15
J. COMMUN. DISORD. 31 (1998), 215–229 © 1998 by Elsevier Science Inc. All rights reserved. 0021-9924/98/$19.00 655 Avenue of the Americas, New York, NY 10010 PII S0021-9924(97)00053-1 VOICE ONSET TIME IN SPANISH-ENGLISH BILINGUALS: EARLY VERSUS LATE LEARNERS OF ENGLISH DIANNE F. THORNBURGH Department of Communicative Disorders, University of Central Florida, Orlando, Florida JOHN H. RYALLS Department of Communicative Disorders, University of Central Florida, Orlando, Florida Thirty-two Hispanic speakers of English were evenly divided into two groups based on whether or not their initial learning of English began prior to, or after the age of 12 years. Each group had an even number of males (16) and females (16). The subjects were recorded producing a protocol of 18 basic speech syllables. The first three repetitions (54 tokens) were chosen for analysis. The 1728 tokens were digitized and measured for voice onset time (VOT). Findings support the hypothesis that the VOT values of Hispanics speaking English differ according to whether initial learning of English began prior to or after the age of 12 years. An analysis of variance (ANOVA) found significant main effects of group, place, voice, and gender. Significant interactions were group by voice, and voice by gender. © 1998 by Elsevier Science Inc. Educational Objectives: The reader will understand voice onset time, one of the important differences between the way Spanish and English are pronounced. The reader will also gain insight into one of the difficulties Spanish speakers have in learning English. KEY WORDS: Language acquisition; Voice; Bilingual speakers INTRODUCTION According to the 1994 Statistical Abstract of the United States (USDUC, 1994), about one out of every ten Americans, or nearly 26 million people, are native speakers of Spanish. Approximately 70% to 75% may remain essen- tially monolingual for several reasons, including lack of access to educational institutions, lack of confidence in their own ability to learn a second language, or occupations and living arrangements that do not require the use of English (Langdon & Merino, 1992). Because the phonological system of English is Address correspondence to John Ryalls, Ph.D., Department of Communicative Disorders, University of Central Florida,12424 Research Parkway, Suite 200-210, Orlando, FL 32826

Voice onset time in spanish-english bilinguals: early versus late learners of english

Embed Size (px)

Citation preview

Page 1: Voice onset time in spanish-english bilinguals: early versus late learners of english

J. COMMUN. DISORD. 31 (1998), 215–229© 1998 by Elsevier Science Inc. All rights reserved. 0021-9924/98/$19.00655 Avenue of the Americas, New York, NY 10010 PII S0021-9924(97)00053-1

VOICE ONSET TIME INSPANISH-ENGLISH BILINGUALS:EARLY VERSUS LATE LEARNERS OF ENGLISH

DIANNE F. THORNBURGH

Department of Communicative Disorders, University of Central Florida, Orlando, Florida

JOHN H. RYALLS

Department of Communicative Disorders, University of Central Florida, Orlando, Florida

Thirty-two Hispanic speakers of English were evenly divided into two groups based onwhether or not their initial learning of English began prior to, or after the age of 12 years.Each group had an even number of males (16) and females (16). The subjects were recordedproducing a protocol of 18 basic speech syllables. The first three repetitions (54 tokens)were chosen for analysis. The 1728 tokens were digitized and measured for voice onset time(VOT). Findings support the hypothesis that the VOT values of Hispanics speaking Englishdiffer according to whether initial learning of English began prior to or after the age of12 years. An analysis of variance (ANOVA) found significant main effects of group,place, voice, and gender. Significant interactions were group by voice, and voice by

gender. © 1998 by Elsevier Science Inc.

Educational Objectives:

The reader will understand voice onset time, one of the importantdifferences between the way Spanish and English are pronounced. The reader will also gaininsight into one of the difficulties Spanish speakers have in learning English.

KEY WORDS:

Language acquisition; Voice; Bilingual speakers

INTRODUCTION

According to the 1994

Statistical Abstract of the United States

(USDUC,1994), about one out of every ten Americans, or nearly 26 million people, arenative speakers of Spanish. Approximately 70% to 75% may remain essen-tially monolingual for several reasons, including lack of access to educationalinstitutions, lack of confidence in their own ability to learn a second language,or occupations and living arrangements that do not require the use of English(Langdon & Merino, 1992). Because the phonological system of English is

Address correspondence to John Ryalls, Ph.D., Department of Communicative Disorders, Universityof Central Florida,12424 Research Parkway, Suite 200-210, Orlando, FL 32826

Page 2: Voice onset time in spanish-english bilinguals: early versus late learners of english

216 THORNBURG and RYALLS

more complex than that of Spanish, many who do desire to learn English mayseek assistance in developing their “communicative competency” (Kayser,1993).

