be regarded as an avoiding strategy for structures they find difficult. As mentioned before (see section 2.2), similar avoiding strategies were also attested in younger children with SLI (de Jong, 1999). For individuals with CHI the picture is less clear, as complex syntax in spontaneous speech is less well studied for this group, especially after childhood.
The previously mentioned study of Huysmans et al. (2014) reported no group differences between CHI and normal hearing adults on measures of syntactic complexity, such as MLU.
2.5. Theories explaining language problems in both SLI and CHI
As becomes apparent from what we described above, the main problems of both individuals with SLI as well as those with CHI lie in the area of morphosyntax. Nevertheless, up to date, there is no consensus on what causes those particular language problems in both groups.
For SLI, many different explanations have been proposed in the literature, and although explanations vary widely in terms of the exact locus of the deficit, we can roughly divide them into two main types of accounts: (1) representational accounts, arguing that SLI is caused by a specific deficit in the representation of linguistic knowledge, and (2) processing accounts, arguing that the underlying cause of SLI stems from problems with the processing of information. The latter group of accounts differs in terms of whether those processing limitations are specific to linguistic information only, or if limitations are more general, including the processing of non-linguistic information as well.
One interesting representational account is the Computational Grammatical Complexity Hypothesis (CGCH) (van der Lely, 2005). The CGCH assumes that children with SLI have difficulties computing “the hierarchical structural organization of sounds into words” (van der Lely, 2005, p. 54) and rather store inflected forms as separate lexical items. The CGCH can account for finding inflected forms for one lexical item and not for another, as storage of lexical items is highly dependent on frequency effects, i.e. inflected forms that appear more often in the input, are likely to be used by children with SLI as well. However, this theory cannot explain why inflected and uninflected forms inconsistently appear within the same lexical item, as was attested by data from Bishop (1994), and also fails to explain why children with SLI are sometimes found to produce overgeneralizations such as „he catched‟, (and recall example (i) in the CHI data from Hammer, 2010) and even produce some correctly inflected forms of nonsense verbs (Leonard, Eyer, Bedore, & Grela, 1997). As such utterances do not appear in the input, they cannot be stored as separate
lexical items and suggest that at least some knowledge of that particular grammatical rule is present in children with SLI (Bishop, 1994).
Also, crucial for the current study, representational accounts cannot explain the problems individuals with CHI experience, as we assume they dispose of a normal working language device. Although we cannot exclude the possibility that SLI and CHI are really two completely different impairments leading to similar problems with language, we take as our starting point that performance between the two groups is so similar because both have, although with different etiologies, difficulties with the processing of linguistic information.
Processing accounts thus have the potential to explain problems in both individuals with SLI and those with CHI. This makes it particularly interesting to compare both groups, as the source of problems is much better identified in individuals with CHI (i.e. hearing loss), and thus may provide cues of what underlies problems of individuals with SLI as well.
Within the stream of processing accounts, theories vary widely in terms of the exact locus of the deficit, for CHI, but especially for SLI. One of those theories is the Surface Account (Leonard et al., 2003, 1997), proposed for SLI. Leonard and his colleagues noted that many of the problems children with SLI experience involve morphemes that are non-salient and unstressed in language (Leonard et al., 2003), such as third person singular –s (English) or –t (Dutch) or past tense morpheme –ed (English). The Surface Account proposes that grammatical errors are caused by the low perceptual salience of the morphemes themselves. Leonard et al. (2003) argue that children with SLI have limited processing capacity, which is taxed when these non-salient speech segments also need to be processed for their grammatical function. The Surface Account can account for cross-linguistic variation in core features of SLI: whereas verbal agreement is difficult for Dutch and English acquiring children (-t and –s, respectively), this does not seem to be the case for children acquiring Romance languages such as Spanish or Italian, in which agreement morphemes often involve a syllable, rather than a single morpheme, and are thus argued to be more acoustically salient (see Table 3) (Leonard, 2009).
Singular Plural
1st person Canto Cantamos
2nd person Cantas Cantan
3rd person Canta Cantan
Table 3: Verbal agreement paradigm for the Spanish verb „cantar‟ (to sing). Adjusted from Anderson
(2001 p. 7).
