Comparing foreign accent and Rosengård Swedish: some hypotheses and initial observations

Petra Bodén^*

1. Introduction

1.1 Purpose and outline

Within the research project ‘Language and language use among young people in multilingual urban settings’ (Lindberg 2004) my role is to describe SMG pronunciation, the pronunciation of ‘Swedish on Multilingual Ground’ (e.g. Rosengård Swedish). Thanks to means placed at my disposal by the Faculty of Humanities at Lund University, I am able to broaden my research field to include also a comparison of SMG and foreign accent. The following is mainly based on the research plans written for both lines of research. After a short introduction to the projects and brief descriptions of the materials, I will discuss the problems involved in distinguishing between SMG and foreign accent, and finally, present a few hypotheses and initial observations.

1.2 Language and language use among young people in multilingual urban settings

In Sweden, an increasing number of adolescents speak Swedish in new, foreign-sounding ways. These new ways of speaking Swedish are primarily found in suburbs and urban districts with a high proportion of immigrant residents, e.g. in Rosengård in Malmö. However, many of the speakers are born in Sweden, or have arrived in Sweden as very young, and have acquired Swedish alongside with their mother tongue (at least since kindergarten). According to popular beliefs, some speakers of Rosengård Swedish do not even have an immigrant background. Therefore, the foreign-sounding features of their speech cannot necessarily be classified as transfer or interference from another language. It is furthermore often claimed that speakers of foreign-sounding Swedish master a standard variety of Swedish too. These claims have lead to the hypothesis that these new foreign-sounding ways of speaking Swedish represent new Swedish varieties (dialects, sociolects or group languages) rather than individual speakers’ interlanguages.

Varieties like Rosengård Swedish are primarily a medium for social functions with other group members (Bijvoet forthc). The group identity is marked by signals of a non-Swedish background. Examples of non-Swedish linguistic features, that functions as such signals, are absence of inversion (V2) in cases where the initial element is not the subject, and a pronunciation that is perceived as foreign-accented by native speakers of Swedish.

The project ‘Language and language use among young people in multilingual urban settings’ has as its overarching goal to describe and analyze SMG. My particular role in the project is to describe the phonetics and phonology of SMG.

1.3 Foreign-accented Swedish

In adult language learners’ Swedish, the pronunciation is often what most clearly stands out as not being native-like. Even learners with a large Swedish vocabulary and an advanced grammar, frequently have numerous non-native features in their pronunciation of Swedish. The prosody of the mother tongue has been argued to be particularly difficult to avoid transferring to the target language (Bannert 1979, Gårding 1974).

Whereas a considerable amount of SLA (Second Language Acquisition) research is dedicated to the morphology, syntax, vocabulary and pragmatic aspects of SLA (see e.g. Håkansson 2003), few studies investigate phonetic and phonological SLA. Broad surveys on the literature that nevertheless exist on the topic can be found in Duncan’s doctoral dissertation Phonetic Imitation, Accent, and the Learner (1997) and Major’s Foreign Accent (2001). Major (2001: 17) draws special attention to the “dearth of SLA research in tone and intonation”.

The purpose of my investigation of foreign-accented Swedish is two-fold. Firstly, I would like to describe prosodic SLA, paying special attention to the melodic aspects (i.e. tone/accent and intonation). Secondly, I hope to gain insights into the complex relationship that exists between foreign accent and SMG. When methodologically possible, UG’s (Universal Grammar’s) role in foreign accent and SMG will be considered too.

2. Materials

2.1 Swedish on Multilingual Ground

The Rosengård Swedish material analyzed in this paper comes from the speech database collected by the research project ‘Language and language use among young people in multilingual urban settings’. During the academic year 2002-2003, the project collected a large amount of comparable data in schools in Malmö, Gothenburg and Stockholm. The speakers are young people who attended the second year of the upper secondary school’s educational program in social science during 2002-2003. The recordings are comprised of both scripted speech and spontaneous speech.

2.2 Swedish as a second language

The recordings of second language learners of Swedish are being made during 2004. So far, mainly recordings of international students at Lund University have been made. The international students have different linguistic backgrounds, but have all begun to learn Swedish as adults. Recordings of foreign doctoral students and researchers in Sweden (who have been speaking Swedish for a considerably longer time than the international students) are also planned. The recording sessions are approximately 30 minutes long and consist of two parts: a read part and a spontaneous (unscripted) part.

