An Acoustic Study of the Estonian Swedish Lateral

Proceedings from FONETIK 2014, Department of Linguistics, Stockholm University

An acoustic study of the Estonian Swedish lateral [ɬ] Susanne Schötz1, Francis Nolan2, Eva Liina Asu3 1Centre for Languages and Literature, Lund University, Sweden 2 Department of Theoretical and Applied Linguistics, University of Cambridge, UK 3 Institute of Estonian and General Linguistics, University of Tartu, Estonia [email protected], [email protected], [email protected]

Abstract Schötz and Kügler (2009), and a study This pilot study investigates the Estoni- of the prosody of compounds by Schötz an Swedish (ES) voiceless lateral [ɬ], and Asu (2013). Yet, the ES sound sys- which is rare among the Scandinavian tem offers plenty of interesting and dialects and is a development mainly of unique material for acoustic phonetic historic /sl/ clusters. Six elderly ES study. One of such is the lateral system speakers were recorded, and the phono- that will be tackled here. logical and phonetic (duration, relative According to H. Lagman (1971: intensity) properties of [ɬ] studied and 175–188), ES has a fairly large number compared to other ES consonants and of laterals consisting of the following: [ɬ] in Icelandic. The results suggest that (1) [l]: a voiced dental or alveolar ES [ɬ] is a single consonant rather than common to all Scandinavian lan- a consonant cluster. It behaves much guages, in most contexts, e.g. lag like initial [s] in duration, although a (team) and vall (ley, mound), tendency to anticipatory voicing in its (2) [ɭ]: a voiced supradental (post al- latter part may point to its ‘approxi- veolar) in the context /r/ + /l/, like mant’ status. Furthermore, ES [ɬ] is many Swedish and Norwegian dia- similar in intensity to the Icelandic [ɬ]. lects (except most with [ʁ]; and It has a phonemic status and it can be Finland Swedish), e.g. farligheter both short and long. Laterals as well as [fɑːɭɪhaɪtɵ(ɹ)] (hazards), other phonetic aspects of ES are in ur- (3) [ɽ]: a voiced retroflex flap in some gent need of further research. contexts, as in some Swedish and Norwegian dialects, e.g. blå [bɽoː] Introduction (blue) and nagel [nɑːɽ] (nail), Until the Soviet Russian occupation at (4) [ɬ]: a voiceless alveolar fricative, the end of WWII there was a substantial used sometimes in the (historic) Swedish speaking population in Estonia contexts /sl-/ and /tl-/, which is which had settled there in medieval more rare among the Scandinavian times from the 13th century onwards. dialects, e.g. slag [ɬoː] (stroke, Swedish was concentrated mainly on blow) and vassle [vaɬː] (whey). islands off the west coast of Estonia and The latter two constitute major devel- on the north-west corner of the Estonian opments in the lateral system, one mainland. Nowadays, Estonian Swedish common to many dialects of Swedish (ES) survives only in a small communi- and Norwegian, and the other idiosyn- ty of elderly emigrants to Sweden and a cratic, or at least very rare. The retro- tiny handful of equally elderly speakers flex flap [ɽ], often referred to as ‘cacu- in Estonia. minal’ or ‘thick’ /l/, is found in initial The sound system of ES has so far clusters (e.g. flytta [fɽɛte] or [fɽɪtː] ‘to mainly been studied in the descriptive move’, and blöt [bɽɑʊt] ‘wet’), medially framework of dialect research (e.g. E. in some words (e.g. mala [mɔɽɐ] ‘to Lagman, 1979). The only two existing grind’), and finally after certain conso- acoustic phonetic studies are an investi- nants (e.g. fågel [føːɽ] ‘bird’, and gation of ES close vowels by Asu, körhjul [keːɾjøːɽ] ‘bicycle’). Swedish

