Spoken Vocabulary Growth and the Segmental Restructuring of Lexical Representations: Precursors to Phonemic Awareness and Early Reading Ability
Jamie L. Metsala
University of Maryland, College Park
Amanda C. Walley
University of Alabama at Birmingham
In this chapter, we present a model of the development of spoken word recognition, the processes involved in matching speech input to lexical patterns stored in memory. Our lexical restructuring model was formulated to account for developmental changes in the structure of spoken word representations and the growth of phonological awareness (see also Metsala, 1997a, 1997b; Walley, 1993b). According to this model, the representations supporting spoken word recognition become increasingly segmental with spoken vocabulary growth, and this change makes possible explicit access to phonemic units. We propose that lexical restructuring is a protracted process that extends into early and even middle childhood. This restructuring is influenced by the words that are known at a given point in time and that must be distinguished from one another for successful recognition. Variations across children in lexical growth and in the restructuring process contribute to individual differences in phonemic awareness, and thus success in learning to read an alphabetic orthography (for a similar position, see Fowler, 1991).
We first outline two theoretical positions on the developmental origins of the phonemic segment, each of which is grounded in research on adult spoken word recognition and infant speech perception. Two models of adult recognition are described, and then we survey studies of infant perception contrasting the earliest inferences drawn from this work concerning the initial status of the phoneme with more recent interpretations. Second, we examine work pointing to the emerging segmental structure of speech representations in late infancy, which has been widely attributed to vocabulary growth; this attribution serves as a primary impetus for our model. Third, we describe the main claims of the model, which is unique in its focus on spoken word recognition in early and middle childhood, and then review empirical findings that support its predictions regarding the protracted emergence of segmental representations. Fourth, we discuss evidence for the proposed link between the emergence of the segment as an implicit, perceptual unit of spoken word representation and processing, and as a more explicit and accessible cognitive unit that can be harnessed for the reading task. Specifically, we consider basic spoken language development and its relation to phonemic awareness, as well as the speech perception difficulties of reading-disabled children. Our overriding goal is to bring together what have often been viewed as separate developmental domains—namely, basic speech processing, phonemic awareness, and early reading. The links across these three domains require further scrutiny, and our model is only a starting point in this endeavor.
THE DEVELOPMENTAL ORIGINS OF THE SEGMENT
A Preformed or Emergent Unit? The Accessibility and Emergent Positions
The empirical relations between phonemic awareness and reading success are now well established. Both children and adults with greater phoneme segmentation and manipulation ability demonstrate superior word reading skill for alphabetic orthographies, which rely on phoneme–letter correspondences. Prereaders and poor readers do not perform very well in phoneme segmentation and manipulation tasks (for a description and analysis of such tasks, and discussion of their relation to early reading, see Stahl & Murray, chap. 3, this volume). Importantly, there are two main theoretical positions regarding the developmental origins of the segment, each of which tends to be associated with a particular stance on the causal relation between phonemic awareness and early reading ability (for elaboration, see Fowler, 1991; Walley, 1993b).
According to the traditional accessibility position, phonemic segments are highly modularized and largely available only for that purpose for which they evolved, namely basic speech processing. Although these units may be present and functional even in infancy, they are not accessible at an explicit, conscious level before reading experience with an alphabetic orthography or with metacognitive development more generally (e.g., Gleitman & Rozin, 1977; Liberman, Shankweiler, & Liberman, 1989; Morais, Cary, Alegria, & Bertelson, 1979; Rozin, 1976).1 This position is largely adevelopmental in that the perceptual units that support speech processing are viewed as essentially preformed. Neither phonemic segments nor the lexical structures they comprise undergo any change in their fundamental nature.
Although many reading researchers agree that experience with an alphabetic system enhances phonemic awareness (see, e.g., Ehri, 1984; cf. Morais et al., 1979; Morais, Bertelson, Cary, & Alegria, 1986), some degree or level of phonological awareness is an important precursor to early reading success (e.g., Share & Stanovich, 1995). There has, however, been little attention paid to the development of the phoneme prior to its emergence as a consciously accessible unit that can be deployed for reading purposes. Studies pointing to the stability of reading-related phonological processes from a young age (see Torgesen & Burgess, chap. 7, this volume) make understanding the origins and growth of the segment in the prereading period a priority. Our focus is on developmental research that speaks to this crucial issue, but that has been largely overlooked by those interested in the acquisition of reading-related abilities.
Research on spoken word recognition and speech perception in childhood supports a more recently influential account of the development of the segment, which we refer to as the “emergent” position. According to this position, the phoneme is not an integral, hard-wired aspect of speech representation and processing. Rather, the phoneme emerges with spoken language experience as a result of interactions between vocabulary growth and performance constraints (e.g., Ferguson, 1986; Jusczyk, 1986, 1993; Lindblom, 1992; Lindblom, MacNeilage, & Studdert-Kennedy, 1983; Nittrouer & Studdert-Kennedy, 1987; for additional references, see Walley, 1993b). In an elaboration of this position, we have stressed the gradual, twofold nature of this change. Over the course of early and middle childhood, the phoneme emerges first as an implicit perceptual unit that is used in basic speech processing, and only later as an explicit unit that can be deployed for reading-related activities (Metsala, 1997a; Walley, 1993b; see also Fowler, 1991).
According to this emergent position, phonemic awareness is not “simply” a problem of accessing underlying units of speech representation, but also is limited by the very nature of these representations, which undergo substantial developmental change. Our Lexical Restructuring Model (LRM): emphasizes the impact of continued vocabulary growth on spoken word recognition in childhood, and what such growth encompasses; and delineates specific relations that might be expected among basic speech processing, phonemic awareness, and early reading ability. To explicate the emergent position and our model, we turn now to models of adult spoken word recognition and then to studies of infant speech perception.
Models of Adult Spoken Word Recognition
Spoken word recognition is widely regarded as a process by which a given word is perceived in the context of other words in memory, or is discriminated from various lexical alternatives (see Luce, 1986). This theoretical assumption is also central to our model. For now, let us consider how it is realized in two models of adult recognition: the cohort model and the Neighborhood Activation Model (for detailed evaluations of these and other models, see Cutler, 1995; Nygaard & Pisoni, 1995). Both models are concerned specifically with spoken word recognition, which has developmental priority over visual word recognition.
In the cohort model (e.g., Marslen-Wilson, 1987, 1989), the emphasis is on the temporal course of spoken word recognition. There are two core components of recognition: the multiple activation or access of all word candidates that are consistent with word-initial speech input (about the first 100 ms to 150 ms of a word, or one to two phonetic segments), and the selection of a particular candidate through the elimination of competitors that are inconsistent with subsequent input. The model therefore accounts for the fact that adult listeners can recognize words on the basis of partial speech input, before they have heard the entire word (e.g., Grosjean, 1980; cf. Bard, Shillock, & Altmann, 1988).
A more elaborate treatment of the structural organization of the lexicon is provided by the Neighborhood Activation Model (NAM; e.g., Goldinger, Luce, & Pisoni, 1989; Luce, 1986). According to NAM, the speed or accuracy with which a target word is recognized depends on its similarity neighborhood structure, which encompasses the number and degree of confusability of words that overlap, on a segmental basis, with the target, and the frequency of the word and the combined frequency of the word’s neighbors.
In both models, recognition involves discrimination among segmentally structured lexical representations that are activated in parallel on the basis of speech input.2 In the selection component of the cohort model, activated word candidates are also monitored in parallel, such that no cost is incurred for the processing of large versus smaller cohorts. In contrast, in NAM the ease of recognizing a word depends on the number and nature of activated, and thus competing, words. In support, adult recognition is better for high-frequency words with few neighbors than for low-frequency words with many neighbors (e.g., Goldinger et al., 1989).
These assumptions are useful for conceptualizing how recognition might differ in young versus older listeners and how development might proceed. In particular, children’s lexical representations may not, at the outset, have the segmental structure needed to support recognition from partial speech input. However, their lexicons grow rapidly, which might precipitate the implementation of more detailed, segmental units of representation. Yet this restructuring may be gradual, because vocabulary growth continues to be substantial throughout much of childhood. Thus, recognition by children and older listeners should differ for some time. We next examine the extent to which these expectations are supported by infant speech perception studies.
Infant Speech Perception
Much of what we know about the development of speech perception comes from studies of infants (for review, see Aslin, Pisoni, & Jusczyk, 1983; Jusczyk, 1995; Werker, 1991). This focus on infancy can be traced to the seminal study of Eimas, Siqueland, Jusczyk, and Vigorito (1971), who found that 1- to 4-month-old infants discriminated synthetic consonant–vowel stimuli varying in voice-onset time (a speech cue that signals, e.g., the difference between bat and pat) in a categorical and thus adultlike manner. That is, infants discriminated stimuli from different phonemic categories (/b/ vs. /p/), but not stimuli from within the same category. Because categorical perception was widely considered to uniquely characterize human perception of speech, these data were interpreted as indicating that phonemic segments and/or their associated features are specified innately, or constitute part of our biological makeup for speech processing.
The finding that young infants discriminate stimuli differing by a single phonemic segment was subsequently extended to a variety of other contrasts (see Aslin et al., 1983; Jusczyk, 1995; Werker, 1991). However, two other results quickly called into question the view that this early capacity is species-and speech-specific. First, similarities were observed in the perception of speech by humans and nonhumans (e.g., Burdick & Miller, 1975; Kuhl & Padden, 1982). Second, similarities in the perception of nonspeech patterns by adults and infants were found (e.g., Jusczyk, Pisoni, Walley, & Murray, 1980; Pisoni, Carrell, & Gans, 1983). These results suggested that basic speech processing is mediated by general auditory, as opposed to speech-specific mechanisms. However, as Miller and Eimas (1994) noted, there are problems in choosing between these theoretical alternatives: Cross-species parallels in perception do not necessarily imply the same underlying mechanism, and a modularized speech processor might be “fooled” by sufficiently complex nonspeech stimuli (cf. Fowler, 1995).
