Linguistic structure co-determines how a speech sound is produced. This study therefore investigated whether the speaker-dependent information in the vowel [aː] varies when uttered in different... Show moreLinguistic structure co-determines how a speech sound is produced. This study therefore investigated whether the speaker-dependent information in the vowel [aː] varies when uttered in different word classes. From two spontaneous speech corpora, [aː] tokens were sampled and annotated for word class (content, function word). This was done for 50 male adult speakers of Standard Dutch in face-to-face speech (N = 3,128 tokens), and another 50 male adult speakers in telephone speech (N = 3,136 tokens). First, the effect of word class on various acoustic variables in spontaneous speech was tested. Results showed that [aː]s were shorter and more centralized in function than content words. Next, tokens were used to assess their speaker-dependent information as a function of word class, by using acoustic-phonetic variables to (a) build speaker classification models, and (b) compute the strength-of-evidence, a technique from forensic phonetics. Speaker-classification performance was somewhat better for content than function words, whereas forensic strength-of-evidence was comparable between the word classes. This seems explained by how these methods weigh between- and within-speaker variation. Because these two sources of variation co-varied in size with word class, acoustic word-class variation is not expected to affect the sampling of tokens in forensic speaker comparisons. Show less
The relative contributions of static and dynamic formant representations to speaker-specificity were investigated in conversational speech and in two vowels varyingin inherent spectral change.... Show moreThe relative contributions of static and dynamic formant representations to speaker-specificity were investigated in conversational speech and in two vowels varyingin inherent spectral change. Using polynomial fits, the contribution of dynamicformant coefficients to speaker-specificity relative to that of the formant interceptwas investigated in the diphthongal vowel [ei] taken from English and Dutch conversationalspeech. The [ei] tokens were sampled from various linguistic contextsand analysed in an LR approach. Results show that formant dynamics containspeaker-specific information in conversational speech even though the high contextualvariation seems to reduce its effect relative to that reported by earlier work.Vowels differ in inherent dynamicity and therefore, the added value of dynamicformant information to speaker-specificity was also compared between vowels differingin inherent spectral change. Using Dutch data, the contribution of formantdynamics to speaker-specificity was compared between [ei] and [aː] tokens producedby the same speakers. Formant dynamics in conversational speech only contributedto speaker-specificity in the diphthong [ei], not in the monophthong [aː]. Show less
Achtergrond en doelDe Nederlandse politie tapt gemiddeld meer dan duizend telefoons per dag af [1]. Forensisch spraakwetenschappers kunnen onderzoeken of de stem op de opname bij de verdachte past.... Show moreAchtergrond en doelDe Nederlandse politie tapt gemiddeld meer dan duizend telefoons per dag af [1]. Forensisch spraakwetenschappers kunnen onderzoeken of de stem op de opname bij de verdachte past. In tegenstelling tot andere biometrische gegevens is variatie binnen een stem inherent aan spraak: dezelfde mond moet immers allerlei verschillende klanken maken. Ook binnen klanken is er onbewust variatie in de uitspraak. En een spreker kan de klank bewust aanpassen. Een klank die relatief robuust lijkt te zijn voor dit soort invloeden is de aarzelingsklank ‘eh’ [2]. In deze studie onderzochten we in hoeverre verschillende taalkundige factoren invloed uitoefenen op de uitspraak van ‘eh’.Methode: data en analyses Ons onderzoek is gebaseerd op 2106 aarzelingen in spontane spraak van 59 Utrechtse studentes [3]. De studentes spraken in hun eerste taal (Nederlands) en in hun tweede taal (Engels). De aarzelingen zijn gemeten op akoestische variabelen zoals duur en toonhoogte. Vervolgens hebben we in R [4] per akoestische variabele linear mixed-effects models gebouwd om te onderzoeken of de