23 Proceedings from FONETIK 2014, Department of Linguistics, Stockholm University and Norwegian [ɽ] is thought to have Ormsö ES were recorded in Stockholm first arisen as a realisation of [rð] se- in 2012. All speakers have been resi- quences in Old Norse in words such as dent in Stockholm since the 1940s. The garðr ‘court(yard)’ and borð ‘board’, main speech material was compiled and to have spread to /l/ in certain con- with the help of the members of the texts. Tiberg (1962), though, claims that Estonian Swedish community Svenska [ɽ] in Estonian Swedish did not origi- Odlingens Vänner (SOV) in Stockholm nate from [rð] sequences. and contained 90 words, most of which The second development, the lat- included liquid consonants in various eral voiceless fricative [ɬ] from historic positions. The words were read in the sequences of /s/+/l/, occurs in words carrier sentence ‘jag sa ___ igen’ (I said such as slipp [ɬɪpː] ‘cloth/rag’ and ___ again). The speakers were asked to vassle [vaɬː] ‘whey’. This sound is used produce three repetitions of each sen- consistently across our speakers, though tence. In this study we focussed on five with realisational variation. It is very words with initial [ɬ(l)]: slag, slita, slipp rare in Swedish varieties, being (to our (=‘rag’), slå, and slak, eleven words knowledge) found elsewhere only in the with initial [f]: fågel, falla, fel, föl, föll, Älvdalska and Orsamål dialects of Da- fall, fil, fall, fylla, fåll, fittla (=‘tickle’), larna in the west of Sweden. It may also and six words with initial [fɽ]: flytta, have been present in the Gam- flaska, flat, fläta, flämta, flicka. malsvenskby dialect found in Ukraine before coming under pressure from more standard forms. The purpose of the present paper is to quantify the phonetic properties of ES [ɬ]. Specifically, we wanted to an- swer the following questions: (a) Is [ɬ] still a complex onset, i.e. a nostalgic [sl–] consonant cluster? An initial audi- tory analysis revealed that many tokens, especially in Rickul ES, have [ɬl], and some tokens showed evidence of a voiceless vocalic portion before the lateral friction, i.e. [hɬl]. Does this suggest a phonological sequence, assimi- lated for manner but not voicing? (b) if [ɬ(l)] is not a cluster, has the original [sl-] become a lateralised [s], or has it become a devoiced [l]?, and (c) Is it an Figure 1. The main dialect areas of Estonian Swedish in the 1930s (from E. Lagman approximant [l̥ ] or a fricative [ɬ]? An ̥ 1979: 2). additional aim ̥was to hypothesise the historical origin̥ of [ɬ]. As additional materials an elicited word list adapted from the word list used in Material and method the Swedish dialect project SweDia Speakers for the present study represent 2000 (Bruce et al., 1999) was used. the Nuckö-Rickul-Ormsö variety of These materials were recorded in Estonian Swedish which forms the Stockholm in 2009 with four Rickul largest dialect area of Estonian Swe- speakers (of whom three were the same dish. As can be seen from Figure 1 as the ones recorded in 2012). From this Nuckö and Rickul lie in the north-west corpus, we selected the following of the Estonian mainland and Ormsö is words: [l]: läs, lus, lös, lat, lott, lass, a small island close to Nuckö. Three ludd, lett, and [s]: särk, såll, saker, sur elderly speakers of each of Rickul and ([søːɹ], used instead of söt for /øː/).

24 Proceedings from FONETIK 2014, Department of Linguistics, Stockholm University

The target words were manually voiceless [ɬ], while one speaker has segmented in Praat (Boersma & Ween- shorter, and less frequent voicing at the ink, 2014), and durations obtained of end of the lateral. There is some ten- the initial singleton consonant or con- dency towards an [h] onset to [ɬ] in both sonant cluster. In the case of [ɬ] we also dialects, which may be interpreted as measured the duration of its potential pre-aspiration or anticipatory devoicing components: [(h) ɬ (l)]. Figure 2 shows of the vowel. an example of an [ɬl] by one Rickul speaker. Phonological and phonetic status of [ɬ] Is [ɬ] a singleton or consonant cluster? Several of the [ɬ] tokens are realised as [(h)ɬl], which is unlike [ɬ] in other languages, like e.g. Zulu and Welsh (Ladefoged & Maddieson 1996; Mark Jones, pers. communication). The duration is also rather long for a singleton (150-250 ms, see Figure 3). This would suggest that [ɬl] is phonologically a consonant cluster. However, two of the Figure 2. Waveform and spectrogram of [ɬl] Ormsö speakers lack voicing, and only for one Rickul speaker. one Rickul speaker has no fully voiceless tokens, which would indicate that Results [ɬl] is phonologically a singleton. Figure Phonetic properties of [ɬ] 4 shows mean durations of the consonant cluster [fɽ], the two singleton con- Mean durations of [(h) ɬ (l)] for Rickul sonants [s] and [f], and the possible and Ormsö ES are shown in Figure 3. consonant cluster [(h)ɬl].