Still, other observations cannot be readily accommodated within the “speech is special” framework. Although young infants can discriminate a wide variety of speech contrasts, older infants do not always perform so impressively even for their native language, and perception may not approach adultlike levels until middle childhood or beyond (e.g., Flege, & Eefting, 1986; Nittrouer & Studdert-Kennedy, 1987; Walley, Flege, & Randazza, 1998). This discrepancy has been attributed in part to the different tasks used to assess perception across age levels (e.g., Jusczyk, 1992; Studdert-Kennedy, 1986). Those paradigms used to assess infant perception have, for the most part, yielded only discrimination data, which are not sufficient to support the claim that speech is perceived at a linguistic as opposed to a more general auditory level. Most important, discrimination could be mediated by holistic processes, and not by the detection of localized, segmental differences.
In fact, there is little positive evidence that infants’ speech representations are, at the outset, structured in terms of segments. In a series of experiments addressing this issue, Jusczyk and colleagues (see Jusczyk, 1992, 1993) showed that young infants are sensitive to the presence of a common syllable, but not a common phoneme in sets of stimuli. In addition, the results of several studies (see Cooper & Aslin, 1989; Jusczyk, 1995) suggest that prosodic information, which is distributed globally throughout the speech waveform, is more likely to capture infants’ attention than segmental information. For example, soon after birth, infants prefer to listen to child-directed speech or “Motherese,” in which prosodic features (such as intonation contour, stress, and pauses) are typically exaggerated, rather than to adult-directed speech (e.g., Cooper & Aslin, 1990; Pegg, Werker, & McLeod, 1992). Newborns also distinguish between passages spoken in their native language versus a foreign one, even when the passages are low-pass filtered, and their segmental content thus largely removed (Bertoncini, Bijeljac-Babic, Jusczyk, Kennedy, & Mehner, 1988). Such early sensitivity to prosodic information may provide infants with a starting point for parsing the stream of speech around them into increasingly smaller and linguistically relevant units, such as clauses, phrases and words (e.g., Hirsh-Pasek et al., 1987; Jusczyk & Aslin, 1995; Jusczyk et al., 1992). The word may, in turn, serve as a crucial point of entry to the segmental level of the native language. When and how, more precisely, do infants gain such entry?
Cross-linguistic studies of infant perception help to provide an answer. These studies suggest that infants from a given language community can initially discriminate a wide range of both native and nonnative contrasts. However, sensitivity to the latter generally wanes over the second 6 months of life—at around the same time that infants are becoming more sensitive to native language sound properties. For example, Werker and Tees (1984) tested infants from Canadian-English speaking homes aged 6 to 8 months, 8 to 10 months, and 10 to 12 months. Although the youngest infants discriminated English, Hindi, and Nthlakampx consonantal contrasts, only some of the infants from the middle age group discriminated the foreign contrasts, and those from the oldest group discriminated the English contrast only. In addition, 9-month-old American and Dutch infants prefer to listen to lists of words from their native language, whereas 6-month-olds do not; however, when the words are low-pass filtered, leaving only prosodic information that is similar for English and Dutch, older infants exhibit no preference, suggesting that they are beginning to focus on the individual sound segments comprising the utterances (Jusczyk, Friederici, Wessels, Svenkerund, & Jusczyk, 1993). More recently, Jusczyk, Luce, and Charles-Luce (1994) showed that, by 9 months of age, infants are sensitive to language-specific phonotactic sequences, including their frequency distribution.
Such enhanced sensitivity to native language contrasts has been attributed by many to the onset of early word learning, when sounds first become interfaced with meaning (but see, e.g., Kuhl, Williams, Lacerda, Stevens, & Lindblom, 1992). This interfacing is, for example, central to Jusczyk’s (1992, 1993) model of word recognition and phonetic structure acquisition (WRAPSA). According to WRAPSA, general auditory analyzers first yield a detailed but transient description of an incoming acoustic signal, and it is this analysis that mediates infants’, adults’, and nonhumans’ basic perception (e.g., discrimination) of various speech and nonspeech patterns. However, crucial to the attainment of word recognition ability is the development of an interpretive scheme, which debuts around 9 to 12 months of age. This interpretive scheme provides a more selective and enduring representation of the speech signal (in the form of syllables, for much of childhood) by weighting information from the preliminary auditory analyzers, such that features critical for signaling meaningful distinctions are given prominence, and those that are irrelevant are rendered less discriminable. Thus, the infant begins to learn that bat means something different than pat, but that physically different renditions of bat are equivalent. Not only does this model account for infants’ and adults’ perception of native and nonnative speech patterns, it also speaks to other requisite (e.g., memory-related) abilities for word recognition (cf. Kuhl, 1993). More generally, this model attempts to explain how infants’ perceptual abilities are relevant to later language learning. As yet there is, however, little empirical evidence directly linking attentional shifts in speech perception to vocabulary development. In particular, such shifts have not, within a given infant or group of infants, been tied to “recognitory comprehension,” or the ability to make a correct behavioral response to a spoken word without extralinguistic, contextual support (e.g., Thomas, Campos, Shucard, Ramsay, & Shucard, 1981). For example, would 9- to 12-month-olds be more likely to respond correctly when told to “Look at the dog!” by various native, as opposed to nonnative, English speakers?
In summary, some researchers have begun to reevaluate the status of the segment as an early perceptual unit, and there has been a growing consensus that phonetic/phonemic segments are emergent units of speech perception (see Studdert-Kennedy, 1993; Walley, 1993a). Yet, the related issues of what constitute the primary structural units of infant perception, the extent to which these are speech-specific, and whether their development proceeds in an active or passive manner have not been resolved. Fortunately, there is additional evidence from studies of toddlers’ first word representations, as well as from another, until recently neglected area—namely, phonetic perception and spoken word recognition in childhood—that supports the emergent position, and the claims of our LRM more specifically.
EARLY WORD REPRESENTATIONS
The first lexical representations of the older infant/toddler (about 9–12 months to 18 months of age) may be holistic or undifferentiated in that there is little intraword differentiation and interword organization. Because the child’s vocabulary is still very small, there would be little need to represent words in any systematic and detailed manner (e.g., as sequentially arrayed phonemic segments). Instead, words might be represented and recognized on the basis of individual salient characteristics or their overall acoustic shape or prosodic structure (e.g., Jusczyk, 1986; Menyuk & Menn, 1979). Representations might also be holistic because the child is in the process of establishing a lexical base. This is a multifaceted task that includes segmenting words from continuous speech, noting correspondences between recurring speech patterns and non-linguistic events, discovering the relevant nonphonological (e.g., semantic and syntactic) features of words, translating sound sequences into articulatory sequences that can be used in production, and so on (e.g., Menyuk & Menn, 1979). All this must be accomplished in addition to any momentary recognition task—matching an input speech pattern to representations already in memory. Systematic and detailed segmental matching may therefore exceed the child’s attentional and/or memory capacity, such that words are instead stored and retrieved as unanalyzed wholes.
These expectations are consistent with the emergent view of the development of the phoneme in that segmental information is not, at the outset, used at the level of perceptual representations that are accessible in real time and relevant to word recognition (see Aslin & Smith, 1988). However, the empirical evidence for the holistic nature of early word recognition is not extensive or direct; rather, it derives from a limited number of perceptual studies in late infancy and toddlerhood, and from studies of early word productions.
Perception
Only a few studies have assessed speech perception in late infancy and toddlerhood—as the child gains entry to the native language proper and begins to show signs of comprehending words. Although the lower limit on this ability is often placed at 9 months, based largely on observational work (e.g., Benedict, 1979; Huttenlocher, 1984), experimental studies of recognitory comprehension yield a more conservative estimate (for discussion, see Walley, 1993b); for example, Thomas et al. (1981) found that 13-month-olds, but not 11-month-olds, looked longer at the referents of words that were familiar, according to maternal report, than they did at the referents of unfamiliar words. In studies concerned with phonetic perception, objects, such as blocks with faces on them, are assigned nonsense names, pairs of which differ only in their initial consonant (e.g., “bak” vs. “mak”) and children are asked to point to one of the objects. Under these conditions, they have difficulty perceiving single phoneme differences. For example, Shvachkin (1973) reported that 18-month-old Russian children could correctly distinguish a “bak” from a “zub” or a “mak” from a “zub,” but not a “bak” from a “mak.” Similar results have been reported for 17- to 22-month-old American children (Garnica, 1973) and even for 3-year-olds (Edwards, 1974). More recently, Werker and Baldwin (see Werker & Pegg, 1992) found that 19-month-olds, but not younger children, look longer at a pictured object when presented with a matching a test word (e.g., dog), as opposed to a minimal pair nonword, a phonetically dissimilar nonword, or a phonetically dissimilar real word (e.g., bog, luf, or car)3 Thus, older infants and toddlers show the beginnings of access to segmental information for highly familiar words, but this ability is certainly not fully developed.
Production
The speech perception data suggest that phonemes, as linguistic units that distinguish sound patterns in terms of their meaning, are less salient for older infants and toddlers than they are for adults. This difference is mirrored in early speech productions, which have proved more tractable for study (albeit primarily through observational, rather than experimental methods; but see, e.g., Gerken, 1994). In the production literature, there is a more long-standing consensus that segments do not serve as early units of speech representation, and that children’s first lexical representations are holistic in nature (see Ferguson, 1986; Studdert-Kennedy, 1986; Suomi, 1993; Walley, 1993b; for models relating the development of production and perception, see Menn & Matthei, 1992; Suomi, 1993; Vihman, 1993).
On the one hand, the first 50-word stage of production (beginning anywhere from 9 to 18 months of age) is characterized by little phonetic variability or “mobility,” such that some words seem “fixed” (Bloch, 1913, cited in Ingram, 1979). Some pronunciations are quite accurate, and more advanced than later ones; others persist in falling short of the rest of the child’s productions (sometimes referred to in the literature as progressive and regressive “phonological idioms,” respectively). On the other hand, some early words are highly variable in form. A classic example is that of K, who offered 10 different renditions of pen within one half-hour (Ferguson & Farwell, 1975); another is Jennika, who produced three phonological variants of blanket on the same day (Ingram, 1979). Thus, there is considerable interword and intraword variability in the child’s early speech, and phonemes do not seem to exist initially as functional, commutable, and “crystallized” elements in his or her phonological repertoire (e.g., Studdert-Kennedy, 1986). Rather, “There is some word-by-word learning to pronounce … in which the whole word is an indivisible target” (Menyuk & Menn, 1979, p. 63).