akoestische vorm van aarzelingen wordt beïnvloed door: 1) de taal; 2) de variant (eh of ehm); 3) de positie; en 4) de spreker.Resultaten en conclusieSprekers verschilden van elkaar in hun uitspraak van ‘eh’ en ‘ehm’, maar de uitspraak bleek ook afhankelijk van de positie, de variant en de taal. Dit resultaat laat zien dat vergelijkend spraakonderzoek naast variatie tussen personen ook rekening te houden heeft met variatie binnen personen. Deze alertheid is ook geboden bij relatief onbewust geuite klanken zoals aarzelingen.Bronnen[1] Ministerie van Justitie en Veiligheid (2018). Jaarverslag en slotwet. Tweede Kamer, vergaderjaar 2018-2019, 35 200 VI, nr. 1.[2] Hughes, V., Foulkes, P., & Wood, S. (2016). Strength of forensic voice comparison evidence from the acoustics of filled pauses. International Journal of Speech, Language and the Law, 23, 99−132.[3] Orr, R., & Quené, H. (2017). D-LUCEA: Curation of the UCU Accent Project data. In: J. Odijk & A. van Hessen (red.), CLARIN in the Low Countries. Berkeley: Ubiquity Press (pp. 177−190).[4] R Core Team (2018). R: A language and environment for statistical computing. Wenen: R Foundation for Statistical Computing. . Show less
Although previous work has shown that some speech sounds are more speaker-specific than others, not much is known about the speaker information of the same segment in different linguistic contexts.... Show moreAlthough previous work has shown that some speech sounds are more speaker-specific than others, not much is known about the speaker information of the same segment in different linguistic contexts. The present study, there- fore, investigated whether Dutch fricatives /s/ and /x/ from telephone dialogues contain differential speaker informa- tion as a function of syllabic position and labial co-articulation. These linguistic effects, established in earlier work on read broadband speech, were first investigated. Using a corpus of Dutch telephone speech, results showed that the telephone bandwidth captures the expected effects of perseverative and anticipatory labialization for dorsal fricative /x/, for which spectral peaks fall within the telephone band, but not for coronal fricative /s/, for which the spectral peak falls outside the telephone band. Multinomial logistic regression shows that /s/ contains slightly more speaker information than /x/ in telephone speech and that speaker information is distributed across the speech signal in a sys- tematic way; even though differences in classification accuracy were small, codas and tokens with labial neighbors yielded higher scores than onsets and tokens with non-labial neighbors for both /s/ and /x/. These findings indicate that speaker information in the same speech sound is not the same across linguistic contexts. Show less
In several languages, filled pauses (i.e. uh, um) have been described as speaker-specific (e.g. Horváth, 2010; Hughes et al., 2016; Künzel, 1997). In ongoing work we evaluate the speaker... Show moreIn several languages, filled pauses (i.e. uh, um) have been described as speaker-specific (e.g. Horváth, 2010; Hughes et al., 2016; Künzel, 1997). In ongoing work we evaluate the speaker-specificity of filled pauses across languages, specifically in speakers with L1 Dutch and L2 English: do speakers maintain their hesitation behavior across languages?So far, filled pauses have been largely ignored in phonetic descriptions of Dutch, whereas they are highly frequent (De Jong, 2017). The most specific paper describes Dutch uh as a schwa (Swerts, 1998). Whereas the same description has been given to filled pauses in English (e.g. Hughes et al., 2016) and Hungarian (Horváth, 2010), phonetic measurements in these languages show between-language differences in mean formant values of over 100 Hz.The data collected to answer our main research question also allows for a phonetic description of Dutch filled pauses. We analyzed 2,962 tokens from 20 male and 59 female native speakers of Dutch. Measurements included type (i.e. uh or um), fundamental frequency, duration, and vowel formants.ReferencesDe Jong, N. (2017). Spreekvaardigheid en vloeiendheid. Vakwerk 11: Selectie uit de lezingen en presentaties conferentie BVNT2, Amsterdam 2017 (pp. 37-45).Horváth, V. (2010). Filled pauses in Hungarian: Their phonetic form and function. Acta Linguistica Hungarica, 57(2-3), 288-306.Hughes, V., Foulkes, P., & Wood, S. (2016). Strength of forensic voice comparison evidencefrom the acoustics of filled pauses. International Journal of Speech, Language and the Law, 23, 99-132.Künzel, H. J. (1997). Some general phonetic and forensic aspects of speaking tempo. International Journal of Speech, Language and the Law, 4(1), 48-83.Swerts, M. (1998). Filled pauses as markers of discourse structure. Journal of pragmatics, 30(4), 485-496. Show less
