Proceedings Fonetik 2009 - Institutionen för lingvistik

Proceedings, FONETIK 2009, Dept. of Linguistics, Stockholm Universityence of the CS Swedish consonants on theseconsonants in Fenno-Swedish. Thus, to the extentthat Fenno-Swedish differs from CS Swedish(or, more generally, any varieties of Swedishspoken in Sweden), a very likely cause ofthe difference is influence from Finnish. In additionto the influence of the Finnish /p t k/ onthe Fenno-Swedish /p t k/, we suggest that anyvariation towards voiceless productions of /b dg/ is also due to Finnish influence.Our results for utterance-initial stops in alanguage in which /b d g/ stops are predominantlyprevoiced are not altogether unprecedented.They resemble the results of Caramazzaand Yeni-Komshian (1974) on CanadianFrench and those of van Alphen and Smits(2004) for Dutch. Caramazza and Yeni-Komshian observed substantial overlap betweenthe VOT distributions of /b d g/ and /p tk/: a large proportion (58%) of the /b d g/ tokenswere produced without prevoicing, while/p t k/ were all produced without aspiration.The authors argued that the Canadian FrenchVOT values are shifting as a result of the influenceof Canadian English. van Alphen andSmits observed that, overall, 25% of the Dutch/b d g/ were produced without prevoicing bytheir 10 speakers, and, as in the present study,there was variation among the speakers: five ofthe speakers prevoiced very consistently, withmore than 90% of their /b d g/ tokens beingprevoiced, while for the other five speakersthere was less prevoicing and considerable inter-speakervariation; one speaker producedonly 38% of /b d g/ with prevoicing. van Alphen& Smits’ list of target words containedwords with initial lenis stops before consonants,which ours did not. They found that theamount of prevoicing was lower when the stopswere followed by a consonant. If we comparethe results for the prevocalic lenis stops in thetwo studies, the results are almost identical(86% prevoicing for van Alphen & Smits, 87%for our speakers). The authors are puzzled bythe question: given the importance of prevoicingas the most reliable cue to the voicing distinctionin Dutch initial plosives, why dospeakers not produce prevoicing more reliably?As a possible explanation to this seeminglyparadoxical situation, van Alphen and Smitssuggest that Dutch is undergoing a soundchange that may be caused or strengthened bythe large influence from English through theeducational system and the media. It may be,however, that van Alphen and Smits’ speakershave also been in contact with speakers of dialectsof Dutch with no prevoicing (and aspiratedstops) or by contact with speakers of German.There is evidence that speakers are verysensitive to the VOTs they are exposed to.Nielsen (2006, 2007) has shown that speakersof American English produced significantlylonger VOTs in /p/ after they were asked toimitate speech with artificially lengthenedVOTs in /p/, and the increased aspiration wasgeneralised to new instances of /p/ (in newwords) and to the new segment /k/. There isalso evidence that native speakers of a languageshift VOTs in their native language as a resultof VOTs in the language spoken around them(Caramazza and Yeni-Komshian, 1974; vanAlphen and Smits, 2004). Sancier and Fowler(1997) show that the positive VOTs in thespeech of a native Brazilian Portuguese speakerwere longer after an extended stay in the UnitedStates and shorter again after an extended stayin Brazil. The authors conclude that the Englishlong-lag /p t k/ influenced the amount of positiveVOT in the speaker’s native BrazilianPortuguese.All of our Fenno-Swedish speakers, like themajority of Fenno-Swedish speakers, are fluentin Finnish (as was observed before and after therecordings). Fenno-Swedish is a minority languagein Finland, and hence for most speakersit is very difficult not to hear and speak Finnish.Consequently, most speakers of Fenno-Swedishare in contact, on a daily basis, with a languagein which there is no aspiration and in whichprevoicing is not often heard. On the basis ofinformation available to us on our speakers’place of birth, age and sex, it is not possible todetect any systematic pattern in the variation inthe degree of voicing in /b d g/ as a function ofthese variables.In brief, the situation in Fenno-Swedishmay be parallel, mutatis mutandis, to that observedin Canadian French and Dutch. Assumingthat prevoicing in Fenno-Swedish /b d g/has been more systematic in the past than it isamong our speakers (which can hardly be verifiedexperimentally), influence from Finnish isan explanation for the variability of prevoicingin Fenno-Swedish that cannot be ruled outeasily. Without such influence it is difficult tosee why speakers would choose to collapsephonemic categories in their native language. 2AcknowledgementsWe are grateful to Mikko Kuronen and Kari64