sign in sign up
<<
Home , Diasystem, Rhotic consonant, Voiceless alveolar trill

‘-atics 6 Paris 7th, 8th november 2019

‘R-atics6 [paʁi] 7th - 8th November 2019

‘R-atics 6 Paris 7th, 8th november 2019

Contents ‘R-atics6 Presentation ...... 3 Location of the conference ...... 4 Accomodation ...... 4 Transportation ...... 4 Thursday 7th: Map of La Maison de la Recherche ...... 5 Friday 8th: Map of Université Sorbonne Nouvelle – Paris 3 ...... 6 Program ...... 7 Abstracts ...... 9 Plenary 1 ...... 9 session – Thursday 7th ...... 9 Plenary 2: ...... 18 Sound Change Session – Thursday 7th ...... 18 Poster Session – Thursday 7th...... 29 Plenary 3 ...... 38 Production Session – Friday 8th ...... 38 Plenary 4 ...... 51 Acquisition Session – Friday 8th...... 52

‘R-atics 6 Paris 7th, 8th november 2019

‘R-atics6 Presentation

The Laboratory of and Phonology of the Université Sorbonne nouvelle – Paris 3 and CNRS organizes the ’R-atics 6 colloquium in Paris the 7th and 8th November 2019. ‘R-atics gathers an international network of researchers working on different issues concerning r sounds and has been previously organized in Nijmegen (The Netherlands), Bruxelles (Belgium), Bozen (Italy), Grenoble (France) and Leeuwarden (Nethrelands). This is an international mid-size specific conference that enhances mutual cooperation in a friendly environment on phonetic, phonological and sociolinguistic aspects of rhotics. The 2019 edition will propose to emphasize, in addition to the usual themes, r's in indigenous and the automatic treatment of rhotic variation.

Invited Speakers  Martine Adda-Decker : LPP, CNRS-UMR 7018, Sorbonne nouvelle & Limsi CNRS  Jim Scobbie: Queen Margaret College, Edinburgh  Koen Sebregts: University of Utrecht  Moges Yigezu: University of Addis Ababa Organizing Comitee  Didier Demolin : LPP, CNRS-UMR 7018, Sorbonne nouvelle  Alexis Dehais Underdown : LPP, CNRS-UMR 7018, Sorbonne nouvelle  Cédric Gendrot : LPP, CNRS-UMR 7018, Sorbonne nouvelle  Naomi Yamaguchi : LPP, CNRS-UMR 7018, Sorbonne nouvelle

You will find here all the information concerning the place, the organization, the program and the participants’ abstracts.

‘R-atics 6 Paris 7th, 8th november 2019

Location of the conference

The conference will take place at a different place each day:  Thursday 7th: La maison de la recherche of the Université Sorbonne Nouvelle – Paris 3 located at 4 rue des Irlandais, 75005, Paris, France. All talks will take place in the room Claude Simon.  Friday 8th: Université Sorbonne Nouvelle – Paris 3 located at 13, rue de Santeuil, 75005, Paris, France. All talks will take place in Room 245 (2nd floor). You will find a map of the two sites and neighborhood on the next page.

Accomodation

Paris offers a lot of accomodations. We recommend you Hôtel Saint Pierre, Hôtel Minerve or Hôtel Maxim that are located near the two conference sites. You can look for a place on Air BnB.

Transportation

Paris has a lot of transports: buses, subway, RER (=suburban train), trams. You can easily purchase tickets (1 for 1,90€; 10 for 14,90€) at a RATP subway desk. To go to: 1) La Maison de la recherche you can:  Take the subway line 10 and stop at Cardinal Lemoine station and walk (10 minutes)  Take the subway line 7 and stop at Place Monge station and walk (10 minutes)  Take the RER B and stop at Luxembourg station and walk (10 minutes) 2) Université Sorbonne Nouvelle – Paris 3 you can:  Take the subway line 7 and stop at Censier Daubenton station and walk (5 minutes)  Take the subway line 10 or RER C and stop at Gare d’Austerlitz Station and walk (10minutes)

‘R-atics 6 Paris 7th, 8th november 2019

Thursday 7th: Map of La Maison de la Recherche

Metro 10

RER B

‘R-atics 6 Paris 7th, 8th november 2019

Friday 8th: Map of Université Sorbonne Nouvelle – Paris 3

Maison de la recherche

Metro 10 RER C

Metro 7 ‘R-atics 6 Paris 7th, 8th november 2019

Program

Thursday 7th (Maison de la recherche) 9h Opening Speech 9h15 Moges Yigezu : Rhotics in Ethiopian languages

Variation - Chair Person : Hans Van de Velde 10h15 Ongoing change in post-independence São Tomé: The use of rhotics as a marker of national identity among young speakers of Santomean Portuguese Marie-Eve Bouchard 10h35 The “rolled” /R/ in French and some regional languages of France Timothée Premat & Philippe Boula de Mareüil 10h55 A sociophonetic exploration of /r/ in the South Cape Yolandi Ribbens-Klein 11h15 Coffee Break

11h45 Uvular and apical rhotic distribution in a basque speaking valley between the spanish and french states Alexander Arcelus Múgica 12h05 A large-scale automatic acoustic analysis of rhoticity in Jane Stuart-Smith & Rachel Macdonald 12h25

Lunch Break

14h Koen Sebregts : That is what we R: phonetics and phonology in rhotic identity Sound Change - Chair Person : Cédric Gendrot 15h What aʁʁʁ you? Reconsidering François Falc’hun’s French /ʁ/ hypothesis through the romance diasystem Fabian ZUK 15h20 Acoustic cues of rhotics degemination in Rome Italian Rosalba Nodari & Chiara Meluzzi Poster Session 15h40 - On the existence of an emotion induced tap /ɾ/ in Coffee Break Efik, Ememobong Udoh - An Experimental Investigation on Rhoticity and 16h00 From apical to dorsal /R/ in Quebec: The role of the mode of /r/- Sandhi in Devon English, Raphael Werner articulation - How many variants of rhotics in Persian? An Marie-Hélène Côté & Mélanie Lancien ultrasound and acoustic study, Reza Falahati 16h20 The Korean rhotic: Are we witnessing the birth of a ? - Toward expanding our understanding of variability in L2 liquid production, Mihoko Mathilde Hutin Teshigawara & Thomas Magnuson 16h40 The quantal change of alveolar /r/ to uvular /ʀ/ - Impact of post-lexical context and speech style Didier Demolin & Hans Van de Velde on word-final /ʁ/ realization in French using large 17h corpora and automatic speech processing, Yaru 18h Poster Session Wu, Martine Adda-Decker, Cédric Gendrot, Lori 19h Diner (included) Lamel

‘R-atics 6 Paris 7th, 8th november 2019

Friday 8th (Unvisersité Sorbonne Nouvelle – Salle 245) 9h30 Martine Adda-Decker : Investigating /R/ production in large corpora Production - Chair Person : Jane Stuart Smith 10h30 Why rhoticity is on everyone's lips: Accounting for the labial gesture in English /r/ Hannah King & Emmanuel Ferragne 10h50 Explaining extreme /r/ weakening in utterance-final position: an ultrasound tongue imaging study of rhotic lingual gesture timing using a spontaneous-speech corpus. Eleanor Lawson & Jane Stuart-Smith 11h10 Onset /R/-ticulation in Mandarin Kaiyue Xing 11h30 Coffee break

11h50 Fuse unless tense: Rhotic- interaction in Elbe-Weser German Daniel Gleim & Sören Eggert Tebay 12h10 The Acoustic Properties of Rhotic Post- in Moroccan Arabic

Aaron Freeman 12h30 Articulatory variation in Southern British English rhotics

Patrycja Strycharczuk, Susan Lloyd & James Scobbie

12h50 Lunch Break

14h James Scobbie: Phonetic and phonological variations in rhoticity from overlapping perspectives of self, other and Acquisition - Chair Person : Naomi Yamaguchi 15h00 (Socio)phonetic variation in the production of rhotics by Italian schoolchildren Chiara Meluzzi 15h20 Acquisition of postvocalic /r/ by Mandarin-speaking learners of English Zeyu Li 15h40 Coffee break

16h The acquisition of the Spanish trill in child and adult heritage speakers Gemma Repiso-Puigdelliura & Ji Young Kim 16h20 Acquiring rhotics in L2 English and L3 Polish – a comparison of adult and child learners Christina Golin, Romana Kopeckova & Ulrike Gut On the role of in the acquisition of Portuguese /ɾ/ by Mandarin native speakers Silke Hamann & Chao Zhou 16h40 R-atics6 closure speech

‘R-atics 6 Paris 7th, 8th november 2019

Abstracts

Plenary 1

Rhotics in Ethiopian languages Moges Yigezu University of Addis Ababa [email protected] This study examines the phonetic, phonological and sociolinguistic aspects of rhotics across Ethiopian languages taking into account sixty-one languages of the Afro-asiatic and Nilo- Saharan phyla. Forty-seven of these languages belong to the Afro-asiatic phylum that include15 Semitic, 16 Cushitic and 16 Omotic languages. The remaining 14 are Nilo-Saharan languages that belong to the Nilotic, Surmic and Koman families. The study surveys the typology of rhotic sounds - their frequency and distribution as well as variability across languages. All the sixty-one languages considered for this study have rhotic with varying place and mannerof articulations although the predominance of dental-alveolar is clear. There is no language for which no rhotic phoneme has been reported. The study moreover investigates the phonological patterning of rhotic consonants and provides articulatory information, allophonic distributions, phonotactic behavior, status of gemination and related phenomenon such as in these languages. The paper further discusses whether rhotics behave as a class of their own with respect to phonological patterning.

Variation session – Thursday 7th

Ongoing change in post-independence São Tomé: The use of rhotics as a marker of national identity among young speakers of Santomean Portuguese Marie-Eve Bouchard1 1Stockholm University [email protected] This paper examines variation in the use of rhotics in Santomean Portuguese. In Portuguese, the distribution of rhotics is determined by structure (Bonet & Mascaró, 1997). However, the emerging variety of Portuguese spoken in São Tomé and Príncipe diverges from the standard norm and shows great variability in its use of rhotics; specifically, Santomeans often use a strong-R in positions that require a weak-r in European and (e.g. [bʁɐ̃ku] instead of [bɾɐ̃ku/]). Sociolinguistic interviews with 56 speakers from the capital of São Tomé and Príncipe and its surroundings provide 5287 data points for analysis. Based on the apparent-time construct (Bailey, 2004; Bailey et al., 1991), variation patterns show a language change in progress in the use of the strong-R, with younger speakers using it more frequently. Within this strong-R category, the rhotic have emerged, a phenomenon that clearly distinguishes two generations (the Santomeans over 50 years old and the ones under 49 years old). This cut- off also marks the rise of the independentist movement, which was an important milestone in the history of São Tomé and Príncipe and the formation of its national identity. This suggests that rhotics also mark the emergence of a still- forming Santomean identity.

References: ‘R-atics 6 Paris 7th, 8th november 2019

Bailey, Guy. (2004). Real and apparent time. In Peter Trudgill, J.K. Chambers, and Natalie Schilling-Estes (eds.), The Handbook of Language Variation and Change. Oxford: Blackwell Publishing. 312–332.

Bailey, Guy, Wikle, Tom, Tillery, Jan, & Sand, Lori. (1991). The apparent time construct. Language Variation and Change 3. 241–264.

Bonet, Eulàlia, & Mascaró, Joan. 1997. On the representation of contrasting rhotics. In F. Martínez-Gil, & A. Morales-Front (eds.), Issues in the Phonology and Morphology of the Major Iberian Languages. Washington D.C.: Georgetown University Press. 103-126.

‘R-atics 6 Paris 7th, 8th november 2019

The “rolled” /R/ in French and some regional languages of France Timothée Premat1 & Philippe Boula de Mareüil2 1 SFL, CNRS & Univ. Paris 8, 2 LIMSI, CNRS & Univ. Paris-Saclay [email protected], [email protected] The archiphoneme /R/ rhotic, whose pronunciation may take many forms (being phonologically under-specified), is at the origin of pioneering work in (Labov, 1972). The front “rolled” [r], apical (pronounced with the tip of the tongue) and inherited from the r, still corresponded to the prestige pronunciation for the French of the seventeenth century (Rouillé, 2008: 11). Since then, it has been largely replaced by a back [ʁ], dorsal (involving the back of the tongue), but the speed and conditions of this sound change may vary as a function of the region. The Parisian origin of the shift from [r] to [ʁ] is fairly consensual, even though its progressive or spontaneous character, between the 17th and the 18th century, remains disputed (Straka, 1979; Lodge, 1997). Until a relatively recent date, however, this innovation has not succeeded in prevailing throughout the country, and an apical [r] may have been maintained in traditional dialects (or regional languages). This work aims at documenting the pronunciation of the /R/ rhotic in (often elderly) speakers who, in addition to French, speak a Gallo-Romance dialect and who produce an apical [r], in a more or less stable fashion, in French and their dialect. In this study, 35 speakers of Gascon, Languedocien, Francoprovençal, Bourguignon and Mainiot (7 speakers per dialect) were investigated: they were recorded reading the same text (Aesop’s fable “The North Wind and the Sun”, about one minute of speech) in French and translating it in their regional language. For each speaker, we measured the percentages of apical [r]s in French and in dialect, thus allowing comparisons, all other things being equal. Each of the 70 versions of the fable considered totals between 20 and 25 occurrences of /R/. Without mentioning the cases where the phoneme is not realised, 1689 /R/s have been analyzed: they were annotated in a binary way (front/back) by the authors, based on perception and inspection of spectrograms. We were also interested in the position in syllable onset or coda, the latter having been described by Morin (2013) and Demolin (1999), among others, as favoring the weakening of /R/. What are the phonological contexts and the social factors (in particular related to the speaker age) favouring the appearance of the [r]? We propose to answer this twofold question here. It turns out that there are more apical [r]s in dialects than in French (up to 88% in the Occitan variety of Gascony vs. up to 44% in the French variety of Burgundy), more so in a syllable onset or an internal coda than in a word-final position, and more so in elderly speakers than in younger ones. By contrast, we did not find any apical [r] in the regional French varieties of speakers under 65 years old. Faced with this situation, we will offer some elements of discussion about the interface between phonology and sociolinguistics.

