1. Voice Onset Time as a Parameter for Identification of Bilinguals Claire Gurski University of Western Ontario London, ON Canada

2. Introduction A recording may be made in one language A recording may be made in one language Suspect refuses to speak that language Suspect refuses to speak that language Can voice onset time be used as a parameter for identification when cross language comparison materials are presented?

3. Background Voice Onset Time (VOT): ‘The interval between the release of a stop consonant occlusion and the onset of the vocal- fold vibration’ and is measured ‘from acoustic displays as the time between the release burst and the first quasi- periodicity in the acoustic signal’. (Keating, 1984) Voice Onset Time (VOT): ‘The interval between the release of a stop consonant occlusion and the onset of the vocal- fold vibration’ and is measured ‘from acoustic displays as the time between the release burst and the first quasi- periodicity in the acoustic signal’. (Keating, 1984)    Burst VOT Beginning of Vocal-fold vibration

4. Voice Onset Time

5. VOT – Language Specific The VOT of stop consonants [p b t d k g] differs between languages The VOT of stop consonants [p b t d k g] differs between languages ‘Plus/minus values of the voicing feature will have different quantitative VOT values in different languages.’ (Keating, 1984) ‘Plus/minus values of the voicing feature will have different quantitative VOT values in different languages.’ (Keating, 1984) Eg. [p] in French has a VOT similar to a [b] in English Eg. [p] in French has a VOT similar to a [b] in English Language Voiceless Plosives Voiced Plosives English Positive VOT Zero VOT French Negative VOT

6. VOT and Identification? Does a person who learns a second language acquire a native-like VOT in their second language or does their native language have an influence? Does a person who learns a second language acquire a native-like VOT in their second language or does their native language have an influence? Does a person have a consistent VOT across languages regardless of differences between the two languages? Does a person have a consistent VOT across languages regardless of differences between the two languages? Is VOT a factor that can be used to help identify voices as being from the same speaker even if they are in two different languages? Is VOT a factor that can be used to help identify voices as being from the same speaker even if they are in two different languages?

7. Methodology - 2 studies Study 1 Eight bilingual speakers – 4 Native French and 4 Native English Eight bilingual speakers – 4 Native French and 4 Native English 2 paragraphs – one beginning in English and ending in French and one beginning in French and ending in English 2 paragraphs – one beginning in English and ending in French and one beginning in French and ending in English Each paragraph read 4 times Each paragraph read 4 times Participants did not have the influence of other languages. Participants did not have the influence of other languages.

8. Participants – Study 1 Participant Number First Language Second Language 20FrenchEnglish 21FrenchEnglish 22EnglishFrench 23EnglishFrench 24FrenchEnglish 25FrenchEnglish 26EnglishFrench 27EnglishFrench

9. Analysis – Study 1 Only intervocalic stops were analyzed to eliminate the differences due to aspirated or unreleased stops Only intervocalic stops were analyzed to eliminate the differences due to aspirated or unreleased stops We found minimal variation between the languages in each of the paragraphs ie. English of the first paragraph and English of the second paragraph We found minimal variation between the languages in each of the paragraphs ie. English of the first paragraph and English of the second paragraph Stats calculated based on the language of use – not its position in relation to the other language Stats calculated based on the language of use – not its position in relation to the other language

10. Study 1 - Bilingual VOT

11. Hybrid System For 7 out of 8 of the speakers, their VOT does not pattern with either the French or the English systems For 7 out of 8 of the speakers, their VOT does not pattern with either the French or the English systems Participants reacted similarly regardless of native language Participants reacted similarly regardless of native language Hybrid system created Hybrid system created Language Voiceless plosives Voices plosives English Positive VOT Zero VOT French Negative VOT Bilingual Positive VOT Negative VOT

12. Hybrid VOT System of Bilinguals French voiced voiceless English voiced voiceless VOT positive VOT zero VOT negative bilinguals

13. VOT and variation between languages T scores – do the VOTs of the two languages of Bilingual individuals differ enough the be significantly different/distinct in the two languages? T scores – do the VOTs of the two languages of Bilingual individuals differ enough the be significantly different/distinct in the two languages? Does their VOT pattern similarly in both languages? Does their VOT pattern similarly in both languages?

