1 Second Language Acquisition of Suprasegmental Phonology Lesley Carmichael 1.0 Introduction New questions need to be asked in second language acquisition (SLA) research: Can second language learners have differential success in acquiring the segmental vs. suprasegmental phonology of a second language (L2)? That is, are segmental and suprasegmental phonology independent aspects of phonological acquisition? And if so, are they necessarily constrained according to the same age-based schedule of acquisition? In this paper, I draw from the SLA literature to demonstrate how work that has already been done indicates a natural division between segmental and suprasegmental phonology in terms of SLA. I propose that second language learners do tend to achieve differential pronunciation success of segmental and suprasegmental components of speech as a factor of the age-related characteristics of their SLA experience. Specifically, suprasegmental acquisition success in an L2 may require an earlier start in life than segmental acquisition success. The decline of an L2 learner’s ability to acquire nativelike intonation appears to begin at a very young age, probably before the decline of the ability to achieve nativelike segmental phonology in an L2. 1.1 Outline of discussion A review of SLA phonology research and the generally accepted separation of phonological and morphosyntactic acquisition lay the foundation for the paper. A critique of SLA research on the acquisition of phonology will present the tendency of SLA 1 2 research to evaluate segmental and suprasegmental phonology together as a single system, and will motivate a natural subdivision between them. I will discuss the importance of intonation in (the perception of) foreign-accented speech, substantiating the need to investigate the acquisition of intonation as an independent linguistic system. Intonation is assessed in terms of its role in the discrimination of languages and its role in assisting the development of other linguistic subsystems. A look at first language acquisition suggests that suprasegmental phonology begins developing earlier than segmental phonology in the development of first language (L1) competence. Combined, these evaluations of L1 and L2 acquisition research suggest that the segmental and suprasegmental components of L2 phonology are subject to different developmental constraints. A new, finer-grained research agenda is proposed to properly characterize the acquisition of L2 phonology and pronunciation. This agenda includes new approaches designed to isolate the respective contributions of segmental and suprasegmental production to foreign accent. The agenda also proposes independent mappings of segmental and suprasegmental production data to age of acquisition data in order to assess whether segmental and suprasegmental phonology have different maturational schedules. 1.2 Terminology The terms suprasegmental phonology, intonation, and prosody are used with various intentions and interpretations across linguistic subdisciplines. In this paper, the term suprasegmental phonology refers to phrase- and discourse-level pronunciation components such as the manipulation of the fundamental frequency of the voice, pausing, 2 3 and the timing of pitch events across phrases, sentences, and discourse units. I will use the term intonation to primarily indicate the manipulation of pitch, but also to broadly include features such as duration and loudness. (The term prosody is often used to refer to the collective arrangement of pitch, loudness, and duration features.) The distinction between intonation and prosody are not critical to this particular discussion of a differentiation between the L2 acquisition of segmental and suprasegmental phonology. It may be the case that a distinction between intonation and prosody is relevant in other discussions of the acquisition of SLA phonology, and will become more relevant as investigation into SLA continues. 2.0 Age of acquisition and second language success Lenneberg (1967) originally theorized that puberty marked the end of a biological critical period for language acquisition. Since then, the acquisition of individual linguistic subsystems (such as phonology and morphology) have been studied as well as language as a whole. The later in life an individual begins to learn a second language, the more elusive nativelike pronunciation of the L2 becomes. This observation is supported by a large body of L2 research correlating the presence and degree of foreign accent in speech with the starting age of L2 acquisition (e.g., Scovel, 1969, 1988; Birdsong, 1999; Flege et al.., 1995; Carmichael, 2000). Overall, this research seems to indicate that L2 acquisition at some point before complete biological maturation is critical for a learner to be able to attain a nativelike command of the pronunciation, morphology, and syntax of the L21. 1 There are attested cases of late L2 learners achieving a nativelike command of the L2. Bongaerts, Planken and Schils (1995) (and the 1997 replication by Bongaerts, van Summeren, Planken, and Schils) found that a small number of late second language learners could pass for native speakers phonologically. Scovel (1988) describes the Joseph Conrad effect, a phenomenon in which L2 learners can pass for natives in the areas of 3 4 2.1 Independent critical periods for SLA of phonology and syntax SLA research shows evidence for differential acquisition success of phonology and syntax relative to the starting point of SLA during a learner’s life span. There is general agreement among SLA researchers that L2 learners who begin learning an L2 before roughly puberty can achieve good syntax (e.g., Patkowski’s [1980] study of ultimate attainment of L2 syntax and age of arrival effects). Many researchers also maintain that the same general biological constraints (i.e., puberty) mark the end of one’s ability to acquire a nativelike pronunciation.2 For example, Patkowski (1990) found a discrete shift in pronunciation ability corresponding to an age of acquisition at about age 15, indicating the end of a critical period for phonological acquisition. Flege, Munro, and MacKay (1995) also proposed a discrete age-related constraint on phonological acquisition, but instead found a continuous decline in ratings of foreign accent with respect to L2 learners’ age of arrival in the L2 environment. Research comparing the perceived nativeness of (both) the grammar and pronunciation attained by L2 learners seems to indicate a natural split between these two aspects of language acquisition. Ioup (1984) found that native English listeners could classify native and nonnative English speakers according to their pronunciation but not by their use of syntax. She later concluded that near-native fluency of syntax is achieved more easily by adults than near-native pronunciation (1987). In Scovel’s (1988) perception study, native speaker judges syntax, morphology, and lexicon upon examination of writing samples. The evaluation of SLA in this paper is based on the prominent trend in SLA research indicating that accent-free speech is generally unattainable for learners beginning the SLA process later in life. 2 This paper is mainly concerned with the performance, or productive, aspect of intonation. It should be noted that infants demonstrate a perceptive sensitivity to intonation before they begin to produce intonation features of their native language (see section 6 for a discussion of intonation in L1, including infants’ perceptual abilities). A gap between the comprehension and production of both phonology and syntax is well attested (see the volume on competence and performance in SLA edited by Brown et al., 1996). 