Rytų aukštaičių vilniškių ilgųjų balsių priegaidžių audicinis tyrimas

Acta Linguistica Lithuanica LIV (2006), 63 – 82 An Auditory Study of the Tonal Accents of Long Vowels in the Eastern Aukštaitian Vilniškiai Dialect Jolita Urbanavičienė Lietuvių kalbos institutas Vytautas Kardelis Vilniaus universitetas The article describes an auditive investigation into the tones of long monophthongs in the north-eastern dialects of the Vilnius Lithuanian dialect area and presents its results. Four points were singled out for an auditive analysis: Adutiškis, Svirkos and Vidutinė, situated along the Lithuanian-Belorussian border, belong to the north-eastern dialects of the Vilnius area; the fourth, Linkmenys, represents a transition between the dialects of the Utena and Vilnius areas and is closer to the former with respect to vocalism. These points have been carefully chosen with the aim of subjectively evaluating the auditive salience of tones in different points of the north-eastern part of the Vilnius dialect area. In investigating the tones, four minimal pairs with front and back vowels and polypthongs were chosen. The experiment was carried out on groups of auditors (consisting of seven persons each) from every investigated point. The statistical results were evaluated with the aid of the programme UKR.PAS, which takes the U criterion into account. The main conclusions from the experiment are the following: in the investigated points of the north-eastern Vilnius area no complete levelling out of tones could be established. Only in sundry cases can we speak of a decreased contrast between the tones. The results of the investigation are sensitive to a set of factors with respect to the auditors: (1) use of dialect, (2) age, (3) education level, (4) musical hearing. Tones are distinguished better by those auditors who have a good command of the dialect and realise tone oppositions in their own utterances. The individual distinguishing features of the dialect have no influence on the identification of tones. The salience of tones in long vowels does not depend on the phonetic properties of the syllabic nucleus either. 63 JOLITA URBANAVIČIENĖ, VYTAUTAS KARDELIS 1. TRENDS IN THE NEUTRALIZATION OF TONAL ACCENTS IN THE EASTERN DIALECTS Lithuanian dialectological studies, based mostly on empirical observations, note a tendency toward the loss of tonal accents in monophthongs in the eastern and southern dialects and in urban colloquial speech (Gerullis 1930: xxxi-ii; LKG 1965: 132; Pakerys 1982: 154; DLKG 2005: 39). It was assumed that in the eastern subdialects the tonal accents of both monophthongs and diphthongs had generally disappeared. In 1968, J. Kazlauskas wrote: “It is entirely wrong to say that in the Eastern Aukštaitian dialects tonal accents have been preserved in stressed diphthongs but leveled in long vowels and diphthongs ie, uo. There can now be no question of any tonal accents in the Eastern Aukštaitian dialects” (Kazlauskas 2000: 14). He proposed writing the words ká.lnas, vaĩ.kas as kà.lnas, vãl.kas. Later, the idea of the complete neutralization of tonal accents was abandoned (see Zinkevičius 1974: 93–94; Kosienė 1982: 61–71). Both the Lietuvių kalbos atlasas (LKA₂ 1982: 125) and the reader Lietuvių kalbos tarmės (LKT 1970: 37) hold that the southern and eastern Dzūkai (in the current classification, the Southern Aukštaitians and Eastern Aukštaitian Vilniškiai) distinguish tonal accents less well than the Kapsai and Zanavykai, but better than the Eastern Aukštaitians. It is thought that, among all Lithuanian dialects, the Samogitians and Western Aukštaitians distinguish tonal accents best. In E. Mikalauskaitė’s view (1975: 80), this is because Samogitians always stress the first component and distinguish tonal accents by the movement of the voice, or pitch. Experimental studies by A. Girdenis refined this claim. It was established that, at least in Samogitian, tonal accents differ in the rate of change in fundamental frequency and in the distribution of energy: the acute’s energy is concentrated at one point in the sound and changes abruptly, whereas the circumflex is not abrupt and is pronounced without concentrating energy (see Girdenis 1974: 186 = 2000: 300; 1996: 82 = 2001: 254; 1998: 38 = 2001: 404; 2003: 272). Eastern and Southern Aukštaitians, by contrast, distinguish the tonal accents of diphthongs or mixed diphthongs well because the position of the stressed sound matters to them. The tonal accents of monophthongs become more alike in these dialects because the movement of the voice—its rise and fall—becomes the same when pronouncing either accent; consequently, the falling accent approaches the rising accent, and vice versa. A. Pakerys examined the perceptual salience of tonal