Category restructuring during second-language speech acquisition.

This study examined the production of English /b/ and the perception of short-lag English /b d g/ tokens by four groups of bilinguals who differed according to their age of arrival (AOA) in Canada from Italy and amount of self-reported native language (L1) use. A clear difference emerged between early bilinguals (mean AOA= 8 years) and late bilinguals (mean AOA= 20 years). The late bilinguals showed a stronger L1 influence than the early bilinguals did on both the production and perception of English stops. In experiment 2, the late bilinguals produced a larger percentage of prevoiced English /b/ tokens than early bilinguals and native English (NE) speakers did. In experiment 3, the late bilinguals misidentified short-lag English /b d g/ tokens as /p t k/ more often than the early bilinguals and NE speakers did. Experiment 4 revealed that the frequencies with which the bilinguals prevoiced /b d g/ in Italian and English were correlated. The observed differences between the early and late bilinguals were attributed to differences in the quantity and quality of English phonetic input they had received, not to a greater likelihood by the early than late bilinguals to establish new phonetic categories for English /b d g/.

The identification of English consonants by native speakers of Italian.

This study examined the identification of English consonants in noise by native speakers of Italian. The effect of age of first exposure to English was evaluated by comparing three groups of subjects who continued to use Italian relatively often but differed according to their age of arrival (AOA) in Canada from Italy (early: 7, mid: 14, late: 19 years). The subjects in the late group made more errors identifying word-initial consonants than subjects in the early group did; however, the effect of AOA was nonsignificant for word-final stops. The effect of amount of native language (L1) use was evaluated by comparing two groups of early bilinguals who were matched for AOA (mean = 7 years) but differed according to self-reported percentage use of Italian (early: 32%, early-low: 8%). The early bilinguals who used Italian often (early) made significantly more errors identifying word-initial and word-final consonants than native English (NE) subjects did, whereas the early bilinguals who used Italian seldom (early-low) did not differ from the NE subjects. The subjects' phonological short-term memory was estimated by having them repeat Italian non-words. This was done in an attempt to identify the source of individual differences. The nonword repetition scores were in fact found to independently account for 15% of the variance in subjects' errors identifying word-final English consonants and 8% of the variance for word-initial consonants.

An investigation of current models of second language speech perception: the case of Japanese adults' perception of English consonants.

This study reports the results of two experiments with native speakers of Japanese. In experiment 1, near-monolingual Japanese listeners participated in a cross-language mapping experiment in which they identified English and Japanese consonants in terms of a Japanese category, then rated the identifications for goodness-of-fit to that Japanese category. Experiment 2 used the same set of stimuli in a categorial discrimination test. Three groups of Japanese speakers varying in English-language experience, and one group of native English speakers participated. Contrast pairs composed of two English consonants, two Japanese consonants, and one English and one Japanese consonant were tested. The results indicated that the perceived phonetic distance of second language (L2) consonants from the closest first language (L1) consonant predicted the discrimination of L2 sounds. In addition, this study investigated the role of experience in learning sounds in a second language. Some of the consonant contrasts tested showed evidence of learning (i.e., significantly higher scores for the experienced than the relatively inexperienced Japanese groups). The perceived phonetic distance of L1 and L2 sounds was found to predict learning effects in discrimination of L1 and L2 sounds in some cases. The results are discussed in terms of models of cross-language speech perception and L2 phonetic learning.

Adults' perception and production of the English vowel /i/.

This study investigated the link between the perception and production of the English vowel /i/ by adult native speakers of English. Participants first produced the vowel /i/ using normal (citation) and careful (hyperarticulated) speech, then completed a method of adjustment task in which they selected their ideal exemplar of /i/. In this perceptual task, 24 of 35 participants had a prototype; the remaining 11 did not, but were retained for comparison. In keeping with the hyperspace effect (K. Johnson, E. Flemming, & R. Wright, 1993), all participants selected perceptual stimuli with F1 and F2 values that were more extreme (i.e., higher and further forward in the vowel space) than those of their normal, citation productions. An analysis of front-back and high-low qualities for the perceptual and production data in Euclidian space revealed that hyperarticulated speech was closer to the perceptual data than citation speech was, but only for participants with relatively clear-cut prototypes. The basis for such individual variation in perception-production links is discussed.