According to many researchers then, it is the word, not the phoneme, that serves as the basic unit of early phonological contrast—a position perhaps most evident in Ferguson’s work (e.g., Ferguson, 1986). The word is primary because it is the simplest nonprosodic unit with which the child can accomplish communicative intent; it represents the first major interface between sound and meaning in development (Studdert-Kennedy, 1986), as well as in online spoken language processing (e.g., Marslen-Wilson, 1987). Some place greater emphasis on the initial importance of larger units, such as intonational phrases (e.g., Moskowitz, 1973; Peters, 1983), whereas others stress the role of the syllable (e.g., Jusczyk, 1993; Menyuk & Menn, 1979); indeed, there may be individual as well as cross-language differences in reliance on different processing units (Cutler, Mehler, Norris, & Segui, 1989; Menyuk & Menn, 1979). Nevertheless, there is general agreement that development entails the increasing differentiation of speech representations, such that “Smaller units (syllables, clusters, segments) become more definitely located within [larger ones, such as the] word” (Ferguson, 1986, p. 50). Ferguson emphasized that this discovery or construction of a phonological system by the child does not center around the phoneme, but proceeds gradually via “lexical diffusion,” or in a word-dependent manner (cf. Jakobson, 1968).
The Vocabulary Growth Spurt and the Beginnings of Segments
Although early phonological development has been characterized as a period in which lexical representations of a holistic nature predominate, one major event that may prompt a change in the nature of these representations is the vocabulary growth spurt, or naming explosion. Whereas the child’s first 50 words or so are acquired slowly and one at a time, after this point (i.e., from about 18 months to 3 years of age), many children show a large and sudden increase in the number of words they can produce and comprehend (see Reznick & Goldfield, 1992). As a growing number of words overlap in their acoustic properties, there should be considerable pressure to implement more fine-grained (e.g., segmental) representations. Such representations would facilitate fast and accurate discrimination of a growing base of lexical alternatives (e.g., Charles-Luce & Luce, 1990; Jusczyk, 1986; Walley, 1993b), and support more efficient articulation (e.g., Lindblom, 1992; Menn, 1983; Studdert-Kennedy, 1986).
With respect to how lexical restructuring proceeds, Jusczyk (1986, 1993) proposed that a word recognition network might first be organized around the spectral onset characteristics of syllables, such that words with similar onsets are located close to one another. This organization, which mirrors the temporal structure of input patterns and capitalizes on the saliency and robustness of word-initial information, would narrow the set of items to be searched and enhance the speed and accuracy of recognition. According to Jusczyk, the development of segmental representations might go forward in a similar manner, but perhaps only later when learning to read, especially as unfamiliar words are encountered (see also Ferguson, 1986; Peters, 1983; Vihman, 1981). Yet greater detail about similar syllables might have to be represented earlier.
Locke (1988) suggested that beginning with the vocabulary growth spurt, children are increasingly likely to encounter minimal pairs of words (e.g., cat and hat) and thus to realize that some words differ in only one part. In support, there is a high degree of synchrony in the number of different initial consonants and the number of words produced by 2-year-olds (Stoel-Gammon, 1991). There is also the early-occurring phenomenon of homonymy, or the use of a single sound pattern to produce two or more phonologically similar words (e.g., [aš] for cheese, sausage, sock, and slipper; Vihman, 1981). According to Ingram (1985), however, homonymy eventually declines because of increases in phonetic inventory size, which, in turn, are the result of vocabulary increases.4 As Schwartz (1988; see also Studdert-Kennedy, 1987) proposed:
[Such] word associations may enable the child to recognize similarities and ultimately differences between the phonetic forms for different words both in perception and in production. This may allow the child to form categories of word structures and perhaps facilitate the extraction of segments or features that are shared by certain words and differentiate others.… [That is, eventually] the child must progress to a point of extracting smaller units. [Homonymy] may be one of the ways in which this progression occurs, (p. 216)
We concur with the increasingly influential notion that vocabulary growth serves as an important mechanism for prompting the differentiation of lexical representations and enhancing the discriminative power of the recognition process. More generally, there is a trend from wholes to parts that can be found across many areas of perceptual-cognitive development (see Aslin & Smith, 1988). However, our LRM focuses on spoken word recognition beyond late infancy and toddlerhood. In addition, it is concerned with how the increasingly segmental nature of representations provides the foundations for explicit segmentation ability and thus early reading success (see also Fowler, 1991).
LATER WORD REPRESENTATIONS
The Lexical Restructuring Model
The first claim of our LRM is that young children recognize words in a more holistic manner than do older children and adults, due to both the continued segmental restructuring of lexical representations and processing limitations that arise from the demands of a rapidly growing vocabulary. Therefore, despite similarities in basic phonetic perception by infants and adults, these should be differences in young children’s and adults’ performance for tasks that more closely resemble word recognition.
Although the emergence of segments may have sudden beginnings or be precipitated by the vocabulary growth spurt, the segmental restructuring of lexical representations may be quite gradual, extending into early and middle childhood (Fowler, 1991; Walley, 1993b). The central reason for such protracted restructuring, according to our model, is that there is still substantial spoken vocabulary growth after age 3 (the upper limit on the growth spurt) and prior to extensive reading and writing experience. That is, there are changes in the absolute size of children’s lexicons, and the rate of vocabulary increase is still quite dramatic over the preschool and early elementary school years. Anglin (1989), for example, estimated that 4-year-old preschoolers know 2,500–3,000 words, first graders know 7,000–10,000 words, and fifth graders know 39,000–46,000 words. In addition, there are increases in the familiarity status of individual words and changes in the organizational structure among words based on their sound-similarity relations or segmental overlap.
The advantages of more segmental, adultlike representations, such as the ability to recognize words from partial speech input, may not be immediately apparent after lexical restructuring. Early decisions based on segmental processing of speech input may tax children’s attention and memory resources and cause recognition errors when new words are presented. Thus, children may continue to rely on holistic processing, or pay greater attention to information distributed throughout the speech input, until vocabulary knowledge becomes more stable.
The second claim of our model is that lexical restructuring does not occur in an all-or-none manner, or on a systemwide basis, but rather is gradual and word-specific, depending on such factors as overall vocabulary size or rate of expansion, as well as the familiarity status and sound-similarity relations among individual words in the child’s lexicon.
A growing vocabulary necessitates recognition among words of increasing similarity in their phonological structure. In the adult word recognition literature, “neighborhood density” has typically been defined in terms of the number of words in the lexicon that differ from a given target word by a one-phoneme substitution, deletion, or addition (Logan, 1992; Luce, 1986). Words with many similar-sounding neighbors reside in “dense” neighborhoods. Thus, recognizing the word big will require a more fine-grained representation for a child who also knows the words bag, bug, bib, bit, dig, and wig, than for a child who does not. Some words do not have many similar-sounding neighbors, even in the adult lexicon (e.g., girl); these words reside in “sparse” neighborhoods. According to our model, words residing in dense neighborhoods will need to be segmentally represented earlier in development than will words from sparse neighborhoods. As new words enter the lexicon, they must be analyzed for their phonological structure and cross-referenced with existing representations to facilitate efficient storage and online recognition. Thus, children’s performance on word recognition tasks should be better (i.e., most adultlike) for words in dense neighborhoods, and developmental differences greatest for words in sparse neighborhoods.
Another important factor in our model is the familiarity status of individual words. Word familiarity depends on at least two factors: experienced frequency and age of acquisition (AOA). Experienced frequency and AOA are correlated, but not identical. For example, cartoon, is acquired early by most children, but is not necessarily heard often relative to many other words; cartilage, on the other hand, is learned later by most people, but could be heard frequently by some (e.g., medical students). In the adult literature, it has been suggested that early-acquired words have more robust or detailed representations than later-acquired ones (e.g., Brown & Watson, 1987), and recent research indicates that AOA, when unconfounded with word frequency, may have a greater influence on recognition performance (e.g., Morrison & Ellis, 1995; but see Garlock, Walley, Randazza, & Metsala, 1998). To date, however, there has been little developmental research examining the relative impact of these two dimensions of word familiarity on recognition. Therefore, in our model, words that are acquired at a young age, as well as words that are heard frequently, are assumed to undergo more extensive segmental restructuring and to be recognized better by children (in a more adultlike manner) than are later-acquired or infrequently encountered words.
The third claim of our model is that the gradual segmental restructuring of lexical representations plays a primary role in the development of explicit segmentation ability (i.e., phonemic awareness). That is, the phoneme emerges first at an implicit level for the perceptual representation and processing of spoken words, and thus only later as a cognitive unit that can be consciously accessed and manipulated. According to this claim then, implicit and explicit segmentation ability should follow similar developmental trends, with the latter delayed relative to the former. Furthermore, because word characteristics, such as familiarity and neighborhood density, are the causal mechanisms underlying restructuring of lexical representations, we might expect to observe a relation in performance across implicit and explicit tasks for specific items (e.g., better performance in both recognition and awareness tasks for high-frequency words with many neighbors than for low-frequency word with few neighbors).
The fourth claim of our model is that deficits in lexical restructuring play a causal role in reading-disabled children’s difficulties with phonological processing, phonemic awareness, and reading ability. If lexical representations do not become segmentalized in a developmentally appropriate manner or time frame, children should be unable to access phonemes and to learn the relation between phonemes and graphemes (i.e., decipher the alphabetic code). Therefore, our model offers a developmental framework for understanding the pervasive phonological processing deficits observed in individuals with reading disabilities. Reading-disabled children should show particular deficits for unfamiliar words or words with few similar sounding neighbors, and their performance on spoken word recognition tasks should resemble that of younger children with similar vocabulary knowledge and lexical representations. To what extent are these claims of our model supported by the existing empirical data?