The idiosyncrasy of filled pauses in L2 English in the context of phonetic convergenceUniversity College Utrecht (UCU) brings together students from different language backgrounds, who use English... Show moreThe idiosyncrasy of filled pauses in L2 English in the context of phonetic convergenceUniversity College Utrecht (UCU) brings together students from different language backgrounds, who use English as a lingua franca. Only a small minority of the students speaks English as a native language (L1); the majority is Dutch (Orr & Quené, 2017). Prior studies have shown that the speakers in this multilingual community converge towards a shared accent of English: after three years on campus, students’ /s/ pronunciations have become more similar (Quené et al., 2017), as has their speech rhythm (Quené & Orr, 2014).This study aims to investigate how this convergence may affect the idiosyncrasy of non-native (L2) English filled pauses ‘uh’ and ‘um’. Filled pauses (FPs) are considered useful features in forensic speaker comparisons, since they are highly speaker-specific (e.g. Künzel, 1997). This not only means that there is variation in how FPs are used by different speakers, but also that they are a consistent feature in an individual’s speech (Braun & Rosin, 2015). Moreover, it is possible that speakers are consistent in using FPs in their L1 and L2, because FPs tends to not be explicitly taught. However, because languages differ in FP use (De Leeuw, 2007), ongoing practice in speaking the L2 may affect their realization, as well as the community’s converging language.To investigate the idiosyncrasy of filled pauses in the context of convergence, 20 Dutch female UCU students were selected from the LUCEA corpus (Orr & Quené, 2017). FPs were segmented from 2-minute English monologues, recorded at the beginning and the end of their Bachelor studies. Together, they produced 680 FPs. The first and second formant of the vowels in ‘uh’ and ‘um’ showed convergence – measured as reduced variance – whereas the third formant and F0 showed more variance over time. Linear mixed-effects models showed that only part of the students changed their FP realizations with time. While changes were minimal, they resulted in a drop in speaker classification performance in cross-time comparisons.References:Braun, A. & A. Rosin (2015). On the speaker-specificity of hesitation markers. Proc. 18th International Congress of Phonetic Sciences: 10-14.De Leeuw, E. (2007). Hesitation markers in English, German, and Dutch. Journal of Germanic Linguistics19: 85-114.Künzel, H.J. (1997). Some general phonetic and forensic aspects of speaking tempo. Forensic Linguistics4: 48-83.Hughes, V. et al. (2016). Strength of forensic voice comparison evidence from the acoustics of filled pauses. International Journal of Speech, Language and the Law 23: 99-132.Orr, R. & H. Quené (2017). D-LUCEA: Curation of the UCU Accent Project data. In Odijk, J. & A. Van Hessen (eds.): CLARIN in the Low Countries: 177-190. London: Ubiquity Press.Quené, H., & R. Orr (2014). Long-term convergence of speech rhythm in L1 and L2 English. Social and Linguistic Speech Prosody 7: 342-345.Quené, H. et al. (2017). Phonetic similarity of /s/ in native and second language: Individual differences in learning curves. The Journal of the Acoustical Society of America 142: 519-524. Show less
Ditewig, S.; Heeren, W.F.L.; Smorenburg, B.J.L.; Quené, H. 2019