References: Demolin, D. (1999). “Some phonetic and phonological observations concerning /R/ in Belgian French”. In: H. Van De Velde & R. Van Hout, (Eds.), ’R-atics: Sociolinguistic, phonetic and phonological characteristics of /r/, pp. 63–73. Brussels: Institut des Langues Vivantes de et Phonétique. ‘R-atics 6 Paris 7th, 8th november 2019

Labov, W. (1972). Sociolinguistic patterns. Philadelphia: University of Pennsylvania Press.

Lodge, A. (1997). Le Français, Histoire d’un dialecte devenu langue. Paris: Fayard.

Morin, Y. C. (2013). “From apical [r] to uvular [ʁ]: what the apico-dorsal r in Montreal French reveals about abrupt sound changes”. In: F. Sánchez Miret & D. Recasens, (Eds.), Studies in phonetics, phonology and sound change in Romance. Munich: LINCOM Europa.

Rouillé, N. (2008). La prononciation de la langue publique aux XVIIème et XVIIIème siècles. Sampzon: Éditions Delatour France.

Straka, G. (1979). “L’histoire de la consonne r en français”. In G. Straka (Ed.): Les Sons et les mots, pp. 465–499. Paris: Librairie C. Klincksieck.

‘R-atics 6 Paris 7th, 8th november 2019

A sociophonetic exploration of Afrikaans /r/ in the South Cape Yolandi Ribbens-Klein1 1 Universität Duisburg-Essen [email protected] This paper is a sociophonetic exploration into variation of Afrikaans /r/ as used by participants from George in the South Cape (Western Cape Province, South Africa). There is a dearth of research on rhotic variation in Afrikaans: alveolar trill is considered to be the supra-local, standard Afrikaans form and uvular realisations of /r/ (e.g. uvular trill or uvular fricative) are regional features (De Villiers, 1970). Both alveolar and uvular realisations occur in the speech of the participants, and the aim of this paper is to provide a qualitative description of the different types of rhotics used by the participants, followed by a quantitative analysis of the /r/ tokens according to genders, age groups, and phonological environments. The ‘family resemblances’ model proposed by Lindau (1985), which connect the different types of rhotics into a class of sounds, is expanded on by Magnuson (2007). I follow Magnuson’s (2007) argument that the contribution of the laryngeal/pharyngeal vocal tract is a facet of rhoticity that should be included with rhotic variants produced in the oral vocal tract. Thus, for the qualitative analysis, I distinguish three main types of Afrikaans rhotics based on place of articulation, namely coronal, dorsal, and radical. Furthermore, I draw on Sebregts’ (2015) approach in his sociophonetic analysis of Dutch /r/. He identified twenty variants based on distinctive acoustic and auditory parameters and articulatory correlates, which he grouped according to places and manners of articulation. For this study, twenty participants were grouped into two genders (female and male), and into two age cohorts: younger than 35, and older than 35. The /r/ tokens were extracted according to four phonological environments (initial; post-consonantal /r/ in onsets; pre- consonantal /r/ in codas; and final) in Praat (Boersma and Weenink, 2019) from picture description tasks performed during sociolinguistic interviews with the participants. For the qualitative analysis, I explored different acoustic measurements, inspired by previous research of /r/ variation in Dutch (e.g. Sebregts, 2015) and Flemish (e.g. Tops, 2009). The qualitative results found fourteen variants, which were grouped into four main sets based on articulation types. The variation encoded includes place of articulation (alveolar – uvular – pharyngeal), as well as manner (trill – tap – fricative – – vowel-like). Quantitatively, coronal articulations were most frequently used, with alveolar trill-tap as the dominant variant. Vowel-like articulations were second, mainly because I included the omitted /r/ in this main set. With dorsal articulations, uvular trill with frication was most frequently used, with the remainder of the variants being uvular trills. In terms of the radical articulations, the weak pharyngeal was the dominant variant. The younger and older groups showed an indication of age-related variation, especially in terms of frication and pharyngeal articulations. Gendered patterns of variation were also observed, where coronal articulations were more frequently used by women than men. Importantly, my study confirms the importance of including pharyngeal articulations as part of a speaker’s rhotic repertoire, and adds insights into the variable realisations of Afrikaans rhotics.

References: ‘R-atics 6 Paris 7th, 8th november 2019

Boersma, P. & Weenink, D. 2019. Praat: doing phonetics by computer [Computer program]. Version 6.0.32, retrieved from http://www.praat.org/.

De Villiers, M. 1970. Afrikaanse Klankleer: Fonetiek, Fonologie en Woordbou. Cape Town: A.A. Balkema.

Lindau, M. 1985. The story of /r/. In V. Fromkin (ed.) Phonetic : Essays in Honor of . Orlando, FL: Academic Press. 157-168.

Magnuson, T.J. 2007. The story of /r/ in two vocal tracts. Proceedings of the 16th International Congress of Phonetic Sciences: 1193-1196.

Sebregts, K. 2015. The Sociophonetics and Phonology of Dutch r. Utrecht: LOT.

Tops, E. 2009. Variatie en Verandering van de /r/ in Vlaanderen. Brussels: VUBPRESS.

‘R-atics 6 Paris 7th, 8th november 2019

Uvular and apical rhotic distribution in a basque speaking valley between the spanish and french states Alexander Arcelus Múgica1 1 Universidad del País Vasco / Euskal Herriko Unibertsitatea [email protected] The Basque is a language which makes phonological distinction between tap and trill rhotics. However, this contrast is only preserved in intervocalic position (Bradley, 2001: 2). Traditionally, both rothic consonants have been described as apicals (Txillardegi, 1980: 161; Hualde 1998 [1991]: 15, 2003). In contrast, uvular pronunciation of trill rhotics is being generalized on the French side of the country due to the influence of the . Mitxelena (1990 [1961]: 328) named this phenomenon as grasseyement français. Uvular rhotics are mostly used on the western side of the French Basque Country (along the coastline) and apical rhotics are better conserved in the east (in the mountains). Garazi valley is a very interesting area to study the expansion of this French influence. This valley is a rural area where most of its urban settlements are located on the French side of the Basque Country with the exception of Luzaide, which belongs to the Spanish state. This politico-administrative situation, together with the geographical distribution of population in these territories, have conditioned the area linguistically. There has existed a social unity between Luzaide and the rest of the towns of Garazi until the second half of the 20th century. However, the political division of the valley into two states has created a social barrier recently. As Auer and Hinskens (1996: 4) point out, “Convergence in the dialect- and dialect-dialect dimensions necessarily leads to divergence at the borders. Particularly on the international level, political borders transgressing old dialect continua have turned into new dialect borders”. Our hypotheses state that, in terms of geography, the political frontier has stopped the expansion of the uvular rhotic. In other words, we do not find any uvular trill in Luzaide because we only have apical trills in Spanish. Regarding age, we consider that oldest people are the ones who still continue to pronounce the apical phoneme. However, adults and young have adopted the uvular pronunciation. Finally, we expect that oldest women have higher rates of apical trills, due to the fact that they have not had the opportunity of getting out of home. Hence, they are more conservative. Regarding methodology, we have decided to use the friends’ friend technique in our fieldwork in order to guarantee the obtainment of spontaneous speech. We have recorded 30 people with the laptop using Audacity program and Sony microphone. We have used Praat to transcribe, label and annotate data (Boersma & Weenick, 2019) and we have created a data- base with Excel. Finally, we have made chi- square, hierarchic conglomerates and multidimensional analysis with SPSS statistical program. Preliminary results are proving initial hypothesis. Aditionally, we have also found that it is very rarely the case that the same informant mixes the two phonemes.

References: Auer, P. & Hinskens, F. (1996). The Convergence and Divergence of Dialects in Europe. New and not so new developments in an old area. Sociolinguistica 10, 1-30. ‘R-atics 6 Paris 7th, 8th november 2019

Boersma, P. eta D. Weenink, (2016). Praat: doing phonetics by computer, Version 5.1. [computer program]: [last consultation: 30-06-19].

Bradley, T. (2001 a). “A typology of rhotic duration and neutralization”, Procedings of the north east linguistic society 31: [last consultation: 30-06-19].

Hualde, J.I. (1998). [1991]. Basque phonology. AEB eta Kanada: Routledge.

Mitxelena, K. (1990) [1961]. Fonética Histórica Vasca. Donostia: Gipuzkoako Foru aldundia.

Txillardegi, (1980). Euskal Fonologia. Donostia: Ediciones Vascas.

‘R-atics 6 Paris 7th, 8th november 2019

A large-scale automatic acoustic analysis of rhoticity in Scottish English Jane Stuart Smith1 & Rachel Macdonald1 1University of Glasgow [email protected], [email protected] Scottish English is typically regarded to be ‘rhotic’, i.e. /r/ in non-prevocalic, postvocalic position, in words such as card, car, and better, is realized as an articulated rhotic (Wells, 1982). Especially in the dialects of Scottish spoken across the Central Belt, the most populated area of Scotland, stretching between Glasgow and Edinburgh, a number of studies since (Romaine, 1979) have provided evidence of socially-stratified changes in rhoticity. Specifically, working-class speakers have been shown to produce auditorily weak /r/s, or even no audible /r/ at all, whilst middle-class speakers produce auditorily strong /r/. A recent auditory-acoustic-articulatory experimental study of coda /r/ in a socially-stratified sample of Glaswegian (anon 2019), has revealed the following robust correlations: the auditorily weaker the /r/, the higher the third , and the longer the delay of the maximum tongue gesture with respect to the offset of voicing (in middle-class adolescents), and vice versa, the auditorily stronger the /r/, the lower F3, and the earlier the tongue body maximum in relation to voicing offset (in working-class young speakers). Alongside this pattern of socially- conditioned polarization of strong–weak /r/ in the Central Belt, bar a few studies, e.g. Jauriberry et al (2012) for Ayr, and Schuetzler (2009) for Scottish Standard English, the degree of rhoticity across contemporary Scottish English dialects is unknown. This paper is the first to consider rhoticity in Scottish English by vowel, dialect, and time period, from evidence collected using a large-scale analysis of coda /r/ from approximately 500 speakers across three Scottish Standard English corpora (Arthur the Rat (a selection), DoubleTalk (SSE speakers), How Stable is the Standard)) and five regional vernacular corpora (Sounds of the City, SCOTS, One Speaker Two Dialects, Brains in Dialogue, Glaswasian)) which together cover at least five broad dialect areas (Glasgow, West, East Coast, North East and Highlands & Islands). Large-scale analysis acoustic analysis of /r/ in a consistent manner across multiple datasets, also controlling for known linguistic and prosodic factors such as, preceding and following context, lexical frequency, speech rate and phrase position, is made possible using the Integrated Speech Corpus ANalysis (ISCAN) tool (e.g. McAuliffe et al, 2019). Automated implementation of static formant measures and dynamic formant tracks for the first three , and especially F3, are used to capture coda /r/ acoustically. Our dataset allows us to consider rhoticity with respect to a range of important dimensions. First, the acoustic correlates of rhoticity are charted across dialects of Scottish English for the first time. Second, recordings of both spontaneous and controlled speech from the late 1940s until the mid-2010s in a number of dialects permit a larger-scale examination of shifts in rhoticity in vernacular and standard Scottish English speech than has been possible before. Third, the overall extent of derhoticisation in Scottish English dialects across the country, is considered in a consistent fashion, using automated analysis. We also reflect on the usefulness and pitfalls, of analysing rhoticity via large-scale automated acoustic speech corpus analysis.

‘R-atics 6 Paris 7th, 8th november 2019

Plenary 2:

That is what we R: phonetics and phonology in rhotic identity Koen Sebregts University of Utrecht [email protected] It has almost become a truism that rhotics are highly variable phonetically, and that whatever it is that unites them must be sought at the phonological level only. This does not mean, however, that there is no phonological r-variation, nor that people have stopped looking for phonetic characteristics in defining rhotics. In this talk, I will explore these two strands of rhotic studies. First, I will consider the question of phonetic vs. phonological allophony: how do we decide if a phonetic variant should be seen as a separate phonological object? Secondly, I will ask whether the search for phonetic links among rhotic classes is still a worthwhile avenue to explore.

Sound Change Session – Thursday 7th

What aʁʁʁ you? Reconsidering François Falc’hun’s French /ʁ/ hypothesis through the romance diasystem Fabian Zuk1 1Université Jean Moulin, Lyon III [email protected] Though the standard french rhotic is uvular today, at the beginning of the twentieth century, the apical “rolled” [r] inherited from latin could still be found in Normandy (Girard & Lyche, 2003), in Canada, and in other non-standard dialects. While many historical phoneticians, Bourciez (1967, §167), Pope (1952, §498) et Martinet (1962, p. 201) maintain that uvular [ʀ] appeared only as of the 17th century, the Breton dialectologist François Falc’hun (1972) made the claim that the coexistence of an apical [r] and a dorsal [ʀ] had its origin in a much earlier state of galloromance, an argument which he made based on the treatment of Breton uvular fricatives as French rhotics in numerous toponyms, ex. Bret. /neχ/ ‘elevation’ represented with a final ⟨r⟩ in numerous modern toponyms:

Nermont (Saint-Coulomb) Pêne Nère (Haute-Pyrénées) Nerville-la-Forêt (Val-d'Oise) Ners (Gard)

If Falc’hun hypothesis never gained traction, its because his old-school dialectological approach could not stand up to the expectations of 70s generative phonology put in place by Chomsky & Halle (1968) and others. Falc’hun’s 1972 article provides no attempt of a synchronic analysis of presumed [r] and [ʀ] distribution in French and its historical varieties. This presentation therefore provides a synchronic analyis of r like phones and their phonological mapping in later gallo-romance. Its a well-known fact of classical latin and of many of its romance descendants that singleton /r/ and geminate /rː/ contrasted phonemicaly in intervocalic position, ex. span. pero /pero/ ‘but’ vs. perro /perːo/ ‘dog’. In the Romance languages, this length contrast has generally evolved into a quality contrast (cf. Bossong, 2008, p. 283). While lenis *[r] vs fortis *[rː] are in near complementary distribution, they contrast in intervocalic position, which can be presented as follows:

‘R-atics 6 Paris 7th, 8th november 2019

Figure 1. #_ _V _ _.C V_ _V .C_ _V _ _ # Italian [rː] [ɾ] [ɾ] / [rː] [ɾ] [ɾ] Castilian [rː] [ɾ] [ɾ] / [rː] [ɾ] [ɾ] Portuguese [ʁː] [ɾ] [ɾ] / [ʁː] [ɾ] [ɾ] Old French [χː] [r] [r] / [χː] [r] [χː] Latin Etymon RARO ORDO RARUM / CARRUM GRANDIS PER

As shown in figure 1, the distribution of allophones in Old French is very close to that of modern Portuguese and the other Romance languages, with the exception of one position: Coda. Coincidently Old French (and arguable Modern French in its resonants) was affected by synchronic rule of voicing neutralization in final position. Furthermore, the specific dorsal pronunciation of fortis */rː/ in galloromance is best explained by clear synchronic voicing and place assimilation rules proper to both adstratal celtic and germanic languages. Finaly, Falc’hun’s schematic hypothesis is supported by modern phonetic data (International Phonetic Association, 1999, p. 78‑81) which clearly shows an allophonic distribution of voiced [ʁ] and voiceless [χ] presenting the possibility for confusion with Breton [χ]. Ultimately, phonetic detail, phonemic considerations from the Romance diasystem and historical sociolingusitics are required to make sense of Falc’huns initial claim that dorsal [ʀ] should be dated to Gallo- romance. ‘R-atics6 is quite obviously the best audience to reflect upon the limits and theoretical implications of this proposal.