4 5 correctly identified nonnatives 97% of the time based on speech but only 47% of the time based on writing samples, suggesting that the spoken language contained stronger evidence of nonnativeness than the grammar did. He claimed that even highly advanced L2 users can be easily identified as nonnative speakers (p. 108). These differential acquisition outcomes have been related to differences in the acquisition experiences of the learners: specifically, the time of life when a second language phonology can be successfully achieved passes earlier than the time of life when a second language syntax can be successfully achieved. 2.2 Maturational effects on second language pronunciation While there is general agreement that people who acquire an L2 before puberty can attain a good, communicative command of syntax (not easily identified as nonnative), there is less accord regarding the relationship between age of acquisition and ultimate pronunciation outcomes. Some work indicates that while most adults retain an accent when they begin learning a second language after puberty, children who begin the process before puberty show little or no foreign accent—and the younger they begin, the better. Oyama (1976) found that L2 learners who were younger than 10 years of age when they began learning their L2 were rated as having the least accent. Tahta, Wood and Loewenthal (1981) observed a marked jump in ratings of foreign accentedness for L2 learners who began learning their L2 at age 12 as compared to those who began at age 11. In Seliger, Krashen and Ladefoged (1975), L2 learners were asked to report on their own degree of foreign accent in the languages they spoke. About 85% of those who 5 6 immigrated to the L2 environment before the age of 10 rated themselves as having no accent in either their first or second language, while about 90% of those who immigrated after the age of 15 rated themselves as having an accent in both languages. Those L2 learners who immigrated between the ages of 10-15 were split almost in half in their selfreports of accent vs. no accent in the L2. Snow and Hoefnagel-Hohle (1978) found that the teens in their study were learning Dutch pronunciation (also morphology and syntax) more quickly than the children were; however, Scovel (1988) points out in his review of research on L2 pronunciation acquisition that those early outcomes do not prove that the teens ultimately achieved greater pronunciation proficiency. Other researchers propose that SLA must begin before even 5-7 years of age to result in accent-free speech. Asher and Garcia (1969) found that 68% of the L2 learners who began learning the L2 before 7 years of age (n=19) were rated as being near-native. Of the L2 learners who began learning the L2 after 12 years of age (n=15), 93% were rated as having a foreign accent. The L2 learners who began learning the L2 between ages 7-12 (n=37) were divided between near-native (41%), slight accent (43%), and definite accent (16%). Fathman (1975) and Williams (1979) found that younger language students retained less accent than adolescents who were just past puberty. Furthermore, younger children have an eventual oral skill achievement advantage over older children. In Fathman (1975), older children (11- to 15-year-olds) initially showed significant advantages over younger children (6- to 10-year-olds) in acquiring second language English. Three years later, the younger children performed better than the older ones on the same pronunciation measures (and also on morphology and syntax). Other research 6 7 confirms this finding, with younger language learners eventually outperforming older learners in oral skills (e.g., Oyama, 1978). Krashen, Long, and Scarcella (1979) reviewed several studies on child and adult SLA and concluded that while older is better in terms of the rate of SLA, younger is better in terms of ultimate attainment. Scovel (1969) said that accent-free speech would be impossible for anyone who learned an L2 after the language faculty had lateralized in the brain. Evidence cited by Krashen (1973) suggested that the lateralization of language occurred around age 5, long before puberty. Krashen interpreted this to mean that lateralization could therefore have nothing to do with a critical period for language acquisition, nor could it present any kind of a biological barrier to the attainment of nativelike pronunciation. (It should be noted that at that time, a single critical period ending around puberty was a fairly new hypothesis.) SLA research since that time seems to not only require a softening of Krashen’s interpretation, but perhaps even a reversal of it. Scovel (1988) actually proposes some kind of “cutoff” for the ability to acquire nativelike L2 phonology at around age 6 or 7. Long (1990) supports this claim in his survey of the evidence for critical periods in language acquisition, saying that in order to attain authentic pronunciation, first exposure to the language must be before age 6 for many people, and by about age 12 for the rest (accounting for some exceptional language learners). The tenets of the Constructionist hypothesis of SLA (Herschensohn, 2000) support observations of early exposure followed by a progressive development of the L2 phonology, resulting in nativelike pronunciation. Constructionism contends that an initial period of feature underspecification is followed by a period in which L2 values are constructed by building 7 8 on other constructions. While presented mainly for the SLA of morphology and syntax, the principles of Constructionism may well account for the development of L2 pronunciation over time. The increasing granularity and sophistication of SLA research over the last 30 years is yielding growing support for a hypothesis of L2 phonological acquisition in which acquisition of the L2 must begin before about age 6 or 7. 3.0 The objects of phonological acquisition research When age-based constraints are proposed for L2 phonological acquisition, exactly what aspects of language are comprised by the term phonological? SLA research on phonological acquisition has focused predominantly on segmental and articulatory phonology, without particular regard to suprasegmental linguistic behavior (e.g., the construction of new perceptual categories for segments: Flege, 1987; Bohn and Flege, 1990; e.g., the production of new sounds: Lado, 1957; Flege, 1987; Bohn and Flege, 1992; Logan, Lively, and Pisoni, 1991) or on a general measure such as “overall foreign accent.” Recent pedagogically-driven work (see section 4.1) and work by Archibald (1998) on second language phonology are notable exceptions. (Archibald [1998] includes an investigation of the acquisition of L2 metrical structure, and offers a thorough account of the interlanguage occurring during SLA.) Various components of phonology interact in pronunciation, and studies of pronunciation acquisition have not considered all subdivisions of phonology as possibly independent contributors to pronunciation accuracy. The L2 acquisition of suprasegmental phonology has not been subjected to 8 9 much independent investigation; specifically, it has not been intentionally separated from other aspects of phonological acquisition. Further, the heretofore limited interest in suprasegmentals as an independent system within a language’s phonology has resulted in studies of segmental acquisition which do not necessarily consider the influence of suprasegmental features. Scovel (1988) pointed out in a criticism of Olsen and Samuels (1973) (in which teen L2 learners were found to have better pronunciation than children L2 learners) that since they did not consider intonation, the question of how “accented” the speakers sounded was really left unanswered. More recent work has suggested that intonation plays a very distinct role in terms of its contribution to a foreign accent (e.g., Magen, 1998, conducted a perception study using speech synthesized to contain purported features of accentedness; and Wennerstrom, 2001, analyzed intonation used by English as a Second Language speakers [see section 4.2]). Various speech-based technologies are also drawing attention to the complexity of linguistic intonation and its independent role in establishing pragmatic relationships and discourse structure (Wichmann, 2000) and synthesizing natural sounding speech (Goldsmith, 1999) (also see Cutler et al., 1997 for a review of prosody in lexical, sentence level, and discourse processing). These new applications indicate a need for new research on intonation as its own sophisticated linguistic system. 