Does strength of phonological representations predict phonological awareness in preschool children?

Previous research has shown a clear relationship between phonological awareness and early reading ability. This article concerns some aspects of spoken language skill that may contribute to the development of phonological awareness, as manifested in rhyme awareness and phoneme awareness. It addresses the hypothesis that phonological awareness abilities are associated with measures that purportedly tap into the strength of phonological representations. We examined rhyme awareness, phoneme awareness, articulatory skill, speech perception, vocabulary, and letter and word knowledge in 40 children, aged 4 to 6, who were just beginning to be exposed to formal reading experiences in private preschools. The children also received cognitive tests and tests of reading ability. The results did not validate strength of phonological representation as a unitary construct underlying phonological awareness more generally, but instead revealed a selective pattern of associations between spoken language tasks and aspects of phonological awareness. Speech perception was closely associated with rhyme awareness measures when age, vocabulary, and letter knowledge were controlled. Children with a less developed sense of rhyme had a less mature pattern of articulation, independent of age, vocabulary, and letter knowledge. Phoneme awareness was associated with phonological perception and production. Children with low phoneme awareness skills showed a different pattern of speech perception and articulation errors than children with strong abilities. However, these differences appeared to be largely a function of age, letter knowledge, and especially vocabulary knowledge. It is now well established that early reading problems are associated with difficulties in phonological awareness – difficulties in phoneme awareness being the clearest case in point (for reviews, see Lyon, 1994; Mann, 1998; Stanovich, 1993; Wagner & Torgesen, 1987). Despite the strong relationship between phonological skills and reading, the nature of the phonological deficits that underlie poor reading remains to be elucidated. Differences in performance on various measures of phonological awareness have been attributed to educational experi 2001 Cambridge University Press 0142-7164/01 $9.50 Applied Psycholinguistics 22:3 302 Foy & Mann: Strength of phonological representations ence (see Alegria, Pignot, & Morais, 1982; Mann & Wimmer, 2001; Morais, Cary, Alegria, & Bertelson, 1979), but this has not always been the case (see Bradley & Bryant, 1978; Lundberg, Oloffsson, & Wall, 1980; Mann, 1984, 1986). It may be that the development of phonological awareness is dependent on a complex interplay between individual differences in endowment and literacy-related experiences. Noting that the link between reading and phonological skill encompasses rhyme awareness, phoneme awareness, speech perception, naming, and vocabulary, several investigators have proposed that the problem for poor readers may be traced back to a bottleneck reflecting inadequate phonological processing skills (see Mann, 1998; McBride-Chang, Wagner, & Chang, 1997; Watson & Miller, 1993), possibly due to ineffective or immature phonological representations in the mental lexicon (e.g., Elbro, 1996; Fowler, 1991; Walley, 1993). Since the literature to date has not reached a consensus as to whether adequacy, effectiveness, maturity, or distinctiveness of phonological representations is the source of the problem, in the present article we put aside this issue and refer to the hypothetical deficiency as one involving “strength of phonological representations.” Our purpose is to examine whether the data support the position that a unitary construct underlies tasks that purportedly measure the qualitative aspects of phonological representations, and that this unitary construct relates to phonological awareness. For example, the qualities of a phonological representation that allow a child to perform a nonword repetition task may or may not be the same as those that are required to distinguish between pairs of words. These tasks, in turn, may or may not be equally well associated with tasks that measure phonological awareness. In this study, we examine the relationship between performance on tasks that purportedly measure the strength of children’s phonological representations and tasks that measure phonological awareness in the preschool years. We consider two different levels of phonological awareness: awareness of rhyme and awareness of phonemes. For children learning to read an alphabetic orthography such as English, phoneme awareness is the most widely replicated and accepted correlate of early reading skill (for reviews, see Adams, 1990; Brady & Shankweiler, 1991; Mann, 1998). While there is considerable evidence that phoneme awareness accrues as a consequence of exposure to letter knowledge and alphabetic print (Alegria et al., 1982; Barron, 1994; Mann, 1986; Mann & Wimmer, 2001), there is also evidence that individual differences can involve something above and beyond such exposure (Bradley & Bryant, 1978; Lundberg et al., 1980; Mann, 1986, 1991). The evidence linking rhyme awareness to reading