References: Bossong, G. (2008). Die romanischen Sprachen: Eine vergleichende Einführung. Hamburg: Buske.

Bourciez, É. (1967). Phonétique française. Paris: Klincksieck.

Chomsky, N., & Halle, M. (1968). The sound pattern of English. New York: Harper & Row.

Falc’hun, F. (1972). De l’"r" gaulois à l’"r" français. In A. Valdman (Éd.), Papers in Linguistics and Phonetics to the Memory of Pierre Delattre (p. 105‑115). Paris: Mouton De Gruyter.

Girard, F., & Lyche, C. (2003). La phonologie du français contemporain dans le Domfrontais: un français en évolution. Tribune Internationale des Langues Vivantes, 33, 166‑173.

International Phonetic Association. (1999). Handbook of the International Phonetic Association: A Guide to the Use of the International Phonetic Alphabet. Cambridge University Press.

Martinet, A. (1962). Linguistique structurale. Annuaires de l’École pratique des hautes études, 95(1), 113‑113. ‘R-atics 6 Paris 7th, 8th november 2019

Pope, M. K. (1952). From Latin to Modern French with Especial Consideration of Anglo– Norman Phonology and Morphology. Manchester: Manchester University Press.

‘R-atics 6 Paris 7th, 8th november 2019

Acoustic cues of rhotics degemination in Rome Italian Rosalba Nodari1 & Chiara Meluzzi2 1 Università degli studi di Siena, 2 Università di Pavia [email protected], [email protected] Rhotics degemination is considered a sociolinguistic marker of the Italian variety spoken in Rome (RI), usually associated with low status speakers and informality (Stefinlongo 1999). Given the particular status of the R variable in conveying socio-indexical information (Scobbie 2006), we analyse if, and how, young Roman speakers produce rhotics degemination in RI. A perceptive test integrates the production results in order to verify which phonetic cues convey the perception of (de)gemination, and whether this variable is associated with prototypical RI speakers. For production experiment, 10 speakers from Rome (5 males and 5 females aged between 23 to 30) participated to a sentence-reading task, with 70 sentences of equal length and controlled prosodic contour. Each sentence contained one token with a singleton and/or geminate /r/, in stressed and/or unstressed condition. 700 tokens were annotated following the protocol by Celata et al. (2016), thus classifying rhotics as either taps, trills, or fricatives according to their spectrogram realization. For quantitative analysis, we relied on preceding vowel and duration to test the consonant gemination (Argiolas 1995, Payne 2005). For perception experiment, 69 Italian native speakers from Rome and other parts of Italy participated to an online test with 12 selected tokens with a geminated rhotic coming from our corpus, mixed with an equal number of fillers. Results show that, for rhotic duration, lexical gemination was statistically significant, and geminated rhotics were longer than singleton rhotics. Speaker’s sex and syllable stress were found to interact in stressed only. As for vowel duration, only syllable stress was statistically significant, with stressed syllable being, as expected, longer than unstressed one; unexpectedly, no effect of lexical gemination was found. Qualitative analysis shows, instead, a more complex picture. Intervocalic geminate rhotics seem to allow a greater range of possibilities: they can be realized mainly as trills, but approximants, fricatives, taps, and combined realizations (trill or tap with a fricative appendix) are found too. However, a great within-speaker variation has also been observed. The perceptive test shows that rhotic plays a major role in conveying the perception of degemination, whereas quantitative phonetic cues are not sufficient per se. Moreover, tokens clearly perceived as degeminated are significantly associated with RI. These perceptual results shed light on the real status of rhotic degemination in RI, leading to an interpretation of this phenomenon as involving a qualitative change, from trill to tap, rather than a matter of consonantal or preceding vowel duration. They confirm also the status of rhotic degemination as a feature associated to RI.

References: Argiolas, F., Macrì, F., Di Benedetto, M. G. (1995), Acoustic analysis of Italian [r] and [l], Journal of the Acoustical Society of America, 97, 3418

Celata, C., Meluzzi, C. and Ricci, I. (2016), The sociophonetics of rhotic variation in Sicilian dialects and Sicilian Italian: corpus, methodology and first results, Loquens, 3,1. ‘R-atics 6 Paris 7th, 8th november 2019

Payne, E. (2005). Phonetic variation in Italian consonant gemination, Journal of the International Phonetic Association, 35, 2, pp. 153-181.

Scobbie, J. (2006). (R) as a variable. In Brown, K. (Ed.), Encyclopedia of Language & Linguistics (2nd ed.), Oxford, Elsevier, pp. 337–344.

Stefinlongo, A. (1999) “Neoromanizzazione” del territorio. La lingua delle scritte murali nell'area metropolitana romana, in Dardano, M. et al. (a cura di), Roma e il suo territorio. Lingua, dialetto, società, Roma, Bulzoni, pp. 267-285.

‘R-atics 6 Paris 7th, 8th november 2019

From apical to dorsal /R/ in Quebec: The role of the mode of articulation Marie-Hélène Côté1 & Mélanie Lancien1 1 Université de Lausanne [email protected], [email protected] Quebec French offers a unique testing ground for the study of /R/ due to the coexistence of apical and dorsal /R/s – respectively in the western and eastern part of the territory – and to the fast “dorsalization” of apical /R/s in the west by the generation born between 1920 and 1960. The study of the regression of apical /R/s, essentially documented in Montréal (Clermont & Cedergren 1979; Cedergren 1985; Tousignant 1987; Santerre 1979, 1982; Sankoff & Blondeau 2007), has revealed the impact of several social factors and the more regular retention of apical /R/s in prevocalic than in postvocalic position. However, a global comprehension of this change requires a wider perspective, geographically (beyond Montréal), as well as temporally (beyond the transition generation). The mode of articulation of /R/ should also be considered in more detail, in addition to place of articulation. The most recent results (Saint-Amant Lamy 2016) show in particular that the regression of apical /R/ is sensitive to the size of the community, smaller ones being affected later. We explore here a subset of the PFC corpus (Durand 2002; Côté 2014), containing 159 speakers (80 women, 79 men, born between 1921 and 1998), distributed in three age groups (senior, intermediate, junior) and 12 survey points across the province of Quebec (6 in each of the eastern and western dialectal areas). Every speaker read two word lists and all /R/s were perceptually coded for place (apical/dorsal) and mode of articulation (trill, flap, voiceless fricative, voiced fricative, approximant, vocalized, retroflex, not realized). A qualitative observation of the data suggests that, in the adoption of the new dorsal variants, western speakers do not realize the same modes of articulation than those observed in the East (whose /R/s were already dorsal). Linear mixed models considering the interactions between the effect of linguistic (position in the word/syllable) and extralinguistic factors (age, east/west) on the mode of articulation of dorsal /R/s (N=18 603) were performed and new results emerge. More specifically, we observe for the senior and intermediate age groups in the West with respect to those in the East: a more frequent use of uvular trills (p<0.05), those trills being replaced by fricatives in the junior groups in both areas (p<0.001 for the difference in the proportions of trills between juniors and seniors, and p<0.01 for the frequency of fricatives); a more frequent use of vocalized /R/s word-finally (p<0.001 for seniors), the proportion of vocalized /R/s dropping in the western junior group in favor of approximants and fricatives (p<0.01 for the difference in the rate of vocalization between juniors and intermediate/senior speakers, and p<0.01 for the proportion of fricatives between juniors and seniors). Those two tendencies suggest that the adoption of the dorsal variants in the West may have been favored by two transitory mechanisms: (1) the transition from apical trills to uvular trills (same mode of articulation) to uvular fricatives (same place of articulation); (2) the “deconsonantization” of /R/, which then allows a dorsal fortition. Beyond the simple adoption of the variants used in the East, the transition in the West from apical to dorsal variants appears to also implicate changes in mode of articulation.

‘R-atics 6 Paris 7th, 8th november 2019

References Cedergren, H. (1985). Une histoire d’R. In Lemieux, M. & Cedergren, H. (ed.), Les tendances dynamiques du français parlé à Montréal. Québec : Gouvernement du Québec, Office de la langue française, Vol. 1, 25-56.

Clermont, J. & Cedergren, H. J. (1979). Les ‘r’ de ma mère sont perdus dans l’air. In Thibault, P. (ed.), Le français parlé : études sociolinguistiques. Edmonton : Linguistic Research, 13-28.

Côté, M.-H. (2014). Le projet PFC et la géophonologie du français laurentien. In J. Durand, G. Kristoffersen & B. Laks, avec la collaboration de J. Peuvergne (eds), La phonologie du français: normes, périphéries, modélisation. Nanterre: Presses Universitaires de Paris Ouest, 175-198.

Durand, J., Laks, B. & Lyche, C. (2002). La phonologie du français contemporain : usages, variétés et structure. In Pusch, C. & Raible, W. (ed.), Romanistische Korpuslinguistik- Korpora und gesprochene Sprache/Romance Corpus Linguistics - Corpora and Spoken Language. Tübingen : Gunter Narr Verlag, 93-106.

Saint-Amant Lamy, H. (2016) Kilomètre zéro: /R/ et le statut épicentral de Montréal dans le domaine linguistique laurentien. Cahiers internationaux de sociolinguistique, 10, 101-134.

Santerre, L. (1979). Les (r) montréalais en régression rapide. In Lavoie, T. (ed.), Les français régionaux du Québec, Numéro spécial de Protée VII(2), 117-132.

Santerre, L. (1982). Des r montréalais imprévisibles et inouïs. Revue québécoise de linguistique, 12(1), 77-96.

Tousignant, C. (1987). La variation sociolinguistique. Modèles québécois et méthode d'analyse. Sillery : Presses de l'Université du Québec.

‘R-atics 6 Paris 7th, 8th november 2019

The Korean rhotic: Are we witnessing the birth of a phoneme? Mathilde Hutin1 1 Structure Formelle du Langage (UMR 7023) [email protected] Korean has a (Cho 1967), realized as a lateral [l] in coda position, as a tapped [ɾ] intervocalically and as a geminated [ll] if preceded by a rhotic itself, as in (1). (1) /muR/ → [mul] water /muR + i/ → [muɾi] water-Nom. /muR + Ro/ → [mullo]water-Instr. This rhotic is subject to several surprising phenomena (Cho 2016): - Like nasals and approximants, it does not block intervocalic voicing, as in (2). (2) /nɔRp-ɔ/ → [nɔlbɔ] it is large /maRk-a/ → [malga] it is clear - It assimilates /t/ regressively and /n/ regressively and progressively, as in (3). (3) /tikɨt-RiɨR/ → [tikɨlliɨl] the letters t and l /ʧin-Ri/ → [ʧilli] truth /taR-naRa/ → [tallaɾa] moon-world - It is nasalized when in word-initial position or in post-coda onset, as in (4). (4) /kjɔŋ-hwe-Ru/ → [kjɔŋhweɾu] Kyunghoe Pavillon /Ru-kak/ → [nugak˺] high pavillon /ok-Ru/ → [oŋnu] Jade Pavillon - It is syncopated when in word-initial position followed by /j, i/, as in (5). (5) /ki-Rju/ → [kiɾju] flow (of air) /Rju-Laŋ/ → [juɾaŋ] floating I argue here that this distribution pattern can be explained using a contour-based approach of phonology (Carvalho 2008, 2014, 2017) derived from the Strict-CV framework (Lowenstamm 1996). I propose that the native Korean rhotic is, essentially, a melody that has no capacity to license a timing slot. It can be expressed only when a timing slot is already licensed by the preceding nucleus as in (6a), thus being realized as [l], or by the preceding and the following nucleus as in (6b), thus being realized as [ɾ] (cf (1)). Given the special status of the association between /R/ and its timing slot in (6a), the intervocalic voicing is not blocked if the rhotic coda is followed by a CV sequence (cf (2)).

Given my proposal, when /R/ appears as an onset, it has to either (i) spread to the preceding coda consonant (if it shares melodic content like /t/, /n/ or /R/, cf (1, 3)), (ii) be strengthened by some extra- melody (i.e. with nasality, cf (4)), or (iii) it cannot surface (i.e. in front of a ‘R-atics 6 Paris 7th, 8th november 2019 palatal vowel or glide, cf (5)). This representation of the native Korean rhotic accounts for the l~ɾ allophony and is consistent with the fact that a rhotic can never be realized as such word- initially. However, rhotics can now be found in word-initial position in loanwords, as shown in (7), and will then be realized as tapped (Kang 2012). (7) Engl. /rækɪt/ → ko. /Rakes/ → ko. [ɾaket˺] Engl. /ræp/ → ko. /Rap/ → ko. [ɾap˺] The claim of this paper is that there is, in the peripheral phonology of Korean – as the one used for loanword adaptation for example – the possibility for /R/ to license its own timing slot. This hypothesis makes the prediction that /R/ in loanwords will be able to appear in front of /j, i/, or after /n/ without assimilating. That is indeed the case, as shown in (8). (8) Engl. /lɪkwɪd/ → ko. /Rikʰwitɨ/ → ko. [ɾikʰwidɨ] Engl. /tɛnrɛk/ → ko. /tʰenRek/ → ko. [tʰenɾek˺] Moreover, it could also be expected that /R/ in loanwords, especially in the unusual positions displayed in (7) and (8), will have particular articulatory and acoustic properties in comparison with native /R/. This hypothesis will be tested and the results shared in due time. If the acoustic data confirm this hypothesis, it will mean that there is in the peripheral phonology of Korean the emergence of a “stronger”, “full” /R/-phoneme.