Some material from the database described above (in section 2.1) is also suitable for investigations of foreign accent.

3. Distinguishing between foreign accent, Rosengård Swedish and standard varieties of Swedish

A comparison of foreign accent and Rosengård Swedish presupposes that a distinction can be made between foreign-accented Swedish and Rosengård Swedish. It also presupposes that a distinction can be made between Rosengård Swedish and other varieties of Swedish that the recorded subjects may speak, e.g. other types of Malmö youth languages. Making these distinctions is not easy, and in some cases, probably not even possible.

There are reasons for thinking about the observed variation as reflecting three distinct ways of speaking Swedish, although a continuum of variation – from foreign-accented Swedish at one end to standard Swedish at the other – probably is a more accurate description. Some of those reasons have already been mentioned, and they will be discussed in more detail below. The popular idea that speakers without an immigrant background can pick up the features of Rosengård Swedish, and learn how to speak it, suggests that Rosengård Swedish is a variety, a way to speak that is restricted by a set of more or less conventionalized rules. Normal word order as an acceptable alternative to V2 word order may be one such rule, and front /r/ sounds as an alternative to the regional, uvular /r/ sounds may be another.

Problems of how to classify a speaker obviously arise when, e.g., the speaker is a second language learner of Swedish and show transfer from her L1, but at the same time appears to have Rosengård Swedish as her target language/variety. She may then also have foreign-sounding features in her speech that cannot be explained as transfer from her own L1.

A perception experiment has been designed to test Malmö teachers’ and pupils’ intuition about what Rosengård Swedish sounds like. The results of the experiment will be reported elsewhere (Hansson & Svensson forthc), as only ten teachers’ test results have been analyzed so far. However, two preliminary findings are relevant to mention here. The first is that Rosengård Swedish is spoken by adolescents without an immigrant background. Even listeners that claimed that Rosengård Swedish is only spoken by people with an immigrant background, judged one of the speakers in the experiment – born in Sweden by Swedish born parents – as a speaker of Rosengård Swedish. The second finding is that at least some speakers of Rosengård Swedish master other varieties of Swedish too. One speaker, occurring in two different stimuli, was judged as a speaker of Rosengård Swedish in one stimulus, and as not being a speaker of Rosengård Swedish in the other. Both these findings lend strong support to the description of Rosengård Swedish as a variety (rather than individual speakers’ interlanguages). It has a set of conventions that are stable enough for someone without a foreign accent to learn, and the foreign-sounding pronunciation can – in contrast to foreign accent – be switched on and off.

4. Hypotheses and initial observations

4.1 Reductions, length and speech rhythm

The perception of speech rhythm is related to repetitive patterns in the spoken language, the recurrence of some kind of speech unit. Pike (1946) suggested that languages have isochronous units of speech, and that the relevant unit is either the syllable (in syllable-timed languages) or the foot or interstress-interval (in stress-timed languages). In other words, the length of each syllable is said to be of more or less equal length in a syllable-timed language, and each interstress-interval of approximately the same length in a stress-timed language. Whereas empirical evidence for isochrony in speech is weak (see Low, Grabe and Nolan (2001) for a survey of the literature), factors such as the presence or absence of vowel reduction, the syllable-structure and word stress have been found to be relevant for the perceived rhythm of a language. In stress-timed languages, e.g., vowel reduction is more frequently found than in syllable-timed languages.

The speech rhythm of the mother tongue is one of the phenomena that can be expected to be transferred to the target language by many language learners. Therefore, it will be investigated in the recorded L2 learners’ Swedish, and it is interesting to also investigate in SMG varieties such as Rosengård Swedish.

In a description of Rinkeby Swedish (a Stockholm variety of SMG), Kotsinas (1990) presents three findings that may be interpreted as evidence of a different rhythm in SMG than in standard Swedish. The first finding is a reported lack of assimilations and reductions in Rinkeby Swedish. Some expected assimilations and reductions across word boundaries are not produced by the SMG speakers; instead of har’nte ‘haven’t’, the speakers say har inte ‘have not’, and so on. The finding is interpreted as evidence of more clearly marked boundaries in SMG, and it is not related to another phenomenon that is also reported, namely the observed hypercorrections (words like ibland ‘sometimes’ are not, as expected, pronounced iblan). However, it may be argued that both findings reflect the same phenomenon: a more infrequent use of reductions and assimilations in Rinkeby Swedish than in standard varieties of Swedish.