is more controversial. Some studies reveal an association between levels of rhyme awareness and early reading skills (see Bryant, 1997; Bradley & Bryant, 1983, 1991), but others offer data that challenges this association (Duncan, Seymour, & Hill, 1997; Muter, Hulme, Snowling, & Taylor, 1998; Muter & Snowling, 1998). Likewise, some studies suggest that rhyme awareness is a developmental antecedent and a precursor of phoneme awareness (see Bradley & Bryant, 1983; Bryant & Goswami, 1987; Treiman & Zukowski, 1991), whereas others note that phoneme awareness can be independent of rhyme awareness (Duncan & Johnston, 1999). One result seems quite clear: rhyme awareness can – Applied Psycholinguistics 22:3 303 Foy & Mann: Strength of phonological representations and does – develop without exposure to a writing system (Liberman, Shankweiler, Fischer, & Carter, 1974; Morais, 1991; Treiman & Zukowski, 1991). Our primary goal in the present study is to document the relation between children’s performance on rhyme and phoneme awareness tasks and their performance on a variety of spoken language tasks that purportedly reflect the qualitative aspects of the phonological representation. These measures include speech discrimination, articulation, rapid naming, nonword repetition, and distinctness tasks. First, we review the evidence that supports the use of these tasks to measure the strength of phonological representation. We then offer a brief review of the evidence linking phoneme awareness to vocabulary and letter knowledge. Measures of phonological strength Speech discrimination. Several lines of evidence have suggested that poor reading ability (and, by implication, poor phonological awareness) relates to poor speech perception. For example, consonant perception in newborns is associated with language competence at 3 to 5 years, which, in turn, is predictive of reading achievement at 8 years (Scarborough, 1990). Children and adults with reading difficulties have difficulties perceiving speech in noise (Brady, Poggie, & Rapala, 1989; Brady, Shankweiler, & Mann, 1983), and children who are poor readers require a longer segment of a gated word in order to perceive it correctly (Metsala, 1997). Following Walley (1993), Metsala (1997) suggested that the perceptual problem associated with poor readers and the concomitant difficulty these children show with phoneme awareness follow from the fact that phonemes are not preformed perceptual units whose conscious accessibility changes with literacy exposure. Instead, she argued, phonemes gradually develop over childhood, as the growth of spoken vocabulary causes lexical representations to become more segmental. Similarly, Manis and his colleagues (Manis et al., 1997) proposed that even small speech perception problems may be very important for learning to read, and that, in certain poor readers, the phonological categories may be “more prone to disruption under stress” (p. 214), such as perception in noise or gating conditions. In their view, another source of stress is the challenge of learning to read. As Manis et al. suggested, if children cannot perceive clear distinctions between phonemes, it will be hard for them to develop representations that can be easily accessed. Problems with accessing phonological representations, in turn, lead to difficulty in segmenting and manipulating phonemes and in learning grapheme–phoneme relationships. Speech perception may also be studied by examining error patterns in tasks where children are asked to discriminate between words that vary along a single perceptual dimension. This is the strategy that we use in the present study. Speech discrimination errors often reflect perceptual distinctions that are particularly difficult for children (e.g., voicing, blends/clusters) (Treiman, 1985; Treiman, Broderick, Tincoff, & Rodriguez, 1998). Additional support for a link between speech perception, phonological representations, and poor reading can be found in studies of children with specific language impairment (SLI). These children are six times more likely to have a Applied Psycholinguistics 22:3 304 Foy & Mann: Strength of phonological representations reading disorder compared to children without the language difficulty (e.g., Ingram, Mason, & Blackburn, 1970), especially when the language delay has not resolved by the age of 51⁄2 (Bishop & Adams, 1990). Children with SLI also have speech perception difficulties. Edwards and Lahey (1998) found that SLI children were less accurate than typically developing children, not because they missed stops or unstressed syllables, but because they made syllable structure, phoneme deletion, and phoneme substitution errors. These authors suggested that children with SLI have more holistic phonological representations than even younger typically developing children, who are able to utilize representations that are relatively distinctive (Edwards & Lahey, 1998). Thus, the speech perception differences between children with SLI and typically developing children have been interpreted as reflecting the difference between immature (holistic) and segmental (mature) representations (Edwards & Lahey, 1998; Manis et al., 1997). Articulation. If we regard perceptual and awareness problems as the