Figure 3. Mean duration of the components of [(h) ɬ (l)] for three Rickul (top) and three Figure 4. Mean duration of initial [ɬ(l)], [fɽ], Ormsö (bottom) speakers. [s] and [f] for three Rickul (top) and three Ormsö (bottom) speakers. The Rickul speakers on average produce a largely voiceless lateral, which The results for [s] should, however, be is often voiced towards the end. Two of seen as tentative since they are based on the Ormsö speakers produce completely only three tokens of each speaker, and

25 Proceedings from FONETIK 2014, Department of Linguistics, Stockholm University as two of the Ormsö speakers did not voicing is characteristic of [l̥ ] (and produce any words with initial [s]. In some voiceless nasals) rather than [ɬ]. both Rickul and Ormsö ES, [s] can be On this basis the ES voiceless lat- as long as [ɬ(l)], suggesting that [ɬ(l)] eral shows signs of anticipatory voic- need not be a cluster. However, dura- ing, i.e. of being an approximant, at tional compensation means that clusters least in Rickul ES. However, impres- need not be longer than singletons, so sionistically the strength of the friction the question is still open, although the suggests ‘fricative’. evidence tends against the consonant We compared intensity ratios be- cluster structure of /sl-/. tween the lateral and the following vowel in ES and Icelandic, which has Is [ɬ(l)] as a lateral [s] or voiceless [l̥ ]? an approximant /l̥ /, often realised as [ɬ], If [ɬ(l)] is not a nostalgic reflex of the as well as a fricative [s]. Figure 6 shows [sl-] consonant cluster but a singleton, that the mean intensity of the following should we regard it as a lateral [s], or vowel relative to [ɬ] is very similar in rather as a voiceless [l]? We compared Ormsö ES and Icelandic, suggesting the duration of [ɬ(l)] to that of [l], and that [ɬ] may have approximant status. [s] in initial positions of the words taken from the word list materials read by the same speakers of Rickul ES, taken from the word list material. The duration for four Rickul speakers are shown in Figure 5.

Figure 6. Mean intensity (dB) of the following vowel relative to [ɬ] for three Ormsö speakers and one speaker of Icelandic. Is [ɬ(l)] a phoneme? Figure 7 shows the waveform and spectrogram of an example minimal pair in Figure 5. Mean duration of initial [ɬ(l)], [l] ES: /valː/ – /vaɬː/. and [s] for four Rickul speakers. The duration of [ɬ(l)] is clearly more similar to the duration of [s] than that of [l], indicating that durationally, [ɬ(l)] may be some kind of lateralised [s]. Is it a [ɬ] or a [l̥ ]? The difference between [ɬ] and [l̥ ] is arguably that devoicing an approximant (e.g. [l̥ ]) needs only enough glottal opening to stop voicing, and gives minimal local friction, whereas a voiceless fricative (e.g. [ɬ]) needs wide opening, and narrower oral constriction, to produce significant local friction. Figure 7. Waveform and spectrogram of the Ladefoged and Maddieson SotWL Estonian Swedish minimal pair /valː/ – /vaɬː/. (1996: 199) imply that anticipatory

