Pietilä, P. & O-P. Salo (eds.) 1999. Multiple Languages – Multiple Perspectives. AFinLA Yearbook 1999.
Publications de l’Association Finlandaise de Linguistique Appliquée 57. pp. 206–210.
THE CATEGORISATION OF /u/, /w/ AND /v/ BY NATIVE SPEAKERS OF FINNISH AND ENGLISH
Maija Peltola, University of Turku
The categorisation of the sounds /u/, /w/ and /v/ by native speakers of Finnish and English was studied with the help of synthesised acoustic stimuli. The aim was to investigate the role that the mother tongue and the proficiency in the target language have on the identification of sounds. The results showed that monolingual Finns categorised the acoustic continuum in accordance with the Finnish phonology, whereas the answers given by advanced students of English revealed that the categorisation was affected by the imposing new system. In fact, students of English identified more sounds as the approximant /w/ in comparison with the group of native speakers of English. This tendency towards hypercorrection was evident in the results. In addition, the fact that the English group categorised a frictionless sound as /v/ shows that the English category /v/
has an allophone which does not contain friction.
Keywords: approximants, categorisation, phonology, acoustic stimulus
1 INTRODUCTION
The phonological structure of the native language functions as the basis for the categorisation of acoustic stimuli into various groups of phonemes (Lado 1957, Wiik 1965, Best & Strange 1992, Kuhl 1993). In a situation of a language contact, viz. during a process of second language learning, the sounds of the target language are categorised according to the model provided by the mother tongue. However, this template does not automatically coincide with the system of the target language and, as a consequence, learning difficulties occur.
In the categorisation of the acoustic continuum ranging from the vowel /u/ through /w/ to /v/, native speakers of Finnish and English react differently to synthesised acoustic stimuli created in order to test the identification of the sounds in a word initial position before a close front
1For more detailed information concerning the acoustic characteristics of the stimuli, see the discussion in Pienimäki 1998.
2 This method is widely utilised; see e.g. Aaltonen 1985. It should be noticed that, since these two experiments were part of a larger series of experiments, the testees also had the alternatives /f/, /l/ and /r/ in the sheets.
vowel /i/. The vowel sound began at 200 ms and lasted until 490 ms, which was, therefore, the overall duration of the stimuli. The first 200 ms were devoted to the actual changes of interest in the continuum /u/ – /w/ – /v/: in Experiment 1, the duration of the voiced pulse was systematically varied and in Experiment 2 the duration of the pre-transitional steady state phase of the formants was studied. In both cases, the same durations were experimented with both low (F1: 600 Hz and F2: 1600 Hz) and high (F1:
1600 Hz and F2: 2400 Hz) formant settings1. The objective of the experiments was to study the effect of both the mother tongue and the proficiency in the target language in the identification process. In the following discussion the aim is to briefly report some of the major findings of the experiment. Firstly, the group consisting of monoligual Finns will be investigated; secondly, the results of advanced students will be reported, and lastly, the native speakers of English will be in the focus of attention.
The members of the test groups were asked to circle the correct alternative from the answer sheet after having heard the stimulus twice, i.e. the experiment was conducted with the forced choice method2.
2 NATIVE LANGUAGE FINNISH I
This group consisted of 34 young native speakers of Finnish (fifth grade).
The members of the group were expected to be acquainted with some basic ideas of the English sound system, e.g. they were familiar with the phoneme/grapheme /w/, but the hypothesis was that the primary system should still be Finnish with its two categories, namely /u/ and /v/. The main result of Experiment 1 was that the testees were sensitive to the frequency of the formant loci: when the second and the third formant were given low values the stimuli were categorised as either /w/ (when the duration of voicing was short), or as the vowel /u/ (when the duration was either long or intermediate). With the high formant settings, however, the answer type containing the initial /v/ was favoured with each variant.
In Experiment 2, the frequencies of the second and the third formants were also of significance. With the low settings, the alternative containing the approximant /w/ was considered to be the correct alternative both with the long and the intermediate duration of transitions. In contrast, when the steady state phase was maximally long, the vowel /u/ was favoured by the majority of the testees. With the high formant values, on the other hand, the most popular answer was /v/: the answers were most unanimous with long transitions and the amount of disagreement augmented as the steady state became longer.
3 NATIVE LANGUAGE FINNISH II
This group consisted of 37 first-year students of English studying at the Department of English, University of Turku. The subjects were supposed to have a good knowledge of the English system both in regard to spelling conventions and phonology. The most interesting result was that – with the low formant values – the alternative /w/ was always the most attractive answer type. The highest level of unanimity was reached with the shortest voicing. In addition, the vowel /u/ was considered to be a possibility in each stimulus type. In the case of the high formant loci the testees categorised the stimuli as /v/ with a consi-derable degree of agreement; the stimulus with the long voicing was considered to be the best exemplar of the category /v/.
In Experiment 2, the testees favoured the alternative /w/ in all cases when the formant transitions began from low frequencies: the highest agreement was in conjunction with the short transitions. The vowel /u/ was also considered to be a possible category. With the high resonance values, the results showed that the approximant /v/ was the most attractive answer with an extremely high level of agreement. In fact, the stimulus containing the long transitions and the short steady state was identified as /v/ by all subjects.