Speech language pathologists are increasingly being consulted by adultnon-native speakers of English who wish to improve their English pronuncia-tion. However, many speech language pathologists possess inadequate knowl-edge of culturally-based data, theory, and clinical standards (Taylor, 1993). Inorder to be “capable of evaluating communication needs of specific popula-tions” [(Goldberg, 1995) p. 48], the speech language pathologist will need tobecome knowledgeable not only about the phonological and grammaticalstructure of English, but also how the client’s native language may influenceacquisition of English (Gandour, 1980).

The attainment of native-like pronunciation for most second-languagelearners appears an elusive goal. Only about 5% of second-language learnersattain “native-speaker competence,” or “productive performance in the secondlanguage . . . identical to that produced by the native speaker of the tongue”[(Ho, 1986) p. 36]. Flege (1984) delineates “adequate pronunciation” (i.e., in-telligible, though heavily accented) from “authentic pronunciation” (i.e., howcompliant the utterance is with second-language phonetic norms, based onphysical measures such as voice onset time).

Global foreign accent is significantly correlated with voice onset time(VOT) production in stop consonants produced by non-native speakers (Flege,1992). Initial position voiceless stops in Spanish are usually produced withoutaspiration (Langdon & Merino, 1992). However, English voiceless stops aretypically aspirated except when they follow /s/ (Kent & Read, 1992). The sub-stitution of Spanish unaspirated stops for English aspirated ones may be mis-taken by a native speaker of English as voiced stops (Nathan, 1987).

English stops also differ significantly from Spanish stops in VOT (Flege,1988). English voiceless stops have significantly longer VOT than do voice-less stops in languages like French and Spanish. VOT values for Englishvoiced stops are similar to those for voiceless stops in French and Spanish.Voiced stops in French and Spanish have voicing prior to articulatory release.Native speakers of French and Spanish must learn to produce English voice-less stops with longer VOT values than those of corresponding stops in theirnative languages.

It appears that language-specific VOT values must be learned for adequateand authentic pronunciation of the second language (Flege, 1988). Flege, Mu-nro, & MacKay (1995) found acoustical differences between native Englishspeakers and native Italian speakers with age of learning (AOL) English asearly as 11 years.

Researchers believe that the optimal learning of a second language is be-tween the ages of 2 and 12 or 13 years (Flege, 1992; Yeni-Komshian, 1995).Lenneberg (1967) hypothesized that second language learning after age 12

Page 3: Voice onset time in spanish-english bilinguals: early versus late learners of english

VOICE ONSET TIME IN BILINGUALS 217

was limited by the decrease in neural plasticity that accompanied lateraliza-tion of cerebral function (Yeni-Komshian, 1995). More recent research hasproposed that the learning of phonemic categories in the first language mayinterfere with formation of phonemic categories in the second language. Inother words, the second language learner may perceive second languagesounds similar to sounds in the first language as allophones of the same pho-nemic category (Flege, 1992a, 1992b). Motivation to learn and use the secondlanguage, adequacy of phonetic input, motoric difficulty with articulation,habit formation, and incorrrect perception are several factors that may influ-ence second language proficiency (Flege, 1988; Flege, Munro, & MacKay,1995).

Considering age as a factor in learning a second language, although mostresearchers agree that “earlier is better” in regard to second language learning(Yeni-Komshian, 1995), some researchers disagree whether or not pronuncia-tion of a second language may totally match that of a native-born speaker(Mack, 1984). Flege (1992a, 1992b) believes that the English sounds /p/, /t/, /k/may be mastered if second language learning begins by age 5 or 6. Yeni-Komshian (1995), in a study of Korean-English bilinguals, found that eventhose with initial learning about age 6 had a perceptible but mild accent. How-ever, in the same study 32 out of 220 subjects (15%) received high overall rat-ings in both English and Korean, demonstrating that it is possible, though notusual, to achieve native-like performance in both languages.