For children with CHI, such theory seems even more straightforward: hearing loss makes it hard to perceive those phonologically non-salient features (Elfenbein et al., 1994; Svirsky, Stallings, Lento, Ying, & Leonard, 2002). Similar to the Surface Account for SLI (Leonard et al., 2003, 1997), Svirsky et al. (2002) formulated this idea for children with CHI as the „Perceptual Prominence Hypothesis‟: the authors investigated English acquiring children with CHI, all CI-users (N = 9, mean age 6;10) and found that their order of acquisition of three grammatical phenomena (copula verbs is and are > plural marker -s > past tense marker -t) was in line with the perceptual prominence of those features: is and are can be more easily distinguished than the single morpheme -s; which in turn is more prominent than -t as the latter is three to four times shorter in duration (Svirsky et al., 2002). This acquisition pattern was in contrast with that found in younger children with TL (N = 9, mean age 4;10) and children with SLI (N = 9, mean age 4;9), who acquired those phenomena in a different order: plural > copula > past tense (Svirsky et al., 2002). Although finding a different pattern in SLI and CHI individuals is problematic for arguing they both experience the same problems with processing linguistic information, there is another problem with the Surface Account for SLI and the Perceptual Prominence Hypothesis for CHI: French acquiring children, both those with SLI and those with CHI, have been found to make more errors with accusative clitics le and la (recall example a), than with definite articles le and la (Delage & Tuller, 2007; Tuller et al., 2011), which are perceptually equal. In a similar vein, English acquiring children with SLI have been found to have difficulties with verbal agreement of third person singular –s, whereas plural noun inflection –s seems less problematic (Bishop, 1994), and also English-acquiring children with CHI have been found to accurately produce the plural morpheme –s much more often than third person singular –s (McGuckian & Henry, 2007).
Although the findings from Svirsky et al. (2002) (i.e. different acquisition pattern, in line with perceptual prominence), and Huysmans et al. (2014) (i.e. alternations between determiners de and het, deviant from the normal acquisition
pattern) suggest that perceptual non-saliency plays an important role in individuals with CHI, it cannot fully explain the difficulties that are found in CHI and SLI subjects. That is also what McGuckian & Henry (2007) conclude from their findings.
Another theory that potentially explains language difficulties in both SLI and CHI is the Vulnerable Markers Hypothesis (VMH) (Bishop, 1994). The VMH has been originally proposed for children with SLI but more recently, also outcomes of adults with CHI were in line with this theory (Huysmans et al., 2014). The VMH makes the assumption that SLI children have limited processing capacity, which causes problems with the acquisition of grammatical rules, especially in vulnerable areas of language (i.e. those areas which typically developing children take long to acquire). As was already pointed out by Leonard et al. (2003), the processing of the grammatical function of certain grammatical features plays an important role. Different from the Surface Account, however, the VMH assumes that morphosyntactic problems go beyond the perceptual saliency of those features, and rather concern features that are non-transparent and hard to derive from the input, regardless of their perceptual saliency: “in assigning grammatical roles such as subject, various cues, such as word order, agreement, case marking, and grammatical prosody, will play a part, with each having an influence in proportion to how reliable and valid a cue is (Bishop, 1994, p. 533). What this quote implies, is that vulnerable areas require more cues from the input in order to build up the grammatical rule.
Viewing language problems in terms of vulnerability, the VMH can also account for cross-linguistic differences: taking gender assignment as an example, the paradigm in a language such as Greek is very consistent and grammatical cues are provided on the article (both definite and indefinite), the noun, and the adjective (m). In contrast, the Dutch gender paradigm (as discussed in section 2.1) is very opaque and no morpho-phonological cues as is the case in Greek (e.g. o on the article, noun, and adjective in the case of masculine) are provided (Unsworth et al., 2014). The vulnerability of gender assignment in Dutch but not in Greek is reflected in the age of acquisition by children with TL: whereas Greek acquiring children master the paradigm already around the age of 3;6 (Tsimpli, 2003) in Unsworth et al., 2014), it takes Dutch acquiring children until the age of 7 in order to acquire it (Blom et al., 2008).