Native Italian speakers' perception and production of English vowels.

This study examined the production and perception of English vowels by highly experienced native Italian speakers of English. The subjects were selected on the basis of the age at which they arrived in Canada and began to learn English, and how much they continued to use Italian. Vowel production accuracy was assessed through an intelligibility test in which native English-speaking listeners attempted to identify vowels spoken by the native Italian subjects. Vowel perception was assessed using a categorial discrimination test. The later in life the native Italian subjects began to learn English, the less accurately they produced and perceived English vowels. Neither of two groups of early Italian/English bilinguals differed significantly from native speakers of English either for production or perception. This finding is consistent with the hypothesis of the speech learning model [Flege, in Speech Perception and Linguistic Experience: Theoretical and Methodological Issues (York, Timonium, MD, 1995)] that early bilinguals establish new categories for vowels found in the second language (L2). The significant correlation observed to exist between the measures of L2 vowel production and perception is consistent with another hypothesis of the speech learning model, viz., that the accuracy with which L2 vowels are produced is limited by how accurately they are perceived.

Adults' perception of native and nonnative vowels: implications for the perceptual magnet effect.

Two experiments were conducted to evaluate the perceptual magnet effect. In Experiment 1, American English speakers representing diverse dialects were presented with a fine-grained set of stimuli (varying in just noticeable differences for F1 and F2) and indicated whether they heard "/i/" or "not/i/," thus delimiting the /i/ portion of the vowel space for individual subjects. Then these same subjects selected their own /i/ prototype with a method-of-adjustment procedure. The data from this experiment were used to synthesize customized prototype and nonprototype stimulus sets for Experiment 2. In Experiment 2, 24 of our original 37 subjects completed a discrimination task for each of three conditions, in which vector stimuli varied from the subject's prototype, the nonprototype, or a foreign vowel (/y/) in 15-mel steps. Subjects displayed higher discrimination, as indexed by d', for the nonprototype condition than they did for both the prototype and the foreign conditions. In addition, discrimination was better for variants further away from the referent in each condition. However, discrimination was not especially poor for stimuli close to subjects' individual prototypes--a result that would have yielded the strongest support for the operation of a magnet effect. This negative finding, together with other aspects of our results, raises problems for any theory of vowel perception that relies solely on "one-size-fits-all" prototype representations.

Lexical familiarity and English-language experience affect Japanese adults' perception of / / and /l/.

This study assessed the influence of subjective lexical familiarity and English-language experience on Japanese adults' accuracy in identifying singleton word-initial tokens of English [symbol: see text] and /l/. The inexperienced Japanese (IJ) subjects had lived in the U.S. for 2 years, whereas the experienced Japanese (EJ) subjects had lived there for 21 years, on average. The native Japanese subjects correctly identified English [symbol: see text] and /l/ tokens less often than did a group of native English (NE) subjects, but they did not differ from the NE subjects in identifying the control consonants /w/ and /d/. The NE subjects, who were at ceiling, showed no effect of subjective lexical familiarity. However, the EJ and IJ subjects correctly identified [symbol: see text] and /l/ tokens more often in words that were more familiar than their minimal pairs than in words that were less familiar than their minimal pairs. The EJ subjects identified liquids more often than did the IJ subjects, but usually less often than the NE subjects. However, the EJ subjects managed to identify [symbol: see text] tokens at rates comparable to the NE subjects' rates in words that wer matched in subjective familiarity to their minimal pair (experiment 1), and when identifying [symbol: see text] tokens that had been edited out of their original word or nonword context (experiment 2).

Age of learning affects rate-dependent processing of stops in a second language.