consequence of inadequate phonological representation, we might also look to problems with speech production as another possible consequence. Although young children’s perception of speech may be superior to their production (Mann, Sharlin, & Dorman, 1985), speech perception and production are often found to be closely related. Thomas and Senechal (1998) found that 3-year-old children who cannot accurately produce /r/ perform less well on phoneme awareness tasks that require manipulation of that phoneme, thus providing a direct link between phoneme awareness and articulation. Speech production errors in typically developing young children often reflect a difficulty in mastering certain perceptual distinctions, such as those between /r/, /l/, and /w/ (Vihman, 1996, p. 162). Children with SLI also present with abnormal speech production histories. These histories include less complex babbling (Whitehurst, Smith, Fischel, Arnold, & Lonigan, 1991), less productive use of complex syllables (Stoel-Gammon, 1989) and consonants (Roberts, Rescorla, Giroux, & Stevens, 1998), fewer vocalizations, and reduced phonetic inventories compared to controls (Paul & Jennings, 1992; Rescorla & Ratner, 1996). Like the findings about speech perception in children with SLI, findings about articulation suggest that they have delayed, rather than deviant, phonological development (Manis et al., 1997; Sussman, 1993). Additional evidence that articulatory impairment could be a consequence of inadequate phonological representations comes from studies of children with speech production difficulties. Children with phonological disorders become resistant to changing incorrect production patterns, suggesting that it is somehow difficult for them to develop representations that are more distinct (Rvachew & Jamieson, 1995). In a lip-reading task, children who were unable to produce target consonants had greater difficulty integrating audible and visible speech than children who had productive control over these consonants (Desjardins, Rogers, & Werker, 1997). The authors interpreted their findings as suggesting that children who misarticulate phonemes have internal representations for these phonemes that lack critical distinguishing phonological features. Applied Psycholinguistics 22:3 305 Foy & Mann: Strength of phonological representations Naming speed. Naming speed, which involves both articulation and lexical access, has been shown to be strongly linked to reading ability. Children with reading disorders are slower in naming simple familiar items (e.g., colors, letters, words) than children without reading disorders (see Bowers & Wolf, 1993; Denckla & Rudel, 1976). Naming speed has also been linked to phonological deficits (e.g., Ben-Dror, Pollatsek, & Scarpati, 1991; Lundberg & Hoien, 1990; Manis, 1985; Murphy, Pollatsek, & Well, 1988). As naming problems in reading-disordered subjects have been associated with slower articulation rates, it has been suggested that lexical access in reading-disordered subjects may be associated with impaired access to individual phonological representations in the lexicon (e.g., Catts, 1989; James, Van Steenbrugge, & Chiveralls, 1994; Raine, Hulme, Chadderton, & Bailey, 1991; Wolff, Michel, & Ovrut, 1990). Katz (1986) suggested that the problem may in part stem from less phonologically complete lexical representations. Nonword repetition. Nonword repetition, assessed during tasks that ask children to repeat novel phoneme sequences, is widely regarded as a measure of phonological processing (Brady et al., 1989; Edwards & Lahey, 1998; Manis et al., 1997). Studies have consistently found that children with reading and language difficulties are impaired in nonword repetition compared to children without such difficulties (Brady et al., 1989; Gathercole & Baddeley, 1993; Kamhi & Catts, 1986; Kamhi, Catts, & Mauer, 1990; Leonard, Schwartz, & Loeb, 1987; Manis et al., 1997), lending support for the view that nonword repetition is a promising predictor of reading difficulties, at least in some children (Edwards & Lahey, 1998; Manis et al., 1997). Several studies have offered the view that poor readers’ difficulties with nonword repetition may reflect a problem with the distinctness or strength of the underlying phonological representations (Edwards & Lahey, 1998; Gathercole, Willis, Baddeley, & Emslie, 1994; Manis et al., 1997). Distinctness. Distinctness – or the degree of separation between a phonological representation and similar-sounding words in the lexicon – appears to be an important factor in the development of language in general and phonological awareness in particular (Godfrey, Syrdal-Laskey, Millay, & Knox, 1981; Katz, 1986; Reed, 1989; Walley, 1993). In the distinctness task we used in this study (Elbro, Borstrom, & Petersen, 1998), children were presented with an imprecise label for a well-known object (e.g., e-phant for elephant) and were asked to provide the correct label. Such a task requires that the child accurately perceive the imprecise label, access the correct label, and execute the corrected response. The phonological distinctness of the representation of vocabulary items in kindergarten was found to contribute significantly to the prediction of phoneme awareness in grade 2 (Elbro