(m) o oreos kipos
the.MASC beautiful.MASC garden.MASC
‘the beautiful garden’
(Unsworth et al., 2014, p. 773)
Although the gender assignment paradigm can also be explained in terms of non-saliency (e.g. Greek masculine marker os is more acoustically salient than Dutch schwa on the adjective), vulnerability is much more suited to explain differences between English plural marker –s and third person singular –s (plural marking in English is less opaque than the paradigm for verbal agreement), and between French articles le and la on the one hand, and accusative clitics le and la on the other.
As the rules concerning vulnerable grammatical features are opaque and difficult, it is generally assumed that more input is needed in order to derive the necessary rules from the input, hence the later age of acquisition in TL. The VMH assumes that children with SLI have extra difficulties with deriving the rules from the input, as limited processing capacity makes it hard for them to handle multiple operations simultaneously, such as processing the necessary information from the input while at the same time storing it (Bishop, 1994). It is thus assumed they have a reduced „uptake‟ of the necessary input in order to derive grammatical rules of such vulnerable features of grammar.
Although children with CHI are not assumed to have limited processing capacities per se, their hearing loss makes it harder to process (some) linguistic information, especially when it is non-salient in the input. We thus consider them to have a reduced uptake of the necessary information from perceptually non-salient input. The VMH can explain why CHI children have more difficulties with accusative clitics than with articles in French: the uptake of le and la, regardless of function, is reduced because of hearing loss. Nevertheless, as the rules for accusative clitics are more computationally complex (Tuller & Delage, 2014), more input is needed to settle those rules, as compared with article assignment. This results in more problems with clitics.
Finally, the VMH can also account for the variability in performance that is often attested in SLI and CHI subjects: it is observed that they sometimes alternate between correct and incorrect use of the same item, depending on the context in
which it appears (Bishop, 1994). Limited processing capacity would not only cause difficulties with acquisition of grammatical features, it also implies that whenever the demand on this capacity increases, e.g. because the linguistic context is more complex, more errors will be made, as the grammatical rules are less well settled.
As we discussed in section 2.4, Huysmans et al. (2014) found that Dutch CHI adults made significantly more errors than adults with TL on determiners, bound morphemes, and the adverb er, but this was only the case during a spoken task and not during a written task. This was explained by the authors in terms of complexity and time pressure: good performance on the written task shows that grammatical knowledge is (at least partially) present, but as a spoken task is more cognitive demanding than a written task, higher error rates are found during the former.
In sum, the VMH seems able to explain most problems of both CHI and SLI subjects, and thus similar outcomes in language production are expected. However, as we already noted, for CHI subjects we still need to make a distinction between perceptual salient and non-salient features. We assume that individuals with CHI have normal processing capacities, but have difficulties processing non-salient features of language. This would lead to reduced uptake only for perceptual non-salient features in the input, regardless of grammatical function, as these are less available to them. It is only when those grammatical features are also vulnerable features (i.e. require more input), that problems are expected. Vulnerable features that are acoustically more salient are assumed not to be problematic for CHI subjects, as the uptake for those features would be similar to that of normal hearing subjects.
In contrast, SLI subjects are not assumed to have problems with the non-saliency of features, but due to limited processing capacities, they will show difficulties in all vulnerable areas of language, regardless of perceptual saliency. This leads to partially different predictions for each group, which we will present in the next section.
3. Aims and Research Questions
Both adolescents with CHI and those with SLI often show persistent problems, mostly in the domain of morphosyntax. For both groups of impairments, the VMH (Bishop, 1994) has been proposed to explain most of the persisting problems
experienced. However, the VMH has not yet been directly tested in a comparative study between CHI and SLI adolescents.
Furthermore, as we are dealing with adolescents, we need to take into account that their language problems are often more subtle than in younger children, and mostly appear in complex or time-pressured situations. As most standardized tests measure isolated skills, we lack an adequate tool to identify language problems after childhood.
The aim of this investigation is therefore twofold: (1) the theoretical aim is to identify language problems in Dutch speaking CHI and SLI adolescents and test the validity of the VMH by directly comparing those two groups; (2) the methodological aim is to investigate the potential of a narrative task in identifying linguistic problems in adolescents. In order to do so, we compared scores on several standardized, age-appropriate language measures to performance on a narrative task.