3.1 Factors measured in previous SLA phonology research Age of acquisition was one factor assessed in an investigation of voice onset time produced by Spanish speakers of English in a study by Flege et al.. (1998). In Riney, Flege and Flege (1998), native English listeners assessed “global foreign accent” change 9 10 and improvement in the production of English [l] and [] by Japanese students attending an American university over a period of four years. They found that the subjects reduced their overall foreign accent during their time in the university and improved in their ability to identify and produce the two liquid sounds. It is unclear exactly what factors the native English respondents were attending to when they made decisions about the global foreign accent they perceived in the Japanese students’ speech. In another study focusing on L2 learners’ ability to perceive nonnative segmental contrasts, Lively et al.. (1994) trained Japanese listeners to discriminate between English [l] and []. Follow-up testing three months later showed that they retained the ability to discriminate the sounds without any further training. They concluded that changes in perception occur from selective attention to the acoustic cues that signal phonetic contrasts. It is tempting to expand this general claim about developing perceptual discrimination to the discrimination of intonational contrasts. Similar research focusing on intonation can help determine whether selective attention also contributes to the perception of intonational contrasts in another language. Claims about critical periods for pronunciation or correlations between age and perceived degrees of foreign accent, as exemplified by the studies mentioned herein, are typically based on L2 learners’ ability to acquire the segmental articulation and phonology of the target language. Others are unspecified in terms of what it is about the speech that is causing listeners to perceive a foreign accent. (A recent exception is a study of the impact of speech rate on perceptions of foreign accent by Munro and Derwing, 2001.) Importantly, this means that most investigations of L2 phonological acquisition to date 10 11 not only have not explicitly considered intonation, they have not explicitly controlled for it either. Flege, Munro, and MacKay (1995) found that foreign accents were detected in the speech of Italian learners of English who began learning English before puberty; however, it was not ascertained what aspects of pronunciation caused the listeners to perceive a foreign accent. Piske, MacKay and Flege (2001) present an excellent review of work examining factors that have been claimed to affect the degree of perceived foreign accent in L2 (e.g., age of SLA, length of residence in the L2 environment, instruction, motivation, etc.). But in their study (presented in the same paper), Piske et al.. continue to use “overall degree of foreign accent” as the measurable outcome. This thorough work can thus only tell us how the factors influenced the amalgamated end product of segment production and intonation. This and many other studies dismiss the evaluation of suprasegmental variation as beyond the scope of their current work; crucially, this variation is not controlled for as a separate variable contributing to the resulting foreign accent. Results which correlate (segmental) pronunciation metrics with perceived foreign accent and age-related acquisition experience may actually reflect L2 learners’ use of intonation as well. 4.0 Intonation is a key component of foreign accent Theoretical linguistics has only engaged in a focused investigation of intonation since the late 1970s (see Cutler et al.., 1997 for a comprehensive literature review). In this time, it has also been acknowledged by language pedagogists and English as a Second Language (ESL) researchers as a critical subcategory of language—one which plays a significant 11 12 role both in the production and perception of language (e.g., James, 1976; Pennington and Richards, 1986; Nunan, 1999). 4.1 Intonation in communicative language teaching Nonnative intonation interferes with the accurate communication of meaning, and a mismatch between intonation and intention can even result in a complete contradiction of a speaker’s intended message (Nash, 1971). James (1976) stressed the importance of intonation as a critical factor in (second) language in his work using a speech visualizer to assist speakers in modeling speech intonation. The best results were achieved by speakers using an audiovisual representation of the target speech, including a pitch contour, along with a feedback display showing speakers’ imitations (better than results achieved with audio or audiovisual presentation but without a feedback display). He claimed that incorrect use of prosodic components can cause not only misunderstanding but a total breakdown in communication, and that his results suggested that intonation is more essential to the acceptability of second language speech than articulation. This suggestion has since been borne out in writings on second language pedagogy. As noted by Nunan in his pedagogical book Second Language Teaching and Learning (1999), improper use of stress, rhythm, and intonation present more difficulty for hearers than does poor articulation. The intentional and primary development of intonation in L2 learning is justified by Abberton et al. (1978), who point out that intonation provides the temporal foundation for the distribution and realization of segmental information. Without the organization 12 13 nativelike intonation provides, properly articulated segments are delivered in a nonnative relation to each other. As approaches to language teaching have moved toward communicative methods (see, for example, Lee and VanPatten, 1995), intonation is taking a more prominent role in the instruction of pronunciation. While L2 teaching methodologies have changed significantly and often over the years3, the advantage of prosodic naturalness to communicative efficacy is becoming recognized in L2 pedagogy (e.g., Nunan, 1999). Pennington and Richards (1986) explicitly state that the positioning of sounds and words in speech is a critical aspect of pronunciation, and they discourage the teaching of linguistic subunits in isolation. 