LIETUVIŲ KALBOTYROS KLAUSIMAI XL (1998)

Rima BACEVIČIŪTĖ

Institute of the Lithuanian Language, Vilnius

DIFFERENCES BETWEEN INHERENT AND POSITIONAL LENGTH OF LOW VOWELS IN THE LUKŠIAI DIALECT

1. Linguistic literature most often reflects the conventional view that the vowels [aˑ, eˑ] < *-an-, *-en and [aˑ, eˑ] < *-à-, *-è- do not differ in any way. This is claimed both for the system of the standard language, where “in stressed syllables, vowels of historical and positional length [aˑ], [eˑ] are pronounced identically” (Pakerys 1995: 33; see also Vaitkevičiūtė 1960: 207, 217; Vaitkevičiūtė 1961: 31, 37; LKG 1: §17, §23; Mikalauskaitė 1976: 17), and for the closely related dialect of the southern Western Aukštaitians, in which “the monophthongs aˑ, eˑ < a, e do not differ in length from the stressed lengthened ā, ē” (Zinkevičius 1966: 80). However, the opposition between vowels of inherent (historical) and positional length has been observed for quite some time in many dialects, including the southern Western Aukštaitian dialects.

The twofold quantity of low vowels was discussed a century ago by A. Baranauskas (Baranovskij 1898: 15) and especially by K. Jaunius, who clearly stated that vowels lengthened by stress “in Kaunas County are of medium length (i.e. somewhat longer than short vowels, but not equal to long ones)” (Jaunius 1970: 83). Observations supporting the existence of an opposition between long and half-long vowels in the Kaunas subdialect are also found in later studies (Girdenis 1971: 205; Zinkevičius 1980: 98; Pakerys 1986: 33), and some of them also mention differences in quality: vowels of inherent length may be more open and perhaps even slightly nasalized (Girdenis 1995: 172; see also 1971: 205; 1981: 128). Vowels of this type have also been studied experimentally. B. Simanavičienė’s conclusions (1993: 46–53) from her study of the Veliuona dialect concerning the dependence of the intonation contour of [aˑ, eˑ] < *-à-, *-è- on a voiced or voiceless context¹ also allow us to consider quantitative differences between positional and inherent vowels. The existence of a quantitative opposition in the Igliauka dialect has been demonstrated by A. Kazlauskienė’s instrumental and auditory study² (1996: 128–130).

¹ According to this study’s data, positional-length vowels have a genuine circumflex intonation contour only next to voiced consonants. After or before a voiceless consonant, the intonation contour is more like an acute.

² It is especially important that the speakers of the dialect themselves perceive this difference. Thus, the words must differ in certain phonologically significant features.

102

This study attempts to investigate experimentally the quantity and quality of low vowels in the Lukšiai dialect. According to the current classification of dialects, this dialect belongs to the northern part of the Kaunas subdialect of Western Aukštaitian (Zinkevičius 1994: 26), while according to K. Jaunius and A. Salys’s older classification it would be the Veliuona dialect of Western Aukštaitian, since it systematically shortens the vowels [o·, e·] in open unstressed endings, e.g.: sā.ku, dā.ri, vāiku, sā.uli, etc. The dialect’s center is the town of Lukšiai (Šakiai District), located 8 km east of Šakiai and 52 km west of Kaunas.

2. The word pairs grąžtas – kraštas, šąla – kãla, patręšęs – patręšęs, patręšus – patręšus were selected for the study, and short sentences illustrating the meanings of the pairs under investigation were constructed. The target words were usually placed in penultimate position: aštrùz grã:štaz bū:davų ‖ vìsas krã:šta(s) sūdėgì ‖ ilgai šā-la šì·met ‖ senėi_įėu kā.la ta._stūba ‖ biškì patrē.šes tas_mē.dis ‖ smārkei patrē.šes ta._pìeva ‖ būvù gerai patrē.šus ta:z_braš kēs ‖ ta_o·bels patrē.šus įėu_būvù. The sentences were presented in random order, with additional sentences irrelevant to the topic under investigation interspersed among them. The material was recorded on magnetic tape from speaker A. Bacevičius, born in 1930, while the sentences by the author of the text were entered directly into the computer’s memory. Both speakers are representatives of the dialect under study. During recording, after listening to a clearly spoken sentence (including those irrelevant to the study), the informant repeated it freely, using the most neutral intonation possible. This made it possible to avoid the unnatural speech intonation that occurs when reading³. The sentences were repeated five times each.