The goal of the present study is threefold. First, to compare the productionof both voiceless and voiced stops by Hispanic speakers of English who beganto learn English prior to age 12 with those whose initial learning of Englishbegan at or after this age. No prior study has targeted both voiceless andvoiced stops, nor compared “early” versus “late” learners of English. Second,although VOT studies of bilingual speakers have included both males and fe-males as subjects, none of them have isolated gender as a potential factor inVOT differences. Third, and perhaps most important, was our immediate goalof providing more objective data on VOT in bilingual speakers with the even-tual goal of developing more informed and more effective therapy for bilin-gual dialect clients.

It was expected that the earlier English is learned as a second language, themore English-like the VOT values for both voiceless and voiced stops. Genderwas not predicted to be a significant factor, despite gender-based physiologicaldifferences relating to speech (Kahane 1978; Titze, 1989). Since typical VOTvalues for English and Spanish are much closer for voiceless stops than voicedstops, it was expected that all subjects would more closely approximate En-glish for voiceless targets. Furthermore since neither Spanish nor English pos-ses a phonemic distinction between negative and short positive VOT valuesfor voiced stops, it was expected that voiced stop targets would result in “hy-brid” VOT values which would fall between the typical long negative values

Page 4: Voice onset time in spanish-english bilinguals: early versus late learners of english

218 THORNBURG and RYALLS

associated with Spanish, and the short positive values associated with English,but still be negative in value. That is, we expected Spanish-English bilingualswould produce voiced stops with shorter negative values than typically pro-duced in Spanish, but still with the prevoicing which characterizes Spanishvoiced stops (but is atypical for English).

METHODS

Subjects

Thirty-two native speakers of Spanish were evenly divided into two groups de-pending on whether their initial learning of English began prior to or after theage of 12 years. Both groups had equal numbers of males and females. Subjectswere either students at the University of Central Florida (UCF), or enrolled inthe university’s Intensive English Program (IEP). IEP subjects were not enrolledat UCF, but were in this country for three months intensive study of English.

In the pre-12 group (

n

5

16) 11 subjects were UCF students and five werein the IEP. Minimum age of learning (AOL) English for this group was 1.5years, maximum AOL 11 years, with mean AOL of 5 years, 7 months. Twostudents in the pre-12 group only (ME 12 and FE 05) stated they learnedSpanish and English simultaneously. In the post-12 group, eight were UCFstudents and eight were in the IEP. Minimum AOL English for this group was12 years, maximum AOL 18 years, with mean AOL of 14 years, 4 months.

Native countries of the 32 subjects were: the United States (4), Puerto Rico(8), Mexico (1), Cuba (1), Venezuela (7), Paraguay (2), Colombia (6), Ecua-dor (1), and El Salvador (2). Although there is some lexical variation acrossthe Spanish spoken in various countries, no variation in VOT by country hasbeen reported in the literature.

Mean estimated daily use of English for the pre-12 group (

n

5

16) was73% with a range from 50% to 90%. For the post-12 group (

n

5

16), mean es-timated daily use of English was 65%, with a range from 25% to 90%.

Stimuli

A protocol of 18 basic speech syllables using the six stops (/p/, /t/, /k/, /b/, /d/,/g/) in the initial position combined with the three extreme vowels (/i/, /

ɑ

/, /u/)was developed. These three vowels, known as point vowels, should be articu-lated in approximately the same locations across all languages (Bradlow,1995). The word stimuli are listed in Table 1.

All 18 syllables were elicited seven times in quasi-random order. The sameinstructions were read to each subject after being seated in the sound-treatedroom. Following a reading for pronunciation practice, seven recordings of the18 syllables were elicited. The first three repetitions (54 tokens) were chosen

Page 5: Voice onset time in spanish-english bilinguals: early versus late learners of english

VOICE ONSET TIME IN BILINGUALS 219

for analysis, unless the subject substituted a phoneme from a different place ofarticulation, mispronounced a vowel, the amplitude was too low to view theburst, or a technical malfunction (e.g., microphone feedback) altered either thevisual or audio display of data.

Tokens from other than the first three repetitions were used for only 120out of the total 1728 tokens, or about 7% of the tokens analyzed. This oc-curred evenly between the pre-12 (two females, five males) and post-12 (fourfemales, three males) groups.

Productions were recorded in an Industrial Acoustic Company (IAC) sound-treated room, using a portable Sony TCD-D3 Digital Audio Tape (DAT) re-corder. A Sony model ECM-901 electret microphone was positioned approxi-mately 6 inches from each subject’s mouth.