As for the theoretical aim, we take as our starting point the following hypotheses: (1) Adolescents with CHI have difficulty perceiving non-salient features of
language, leading to reduced uptake of those features. However, they are not limited in their processing capacities.
(2) Adolescents with SLI have limited processing capacities. However, they have no difficulty perceiving non-salient features of language.
With regard to the methodological aim of this study we formulated the following hypothesis:
(3) Standardized language tests are not sensitive enough to identify language problems in adolescents, as they test isolated skills in a less complex context. Narrative tasks reveal more subtle language impairments, as its processing cost is higher.
Those hypotheses lead to the following predictions:
a) For adolescents with CHI problems are predicted for non-salient features in language that are also vulnerable (i.e. require more input and uptake). We thus expect difficulties with verbal agreement, past tense, adverbs and determiners. We do not expect difficulties with word order, pronouns,
prepositions and conjunctions as adolescents with CHI can perceive them well.
b) For adolescents with SLI, we predict problems in most of the vulnerable areas of language, regardless of perceptual saliency. On top of problems with verbal agreement, past tense, adverbs and determiners, we also expect problems with word order, pronouns, prepositions and conjunctions.
c) With regards to hypothesis (3), we predict similar performance on the standardized tests for all three groups (SLI, CHI and TL) as they test isolated skills. However, we predict differences in performance between TL and impaired groups on the narrative task, especially in terms of morphosyntax.
No previous studies compared lexical performance and disfluencies in speech samples of individuals with CHI and SLI. As the narrative task provides us with additional information about performance in other areas than morphosyntax, we also explored our data for lexical performance, as well as for disfluencies during the task. As those areas are rather unexplored in adolescents‟ language, no clear predictions were made.
4. Method
4.1. Participants
Three groups of Dutch monolingual adolescents aged 15;0 (years;months) up to 17;11 participated in this study: (1) 10 adolescents with CHI (M = 15,95; SD = 0,58); (2) 11 adolescents with SLI (M = 16,26; SD = 0,82); and (3) 5 TL adolescents (M = 16,12; SD = 0,92). More details about the education and gender of the participants in each group can be found in Table 4. In addition, data of 28 adolescents with TL between 14;1 and 16;9 years old from another study (Duinmeijer, 2016) were included. Those participants were tested with the same standardized language test as used in this study, but no data from a narrative task was available. Therefore, an extra TL group, including the 5 participants recruited for this study and the 28 additional participants from the study from Duinmeijer (2016)
was added to Table 4. The TL+CELF group consisted of 33 participants (M = 14,87; SD = 0,76).
CHI SLI TL TL + CELF
Education type 6 havo mavo vmbo praktijk 3 3 3 1 2 1 7 1 2 2 1 0 3 12 14 4 Education level 2nd year 3th year 4th year unknown 2 6 2 0 2 7 2 0 1 0 4 0 11 9 5 8 Gender female male 6 4 1 10 3 2 14 19
Table 4: Metadata for each group of participants.
Participants from the impaired groups were recruited via a special needs high school for children with communication problems (Dutch „cluster 2‟) in the region of Amsterdam. TL peers were recruited via a mainstream Dutch high school in the region of Amsterdam, and via personal contacts of the researcher. Participants in all groups were selected from the same four types of education7. No form of incentive was received by any of the participants, and consent of their parents was obtained prior to testing. The project was approved by the Ethics Committee of the Faculty of Humanities from the University of Amsterdam.
Participants in the CHI group varied largely in degree of hearing loss and hearing intervention. An overview of the data is given in Table 5.
