Dialect variation in Hungarian vowel pronunciation: Atlas data compared to acoustic measurements

References –

Bailey, Guy, Jan Tillery & Claire Andres 2005. Some Effects of Transcribers on Data in Dialectology. American Speech 80. 3–21.
Brissos, Fernando 2014. New insights into Portuguese central-southern dialects: understanding their present and past forms through acoustic data from stressed vowels. Journal of Portuguese Linguistics 13. 63-115.
Deme, László – Imre Samu (eds.) 1968–1977. A magyar nyelvjárások atlasza 1–6. Budapest: Akadémiai Kiadó.
Imre, Samu 1971. A mai magyar nyelvjárások rendszere. Budapest: Akadémiai Kiadó.
Jones, Jonathan A. & Margaret E. L. Renwick. 2021. Spatial analysis of sub-regional variation in Southern US English. Journal of Linguistic Geography 9. 86–105.
Labov, William, Sharon Ash, Charles Boberg 2006. The Atlas of North American English: Phonetics, Phonology and Sound Change. Berlin/New York: Mouton de Gruyter.
Purnell, Thomas 2013.Hearing the American language change: The state of DARE recordings. American Speech 88: 275–301.
Thomas, Erik R. 2018. Acoustic Phonetic Dialectology. In: Charles Boberg, John Nerbonne & Dominic Watt (eds.) The Handbook of Dialectology John Hoboken, NJ: Wiley Blackwell. 314–329.
Vargha, Fruzsina S. 2017. A nyelvi hasonlóság földrajzi mintázatai. Magyar nyelvjárások dialektometriai elemzése. Budapest: Magyar Nyelvtudományi Társaság.
Vargha, Fruzsina S. 2020. Magánhangzó-minőségek meghatározása nyelvatlaszadatok alapján. Magyar Nyelv 116. 442–461.

 Abstract –

Research on dialect variation has been long concerned about the reliability of the data of earlier studies, especially those produced on the phonetic level relying merely on the perception of fieldworkers. To tackle this issue, we assessed and compared two data sources of vowel quality in this study: transcribed dialect data and formant measurements carried out on archived audio recordings.

Descriptions and categorisations of Hungarian dialects are largely based on phonetic transcriptions, especially dialect atlas data. Dialect areas are mainly determined by differences in the pronunciation of vowels. The outcomes of previous quantitative studies on Hungarian dialect variation have all been impacted by the ways fieldworkers transcribed vowel qualities (e.g., Imre 1971, Vargha 2017). Although dialect transcriptions contain fine phonetic details (with the heavy use of diacritics) and can capture and describe subtle differences, transcribed datasets are seen as subjective, reflecting the researchers’ preconceptions, individual practices, or the presumed patterns of dialect variation (Baley et al. 2005). To challenge the problems with transcribed atlas data, a possible answer is the acoustic analysis of dialect recordings and the presentation of formant values on dialect maps (Labov et al. 2006).

In our study, we deal with transcribed data (collected between 1950 and 1964) from the Atlas of Hungarian Dialects (Deme & Imre 1968–1977) as well as recordings (mainly spontaneous speech data) made at the same investigation points between 1960 and 1964, during the field work of the same atlas. The usefulness of older recordings in the study of dialect change was demonstrated in several studies (e.g. Brissos 2014, Purnell 2013, Thomas 2018, Jones & Renwick 2021), and the acoustic analysis of the field recordings can potentially lead to the verification of the transcribed atlas data as well. We measured vowel formants in these recordings.

To contrast our formant measurements with the subjectively transcribed data, we consider the latter as representations of what we call “perceived vowel qualities”. We estimated typical F1 values (corresponding to the openness) at each atlas location in function of the frequency of different vowel variants for the three most open vowels: /aː/, /ɔ/ and /ɛ/ (for the method see Vargha 2020). In our analysis, we compared these perceived vowel qualities (computed from transcribed atlas data and expressed as numerical values) with the results of acoustic measurements.

Our findings suggest that, while there is a strong correspondence between the two types of vowel quality, some differences also appear. Perceived vowel qualities of /ɛ/ and long /aː/ as derived from transcribed atlas data reflect relative vowel distances in the acoustic space rather than absolute vowel qualities. We infer from these findings that contrasting transcribed atlas data with formant measurements provides new potential interpretations for our old datasets.