ANALYSIS

Recordings were digitized onto disk with 12 bits of resolution using BLISSsoftware (Mertus, 1989) at a sampling rate of 10 kHz, after low-pass filteringat 4.9 kHz to avoid aliasing. The BLISS software package was implementedon an AT-compatible microcomputer equipped with a Data Translation DT-2801A analogue/digital converter. VOT measurements were performed usingboth auditory and visual cues.

VOT was measured on the basis of both the oscillographic display and theaudio signal, in accord with standard procedures. For voiceless tokens, VOTwas measured from the burst to the onset of periodic phonation. This durationis expressed as a positive number (Lisker & Abramson, 1964). For voiced to-kens, the same procedure was followed, except in cases of prevoicing. Forprevoiced tokens, the VOT interval is measured from the onset of voicing tothe burst and thus is a negative value. Figures 1, 2 and 3 present sample VOTmeasurements for a /p/, a positive VOT /b/, and a negative VOT (i.e., pre-voiced) /b/.

Intrajudge and interjudge reliability were assessed. All 32 subjects werelisted alphabetically by last name. A table of random numbers was used tochoose six subjects for reliability, with three subjects chosen for intrajudge re-liability and three for interjudge reliability. Intrajudge reliability measureswere completed by the first investigator, while interjudge reliability measures

Table 1.

Word List for Voice Onset Time Measurements

/p/ /b/ /t/ /d/ /k/ /g/

/i/ pea bee tea Dee key Gee/a/ pot bog tot dot cot got/u/ pooh boo too do coo goo

Page 6: Voice onset time in spanish-english bilinguals: early versus late learners of english

220 THORNBURG and RYALLS

were assessed by the second author and two other graduate students with con-siderable training and experience measuring VOT. All 54 tokens for each ofthe six subjects chosen for reliability were reanalyzed on the same system (54tokens

3

6 subjects

5

324 tokens reanalyzed). The original and reanalyzedVOT measurements for each token were then compared to determine if theydiffered by no more than 8 milliseconds. Comparisons revealed second mea-sures to be within 8 ms of the original measures 93% of the time for the intra-judge reliability; and 90% of the time for the interjudge reliability.

RESULTS

Tables 2 and 3 present VOT averages across vowels for each of the stops forthe “early” and “late” group, respectively. An ANOVA revealed significantmain effects for group [

F

(1,30)

5

15.06;

p

,

.0001], for gender [

F

(1,30)

5

121.58;

p

,

.0001], for voice [

F

(1,30)

5

2068.48;

p

,

.0001] for place-of-articulation [

F

(2,30)

5

67.44;

p

,

.0001]. Significant interactions were

Figure 1. An example of a VOT measure for a /p/ production.

Figure 2. An example of a measure for a positive VOT /b/ production.

Page 7: Voice onset time in spanish-english bilinguals: early versus late learners of english

VOICE ONSET TIME IN BILINGUALS 221

found for group-by-voice [

F

(1,30)

5

18.52;

p

,

.0001], and for voice-by-gender [

F

(1,30)

5

104.07;

p

,

.0001].

Main Effects

The main effect for group [

F

(1,30)

5

15.06;

p

,

.0001] is significant, thussupporting the hypothesis that the VOT values of native speakers of Spanishspeaking English differ according to whether initial learning of English beganprior to or after the age of 12 years. The mean differences in VOT values foreach cognate pair of stops was greater for those who learned English prior to

Figure 3. An example of a measure for a negative VOT (prevoiced) /b/ production.

Table 2.

Average Voice Onset Time: Pre-12 “Early” Group

Subject /p/ /t/ /k/ /b/ /d/ /g/

FE 01 105 105 109

2

3 33 37FE 02 74 83 79 9 23 26FE 03 85 95 96 18 31 24FE 04 25 77 80

2

99

2

118

2

83FE 05 84 114 99

2

45 21 13FE 06 46 78 67

2

69 14 19FE 07 81 102 104

2

36 11 14FE 08 70 82 85

2

26 21 19

ME 01 54 72 79

2

129

2

136

2

123ME 02 146 156 148

2

99

2

77

2

45ME 03 17 24 41

2

98

2

78

2

69ME 04 59 106 109

2

20

2

1 1ME 05 76 76 89

2

38 10 1ME 06 44 62 71

2

81

2

54 1ME 07 31 59 69

2

120

2

119

2

101ME 08 65 76 94

2

88

2

30 40

Page 8: Voice onset time in spanish-english bilinguals: early versus late learners of english

222 THORNBURG and RYALLS

age 12, than for those who began to learn English at age 12 or later. Table 4gives group VOT differences for each cognate pair.