References: Carvalho, Joaquim Brandão de. 2008. From Positions to Transitions: A Contour-based Account of Lenition and Fortition. J. B. de Carvalho, T. Scheer & P. Ségéral (eds.) Lenition and Fortition, 415-445. Berlin: Mouton de Gruyter

Carvalho, Joaquim Brandão de. 2014, C/V interactions in Strict CV. S. Bendjaballah, N. Faust, M. Lahrouchi & N. Lampitelli (eds.), The form of structure, the structure of form. Essays in honor of Jean Lowenstamm. Amsterdam: John Benjamins, 123-138

Carvalho, Joaquim Brandão de. 2017. Deriving sonority from the structure, not the other way round: A Strict CV approach to consonant clusters. The Linguistic Review 34, 589-614

Cho, S-B. 1967. A Phonological Study of Korean. Acta Universitatis Upsaliensis, Studia Uralica et Altaica Upsaliensia 2. Almqvist and Wiksells, Uppsala.

Cho, Young-mee Yu. 2016. Korean Phonetics and Phonology. Oxford University Press Online. Online publication date: Oct. 2016

Kang, Yoonjung. 2012. The Adaptation of English Liquids in Contemporary Korean: a Diachronic Study. Catalan Journal of Linguistics 11, 41-63

Lowenstamm, Jean. 1996. CV as the only syllable type. B. Laks, J. Durand (eds), Currents Trends in Phonology. Models and Methods, Salford, ESRI, 1996, p. 419-441.

‘R-atics 6 Paris 7th, 8th november 2019

The quantal change of alveolar /r/ to uvular /ʀ/ Didier Demolin1 & Hans Van de Velde2 1Laboratoire de Phonétique et de Phonologie (CNRS / Sorbonne Nouvelle), 2 Fryske Akademy Leeuwarden [email protected], [email protected] In recent years, the variability of rhotics has drawn the attention of phoneticians, phonologists, dialectologists and sociolinguists (e.g., Van de Velde & Van Hout (2001, Spreafico & Vieti 2013). At the one hand, /r/ appears to be extremely variable in a (growing) number of languages, at the other hand, /r/ seems to be frequently involved in processes of language change (both as a variable and as a conditioning factor). Many internal and external factors are claimed to contribute to the origin and spread of the change from alveolar /r/ to uvular /ʀ/. However, there is no strong explanation for the origin of this change, nor for its unidirectional character. In this paper we provide a new phonetic explanation for the change from an alveolar to a uvular trill. In this paper we show that the lenition of an alveolar /r/ creates a zone of instability in the anterior region of the VT because the focal point associating F3 and F4, which is characteristic of an alveolar constriction, is affected. This results in a weakening of the articulation and a lowering of F2, F3 and F4 in this zone. This effect is even stronger if there is a perturbation on the lips. The lowering and weakening of the tongue tip not only affects the F-•pattern of the consonant but also changes the length of the front part of the VT. The next focal point where two formants approximate is then further back in the uvular region of the vocal tract where F3 and F4 approximate, although not as strongly as in the alveolar/post- alveolar region. This F3/F4 association is due to the fact that the tongue position is already in a back position for an alveolar trill and favors this formant association and a potential perceptual reinterpretation. If the aerodynamic conditions are maintained, i.e. with a substantial difference between intraoral (Po) and atmospheric (Pa) pressure, then the conditions for a shift to a uvular trill are met. Indeed to produce a trill the Po/Pa difference must be sustained above a certain threshold during the whole consonant. This is for the Bernoulli effect to occur for the duration of the consonant. Thus, an abrupt change in the F-pattern combined with an aerodynamic threshold account for the change from alveolar /r/ to uvular /ʀ/.

References: Brescancini, Claudia and Valéria N. O. Monaretto (2008). 'Os róticos no sul do Brasil: panorama e generalizações', Signum: Estudos Linguisticos 11, 2:51–66.

Demolin, D., Franchetto, B. and Fausto, C. (2015). Uvular taps in Kuikuro. ms.

Engstrand, O., Frid, J. and Lindblom, B. (2007). A perecptual bridge between coronal and dorsal /r/. In M.-­‐J. Solé, P. S. Beddor and M. Ohala. Experimental Approaches to Phonology. Oxford University Press. 175-­‐191.

Lindau, M. 1985. The story of /r/. In Fromkin, V. A. (ed.), Phonetic Linguistics. Essays in honour of Peter Ladefoged, Orlando: Academic Press. 157-­‐68. ‘R-atics 6 Paris 7th, 8th november 2019

Magnuson, T.J. 2007. The story of /r/ in two vocal tracts. Proceedings of the 16th International Congress of Phonetic Sciences: 1193-1196.

Rennicke, L. (2015). Variation and change in the rhotics of Brazilian Portuguese. PhD University of Helsinki.

Sebregts, K., 2015. The sociophonetics and phonology of Dutch /r/, Utrecht, LOT. (1935- 1993), PhD thesis, Katholieke Universiteit Nijmegen.

Spreafico, L & A. Vietti (eds.) (2013). Rhotics: New Data and Perspectives. Bolzano University Press.

Tops, E., 2009. Variatie en verandering van de /r/ in Vlaanderen, Brussel, VUB Press.

Van De Velde, H., 1996. Variatie en verandering van de gesproken Standaard-Nederlands (1935-1993), PhD thesis, Katholieke Universiteit Nijmegen.

Van de Velde, H. & R. van Hout (eds.) (2001). 'r-atics: sociolinguistic, phonetic and phonological characteristics of /r/. Speciaal nummer van Etudes & Travaux 4, ILVP, Bruxelles.

Van de Velde, H., E. Tops & R. Van Hout (2013). The spreading of uvular [R] in Flanders. In L. Spreafico & A. Vietti (eds.), Rotics: New Data and Perspectives. Bolzano University Press, 225-248.

‘R-atics 6 Paris 7th, 8th november 2019

Poster Session – Thursday 7th

On the existence of an emotion-induced tap /ɾ/ in Efik Ememobong Udoh1 1 University of Uyo, Nigeria [email protected] Efik is a Central Lower Cross language which does not have the alveolar tap /ɾ/ in its inventory. Instead, the language has the trill /r/ which occurs in two restricted positions – word medial position, as an onset [CV.rV(V)]; and as a co-onset element in the only consonant cluster sequence [CrV(V)]. The study explores the fact that, the native speaker of Efik does not consistently articulate the alveolar trill during some emotional bout. Seven emotional variables are used to observe the behaviour of the alveolar trill in Efik, viz: fear, anger, joy, pride, sadness, surprise and normal. It has been observed thus far, that when speakers of Efik are angry, they tend to replace the alveolar trill with an unattested tap. This mainly happens only in the CrV(V) syllable. This is work in progress, and we are targeting the study of one hundred and forty Efik language consultants (twenty subjects per emotion) selected by stratification, based on speech event. Thus far, we have not observed this phenomenon during other emotional states apart from anger. With readings from Praat spectrograph readings, we will be able to establish the difference between a trill and the replacing rhotic sound during anger (and any other emotional state). We are also looking at the segmental combinations in the cluster, to ascertain the specific phonological conditioning environment that combines with emotion to cause a change from the trill to a tap in the CrV(V) sequence.

‘R-atics 6 Paris 7th, 8th november 2019

An Experimental Investigation on Rhoticity and /r/- Sandhi in Devon English Raphael Werner1 1Julius-Maximilians-Universität Würzburg [email protected] Rhoticity, i.e., the presence or absence of non-prevocalic /r/, is a variable frequently analyzed in sociolinguistic studies. Research combining rhoticity with /r/-sandhi, i.e., linking and intrusive /r/, is relatively scarce, as rhoticity has often been described as being in complementary distribution with /r/-sandhi (Giegerich, 1999). However, these features are rarely categorical, and can both be present in single speakers (Hay & Sudbury, 2005; Barras, 2018). This experimental study examines the current degrees of rhoticity and /r/-sandhi in Devon in the South West of England and sheds further light on the variability and interrelations of these features. Though traditionally considered a rhotic isle in non-rhotic RP (Wells, 1982; Collins & Mees, 2003), recent research conducted at various locations indicates that the South West is in the process of moving towards non-rhoticity (Williams, 1991; Sullivan, 1992; Jones, 1998; Dudman, 2000; Piercy, 2007, 2012; Grossenbacher, 2016; Barras, 2018; Malarski, forthcoming). This makes Devon, situated in the lower part of the region, the ideal research area, since an intermediate system that contains both phenomena to differing degrees is to be expected due to this shift. The present study aims to further support the research done on non-prevocalic /r/ and /r/- sandhi in the South West, to test internal and external factors governing the use of rhoticity, linking /r/, and intrusive /r/, and, to examine the relationships between these features, as has been done for New Zealand English (Hay & Sudbury, 2005) and the dialect spoken in Lancashire (Barras, 2018). To that end, 17 native speakers of English from Devon have been recorded during sociolinguistic interviews and reading tasks. The collected data are analyzed using mixed effects regression models in Rbrul. The model contains internal (preceding vowel, morphological position of /r/, stress, word type) and external (age, gender, task) factors. Results show that Devon is becoming non-rhotic with higher rates of residual rhoticity in older speakers. The degree of overall rhoticity reaches 13.2% while mean realization rates for /r/-sandhi are comparable to RP. Significant factors influencing the production of these features have been found to be preceding vowel, position of /r/, stress, and age group for non- prevocalic /r/, preceding vowel and task for linking /r/, and stress in the following syllable for a subset of intrusive /r/. Moreover, rhoticity and linking /r/ are shown to be positively correlated, rhoticity and intrusive /r/ to have a negative correlation, and linking and intrusive /r/ to not be correlated under the premises of this study. This project contributes to the study of rhoticity and dialectal variation in the South West reflecting its shift towards /r/-lessness, including factors relevant to the production of these /r/-phenomena in that area. Finally, it reveals some of the conditions of interactions between the discussed features and shows that rhoticity and /r/-sandhi are not mutually exclusive for all speakers.

References: ‘R-atics 6 Paris 7th, 8th november 2019

Barras, W. (2018). Residual Rhoticity and Emergent r-sandhi in the North West and South West of England. Different Approaches to Hiatus-Resolution? In N. Braber & S. Jansen (Eds.), Sociolinguistics in England (pp. 363-392). London: Palgrave Macmillan UK. Collins, B. & Mees, I. M. (2003). Practical Phonetics and Phonology: A Resource Book for Students. London: Routledge.

Dudman, K. (2000). Loss of rhoticity and long mid monophthongisation in St Ives, Cornwall (Unpublished BA thesis). Essex University.

Giegerich, H. J. (1999). Lexical strata in English: Morphological causes, phonological effects. Cambridge: Cambridge University Press.

Grossenbacher, S. (2016). From East to West? Dialect Diffusion between Swindon and Bristol. (Unpublished MA Thesis). University of Bern.

Hay, J. & Sudbury, A. (2005). How Rhoticity Became /r/-sandhi. Language 81 (4), pp. 799- 823. DOI: 10.1353/lan.2005.0175.

Jones, J. (1998). Phonological and lexical change in the dialects of East Devon and West Somerset 1945-1995 (Unpublished PhD Thesis). University of Wales, Swansea.

Malarski, K. (forthcoming). Loss of rhoticity in Cornwall, Devon and Dorset, (Unpublished PhD Thesis). Adam Mickiewicz University, Poznan.

Piercy, C. (2007). A quantitative analysis of rhoticity in Dorset. evidence from four locations of an urban to rural hierarchy of change. CamLing, pp. 199-206. Available online at http://www.ling.cam.ac.uk/camling/Abstracts/Piercy.html.

Piercy, C. (2012). A Transatlantic Cross-Dialectal Comparison of Non-Prevocalic /r/. University of Pennsylvania Working Papers in Linguistics 18 (2), Art. 10. Available online at https://repository.upenn.edu/pwpl/vol18/iss2/10/.

Sullivan, A. E. (1992). Sound change in progress: A study of phonological change and lexical diffusion with reference to and r-loss in the speech of some Exeter school children. Exeter Linguistic Studies 17. Exeter: University of Exeter Press.

Wells, J. C. (1982). Accents of English. An Introduction. Cambridge: University Press (Vol. 1). Williams, M. (1991). Post-vocalic (r) in the urban speech of the Isle of Wight. In J. Holmes (Ed.), Wellington Working Papers in Linguistics 3, pp. 56-66. Available online at https://www.victoria.ac.nz/lals/resources/Publications/wwp/1WWP1991v3.pdf

‘R-atics 6 Paris 7th, 8th november 2019

How many variants of rhotics in Persian? An ultrasound and acoustic study Reza Falahati1 1 University of Applied Sciences, Amsterdam [email protected] This study investigates the distribution of allophonic variants of rhotics in Persian. There is no consensus over the classification of this sound among the scholars in the field of Persian phonology. This is mainly due to the fact that some of these studies consider the environment in their classification as well or some are simply impressionistic research. Samareh (2002), for example, considers both the vocalic environment and the position of /r/ in his impressionistic description and classification of Persian rhotics. According to him, rhotics in Persian could be realized as either trills or fricatives word-finally, depending on the preceding vocalic environment, and as taps or approximants intervocalically. However, Carr (1993) considers three allophones for the phoneme /r/ in Persian: a voiced alveolar tap, a , and a voiced alveolar trill. The voiced alveolar tap is realized intervocalically, the voiceless alveolar trill appears word-finally, and the voiced alveolar trill elsewhere. In one of the few studies on Persian rhotics using acoustic analysis, Rafat (2010) found that the five native speakers in her study used the alveolar fricative as the prominent variant in all/most positions: word-initially, intervocalically, and word-finally. This is quite unexpected compared to the results of the earlier impressionistic studies. Her results also showed that the production of a trill as a singleton exists only word-finally. Falahati and Ooijevaar also in their preliminary study have reported fricatives as the most frequent variant of rhotic in Persian followed by approximants, taps, and trills. The main objective of the current research is to provide articulatory and acoustic data in order to provide a full picture of the allophonic variants of this sound in Persian. This would also allow us to make a comparison between the findings of our study and the earlier descriptions of Persian rhotics. Audio and real time ultrasound video recordings were made from native speakers of Persian. The software AAA and Praat are used for the segmentation, annotation, tongue contour tracing, and acoustic measurements of the rhotics.