Empirical Evidence for Gradual Lexical Restructuring
Phonetic Perception in Childhood
Studies of phonetic perception by children support the first claim of our model that speech/lexical representations are initially holistic and only gradually become segmentally structured during early and middle childhood. Despite the impressive discrimination capacities of infants, a number of studies have indicated that phonetic perception remains incomplete or nonadultlike for several years (for review, see Barton, 1980). For example, children aged 2 years, 11 months to 3 years, 5 months were asked to discriminate the pairs rake/lake, wake/rake, and wake/bake using either recorded, synthetically produced, or live stimuli (Strange & Broen, 1981). All of the children performed above chance on all three contrasts in at least one of the stimulus conditions. Nevertheless, even in this highly constrained testing situation, there was a high degree of variability among subjects. Along similar lines, Zlatin and Koenigsknecht (1976) examined discrimination of voiced and voiceless stop consonants. Two-year-olds required a greater difference for successful discrimination than did 6-year-olds or adults. Even 6-year-olds required a greater difference than did adults when the voiced/voiceless discrimination involved a velar pair (e.g., goat/coat).
More recently, Nittrouer and Studdert-Kennedy (1987) showed that 3- to 5-year-olds were less sensitive than were 7-year-olds and adults to frequency information in fricative noise when identifying stimuli along an /s–š/ continuum. Identification functions were also shallower for children when compared to adults, indicating that their phoneme categories are less well defined (see also Burnham, Earnshaw, & Clark, 1991; Walley et al., 1998). A follow-up study revealed that 5-year-olds’ productions were more affected than were 7-year-olds’ and adults’ by vocalic context, or coarticulated information distributed across the speech waveform (Nittrouer, Studdert-Kennedy, & McGowan, 1989). Moreover, studies with disyllabic stimuli indicate that children under the age of 7 years may attempt to extract syllables rather than phonetic segments from the speech stream (Nittrouer, 1992). In some cases, phonetic development may not be complete until young adulthood. Flege and Eefting (1986) showed that the /da/-/ta/ boundary occurred at a longer mean voice-onset time (VOT) value in both English-and Spanish-speaking adults when compared to their 9-year-old counterparts. These boundary differences were still observed when English-speaking adults were compared with 11-, 13-, and 17-year-olds.
Spoken Word Recognition in Childhood
Studies of the development of spoken word recognition support the first and second claims of LRM—namely, that recognition shifts gradually through middle childhood from being based on holistic to segmental processes, and that this shift is a result of changes in the characteristics of individual words, such as familiarity and neighborhood density. Next we highlight supporting results from three tasks that have been used to study speech representation and spoken word recognition in children.
Similarity Judgments. Similarity judgments are obtained from classification tasks in which subjects are typically given triads of speech stimuli and asked to select the pair that best goes together, and provide some indication of what speech information is most salient to listeners. Treiman and Baron (1981) employed this sort of task with kindergartners, first graders, and adults and presented triads such as /bI/, /ve/, and /bo/. Adults made more common phoneme classifications (e.g., judged /bI/ and /bo/ as most similar), whereas children made more global similarity classifications (e.g., judged /bI/ and /ve/ as similar), even when they were specifically reinforced for making common phoneme classifications. These findings were extended by Treiman and Breaux (1982) to three-phoneme syllables (e.g., /bIs/, /diz/, /bun/) in both training and memory procedures. In the first experiment, similarity training was more effective for preschoolers when compared to phoneme training. Phoneme training was more effective for adults. In a second experiment, subjects were taught the three syllables as the names of three animals. When tested immediately after the training phase, children were more likely to confuse syllables that were globally similar. Adults were more likely to confuse syllables that shared a phoneme (see also Walley, Smith, & Jusczyk, 1986).
More recently, Byrne and Fielding-Barnsley (1993) showed that similarity effects may underlie children’s apparent recognition of phoneme relations. For example, when children are asked whether bowl or shed begins with the same sound as beak, children sometimes give the adultlike answer bowl However, beak and bowl not only share a common initial phoneme, they are also more globally similar than are beak and shed. To address the confounding of these variables, two versions of this phoneme identity task were developed. In the first version, common phoneme relations and overall similarity relations were purposely confounded such that the words which shared a common phoneme were also more similar on global ratings (e.g., beak, bowl, shed). In the second version, one of the test words again shared a common phoneme with the target, but the foil was more globally similar (e.g., beak, bowl, dig). Kindergartners performed better when phoneme identity was confounded with global similarity, suggesting that children do base their judgments on overall acoustic characteristics. Byrne and Fielding-Barnsley also showed that performance on the unconfounded version of this task was more strongly related to pseudoword reading and spelling ability than was performance on the confounded version. Thus, phonemes are not very salient to young children, but children’s growing sensitivity to this unit is important to later reading ability.
Mispronunciation Detection. Mispronunciation detection tasks involve presenting listeners with intact and mispronouced words in stories, sentences, or in isolation and asking them to detect which words are mispronounced. Adults detect mispronunciations in word-initial position more accurately but more slowly than noninitial mispronunciations (e.g., Cole & Perfetti, 1980; Walley & Metsala, 1990). This pattern of performance has been interpreted as indicating that adults pay particular attention to the first one or two phonemic segments in a word during recognition and lexical access; less attention is paid to the middles and ends of words, because the identity of segments in these positions is highly constrained by word-initial information (for additional references, see Marslen-Wilson, 1989; Walley, 1993b).
According to our model, however, lexical representation and processing are more holistic in nature for young children versus older children and adults. This is primarily because overall vocabulary growth in early through middle childhood is still substantial, and the familiarity and neighborhood similarity relations among individual words in the young child’s lexicon are very dynamic. Therefore, word beginnings are not very predictive and young children must pay more attention to information throughout a word in order to identify it.
In support, children’s mispronunciation identifications are less influenced by the position in a word of an error than adult’s performance, especially when words are presented in isolation (e.g., Cole & Perfetti, 1980; Walley, 1987; Walley & Metsala, 1990). Walley (1987), for example, assessed mispronunciation detection for isolated words presented either with or without a picture referent. When pictures were present, 5-year-olds showed greater sensitivity to word-initial than word-final mispronunciations, and greater sensitivity to word-initial mispronunciations than did 4-year-olds. No position effect was observed for 5-year-olds in the no-picture condition or for 4-year-olds in either condition (i.e., when lexical constraints were weaker). In a perceptual judgment task, Walley (1988) presented 5-year-olds and adults with isolated words in which white noise either replaced or was added to phonemes in the initial, medial, or final syllables. Adults rated words with either noise replacing or added to initial phonemes more poorly than words with noninitial disruptions, whereas children’s ratings were not influenced by the positional manipulation and were higher than adults’ for words with initial disruptions.
More recently, we examined children’s and adults’ detection of errors in words varying in age-of-acquisition or AOA (Walley & Metsala, 1990; see also Walley & Metsala, 1992). We used adults’ subjective estimates of AOA to select our stimuli, because previous research indicated that these estimates are reliable and valid. For example, when adults estimate that a word is learned at age 3, then 3-year-olds can correctly select a corresponding picture. We divided our stimuli into three AOA categories—”early,” “current,” and “late” vis-à-vis the age of our youngest, 5-year-old subjects (e.g., policeman, propeller, and pavilion). All words were of low frequency according to adult objective counts and were matched across AOA categories in terms of various structural characteristics (e.g., initial consonant); a similar mispronunciation was then constructed for items across AOA categories (e.g., /t/ for /p/ in the words mentioned previously).
Our subjects heard both mispronounced and intact versions of each word and indicated whether a given stimulus was said “right” or “wrong” by pointing to either a happy or a sad face. We found that 5- and 8-year-olds were as sensitive (as indexed by d′) to errors in early words as were adults. Eight-year-olds were as sensitive as adults to errors in current words, whereas 5-year-olds were less able to discriminate the mispronounced and intact versions of current and late words. Each of the age comparisons was significant for the late words. These findings indicate that recognition for young children varies as a function of word familiarity, with performance being most adultlike for early-acquired words.
Gating Paradigm. In the gating paradigm, listeners are presented with increasing amounts of speech input from word onset and asked to identify the target word after each “gated” trial. For example, the listener may hear the initial 100 msec of the word on the first trial, then the initial 150 msec, 200 msec, and so on for subsequent trials, until all of the word has been presented. Adults can identify many words on the basis of partial input, and recognize high-frequency words and words in context on the basis of less input than low-frequency words or words without context (Grosjean, 1980). In fact, high-frequency, one-syllable words in sentences can be recognized from as little as 150 msec of input, or less than half the speech signal corresponding to the entire word.
Several developmental studies employing this paradigm have shown that children aged 5 to 7 years need more speech input than do adult listeners to correctly recognize words (e.g., Elliott, Hammer, & Evan, 1987; Walley, 1988; Walley, Michela, & Wood, 1995). This age effect is found even for one-syllable words that are highly familiar to young children (Elliott et al., 1987). However, these are systematic improvements in recognition performance during childhood, presumably as a result of increases in vocabulary knowledge; thus, although first graders need more input than adults to recognize gated words, they need less input than kindergarteners (Walley et al., 1995).
More recently, Metsala (1997a) examined word recognition by 7-, 9-, 11-year-olds, and adults using the gating task as a function of word frequency and neighborhood density. Overall, the amount of speech input needed to recognize the test words decreased systematically with age. Of more interest however, age interacted with lexical characteristics. Specifically, the two youngest age groups needed more speech input to recognize low-frequency words and words from sparse neighborhoods than did the oldest group of children and adults. For high-frequency words from sparse neighborhoods, 7-year-olds needed more speech input than did 11-year-olds, and all three child groups needed more input than did the adults. There were no developmental differences in the recognition of high-frequency words from dense neighborhoods.
These findings support the first two claims of our model. First, changes in spoken word recognition continue into early childhood and are related to an individual word’s frequency and neighborhood structure. Even the youngest group of children recognized high-frequency words in dense neighborhoods on the basis of as little speech input as adults. Second, the observed frequency by neighborhood density interaction across age groups suggests that word frequency and neighborhood structure underlie or prompt segmental restructuring and processing of a lexical item. Specifically, high-frequency words are segmentally represented/processed and therefore recognized on the basis of minimal speech input, except when there are many online competitors (i.e., in dense neighborhoods). On the other hand, low-frequency words can be recognized on the basis of little input only if they reside in a dense neighborhood. Presumably these words have experienced pressure to undergo segmental restructuring to a greater extent or earlier than low-frequency words with few neighbors.