The aim of this study was to assess the effect of speaking rate changes on the perception of English stop consonants by four groups of subjects: English and Spanish monolinguals, 'early' Spanish/English bilinguals who learned English in childhood, and 'late' bilinguals who learned English in adulthood. Subjects identified, and then later rated for goodness as exemplars of the English /p/ category, the members of two voice onset time (VOT) continua. The English monolinguals identified a well-defined range of VOT stimuli as English /p/, and stimuli with longer VOT values as 'exaggerated' instances of English /p/. Their goodness ratings increased as VOT increased, then showed a systematic decrease as VOT began to exceed values typical for English /p/. The English monolinguals' goodness ratings also varied systematically as a function of speaking rate, which was simulated in the two continua by varying syllable duration. The Spanish monolinguals, on the other hand, failed to consistently identify any of the stimuli as English /p/. Although speaking rate influenced their goodness ratings, the Spanish monolinguals' rate effects differed significantly from the English monolinguals'. The early bilinguals resembled the English monolinguals, and differed from the Spanish monolinguals to a greater extent than did the late Spanish/English bilinguals. This was taken as support for the hypothesis that early bilinguals are more likely than are late bilinguals to establish new phonetic categories for stop consonants in a second language.

Speaking rate effects on stops produced by Spanish and English monolinguals and Spanish/English bilinguals.

Four groups of 10 subjects each (English and Spanish monolinguals, and two groups of Spanish/English bilinguals) produced Spanish or English sentences at speaking rates designated 'normal', 'slow', and 'fast'. Voice onset time (VOT) was measured in word-initial tokens of /p/ and /t/ found in sentence-initial, -medial, and -final words. The four groups produced comparable changes in sentence duration across the three rates. The speaking rate changes exerted less effect on the VOT in stops spoken by the Spanish than the English monolinguals. Moreover, whereas English monolinguals produced /p,t/ (with shorter VOT at a fast than at a normal rate, many Spanish monolinguals showed a trend in the opposite direction. As expected, all 10 early bilinguals produced English stops with VOT values that were similar to the English monolinguals'. They also showed speaking rate effects on VOT that were similar to those observed for the English monolinguals. The late bilinguals, who had begun learning English as adults, showed smaller effects of speaking rate on VOT than did the English monolinguals. Their mean VOT values for English stops spanned a wide range of values. Just 3 of the late bilinguals produced English stops with VOT values that fell within the range of values observed for the English monolinguals.

Factors affecting strength of perceived foreign accent in a second language.

This study assessed the relation between non-native subjects' age of learning (AOL) English and the overall degree of perceived foreign accent in their production of English sentences. The 240 native Italian (NI) subjects examined had begun learning English in Canada between the ages of 2 and 23 yr, and had lived in Canada for an average of 32 yr. Native English-speaking listeners used a continuous scale to rate sentences spoken by the NI subjects and by subjects in a native English comparison group. Estimates of the AOL of onset of foreign accents varied across the ten listeners who rated the sentences, ranging from 3.1 to 11.6 yr (M = 7.4). Foreign accents were evident in sentences spoken by many NI subjects who had begun learning English long before what is traditionally considered to be the end of a critical period. Very few NI subjects who began learning English after the age of 15 yr received ratings that fell within the native English range. Principal components analyses of the NI subjects' responses to a language background questionnaire were followed by multiple-regression analyses. AOL accounted for an average of 59% of variance in the foreign accent ratings. Language use factors accounted for an additional 15% of variance. Gender was also found to influence degree of foreign accent.

The perception of English and Spanish vowels by native English and Spanish listeners: a multidimensional scaling analysis.

A group of Spanish- and English-speaking listeners participated in a multidimensional scaling (MDS) study examining perceptual responses to three Spanish and seven English vowels. The vowel stimuli represented tokens of Spanish /i/, /e/, and /a/ and English /i/, /I/, /eI/, /epsilon/, /ae/, /lambda/, and /a/. Each vowel had been spoken by three monolingual talkers of Spanish or English and all possible vowel pairs (405 pairs) were presented to listeners (excluding pairs representing the same vowel category). Thirty monolingual English listeners and thirty native Spanish listeners who had learned English as a second language rated these vowel pairs on a nine-point dissimilarity scale. These perceptual distances were then analyzed using then individual-differences version of ALSCAL. Results demonstrated that the English monolinguals used three underlying dimensions in rating vowels while the Spanish-English bilinguals used just two. The most salient perceptual dimension for both groups distinguished vowel height. However, for the English listeners, this dimension was most significantly correlated with duration and indicated a language-dependent sensitivity to this phonetic feature. The second dimension for the English listeners represented a front-back distinction, while the third reflected a central/noncentral distinction. For the Spanish listeners, the second dimension was less easily interpreted. However, the perceptual data for the Spanish listeners was more interpretable in terms of the distribution of the vowels in the two-dimensional perceptual plane. The vowels were distributed in terms of three separate vowel clusters, each cluster near the location of a Spanish vowel.(ABSTRACT TRUNCATED AT 250 WORDS)