4.2 Experimental assessments of L2 intonation These tendencies in teaching are corroborated by analytical work with L2 learners. Wennerstrom (2001) analyzed natural dialogues between native English speakers and ESL learners within the framework of a phonological model of intonation, the Tones and Break Indices (ToBI) model of standard varieties of English (see Silverman et al., 1992, for a description). She assessed learners’ use of pitch to disambiguate new and given information in a discourse and to signal turn taking. Learners who were rated by native English speakers as sounding more fluent were better able to manipulate pitch to indicate both turn taking and relationships among discourse elements. Wennerstrom concluded that proper use of intonational features is an essential aspect of fluency. In addition, she 3 A number of L2 teaching methodologies have been developed over the last 50 years, including Situational Language Teaching, Audiolingualism, Total Physical Response, and Communicative Language Teaching. Different methods prescribe different roles for teachers and students and varying degrees of interactivity. (Richards and Rogers, 2001) 13 14 asserted that intonation is clearly part of a grammatical system and not merely a stylistic aspect of language and communication. Wennerstrom’s study is part of a growing body of work indicating that intonation figures prominently in the perception of foreign accented speech. Pedagogically driven work, as indicated earlier, positions intonation as a critical foundation for good L2 pronunciation and communication. Experimental phonetics and phonology create exciting possibilities for assessing the contribution of intonation to the perception of foreign accent in a more controlled way. Magen (1998) independently manipulated various segmental and suprasegmental aspects of foreign accented speech through synthetic alteration and presented the resulting speech stimuli to native speakers for judgments of accentedness. In addition to adjusting segmental features such as vowel reduction, consonant aspiration, and vowel tensing on Spanish-accented utterances, she resynthesized the utterances with native English intonation contours. The results of her perception experiment (n=10 listeners; n=96 stimulus utterances) indicated that the fundamental frequency (f0) behavior was the most influential cue (more so than all segmental cues) to the perception of a foreign accent in the speech. The results of my research on the contribution of L1 segmental phonology to a foreign accent in an L2 (Carmichael, 2000) suggested that intonation is so strong a cue to foreign accent that nonnative use of prosodic features can override accurate segmental phonology to a native ear. In the study, it appeared that nonnative intonation on monosyllabic utterances overrode accurate segmental production when listeners were asked to rate the 14 15 speech in terms of the degree of foreign accent they perceived. Speech tokens with nearnative segmental characteristics but an intonation contour lacking the typical American English declarative intonation phrase features received average ratings of “moderately heavy” to “heavy” accent (3-4 on a 5-point scale [0=no accent; 4=heavy accent]; n=69 respondents; n=192 stimulus tokens). In contrast, tokens with similar segmental features but an English-like intonation contour received average ratings of “little” to “no” accent (0-1 on the same 0-4 scale). 4.3 Intonation as an independent factor in the perception of foreign accent The development of L2 intonation is undoubtedly part of phonological acquisition, as it is a spoken characteristic of language with its own rules and meaningful contribution to the language being spoken. Research is just beginning on the possible independence of intonation within the boundaries of phonological acquisition. It is unknown whether segmental and suprasegmental phonologies are necessarily constrained by the same maturational schedule (see section 6 for a discussion of intonation in L1). There has also been only limited discussion or experimentation thus far about the individual contributions of segmental vs. suprasegmental aspects of speech to the presence of a foreign accent (or to nativelike pronunciation, depending on the approach). Pitch proved to be a stronger cue to the presence of foreign accent than segmental features in Magen (1998), and intonation also appeared to have an impact on perception ratings in Carmichael (2000) (see section 4.2). Recent work shows that the intonational structure of an utterance is not isomorphic with syntactic structure (see Shattuck-Hufnagel and Turk, 1996 for a discussion of the role of prosodic structure in sentence representation), and it 15 16 requires its own parsing (Beckman, 1996). A system which is independently functional and necessitates its own processing may make a unique contribution to the perception of foreign accent. Intonation needs to be independently investigated to determine whether excellent command of suprasegmental features outweighs inaccurate segmental phonology in a native listener’s assessment of an L2 learner’s speech. It may be the case that phrase-level intonation can even cause the perception of an authentic accent (as in Carmichael, 2000, in which tokens containing measurably non-native segments and nativelike intonation were perceived to have little foreign accent). 5.0 Intonation as a perceptual discriminant between languages The perceptual import of intonation is also indicated by its robustness as a perceptual discriminant between languages. A neural network model was employed by Cummins et al. (1999) to empirically distinguish ten languages by their prosodic systems alone. Synthesized f0 contours of spontaneous speech from native Cantonese, Japanese, and English speakers provided sufficient information for adult native listeners to correctly identify these languages from each other in a study done by Ohala and Gilbert (1978). Even infants are able to discriminate between languages based on features of prosody alone, as show in a variety of experiments by Jusczyk (1995), Mehler et al. (1997), Nazzi et al. (1998), and Ramus and Mehler (1999). Jusczyk (1995) presented speech which had been low-pass filtered at 400 Hz (with the intention of removing segmental information and retaining the prosodic characteristics of the speech) to infants and found that at only 4.5 months of age, they indicated a preference for their native ambient language over other languages. Low-pass filtered speech was also used in discrimination tasks presented 16 17 by Mehler et al. (1997) and Nazzi et al. (1998). In their experiments, newborns were shown to discriminate between pairs of languages based only on the low-pass filtered speech, and they concluded that newborns were particularly sensitive to one aspect of suprasegmental phonology, linguistic rhythm (the syllable-, mora-, or stress-timing of a language as described by Pike, 1945; Abercrombie, 1967; and Ladefoged 1975). In a series of four experiments conducted by Ramus and Mehler (1999), a resynthesis technique was used to selectively degrade cues to specific segments but preserve the full frequency range of the speech signal (thus preventing the inherent unnaturalness and incompleteness of low-pass filtered speech from being a factor in the experiments). Consonant and vowel qualities were controlled in each experiment to allow for an assessment of whether suprasegmental information is cued throughout the signal (in the form of, e.g., energy bursts in particular frequency ranges or pitch excursions reflected in higher frequencies). Four sets of resynthesized stimuli were created to examine which aspects of speech were needed to discriminate between English and Japanese: a) phonotactics and prosody, b) prosody, c) rhythm only, and d) f0 only. Syllabic rhythm and intonation (f0) each independently emerged as sufficient discriminants between the two languages for adult French speakers. 