Early learners of English appear to contrast VOT to a greater degree thando late learners. That is, those who learned English prior to age 12 appear todifferentiate voiceless from voiced stops to a greater degree than do those whobegan to learn English after this age.

The significant main effect of gender can be seen in Table 5. Males in bothgroups contrasted voicing (i.e., voiceless versus voiced VOT values) to agreater degree than did females. Swartz (1992) also found a significant effectof gender in VOT productions of alveolar stops in monolingual speakers ofEnglish.

Females may employ other acoustic cues to voicing to a greater degree.VOT or related measures such as aspiration cue voicing contrast for syllable-initial stops (Kent & Read, 1992). More females than males in either group as-

Table 3.

Voice Onset Time: Post-12 “Late” Group

Subject /p/ /t/ /k/ /b/ /d/ /g/

FL 01 62 65 71

227 212 35FL 02 13 41 63 244 29 31FL 03 64 68 108 15 34 46FL 04 52 64 84 230 25 29FL 05 34 75 70 216 10 30FL 06 41 40 51 246 264 215FL 07 39 56 74 269 297 267FL 08 36 67 81 245 225 37

ML 01 61 62 82 275 11 2ML 02 27 53 69 298 247 225ML 03 103 105 119 256 44 18ML 04 41 40 71 2149 2139 2119ML 05 42 56 60 254 269 216ML 06 61 75 78 1 247 239ML 07 29 41 59 258 262 250ML 08 52 64 76 293 286 286

Table 4. Average Voice Onset Time Differences by Group (in ms)

Cognate pairs

Group /p/ vs. /b/ /t/ vs. /d/ /k/ vs. /g/

Pre-12 125 114 108Post-12 101 96 88

Page 9: Voice onset time in spanish-english bilinguals: early versus late learners of english

VOICE ONSET TIME IN BILINGUALS 223

pirated voiced tokens. Eighty-eight percent of females in both groups (i.e., 7out of 8 subjects in each group) aspirated one or more voiced stop, comparedto 38% of males in the pre-12 group (i.e., 3 out of 8 subjects) and 63% ofmales in the post-12 group (i.e., 5 out of 8 subjects). The majority of aspiratedvoiced tokens were /di/, /dɑ/, /du/; /gi/, /gɑ/, /gu/. Only one subject aspirated/bɑ/. No subject aspirated /bi/ or /bu/.

A possible explanation for the presence of both prevoicing and aspirationmay be the expenditure of extra effort to enhance the contrastiveness of thevoiceless-voiced distinction (Nathan, 1987).

The main effect of place [F (2.30) 5 67.44; p , .0001] was significant. Asdisplayed in Table 6, VOT values for each separate voiceless or voiced cate-gory get larger with more posterior the place of articulation. This finding isconsistent with findings by Lisker & Abramson (1964), Zlatin (1974), andVolaitis and Miller (1992). Also, mean VOT value differences for each cog-nate pair got smaller with the more posterior place of articulation (see Table6). There was more contrast between /p/, /b/ (VOT difference: 114 ms), lessbetween /t/, /d/ (VOT difference: 106 ms), and least between /k/, /g/ (VOT dif-ference: 99 ms).

The main effect of voice [F (1,30) 5 2068.48; p , .0001] can be explainedby comparisons of voiced tokens by pre-12 and post-12 learners of English.That the earlier learners’ voiced productions had less negative VOT valuesthan those of late learners (See Table 6) appears true for /d/ tokens (228 mscompared with 235 ms), but not true for /b/ (258 ms compared with 253 ms)or /g/ (219 ms compared with 212 ms). Both groups’ productions for /g/ to-kens were least negative and were the only voiced productions within the 220to 120 millisecond range for English voiced stops (Kent & Read, 1992). That

Table 5. Voice Onset Time Differences by Gender (in ms)

Subject /p/ vs. /b/ /t/ vs. /d/ /k/ vs. /g/

Females 89 84 71Males 136 126 125

Table 6. Voice Onset Time Comparisons by Place of Articulation

Bilabials Alveolars Velars

Group /p/ /b/ /t/ /d/ /k/ /g/

Pre-12 67 258 86 228 89 219Post-12 48 253 61 235 76 212All subjects 58 256 74 232 83 216

Page 10: Voice onset time in spanish-english bilinguals: early versus late learners of english

224 THORNBURG and RYALLS

/g/ productions fell within the range for English voiced stops is consistent withfindings by Nathan (1987).