Previous adult studies have been inconsistent with respect to the effect of neighborhood density. In naming latency tasks, Luce (1986) found an advantage for all words in sparse versus dense neighborhoods (see also Goldinger, Luce, & Pisoni, 1989). According to his Neighborhood Activation Model, words from sparse neighborhoods have the least online competition and should therefore be easiest to recognize. This main effect of neighborhood density is inconsistent with Metsala’s (1997a) observed frequency by neighborhood density interaction. This interaction has, however, also been observed in adults’ lexical decision performance (Luce, 1986) and in studies of visual word naming (e.g., Andrews, 1992; Seidenberg & McClelland, 1989). The effects of these two factors on children’s spoken word recognition needs further examination (see Garlock et al., 1998).
Taken together, the developmental research suggests that even though children begin to be sensitive to segmental information with the onset of the vocabulary growth spurt, segmental restructuring continues into middle childhood and processing of spoken words remains relatively holistic into the early school years. This more holistic processing occurs despite the fact that the point at which a given word can be structurally differentiated from others on a left-to-right, segmental basis occurs earlier in the smaller lexicons of young children. In a study by Charles-Luce and Luce (1990), 5- and 7-year-olds’ production lexicons were found to consist predominantly of short, high-frequency words, the recognition of which might well be accomplished by global, as opposed to more segmental and thus capacity-demanding processing (cf. Dollaghan, 1994; Logan, 1992). However, the perceptual evidence we have reviewed indicates that by about 4 or 5 years of age, children are not completely unable to use partial speech input and do not need to hear all of a word in order to recognize it (consistent with an emerging ability to use segments or “chunks” of segments). In support of vocabulary expansion as the basis for a shift from global to more segmentally based processing, Charles-Luce and Luce found that as the size of the lexicon increases, so too does the number of similar neighbors that a given word has. There is, in turn, a trend toward denser similarity neighborhoods that goes hand in hand with developmental increases in the ability to organize and attend to the segmental and temporal properties of words.
THE LINKS BETWEEN LEXICAL RESTRUCTURING, PHONEMIC AWARENESS, AND READING ABILITY
The third claim of our LRM is that explicit segmentation ability depends on the emerging segmental structure of lexical representations. That is, phonemic awareness is limited by the very nature of the phonological representations being accessed and manipulated; if representations are not adequately segmented, it should not be possible for children to perform cognitive operations on individual phonemes. We therefore predict a close relationship between changing lexical representations and phonemic awareness, with the latter delayed relative to the former. As well, beginning phonemic awareness should depend on language development specifically, rather than on general metacognitive ability or reading experience.
The fourth claim of our model is that the phonological deficits in reading disabilities are due, in part, to atypical development in lexical restructuring. Next, we discuss the empirical support for these two claims of the model and thus attempt to integrate findings across largely disparate bodies of literature. In particular, we relate the model to empirical findings in the following areas: the relationship between language development and phonological awareness, and the link between implicit speech processing difficulties in reading-disabled children and lexical restructuring.
Language Development and Phonological Awareness
What does research reveal about the relationship between the development of basic speech and emerging phonological awareness skills? Only a few studies have examined this relationship empirically. One area of investigation has attempted to ascertain whether language and metalinguistic skills interact or develop autonomously (e.g., Smith & Tager-Flusberg, 1982; Webster & Plante, 1995). The logic is that if basic language ability and metalinguistic abilities are intertwined, then there should be strong interrelationships between measures of each—the interaction hypothesis. Alternatively, metalinguistic awareness might develop as a function of metacognitive abilities that arise in middle childhood and be independent of basic language ability—the autonomy hypothesis. These two hypotheses parallel the emergent versus accessibility positions outlined earlier, although the interaction–autonomy issue has been applied more broadly (e.g., to the domain of syntactic development).
Webster and Plante (1995) assessed the relationship between productive phonology and phonological awareness in a longitudinal study of 45 children over a 2- to 3-year period. Beginning at age 3½ children’s productive phonology was measured every 6 months using the Khan-Lewis Phonological Analysis test, which provides a measure of the intelligibility of a child’s speech. Importantly, children were selected to represent a wide range of articulatory abilities at study onset, yet performed in the normal range on tests of nonverbal intelligence, receptive language, and speech and hearing discrimination. Two metaphonological tasks, also administered every 6 months, consisted of rhyme and alliteration detection. The three main findings were: (a) children with underdeveloped productive phonologies were much less likely to perform at criterion on the two measures of phonological awareness; (b) increases over time on the production measure predicted significant increases in the likelihood of reaching criterion on the two metaphonological tasks—specifically, as production improved, there was an exponential growth in phonological awareness; and (c) change in rank on the phonological measure, rather than age, predicted increasing phonological awareness. These investigators concluded that “The development of primary phonology is a causal factor in the development of phonological awareness, but not the reverse” (p. 54).
Similar studies with young prereaders support the interaction hypothesis, namely that metaphonological skills develop as a function of basic phonological development. For example, Chaney (1994) examined the relationship among social class, language development, metalinguistic awareness, and emergent literacy in 4-year-old preschoolers (see also Chaney, 1992). She found that a standardized test of basic language development (the Preschool Language Scale—Revised) was most highly associated with performance on a composite test of phonological awareness (consisting of phoneme synthesis, rhyme detection, mispronunciation detection, and initial sound detection identification). Measures of social and family literacy accounted for unique variance beyond age and language ability in measures of overall print awareness, alphabetic concepts, and concepts about books. However, these latter variables did not predict additional variance in the phonological awareness composite, once age and language had been taken into account.
Chaney’s (1992, 1994) findings support a general link between basic language ability and phonological awareness. She concluded that “The young child’s success in solving metalinguistic problems is not so dependent on an ability to step back from meaning as on the state of linguistic knowledge of a particular linguistic structure …” (Chaney, 1992, p. 512). Webster and Plante’s (1995) findings concerning the relationship between productive phonology and phonological awareness support the more specific hypothesis of our LRM—namely, that there is a strong causal relationship between basic phonological development and the emergence of phonological awareness tasks within individual children. Further, Chaney (1992) found that 4-year-olds’ spoken vocabulary scores were significantly related to a composite phonological awareness score, which is also consistent with our model.
Metsala and Stanovich (1995) examined the relationship between overall vocabulary development, lexical representation (word-nonword status), and phonemic awareness. Sixty preschoolers (4- and 5-year-olds) performed better on two measures of phonemic awareness (isolating initial phonemes and onset-rime blending) for words than for nonwords. Furthermore, there were (age-partialed) correlations between a receptive vocabulary measure, as in Chaney’s (1992) sample of 4-year-olds, and in Bowey and Patel’s (1988) sample of 5- and 6-year-olds. These findings support the prediction of our model that rudimentary phonemic awareness develops prior to the onset of reading and provide preliminary evidence for a specific link between vocabulary knowledge and phonemic awareness. Vocabulary size and familiarity of individual lexical items both play the role in our model of propelling the segmental restructuring of lexical representations, and it is primarily this restructuring that mediates beginning phonemic awareness. These claims are supported by the finding that children with larger vocabularies performed better on phonemic awareness tasks and that performance on these tasks was better for familiar than unfamiliar items (cf. Bowey & Francis, 1991).
Metsala and Stanovich (1995) also examined the relationship among nonword repetition, vocabulary growth, and phonemic awareness. Previously, nonword repetition, especially for “unwordlike” nonwords, has been thought of as a measure of phonological short-term memory capacity (e.g., Gathercole, 1995; Gathercole & Baddeley, 1993; cf. Snowling, Chiat, & Hulme, 1991). Gathercole, in particular, proposed that phonological short-term memory influences vocabulary growth in young children. In contrast, Metsala and Stanovich suggested that it is the degree of segmental structure of lexical items that influences nonword repetition ability. The reasoning is that segmental representations are necessary for a nonword to be adequately represented in the first place. They maintained that correlations between nonword repetition accuracy and vocabulary size reflect this mediating influence of segmental representations. In support, they found that a measure of explicit segmental structure (i.e., phonemic awareness) had a unique relationship with vocabulary size, even after age and nonword repetition scores were partialed. However, nonword repetition did not predict unique variance in vocabulary, after age and a phonemic awareness measure were partialed. These findings are consistent with our proposal that extent of lexical restructuring, or segmental representation, influences implicit phonological tasks (such as nonword repetition) and is related to the development of a child’s vocabulary.
Bowey and Patel (1988) assessed two hypotheses that are relevant to these issues. By the first, independent metalinguistic hypothesis, metalinguistic ability presupposes cognitive control that does not develop until middle childhood. By the second, general language development hypothesis, metalinguistic skills are highly correlated with basic spoken language development, more so than with any general cognitive factor. Both phonological and syntactic awareness tests were employed as metalinguistic measures, word reading and reading comprehension tests as reading measures. Bowey and Patel found that metalinguistic awareness did not account for additional unique variance in word reading or reading comprehension after basic language development was entered into the regression analysis. Similarly, basic language development did not account for variance in beginning word reading or comprehension after variance due to metalinguistic awareness was controlled. It was therefore concluded that although metalinguistic awareness does contribute to reading development, such awareness is dependent on basic language development.
In summary, children as young as 3 to 4 years of age can perform many tasks that are traditionally used to measure phonological awareness. This is a powerful argument against the view that metalinguistic skills develop solely as a function of the ability to reflect upon the structure of language, independent of its meaning, only after approximately 5 or 6 years of age. Instead, there is evidence for a strong relation between early vocabulary development and the emergence of phonological awareness ability. These findings are consistent with our LRM, but certainly additional research is needed to strengthen its viability. To this end, we are, for example, currently examining performance over the preliteracy and early literacy periods on both implicit spoken word recognition tasks and explicit phonemic awareness tasks for words that vary in age of acquisition, frequency of occurrence, and neighborhood structure (Garlock et al., 1998). Such studies will help us better understand the emergence of phonemic awareness and early reading success (see also McBride-Chang, 1995).
The Lexical Restructuring Model and Reading Disabilities
Reading-disabled children display phonological processing deficits for diverse tasks that do not require explicit access to or manipulation of phonemes (for review see Siegel, chap. 6, this volume). We believe that examining such findings from a developmental perspective may help to explain the implicit or phonological core deficits in reading disabilities.