6.0 Intonation in L1 It is interesting that intonation carries such weight in marking speakers as nonnative, and that it is a notoriously difficult aspect of an L2 to develop. The young human brain may be “wired” in some way to develop language (see Kuhl, 1989, for a discussion of infants’ sophistication in perceiving speech sounds); could the persistence of foreign accent in L2 17 18 learners’ use of intonation be related to an early and quick development schedule for suprasegmental acquisition? Research on the development path of first language acquisition indicates that infants are sensitive to suprasegmental properties of language, such as rhythm and intonation, long before they are responsive to segmental ones. Experiments conducted by Ramus and Mehler (1999) indicate that newborns as young as 2-5 days old can discriminate between languages based on prosodic factors but not on phonotactic ones. These results are consistent with Jusczyk’s evidence that infants do not indicate familiarity with their native language’s phonotactics until they are 6-9 months old (Jusczyk et al., 1993). 6.1 The development of language subsystems in L1 While the acquisition of intonation begins very early in life, the sophistication of a native speaker’s L1 intonation is said to develop over time. According to Neufeld and Schneiderman (1980), the linguistic system of intonation is in place quite early in life, and subsequent advances are those of degree and not kind. An analogue to this concept is the ability of adult L1ers to continue to learn new lexical items throughout life (after the L1 has been fully acquired or developed, including the structural principles governing possible lexical items). When new lexical items are learned, they conform to the phonological and grammatical constraints that have already been acquired. Neufeld and Schneiderman divide the development of ultimate attainment of intonation into primary and secondary competence. Primary competence describes the typical 5-year-old, who has acquired the intonational phonology of the L1 and speaks without an accent, but whose use of intonation is less sophisticated than that of an adult. Children at this stage 18 19 can distinguish between sentence types (such as declarative, imperative, or interrogative) and can use intonation to signal a variety of affective states. Secondary prosodic competence develops into adolescence and adulthood and enables the speaker to signal more subtle affective states and to use language in a more socially appropriate manner. Thus, the building blocks of intonational phonology appear to be in place by about age 5, with the ability to use new and more complicated grammatical constructions emerging as they move into adulthood (Berko-Gleason, 1993). Neufeld-Kaiser examined children’s ability to identify changes in a pronoun’s referent which were signaled by stress alone. The children demonstrated sensitivity to the stress to near adult level by age 5 when the contrastively stressed pronouns were part of a transitive context (e.g., “The turtle jumps over the frog, and then the chicken comes along and jumps over HIM.”). The children (and also the adults) did not perform as well, however, when the target pronouns were part of a dative construction (e.g., “Give the turtle to the frog; now give ‘im to the chicken; now give HIM to the lion.”). Neufeld-Kaiser concluded that their difficulty was due to their working memory capacity, not because they were unable to use the stress information properly (Neufeld-Kaiser, 1995).4 A comparison between prosodic and syntactic development may weaken Neufeld and Schneiderman’s assessment of prosodic development, however. Many syntactic constructions are not produced by native speakers until well into adolescence, yet some researchers claim that syntactic competence is in place in early childhood (e.g., Slobin, 1992). Are later changes in use of syntax, such as the production of new constructions, changes of degree and not kind? Is the ability to Neufeld-Kaiser refers to Daneman and Carpenter’s (1980) working memory measurement for language, the reading span test, which showed considerable differences between individuals’ ability to recall linguistic information. Also, MacDonald, Just, and Carpenter (1992) found that individuals with low reading span scores responded more slowly to stimuli and interpreted ambiguous sentences incorrectly. 4 19 20 signal subtle distinctions with intonation any different than the ability to use syntax to indicate more subtle grammatical relationships? Intonation is expressed by the manipulation of continuous variables (such as f0 or amplitude), whereas the expression of different syntactic constructions is discontinuous in nature (such as morphology, or the presence, absence, or relative location of words). The continuous nature of the components of intonation should not be misinterpreted to indicate that development of the intonation system is one of degree and not kind. The underlying structure of intonation behavior may well be different in kind from intonation behavior used just a few years prior, even though the signal expressing the intonation appears to have changed by degree. The lack of sophisticated grammatical or prosodic structures may indicate that they have not yet been acquired; or it may only indicate a lesser amount of confidence about using them on the part of the speaker (i.e., the constructions are often avoided even though they are acquired). 6.2 The acquisition schedule of linguistic subsystems in L1 The acquisition of intonation appears to have already begun within days of birth, according to the research cited in section 7. Stark (1980) identifies five stages of infant vocalizing preceding meaningful language use. The final of these stages is a period of non-reduplicated babbling (9-18 months). Stark claims that up to and during this period, children can use the intonation and stress patterns of their ambient language, but not the segmental phonology. According to Jusczyk et al. (1993), perceptual sensitivity to the phonotactics of the native language appears at 6-9 months of age, but Stark maintains that at no point during the five stages of vocalizing are the segments produced by the children 20 21 limited to those of the linguistic environment (1980). In terms of syntax, the acquisition of the native syntax is considered complete by about the age of four (Slobin, 1992). However, the productive grammar at that time isn’t replete with the variety and functional appropriateness of the adult language; it continues to develop (Kuczaj, 1999). These observations of L1 acquisition could be interpreted to show that these separate systems within the language each have their own schedule of acquisition. The onset of productive ability to use the suprasegmental phonology in L1 appears to begin the earliest, followed by segmental and articulatory features of pronunciation, and then syntax, with each of these facets of language acquired within some limited amount of time by the child. 6.3 Prosodic bootstrapping? Infants’ early sensitivity to their native language prosody may act as a bootstrap for the acquisition of syntactic structures. Basic prosodic competence would have to be in place quite early in life in order to serve as a bootstrap to the acquisition of syntax or information structure. Cutler et al. (1997) point out that prosody and syntax are not isomorphic; however, after a thorough review of research on prosody in the computation of syntactic structure, they conclude that listeners prefer syntactic analyses which are in accord with the prosodic information they hear. Morgan and Demuth’s volume on prosodic bootstrapping suggests that prosody plays some role in assisting children in the acquisition of other aspects of language (1996). Jucszyk and Kemler Nelson (1996) suggest that prosodic groupings may provide a perceptual precursor to the discovery of syntactic units. Venditti, Jun, and Beckman (1996) claim that prosody serves as a cue to 21 22 information structure rather than syntactic structure. Through a comparison of the mappings between prosodic structure and linguistic categories in Japanese, Korean, and English, they found a better fit between prosody and