The recordings were later transferred from magnetic tape to the computer’s memory and processed using “Kalbam44,” P. Kasparaitis’s speech-sound analysis program. The program displays an oscillogram of the input signal, which can be segmented into individual sounds; the program automatically measures the duration of the marked sound interval. The same program also enables spectral analysis of sounds: the screen displays a dynamic spectrogram of the signal, changes in formant trajectories, and a spectral slice of the segment.

3. The data were evaluated statistically. After measuring duration, the arithmetic mean x̄, the standard deviation s, indicating how far the values of the feature under investigation lie from the mean, and the 95% confidence interval were calculated. In addition, the significance of the results was tested using Student’s t-test. These calculations were performed on a computer using A. Girdenis’s program.

3.1. The results obtained are presented in Tables 1–2.

³ Plg. I. Mažiulienės (1996: 46–47) darbo metodiką.

103

Vowel duration in individual words

Table 1

Examples | n | x̄ (ms) | s (ms) | 95% confidence interval | t statistic | 10 | 203 | 29 | 182 ÷ 224 | 5.42 > 3.92 edge | 10 | 151 | 8 | 145 ÷ 157 freezes | 10 | 209 | 13 | 199 ÷ 218 | 6.40 > 3.92 chops | 10 | 173 | 12 | 165 ÷ 182 half-frozen | 10 | 203 | 23 | 185 ÷ 219 | 6.37 > 3.92 half-frozen | 10 | 152 | 9 | 145 ÷ 159 salted | 10 | 223 | 17 | 210 ÷ 235 | 9.56 > 3.92 salted | 10 | 161 | 11 | 153 ÷ 169

Table 2

Aggregated vowel duration

Examples | n | x̄ (ms) | s (ms) | 95% confidence interval | t [aˑ] < *-an- | 20 | 206 | 22 | 195 ÷ 216 | 7.29 > 3.57 [aˑ] < *-ā- | 20 | 162 | 15 | 155 ÷ 169 [eˑ] < *-en- | 20 | 213 | 22 | 202 ÷ 223 | 10.18 > 3.57 [eˑ] < *-ē- | 20 | 156 | 11 | 150 ÷ 160 longer vowels | 40 | 209 | 9 | 194 ÷ 224 | 7.18 > 5.96 shorter vowels | 40 | 159 | 10 | 143 ÷ 176

We see that there is a clearly significant duration difference between vowels of historical and positional length: the confidence intervals nowhere intersect or overlap. The values of the Student's t-test also show that this difference is significant, since in every case they exceed the critical value at the 99.9% confidence level.

The quantitative difference is even clearer when we aggregate the results (see Table 2): the confidence intervals again nowhere intersect, and the values of the Student's t-test are much greater than the critical value (twice as large for back vowels and almost three times as large for front vowels!). The overall ratio of the mean duration of shorter to longer vowels is 1:1.3.

3.2. The quantitative difference between vowels of inherent and positional length is also clearly shown by the scatterplot. If we plot the durations (ms) of vowels of positional length on the ordinate axis of the first quadrant of a coordinate system, and those of the corresponding vowels of inherent length in each pair⁴ on the abscissa axis, we obtain a graph that illustrates these differences in vowel quantity very well (see Fig. 1).

⁴ For clarity, the origin of the coordinate system is set at 100 ms rather than 0 ms.

104

Fig. 1. Scatterplot of vowel durations

All the points lie on one side of the diagonal, and some are even very far from it—thus, the two variants differ markedly.

4. An attempt was also made to analyze the qualitative parameters of these vowels. The program “Kalbam44” was used to analyze spectral sections of sound segments; the formant averages were then assessed using a special formant evaluation program developed by A. Girdenis, which, according to certain formulas (see Piotrovskij 1960: 29), calculates compactness (C¹⁰3), bemolization (b¹⁰3), and tonality (T) indices (a higher T indicates a higher vowel timbre, while a lower, especially negative, value indicates a lower one).

4.1. Since the distinctive features of sonorant sounds are determined by the first two formants, while the others often convey only the speaker’s individual characteristics or expressive content, analysis of the acoustic properties of sounds relies primarily on the first two. The spectra of male and female voices were examined separately because the acoustic characteristics of men’s and women’s voices differ considerably, although certain stable relationships remain. The results of the spectral analysis of male and female voices in this study are likewise analogous, except that the formants of the female voice are substantially higher owing to its higher timbre. Therefore, we present the measurements of the male voice, which provide a clearer picture. It is also worth discussing separately the vowel quality associated with front and back articulation.