et al., 1988). Literacy-related skills linked to phonological processing Vocabulary. Consistent findings show that vocabulary skills – especially productive vocabulary skills – have been related to the development of reading Applied Psycholinguistics 22:3 306 Foy & Mann: Strength of phonological representations ability (for reviews, see Bowers & Wolf, 1993; Wolf, 1999). Vocabulary knowledge may interact with qualities of phonological representations. For example, dyslexic adults often confuse similar-sounding words in a vocabulary task (Elbro, Nielsen, & Petersen, 1994), suggesting that phonological distinctiveness may be an important factor. Others have noted that speech perception may contribute to literacy by facilitating vocabulary development (Sawyer & Butler, 1991; Walley, 1993). Gathercole and Baddeley (1989) showed that vocabulary was a significant predictor of nonword repetition accuracy, implying that existing lexical representations may facilitate nonword repetition abilities or that the factors that contribute to vocabulary development also affect nonword repetition skills. We included a vocabulary measure in our battery so that we could gauge whether vocabulary mediates any relationship between our spoken language tasks and phonological awareness. Letter knowledge. Several lines of evidence point to the significance of letter knowledge in the acquisition of reading. Illiterate adults have poor phoneme awareness, supporting the position that phoneme awareness is a consequence of learning to read an alphabetic orthography (Morais et al., 1979). A refinement by Barron (e.g., 1991, 1994), termed the proto-literacy hypothesis, proposes that the critical level of literacy for the inducement of phonemic awareness is merely letter knowledge as opposed to decoding ability. Barron suggested that a child who, for example, learns that A is /ae/ may be aware of /ae/ as an invariant segment of words, yet may be unable to recover the sequence of phonemes that a sequence of letters transcribes. Consistent with this proposal, Mann and Wimmer (2001) showed that phoneme awareness is almost completely lacking among German kindergartners, who identified only 28% of the letters, whereas it is quite well developed among American children, who identified 94% of the letters. In their data, there was an even closer connection between reading ability (in particular, phonological recoding ability) and phoneme awareness than between letter knowledge and phoneme awareness. A similar observation was reported by Bowey (1994). Mann and Wimmer also observed substantial individual variation among the kindergarten children. For example, one German child without any letter knowledge exhibited perfect performance on the phoneme judgment task. Nevertheless, the data, on average, are consistent with the possibility that phonological awareness is related to letter knowledge and with Barron’s (1994) suggestion that knowledge of letter names may actually induce phoneme awareness. Previous research findings are also consistent with a view that both the acquisition of letter names and the emergence of phonological awareness are aided by a change in the strength of phonological representation. This change may allow children to distinguish among letter names and other highly similar words in the lexicon (see Walley, 1993); it may be easier for them to map orthography onto phonological representations that are well specified (Elbro et al., 1994). These considerations led us to include measures of letter knowledge and reading so that we could account for any relationship that might emerge between the speech and phonological awareness tasks. Applied Psycholinguistics 22:3 307 Foy & Mann: Strength of phonological representations Objectives of the present study The literature suggests that there is a strong relation between phonological awareness and performance on tasks that purportedly tap into the strength of phonological representations. Few studies, however, have examined this relation in children just beginning to be exposed to the alphabetic principle, few have examined multiple tasks, and few have considered whether the relation depends on the type of awareness being measured. We attempt to identify how rhyme awareness and phoneme awareness relate to several spoken language tasks that purportedly measure phonological strength. We also measure vocabulary and letter knowledge, given their link to phonological awareness as well as to some of the spoken language tasks we employ. We examine the following hypotheses. First, tasks that purportedly measure the strength of the phonological representations should be related to each other and to the children’s phonological awareness. Second, the relation between the strength of phonological representations and performance on phonological awareness tasks should be a function of the level of awareness being assessed (rhyme vs. phoneme), especially given that phoneme awareness is clearly tied to literacy experience. Third, levels of performance as well as error patterns on spoken language tasks should be related to level of phonological awareness. Errors for less aware children should reflect holistic phonological representations, resembling those previously observed in children with SLI.