pragmatics than between prosody and syntax. Dresher (1996) and Fisher and Tokura (1996) also conclude that prosody does not map transparently onto syntax. Fernald and McRoberts (1996) point out that indirect evidence for prosodic bootstrapping exists, but no direct evidence exists. Only the results of manipulating the syntax-prosody relationship during the actual language acquisition process could provide unequivocal evidence for or against the prosodic bootstrapping hypothesis (p. 365). They also bring up a problem with the way prosodic bootstrapping has been investigated. Most studies evaluate the prosody-syntax relationship by estimating the probability of a prosodic cue given a particular syntactic structure [p(cue|structure)]. With this estimation, the probability could appear high even when the same cue can be used to signal other constructions. The probability estimation ought actually to be based on the probability of the syntactic structure given the occurrence of the prosodic cue [p(structure|cue)]; that is, the probability that the prosodic cue led to the correct conclusion regarding the syntactic structure (Fernald and McRoberts, 1996). A dynamic systems perspective is proposed by Hirsch-Pasek and Tucker (1996) for the role prosody plays in language acquisition. They view prosody as one of a set of mutually informing systems of language input that “coordinately enable the induction of grammar” (p. 464). The weighting of each system can change over time as the different systems interact in the acquisition process. 22 23 7.0 The acquisition of intonation in L1 vs. L2 The path of L1 acquisition may be characterized as a sequence of sensitive period onsets (i.e., the acquisition of one linguistic subsystem must begin before the next one can begin, possibly implying a sequential bootstrapping mechanism). Infants demonstrate sensitivity to intonation very early, but they begin developing segmental phonology before they have attained a level of competence in their intonational phonology. If L1 is subject to maturational or age-based constraints on acquisition, is it the case that L2 is subject to a similar maturational schedule? In review, we find evidence in the literature that syntax acquired after puberty often reveals itself as nonnative. (Note Scovel’s [1988] finding that 47% of non-native speakers were identified as non-native according to their writing samples as compared to 97% by their speech samples. It is possible that the writing task afforded more time and/or attention to the grammar than an on-demand spoken task may have, enabling L2 learners to self-correct; therefore, the spoken syntax may well be strongly nonnative.) There is also evidence that pronunciation acquired after ages as young as 6 or 7 will generally retain a foreign accent. Where does intonation fit into the L2 acquisition schedule? In L1, infants seem to begin to acquire their native intonation extremely early; is it the case that to acquire nativelike intonation in an L2, learning must also begin before some early point in maturation? The evidence of differential age-related acquisition constraints for L2 pronunciation and syntax, combined with observations that intonation is generally the most difficult aspect of an L2 to produce with nativelike mastery, suggest that in order for L2 learners (non-exceptional) to achieve nativelike intonation, they would have 23 24 to begin their exposure to and use of the L2 at an earlier age than would be necessary to achieve nativelike segmental outcomes. The traditional analysis of phonological acquisition must be subdivided into suprasegmental and segmental phonological acquisition to account for both L1 acquisition behavior and age-related pronunciation attainment in L2. This reanalysis mandates a new research agenda to explicitly and separately address the acquisition of segmental vs. suprasegmental acquisition. 8.0 New research agenda for L2 phonological acquisition: Independent investigations of segmental vs. suprasegmental acquisition In order to characterize the acquisition of segmental vs. suprasegmental aspects of L2 phonology, new research assessing their separate contributions to L2 pronunciation must be done. Assessing segmental and suprasegmental aspects of L2 acquisition separately is a nontrivial task. Finding L2 learners who only show non-native pronunciation along one of those dimensions is probably impossible. It is also the case that research into intonation is relatively new, and it’s not entirely clear how suprasegmental behavior would be evaluated and classified in terms of nativeness. A common strategy in SLA is to rely upon native speaker judgments to determine the nativeness of speech. In fact, native speaker impressions are important in speech assessment across disciplines: It is common to find terms such as “acceptability” used even when referring to the quality of synthetic speech. Native speaker judgments have helped SLA researchers learn a great deal about the SLA process, as evidenced by many of the studies cited herein. Thus, the critical requirement of a new research agenda is the 24 25 development of a way to collect native speaker judgments of segmental and suprasegmental aspects of L2 speech separately, and then organize those judgments in terms of the age-related acquisition experiences of the L2 learners. 8.1 Experimental isolation of intonation Modern signal processing technology allows us to manipulate features of pronunciation independently, keeping other features essentially unchanged. For example, the Pitch Synchronous Overlap and Add [PSOLA] resynthesis technique can be applied to a speech stream whose f0 values have been altered. PSOLA and other types of resynthesis can be done using Praat, a phonetic analysis system developed by Boersma and Weenink, 2000. By using this kind of technology, SLA researchers can independently model and synthesize foreign-accented segmental or suprasegmental features while preserving all other aspects of an original, native speech stream. The resynthesized speech samples can then be presented to native speakers for judgment of degree of foreign accent. With this technology, different aspects of pronunciation can be independently evaluated for their contribution to the perceived foreign accent in the speech. The judgments will reflect only the contribution of the non-native features specifically manipulated into the signal since all other aspects of the native speech are held constant. The reverse process can also be done, using nonnative speech as a baseline and manipulating different aspects of the pronunciation to be nativelike (cf. Magen, 1998). Nonnative speech can also be low-pass filtered to reduce segmental distinctions and preserve mostly only prosodic features. This would allow for native speaker judgments to 25 26 be based mostly on suprasegmental information. This method is not ideal for a variety of reasons: The degraded signal is highly unnatural, which could affect judgments. Further, low-pass filtering only reduces segmental distinctions; it does not remove them. The frequency band up to 400 Hz (a typical low-pass filter used in intonation research) contains features which redundantly indicate segmental information. For example, low vowels are typically accompanied by a slight lowering of f0, and high vowels are typically accompanied by a slight raising of f0. Different segments create more energy in some frequencies than others, so segments which contain more energy in the lower frequencies will be better represented in a low-pass filtered sample than other segments. Furthermore, f0 is indicated across frequencies as harmonics (energy at integer multiples of the f0). Low-pass filtering may be removing some of the critical cue information to the f0—which is precisely the aspect of the signal the filtering is intended to preserve. However, low-pass filtering could be used in conjunction with other tasks to help clarify the independent contribution of intonation to the degree of foreign accent perceived in L2 learners’ pronunciation. 