Table 3

Formant values for the vowels [aː] < *-an- and [aː] < *-ā- in a male voice⁵

For example F₁ (Hz) F₂ (Hz) F₃ (Hz) C.10³ T gržtas 675 1326 2410 888 34 kraštas 685 1353 2320 889 61 šala 675 1320 2564 888 3 kala 703 1340 2525 893 5

⁵ Cf. female voice: grąžtas F₁=728 Hz, F₂=1410 Hz; kraštas F₁=785 Hz, F₂=1498 Hz; šala F₁=728 Hz, F₂=1369 Hz; kala F₁=785 Hz, F₂=1398 Hz.

105

As we know, all back vowels are low in timbre, which is determined by the broad and large oral resonator formed when the tongue is retracted. This is also indicated by the second formant, which lies in the low-frequency range. However, the F₂ of vowels of the [a] type with inherent length is still slightly lower than the F₂ of vowels with positional length. Thus, vowels of historical length are lower in timbre, which is determined by greater retraction of the tongue. Vowels with inherent length also have a slightly lower first formant, and are therefore somewhat higher in vowel height and more open. The lower first and second formants, both farther removed from F₃, are also a feature of the nasalization of vowels [a·] <*-an-, so we can state that spectral analysis confirms the supposition that a certain degree of nasalization has persisted in these vowels. The values of the third formants also differ: they are considerably higher in vowels of historical length.

In terms of spectral spread, low vowels are the most compact, because neither formant lies very far from the spectral midpoint (≈ 1000 Hz). Vowels of historical length are slightly less compact, but lower-pitched (see Table 3). These vowels are neutral with respect to flatness (b.10³ =168), since the lips are not involved in their articulation.

We can also see the relationships between the formants in the graph (see Fig. 2).

Fig. 2. Spectra of vowels [a·] < *-an- (----) and [a·] < *-a- (—) in the words grąžtas and kraštas⁶

4.2. The vowels [e·] and [æ·] have rather unstable articulation, and their acoustic properties consequently vary as well. The vowel [æ·] is particularly unstable: at the beginning of its articulation, F₁ and F₂ lie within the range of the formants of [e·], whereas at the end the sound becomes very similar to [a·]—open, more compact, and lower in timbre. Therefore, the segmental spectrum of these vowels was measured at three points: at the beginning, middle, and end of the sound. The calculations are shown in the table.

⁶ A similar graphic representation would also be obtained for the words šala – kala.

106

Formant values in male speech for vowels [e·] < *-ėn/[e·] < *-ė and [æ·] < *-æn/[æ·] < *-àè-vyro⁸

Examples | F₁ (Hz) | F₂ (Hz) | F₃ (Hz) | C.10³ | T patrėšęs | 576 / 605 / 626 | 1930 / 1787 / 1621 | 2580 / 2529 / 2379 | 828 / 842 / 857 | 399 / 320 / 247 patrėšęs | 587 / 615 / 610 | 1775 / 1714 / 1605 | 2471 / 2414 / 2326 | 839 / 849 / 855 | 337 / 269 / 259 patrėšus | 613 / 690 / 677 | 1627 / 1412 / 1351 | 2451 / 2349 / 2353 | 853 / 885 / 887 | 246 / 90 / 59 patrėšus | 623 / 627 / 678 | 1555 / 1448 / 1377 | 2363 / 2297 / 2249 | 860 / 869 / 886 | 216 / 163 / 95

We see that the formant characteristics of all [e]-type vowels change over the course of articulation. The first formant rises almost consistently, while the second falls. This is especially apparent before a hard consonant (in the /—C position).

5.2.1. Comparing low vowels in the /—C’ position, we see that the F₂ of vowels of historical length remains higher throughout articulation than that of vowels with positional length—they are higher in timbre, articulated further forward, and more tense (their formants lie farther from the “neutral” speech spectrum). The vowel with positional length is less tense, more back, and lower in timbre.⁹ At the beginning of articulation, the vowel with positional length has a higher first formant, but toward the end the F₁ of the vowel with inherent length rises higher; that of the positional vowel drops slightly. The vowel of historical length has a considerably higher third formant throughout articulation.

The compactness of these vowels also varies. At the beginning and middle of articulation, the vowel with positional length is more compact, whereas at the end the vowel with inherent length is more compact. These vowels are highly unstable in terms of tonality, but however much they vary, the differences between historical and positional long vowels remain. At the beginning and middle of articulation, the historical vowel has a much higher pitch; at the end, the vowel with positional length has a slightly higher tonality index.