Kotsinas’ (1990) third finding has to do with vowel lengths in Rinkeby Swedish. In Swedish, a phonological distinction between long and short vowels is made. Kotsinas observes that her SMG speakers in Rinkeby produce too long short vowels and too short long vowels (too long and too short if compared with the vowels of other speakers in the same region, i.e. Stockholm). However, since Scanian, the dialect spoken in the southernmost part of Sweden, is characterized by the same small difference in duration between phonologically short and long vowels, this feature of Rinkeby Swedish does not necessarily play a role for the variety’s foreign “sound”.

Kotsinas’ findings – the observed lack of assimilations and reductions where such are expected in unstressed syllables, and the shortened long vowels and prolonged short vowels in stressed syllables – are, nevertheless, very interesting. It is possible that they reflect a different speech rhythm in SMG, a tendency towards producing syllables of approximately the same length, i.e. towards syllable-timing. Other stress-timed languages have been reported to exist in syllable-timed variants in areas where the language comes into contact with other languages. Singapore English (Low et al. 2001) and Hawaiian English (Major 2001) are examples of syllable-timed variants of the otherwise stressed-timed English language.

The lack of reductions in unstressed syllables, the perceived assignment of almost equal weight to stressed and unstressed syllables, gives rise to a staccato-like rhythm in SMG (in Kotsinas works referred to as “stötighet”). The staccato-like rhythm is interesting because it has also been observed in the so-called ‘Nuuk Danish’ in Greenland (Jacobsen 2000) and in the so-called ‘multi-ethnolect’ in Copenhagen (Quist 2000). Nuuk Danish is spoken by monolingual Danish-speaking adolescents in Greenland, and the Copenhagen multi-ethnolect by adolescents in Copenhagen with an immigrant background. Interestingly enough, the staccato-like rhythm in the Copenhagen multi-ethnolect is not the result of syllable-timing. At a seminar in 2003, given at the IAAS at Copenhagen University, Quist played a number of audio files illustrating the multi-ethnolect. The seminar participants – linguists and phoneticians – heard no syllable-timing in the recordings. The staccato-like rhythm would instead appear to be related to the many unstressed syllables that were deleted, and the resulting string of closely positioned stressed syllables. Quist’s own notes from the seminar can be found at her home page (Quist 2003).

Whereas measuring the degree of stress- or syllable-timing in SMG is not a trivial matter, attempts will be made to do so in the material described in section 2.2. We will follow Low et al. (2000) and measure the durations of vowels and the duration of intervals between vowels in a read text. A normalised Pairwise Variability Index will then be computed for each type of measurement, and finally be compared with the indexes of typical stress- and syllable-timed languages. The subjects in Low et al. (2001) read the ‘North Wind and the Sun’ which is the English version of the same text that the recorded L2 speakers of Swedish are asked to read (Nordanvinden och solen). Since SMG is not a variety likely to be used in reading, the same method cannot be used for investigating syllable-timing in Rosengård Swedish.

4.2 Word accents, final rises and sentence intonation

Swedish is a language with a lexically and morphologically conditioned distinction of accent type. A word’s primary stressed syllable is associated with an accent – a ‘word accent’ – which is either acute (accent I) or grave (accent II). Phonetically, the difference between accent I and II is one of F0 peak timing. In all dialects of Swedish (except Finland Swedish), the F0 peak of accent I has an earlier alignment with the stressed syllable than accent II (Bruce and Gårding 1978).

It is a well known fact that L2 speakers of Swedish have difficulties in perceiving and producing the word accent distinction (Bannert 1979). Below, a speaker from the database described in section 2.2 (i.e. an L2 speaker of Swedish) has produced the minimal pair anden ‘the duck’ – anden ‘the spirit/genie’. Although the L2 speaker’s accent in ‘spirit/genie’ is fairly close to a southern Swedish pronunciation, see Figure 1, it is most likely simply a consequence of a positive transfer, a “free ride”. The same transfer can be observed in ‘duck’. There it has to be classified as a negative transfer, however. The speaker’s mother tongue is German.