8.2 Examining L1-L2 pairings The two experimental methods mentioned in section 8.1 will also allow investigation into the pronunciation outcomes of specific L1-L2 language pairs to determine what role, if any, the specific L1 plays in contributing to the perception of foreign accent. Johnson and Newport (1989) argued that their work on a maturational model of L2 (based on a grammaticality judgment task involving Chinese and Korean L2 learners of English) should extend to any L1-L2 pairing. However, Birdsong and Molis’ (2001) replication of 26 27 that study, involving Spanish learners of English, suggested that the L1-L2 pairing may be a factor in eventual L2 acquisition. Piske et al.. (2001) also question whether the relative contribution of segmental vs. suprasegmental phonology to foreign accent varies as a function of the L1-L2 pairing. Using phonetic analysis technologies, the pronunciation features of an L1 speech stream can be mechanically adjusted to match those of various L2s. The resulting resynthesized speech samples could then be presented for native speaker judgments to determine the relative influence of the L1 on different L2s. The experiment can be repeated using different L1 speech streams to show the relative influence of different L1s on the same L2. 8.3 Modeling intonation Background preparation for the kind of work envisioned in sections 8.1 and 8.2 includes a thorough understanding of the intonation systems of the L1 and L2. Research is currently underway to formulate phonological and phonetic representations of many languages’ intonation systems. The phonological Tones and Break Indices (ToBI) intonation model has been applied to at least 10 languages to date (see http://www.ling.ohio-state.edu/~tobi/), resulting in phonological models of the categorical possibilities of intonation behavior in each language. Phonetic models such as Tilt (Taylor, 2000) and INTSINT (Campione et al., 2000) are being used crosslinguistically to evaluate the acoustic outcomes of intonation behavior. These representations can be used to identify the relevant aspects of the foreign-accented intonation to model on the otherwise native speech signal. 27 28 It should be noted here that separate aspects of pronunciation may actually have a synergistic effect, amplifying the perceived degree of foreign accent when combined. Or, the naturalness of a full speech signal may override the nonnativeness to some degree. Therefore, it is important to evaluate the pronunciation of unaltered L2 speech alongside examinations of any single component of pronunciation. 8.4 Mapping L2 intonation to L2 acquisition experiences Finally, any methods used must yield measures that can be organized in terms of the SLA experiences of the speakers after whom each speech sample’s segmental or suprasegmental features were modeled. Important SLA markers include the age of first exposure to the L2, the age of arrival in the L2 immersion environment, and the age at which using the L2 became important or necessary (since age of arrival may not entail the actual onset of L2 use; see Carmichael, 2000 for a discussion). Combining new technological methods such as those outlined here with the well-attested practice of using native speaker judgments will initiate the examination of L2 segmental and suprasegmental acquisition separately. Subsequent methodological refinements will determine whether and how these aspects of L2 phonology are differentially acquired. 28 29 References Abberton, E. and A. Fourcin. (1978). Intonation and speaker identification. Language and Speech, 21:305-318. Abercrombie, D. (1967). Elements of general phonetics. Chicago, IL: Aldine. Archibald, John. (1998). Second language phonology. Philadelphia: John Benjamins. Asher, J.J. and R. Garcia (1969). The optimal age to learn a foreign language. Modern Language Journal, 53:5. 334-341. Beckman, M.E. (1996). The parsing of prosody. Language and Cognitive Processes, 11, 17-67. Berko-Gleason. J. (1993). The Development of Language. 3rd Edition. New York: Macmillan Publishing. Birdsong, D. (1989). Metalinguistic performance and interlanguage competence. New York: Springer-Verlag. Birdsong, D. (Ed.). (1999). Second language acquisition and the critical period hypothesis. Mahwah, NJ: Lawrence Erlbaum. Birdsong, D., and M. Molis. (2001). On the evidence for maturational constraints in second-language acquisition. Journal of Memory and Language, 44, 235-249. Boersma, P. and D. Weenink. (2000). Praat: a system for doing phonetics by computer. http://www.fon.hum.uva.nl/praat/. Bohn, O.-S. and J.E. Flege. (1990). Interlingual identification and the role of foreign language experience in L2 vowel perception. Applied Psycholinguistics 11, 303-328. Bohn, O.-S. and J.E. Flege. (1992). The production of new and similar vowels by adult German learners of English. Studies in Second Language Acquisition 14, 131-158. Bongaerts, T., B. Planken, and E. Schils. (1995). Can late starters attain a native accent in a foreign language? A test of the critical period hypothesis. In Eds., D. Singleton and Z. Lengyel, The Age Factor in Second Language Acquisition. Clevedon: Multilingual Matters. Bongaerts, T., C. van Summeren, B. Planken, and E. Schils. (1997). Age and ultimate attainment in the pronunciation of a foreign language. Studies in Second Language Acquisition, 19(4), 447-65. 29 30 Braine, M. (1963). The ontogeny of English phrase structure: the first phrase. Language 39:1-13. Brown, G., K. Malmkjaer, and J. Williams. (1996). Performance and competence in second language acquisition. Cambridge, MA: Cambridge University Press. Campione, Estelle, Daniel Hirst and Jean Véronis. (2000). Automatic stylisation and symbolic coding of F0: implementations of the INTSINT model. In Ed., A. Botinis, Intonation. Research and Applications. Dordrecht: Kluwer. Carmichael, L. (2000). Measurable degrees of foreign accent: a correlational study of production, perception, and acquisition. M.A. thesis, University of Washington. Cummins, F., F. Gers, and J. Schmidhuber. (1999). Comparing prosody across languages. IDSIA Technical Report. Cutler, A., D. Dahan, and W. van Donselaar. (1997). Prosody in the comprehension of spoken language: A literature review. Language and Speech, 40:141-201. Daneman, M. and P.A. Carpenter. (1980). Individual differences in working memory and reading. Journal of Verbal Learning and Verbal Behavior, 19: 450-66. Fathman, A. (1975). The relationship between age and second language productive ability. Language Learning, 25:245-253. Flege, J.E. (1987). Production and perception of English stops by native Spanish speakers. Journal of Phonetics, 15, 67-83. Flege, J.E., J.M.Munro, and I.R.A. Mackay (1995). Factors affecting strength of perceived foreign accent in a second language. Journal of Acoustical Society of America, 97:5, 3125-3134. Flege, J.E., E.M. Frieda, A.C. Walley and L.A. Randazza. (1998). Lexical factors and segmental accuracy in second language speech production. Studies in Second Language Acquisition, 20:155-187. Flege, J.E., G. Yeni-Komshian, and S. Liu. (1999). Age constraints on second-language acquisition. Journal of Memory and Language, 41, 78-104. Gilbert, Judy. (1980). Prosodic development: some pilot studies. In Eds., Scarcella, Robin C. and Stephen D. Krashen. Research in Second Language Acquisition Selected Papers of the Los Angeles Second Language Acquisition Research Forum. Rowley, MA: Newbury House Publishers, Inc. Goldsmith, J. (1999). Dealing with prosody in a text to speech system. International Journal of Speech Technology 3: 51-63. 