Effects of speaking rate changes on native and nonnative speech production.

Twenty monolingual English speakers and 40 native Spanish speakers, who were either relatively proficient or nonproficient in English, produced English /p/ at self-selected normal and fast speaking rates. The native English (NE) subjects showed much the same rate effect on voice onset time (VOT) seen in earlier studies. Native Spanish (NS) subjects who were relatively proficient in pronouncing English used fewer short-lag stops and showed an English-like rate effect on VOT. NS subjects who were relatively nonproficient in English, on the other hand, used more short-lag stops and did not show a significant speaking rate effect on VOT, although they did demonstrate speaking rate effects on vowel duration and phrase duration. Among the NS subjects were 21 whose VOT values closely matched those of the NE subjects at both speaking rates, and who showed a rate effect similar to that found for the NE subjects. Their ability to modify VOT across speaking rates suggested that these NS subjects may have established a phonetic category for English /p/.

Japanese adults can learn to produce English /r/ and /l/ accurately.

As is well known, Japanese adults who have just begun to learn English often err in producing /r/ and /l/ because their native language does not possess such liquid consonants. The aim of this study was to determine if Japanese adults eventually learn to produce /r/ and /l/ accurately in words like read and lead. Liquids spoken by 12 native Japanese speakers who had lived in the United States for an average of two years were often misidentified by native English-speaking listeners. Their productions of /r/ and /l/ also received much lower (and thus foreign-accented) ratings than did the native English speakers' liquids. On the other hand, liquids produced by native Japanese speakers who had lived in the United States for an average 21 years were identified correctly in forced-choice tests. This held true for liquids in words that had been read from a list as well as for words that had been spoken spontaneously. The ratings of liquids produced by 10 of the 12 experienced Japanese speakers fell within the range of ratings obtained for the 12 native English speakers. These findings challenge the widely accepted view that segmental production errors in a second language arise from the inevitable loss of ability to learn phonetic segments not found in the native language.

Native speakers of Spanish show rate-dependent processing of English stop consonants.

English monolinguals and native Spanish speakers of English used a 9-point scale to rate syllable-initial stops for goodness as realizations of the English /p/ category. Voice onset time (VOT) was varied in a set of short-duration ('fast-rate') consonant-vowel (CV) stimuli, and in a set of longer-duration ('slow-rate') CV stimuli. The VOT values ranged from values typical for English /b/ to values which exceeded those typical for English /p/. Results for the native English (NE) subjects replicated those obtained previously using the same two continua. Goodness ratings systematically increased, then decreased as VOT values extended beyond the range typical for English /p/. The NE subjects gave their highest ratings to stimuli with VOT values of about 50 ms. For stops with longer VOT values, their ratings were higher for stimuli in the slow-rate than fast-rate continuum. The native Spanish (NS) subjects were assigned to two subgroups based on degree of foreign accent in their pronunciation of English sentences. Both proficient and nonproficient NS subjects gave their highest ratings to stimuli with much the same VOT values as the NE subjects, even though /p/ is realized with short-lag VOT in Spanish. The nonproficient, but not the proficient NS subjects showed significantly smaller rate effects on goodness ratings than did the NE subjects. However, the subjects in both NS groups gave significantly higher ratings to stimuli with short-lag VOT values than did the NE subjects. The results are discussed in terms of the hypothesis that adult learners of a second language may establish new perceptual phonetic categories for phonetic segments not found in the native language.

Auditory and categorical effects on cross-language vowel perception.