Figure 1 F0 contours of the minimal pair anden ‘the duck’ (top line) – anden ‘the spirit’ (bottom line) as produced by Peter (code name). The F0 contour in accent II words in southern Swedish has three turning points^[1]: a L at the beginning of the stressed syllable, a H at the syllable boundary, and a L at the end of the post-tonic syllable. A final L turning point is not easily observed in Peter’s speech.

Somewhat surprising, the investigations of the Rosengård Swedish material so far reveal that Rosengård Swedish word accents are representative of the Malmö region. In other words, there is a F0 fall in the stressed syllable of accent I words, and a F0 rise in the stressed syllable of accent II words. Thus, the most obvious (but perhaps not the most perceptually salient) way of melodically signalling a non-Swedish background is left unused.

The speech melody of Rosengård Swedish is, nevertheless, perceived as foreign-sounding. The foreign-sounding speech melody appears to be related to another, re-occurring pattern in the material: a suspended downstep or possible upstep, that sometimes ends with a final rise, see Figure 2.

Figure 2 F0 contour of Fy, ja(g) Èdissa(de) Èalla ‘Hey, I dissed everyone’. The words dissade and alla are associated with grave accents/accents II (here transcribed as HL accents^[2]), and the phrase ends with a final rise (H%). The speaker is E04.

The above-mentioned intonation pattern is used by subjects with different linguistic backgrounds (i.e. with different L1s), including by a subject without immigrant background (who nevertheless was judged as speaking Rosengård Swedish in the perception experiment presented above). Some evidence to support an awareness of the intonation pattern as a feature of Rosengård Swedish is given in (1). The recorded subject (D49) describes Rosengård Swedish. She illustrates Rosengård Swedish with an example that contains normal word order (idag vi lovade), and then repeats it with a suspended downstep.

1. Det är ofta så, typ i Rosengård.

Idag vi lovade. Så snackar ni. Idag vi lovade.

Absense of Suspended

Inversion (V2) downstep

‘It’s often so, like in Rosengård. Today we promised. That’s how you talk. Today we promised.’

(Speaker D49)

Whether or not the observed pattern is identified as a feature of Rosengård Swedish by speakers of Rosengård Swedish as well as by speakers of other varieties of Swedish is of great interest to us. In this first stage of the study of SMG pronunciation, potential features of the SMG varieties are sought in the material. We have to make use both of our knowledge of typical regional features (so that they can be singled out) and of our intuitions about SMG in the search, and it is important to look for support for the conclusions drawn in the perception experiment’s results. Stimuli produced by E04 and D49, the speakers in the examples above, were classified as Rosengård Swedish by a majority of the listeners in the perception experiment. E04 appeared in two stimuli, but the listeners only judged the stimulus containing an example of the above-mentioned intonation pattern as Rosengård Swedish.

The phonological analysis of the intonation pattern is not yet complete, and it remains to see if it has an origin in any of the L1s of the immigrants in Rosengård. At several seminars, it has been suggested to me that the speech melody in the examples above is influenced by Arabic. Confirmation of such an Arabic influence was, however, not obtained when the sound examples were played to an Arabic-speaking prosodist (Sam Hellmuth, personal communication)^[3]. An Arabic-influenced speech melody in SMG has some support in popular beliefs (Borda-Pedreira 2003).

We do not yet know if the observed intonation pattern has a counterpart in the materials recorded in Gothenburg and Stockholm. Kotsinas (1990) has described a particular final rise (stigande slutton) as a typical feature of Rinkeby Swedish. Although the pattern described above is not a phenomenon as local as a final H, it is possible that the described patterns are one and the same.

5. Future work

Rosengård Swedish and other similar varieties have an obvious relation to foreign-accented Swedish. The comparison of SMG and foreign-accented Swedish is expected to prove fruitful for both lines of research. The description of SMG cannot be completed without its influences from other languages becoming known, i.e. without further studies of transfer in SLA. Likewise, the investigations of SMG may shed light on the role of universal grammar in SLA, thereby improving our understanding of foreign accent.

Acknowledgements

This research has been supported by means from the ‘Faculty of Humanities’ research funds 2003’ (Lund University) and the research project ‘Language and language use among young people in multilingual urban settings’ (financed by the Bank of Sweden Tercentenary Foundation, Grant No. 2000-5124:01).

References

Bannert, Robert. 1979. Ordprosodi i invandrarundervisningen. Praktisk lingvistik 3. Lund: Department of Linguistics and Phonetics, Lund University.