Phonological deficits in reading-disabled children have variously been attributed to less robust, underspecified, or degraded phonological representations. We propose that segmental restructuring of lexical items does not progress in a developmentally appropriate manner for such reading-disabled children. Many of the deficits of reading-disabled children do, in fact, parallel spoken word recognition in younger children, for whom more holistic processing of speech input appears to be the norm. First, reading-disabled children rely on lexical context to recognize ambiguous word onsets to a greater degree than do nondisabled readers (Reed, 1989; Snowling, Goulandris, Bowlby, & Howell, 1986). Similarly, 3- to 5-year-olds’ production and perception of syllables and individual phonemes are more influenced by contextual factors than are those of 7-year-olds and adults (see Nittrouer, 1992). Second, younger children require more speech input than do older children and adults to recognize spoken words (e.g., Elliott et al., 1987; Walley et al., 1995), as do reading-disabled children (Metsala, 1997b). Third, young children’s recognition performance differs most from older children’s and adults’ for words that are least familiar (e.g., Metsala, 1997a; Walley & Metsala, 1990, 1992). Similarly, reading-disabled subjects perform most poorly when stimuli are unfamiliar, multisyllabic, or degraded (e.g., Brady, Shankweiler, & Mann, 1983). Finally, reading-disabled children (e.g., De Weirdt, 1988; Godfrey, Syrdal-Lasky, Millay, & Knox, 1981) and young children (e.g., Walley et al., 1998) both display shallower slopes in their identification functions for phonetic categories, indicating that these categories or segments are more broadly tuned than they are for normal readers and older listeners.
In order to adduce more definitive support for the fourth claim of our LRM, Metsala (1997b) tested whether spoken word recognition differed between reading-disabled and normally achieving children as a function of word frequency and neighborhood density. On a speech gating task, reading-disabled children in Grades 1 through 6 needed more input from word-onset to recognize words with few similar sounding neighbors than did same-age peers. For the youngest children in the sample, the amount of speech input needed predicted individual differences in word and pseudoword reading beyond the variance accounted for by overall vocabulary size and phonemic awareness measures. Young reading-disabled children displayed this variation in lexical processing even when matched for vocabulary size with nondisabled peers.
Metsala’s (1997b) observation that individual differences on the gating task were predictive of young children’s reading, but not older children’s, supports two conclusions. First, there appears to be a direct link between spoken word recognition and reading performance. This conclusion is supported by recent connectionist models of the reading process, which demonstrate that inadequately specified phonological representations impair generalization of the alphabetic code (i.e., the statistical regularities between phonemes and graphemes) to novel word or pseudoword input—precisely the same problem experienced by disabled readers (e.g., Brown, 1997; Metsala & Brown, 1998). Second, the speech–reading relationship may be developmentally limited (see Stanovich, 1986). That is, segmental representations and processing of speech input may affect developmentally early attempts to decipher the alphabetic code for reading. Later word decoding in reading may reflect this early experience, more than continuing deficits in underlying cognitive/perceptual abilities. Certainly, more research is needed to substantiate the causal relationship between segmental representations in speech and reading disabilities (see also, Elbro, 1996; Elbro, Nielsen, & Peterson, 1994; Snowling et al., 1991).
CONCLUSIONS
We have outlined two very different positions regarding the developmental origins of the phonemic segment. By the accessibility position, the phoneme is functional in early infancy as a unit of speech perception, but is inaccessible for conscious manipulation until before substantial reading experience/exposure. By the emergent position, phonemes are not preformed perceptual units and only develop gradually over childhood. As we have shown, there is a growing body of evidence to support this position and our LRM more specifically. According to this model, lexical representations become more segmental in early through middle childhood primarily as a result of spoken vocabulary growth (i.e., changes in absolute vocabulary size, familiarity of individual lexical items, and neighborhood structures). Thus, the development of phonemic awareness abilities is limited by the very nature of those speech representations being accessed and manipulated. At the same time, then, our model provides a framework for understanding the phonological processing deficits in reading disabilities, although clearly further empirical support is needed in this area and other respects.
We also view our model as having important implications for future applied research on phonemic awareness, beginning reading, and reading disabilities. The programs described in this volume that impact phonemic awareness and beginning reading and spelling may be the very programs that impact lexical restructuring. One important implication of our model, then, is that early attention to vocabulary growth may be a key component in the early detection of children at risk for reading difficulties. The stability of phonological processing through the preschool years and onward (see Torgesen & Burgess, chap. 7, this volume) should serve as a strong impetus for finding earlier interventions for promoting successful reading acquisition by more children. With a better understanding of lexical restructuring in childhood, earlier identification and intervention strategies may be found.
ACKNOWLEDGMENTS
The authors wish to thank Rebecca Treiman for her comments on an earlier version of this manuscript, and Vicki Garlock for her input into our formulation of these research issues. Support for the preparation of this chapter was provided by a grant from the National Reading Research Center to the first author, and by grant HD30398 from the National Institute for Child Health and Human Development to the second author. Inquiries can be sent to either author at the following addresses: Amanda C. Walley, Department of Psychology, University of Alabama at Birmingham, UAB Station, Birmingham, AL 35294-1170 (e-mail: [email protected]); Jamie L. Metsala, Department of Human Development, University of Maryland, College Park, MD 20742 (e-mail: [email protected]).
REFERENCES
Andrews, S. (1992). Frequency and neighborhood effects on lexical access: Lexical similarity of orthographic redundancy? Journal of Experimental Psychology: Learning, Memory, and Cognition, V18(2), 234–254.
Anglin, J. M. (1989). Vocabulary growth and the knowing-learning distinction. Reading Canada, 7, 142–146.
Aslin, R. N., Pisoni, D. B., & Jusczyk, P. W. (1983). Auditory development and speech perception in infancy. In M. M. Haith & J. J. Campos (Eds.), Carmichael’s manual of child psychology: Vol. II. Infancy and the biology of development (pp. 573–687). New York: Wiley.
Aslin, R. N., & Smith, L. B. (1988). Perceptual development. Annual Review of Psychology, 39, 435–473.
Bard, E. G., Shillock, R. C., & Altmann, G. T. M. (1988). The recognition of words after their acoustic offsets in spontaneous speech: Effects of subsequent context. Perception & Psychophysics, 44, 395–408.
Barton, D. (1980). Phonemic perception in children. In G. H. Yeni-Komshian, J. F. Kavanagh, & C. A. Ferguson (Eds.), Child phonology: Vol. 2, perception (pp. 97–116). New York: Academic.
Benedict, H. (1979). Early lexical development: Comprehension and production. Journal of Child Language, 6, 183–200.
Bertoncini, J., Bijeljac-Babic, R., Jusczyk, P. W., Kennedy, L. J., & Meher, J. (1988). An investigation of young infants’ perceptual representations of speech sounds. Journal of Experimental Psychology: General, 117, 21–33.
Bowey, J. A., & Francis, J. (1991). Phonological analysis as a function of age and exposure to reading instruction. Applied Psycholinguistics, 12, 91–121.
Bowey, J. A., & Patel (1988). Metalinguistic ability and early reading achievement. Applied Psycholinguistics, 9, 367–383.
Brady, S., Shankweiler, D., & Mann, V. (1983). Speech perception and memory coding in relation to reading ability. Journal of Experimental Child Psychology, 35, 345–367.
Brown, G. D. A. (1997). Developmental and acquired dyslexia: A connectionist comparison. Brain and Language, 59, 207–235.
Brown, G. D. A., & Watson, F. L. (1987). First in, first out: Word learning age and spoken word frequency as predictors of word familarity and word naming latency. Memory & Cognition, 15, 208–216.
Burdick, C. K., & Miller, J. D. (1975). Speech perception by the chinchilla: Discrimination of sustained /a/ and /i/. Journal of the Acoustical Society of America, 58, 415–427.
Burnham, D. K., Earnshaw, L. J., & Clark, J. E. (1991). Development of categorical identification of native and non-native bilabial stops: Infants, children and adults. Journal of Child Language, 18, 231–260.
Byrne, B., & Fielding-Barnsley, R. (1993). Recognition of phoneme invariance by beginning readers: Confounding effects of global similarity. Reading and Writing, 6, 315–324.
Chaney, C. (1992). Language development, metalinguistic skills, and print awareness in three-year-old children. Applied Psycholinguistics, 13, 485–514.
Chaney, C. (1994). Language development, metalinguistic awareness, and emergent literacy skills of 3-year-old children in relation to social class. Applied Psycholinguistics, 15(3), 371–394.
Charles-Luce, J., & Luce, P. A. (1990). Similarity neighborhoods of words in young children’s lexicon. Journal of Child Language, 17, 205–215.
Cole, R. A., & Perfetti, C. A. (1980). Listening for mispronunciations in a children’s story: The use of context by children and adults. Journal of Verbal Learning and Verbal Behavior, 19, 297–315.
Cooper, R. P., & Aslin, R. N. (1989). The language environment of the young infant: Implications for early perceptual development. Canadian Journal of Psychology, 43, 247–265.
Cooper, R. P., & Aslin, R. N. (1990). Preference for infant-directed speech in the first month after birth. Child Development, 61, 1584–1595.
Cutler, A. (1995). Spoken word recognition and production. In J. L. Miller & P. D. Eimas (Eds.), Speech, language, and communication (pp. 97–136). San Diego, CA: Academic.
Cutler, A., Mehler, J., Norris, D., & Segui, J. (1989). Limits of bilingualism. Nature, 340, 229–230.
De Weirdt, W. (1988). Speech perception and frequency discrimination for good and poor readers. Applied Psycholinguistics, 9, 163–183.
Dollaghan, C. A. (1994). Children’s phonological neighborhoods: Half empty or half full? Journal of Child Language, 21, 257–271.
Edwards, M. L. (1974). Perception and production in child phonology: The testing of four hypotheses. Journal of Child Language, 1, 205–219.
Ehri, L. C. (1984). How orthography alters spoken language competencies in children learning to read and spell. In J. Downing & R. Valtin (Eds.), Language awareness and learning to read (pp. 119–147). New York: Springer-Verlag.