References:

Carr, P. 1993. Phonology. London: Macmillan.

Falahati, R. & Ooijevaar, E. .2017. The distribution of rhotic variants in Persian: an ultrasound tongue imaging and acoustic study. AISV, Pisa, Italy.

Rafat, Y. 2010. A sociophonetic investigation of rhotic in Persian. Iranian studies, 43, pp. 667- 682.

Samareh, Y. 2002. Ava Shenasiye Zabane Farsi. Tehran: Markaze Nashre Daneshgahi.

‘R-atics 6 Paris 7th, 8th november 2019

Toward expanding our understanding of variability in L2 liquid production Mihoko Teshigawara1 & Thomas Magnuson2 1 Komazawa University, 2 Freelance [email protected], [email protected] Japanese has a single liquid, and Japanese learners have difficulty acquiring contrasts across liquids in other languages such as English: a topic that has attracted numerous researchers’ attention (e.g., Riney et al., 2000, and literature cited therein). A 42-month longitudinal study, Riney et al. (2000) collected both spontaneous and read speech from 11 Japanese learners of English at two different times. The study focused on the substitutions of the two English liquids /l, r/ with the Japanese liquid /ɾ/, or alveolar flap. They found that /ɾ/ more often substituted for English /l/s than /r/s. The study did not discuss cases involving substitutions among English /l/s and /r/s (e.g., right for light and vice versa). In a more recent preliminary study, Teshigawara (2017) described the L2 English pronunciations of 20 Japanese speakers with proficiency of CEFR B2 or above (comprising half of the study’s available data) and found around half of the speakers had problems distinguishing these two phonemes, with the majority replacing English /l/s with /r/s. The present study revisits the complete dataset to give a fuller account of Teshigawara’s (2017) data based on auditory analysis. With an increased total number of tokens analyzed, the overall tendency did not change. Even at this proficiency level, more than half (11 speakers) were found to have nonstandard /l, r/ pronunciations for 30% or more tokens. Among those speakers, two distinctive populations were identified with different preferences: in pronunciation of the /l, r/ contrast: 1) 5 /r/- dominant speakers, who replaced English /l/’s with /r/’s; and 2) 3 /l/-dominant speakers who did the opposite (i.e., replacing English /r/s with /l/s). In addition, another 3 presumably less- skilled speakers had more than half nonstandard pronunciations for both /l, r/. Across these groups, speakers with fossilized pronunciation of the word “English” [ɪŋɡɹɪʃ] were observed. In addition, the /r/-dominant speakers were also found to exhibit t/d flapping and extra rhoticization of not followed by /r/ (e.g., Ghana became [ɡa˞ːnɚ]), suggestive of association of these features with North American Englishness, a presumable target for these speakers. Different L2 liquid production patterns among speakers with a common L1 leads to the question: why? Could it be that the difference is associated with different L1 liquid production patterns? To investigate this, we present initial findings from a preliminary study underway which examines Japanese English learners’ L1 and L2 liquid production. The Japanese /ɾ/ phoneme covers a variety of liquid realizations, including English /r/-like (i.e., [ɹ]) sounds as well as laterals, which is the case in not only interspeaker but intraspeaker observations (e.g., Magnuson, 2009, 2011). Magnuson’s studies suggest it may be possible to categorize L1 Japanese liquid pronunciation patterns, leading to the intriguing possibility that those patterns may influence L2 tendencies. We anticipate sharing results from a preliminary experiment with 10 Japanese speakers of English along with the expanded analysis of the dataset from Teshigawara (2017).

References: Riney, T., Takada, M., & Ota, M. (2000). Segmentals and global foreign accent: The Japanese flap in EFL. TESOL Quarterly, 34, 4, 711–737. ‘R-atics 6 Paris 7th, 8th november 2019

Teshigawara, M. (2017). Intelligibility and impressions of English spoken by Japanese speakers of English: A perceptual experiment with listeners from five language backgrounds (Part 2). The semiannual periodical of the Faculty of Arts and Sciences Department of English and Department of Foreign Languages, 23, 1–24.

Magnuson, T. J. (2009). What /r/ sounds like in Kansai Japanese: A phonetic investigation of liquid variation in unscripted discourse. MA thesis, University of Victoria.

Magnuson, T. (2011). Realizations of /r/ in Japanese talk-in-interaction. Proceedings of ICPhS XVII, 1306–1309.

‘R-atics 6 Paris 7th, 8th november 2019

Impact of post-lexical context and speech style on word-final /ʁ/ realization in French using large corpora and automatic speech processing Yaru Wu1,2, Martine Adda-Decker1,3, Cédric Gendrot1, Lori Lamel3 1Laboratoire de Phonétique et Phonologie (UMR7018, CNRS-Sorbonne Nouvelle), France, 2Modèles, Dynamiques, Corpus (MoDyCo), UMR 7114, CNRS, 3LIMSI, CNRS, Université Paris-Saclay [email protected], [email protected], cedric.gendrot@sorbonne- nouvelle.fr, [email protected] Post-consonantal word final /ʁ/ in French is known to be unstable [1,2]. In production, French Cʁ# clusters (eg. quatre /katʁ/, four) tend to have several realizations. Speakers could pronounce Cʁ# clusters as they are (quatre [katʁ], canonical form), insert a schwa (quatre [katʁə]), delete the /ʁ/ (quatre [kat]), or even delete the /ʁ/ and insert schwa all together (quatre [katə]). Although Cʁ# cluster variants are frequently observed in daily French, little is known about their realization according to the post-lexical context (cross-word context) and speaking style. Here, we examine the influence of post-lexical (#C vs. #V) context on post- consonantal word final /ʁ/ realization in French with a special focus on the potential impact of speech style. Three corpora were used: ESTER [3], ETAPE [4] and NCCFr [5]. ESTER is composed ofbroadcast news. ETAPE contains journalistic conversations and debates. NCCFr (Nijmegen Corpus of Casual French) comprises conversations between friends. Automatic alignments using the LIMSI speech recognition system in forced alignment mode with optional /ʁ/ and schwa variants [6][7] were used for this study. A subset of the alignments was verified manually by a phonetician. Alignment evaluation was carried out using Cohen's kappa [8]. The two alignments have almost perfect match on /ʁ/ (Kappa = 0.832) and substantial agreement on schwa (Kappa = 0.704). The absence/presence of /ʁ/ and /ə/ were determined by comparing the reference (phonological transcription of Lexique380 [9]) to the aligned variant (speaker’s pronunciation). Statistical analyses were done using generalized linear mixed model [10,11]. Figure 1 shows the alignment results for the canonical form (CAN) and its variants. The two most commonly used variants are "CAN + ə " and "CAN - ʁ" when the following post-lexical context is #C. In contrast, if the next word starts in a vowel, the most commonly used variant is "CAN", without modification. Interestingly, for #V, about 20% of /ʁ/ still disappear during automatic alignment. Variant "CAN - ʁ + ə" is almost never used (≈0%). Figure 2(a-c) suggests that the relative use of the variants depends on the corpus. However, "CAN - ʁ + ə" is almost never used (≈0%), regardless of the corpus. With regard to the formal journalistic corpus ESTER, "CAN + ə" has the highest alignment rate in #C context (65%) and "CAN" is the most frequently selected form when the post-lexical context is a vowel (78%). The same trend is observed with the conversational journalistic corpus ETAPE. Nevertheless, the gap between "CAN + ə" and "CAN - ʁ" is very small in #C context in ETAPE (7% difference) compared to ESTER (50% difference). It can also be noted that the /ʁ/-drop rate increases significantly in #V context for NCCFr, compared to ETAPE. Interestingly, for casual speech (NCCFr) "CAN - ʁ" has the highest alignment rate (69%) in #C context and the same "CAN - ʁ" variant exceeds 40% (41%) for #V context. Interestingly, "CAN" (canonical form) has the highest overall alignment rate among the four variants in #V context, regardless of the corpus. ‘R-atics 6 Paris 7th, 8th november 2019

Figure 1: Alignment rates of the 4 tested pronunciation variants (CAN, CAN + ə, CAN - ʁ, CAN - ʁ + ə) for #C (in brown) and #V (in blue) post-lexical contexts, all data pooled.

Figure 2: Alignment rates of the 4 tested pronunciation variants (CAN, CAN + ə, CAN - ʁ, CAN - ʁ + ə) for #C (in brown) and #V (in blue) post-lexical contexts and from left to right for the 3 corpora ESTER, ETAPE, NCCFr.

References: [1] Côté, M. H. (2004). Consonant cluster simplification in Québec French. Probus, 16(2), 151-201.

[2] Laks, B. (1977). Contribution empirique à l'analyse socio-différentielle de la chute de /r/ dans les groupes consonantiques finals. Langue française, (34), 109-125.

[3] Galliano, S., Geoffrois, E., Gravier, G., Bonastre, J. F., Mostefa, D., & Choukri, K. (2006).

Corpus description of the ESTER evaluation campaign for the rich transcription of French broadcast news. In Proceedings of LREC (Vol. 6, pp. 315-320).

[4] Gravier, G., Adda, G., Paulson, N., Carré, M., Giraudel, A., & Galibert, O. (2012). The ETAPE corpus for the evaluation of speech-based TV content processing in the French language. In LREC-Eighth international conference on Language Resources and Evaluation.

[5] Torreira, F., Adda-Decker, M., & Ernestus, M. (2010). The Nijmegen corpus of casual French. Speech Communication, 52(3), 201-212.

[6] Gauvain, J. L., Lamel, L., & Adda, G. (2002). The LIMSI broadcast news transcription system. Speech communication, 37(1), 89-108. ‘R-atics 6 Paris 7th, 8th november 2019

[7] Adda-Decker M., Lamel L. (2000). The use of lexica in automatic speech recognition. In Lexicon Development for Speech and Language Processing, F. van Eynde & D. Gibbon (Eds.), Kluwer Academic Publishers.

[8] Gamer, M., Lemon, J., Fellows, I., & Singh, P. (2012). irr: Various coefficients of interrater reliability and agreement. R package version 0.84, 137.

[9] New, B., Brysbaert, M., Veronis, J., & Pallier, C. (2007). The use of film subtitles to estimate word frequencies. Applied psycholinguistics, 28(4), 661-677.

[10] Bates, D., Mächler, M., Bolker, B., & Walker, S. (2014). Fitting linear mixed-effects models using lme4. arXiv preprint arXiv:1406.5823.

[11] Team, R. C. (2014). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2014.

‘R-atics 6 Paris 7th, 8th november 2019

Plenary 3

Investigating /R/ production in large corpora Martine Adda-Decker Laboratoire de phonétique et Phonologie (UMR7018, CNRS – Sorbonne Nouvelle) [email protected] Rhotics are very frequent in French, both in lexicon and in speech production. In this contribution, we will focus on large lexica and corpora to compute their distributional frequencies with a focus on embedding consonant clusters. We then present methodologies derived from automatic speech processing to study their realisation in terms of presence/absence in the speech signal, and in terms of realised allophones. A comparison between different speech production situations highlights the differences in /ʁ/ production, in particular as a function of the embedding context. The access to large speech collections enables us to give a quantitatively more precise picture of rhotics production variation. It also questions the role of the brain in restoring missing segments in spontaneous speech.

Production Session – Friday 8th

Why rhoticity is on everyone’s lips: Accounting for the labial gesture in English /r/. Hannah King1 & Emmanuel Ferragne2 1 CLILLAC-ARP, Université de Paris Diderot, 2 Laboratoire de phonétique et Phonologie (UMR7018, CNRS – Sorbonne Nouvelle) [email protected], [email protected] The approximant consonant /r/ has been described as one of the most com- plex phones in the [1]. Articulatory variation has been widely observed, particularly with regards to tongue body shapes, which range from tip down bunched to curled up retroflex [2, 3]. Despite the array of articulatory variation detailed in the literature, notably in rhotic varieties of English, the acoustic profile for /r/ remains remarkably stable; the most salient feature being a very low F3. Articulatory-acoustic models have associated the low F3 with the front cavity, i.e. between the palatal constriction and the lips [4, 5]. Extending the front cavity - through a more posterior palatal constriction, the addition of a sublingual space, or increased lip protrusion - may contribute to the lowering of F3. As bunched /r/ has a low tongue tip, it has negligible sublingual space, contrary to retroflex /r/. [5] posit a “trading relation” between the existence of a sublingual space and the placement of the palatal constriction: bunched /r/ was more posterior than retroflex. A similar trading relation may be possible for lip protrusion. However, there is no study to date that investigates this idea. Although it is generally agreed that /r/ may be accompanied by a labial constriction, the lips tends to be overlooked, and their precise contribution is unknown. The lips are of particular interest in the English spoken in England (Anglo-English) because labiodental variants [V], which supposedly lack an ob- vious tongue body gesture, are becoming increasingly common [6]. We present acoustic and articulatory data (via ultrasound tongue imaging and lip videos) from 24 Anglo-English speakers which suggest that lip protrusion allows speak- ers to maintain a stable acoustic output across different tongue configurations. We observe significantly more lip protrusion in bunched tongue configurations than retroflex ones. Furthermore, evidence from hyperarticulated productions of /r/ reveals that speakers actively control articulatory parameters to enhance the discriminabil- ity of /r/, notably via increased lip protrusion and retroflexion. We engaged speakers in error resolution with a simulated “silent speech” recognition pro- gramme where /r/ was “incorrectly” identified as /w/ or /l/ ‘R-atics 6 Paris 7th, 8th november 2019 one third of the time. As a result, speakers adapted their productions of /r/ to obtain signifi- cantly lower F3 values. Regardless of , lip protrusion significantly increased in hyperarticulation. However, statistical models still predict more lip protrusion in bunched than in retroflex /r/. Retroflex /r/ can undergo fur- ther enhancement through increased curling back, thus increasing the size of the sublingual cavity. We suggest that as this strategy is not available to bunchers, lip protrusion is used as an alternative means to increase the size of the front cavity, and therefore the discriminability of /r/. Our results indicate that lip protrusion is an articulatory mechanism speak- ers employ to enhance rhoticity. As a result, the spread of labiodental variants may be a direct consequence of the perceptual prominence of lips: the labial gesture is perceptually salient enough to supersede the lingual one. Our future research will consider to what extent visual cues may influence the perception of /r/.