Individual variation in filled pauses in the native and second language Meike de Boer & Willemijn Heeren Leiden University Centre for Linguistics Hesitation behavior is a relatively unconscious... Show moreIndividual variation in filled pauses in the native and second language Meike de Boer & Willemijn Heeren Leiden University Centre for Linguistics Hesitation behavior is a relatively unconscious part of language [1], which shows much between-speaker variation [2−4]. Furthermore, individuals are rather consistent in how they hesitate in their native language [2, 5]. This study investigates between-speaker variation in hesitation behavior in the first (L1) and second (L2) language, and within-speaker consistency of filled pauses across languages. In Dutch and English, two fillers are mainly used to express hesitation: uh and um. However, their exact phonetic realization and the ratio between the two are different for these languages [6, 7]. Flege’s Speech Learning Model [8] says that L2 learners only adapt their pronunciation when they perceive a difference between the L1 and L2. Therefore, we expect that Dutch speakers of English more clearly adapt their uh:um proportions than their vowel formants of the uh/um vowels. For other pronunciation features of uh and um, e.g. duration and fundamental frequency (F0), we expect speakers to be consistent across languages [9, 2]. We investigated the speech of 40 Dutch students of University College Utrecht (20 females; 20 males). The speakers were selected from the Longitudinal Corpus of University College English Accents (LUCEA), collected by Orr and Quené [10]. Students from University Colleges have advanced L2 proficiency. Preliminary results show substantial between-speaker variation in the filled pauses uh and um in both Dutch and English. The within-speaker consistency was low where expected: when speaking English, students used the um variant more often than in Dutch. Also, the vowel quality of their filled pauses was pronounced more open and more backwards in English than in Dutch. According to the SLM, this suggests that differences in vowel realization between Dutch and English were sufficiently salient to these speakers, as were the different uh:um ratios. As expected, filled pauses’ durations and F0 remained relatively stable across languages. References [1] Clark, H. H., & Fox Tree, J. E. (2002). Using uh and um in spontaneous speaking. Cognition, 84, 73−111. [2] Braun, A., Rosin, A. (2015). On the speaker-specificity of hesitation markers. Proc. 18th ICPhS Glasgow, 731−736. [3] Hughes, V., Wood, S., & Foulkes, P. (2016). Filled pauses as variables in forensic voice comparison. Int. J. Speech Lang. Law, 23, 99−132. [4] McDougall, K., & Duckworth, M. (2017). Profiling fluency: An analysis of individual variation in disfluencies in adult males. Speech Comm., 95, 16−27. [5] Künzel, H. F. (1997). Some general phonetic and forensic aspects of speaking tempo. For. Linguist., 4, 48−83. [6] De Leeuw, E. (2007). Hesitation markers in English, German, and Dutch. J. Germ. Ling. 19, 85–114. [7] Wieling, M., Grieve, J., Bouma, G., Fruehwald, J., Coleman, J., & Liberman, M. (2016). Variation and change in the use of hesitation markers in Germanic languages. Language Dynamics and Change, 6, 199−234. [8] Flege, J. E. (1995). Second language speech learning: Theory, findings, and problems. In: Strange, W. (Ed.), Speech Perception and Linguistic Experience: Issues in Cross-Language Research. York: York Press, 233−277. [9] Kolly, M. J., Leemann, A., De Mareüil, P. B., & Dellwo, V. (2015). Speaker-idiosyncrasy in pausing behavior: Evidence from a cross-linguistic study. Proc. 18th ICPhS Glasgow, 294−299. [10] Orr, R., & Quené, H. (2017). D-LUCEA: Curation of the UCU Accent Project data. In: Odijk, J., & Van Hessen, A. (Eds.), CLARIN in the Low Countries. London: Ubiquity Press, 177–190. Show less
We investigated the speaker-specificity of filled pauses across languages and time. The filled pauses uh and um contain speaker-specific information in a speaker’s native language. Since speakers... Show moreWe investigated the speaker-specificity of filled pauses across languages and time. The filled pauses uh and um contain speaker-specific information in a speaker’s native language. Since speakers are relatively unaware of their hesitation behavior, it might transfer from their first (L1) to their second language (L2). We examined filled pauses using several phonetic-acoustic features in spontaneous L1 Dutch and L2 English speech of 20 female speakers, recorded at two times, three years apart.Using linear mixed-effects models, we found that speakers differ in their first and second formants of uh and um in L1 versus L2, while duration and fundamental frequency remain stable. Speaker classification models trained on filled pauses in one language perform worse – but still relatively well – on the other language. With the exception of a few speakers, hesitation behavior remained stable over time. In spite of L1-L2 differences, some speaker characteristics of filled pauses remain. Show less