English monolinguals and native Spanish speakers of English rated the dissimilarity of tokens of two Spanish vowel categories, two English vowel categories, or one Spanish and one English vowel category. The dissimilarity ratings of experienced and inexperienced Spanish subjects did not differ significantly. For both the native Spanish and English subjects, perceived dissimilarity increased as the distance between vowels in an F1-F2 acoustic space increased. This supported the existence of a universal, sensory-based component in cross-language vowel perception. The native English and Spanish subjects' ratings were comparable for pairs made up of vowels that were distant in an F1-F2 space, but not for pairs made up of vowels from categories that were adjacent in an F1-F2 space. The inference that the differential classification of a pair of vowels augments perceived dissimilarity was supported by the results of experiment 2, where subjects rated pairs of vowels and participated in an oddity discrimination task. Triads in the oddity task were made up of tokens of vowel categories that were either adjacent (e.g., /a/-/ae/-/a/) or nonadjacent (e.g., /a/-/i/-/i/) in an F1-F2 space. The native English subjects' discrimination was better than the native Spanish subjects' for adjacent but not nonadjacent triads. The better the Spanish subjects performed on adjacent triads--and thus the more likely they were to have differentially classified the two phonetically distinct vowels in the triad--the more dissimilar they had earlier judged realizations of those two categories to be when presented in pairs. Results are discussed in terms of their implications for second language acquisition.

Production and perception of a novel, second-language phonetic contrast.

Four experiments, all of which focused on vowel duration, assessed Chinese subjects' production and perception of the contrast between /t/ and /d/ in the final position of English words. Vowel duration was measured in minimal pairs in the first experiment. The stimuli in natural-edited beat-bead and bat-bad continua in which vowel duration varied in 20-ms steps were then presented to native English and Chinese subjects in a forced-choice test, in an experiment using the method of adjustment, and in an imitation task. The non-natives who learned English in childhood closely resembled native speakers in all four experiments. Three groups of non-natives who had learned English as a second language in adulthood, on the other hand, differed from the native speakers. The late learners produced significantly longer vowels in words ending in /d/ than /t/. However, the late learners' vowel duration differences were much smaller than the native speakers', and were correlated significantly with degree of foreign accent in English. The late learners differed from the native speakers in several ways in the two perception experiments, and also in the imitation task. The pattern of significant and nonsignificant between-group differences, but not data for individual subjects, was consistent with the hypothesis that L2 (second language) production accuracy is limited by the adequacy of perceptual representations for sounds in the L2.

Production of the word-final English /t/-/d/ contrast by native speakers of English, Mandarin, and Spanish.

The primary aim of this study was to determine if adults whose native language permits neither voiced nor voiceless stops to occur in word-final position can master the English word-final /t/-/d/ contrast. Native English-speaking listeners identified the voicing feature in word-final stops produced by talkers in five groups: native speakers of English, experienced and inexperienced native Spanish speakers of English, and experienced and inexperienced native Mandarin speakers of English. Contrary to hypothesis, the experienced second language (L2) learners' stops were not identified significantly better than stops produced by the inexperienced L2 learners; and their stops were correctly identified significantly less often than stops produced by the native English speakers. Acoustic analyses revealed that the native English speakers made vowels significantly longer before /d/ than /t/, produced /t/-final words with a higher F1 offset frequency than /d/-final words, produced more closure voicing in /d/ than /t/, and sustained closure longer for /t/ than /d/. The L2 learners produced the same kinds of acoustic differences between /t/ and /d/, but theirs were usually of significantly smaller magnitude. Taken together, the results suggest that only a few of the 40 L2 learners examined in the present study had mastered the English word-final /t/-/d/ contrast. Several possible explanations for this negative finding are presented. Multiple regression analyses revealed that the native English listeners made perceptual use of the small, albeit significant, vowel duration differences produced in minimal pairs by the nonnative speakers. A significantly stronger correlation existed between vowel duration differences and the listeners' identifications of final stops in minimal pairs when the perceptual judgments were obtained in an "edited" condition (where post-vocalic cues were removed) than in a "full cue" condition. This suggested that listeners may modify their identification of stops based on the availability of acoustic cues.