References: [1] M. Adler-Bock, B. M. Bernhardt, B. Gick, and P. Bacsfalvi, “The use of ultrasound in remediation of North /r/ in 2 adolescents,” American Journal of Speech-Language Pathology, vol. 16, no. 2, pp. 128–139, 2007. [2] P. Delattre and D. C. Freeman, “A dialect study of American r’s by x-ray motion picture,” Linguistics, vol. 6, no. 44, pp. 29–68, 1968. [3] M. K. Tiede, S. E. Boyce, C. K. Holland, and K. A. Choe, “A new taxonomy of American English /r/ using MRI and ultrasound,” The Journal of the Acoustical Society of America, vol. 115, no. 5, pp. 2633–2634, 2004. [4] Alwan, S. Narayanan, and K. Haker, “Toward articulatory-acoustic models for liquid approximants based on MRI and EPG data. Part II. The rhotics,” The Journal of the Acoustical Society of America, vol. 101, no. 2, pp. 1078–1089, 1997. [5] Y. Espy-Wilson, S. E. Boyce, M. Jackson, S. Narayanan, and A. Alwan, “Acous- tic modeling of American English /r/,” The Journal of the Acoustical Society of America, vol. 108, no. 1, pp. 343–356, 2000. [6] G. Docherty and P. Foulkes, “Variability in (r) production-instrumental perspec- tives,” ’R-atics: Sociolinguistic, phonetic and phonological characteristics of /r/. Etudes & Travaux 4, pp. 173–184, 2001. ‘R-atics 6 Paris 7th, 8th november 2019

Explaining extreme /r/ weakening in utterance-final position: an ultrasound tongue imaging study of rhotic lingual gesture timing using a spontaneous-speech corpus. Eleanor Lawson1 & Jane Stuart Smith2 1 Queen Margaret University, Edinburgh, 2 University of Glasgow [email protected], [email protected] Audio-acoustic studies of derhoticisation in Scottish English have identified weakening and apparent loss of postvocalic /r/ mainly in working-class speech. Many of these studies also note that the greatest degrees of weakening occur in utterance-final position, see e.g. (Romaine, 1978; Speitel & Johnston, 1983; Jauriberry, Sock, Hamm, & Pukli, 2012). The current study investigates the impact of the preboundary location of postvocalic rhotics (i.e. before a syllable, word or utterance boundary) on the timing of their anterior lingual gestures, and on audible strength of rhoticity, using a socially-stratified conversational ultrasound- tongue-imaging (UTI) speech corpus. 14 Glaswegian adolescents aged 12-13 (8 middle-class (MC), 6 working-class (WC), 7 male and 7 female) were recorded carrying out a spot-the-difference task in same-sex dyads. Sampled UTI recordings around 12 minutes long were made using two DP2200 video-based ultrasound machines, with probe-stabilizing headsets and Articulate Assistant Advanced (AAA) software (A. Wrench, 2012). A mean 23 analysable postvocalic /r/-word tokens in stressed and unstressed syllables were obtained per speaker. For each /r/ token gesture lag was measured – a normalised durational measure between the visually-determined maximum of the anterior lingual /r/ gesture and either voicing offset or onset of a following consonant – two events that could potentially audibly mask the /r/ articulation. Each /r/ word was also segmented, exported, anonymized, randomized and rated for rhotic strength on a 5-point index – lower scores for weaker-sounding /r/s. Mixed effects modelling was carried out for the articulatory and audio variables. Fixed factors were: (1) post-rhotic linguistic boundary (syllable, word, utterance); (2) post-rhotic consonant place (dental, alveolar etc,); (3) syllable stress; (4) speaker social class; (5) speaker sex. Random factors were speaker and word. A correlational analysis was also carried out for the two dependent variables. In this spontaneous speech dataset, syllable rimes were found to be significantly longer in utterance- final position across social-class groups, a well-known prosodic phenomenon (Lehiste, 1973; Wightman, Shattuck-Hufnagel, Ostendorf, & Price, 1992). gesture lag was also found to be significantly longer in utterance-final position than before syllable or word boundaries for both social class groups. However, WC speakers showed greater degrees of gesture lag than MC speakers in all boundary conditions and gesture lag for WC /r/ in utterance-final position was particularly long; occurring, on average, after voicing offset, or after the onset of a following consonant. WC /r/ in utterance-final position also had significantly lower /r/-index scores than MC /r/ in this position and in fact a correlational analysis showed a significant negative correlation between gesture lag and /r/-index score; the greater the gesture lag, the weaker-sounding the /r/. Following consonant place was also found to have a significant impact on gesture lag, with earlier /r/ maxima - shorter gesture lags - occurring before coronal consonants [ð, s, t, ʃ etc.] and later /r/ maxima before non-coronal [ɡ, w etc.] and nonlingual [b, m, f etc.] consonants. The findings of this study suggest an /r/ weakening mechanism involving an interaction between lingual articulation, prosody and social-indexical factors. ‘R-atics 6 Paris 7th, 8th november 2019

References: Jauriberry, T., Sock, R., Hamm, A., & Pukli, M. (2012). Rhoticite et derhoticisation en anglais ecossais d'ayrshire. Paper presented at the Proceedings of the Joint Conference JEP- TALN- RECITAL, 89-96.

Lehiste, I. (1973). Rhythmic units and syntactic units in production and perception. Journal of the Acoustical Society of America, 54(7), 1228-1234.

Romaine, S. (1978). Postvocalic /r/ in scottish english: Sound change in progress. In P. Trudgill (Ed.), Sociolinguistic patterns in british english (pp. 144-158)

Speitel, H. H., & Johnston, P. A. (1983). A sociolinguistic investigation of edinburgh speech. (No. End of Grant Report). Economic and Social Research Council.

Wightman, C. W., Shattuck-Hufnagel, S., Ostendorf, M., & Price, P. J. (1992). Segmental durations in the vicinity of prosodic phrase boundaries. Journal of the Acoustical Society of America, 91(3), 1701-1717.

Wrench, A., & Hardcastle, W. (2000). A multichannel articulatory speech database and its application for automatic speech recognition. Paper presented at the Proceedings of the 5th Seminar on Speech Production, may 1-4, Kloster Seeon, Germany. 305-308.

Wrench, A. (2012). Articulate assistant advanced user guide [computer software]. Edinburgh: Articulate Instruments Ltd.

‘R-atics 6 Paris 7th, 8th november 2019

Onset /r/-ticulation in Mandarin Kaiyue Xing1 1 The University of Manchester [email protected] There are conflicting reports in the literature concerning the articulation of onset /r/ in Mandarin Chinese. It is reported as post-alveolar fricative /ʐ/ by Gao (1948) and Chen, et al (2017), but as retroflex approximant /ɻ/ by Wang (1983) and Chao (1968). This study presents new articulatory data on the realisation of onset /r/ in Mandarin, which show additional aspects of /r/-variation conditioned by both prosodic position and vocalic environment. The test items included /r/ in word-initial and word-medial position, the latter derived through compounding. The word-initial /r/ was followed by the vowels /ʅ/ (apical vowel), /u/, /ɤ/, /au/ and /əu/. The word-medial /r/ was set in of context of /Cau.rəu/. The onset of the first syllable was /tʰ/, /k/ or /n/. Eight males and eight females speaking Beijing dialect participated in the ultrasound recording and read six repetitions of eight tokens of onset /r/. The midsagittal tongue contours were automatically tracked throughout the recording with hand correction as necessary. The analysis below focuses on ultrasonic corresponding to the maximum constriction of /r/. The current data confirmed the presence of both /r/-variants mentioned in previous literature, subject to inter-speaker variation. The /r/ can be produced with retroflexion (bending backwards of the tongue tip) for some speakers, and without retroflexion (the grooving of tongue blade or a flat tongue) for other speakers, as shown in given examples in Figure 1. The retroflex gesture corresponded to an approximant quality with little or partial frication in the acoustics, while the non-retroflex showed stronger frication level and longer frication period. However, both tongue configurations were characterised by the merging of F2 and F3 of the adjacent vowels during the onset /r/ portion. In addition, the realisation of onset /r/ is conditioned by the following vowel. When followed by the vowel /au/ (“ao” in ), the word-initial /r/ was never fricated. Articulatorily, the tongue front was lowered and tongue body was slightly retracted in “rao”, compared to any other vocalic contexts, regardless of the tongue configuration of the onset /r/, as shown in Figure 2. Onset /r/ is subject to further variation induced by prosodic position. As shown in Figure 3, compared with word-initial onset /r/, the word-medial onset /r/ was characterised by no frication and the split of F2 and F3, which was consistent with a slightly retracted tongue body. In sum, the lack of frication for the onset /r/ in “rao” and word-medial position indicated that the more fricated variants alternated with the more approximantal variant. However, this dramatic acoustic shift involved relatively little variation in the articulation. Additionally, the onset /r/ in these two cases presented a third distinct option of /r/ realisation with approximant quality and non-retroflex gesture. The present data reveal that the realisation of Mandarin onset /r/ is conditioned by different speakers. The prosodic segmental contexts and vocalic environment can trigger the alternation between fricated variants and canonical approximant rhotics. I discuss this in the context of defining fricated variants as rhotics.

‘R-atics 6 Paris 7th, 8th november 2019

Figure 1. Tongue shapes of the word-initial onset /r/ in /rəu/ and their corresponding spectrograms for the speaker M03 (left) and M05 (right).

Figure 2. Smoothed tongue shapes of word-initial onset /r/ at the maximum constriction followed by different vowels for M03 (left) and M02 (right). The red lines present the tongue shape followed by the vowel /au/.

‘R-atics 6 Paris 7th, 8th november 2019

Figure 3. Onset /r/ in word-initial and word-medial position (upper panel) produced by speaker F04 and the corresponding tongue shape comparisons at the maximum constriction (lower panel).

References: Chao, Yuenren. (1968). A grammar of spoken Chinese. Berkeley, CA: University of California Press.

Chen, S., Mok, P. P. K., Tiede, M., Chen, W. R., & Whalen, D. H. Investigating the production of Mandarin rhotics using ultrasound imaging. Ultrafest VIII, 123(6), 17.

Gao. Benhan. Zhao.Renyuan & Li fanggui (1948). The study of Chinese Phonology. Shangwuyinshu Press. Wang. Li. (1983). The acoustics of Rimu and Consonant Table of Putonghua, Chinese (1).

‘R-atics 6 Paris 7th, 8th november 2019

Fuse unless tense: Rhotic-vowel interaction in Elbe-Weser German Daniel Gleim1 & Sören Eggert Tebay1 1Leipzig University [email protected], [email protected] The particularities of coda-ʁ vocalisation in the northern German varieties have been recognised, however, so far they not been subject to phonetic measurements and a detailed theoretical analysis. Our acoustic study shows that coda-ʁ vocalization depends on tenseness of the preceding vowel. We claim that /ʁ/ is underlyingly specified as [–t(ense)] (cf. Wiese 2000). The diverging behaviour of tense and lax vowels, including the typologically unexpected tensing, can than be attributed to OCP induced dissimilation. Data: In most German varieties, consonantal coda-ʁ is illicit and triggers a wide array of repairs (Hall, 1993; Wiese, 2000). In northern varieties, these repairs are less uniform and depend largely on the preceding vowel. We focus on new data from the Elbe-Weser variety of colloquial northern German (EW). Here, tense vowels are stable before vocalised /ʁ/ (1). Lax vowels on the other hand fuse into a long, lax vowel. Non-low lax vowels optionally escape fusion by tensing (2). High vowels are also lowered and rounded (cf. Krech 1982). We will disregard schwa.

Speakers are on the verge of neutralizing height and tenseness contrasts before /ʁ/, already completed for low vowels (Wiese, 2000). Acoustic Study: We recorded 120 items from two native speakers of EW in a carrier sentence from two conditions. In the first condition vowels appeared before coda-ʁ , whereas they were followed by an onset-ʁ in the second. Each vowel occured in both. F1 and F2 were measured at 25%, 50% and 75% of the vowel length with Praat (Boersma & Weenink, 2017). The results of this study show that high lax vowels have a higher F2 (i.e. are lowered) before coda-ʁ, as shown in (3). We additionally analyzed diphthongization as the sum of the euclidian distances between points of measurement and found that high lax vowels are least likely to surface as diphthongs, whereas tense mid vowels are most likely. This supports our impressionistic data in (1) and (2). We assume that consonants may be (language-specifically) specified for vocalic features. More precisely, in EW the rhotic /ʁ/ is specified as [-t]. In addition, we adopt OCP constraints relativized to domains in an OT system (Suzuki, 1998). Analysis: /ʁ/-vocalization is triggered by the constraint *ʁ]σ against consonantal rhotics in codas. EW favours [5] as the vocalic counterpart of /ʁ/. Nevertheless, lax vowels undergo fusion with /ʁ/, because they would otherwise violate the OCP constraint against two [-t] segments in the rime of a syllable. Another way to avoid the OCP violation is dissimilation. The OCP constraint is not activ for tense vowels, since they are underlyingly [+t]. Fusion will therefore never become optimal for such an input. The source of rounding in /Iʁ/ → [œ:] remains unclear. ‘R-atics 6 Paris 7th, 8th november 2019

We provide new data and acoustic evidence for complex interactions between rhotics and vowel. The analysis of ʁ-vocalisation in EW adds to the evidence for vocalic features on consonants (cf. Kostakis 2015). Both tensing and fusion are strategies to avoid an OCP(-tense) violation.

References: Boersma, Paul & David Weenink. 2017. Praat: Doing phonetics by computer. Software. Hall, Tracy Alan. 1993. The phonology of German R. Phonology 10(1). 83–105.

Kostakis, Andrew. 2015. Height, frontness and the special status of/x/, /r/ and /l/ in Germanic language history: Indiana University dissertation.

Krech, Eva-Maria. 1982. Großes Wörterbuch der deutschen Aussprache. Leipzig: Bibliographis- ches Institut.

Suzuki, Keiichiro. 1998. A typological investigation of dissimilation. Tucson, AZ: University of Arizona dissertation.

Wiese, Richard. 2000. The phonology of German. Oxford: Oxford University Press. Originally published: 1996.