26 Proceedings from FONETIK 2014, Department of Linguistics, Stockholm University

The existence of minimal pairs in our consonant rather than a consonant clus- material, e.g. /lɑː/ ‘lag’ (team) – /ɬɑː/ ter [sl-], which behaves like an [s] in ‘slå’ (hit, blow) and /valː/ ‘vall’ duration, but like a [l̥ ] in intensity, sug- (mound) – /vaɬː/ ‘vassle’ (whey) suggesting that it has approximant status. gests that [ɬ] has a phonemic status. However, three speakers of each sub- Moreover, it can be short [ɬ] as in e.g. variety are too few for any broader gen- [ɬɑː] as well as long [ɬː] as in e.g. [vaɬː]. eralisations. Remaining questions may be answered by including a larger num- Historical origin of [ɬ] ber of speakers and other sub-varieties ES has undoubtedly historically been of ES. In the project Estonian Swedish influenced by its closest contact lan- Language Structure we intend to record guage Estonian (e.g. Danell, 1905: 34, additional speakers of different ES dia- ref. in H. Lagman, 1971: 13). As Esto- lects and also use previously recorded nian prefers simple onsets, it is tempt- materials to further investigate the pho- ing to hypothesize that somehow the netic and phonological properties of ES. emergence of [ɬ] for [sl-] might have to do with Estonian influence. On the oth- Conclusions er hand, the sound is not present in the Estonian Swedish has a rich and unusu- Estonian phonemic inventory, and the al set of liquids, as shown in Figure 8. hypothesis is less persuasive than it might be given that we are not aware of any other processes of cluster simplifi- cation in ES like those evident in bor- rowings into Estonian from Germanic languages. Figure 8. Estonian Swedish liquids. Also for this solution to generalize, dialects like Älvsdalska, – one of the The most unusual liquid among Scan- few other Swedish dialects with [ɬ] – dinavian languages is the voiceless lat- would have had to be in contact with eral [ɬ]. Although some questions re- e.g. Sami, which is not self-evidently main, in ES [ɬ] appears to be likely in the relevant timescale. a) a singleton consonant rather than a It is therefore more likely that the nostalgic reflex of a cluster, seeds of [ɬ] may be inherent in variation b) a voiceless fricative in duration, in [sl-] more generally. To test this, a behaving much like initial [s], alt- search was made of the DyViS database hough the partial final voicing may (see Nolan et al., 2009). Perhaps sur- point to ‘approximant’ status, simi- prisingly, in approximately 25 hours of lar to Icelandic voiceless laterals, interviews the interviewees provided c) optionally anticipatorily voiced, only 16 examples of word-initial /sl-/. d) a phoneme, as it appears both short Of these, all had identifiable voiceless and long in minimal pairs like /lɑː/ lateral events following the [s], and – /ɬɑː/ and /valː/ – /vaɬː/. nearly half (7/16) lacked a subsequent Acknowledgements identifiable voiced lateral episode. This suggests that a voiceless lateral is latent The authors are very grateful to Júlía in any [sl-] sequence. Baldursdóttir for her help with data analysis and for recording the Icelandic Discussion and future work words. We would also like to thank our The results of this pilot study suggest Estonian Swedish informants and Sven- that [ɬ] is a phoneme in ES, which can ska Odlingens Vänner in Stockholm. be realised as [ɬ], [ɬl], or [hɬl] in the The work on this paper was supported Rickul and Ormsö sub-varieties. More- by the project Estonian Swedish Lan- over, the duration and intensity meas- guage Structure (ESST) (Swedish Re- urements suggest that [ɬ] is a singleton search Council, grant no. 2012-907).

27 Proceedings from FONETIK 2014, Department of Linguistics, Stockholm University

References Kulturföreningen Svenska Asu, E.L., Schötz, S., & Kügler, F. Odlingens Vänner. (2009). The acoustics of Estonian Lagman H. (1971). Svensk-estnisk Swedish long close vowels as com- språkkontakt. Studier över estnis- pared to Central Swedish and Fin- kans inflytande på de estlandssven- land Swedish. Proceedings of ska dialekterna. Stockholm. Fonetik 2009, Dept. of Linguistics, Nolan, F., McDougall, K., de Jong, G. Stockholm University. & Hudson, T. (2009). The DyViS Boersma, P. & Weenink, D. (2014). database: style-controlled record- Praat: doing phonetics by computer ings of 100 homogeneous speakers [Computer program] Ver. 5.3.76, for forensic phonetic research. In- retrieved from www.praat.org/. ternational Journal of Speech, Bruce G., Elert C.-C., Engstrand O., & Language and the Law 16(1), 31- Eriksson A. (1999). Phonetics and 57. phonology of the Swedish dialects Schötz, S. & Asu, E.L. (2013). An – a project presentation and a data- acoustic study of accentuation in base demonstrator. Proceedings of Estonian Swedish compounds. In ICPhS 99 (San Francisco), 321– Asu and Lippus (Eds.), Nordic 324. Prosody: Proceedings of the XIth Ladefoged P. & Maddieson I. (1996). Conference, Tartu 2012. 343–352. The Sounds of the World’s Lan- Peter Lang, Frankfurt am Main. guages. Oxford: Blackwell. Tiberg N. (1962). Estlandssvenska Lagman E. (1979). En bok om Estlands språkdrag. Lund: Carl Bloms Bok- svenskar. Estlandssvenskarnas tryckeri A.-B. språkförhållanden. 3A. Stockholm: