City Research Online a Prosodically-controlled Word and Nonword Repetition Task for 2-4 Year Olds: Evidence from Typically Developing Children Nonword Repetition and Language Measures Nonword Repetition as a Potential Clinical Marker

An association has been found between nonword repetition and language skills in school-aged children with both typical and atypical language development (Dollaghan & Campbell, 1998; Ellis Weismer et al., 2000; Gathercole & Baddeley, 1990; Montgomery, 2002). This raises the possibility that younger children’s repetition performance may be predictive of later language deficits. In order to investigate this possibility, it is important to establish that elicited repetition with very young children is both feasible and informative. To this end, a repetition task was designed and carried out with 66 children aged 2-4. The task consisted of 18 words and 18 matched nonwords that were systematically manipulated for length and prosodic structure. In addition, an assessment of receptive vocabulary was administered. The repetition task elicited high levels of response. Total scores as well as word and nonword scores were sensitive to age. Lexical status and item length affected performance regardless of age: words were repeated more accurately than nonwords, and one-syllable items were repeated more accurately than two-syllable items, which were in turn repeated more accurately than three-syllable items. The effect of prosodic structure was also significant. Whole syllable errors were almost exclusive to unstressed syllables, with those preceding stress being most vulnerable. Performance on the repetition task was significantly correlated with performance on the receptive vocabulary test. Since this repetition task was effective in eliciting responses from most of the 2 to 4-year-old participants, tapped developmental change in their repetition skills, and revealed patterns in their performance, it has the potential to identify deficits in very early repetition skills that may be indicative of wider language difficulties. Word and nonword repetition 3 A prosodically-controlled word and nonword repetition task for 2-4 year olds: Evidence from typically developing children The repetition of nonwords has gained remarkable status as a measure of children’s language over the last decade. It started its research life as a tool developed by Gathercole and Baddeley (1989) to evaluate their hypothesis that phonological short-term memory (STM) plays an important role in vocabulary acquisition. Nonword repetition was assumed to provide a measure of phonological STM. This assumption has not gone unquestioned; it has been a catalyst for debate and research on what nonword repetition actually measures. However, the current high profile of nonword repetition testing is due more to the relationships that have been found between nonword repetition and a range of other language abilities and disabilities. Nonword Repetition and Language Measures A number of studies have revealed correlations between nonword repetition and receptive and expressive vocabulary size, as well as indices of speech output including repertoire of vocabulary, utterance length and grammatical complexity, in groups of typically developing children ranging from 3-5 years of age (Adams & Gathercole, 1995, 2000; Gathercole & Adams, 1993; Gathercole & Baddeley, 1989). Gathercole and Baddeley’s (1990) investigation of nonword repetition in a small group of children with language disorders extended the evidence, revealing significant differences between the performance of these children and the performance of verbally and non-verbally matched controls. These findings have been corroborated by a growing number of studies. All have reported nonword repetition deficits and correlations between nonword repetition and measures of language in children with language impairment (e.g., Botting & ContiRamsden, 2001; Dollaghan & Campbell, 1998; Edwards & Lahey, 1998; Ellis Weismer et Word and nonword repetition 4 al., 2000; Montgomery, 1995, 2002; Norbury, Bishop & Briscoe, 2002; Stothard, Snowling, Bishop, Chipchase & Kaplan, 1998), and even in children whose language difficulties appear to have resolved (Bishop, North & Donlan, 1996; Conti-Ramsden, Botting & Faragher, 2001). Nonword Repetition as a Potential Clinical Marker Bishop et al. (1996) was the first study to establish that a deficit in nonword repetition is significantly heritable, leading to the suggestion that nonword repetition “provides a marker of the phenotype of developmental language impairment” (p. 156). Other indices of language have also been proposed as clinical markers, for example, children’s production of finite verb forms. Bedore and Leonard (1998) found that a finite verb morpheme composite was highly accurate in discriminating children with SLI aged 3;7-5;9 (year;month) from age-matched controls, in keeping with the findings of Rice, Wexler and Cleave (1995). When Conti-Ramsden et al. (2001) compared four proposed clinical markers, they found that the overall accuracy of nonword repetition in identifying 11-year-olds with SLI was high – marginally poorer than sentence recall, but considerably better than past tense and third person singular marking. It has been argued that a viable clinical marker should be largely independent of IQ (Bishop et al., 1996). Nonword repetition meets this criterion: a number of studies have found it to be largely independent of performance IQ in children with both typical and atypical language development (Conti-Ramsden et al., 2001; Ellis Weismer et al., 2000). A further advantage for nonword repetition is that it appears to be a culturally unbiased measure (Dollaghan & Campbell, 1998), unlike other measures of language that are known to be associated with socioeconomic factors (Hart & Risley, 1995). Burt, Holm and Dodd (1999) found no difference in nonword repetition performance between children from upper middle class and working class areas in the UK. Similarly, Ellis Weismer et al. (2000) found Word and nonword repetition 5 no difference between ethnic groups represented in their USA study, and point out “the potential usefulness of processing-based measures generally, and nonword repetition tasks specifically, in providing culturally nonbiased assessments of linguistic abilities” (p. 874). Less attention has been given to gender, but Burt et al.’s (1999) investigation found no gender differences. This contrasts reported gender effects on measures of early language development, particularly vocabulary (Fenson et al., 2000; Rescorla & Alley, 2001). In summary, nonword repetition has strong potential as a processing-based clinical marker, uncontaminated by factors known to affect performance on most knowledge-based measures of language. Nonword Repetition as a Potential Predictor The findings regarding the relationship between deficits in nonword repetition and language deficits have significant implications for development of children younger than those that have been studied. The bulk of evidence to date comes from typically developing children whose language development is virtually complete. The majority of studies with English-speaking children have used either the Children’s Test of Nonword Repetition (Gathercole & Baddeley, 1990), which is standardized for children aged 4-8, or a test designed by Dollaghan and Campbell (1998) that was originally carried out with languageimpaired children aged 6;0 to 9;9 and their age-matched controls. There is very limited evidence from children under the age of 4. If nonword repetition is a marker for SLI in older children, we might reasonably expect it to be a predictor of language impairment in younger children. This could be particularly important in the assessment of language in the very early years, as it is currently difficult to diagnose language problems at an early age. Follow-up studies of very young children who show delay or impairment in language development illustrate the difficulties of early diagnosis. Bishop and Edmundson (1987) followed up children who presented with impairment on a range of standardized language Word and nonword repetition 6 assessments, in order to investigate the predictiveness of these assessments. They found that performance on these tests at 4 years old was predictive of outcome at 51⁄2, with a story retell task proving the most reliable discriminator. However, as they point out, their findings applied to a narrow range of ages: “one could not generalise [these findings] to 3-year-olds where inability to tell a story might have a different significance” (p.170). Studies following up younger children identified as late talkers at 24-31 months have found some, but not all, early measures to be predictive of later language skills (Rescorla, Dahlsgaard, & Roberts, 2000; Rescorla, 2002). For example, Rescorla’s Language Development Survey at 2 to 3 years was significantly associated with language and reading scores at 6 to 9 years (Rescorla, 2002). However, while measures such as this afford good group predictive validity, they are less informative about individual outcomes and ‘clinical caseness.’ It may be that assessments tapping emergent skills key to language development will provide better individual and clinical prediction than assessments of language itself. In the search for key emergent skills, repetition is a promising avenue to explore. It may provide a better indicator of which children with delayed language will have persistent and specific problems with language. In order to investigate the predictive value of repetition in younger children, we need a test that is applicable to younger children. The goal of the present study was to develop a repetition task for typically developing 2 to 4-year-old children, which shows differences within this age range. In designing such a test, it was important to take account of factors that may affect nonword repetition, and whose effects may be informative about children’s phonological processing. Key Factors in Test Design One factor known to be influential is item length. This has been systematically varied in existing tests of nonword repetition, and studies using these tests have consistently revealed Word and nonword repetition 7 that length affects performance (e.g., Dollaghan & Campbell, 1998; Gathercole, Willis, Baddeley, & Emslie, 