30 31 Herschensohn, J. (2000). The Second Time Around—Minimalism and L2 Acquisition. Philadelphia: John Benjamins. Ioup, G. (1984). Is there a structural foreign accent? A comparison of syntactic and phonological errors in second language acquisition. Language Learning, 34, 2, 1-17 Ioup, G. and A. Tansomboon (1987). The acquisition of tone: a maturational perspective. In Eds., Georgette Ioup and Steven H. Weinberger, Interlanguage Phonology the Acquisition of a Second Language Sound System. Cambridge, MA: Newbury House Publishers. James. E. (1976). The acquisition of prosodic features of speech using a speech visualizer. IRAL, 14, 227-243. Jusczyk, P.W., A.D. Friederici, J. Wessels, V.Y. Svenkerud, and A.M. Jusczyk. (1993). Infants' sensitivity to the sound patterns of native language words. Journal of Memory and Language, 32, 402-420. Jusczyk, P. W. (1995). Language acquisition: speech sounds and the beginning of phonology. In Eds., J. L. Miller and P. D. Eimas, Speech, Language, and Communication. San Diego, CA: Academic Press. Krashen, S.D. (1973). Lateralization, language learning and the critical period: some new evidence. Language Learning, 23, 63-74. Kuhl, Patricia K. (1989). On babies, birds, modules, and mechanisms: A comparative approach to the acquisition of vocal communication. In Eds., Robert J. Dooling, Stewart H. Hulse, The comparative psychology of audition: Perceiving complex sounds. Hillsdale, NJ: Lawrence Erlbaum Associates, Inc. Kuczaj II, Stan A. (1999). The world of words: Thoughts on the development of a lexicon. In Ed., M. Barrett, The Development of Language. Hove, East Sussex, U.K.: Psychology Press. Ladefoged, P. (1975). A course in phonetics. New York: Harcourt Brace Jovanovich. Lado, R. (1957). Linguistics Across Cultures. Ann Arbor: University of Michigan Press. Lee, James F. and Bill VanPatten.(1995). Making Communicative Language Teaching Happen. New York: McGraw-Hill, Inc. Lenneberg, E. (1967). Biological foundation of language. New York: John Wiley. 31 32 Lively, Scott E., David B. Pisoni, Reiko A. Yamada, and Yoh'ichi Tohkura. (1994). Training Japanese listeners to identify English /r/ and /l/. III. Long-term retention of new phonetic categories. Journal of the Acoustical Society of America, 96, 4:2076-2087. Logan, J.S., S.E. Lively, and D.B. Pisoni. (1991). Training Japanese listeners to identify English /r/ and /l/: A first report. Journal of the Acoustical Society of America, 89, 874 886. Long, M. (1990). Maturational constraints on language development. Studies in Second Language Acquisition, 12, 251-285. MacDonald, M.C., M.A. Just, and P.A. Carpenter. (1992). Working memory constraints on the processing of syntactic ambiguity. Cognitive Psychology, 24, 56-98. Magen, H.S. (1998). The perception of foreign accented speech. Journal of Phonetics, 26, 381-400 Mehler, J., J. Bertoncini, E. Dupoux, and C. Pallier. (1996). The role of suprasegmentals in speech perception and acquisition. In Eds., T. Otake and A. Cutler, Phonological Structure and Language Processing: Cross-linguistic Studies (pp.145-169), New York: Mouton de Gruyter. Morgan, J. L., K. Demuth (Eds.). (1996). Signal to syntax : bootstrapping from speech to grammar in early acquisition. Mahwah, NJ: Lawrence Erlbaum Associates, Inc. Munro, Murray J. and Tracey M. Derwing. (2001). Modeling perceptions of the accentedness and comprehensibility of L2 speech. The role of speaking rate. Studies in Second Language Acquisition, 23:451-468. Nash, R. (1971). Phonemic and prosodic interference and their effects on intelligibility. International Congress of Phonetic Sciences. 7, 138-139. Nazzi, T., J. Bertoncini, and J. Mehler. (1998). Language discrimination by newborns: Towards an understanding of the role of rhythm. Journal of Experimental Psychology: Human Perception and Performance, 24, 756-766. Neufeld, G.G. and E.I. Schneiderman. (1980). Prosodic and articulatory features in adult language learning. In Eds., R. Scarcella and S. Krashen, Research in Second Language Acquisition. Rowley, MA: Newbury House. Nunan, David. (1999). Second language teaching and learning. Boston, MA: Heinle and Heinle Publishers. Ohala, J. and J. Gilbert. (1978). Listener’s ability to identify languages by their prosody. Report of the Phonology Lab II. University of California, Berkeley. 32 33 Oyama, S. (1976). A sensitive period in the acquisition of a non-native phonological system. Journal of Psycholinguistic Research, 5, 261-285. Oyama, S. (1978). The sensitive period and comprehension of speech. Working Papers on Bilingualism, 6, 1-17. Patkowski, M.S. (1980). The sensitive period for the acquisition of syntax in a second language. Language Learning, 30, 449-472. Patkowski, M.S. (1990). Age and accent in a second language: A reply to James Emil Flege. Applied Linguistics, 11:1, 73-89. Pennington M.C. and J.C. Richards. (1986). Pronunciation revisited. TESOL Quarterly, 20, 207-225. Pike, K. (1945). The intonation of American English. Ann Arbor, MI: University of Michigan Press. Pinker, S. (1987). The bootstrapping problem in language acquisition. In Ed., B. MacWhinney, Mechanisms of Language Acquisition. Hillsdale, NJ: Erlbaum. Ramus, F., and J. Mehler. (1999). Language identification with suprasegmental cues: A study based on speech resynthesis. JASA, 105, 512-521. Richards, J. C., and T.S. Rodgers. (2001). Approaches and Methods in Language Teaching (2nd ed.). Cambridge: Cambridge University Press. Riney, Timothy J. and James E. Flege. (1998). Changes over time in global foreign accent and liquid identifiability and accuracy. Studies in Second Language Acquisition, 20, 213-243. Scovel, T. (1969). Foreign accents, language acquisition, and cerebral dominance. Language Learning, 19, 245-54. Scovel, T. (1988). A time to speak: a psycholinguistic inquiry into the critical period for human speech. Rowley, MA: Newbury House. Seliger, H., S. Krashen, and P. Ladefoged. (1975). Maturational constraints in the acquisition of a native-like accent in second language learning. Languages Sciences, 36, 20-22. Shattuck-Hufnagel, S. and A. Turk. (1996). A prosody tutorial for investigators of auditory sentence processing. Journal of Psycholinguistic Research, 25: 193-247. Silverman, K., M. E. Beckman, J. Pitrelli, M. Ostendorf, C. Wightman, P. Price, J. Pierrehumbert, and J. Hirschberg. (1992). ToBI: a standard for labeling English prosody. 33 34 Proceedings of the 2nd International Conference on Spoken Language Processing (ICSLP), 2:867-70. Slobin, D. I. (Ed.) (1985a/1992). The crosslinguistic study of language acquisition. Vols. 1-2, 1985; Vol. 3, 1992. Hillsdale, NJ: Erlbaum. Snow, C. and M. Hoefnagel-Hohle (1978). The critical age for language acquisition: Evidence from second language learning. Child Development, 49, 1114-1128. Stark, R. (1980). Stages of speech development in the first year of life. In Eds., G. YeniKomshian, J. Kavanaugh, and C. Ferguson, Child phonology, Vol.1. New York: Academic Press. Tahta, S., M.Wood, and K. Loewenthal (1981). Age changes in the ability to replicate foreign pronunciation and intonation. Language and Speech, 24, 4, 363-372. Taylor, Paul. (2000). Analysis and synthesis of intonation using the tilt model. Journal of the Acoustical Society of America, 107, 3, 1697-1714. Wennerstrom, Ann. (2001). The music of everyday speech prosody and discourse analysis. New York: Oxford University Press. Wichmann, Anne. (2000). Intonation in text and discourse beginnings, middles, and ends. New York: Longman. Williams, L. (1979). The modification of speech perception and production in secondlanguage learning. Perception and Psychophysics, 26, 95-104. http://www.ling.ohio-state.edu/~tobi/ 34