Talker and listener effects on degree of perceived foreign accent.

Four experiments were carried out to examine listener- and talker-related factors that may influence degree of perceived foreign accent. In each, native English listeners rated English sentences for degree of accent. It was found that degree of accent is influenced by range effects. The larger the proportion of native (or near-native) speakers included in a set of sentences being evaluated, the more strongly accented listeners judged sentences spoken by non-native speakers to be. Foreign accent ratings were not stable. Listeners judged a set of non-native-produced sentences to be more strongly accented after, as compared to before, they became familiar with those sentences. One talker-related effect noted in the study was the finding that adults' pronunciation of an L2 may improve over time. Late L2 learners who had lived in the United States for an average of 14.3 years received significantly higher scores than late learners who had resided in the United States for 0.7 years. Another talker-related effect pertained to the age of L2 learning (AOL). Native Spanish subjects with an AOL of five to six years were not found to have an accent (i.e., to receive significantly lower scores than native English speakers), whereas native Chinese subjects with an average AOL of 7.6 years did have a measurable accent. The paper concludes with the presentation of several hypotheses concerning the relationship between AOL and degree of foreign accent.

The interlingual identification of Spanish and English vowels: orthographic evidence.

When someone who is learning a second language (L2) produces a sound in the L2 using a familiar, native-language (L1) category, the L2 sound is said to have been "identified with" an L1 sound. Although interlingual identification exerts a powerful influence on L2 pronunciation, it is still poorly understood. Orthographic classification was used here to assess the interlingual identification of Spanish and English vowels. Sixty native speakers of Spanish in three experiments judged the vowels /i/, /I/, /e/, and /ae/ in multiple tokens of English words ("beat", "bit", "bet", "bat") spoken by ten native speakers of American English. The subjects labelled each English vowel by circling one of the five letters used to spell the vowel phonemes of Spanish (viz. less than i greater than, less than e greater than, less than a greater than, less than o greater than, less than u greater than) or by circling "none" if they thought they had heard a vowel not found in Spanish. Subjects who spoke English as an L2 used the "none" label more often than did Spanish monolinguals, suggesting that L2 learning heightens bilinguals' awareness of cross-language phonetic differences. Experienced Spanish speakers of English did use the "none" label more often than did inexperienced subjects (42% vs. 18%). A few subjects used the "none" label consistently for /ae/ and /I/, suggesting that they may have regarded these vowels as "new" (i.e., non-Spanish). However, the group data provided little support for the hypothesis that the adult Spanish learners of English treated either /ae/ or /I/ as new. The great majority of subjects, even those highly experienced in English, identified English /ae/ with their Spanish /a/.

Age of learning affects the authenticity of voice-onset time (VOT) in stop consonants produced in a second language.

This study examined whether Spanish-English bilinguals are able to fully differentiate Spanish and English /t/ according to voice-onset time (VOT) if they learn English as a second language (L2) in early childhood. In experiment 1, VOT was measured in Spanish words spoken by Spanish monolinguals, in English words spoken by English monolinguals, and in Spanish and English words spoken by bilinguals who learned English either as young children or as adults. As expected, the Spanish monolinguals produced /t/ with considerably shorter VOT values than the English monolinguals. Also as expected, the late L2 learners produced English /t/ with "compromise" VOT values that were intermediate to the short-lag values observed for Spanish monolinguals and the long-lag values observed for English monolinguals. The early learners' VOT values for English /t/, on the other hand, did not differ from English monolinguals' VOT. The same pattern of results was obtained for stops in utterance-medial position and in absolute utterance-initial position. The results of experiment 1 were replicated in experiment 2, where bilingual subjects were required to produce Spanish and English utterances (sentences, phrases, words) in alteration. The results are interpreted to mean that individuals who learn an L2 in early childhood, but not those who learn an L2 later in life, are able to establish phonetic categories for sounds in the L2 that differ acoustically from corresponding sounds in the native language. It is hypothesized that the late L2 learners produced /t/ with slightly longer VOT values in English than Spanish by applying different realization rules to a single phonetic category.