Eesti keele lühikeste klusiilide häälduse variatsioon ja seda mõjutavad tegurid
Variation in the pronunciation of Estonian short plosives and its affecting factors

Summary/Abstract: Plosives in Estonian have been considered voiceless. However, analysis has shown that short plosives tend to get at least partially voiced and otherwise reduced in connected speech. This seems to be quite a universal tendency in different languages. The present paper investigates short plosives in intervocalic position in most frequent content words. Phonetic materials were extracted from the Phonetic Corpus of Estonian Spontaneous Speech. Patterns in the reduction of plosives and possible influences of stress and vowel context were investigated. Two kinds of allophones emerged – those that were fully realised and had a distinguishable release burst, and the reduced ones that had lost the burst phase. The amount of reduced tokens differentiated the velar plosive [k] from others. As [p] and [t] both had over 65% of fully realised tokens, but over 60% of [k] tokens were reduced. [k] also had most different allophones. Among fully realised tokens there were voiceless, partially voiced, and fully voiced allophones. The voiceless allophone was the rarest, 19% tokens of [p] and only 10% of [t] and [k] were voiceless. Most frequent allophones among phonemes were partially voiced for [p] (29%), fully voiced for [t] (37%), and reduced voiced for [k] (47%). Closure durations were related to place of articulation. [p] had the longest average durations and [t] the shortest. Across all tokens [k] and [t] had similar average durations but within allophones durations were closer between [k] and [p]. Burst durations were the longest, around 30 ms for [k] and almost the same duration, between 21–23 ms, for [p] and [t], with the exception of the voiceless allophone [t], which was 27 ms. Closure durations differed significantly between [p] and [t] and burst durations were significantly different between [k] and other phonemes. Stressed positions included both lexical and contrastive stresses. Stress had some effect on the allophonic distribution but almost none on durations. As expected, there were more voiceless and partially voiced tokens in stressed position and more reduced tokens and total loss in unstressed position. Differences were the biggest for [t] and the smallest for [k]. Durations differed very little, whereas none were statistically significant. Vowel context had some influence on allophonic distribution. The influence was the biggest on [t]. Overall, there were more fricative and approximant tokens around [i].[t] had more partially voiced tokens and less voiced tokens around labial vowels [o, u]. There were more reduced [k] tokens around [a] and [i]. On durations the vowel context again influenced [t] the most. Durations between all vowel contexts were statistically different for [t] (p<.01); the longest durations appeared after [i]. In general, the present study confirms the results of the previous ones. Allophonic distribution is very similar to the previous study of Estonian spontaneous speech. Closures were at least partially voiced in most cases which refer to carry-over voicing of the previous vowel. Vocal cord vibration stops for a very brief time or does not stop at all during short closure times. Burst durations appear to be longer in spontaneous speech than in read speech. Little influence of stress is in accordance with findings in the studies on Estonian and some other languages. Vowel influences were dependent on the place of articulation. Bilabial [p] was the least affected both in allophonic variation and in durations. Velar [k] was influenced by the vowel context but it mostly occurred in whole as extensive reduction; different vowels had more effect on the allophonic variation than in the case of [p] but durations were almost unaffected. Influences on [t] mostly occurred as significant duration differences; yet, also some differences in allophonic variation occurred.

• Details
• Contents