InteractionsInteractions of group by voice [F (1.30) 5 18.52; p , .0001], voice by gender[F (1.30) 5 104.07; p , .0001], and voice by vowel [F (2.30) 5 11.44; p ,.0001] were significant. The group by voice interaction indicates that the pre-12 group had less negative (i.e., more English-like) mean VOTs only for /d/productions. Mean VOT values of both groups for /g/ tokens were the onlyproductions within the 220 to 120 ms range for English voiced stops (Kent& Read, 1992), although the /d/ productions by the pre-12 group, at a meanVOT value of 228, were just outside this range.

Table 7 displays the interaction voice by gender. Males, as a group, havemore negative VOT values than do females. None of the voiced productionsby males fell within the 220 to 120 ms range. Mean VOT values for produc-tions of /d/ and /g/ by females fell within the 220 to 120 ms range.

The voice by vowel interaction indicates that /d/ followed by /u/, and /g/followed by either /i/, /ɑ/, or /u/ produced productions in which VOT fellwithin the 220 to 120 ms range for English voiced stops. For the voicelessstops, /p/ and /t/ had the longest mean VOT values when followed by /u/, /i/,and /ɑ/, in that order. For /k/ tokens, /i/ produced slightly longer mean VOTvalues than did /u/, at 91 ms versus 89 ms. /ɑ/ tokens had the least long meanVOT values at 69 ms.

Furthermore, /i/ and /u/ resulted in longer mean VOT times than did /ɑ/ forthe stops preceding them regardless of place of articulation. This finding isconsistent with that of Klatt (1975), Weismer (1979), and Flege (1991) forvoiceless stops.

DISCUSSION The results of this study disagree with the assertion by Flege (1992a) that pre-12 learners, but not those whose learning begins at or after this age, can “au-thentically” produce English voiceless stops, according to his operational def-inition. Twenty-nine out of 32 subjects had mean VOT values for each of the

Table 7. Interaction Voice by Gender

Gender /b/ /d/ /g/

All males 279 255 242All females 232 28 12

Page 11: Voice onset time in spanish-english bilinguals: early versus late learners of english

VOICE ONSET TIME IN BILINGUALS 225

three voiceless stops that fell within the range for English voiceless stops (i.e.,125 to 1100 ms: Kent & Read, 1992).

All 32 subjects in this study produced aspiration for at least 77% of the /p/tokens, for at least 98% of the /t/ tokens, and for 100% of the /k/ tokens. Flege& Eefting (1988) found that Spanish bilinguals were able to produce the En-glish voiceless stop /t/ with aspiration. The results of this study extend thesefindings.

Pre-12 learners had more contrast in VOT than did the post-12 group. Itdoes not necessarily follow, however, that productions by early learners werenecessarily more authentic than those by late learners. Keating (1984) hypoth-esized that monolingual speakers of languages such as Polish, French, andSpanish “attempt to enhance the contrast” between prevoiced /b/, /d/, /g/ andvoiceless unaspirated stops /p/, /t/, /k/. English monolinguals contrast voicedstops that have VOTs in a 40 ms range around the burst (i.e., from 220 to120 ms), and voiceless stops in the 125 to 1100 ms range. /p/, /t/, /k/ in ini-tial position are typically aspirated in English. It may be that the most authen-tic productions of English stops by native speakers of Spanish have VOTs forvoiced stops in the 220 to 120 ms range that contrast with VOTs for voice-less stops in the 125 to 1100 ms range.

An examination of individual data revealed that 84%, or 27 out of 32 sub-jects (16 females, 11 males), had at least one token that had VOT contrasts be-tween the target ranges. These tokens would meet Flege’s (1984) criteria forauthentic productions (i.e., compliant with English phonetic norms).

A Spearman rank coefficient of correlation (r 5 2.13178, p . .305) indi-cated no significant correlation between age of initial learning of English andthe overall mean amount of VOT difference. Although the pre-12 group con-trasted VOT to a greater degree than did the post-12 group, other factors mayinfluence the amount of second language proficiency. The motivation to useEnglish relative to use of Spanish (Yeni-Komshian, 1995), and the authentic-ity of phonetic models to be emulated (Flege, 1992a) are two such factors.Only 66% of subjects (i.e., 21 out of 32) estimated their daily usage of Englishto be at least 80%. The average estimated daily use of English by the pre-12group was 73%, with range from 50% to 90%; for the post-12 group the aver-age was 65%, with range from 25% to 90%. An analysis of variance revealedno significant difference between groups in voicing (e.g., most voiced tokenswere prevoiced by both groups).

Age of learning a second language appears to be one critical factor in sec-ond language acquisition. Other factors such as interaction with native speak-ers, motivation to learn the second language, and amount of daily usage alsomay play a role in second language proficiency. A limitation of the presentstudy is that it did not control for factors other than age.

This study compared productions of voiceless and voiced stops in 18 basicspeech syllables by two groups of Hispanic speakers of English: those whose

Page 12: Voice onset time in spanish-english bilinguals: early versus late learners of english

226 THORNBURG and RYALLS

initial learning of English began prior to age 12 with those who learned En-glish at or after this age. Several additional issues should be targeted by futureresearch: perception versus production, how second language proficiency af-fects first language production, and VOT measures of running speech com-pared to those of read CV syllables. However, this study represents anotherstep in the direction of more detailed acoustic studies of Spanish-Englishbilinguals.

This research was conducted as part of a master’s thesis research project by the first author, underthe supervision of the second author, in the Department of Communicative Disorders at the Uni-versity of Central Florida in Orlando. We are grateful to Ms. Julie Pepe of the Statistics Depart-ment of UCF for providing the statistical analyses. The authors express their appreciation to allwho assisted in this research project, especially to the members of the Hispanic American StudentAssociation and their adviser Ms. Justina Gonzales-Marti; to Mrs. Myrna Creasman (director ofthe university’s Intensive English Program) and her colleagues. We also thank all the subjectswho participated for their time and cooperation.

REFERENCESBradlow, A.R. (1995). A comparative study of English and Spanish vowels.

Journal of the Acoustical Society of America, 97(3), 1916–1924.

Flege, J.E. (1984). The detection of French accent by American listeners.Journal of the Acoustical Society of America, 76(3), 692–707.

Flege, J.E. (1988). The production and perception of speech sounds. In H.Winitz (Ed.), Human communication and its disorders: A review—1988(pp. 224–401). Norwood, NJ: Ablex Publishing Corporation.

Flege, J.E. (1991). Age of learning affects the authenticity of VOT in stopconsonants produced in a second language. Journal of the Acoustical Soci-ety of America, 89(1), 395–411.

Flege, J.E. (1992). Speech learning in a second language. In C.A. Ferguson, L.Mann, & C. Stoel-Gannon (Eds.), Phonological development: Models, re-search, implications (pp. 565–603). Timonium, MD: New York Press, Inc.

Flege, J.E., & Eefting, W. (1988). Imitation of a VOT continuum by nativespeakers of English and Spanish: Evidence for phonetic category formation.Journal of the Acoustical Society of America, 83(2), 729–740.

Flege, J.E., Munro, M.J., & MacKay, I.R. (1995). Effects of age of second-language learning on the production of English consonants. Speech Com-munication, 16(1), 1–26.

Gandour, J. (1980). Letter to the editor & ASHA response. Journal of Speechand Hearing Disorders, 45(1), 133–138.

Page 13: Voice onset time in spanish-english bilinguals: early versus late learners of english

VOICE ONSET TIME IN BILINGUALS 227

Goldberg, B. (1995). The transformation age: the sky’s the limit. ASHA,27(1), 45–48.

Ho, D.Y.F. (1986). Two contrasting positions of second-language acquisition:A proposed solution. International Review of Applied Linguistics, 24(1),35–47.

Kahane, J. (1978). A morphological study of the human prepubertal and pu-bertal larynx. American Journal of Anatomy, 151(1), 11–20.

Kayser, H. (1993). Hispanic cultures. In D.E. Battle (Ed.), Communicationdisorders in multicultural populations. Boston, MA: Andover Medical Pub-lishers.

Keating, P. (1984). Phonetic and phonological representation of stop conso-nant voicing. Language, 60(2), 286–319.

Kent, R.D., & Read, C. (1992). The acoustic analysis of speech. San Diego,CA: Singular Publishing Group, Inc.

Klatt, D. (1975). VOT, frication, and aspiration in word-initial consonantclusters. In R.J. Baken & R.G. Daniloff (Eds.), Readings in the clinicalspectrography of speech. San Diego, CA: Singular Publishing Group, Inc.

Langdon, H.W., & Merino, B.J. (1992). Acquisition and development of a secondlanguage in the Spanish speaker. In H.W. Langdon & L.-R.L. Cheng (Eds.),Hispanic children and adults with communication disorders: Assessment andintervention (pp. 132–167). Gaithersburg, MD: Aspen Publishers, Inc.

Lenneberg, E.H. (1967). Biological Foundations of language. New York, NY:Harcourt Brace.

Lisker, L., & Abramson, A. (1964). A cross-language study of voicing in ini-tial stops: Acoustical measurements. In R.J. Baken & R.G. Daniloff (Eds.),Readings in the Clinical Spectrography of Speech. San Diego, CA: SingularPublishing Group, Inc.

Mack, M. (1984). Early bilinguals: How monolingual are they? In M. Paradis& I. Lebrun (Eds.), Early Bilingualism and Child Development (pp. 161–173). Lisse, Netherlands: Swets and Zeitlinger, B.V.

Mertus, J. (1989). BLISS manual. Department of Cognitive and LinguisticSciences, Providence, RI: Brown University.

Nathan, G.S. (1987). On second-language acquisition of voiced stops. Journalof Phonetics, 15(3), 313–322.

Swartz, B.L. (1992). Gender differences in voice onset time. Perceptual andMotor Skills, 75, 983–992.

Page 14: Voice onset time in spanish-english bilinguals: early versus late learners of english

228 THORNBURG and RYALLS

Taylor, O. (1993). Foreword. In D.E. Battle (Ed.), Communication disordersin multicultural populations (pp. xii, xiii) Boston, MA: Andover MedicalPublishers.

Titze, I.R. (1989). Physiological and acoustic differences between male and fe-male voices. Journal of the Acoustical Society of America, 85(4), 1699–1707.

United States Department of Commerce, Bureau of the Census (1994). Statis-tical abstract of the United States. Washington, DC: United States Govern-ment Printing Office.

Volaitis, L.E., & Miller, J.L. (1992). Phonetic prototypes: Influence of placeof articulation and speaking rate on the internal structure of voicing catego-ries. Journal of the Acoustical Society of America, 92(2), 1:723–735.

Weismer, G. (1979). Sensitivity of VOT measures to certain segmental fea-tures in speech production. Journal of Phonetics, 7(2), 197–204.

Yeni-Komshian, G. (1995). What happens to our first language when we learna second language? Paper presented at the Annual Meeting of the AmericanAssociation for the Advancement of Science, Atlanta, GA, February 17, 1995.

Zlatin, M.A. (1974). Voicing contrast: Perceptual and productive VOT character-istics of adults. Journal of the Acoustical Society of America, 56(3), 981–994.

CONTINUING EDUCATION

Voice Onset Time in Spanish-English Bilinguals: Early versus Late Learners of English

QUESTIONS

1. Voice Onset Time relates to which of the following speech sound con-trasts?a. Consonants and vowelsb. Stops and fricativesc. Voiced and voicelessd. High versus lowe. Place versus manner

2. How are voiced stop consonants pronounced differently in Spanish thanthey are in English?a. They have a negative VOTb. They are “prevoiced”c. They have a positive VOTd. a and b abovee. None of the above

Page 15: Voice onset time in spanish-english bilinguals: early versus late learners of english

VOICE ONSET TIME IN BILINGUALS 229

3. How was VOT measured in this study?a. On a sound spectrographb. On a digital tape recorderc. On a computerd. On an intensity metere. On a stop watch

4. What was the difference found between early versus late learners of En-glish?a. Early learners produced less VOT contrastb. Early learners produced more VOT contrastc. Early learners produced English-like VOTd. Only early learners produced VOTe. Only late learners produced VOT

5. What were other factors affecting second language learning, besides age oflearning, discussed in the article:a. Social classb. Motivationc. Interaction with native speakersd. a and b abovee. b and c above