‘R-atics 6 Paris 7th, 8th november 2019

The Acoustic Properties of Rhotic Post-Velarization in Moroccan Arabic Aaron Freeman1 1 University of Pennsylvania [email protected] In Arabic, contrastive post-velar secondary articulations distinguish a class of consonants known as ‘emphatics’ from their non-post-velarized counterparts. While this post-velarization feature is only fully contrastive in the case of coronal , and non-coronals sporadically exhibit post-velarization as well in a variety of Arabic dialects. The ‘emphatic R,’ typically transcribed as [rˤ], is the most widely distributed and distinctive of these marginal post- velarized variants (Younes 1994). This study investigates the phonetic properties of [rˤ] in the Fes variety of Moroccan Colloquial Arabic, showing that it is associated with a wider range of acoustic behavior than the emphatic coronal obstruents, and, in some instances, has more phonetic similarity to uvular than to pharyngealized consonants. The most distinctive acoustic effect associated with Arabic post-velar consonants is coarticulatory backing and/or lowering of adjacent vowels. Raised F1 alone is associated with the lower pharyngeal consonants [ʕ ħ], while uvular and ‘emphatic’ consonants are associated with a combination of raised F1 and lowered F2 (Ghazeli 1977). For the emphatic coronal obstruents, this effect has been phonologized into a long-range process known as ‘emphasis spread’. Uvulars and segments such as [rˤ] are sometimes also reported to cause emphasis spread (Heath 1987), but in some poorly-defined ‘attenuated’ way. To better understand the behavior of these segments, I conducted an acoustic analysis of post-velar coarticulation in the recorded speech of 23 Fessi Arabic speakers. My results indicate that the [q] causes consistently less F2 lowering of adjacent vowels than do emphatic coronal obstruents, approximately 200 Hz as compared to 400 Hz in the case of /a/. The uvular stop was also not observed to trigger long-range emphasis spread. [rˤ] was not found to behave as predictably. In certain words, it caused the same pattern of post-velar coarticulation as [q]. In other cases, it triggered the more extreme F2 lowering and long-range spread associated with emphatic coronal obstruents. Fig. 1 illustrates the uvular-type [rˤ] in the word [ʒrˤanat] ‘frogs’, in which the F2 of /a/ after [rˤ] is not significantly different from F2 of /a/ after [q] in [qabla] (t(27)=0.32, p=.08), but is significantly different from /a/ after both [t] and [tˤ]. In contrast, Fig. 2 illustrates that in the words [farˤ] and [barˤ], mean F2 of /a/ is closer to that in [batˤ] than to that in [baqi]. This difference is not simply a directional asymmetry, but proved to be a lexically conditioned effect which varied between speakers. Rhotic post-velarization, then, has acoustic properties which are similar sometimes to those of uvular consonants, and sometimes to those of other post-velarized coronals. This irregularity is difficult to explain, but may be related to the articulatory variability typical of rhotics (Lindau 1985), including in the Fessi dialect (Hachimi 2007). A wider range of articulatory targets, combined with a representationally ambiguous phonological distribution, may lead to systematic inconsistency in the acquisition of [rˤ]. More research will be needed to fully understand this peculiar and understudied phenomenon in Arabic.

‘R-atics 6 Paris 7th, 8th november 2019

‘R-atics 6 Paris 7th, 8th november 2019

Articulatory variation in Southern British English rhotics Patrycja Strycharczuk1, Susan Loyd2 & James Scobbie2 1The University of Manchester, 2Queen Margaret University [email protected], [email protected], [email protected] Southern British English (SBE) rhotics are well-studied, but mostly from the point of view of whether they occur or not in specific contexts. In contrast, the quality of the /r/ in this variety has received less empirical attention. It is commonly described as a post-alveolar approximant [2], although other variants, e.g. labio-dental, have also been noted [3, 5]. In this paper, we present ultrasound data from 36 speakers of SBE, documenting the presence of multiple articulatory /r/-variants in SBE, including a plain approximant, but also retroflex and bunched /r/, previously observed in rhotic varieties of English [4, 1]. We collected the data during a public outreach event, featuring a live demo of ultrasound speech imaging. Visitors were invited to look at their tongue using ultrasound (Echo-B system from Articulate Instruments). Some of them were recorded reading two sentences, subject to verbal consent. For these speakers, we recorded the ultrasound data, captured using a hand- held probe, as well as time-aligned audio data, captured with a lapel microphone. The stimuli were designed to include rhotics in a variety of prosodic and morphosyntactic contexts. 38 speakers of SBE, aged 16–78, donated their data, including recordings and basic demographic information: age, gender, geographical background, education and occupation. Two were discarded due to poor ultrasound image. Our analysis is based on visual inspection of the ultrasound images, selected at the point of maximum /r/ constriction. Each instance of /r/ was classified as a particular articulatory variant. As we generally found very little intra-speaker variation, we classified each speaker according to the main variant they used. Examples are in Figure 1. 13 speakers realised /r/ as a plain approximant. Of the remaining 23, 10 had a retroflex /r/, and 10 had a bunched /r/. In three cases, the /r/ was difficult to classify as either retroflex or bunched, as the tongue tip was not clearly visible, although the configuration of the pre-dorsum suggested that it was not a plain approximant. Based on binomial logistic regressions, the likelihood of having a plain approximant in- creased with age (p<0.001), and younger speakers were more likely to produce tip-down /r/ (p<0.05). The two findings are correlated, since plain approximants are invariably tip-up. Our data suggest an apparent-time decrease in the use of plain approximant /r/ in SBE, while retroflex and bunched /r/ are increasingly used by younger speakers. This change in progress could signify a shift towards a new acoustically defined /r/-variant, associated with covert variation. Alternatively, it could be seen as a shift towards increased lingual variation, whose acoustic consequences are obscured by the accompanying labial gesture. We consider these two hypotheses in the context of diverging acoustic predictions they make. We also provide a methodological evaluation of collecting articulatory data in the context of a public event. There is a methodological compromise between being able to survey a relatively large number of participants in a short period of time, and acquiring high-quality articulatory data. We discuss this, focusing on the analytical possibilities for the data we collected.

‘R-atics 6 Paris 7th, 8th november 2019

Bunched

Retroflex

Plain approximant

Figure 1: Examples of the articulatory variants in our data

References: [1] Suzanne Boyce et al. “Diversity of tongue shapes for the American English rhotic liquid”. In: Proceedings of the 18th International Congress of Phonetic Sciences, Glasgow. 2015. [2] Paul Carley, Inger M Mees, and Beverley Collins. English phonetics and pronunciation practice. Abingdon, UK/New York, NY: Routledge., 2017. [3] Paul Foulkes and Gerard J Docherty. “Another chapter in the story of /r/: ‘Labiodental’ variants in British English”. In: Journal of Sociolinguistics 4 (2000). [4] E. Lawson, J.M. Scobbie, and J. Stuart-Smith. “The social stratification of tongue shape for postvocalic /r/ in Scottish English”. In: Journal of Sociolinguistics 15 (2011). [5] Eleanor Lawson et al. “Seeing Speech: an articulatory web resource for the study of Phonetics”. In: University of Glasgow, 2018.

‘R-atics 6 Paris 7th, 8th november 2019

Plenary 4

Phonetic and phonological variations in rhoticity from overlapping perspectives of self, other and language Jim Scobbie Queen Margaret University [email protected] The phonetic nature of rhoticity is highly variable in many languages, in part because a wide variety of phonetic productions are labelled as “rhotic”, forming a blended family with a range of articulatory and acoustic characteristics. There are also often interesting patterns of phonotactic distribution, of position-specific contrast among rhotics. Sociophonetic variation and varied patterns of acquisition are also fruitful areas of study when approaching rhoticity. Finally, phonological change provides another locus of variation, as aspects of a system are phonologised, change phonetically, change phonological category e.g. from consonant to vowel or vice versa, merge, or develop into new contrasts. It can be hard to distinguish phonology from phonetics in the simplest of cases. Some phenomena can be fruitfully viewed as being in the overlap between these interpretations of phenomena, rather than being necessarily allocated exclusively to one domain or the other. Rhotics seem to be a particularly problematic – and hence useful – class. In previous work with Stuart-Smith and Lawson, I have adopted a neo-Ohala model of a speaker-hearer triangle (rather than a Speech Chain model) as a way of understanding apparently “mismatches” in articulatory-acoustic characteristics of /r/ (as well as to describe covert articulations in both acquisition and in clinical disorders). In this talk I will consider the phonetics-phonology interface in an augmented tri-dimensional space that spans both phonetics and phonology, conceptualised as three intersecting perspectives: (a) the self (b) others and (c) language. Consideration of The Self leads to a focus on speech planning and production, articulation, L1 phonology, and acquisition. Consideration of Others provides a primarily social perspective, one that is acoustic (and visual), perceptual and pluralistic. The language perspective most clearly focuses on phonological contrast, lexical functionality, typology, literacy, and abstract representation. Overlap between these perspectives is of course, the norm: a “pure” single viewpoint or single interface is rarely attainable, which is why traditionals approaches deriving surface forms from underlying representations (or notational variants thereof) are partial and flawed. I will focus on rhotics in English, using the model to interpret some well-known contemporary issues such as articulatory variation in /r/, lenition of coda /r/, rhoticisation in onsets, and the phonological implications of such variation. I will also present some new empirical data arising from work with Schaeffler on bi- dialectalism. Teenage children of adults with incompatible systems face a challenge in developing and crystallising their own system, or systems. For example – faced with rhotic and non-rhotic accents of English, what do people do? Just as importantly for us: what CAN they do? A greater acceptance of variation and our attempt to force phonetic H-and-H concepts into phonology led us to use more flexible approaches to laboratory-based elicitation. Data therefore showing a wide range of variable behaviours, linking back to the more complex multi-dimensional approach of multiple interfaces between self, other and language. The speakers’ sometimes inconsistent and stereotype-using behaviours show interesting patterns in their successful and failed attempts to produce appropriate rhotic or non-rhotic forms in familiar canonical accents.

‘R-atics 6 Paris 7th, 8th november 2019

Acquisition Session – Friday 8th

(Socio)phonetic variation in the production of rhotics by Italian schoolchildren Chiara Meluzzi1 1 Università di Pavia [email protected] In this work, we aim at investigating how rhotics are produced by Italian schoolchildren aged between 6 and 10 y.o. In particular, we are interested in seeing whether the phonological opposition between singleton and geminated rhotics (Payne 2005) is conveyed only by phonetic durational cues (e.g., rhotic and preceding vowel duration) and/or by rhotic type, with singletons commonly produced as taps and geminate rhotics appearing as trills, similar to what happens for other languages such as Spanish (Hualde 2004). We recorded 80 Italian mother-speakers from North-West Italy performing a naming task of pictures presented on a map, with 20 rhotics occurring as singletons, geminates, and in pre- and post-consonantal position. To avoid the production of different lexical items from the ones expected, we add the target word under the picture. Recordings were conducted in a separated room within the elementary school children attended during the normal hours of lessons, one pupil at a time. Recordings were made with a TASCAM DR-40 recorder with an external Sony microphone, in .wav format with a sampling rate of 44.1 KHz. Besides phonetic data, we also collected sociolinguistic questionnaires that pupils compiled at home with their parents, in order to map each participants’ social status and personal origins from different areas of the Italian peninsula. The 1600 tokens were manually isolated and annotated in PRAAT by following the protocol provided in Celata et al. (2016): rhotics are considered to be produced as sequences of gestures of constrictions and apertures, being either monophasic, in case of intervocalic taps, or multi- phasic, such as in case of pre- and post-consonantal singleton rhotics or with trills. The ongoing analysis is thus focusing on two different elements of rhotic production: rhotic type and durational cues. We wanted in particular to answer three research questions: 1) If rhotic type varies in different phonotactic contexts, and, if so, whether the expected opposition singleton/tap vs. geminated/trill is systematically acquired. 2) If the durational cues associated with gemination are also affecting rhotics in singleton vs. geminated contexts. 3) If sociolinguistic variables, in particular age of speakers’ and their origins, play a role in shaping variation in the production of rhotics. Mixed model analysis will be performed in IBM SPSS 20 to determine the impact of the different variables. We expect that speakers’ age affect rhotic type production, especially in the case of trills in geminated position, which are notoriously quite difficult to acquire in many languages (e.g., Carballo & Mendoza 2000). Moreover, we also expected an influence of age in the acquisition of phonetic cues associated to gemination. Finally, it is up to investigation if pupils’ dialectal background derived from their familiar origins influences rhotic production, since dialectal differences are quite attested in literature (Rohlfs 1996).

References: Carballo, G. & Mendoza, E. (2000) Acoustic characteristics of trill productions by groups of Spanish children. Clinical Linguistics & Phonetics 14: 587-601. ‘R-atics 6 Paris 7th, 8th november 2019

Celata, C., Meluzzi, C. and Ricci, I. (2016), The sociophonetics of rhotic variation in Sicilian dialects and Sicilian Italian: corpus, methodology and first results. Loquens, 3,1.

Hualde, J. (2004) Quasi-phonemic contrasts in Spanish. In Vineeta Chand, Ann Kelleher, Angelo J. Rodriguez, & Benjamin Schmeiser (eds.), WCCFL 23: Proceedings of the 23rd West Coast Conference on Formal Linguistics, 374-398.

Payne, E. (2005), Phonetic variation in Italian consonant gemination. Journal of the International Phonetic Association 35, 2, pp. 153-181.

Rohlfs, G. (1966), Grammatica storica della lingua italiana e dei suoi dialetti. Torino, Einaudi.