The phoneme /h/ is absent in French and its acquisition has been described as being difficult for second language learners of Dutch, a language with /h/ in its phoneme inventory. In this study,... Show moreThe phoneme /h/ is absent in French and its acquisition has been described as being difficult for second language learners of Dutch, a language with /h/ in its phoneme inventory. In this study, several factors were examined that may affect the production of /h/ by Belgian-French learners of Dutch. Specifically, the factors included in this exploratory study were (1) L1-to-L2 transfer, (2) semantic contrastiveness, (3) the monitoring of one’s speech, and (4) educational grade. L1-to-L2 transfer was operationalized as the effect of liaison/elision contexts on /h/-production. The expectation was liaison contexts might transfer and would therefore hinder /h/-production. Semantic contrasts in minimal pairs including an h-initial word would elicit more /h/-productions if that word was contrasted with an empty onset than an onset (oor-hoor) filled by some other consonant (hand-tand). If a speaker pays more attention to his/her speech in an increased-monitoring task, the speaker is expected to produce /h/ more often, and finally it was expected that increased exposure to Dutch would result in more correct productions.In a cross-sectional study, students from the first, third and sixth grades of secondary education (60 in total, aged between 12 years and 19 years old) took part in two reading-aloud tasks, which were assumed to differ in the degree of speech monitoring they require. The first task was a text, with which L1-to-L2 transfer was assessed, and the second a list of minimal pairs containing h-onsets contrasting with either empty or filled onsets. Monitoring was assessed by comparing results between reading tasks.Results showed that increased monitoring positively influenced the numbers of [h]s produced, but that L1-to-L2 transfer of liaison/elision contexts did not occur. A small difference between conditions was found, but in the opposite direction. There was large between-learner variability and no performance increase with amount of exposure from first to sixth grade. Overall, performance left much room for improvement relative to native Dutch speakers and to the learners’ teacher. Further research is needed to better understand the development of French-speaker learners’ production of Dutch /h/. Show less
We introduce a targeted language game approach using the visual world, eye-movement paradigm to assess when and how certain intonational contours affect the interpretation of utterances. We created... Show moreWe introduce a targeted language game approach using the visual world, eye-movement paradigm to assess when and how certain intonational contours affect the interpretation of utterances. We created a computer-based card game in which elliptical utterances such as "Got a candy" occurred with a nuclear contour most consistent with a yes-no question (H* H-H%) or a statement (L* L-L%). In Experiment 1 we explored how such contours are integrated online. In Experiment 2 we studied the expectations listeners have for how intonational contours signal intentions: do these reflect linguistic categories or rapid adaptation to the paradigm? Prosody had an immediate effect on interpretation, as indexed by the pattern and timing of fixations. Moreover, the association between different contours and intentions was quite robust in the absence of clear syntactic cues to sentence type, and was not due to rapid adaptation. Prosody had immediate effects on interpretation even though there was a construction-based bias to interpret "got a" as a question. Taken together, we believe this paradigm will provide further insights into how intonational contours and their phonetic realization interact with other cues to sentence type in online comprehension. Show less