Eimas, P. D., Siqueland, E. R., Jusczyk, P. W., & Vigorito, J. (1971). Speech perception in early infancy. Science, 171, 304–306.
Elbro, C. (1996). Early linguistic abilities and reading development: A review and a hypothesis. Reading and Writing: An Interdisciplinary Journal, 8, 453–485.
Elbro, C., Nielsen, I., & Peterson, D. K. (1994). Dyslexia in adults: Evidence for deficits in non-word reading and in the phonological representation of lexical items. Annals of Dyslexia, 44, 205–226.
Elliott, L. L., Hammer, M. A., & Evan, K. E. (1987). Perception of gated, highly familiar spoken monosyllabic nouns by children, teenagers, and older adults. Perception & Psychophysics, 42, 150–157.
Ferguson, C. A. (1986). Discovering sound units and constructing sound systems: It’s child’s play. In J. S. Perkell, & D. H. Klatt (Eds.), Invariance and variability in speech processes (pp. 36–51). Hillsdale, NJ: Lawrence Erlbaum Associates.
Ferguson, C. A., & Farwell, C. B. (1975). Words and sounds in early language acquisition. Language, 51, 419–439.
Flege, J. E., & Eefting, W. (1986). Linguistic and developmental effects on the production and perception of stop consonants. Phonetica, 43, 155–171.
Fowler, A. E. (1991). How early phonological development might set the stage for phonological awareness. In S. Brady & D. Shankweiler (Eds.), Phonological processes in literacy: A tribute to Isabelle Y. Liberman (pp. 97–117). Hillsdale, NJ: Lawrence Erlbaum Associates.
Fowler, C. A. (1995). Speech production. In J. L. Miller & P. D. Eimas (Eds.), Speech, language, and communication (pp. 29–61). San Diego, CA: Academic.
Garlock, V. M., Walley, A. C., Randazza, L. A., & Metsala, J. L. (1998). Effects of word frequency, age-of-acquisition and neighborhood density on spoken word recognition and phoneme awareness in children and adults. Manuscript in preparation.
Garnica, O. K. (1973). The development of phonemic perception. In T. E. Moore (Ed.), Cognitive development and the acquisition of language (pp. 215–222). New York: Academic.
Gathercole, S. E. (1995). Is nonword repetition a test of phonological memory or long-term knowledge? It all depends on the nonwords. Memory & Cognition, 23(1), 83–94.
Gathercole, S. E., & Baddeley, A. D. (1993). Working memory and language. Hove, England: Lawrence Erlbaum Associates.
Gerken, L. (1994). Sentential processes in early childhood language: Evidence from the perception and production of function morphemes. In J. C. Goodman & H. C. Nusbaum (Eds.), The development of speech perception (pp. 271–298). Cambridge, MA: MIT Press.
Gleitman, L. R., & Rozin, P. (1977). The structure and acquisition of reading I. Relations between orthographies and the structure of language. In A. S. Reber, & D. L. Scarborough (Eds.), Toward a psychology of reading. Hillsdale, NJ: Lawrence Erlbaum Associates.
Godfrey, J. J., Syrdal-Lasky, A. K., Millay, K. K., & Knox, C. M. (1981). Performance of dyslexic children on speech perception tests. Journal of Experimental Child Psychology, 32, 401–24.
Goldinger, S. D., Luce, P. A., & Pisoni, D. B. (1989). Priming lexical neighbors of spoken words: Effects of competition and inhibition. Journal of Memory and Language, 28, 501–518.
Grosjean, F. (1980). Spoken word recognition processes and the gating paradigm. Perception & Psychophysics, 28, 267–283.
Hirsh-Pasek, K., Kemler Nelson, D. G., Jusczyk, P. W., Wright-Cassidy, K., Druss, B., & Kennedy, L. (1987). Clauses are perceptual units for young infants. Cognition, 26, 269–286.
Huttenlocher, J. (1984). Word recognition and word production in children. In H. Bouma & D. G. Bouwhuis (Eds.), Attention and performance: Control of language processes (pp. 447–456). Hillsdale, NJ: Lawrence Erlbaum Associates.
Ingram, D. (1979). Phonological patterns in the speech of young children. In P. Fletcher & M. Garman (Eds.), Language acquisition (pp. 133–148). Cambridge, England: Cambridge University Press.
Ingram, D. (1985). On children’s homonyms. Journal of Child Language, 12, 671–680.
Jakobson, (1968). Child language, aphasia, and phonological universals (A. R. Keiler, Trans.). The Hague, Netherlands: Mouton.
Jusczyk, P. W. (1986). Toward a model of the development of speech perception. In J. S. Perkell & D. H. Klatt (Eds.), Invariance and variability in speech processes (pp. 1–33). Hillsdale, NJ: Lawrence Erlbaum Associates.
Jusczyk, P. W. (1992). Developing phonological categories from the speech signal. In C. A. Ferguson, L. Menn, & C. Stoel-Gammon (Eds.), Phonological development: Models, research, implication (pp. 17–64). Parkton, MD: York.
Jusczyk, P. W. (1993). From general to language-specific capacities: The WRAPSA model of how speech perception develops. Journal of Phonetics, 21, 3–28.
Jusczyk, P. W. (1995). Language acquisition: Speech sounds and the beginning of phonology. In J. L. Miller & P. D. Eimas (Eds.), Speech, language, and communication (pp. 263–301). San Diego, CA: Academic.
Jusczyk, P. W., & Aslin, R. N. (1995). Infants’ detection of the sound patterns of words in fluent speech. Cognitive Psychology, 29, 1–23.
Jusczyk, P. W., Friederici, A. D., Wessels, J., Svenkerund, V. Y., & Jusczyk, A. M. (1993). Infants’ sensitivity to the sound patterns of native language words. Journal of Memory and Language, 32, 402–420.
Jusczyk, P. W., Hirsh-Pasek, K., Kemler Nelson, D. G., Kennedy, L. J., Woodward, A., & Piwoz, J. (1992). Perception of acoustic correlates of major phrasal units by young infants. Cognitive Psychology, 24, 252–293.
Jusczyk, P. W., Luce, P. A., & Charles-Luce, J. (1994). Infants’ sensitivity to phonotactic patterns in the native language. Journal of Memory and Language, 33, 630–645.
Jusczyk, P. W., Pisoni, D. B., Walley, A., & Murray, J. (1980). Discrimination of relative onset time of two-component tones by infants. Journal of the Acoustical Society of America, 67, 262–270.
Kuhl, P. K. (1993). Infant speech perception: A window on psycholinguistic development. International Journal of Psycholinguistics, 9, 33–56.
Kuhl, P. K., & Padden, D. (1982). Enhanced discriminability at the phonetic boundaries for the voicing feature in macaques. Perception & Psychophysics, 32, 542–550.
Kuhl, P. K., Williams, K. A., Lacerda, F., Stevens, K. N., & Lindblom, B. (1992). Linguistic experience alters phonetic perception in infants by 6 months of age. Science, 255, 606–608.
Liberman, I. Y., Shankweiler, D., & Liberman, A. M. (1989). The alphabetic principle and learning to read. In D. Shankweiler, & I. Y. Liberman (Eds.), Phonology and reading disability: Solving the reading puzzle (pp. 1–33). Ann Arbor: University of Michigan Press.
Liberman, I. Y., Shankweiler, D., Liberman, A. M., Fowler, C. A., & Fischer, F. W. (1977). Phonetic segmentation and recoding in the beginning reader. In A. S. Reber & D. L. Scarborough (Eds.), Toward a psychology of reading. (pp.207–225). Hillsdale, NJ: Lawrence Erlbaum Associates.
Lindblom, B. (1992). Phonological units as adaptive emergents of lexical development. In C. A. Ferguson, L. Menn, & C. Stoel-Gammon (Eds.), Phonological development: Models, research, implications (pp. 131–163). Timonium, MD: York.
Lindblom, B., MacNeilage, P., & Studdert-Kennedy, M. (1983). Self-organizing processes and the explanation of phonological universals. In B. Butterworth, B. Comrie, & O. Dahl (Eds.), Explanations of linguistic universals. The Hague, Netherlands: Mouton.
Locke, J. L. (1988). The sound shape of early lexical representations. In M. D. Smith & J. L. Locke (Eds.), The emergent lexicon: The child’s development of a linguistic vocabulary (pp. 3–18). New York: Academic.
Logan, J. S. (1992). A computational analysis of young children’s lexicons. Research on spoken language processing, technical report no. 8. Bloomington, IN: Department of Psychology, Speech Research Laboratory.
Luce, P. (1986). Neighborhoods of words in the mental lexicon. Research on speech perception, technical report no. 6. Bloomington, IN: Department of Psychology, Speech Research Laboratory.
Marslen-Wilson, W. D. (1987). Functional parallelism in spoken word recognition. Cognition, 25, 71–102.
Marslen-Wilson, W. D. (1989). Access and integration: Projecting sound onto meaning. In W. D. Marslen-Wilson (Ed.), Lexical representation and process (pp. 3–24). Cambridge, MA: MIT Press.
McBride-Chang, C. (1995). Phonological processing, speech perception, and reading disability. Educational Psychologist, 30(3), 109–121.
Menn, L. (1983). Development of articulatory, phonetic and phonological capabilities. In J. B. Gleason (Ed.), The development of language (pp. 61–98). Columbus, OH: Merrill.
Menn, L., & Matthei, E. (1992). The “two-lexicon” account of child phonology: Looking back, looking ahead. In C. A. Ferguson, L. Menn, & C. Stoel-Gammon (Eds.), Phonological development: Models, research, implications (pp. 211–247). Timonium, MD: York.
Menyuk, P., & Menn, L. (1979). Early strategies for the perception and production of words and sounds. In P. Fletcher, & M. Garman (Eds.), Language acquisition (pp. 49–70). Cambridge, England: Cambridge University Press.
Metsala, J. L. (1997a). An examination of word frequency and neighborhood density in the development of spoken-word recognition. Memory & Cognition, 25(1), 47–56.
Metsala, J. L. (1997b). Spoken word recognition in reading disabled children. Journal of Educational Psychology, 89(1), 159–169.