‘R-atics 6 Paris 7th, 8th november 2019

Acquisition of postvocalic /r/ by Mandarin-speaking learners of English Zeyu Li1 1University of Münster [email protected] Since rhoticity arguably constitutes “the most important phonotactic difference among accents of English” (Melcher and Shaw, 2003: 19), and a multitude of linguistic and extra-linguistic constraints is known to modulate its realisation, postvocalic-/r/ forms an instructive linguistic marker for investigating the dynamics of second language (L2) phonological acquisition. While the realization of the /r/ variant has been frequently examined from both diachronic and synchronic perspectives on English world-wide (e.g. Becker 2014; Chand 2010; and Li and Kabak 2017), little is known about its perception and production by L2 English learners (but see Sönning 2017 and Colantoni & Steele 2007 for the production of coda /r/). This study thus investigates the acquisition of English postvocalic-/r/ by advanced Mandarin- speaking learners of English, whose first language (L1) has various degrees of rhoticity depending on their places of origin in China. A total of 47 subjects (31f, 16m; aged 17-20) participated in a series of speech perception and production tasks twice, before and after six months of phonological training (T1 and T2). The perceptual ability of the learners was assessed via an oddity discrimination task. Their production was elicited by means of a picture-naming and a reading task. The /r/ tokens (n = 3344) were analysed auditorily by two phonologically- trained coders (inter-coder agreement: 94%). This study further draws on the variationist sociolinguistic approach and controls for potential external (e.g. gender) and internal (e.g. preceding vowels) factors predicting patterns of rhoticity. The data were analyzed by mixed-effects logistic regression models in R. Concerning perception, the results showed that the discrimination ability of the L2 learners improved significantly from T1 to T2, albeit with inter- and intra-learner variability. Preceding phonetic context was shown to be the only significant factor conditioning perceptual patterns of rhoticity following the hierarchy of SQUARE > FORCE > NEAR > START > LETTER, NURSE. Both the acoustic salience of various preceding environments (duration and modulation of F2 and F3) and the cross-linguistic similarity of coda /r/ between Mandarin and English can possibly account for the pattern. The latter is consistent with the hypothesis predicted by the Speech Learning Model (Flege 1995), which predicts that L2 learners are able to establish new categories for sounds absent in their L1. With respect to production, no significant changes were found between the L2 learners’ production at T1 and T2. The results further revealed that their production was controlled by the interplay of L1- and L2-related factors over time. More specifically, the learners’ L2 (NORM PREFERENCE) overrode their L1 (PLACE OF ORIGIN and STYLISTIC TRANSFER) and exerted stronger impact on their production performance at T2. Like perception, the production of the learners was also constrained by preceding contexts in that back vowels disfavour the presence of coda /r/ relative to front vowels. This production pattern in the advanced L2 learners of English contradicts the ascertained perceptual pattern, however, and thus poses a challenge to the purported perception-production link in L2 speech learning (Isbell 2016). In addition, female speakers were found to be more rhotic than males.

References: ‘R-atics 6 Paris 7th, 8th november 2019

Becker, K. (2014). (r) we there yet? The change to rhoticity in New York City English. Language Variation and Change, 26(2), 141-168.

Chand, V. (2010). Postvocalic (r) in urban . English World-Wide, 31(1), 1-39.

Colantoni, L., & Steele, J. (2007). ACQUIRING /[alveolar approximant]/ IN CONTEXT. Studies in Second Language Acquisition, 29(3), 381-406.

Flege, J. E. (1995). Second language speech learning: Theory, findings, and problems. Speech perception and linguistic experience, 233-277.

Isbell, D. R. (2016). The Perception-Production Link in L2 Phonology. MSU Working Papers in Second Language Studies, 7(1).

Li, Z., & Kabak, B. (2018). Rhoticity in Chinese English: An experimental investigation on the realization of the variant (r) in an Expanding Circle variety. Alicante Journal of English Studies, (30), 61-91.

Melcher, G., & Shaw, P. (2003). World Englishes. London: Arnold.

Sönning, L. (2017). Phonological variation in German learner English. (Unpublished PhD dissertation). University of Bamberg, Germany.

‘R-atics 6 Paris 7th, 8th november 2019

The acquisition of the Spanish trill in child and adult heritage speakers Gemma Repiso-Puigdelliura1 & Ji YoungKim1 1University of California, Los Angeles [email protected], [email protected] The present study explores the development of the Spanish trill /r/ by US Spanish heritage speakers (SHSs). /r/ is produced with 2-3 brief contacts between the tongue tip and the (Hualde, 2005), which requires complex coordination of articulators and a sufficient amount of oropharyngeal pressure (Solé, 2002; Lewis, 2003). Thus, typically developing monolingual children do not have full command of this sound until the age of 7 (Bosch, 1983). With respect to SHSs, studies have shown that adult SHSs tend to produce /r/ with a single occlusion using the same articulation as the apico-alveolar tap /ɾ/, although they maintain the /r/-/ɾ/ contrast using durational cues (Amengual, 2016; Henriksen, 2015). The present study compares semi-spontaneous speech of child and adult SHSs to further understand how SHSs acquire the Spanish /r/. 16 college-aged adults (M=20.1 years, SD=1.7) and 20 children (M=10 years, SD=1.2) participated in the study. /r/ was elicited using a wordless picture book. All instances of /r/ were annotated by two coders based on the categories adapted from Rose (2010) (i.e., trill, approximant trill, tap, approximant tap, perceptual tap, tap+fricative, fricative, other) (K=0.868, p<0.001). Moreover, the number of occlusions, segment duration, and the duration of the first aperture of /r/ were analyzed. Results of preliminary data with 6 child SHSs and 6 adult SHSs showed that the adults produced significantly higher rates of target-like /r/ (i.e., trill) (Mean=32.11%, SD=46.8%) than the children (Mean=5.98%, SD=23.82) (Estimate=-2.524, SE=1.272, z=-1.984, p<0.05) and their /r/ was produced with a significantly higher number of occlusion (Mean=1.41, SD=0.95) than that of the children (Mean=0.72, SD=0.82) (Estimate=-2.524, SE=1.272, t=- 1.984, p<0.05). With regard to the duration of /r/, no significant difference was found between the children (Mean=66.1 ms, SD=38.6 ms) and the adults (Mean=66.4ms, SD=35.1ms). Similarly, the duration of the first aperture of target-like /r/ of the children (Mean=22.4ms, SD=9.6ms) and the adults (Mean=21.8ms, SD=6.7ms) did not significantly differ from each other. These durational values are within the ranges found for non-heritage Spanish native speakers, both children (Carballo & Mendoza, 2000) and adults (Henriksen, 2015). Our preliminary data showed that, while the child and adult SHSs in the present study articulated /r/ in a non-target-like manner (i.e., low trill rates and production of less than two occlusions) (supporting the account of acquisition without mastery [Montrul, 2016, 2018]), the adults performed better than the children (supporting delayed acquisition [Flores et al., 2017]). Further analyses will be conducted on the complete dataset to test the robustness of these patterns.

References: Amengual, M. (2016). Acoustic correlates of the Spanish tap-trill contrast: Heritage and L2 Spanish speakers. Heritage Language Journal, 13(2), 88-112.

Bosch, L. (1983). El desarrollo fonológico infantil: Una prueba para su evaluación. Anuario de Psicología, 28, 85–114. ‘R-atics 6 Paris 7th, 8th november 2019

Flores, C., Santos, A.L., Jesus, A., & Marques, R. (2017). Age and input effects in the acquisition of mood in Heritage Portuguese. Journal of Child Language, 44(4), 795-828.

Henriksen, N. C. (2015). Acoustic analysis of the rhotic contrast in Chicagoland Spanish: An intergenerational study. Linguistic Approaches to Bilingualism, 5(3), 285-321.

Hualde, J. I. (2005). The Sounds of Spanish with Audio CD. Cambridge University Press.

Lewis, A. M. (2004). Coarticulatory effects on Spanish trill production. In Proceedings of the 2003 Texas Linguistics Society Conference (Vol. 116, p. 127). Somerville, MA: Cascadilla Proceedings Project.

Montrul, S. (2016). The acquisition of heritage languages. Cambridge, UK: Cambridge University Press.

Montrul, S. (2018). Heritage language development: Connecting the dots. International Journal of Bilingualism, 22(5), 530-546.

Rose, M. (2010). Differences in discriminating L2 consonants: A comparison of Spanish taps and trills. In Selected proceedings of the 2008 second language research forum (pp. 181-196). Somerville, MA: Cascadilla Proceedings Project.

Solé, M. (2002). Aerodynamic characteristics of trills and phonological patterning. Journal of Phonetics, 30: 655-88. ‘R-atics 6 Paris 7th, 8th november 2019

Acquiring rhotics in L2 English and L3 Polish – a comparison of adult and child learners Christina Golin1, Romana Kopeckova1 & Ulrike Gut1 1 University of Münster [email protected], [email protected], [email protected] The acquisition of rhotics of a foreign language has been shown to pose difficulties for learners (e.g. Syed 2013, Colantoni & Steele 2008, Wieden & Nemser 1991), but accuracy seems to increase with years of instruction and the choice of the substitute changes during development. We investigated speakers of L1 Standard High German, which has the uvular voiced fricative [ʁ], with L2 British English, which has the voiced post-alveolar approximant [ɹ], and L3 Standard Polish, which has the alveolar tap [ɾ] or trill [r]. The aim of the longitudinal study was to investigate (i) how learners of two foreign languages with different proficiency levels acquire the rhotic sounds in each of them, (ii) whether the learning trajectories in the two foreign languages differ, and (iii) whether adults outperform children. The participants were 7 adult (aged 21-39) and 9 child (aged 12-13) learners of English (starting age 9 and 6 years respectively) and of Polish (absolute beginners). They acquired both English and Polish in a formal learning setting, with exposure to a variety of native speaker productions. The learners were recorded doing a picture naming and a delayed repetition task containing the rhotics both word- initially as a singleton and word-medially. The data included their L1 German, L2 English and L3 Polish after five weeks of instruction in Polish (T1), after ten weeks (T2) and after 10 months of learning Polish (T3). A total of 1328 tokens of rhotics were analysed auditorily and acoustically (consonant duration, the median of the harmonics-to-noise ratio, mean F3). The findings showed that, as a group, the adults produced the rhotic more accurately than the children in both English and Polish. Very little increase in accuracy is seen in either group for either of the languages, but substitutes for the rhotic change. For the adults, the pronunciation of the rhotic in their L1 German becomes more variable at T3, albeit always within the phoneme inventory of their L1, for example including voiceless velar and uvular fricatives. This is also true for the adults’ substitutes in the production of L2 English and L3 Polish rhotics. In contrast, the children demonstrated a range of realisations, with some L1 German rhotic productions traceable to their L2 and L3 (e.g. [w], [ɹ ] and taps respectively) and some L3 Polish rhotics to their L2, especially at the last testing time. As such, the children showed a greater degree of permeability among all of their languages in the rhotic productions than the adults did. In their L2 English, both learner groups showed a high confusion rate between [/ɹ ]/ and [/w]/, which decreased more for the adults between T1 and T3 (from 20% to 13%, as compared to the children’s 26% to 23%). Overall, it appears that the manner of a rhotic in the foreign languages of the child and adult learners was produced faithfully more often than its place.

References: Colantoni, Laura & Jeffrey Steele. 2008. Integrating articulatory constraints into models of second language phonological acquisition. Applied Psycholinguistics 29(3). 489–534. ‘R-atics 6 Paris 7th, 8th november 2019

Syed, Nasir A. 2013. Acquisition of English [ɹ] by adult Pakistani learners. In Lorenzo Sprafico & Alessandro Vietti (eds.), Rhotics: New data and perspectives, 41–56. Bozen: Bolzano University Press.

Wieden, Wilfried & William Nemser. 1991. The pronunciation of English in Austria. Tübingen: Narr.

‘R-atics 6 Paris 7th, 8th november 2019

On the role of orthography in the acquisition of Portuguese /ɾ/ by Mandarin native speakers Silke Hamann1 & Chao Zhou2 1 University of Amsterdam, 2 Universidade de Lisboa [email protected], [email protected] The European Portuguese (EP) tap /ɾ/ is notoriously difficult for L1-Mandarin learners. Perception experiments with naïve Mandarin listeners showed that EP /ɾ/ is perceived as native Mandarin /l/ rather than the native retroflex rhotic /ɻ/ (authors), presumably because of its relatively high third formant and short duration. In addition, Mandarin learners of EP display an asymmetry between onset and coda position with respect to replacement strategies: While in onset the lateral is the norm, in coda learners either do not realize the tap at all, replace it with /l/, insert schwa (realizing the syllable /lə/) or use the retroflex /ɻ/ (Zhou 2017, Liu 2018; also Patience 2018 on Mandarin learners of Spanish /ɾ/). In this presentation, we argue that replacement with the retroflex rhotic is due to orthographic influence. We provide a formal account of the observed replacement strategies, employing a generative model that integrates not only phonetics and phonology (BiPhon; Boersma 2007) but also orthography (reading grammar by Hamann & Colombo 2017). In a first step, the perception results are formalized with the help of cue constraints, evaluating the mapping between auditory and phonological surface form, and a phonotactic constraint that bans /l/ in Mandarin coda position (not illustrated here). This interaction of constraints accounts for the onset/coda asymmetry exemplified in (1):

auditory input perceived as EP orthography meaning (1) onset [duɾu] /dulu/ ‘hard’ coda [pɔɾku] /pɔku/, ‘pig, pork’ /pɔləku/ The perceived phonological forms are stored in the lexicon, and are then activated in production. Since Mandarin learners acquire the sound structure in parallel with the written form, as many L2 learners do, we assume that these learners lexically store the written form of each EP word along with its phonological form. In the production process, both lexical forms are activated and compete in the creation of an output. This second step in our modelling is formalized in (2) where underlying form and orthographic form (in green) both serve as input. The grapheme represents both EP /ɾ/ and Mandarin /ɻ/ (in Pinyin), and the mapping between and surface /ɻ/ is regulated by the L1 orthographic constraint in green. Whereas in (2a) faithfulness to stored |l| or |lə| is more important than the orthography (IDENT and MAX high ranked), written can show its influence when the phonological form does not have a corresponding sound, as in (2b). The present account of native orthographic influences on the L2 acquisition ‘R-atics 6 Paris 7th, 8th november 2019

of the EP tap by Mandarin learners is transferable to other cases of orthographic influences on L2 rhotic realizations, but could also provide insights into orthography as possible unifying factor in phonetically diverse L1 rhotic categories.

References: Boersma, Paul (2007). Some listener-oriented accounts of h-aspiré in French. Lingua 117: 1989–2054.

Liu, Wen (2018). Aquisição da Vibrante Simples [r] pelos Alunos Chineses Aprendentes de Português como Língua Estrangeira. MA dissertation, University of Macau.

Hamann, Silke & Ilaria Colombo (2017). A formal account of the interaction of orthography and perception: English intervocalic consonants borrowed into Italian. Natural Language & Linguistic Theory 35: 683–714.

Patience, Matthew (2018). Acquisition of the tap-trill contrast by L1 Mandarin–L2 English– L3 Spanish Speakers. Languages 3. doi:10.3390/languages3040042

Zhou, Chao (2017). Contributo para o estudo da aquisição das consoantes líquidas do português europeu por aprendentes chineses. MA dissertation, University of Lisbon.