1994). Another factor known to affect nonword repetition is wordlikeness: nonwords that are more word-like are easier for children to repeat than nonwords that are less word-like (Gathercole et al., 1994; Gathercole & Martin, 1996). In the case of very young children whose exposure to words is relatively recent, it may be that children do not have sufficient familiarity with real words for these to influence performance on nonwords. In order to investigate effects of familiarity on repetition at this early stage, we compared actual words and phonologically matched nonwords, rather than comparing more and less word-like nonwords. In contrast to length and wordlikeness, the prosodic structure of items has rarely been investigated. Yet it is well established that prosodic factors are particularly influential in early development. Echols (1996) highlights the effects of prosodic structure on syllable omission in early word production. Children acquiring English show a preference for trochaic foot structure, in which a stressed (strong) syllable is followed by an unstressed (weak) syllable (SW). They are far more likely to omit weak syllables when these precede a strong syllable (WS) and therefore fall outside this foot structure. For example, children will almost certainly preserve the unstressed syllable in the trochaic form tiger, but they may well omit the unstressed syllable preceding stress in guitar. Kehoe (1997) provides further evidence of children’s sensitivity to the prevalent stress patterns of English. In an investigation of 22 to 34-month-olds’ imitation of real and novel words, Kehoe found that stress errors were more frequent in items that had less typical stress patterns. Similar effects of prosodic factors have been observed in the imitation of function morphemes by normally developing children (Gerken, 1991, 1994) and children with SLI (McGregor & Leonard, 1994). Yet just one study of word and nonword repetition in children with language Word and nonword repetition 8 impairment has considered the effects of prosodic structure on syllable omission. In this study, Sahl n et al. (1999) examined the vulnerability of unstressed syllables in repetition of words and nonwords by Swedish 5-year-olds with language impairment. Omission of whole syllables was found to be rare overall, but interestingly, there were six times more omissions of pre stress syllables than post stress syllables. In the light of these findings, prosody was considered an important factor in the design of a pre school repetition test. By systematically manipulating the prosodic structure of items in our task, we may expose patterns of strength and weakness in typically developing children, with some prosodic structures eliciting more, or more gross, errors than others. A common finding in children with SLI is that their language shows disproportionate effects of factors that influence normal development (Chiat, 2000; Johnston & Schery, 1976). If we extend this observation to nonword repetition, we might expect prosodically difficult structures to be disproportionately difficult for children with SLI. Articulatory complexity is yet another factor that is bound to influence performance in the target age group, but one that we are not aiming to tap. In order to control for its effects, we have kept items as articulatorily simple as possible, for example, by avoiding clusters. However, the limitations of children’s early vocabularies meant that we were not always able to achieve this. Finding items that met our prosodic requirements led us to include some that were more articulatorily complex than we would have liked, e.g. computer and its matched nonword tonkyooper . In order to avoid effects of articulatory complexity in our analysis of children’s performance, we have attempted to match the segmental content of items in different prosodic categories. More importantly, we have allowed for articulatory simplification in our scoring of children’s repetition. Taking these considerations into account, we developed a repetition test consisting of words and phonologically matched nonwords that are systematically varied for length and Word and nonword repetition 9 prosodic structure. This paper reports the performance of a group of 2 to 4-year-olds on this test, analysing the effects of one between-subject variable (age), and three within-subject variables (word status, length, prosodic structure). Our predictions were: 1. Age: 3-year-olds would perform significantly better than 2-year-olds. 2. Word status: Word scores would be significantly higher than nonword scores. 3. Length: Scores on one-syllable items would be significantly higher than scores on twosyllable items, which would be higher than scores on three-syllable items. 4. Prosodic structure: Unstressed syllables would be significantly more vulnerable to omission than stressed syllables, and unstressed syllables outside a trochaic foot (i.e., preceding stress) would be more vulnerable than those within a trochaic foot (i.e., following stress). In addition, a test of receptive vocabulary was administered for comparison with performance on our word and nonword repetition test. Method