ID PTA (dB HL)
Left Right Hearing aid Age (y;m) Education (level)
CHI1 83,75 108,75 CI right 15;6 SN havo (3)
CHI2 30,00 38,75 No HA 16;7 SN vmbo (3)
CHI3 57,50 58,75 HA both sides 16;11 SN vmbo (3)
CHI4 120 113 CI left 16;3 SN mavo (3)
CHI5 112,50 101,25 CI left 15;4 SN praktijk (2)
CHI6 36,25 38,75 No HA 16;3 SN mavo (4)
CHI7 78,75 97,50 HA both sides 15;5 SN havo (3) CHI8 91,25 87,50 HA both sides 15;4 SN vmbo (2)
CHI9 - - HA both sides 15;7 MS mavo (3)
CHI10 57,50 57,50 HA both sides 16;4 MS havo (4)
7
Based on personal abilities, several different types of higher education exist in The Netherlands, ranging from vocational studies („praktijk‟ and „vmbo‟), and vocational studies combined with more theoretical subjects („mavo‟), up to more general („havo‟) and pre-university studies („vwo‟, not
Table 5: Audiometric details of the CHI group. PTA = Pure Tone Average, based on hearing thresholds on the frequencies 500 - 4000 Hz, with best ear unaided indicated in bolt (for participant 209, no audiometric data was available); CI = Cochlear Implant; HA = Hearing Aid; SN = special needs education; MS = mainstream education.
4.2. Materials
Participants were assessed with five different tests, which are described in the following paragraphs. Four of the tests were part of the CELF-4-NL, Clinical Evaluation of Language Fundamentals-Fourth Edition, Dutch version (Semel et al., 2008), a standardized language test, often used in the clinical practice. For each of these subtests norm scores are available. Together, those four subtests also form the
Core Language Score, for which separate norm scores are available as well.
Word Definitions (Definities van Woorden, CELF-4-NL; max. score 50): the
participant is given a word, which is used in a sentence by the examiner, and is then asked to tell what this word means. The test consists of 25 items, of which 24 are nouns, one item is a verb. For each item a maximum of two points can be obtained.
Formulated Sentences (Zinnen Formuleren, CELF-4-NL; max. score 40): the
participant is asked to formulate a sentence with a given word, which matches a given picture. The test consists of 20 items from different word classes, such as nouns, adjectives, adverbs, verbs and conjunctions. The participant is explicitly asked to use the given form of a word (e.g. when a plural form is given, the use of a singular form is scored incorrectly, as is the case for a present tense form of a verb when the past tense is given). For each sentence, a maximum of two points can be acquired. One point is scored e.g. when the given word is correctly used, but the sentence in which it is used contains a morphological error, or when the sentence is not related to the given picture.
Sentence Repetition (Zinnen Herhalen, CELF-4-NL; max. score 93): the
participant repeats a sentence that is given by the examiner. The test consists of 31 items, varying in length and syntactic complexity. A maximum of three points can be obtained per test item.
Word Classes – Receptive and Expressive (Woordcategorieën –
Receptief/Expressief, CELF-4-NL; max. score 40): the participant is presented with four words, of which two belong together. He/she is asked which of the two words belong together (receptive), and why he/she thinks that is the case (expressive). The test consists of 20 items; for each part of the test (receptive and expressive), one
point can be obtained. When the participant received zero points on the receptive part, he/she automatically received zero points for the expressive part as well.
In addition, all participants were administered a narrative generation task, named Frog, where are you? (Mayer, 1969): In this task, the participant was asked to tell the story matching a wordless picture book, consisting of 24 pictures. Before starting, the examiner introduced the protagonists (a boy, a dog, and a frog), while pointing to the characters on the first page of the book. After, the participant was given some time to look through the pictures and familiarize with the story. Then, the participant was asked to tell the story with the support of the pictures. The participant was explicitly asked to tell the story as if the examiner cannot see the pictures, e.g. to avoid the use of pointing to the pictures instead of telling what is seen. During the narrative task, an audio-recording was made to make post-hoc transcription possible. Transcription was done in CLAN according to CHAT conventions (MacWhinney, 2000).
Five different analyses were computed for the narrative task: (1) morpho-syntactic complexity, by means of MLU; (2) lexical diversity, by means of type/token ratio (TTR); and error rates in the following main categories: (3) morphosyntactic errors; (4) lexical errors; and (5) fluency errors. As the narratives of the participants differed widely in length, all errors were normalized into a ratio of the observed errors relative to the total number of contexts in which such error could occur. Table 6 shows a detailed overview of the different errors per category that we coded for, as well as how the ratios were calculated.
Originally, we also included a category with pragmatic errors. However, as the task design allowed participants to point to the pictures, no reliable interpretation could be given to the use or avoidance of referents. Therefore, this category was excluded.