14. T test significance scores Participantpbtdkg 20 – Fr/Eng 0.5600.1550.0550.6730.3320.156 21 – Fr/Eng 0.0800.6480.3280.5340.6780.084 22 – Eng/Fr 0.1120.0760.0000.0050.1100.450 23 – Eng/Fr 0.3470.2800.0040.3290.9610.686 24 – Fr/Eng 0.8490.3200.0140.2190.9540.714 25 – Fr/Eng 0.8380.9540.4630.1580.3320.052 26 – Eng/Fr 0.1560.0000.2280.4260.0040.001 27 – Eng/Fr 0.0750.1100.6290.5530.147N/A Generally, the difference between the French and English VOT of Bilingual speakers does not vary considerably when they change languages

15. Questions to address How does the VOT of Bilingual speakers compare to that of Monolingual speakers How does the VOT of Bilingual speakers compare to that of Monolingual speakers Are there significant differences statistically that would allow for identification? Are there significant differences statistically that would allow for identification?

16. Methodology - 2 studies Study 2 4 Monolingual French Speakers 4 Monolingual French Speakers 4 Monolingual English Speakers 4 Monolingual English Speakers Read ‘The North Wind and the Sun’ 2 times in their native language Read ‘The North Wind and the Sun’ 2 times in their native language Described a comic for spontaneous production (not analyzed for this study) Described a comic for spontaneous production (not analyzed for this study)

17. Participants – Study 2 The VOT of the Monolingual speakers was compared to that of the Bilingual speakers from Study 1 The VOT of the Monolingual speakers was compared to that of the Bilingual speakers from Study 1 ParticipantLanguageParticipantLanguage 1French16English 2French17English 6French18English 7French19English

18. Analysis – Study 2 Only intervocalic stops were analyzed to eliminate the differences due to aspirated or unreleased stops Only intervocalic stops were analyzed to eliminate the differences due to aspirated or unreleased stops Smaller samples were recorded decreasing the number of tokens Smaller samples were recorded decreasing the number of tokens Controlled environment with Marantz recording equipment Controlled environment with Marantz recording equipment

19. Study 2 – Monolingual VOT The VOT of Monolingual speakers follows the general patterns of French and English The VOT of Monolingual speakers follows the general patterns of French and EnglishLanguage Voiceless Plosives Voiced Plosives English Positive VOT Zero VOT French Negative VOT

20. Statistical Analysis ANOVA tests comparing the groups – Monolingual English, French and Bilingual French/English and English/French ANOVA tests comparing the groups – Monolingual English, French and Bilingual French/English and English/French Are the groups significantly distinct? Are the groups significantly distinct?

21. ANOVA – Four Groups Eng, Fr, Eng/Fr, Fr/Eng All are significant (except g – but only a limited number of tokens) All are significant (except g – but only a limited number of tokens) Therefore, there is a difference between the VOT of English, French, English/French Bilinguals and French/English Bilinguals Therefore, there is a difference between the VOT of English, French, English/French Bilinguals and French/English Bilinguals FactorSignificance p.049 b.000 t.015 d.051 k.025 g.100

22. ANOVA – English Monolinguals with French Monolinguals The difference in VOT of English and French Monolinguals is significant in all cases The difference in VOT of English and French Monolinguals is significant in all cases There were no occurrences of intervocalic p so inferences are made based on the other voiceless plosives There were no occurrences of intervocalic p so inferences are made based on the other voiceless plosives FactorSignificance b.000 t.004 d.001 k.052 g.034

23. ANOVA – Fr/Eng Bilinguals with Eng/Fr Bilinguals The averages of their French and English tokens. The averages of their French and English tokens. The only significant is b but if you look at the numbers, only participant 22 deviated and this would cause the significance. The only significant is b but if you look at the numbers, only participant 22 deviated and this would cause the significance.-.051924-.054261-.008036-.033747-.046522-.054442-.027603-.024800 Because there are no other significant differences this shows that the bilingual speakers pattern similarly regardless of first language Because there are no other significant differences this shows that the bilingual speakers pattern similarly regardless of first language FactorSignificance p.216 b.003 t.906 d.428 k.511 g.290