5.2.2. Comparing these low vowels in the /—C position, we see that their instability is even more pronounced. At the beginning of articulation, the difference between the first formants is small: the F₁ of the vowel with positional length is slightly higher, but thereafter it rises only very slightly, whereas the F₁ of the vowel with inherent length rises much higher. In the middle of the segment, the formants differ by more than 60 Hz. At the end of articulation, however, the difference di-

Table 4: ⁷

⁷ The three formant figures indicate measurements at the beginning, middle, and end.

⁸ The female voice shows the same formant relationships.

⁹ In terms of quantity, these vowels differ the most.

107

…completely disappears. Only at the beginning is the historical long vowel F₂ higher, as in the /—C’ position; later, the formants move into the low-frequency region, the sounds become low in timbre, and their formant relationship is the same as in [a]-type vowels: the second formant of the positional long vowel is higher. Thus, at the end of the segment, the historical long vowel [æː] has the lower timbre. The historical long vowel has a higher third formant throughout.

At the beginning, the positional long vowel is more compact; in the middle, the inherent long vowel is more compact, while at the end the difference is indistinct. The tonality index clearly indicates both the dynamism of the sounds and their differences (see Table 4). At the beginning, again, as in the /—C’ position, the historical long vowel has a much higher pitch, whereas in the middle and at the end of the segment the positional long vowel has a higher pitch, as is also the case with [a]-type vowels.

In this case, the flatness index likewise shows no differences (b.10³=167), since the lips are also not involved in articulating these vowels.

The graphs of changes in formant trajectories show corresponding relationships as well (see Figs. 3 and 4).

3 pav. Balsių [eː] < *-eɪ- (---) ir [eː] < *-è- (—) formančių trajektorijų kitimas pozicijoje /—C.

4 pav. Balsių [æː] < *-æɪ- (---) ir [æː] < *-à- (—) formančių trajektorijų kitimas pozicijoje /—C.

108

6. Differences between inherent and positional long low vowels are also shown by the graphs of fundamental pitch and intensity (see Figs. 5–10). The graphs were produced using the WINCECIL22 computer speech-sound analysis program.

Fig. 5. Intensity of vowel [aː] < *-ãi (in conventional dB) and fundamental tone (Hz on a logarithmic scale) in the word grã.stas

Fig. 6. Intensity of vowel [aː] < *-à- intensity and fundamental tone in the word krā.štas

Fig. 7. Intensity and fundamental pitch of vowel [eː] *-en- < in the word patré·šęs

109

Fig. 8 Vowel [eː] < *-ė- intensity and fundamental frequency in the word patrė.šęs

Fig. 9 Vowel [æː] < *-ǣn- intensity and fundamental frequency in the word patrė.šus

Fig. 10 Vowel [æː] < *-æ̀- intensity and fundamental frequency in the word patrė.šus

We can see that the fundamental-tone curves of both front and back vowels of historical and positional length differ: the curve of all historical vowels is rising, while that of positional vowels is falling or rising–falling. Near the end of the sound, the fundamental-tone curve of back vowels rises somewhat higher; thus, the fundamental-tone range of these vowels is wider (see Fig. 5). Among the vowels of positional length, the curve of the vowel [æː] falls lowest at the end of the sound—the fundamental-tone range is the widest.

110

Vowels of historical and positional length also differ in the location of the fundamental-tone peak: for historical vowels it is closer to the end of the sound, and for positional vowels closer to its beginning.

The intensity of these vowels varies similarly: the intensity curves of almost all historical vowels rise slightly, whereas those of positional vowels fall. The intensity peaks of positional vowels occur earlier than the intensity peaks of inherent vowels.

6.1. Experimental studies of the quantity and quality of low vowels in the Lukšiai dialect show that there are indeed significant differences between them. Measurements of vowel duration clearly confirm that in the Lukšiai dialect, vowels of inherent length are longer than vowels of positional length in a ratio of 1:1.3. Statistical data also show this.

6.2. There are also qualitative differences between low historical and positional vowels. Back-row vowels of inherent length have a lower timbre, are slightly less compact, and have a lower pitch than the corresponding vowels of positional length. The lower values of both their first formants clearly indicate that nasalization has been retained. Front-row vowels of inherent length have a higher timbre, are more compact, and have a higher pitch than positional vowels; since their articulation is variable (especially the vowel [zɛː]), their quality at the end of pronunciation is very similar to that of [a]-type vowels.