Bijvoet, Ellen. forthc. ‘Attitudes towards “Rinkeby Swedish”, a group variety among adolescents in multilingual suburbs’. In: Kari Fraurud & Kenneth Hyltenstam (eds), Multilingualism in Global and Local Perspectives. Papers from the 8th Nordic Conference on Bilingualism. November 1-3, 2001, Stockholm Rinkeby. Stockholm: Rinkeby Institute of Multilingual Research.

Borda-Pedreira, Joakim. 2003. ‘Invandrarböcker en korrekt kassako’. Lundagård 7, 18.

Bruce, Gösta & Eva Gårding. 1978. ‘A Prosodic Typology for Swedish Dialects’. In: Eva Gårding, Gösta Bruce & Robert Bannert (eds), Nordic Prosody: Papers from a symposium, 219-228. Malmö: Department of Linguistics and Phonetics, Lund University.

Gårding, Eva. 1974. ‘Den efterhängsna prosodin’. In: Ulf Teleman & Tor G. Hultman (eds), Språket i bruk, 50-71. Lund: LiberLäromedel.

Hansson, Petra. 2003. Prosodic Phrasing in Spontaneous Swedish. Travaux de l’institut de linguistique de Lund 43. Doctoral dissertation. Lund: Department of Linguistics and Phonetics, Lund University.

Hansson, Petra & Gudrun Svensson. forthc. ‘Listening for Rosengård Swedish’. In: Proceedings FONETIK 2004, The Swedish Phonetics Conference, May 26-28 2004. Stockholm: Department of Linguistics, Stockholm University.

Håkansson, Gisela. 2003. Svenska i och utanför Sverige. Department of Lingusitics and Phonetics, Lund University. http://www.ling.lu.se/ projects/Svenska.html (accessed 02-03-2004).

Jacobsen, Birgitte. 2000. ‘Sprog i kontakt. Er der opstået en ny dansk dialekt i Grønland? En pilotundersøgelse’. Grønlandsk kultur- og samfundsforskning 98/99. Ilisimatusarfik/Atuakkiorfik 2000, 37-50.

Kotsinas, Ulla-Britt. 1990. ‘Svensk, invandrarsvensk eller invandrare? Om bedömning av ”främmande” drag i ”ungdomsspråk”’. In: Gunnar Tingbjörn, (ed), Andra symposiet om svenska som andraspråk i Göteborg 1989, 244-274. Gothenburg: Gothenburg University.

Low, Ee Ling, Ester Grabe & Francis Nolan. 2001. ‘Quantitative characterisations of speech rhythm: Syllable-timing in Singapore English’. Language and Speech 43 (4), 377-401.

Lindberg, Inger. 2004. Language and language use among young people in multilingual urban settings 2002 - 2006. Institutet för svenska som andraspråk, Gothenburg University. http://svenska.gu.se/isa/ (accessed 24-02-2004).

Major, Roy C. 2001. Foreign Accent: The Ontogeny and Phylogeny of Second Language Phonology. Mahwah, New Jersey & London: Lawrence Erlbaum Associates Publishers.

Markham, Duncan. 1997. Phonetic Imitation, Accent and the Learner. Travaux de l’institut de linguistique de Lund 33. Doctoral dissertation. Lund: Lund University Press.

Pike, Kenneth. 1945. The Intonation of American English. Ann Arbor: University of Michigan Press.

Quist, Pia. 2000. ‘Ny københavnsk ‘multietnolekt’. Om sprogbrug blandt unge i sprogligt og kulturelt heterogene miljøer’. Danske Talesprog. Copenhagen: C.A. Reitzels Forlag, 143-212.

Quist, Pia. 2003. Multietnolekt og prosodi. Nordisk Forskningsinstitut Copenhagen University. http://www.piaquist.dk/aktuelt.htm#prosodi (accessed 03-03-2004).

---

^* Affiliations: Department of Linguistics and Phonetics and Department of Scandinavian Languages, Lund University.

^[1] Note that a turning point is not necessarily a reflex of a tone, but simply a phonetic turning point accessible to empirical observation in the F0 contour.

^[2] The phonological analysis that motivates this way of transcribing southern Swedish grave accents is discussed in Hansson (2003), pp 94-97.

^[3] Her expertise obviously does not include all Arabic dialects, however.