24. ANOVA – French and English of Fr/Eng and Eng/Fr Bilinguals There are no significant differences between the French and English of Bilingual speakers There are no significant differences between the French and English of Bilingual speakers Therefore the VOT of Second language learners does not vary considerably between their two languages Therefore the VOT of Second language learners does not vary considerably between their two languages FactorSignificance p.195 b.283 t.251 d.990 k.508 g.968

25. ANOVA – Eng with Eng/Fr Bilinguals The English Monolingual Speakers are not significantly different from the average of the English/French Bilinguals The English Monolingual Speakers are not significantly different from the average of the English/French Bilinguals FactorSignificance b.045 t.564 d.790 k.324 g.895

26. ANOVA – Eng with Eng of Eng/Fr Bilinguals The English of English/French Bilingual speakers is not significantly different from the English of monolingual speakers The English of English/French Bilingual speakers is not significantly different from the English of monolingual speakers FactorSignificance b.460 t.322 d.682 k.901 g.586

27. ANOVA – Eng with Eng of Fr/Eng Bilinguals The voiced stops are significantly different between English monolingual speakers and the English of native French speakers. The voiced stops are significantly different between English monolingual speakers and the English of native French speakers. French/English bilingual speakers do not acquire a native like English VOT French/English bilingual speakers do not acquire a native like English VOT If offender has a native like English VOT and suspect does not, may be an indication towards a non-identification If offender has a native like English VOT and suspect does not, may be an indication towards a non-identification FactorSignificance b.001 t.692 d.051 k.986 g.005

28. ANOVA – Fr with Fr/Eng Bilinguals French with Fr/Eng Bilinguals is generally significant but English with Eng/Fr Bilinguals is not French with Fr/Eng Bilinguals is generally significant but English with Eng/Fr Bilinguals is not The French Bilingual speakers vary from the monolingual French pattern which is not the case for their English counterparts. The French Bilingual speakers vary from the monolingual French pattern which is not the case for their English counterparts. FactorSignificance p.009 b.007 t.008 d.074 k.038 g.305

29. ANOVA – Fr with Fr of Fr/Eng Bilinguals In each case (disregard g due to limited tokens) the French VOT of the bilinguals is significantly different from the French of Monolingual speakers In each case (disregard g due to limited tokens) the French VOT of the bilinguals is significantly different from the French of Monolingual speakers This is to be expected since Bilingual Fr/Eng speakers range from Positive to negative in their VOT This is to be expected since Bilingual Fr/Eng speakers range from Positive to negative in their VOT FactorSignificance p.006 b.014 t.002 d.035 k.032 g.161

30. ANOVA – Fr with Fr of Eng/Fr Bilinguals Again we have significant differences between the French of Monolingual speakers and the French of Second language speakers Again we have significant differences between the French of Monolingual speakers and the French of Second language speakers [d] and [g] are not significant but again, participant 22 who is the only outlier may be causing this insignificance [d] and [g] are not significant but again, participant 22 who is the only outlier may be causing this insignificance FactorSignificance p.006 b.008 t.027 d.135 k.007 g.092

31. Conclusions There is still much research to be done before VOT can be deemed useful in a forensic situation There is still much research to be done before VOT can be deemed useful in a forensic situation Use of Hybrid system regardless of native language Use of Hybrid system regardless of native language Language Voiceless plosives Voices plosives English Positive VOT Zero VOT French Negative VOT Bilingual Positive VOT Negative VOT

32. Conclusions Generally, VOT does not vary significantly between the languages of Bilingual speakers Generally, VOT does not vary significantly between the languages of Bilingual speakers This is to say that the VOTs of a /p/ spoken by a bilingual speaker in French or English will have not be significantly different This is to say that the VOTs of a /p/ spoken by a bilingual speaker in French or English will have not be significantly different There are significant differences between the languages of Bilingual speakers and that of their Monolingual counterparts There are significant differences between the languages of Bilingual speakers and that of their Monolingual counterparts

33. Areas of Further Research Continued Statistical Analysis Continued Statistical Analysis Age of language learning Age of language learning Level of fluency Level of fluency Multilingual or bilingual in other languages Multilingual or bilingual in other languages Effects of nervousness Effects of nervousness Reading vs. Spontaneous speech Reading vs. Spontaneous speech Effects of the telephone Effects of the telephone Canadian French affrication Canadian French affrication Larger number of participants with similar voice qualities Larger number of participants with similar voice qualities Vocal Disguise Vocal Disguise