Proceedings Fonetik 2009 - Institutionen för lingvistik

Proceedings, FONETIK 2009, Dept. of Linguistics, Stockholm Universityocclusion duration and quantity as independentvariables, 96.5% of the tokens were correctlyclassified (99.4% of /p t k/ and 93.1% of /b dg/). For group A, with the same independentvariables, all stops were correctly classified(100.0%). For both groups A and B, the orderof strength of the independent variables asseparators of the two categories was: voicingduration > closure duration > quantity.Phonological conclusionsFenno-Swedish contrasts voiced /b d g/ withvoiceless unaspirated /p t k/. On the basis of ouracoustic measurements and some knowledge ofhow these are related to glottal and supraglottalevents, we conclude that the contrast in Fenno-Swedish is one of [voice] vs no laryngeal specification.Note that our definition of voicedstops refers, concretely, to the presence of considerableprevoicing in utterance-initial stops,and to extensive voicing during the occlusion inother positions with, at most, a short positiveVOT after occlusion offset. What we refer to asvoiceless stops, in turn refers, again concretely,to short positive VOT in utterance-initial stops,to voiceless occlusion in medial stops and finalstops (allowing for a very short period of voicingat the beginning of the occlusion, if the precedingsegment is voiced), with at most a veryshort positive VOT.Suomi (1980: 165) concluded for the Finnishvoiceless unaspirated /p t k/ that their “degreeof voicing [is] completely determined bythe supraglottal constrictory articulation”.These stops have no glottal abduction or pharyngealexpansion gesture, a circumstance thatleads to voicelessness of the occlusion (p.155ff). Despite the different terminology, thisamounts to concluding that the Finnish /p t k/have no laryngeal specification. Thus in the twostudies, thirty years apart, essentially the sameconclusions were reached concerning Fenno-Swedish and the Finnish /p t k/.Stop clustersFour cluster types investigated: (1) /kt/, /pt/ (asin läkt, köpt), (2) /kd/, /pd/ (as in väckte, köptewhich, on a generative analysis, are derivedfrom vä/k+d/e and kö/p+d/e), (3) /gt/ (as invägt, byggt) and (4) /gd/ (as in vägde). Clusters(1) – (2) were always almost completely voiceless,and consequently there is no phonetic evidencethat the two types are distinct. The cluster/gd/ was usually nearly fully voicedthroughout. But in the realisation of the /gt/cluster there was again much variation amongthe speakers. The /t/ was always voiceless, butthe /g/ ranged from fully voiceless (in 43% ofthe tokens) to fully voiced (33%), and only thebeginning of /g/ was voiced in the remaining24% of the tokens. For two speakers all six tokensof /g/ were fully voiceless, for threespeakers four tokens were fully voiceless, andfor five speakers, on the other hand, four ormore tokens were fully voiced. As an exampleof intra-speaker variation, one speaker producedtwo fully voiced, two partially voicedand two fully voiceless /g/ tokens. On thewhole, the speakers used the whole continuum,but favoured the extreme ends: /g/ was usuallyeither fully voiced or fully voiceless, the intermediatedegrees of voicing were less common.This dominant bipartite distribution of tokensalong a phonetic continuum is very differentfrom the more or less Gaussian distribution oneusually finds in corresponding studies of a singlephonological category.Contact influence?Historically, Finnish lacks a laryngeal contrastin the stop system, the basic stops being /p t k/,which are voiceless unaspirated. In the past, allborrowed words were adapted to this pattern,e.g. parkki ‘bark’ (< Swedish bark), tilli ‘dill’(< Sw. dill), katu ‘street’ (< Sw. gata). StandardSpoken Finnish (SSF) also has a type of /d/which is usually fully voiced. However, this /d/is not a plosive proper, but something betweena plosive and a flap, and is called a semiplosiveby Suomi, Toivanen and Ylitalo (2008). Itsplace is apical alveolar, and the duration of itsocclusion is very short, about half of that of /t/,ceteris paribus (Lehtonen 1970: 71; Suomi1980: 103). During the occlusion, the locationof the apical contact with the alveoli alsomoves forward when the preceding vowel is afront vowel and the following vowel is a backvowel (Suomi 1998).What is now /d/ in the native vocabulary,was a few centuries ago /ð/ for all speakers.When Finnish was first written down, themostly Swedish-speaking clerks symbolised /ð/variably, e.g. with the grapheme sequence. When the texts were read aloud, againusually by educated people whose nativetongue was Swedish, was pronounced asit would be pronounced in Swedish. At thesame time, /ð/ was vanishing from the vernacular,and it was either replaced by other conso-62