Metsala, J. L., & Brown, G. D. A. (1998). Normal and dyslexic reading development: The role of formal models. In C. Hulme & R. M. Joshi (Eds.), Reading and spelling: Development and disorder (pp. 235–261). Hillsdale, NJ: Lawrence Erlbaum Associates.
Metsala, J. L., & Stanovich, K. E. (1995, April). An examination of young children’s phonological processing as a function of lexical development. Paper presented at the Annual American Educational Research Association, San Francisco, California.
Miller, J. L., & Eimas, P. D. (1994). Observations on speech perception, its development, and the search for a mechanism. In J. C. Goodman & H. C. Nusbaum (Eds.), The development of speech perception (pp. 271–298). Cambridge, MA: MIT Press.
Morais, J., Bertelson, P., Cary., L., & Alegria, J. (1986). Literacy training and speech segmentation. Cognition, 24, 45–64.
Morais, J., Cary, L., Alegria, J., & Bertelson, P. (1979). Does awareness of speech as a sequence of phones arise spontaneously? Cognition, 7, 323–331.
Morrison, C. M., & Ellis, A. W. (1995). Roles of word frequency and age of acquisition in word naming and lexical decision. Journal of Experimental Psychology: Learning, Memory, and Cognition, 21, 116–133.
Moskowitz, A. I. (1973). The acquisition of phonology and syntax. In K. K. J. Hintikka, J. M. E. Moravsik, & P. Suppes (Eds.), Approaches to natural language (pp. 48–84). Dordrect, The Netherlands: Reidel.
Nittrouer, S. (1992). Age-related differences in percepetual effects of formant transitions within syllables and across syllable boundaries. Journal of Phonetics, 20, 1–32.
Nittrouer, S., & Studdert-Kennedy, M. (1987). The role of coarticulatory effects in the perception of fricatives by children and adults. Journal of Speech and Hearing Research, 30, 319–329.
Nittrouer, S., Studdert-Kennedy, M., & McGowan, R. S. (1989). The emergence of phonetic segments: Evidence form the spectral structure of fricative-vowel syllables spoken by children and adults. Journal of Speech and Hearing Research, 32, 120–132.
Nygaard, L. C., & Pisoni, D. B. (1995). Speech perception: New directions in theory and research. In J. L. Miller & P. D. Eimas (Eds.), Speech, language, and communication (pp. 63–93). San Diego, CA: Academic.
Pegg, J. E., Werker, J. F., & McLeod, P. J. (1992). Preference for infant-directed over adult-directed speech: Evidence from 7-week-old infants. Infant Behavior and Development, 15, 325–345.
Peters, A. M. (1983). The units of language. Cambridge, England: Cambridge University Press.
Pisoni, D. B., Carrell, T. D., & Gans, S. J. (1983). Perception of the duration of rapid spectrum changes in speech and nonspeech signals. Perception & Psychophysics, 34, 314–322.
Pisoni, D., & Luce, P. (1987). Acoustic-phonetic representations in word recognition. Cognition, 25, 21–52.
Reed, M. A. (1989). Speech perception and the discrimination of brief auditory cues in reading disabled children. Journal of Experimental Child Psychology, 48, 270–292.
Reznick, J. S., & Goldfield, B. A. (1992). Rapid change in lexical development in comprehension and production. Developmental Psychology, 28, 406–413.
Rozin, P. (1976). The evolution of intelligence and access to the cognitive unconscious. In J. Sprague & A. N. Epstein (Eds.), Progress in psychobiology and physiological psychology (Vol. 6, pp. 245–280). New York: Academic.
Schwartz, R. G. (1988). Phonological factors in early lexical development. In M. D. Smith & J. L. Locke (Eds.), The emergent lexicon: The child’s development of a linguistic vocabulary (pp. 185–218). San Diego, CA: Academic.
Seidenberg, M. S., & McClelland, J. L. (1989). A distributed, developmental model of word recognition and naming. Psychological Review, 96(4), 523–568.
Share, D., & Stanovich, K. E. (1995). Cognitive processes in early reading development: Accommodating individual differences into a model of acquisition. In J. S. Carlson (Ed.), Issues in education: Contributions from psychology (Vol. 1, pp. 1–57). Greenwich, CT: JAI.
Shvachkin, N. K. (1973). The development of phonemic speech perception in early childhood. In C. A. Ferguson & D. I. Slobin (Eds.), Studies of child language development (pp. 91–127). New York: Holt, Rinehart & Winston.
Smith, C. L., & Tager-Flusberg, H. (1982). Metalinguistic awareness and language development. Journal of Experimental Child Psychology, 34, 449–468.
Snowling, M. J., Chiat, S., & Hulme, C. (1991). Words, nonwords and phonological processes: Some comments on Gathercole, Willis, Emslie, and Baddeley. Applied Psycholinguistics, 12, 369–373.
Snowling, M. J., Goulandris, N., Bowlby, M., & Howell, P. (1986). Segmentation and speech perception in relation to reading skill: A developmental analysis. Journal of Experimental Child Psychology, 41, 489–507.
Stanovich, K. E. (1986). Mathew effects in reading: Some consequences of individual differences in the acquisition of literacy. Reading Research Quarterly, 21(4), 360–407.
Stoel-Gammon, C. (1991). Normal and disordered phonology in two-year-olds. Topics in Language Disorders, 11, 21–32.
Strange, W., & Broen, P. A. (1981). The relationship between perception an production of/w/, /r/, and /l/. Journal of Experimental Child Psychology, 31(1), 81–102.
Studdert-Kennedy, M. (1986). Sources of variability in early speech development. In J. S. Perkell, & D. H. Klatt (Eds.), Invariance and variability in speech process (pp. 58–76). Hillsdale, NJ: Lawrence Erlbaum Associates.
Studdert-Kennedy, M. (1993). Discovering phonetic function. Journal of Phonetics, 21, 147–155.
Suomi, K. (1993). An outline of a developmental model of adult phonological organization and behavior. Journal of Phonetics, 21, 29–60.
Thomas, D., Campos, J. J., Shucard, D. W., Ramsay, D. S., & Shucard, J. (1981). Semantic comprehension in infancy: A signal detection approach. Child Development, 52, 798–803.
Treiman, R., & Baron, J. (1981). Segmental analysis ability: Development and relation to reading ability. In G. E. MacKinnon & T. G. Waller (Eds.), Reading research: Advances in theory and practice (Vol. 3, pp. 159–198). New York: Academic.
Treiman, R., & Breaux, A. M. (1982). Common phoneme and overall similarity relations among spoken word syllables: Their use by children and adults. Journal of Psycholinguistic Research, 11, 581–610.
Vihman, M. M. (1993). Variable paths to early word production. Journal of Phonetics, 21, 61–82.
Walley, A. C. (1987). Young children’s detections of word-initial and -final mispronunciations in constrained and unconstrained contexts. Cognitive Development, 2, 145–167.
Walley, A. C. (1988). Spoken word recognition by young children and adults. Cognitive Development, 3, 137–165.
Walley, A. C. (1993a). More developmental research is needed. Journal of Phonetics, 21, 171–176.
Walley, A. C. (1993b). The role of vocabulary development in children’s spoken word recognition and segmentation ability. Developmental Review, 13, 286–350.
Walley, A. C., Flege, J. E., & Randazza, L. A. (1997). Effects of lexical status on the perception of native and non-native vowels: A developmental study. Manuscript in preparation.
Walley, A. C., & Metsala, J. L. (1990). The growth of lexical constraints on spoken word recognition. Perception & Psychophysics, 47, 267–280.
Walley, A. C., & Metsala, J. L. (1992). Young children’s age-of-acquisition estimates for spoken words. Memory & Cognition, 20(2), 171–182.
Walley, A. C., Michela, V. L., & Wood, D. R. (1995). The gating paradigm: Effects of presentation format on spoken word recognition by children and adults. Perception & Psychophysics, 57(3), 343–351.
Walley, A. C., Smith, L. B., & Jusczyk, P. W. (1986). The role of phonemes and syllables in the perceived similarity of speech sounds for children. Memory & Cognition, 14, 220–229.
Webster, P. E., & Plante, A. S. (1995). Productive phonology and phonological awareness in preschool children. Applied Psycholinguistics, 16, 43–57.
Werker, J. F. (1991). The ontogeny of speech perception. In I. G. Mattingly & M. Studdert-Kennedy (Eds.), Modularity and the motor theory of speech perception. Hillsdale, NJ: Lawrence Erlbaum Associates.
Werker, J. F., & Pegg, J. E. (1992). Infant speech perception and phonological acquisition. In C. A. Ferguson, L. Menn, & C. Stoel-Gammon (Eds.), Phonological development: Models, research, implications (pp. 285–311). Timonium, MD: York.
Werker, J. F., & Tees, R. C. (1984). Cross-language speech perception: Evidence for perceptual reorganization during the first year of life. Infant Behavior and Development, 7, 49–63.
Zlatin, M. A., & Koenigsknecht, R. A. (1976). Development of the voicing contrast: A comparsion of voice onset time in stop perception and production. Journal of Speech and Hearing Research, 19, 78–92.
__________
1More specifically, the prereader’s difficulty in explicitly attending to phonemes has been attributed to the contextually dependent manner in which these units (more so than syllables, e.g.) are encoded in the speech waveform (e.g., Gleitman & Rozin, 1977; Liberman, Shankweiler, Liberman, Fowler, & Fischer, 1977). However, this same factor may also underlie more fundamental developmental differences in both speech perception and production (e.g., Nittrouer & Studdert-Kennedy, 1987; Nittrouer et al., 1989).
2Although there is considerable evidence that adult’s lexical representations are segmentally structured (see Pisoni & Luce, 1987), this view is not uncontroversial (see walley, 1993b).
3This study employed a single picture presentation format, rather than the four-alternative format used by Thomas et al. (1981); however, the phonetic similarity of targets and foils was not systematically varied in the earlier study—a difference that could account for the discrepancy in the age estimates obtained.
4The emergence of segments would also promote further vocabulary growth, in that a limited pool of highly productive elements with recombinatorial power would support more efficient storage of lexical items in memory (e.g., Lindblom, 1992).