6.3. The differences between low vowels of inherent length and positional length are also shown by the intensity and fundamental-frequency curves: those of inherent vowels rise slightly, while those of positional vowels fall.

6.4. These conclusions support the reliable assertion that the Lukšiai dialect does indeed have two types of non-short low vowels.¹⁰

Received 1998 07 02

LITERATŪRA

Baranovskij 1898: Барановский А. Заметки о литовскомъ языке и словаре. Санктпетербургъ: Типогр. Имп. Акад. наукъ.

Girdenis A. 1971: [Rec.] Lietuvių kalbos tarmės (chrestomatija). – Baltistica 7 (2), 201–209.

Girdenis A. 1981: Fonologija, Vilnius: Mokslas.

Girdenis A. 1995: Teoriniai fonologijos pagrindai, Vilnius.

Jaunius K. 1970: Kauniškių tarmė. – Drotvinas V., Grinaveckis V. Kalbininkas Kazimieras Jaunius, Vilnius: Mintis, 81–84.

LKG 1 – Lietuvių kalbos gramatika, 1: Fonetika ir morfologija / Ats. red. K. Ulvydas, Vilnius: Mintis, 1965.

Kazlauskienė A. 1996: Dve jop a žemutinių balsių kiekybė Igliaukos šnektoje. – Kalbotyra 45(1).

Mažiulienė I. 1996: Centrinės šiaurės žemaičių tarmės prozodija. – Kalbotyra 45 (1), 30–115.

Mikalauskaitė E. 1975: Lietuvių kalbos fonetikos darbai, Vilnius: Mokslas.

¹⁰ Baigdama norėčiau už vertingas pastabas padėkoti prof. habil. dr. A. Girdeniui.

111

Pakerys A. 1986: Lietuvių bendrinės kalbos fonetika, Vilnius: Mokslas.

Pakerys A. 1995: Lietuvių bendrinės kalbos fonetika, Vilnius.

Piotrovskij 1960 – Пиотровский Р. Г. Еще раз о дифференциальных признаках фонем. – Вопросы языкознания, №6, 24–38.

Simanavičienė B. 1993: Dvejopa [ā, ē] < *-ā-, *-ē- priegaidė veliuoniečių šnektoje. – Kalbotyra 42 (1), 46–53.

Vaitkevičiūtė V. 1960: Lietuvių kalbos balsių ir dvibalsių ilgumas, arba kiekybė. – Lietuvių kalbotyros klausimai 3, 207–217.

Vaitkevičiūtė V. 1961: Lietuvių literatūrinės kalbos balsinės ir dvibalsinės fonemos. – Lietuvių kalbotyros klausimai 4, 19–47.

Zinkevičius Z. 1966: Lietuvių dialektologija, Vilnius: Mintis.

Zinkevičius Z. 1980: Lietuvių kalbos istorinė gramatika 1, Vilnius: Mokslas.

Zinkevičius Z. 1994: Lietuvių kalbos dialektologija, Vilnius: Mokslo ir enciklopedijų leidykla.

UNTERSCHIEDE DER HISTORISCH SOWIE DER POSITIONELL NIEDRIGEN LANGVOKALE IM DIALEKT VON LUKŠIAI

Zusammenfassung

In der linguistischen Literatur wird behauptet, daß sich in der litauischen Gemeinsprache und den Dialekten die historisch sowie positionall niedrigen Langvokale nicht unterscheiden. Der vorliegende Beitrag begründet die quantitativen und qualitativen Unterschiede dieser Vokale in der Mundart von Lukšiai.

Die experimentellen Untersuchungen und statistischen Berechnungen zeigen, daß die Dauer historisch sowie positionell niedrigen Langvokale im Verhältnis 1:1,3 steht. Auch die Qualität dieser Vokale weist Klangunterschiede auf: die historischen Langvokale der hinteren Reihe weisen einen niedrigeren Timbre auf, sind weniger kompakt, zeichnen sich durch einen niedrigeren Ton aus. Die beiden ersten niedrigeren Formanten weisen die erhaltene Nasalierung auf. Die historischen Vokale der vorderen Reihe weisen einen höheren Timbre auf, sind kompakter, zeichnen sich durch einen höheren Ton aus.

Von den Unterschieden der historisch sowie positionell niedrigen Vokale zeugen auch die Kurven der Intensität sowie des Haupttones: bei den historisch niedrigen Vokalen klingen die Schwingungen ein wenig auf, während sie bei den positionsbedingten abklingen.

112