4 NATIVE LANGUAGE ENGLISH
The group consisting of 15 native speakers of English gave the most homogeneus answers. In Experiment 1, all informants favoured the approximant /w/ with low values for the formants and the short and intermediate durations of the voiced pulse; with the long voicing this
alternative was favoured by the great majority with the vowel /u/ receiving only 20 % of the answers. With the high formant values all testees gave the answer type containg the initial approximant /v/ in each case.
In Experiment 2, the testees favoured the phoneme /w/ with all durations of the transitions when the formants were set to be low.
Interestingly, some testees even chose the syllable containing the initial /u/.
However, with the high formant loci the most popular alternative was /v/; in fact, this phoneme was chosen by all testees when the transitions were either intermediate or long.
5 CONCLUSIONS
The results of the experiments reported above showed that both the native language and the proficiency in a foreign language have an effect on the categorisation/identification of linguistic stimuli. The group of nearly monolingual Finns considered the alternative with the initial vowel /u/ to be a possible choice in many of the stimuli in Experiment 1; advanced learners of English, on the other hand, did not favour this alternative. This shows that a high level of know-ledge of the target language alters the response to acoustic stimuli. In fact, the results of the advanced group showed a tendency towards hypercorrection (Itkonen 1966: 200) or (over)- generalization (James 1988: 38–39), since these testees favoured the alternative /w/ even more unanimously than the native speakers of English.
In addition, the group of advanced learners failed to give as unanimous anwers as the other groups, which implies that there is an amount of fluctuation in the categories as the testees are gaining competence in a foreign language. The results of the group of native speakers of English had two major aspects: firstly, the categorisation differed from the Finnish groups in that the vowel /u/ was extremely rarely chosen by these testees, and secondly, the English system seems to contain a frictionless allophone of either the approximant /w/ or the fricative /f/. The former of these sounds is described as a labio-velar semi-vowel (Gimson 1975: 216–217) or labial- velar central gliding approximant (Morris-Wilson 1992: 117), and the latter is considered to be a voiced fricative by definition. This means that the synthesised glide sound (Lehiste & Peterson 1967) functions as a variant of these phonemes. Altogether, the results of this experiment show that the phonological systems of the two languages in contact differ in an important manner with respect to the categorisation of this acoustic continuum. The differences reflect some of the well-known difficulties that
Finnish learners of English have as they struggle to gain competence in the target language. One of the most well-known examples of the learning obstacles is the difficulty that Finnish learners of English have with the pronunciation of the word initial sounds in the phrase very well.
In the future, it is my intention to study the perception of phonemes with the help provided by advances in modern technology. The new method enables one to investigate the process of categorisation/- identification as it occurs on the subconscious level, i.e. in the brains of the speakers (see e.g. Näätänen 1992, Näätänen & Alho 1997, Näätänen et al. 1997). It is my conviction that this approach has a lot to offer for the theory of sound perception in a cross-linguistic situation.
References
Aaltonen, O. 1985. The effect of relative amplitude levels of F2 and F3 on the categorization of synthetic vowels. Journal of Phonetics 13, 1–9.
Best, C. T. & W. Strange 1992. Effects of phonological and phonetic factors on cross- language perception of approximants. Journal of Phonetics 20, 305–330.
Gimson, A. C. 1975. Introduction to the pronunciation of English. Bristol: J.W. Arrow- smith.
Itkonen, E. 1966. Kieli ja sen tutkimus. Helsinki: WSOY.
James, A. 1988. The Acquisition of second language phonology. Tübingen: Günter Narr Verlag.
Kuhl, P. K. 1993. Developmental speech perception. Implications for models of language impairment. Annals of the New York Academy of Science 682, 248– 263.
Lado, R. 1957. Linguistics across cultures. Michigan: University of Michigan Press.
Lehiste, I. & G. E. Peterson. 1967. Transitions, glides, and diphthongs. Journal of the Acoustical Society of America 33, 268–277.
Morris-Wilson, I. 1992. English segmental phonetics for Finns. Helsinki: Finn Lectura.
Näätänen, R. 1992. Attention and brain function. New Jersey: Erlbaum.
Näätänen, R. & K. Alho 1997. Mismatch negativity – The measure for central sound representation accuracy. Audiology & Neuro-Ontology 2, 341–353.
Näätänen, K., A. Lehtokoski, M. Lennes, M. Cheour, M. Huotilainen, A. Iivonen, M.
Vainio, P. Alku, R. J. Ilmoniemi, A. Luuk, J. Allik, J. Sinkkonen & K. Alho 1997.
Language-specific phoneme representations revealed by electric and magnetic brain responses. Nature 385, 432–435.
Pienimäki, M. 1998. Finnish and English approximants: The Acoustic continuum.
Linguistica Uralica 34, 258–263.
Wiik, K. 1965. Finnish and English vowels. Turku: University of Turku.