accents and their acoustic features in greater detail (1982: 151–156). He found that tonal accents are not completely neutralized in any dialect. The tonal accents of diphthongs or mixed diphthongs beginning with the vowels [a], [e] are distinguished best. The least well identified are acute diphthongs beginning with 64 Rytų aukštaičių vilniškių ilgųjų balsių priegaidžių audicinis tyrimas [i][i], [u] (apparently because these sounds tend to retain lax articulation even in stressed position; cf. DLKG 2005: 39). The perceptual salience of tonal accents in monophthongs and polyphthongs [ie], [uo] is also low: Samogitians distinguish the accents of such syllables best, Western Aukštaitian Kauniškiai less well, and Southern and Eastern Aukštaitians worst. A. Pakerys argues that the salience of tonal accents in the Western Aukštaitian and Samogitian dialects depends much less on the quality of the syllable nucleus than it does in the Southern or Eastern Aukštaitian dialects (see Pakerys 1982: 154). There have been few studies, especially experimental ones, of the Eastern Aukštaitian Vilniškiai subdialect. Based on empirical observations, dialectologists indicate that the neutralization of tonal accents across the Eastern Aukštaitian area is not uniform. The farther east one goes, the more the acute and circumflex resemble the falling accent; however, at the easternmost margins—in Adùtiškis, Tverėčius, Lazūnai, and Zietela—the two accents are again clearly distinguished (Zinkevičius 1966: 33). Empirical studies of other peripheral areas confirm this tendency: in Dievėniškės and Gervėčiai, the tonal accents of both stressed and pre-stressed long vowels are clearly audible (Kardelytė 1975: 13; Mikulėnienė, Morkūnas 1997: 16; DvŽ 2005: XVI). An experimental analysis of the acoustic features of stressed vowels in one of the peripheral Vilniškiai areas (Svirkos) found that acute and circumflex monophthongs differ in a bundle of features: 1) duration (falling-accent vowels are longer); 2) peak values of fundamental frequency and intensity (when acute vowels are pronounced, the peaks in loudness and pitch are reached earlier than when circumflex sounds are articulated); 3) quality (falling-accent vowels retain more of their inherent qualities)1 (see Urbanavičienė 2005: 158). Thus, it may be assumed that the perceptual salience of tonal accents may vary within the Vilniškiai area itself: it is more salient in the eastern area closer to Belarus and less salient in the west. Whether this hypothesis is correct can be confirmed 1 In a similar study of the acoustic features of monophthong tonal accents in standard Lithuanian, R. Kudirka (2005: 1–21) found that acute and circumflex vowels do not differ statistically in duration, fundamental-frequency or intensity values, or formant structure. Only general systematic patterns were observed: circumflex sounds are longer and have a wider range than acute vowels. These results can be viewed as manifestations of the neutralization of tonal accents characteristic of urban residents’ colloquial speech (the speakers selected were younger Kaunas residents; cf. also DLKG 2005: 39). 65 Jolita Urbanavičienė, Vytautas Kardelis only subjective studies—auditory experiments. The studies described in this article aim to: (1) determine how listeners identify the differences between the acute and circumflex tones in their native dialect; (2) compare the degree of distinctiveness of the tones in different Vilnian dialect points. Auditory studies are considered subjective because they depend heavily on the informants selected (see also Pakerys 1982: 154). Moreover, comparisons of results from different subdialects—which are also influenced by the phonetic, phonological, and prosodic features of those subdialects—can only be made conditionally. Such experiments, however, provide authentic data on dialect speakers’ auditory perception of tones and complement objective—instrumental—studies; the validity of phonological hypotheses is demonstrated by both auditory and instrumental experiments. 2. MORE IMPORTANT PHONETIC FEATURES OF THE STUDY POINTS The tones of four points were studied auditorily. Three of these points—Adutiškis, Svirkos, and Vidutinė—lie along the Lithuanian-Belarusian border and belong to the northeastern Vilnian dialect, while the fourth, Linkmenys, lies at the boundary between the Utenian and Vilnian subdialects and, in terms of its vowel system, is closer to the Utenian dialect² (see Fig. 1). These differing points on the periphery of the Vilnian dialect were deliberately selected in order to subjectively assess the auditory salience of tones in several locations within the subdialect. The areas in which the study points are located differ from one another in vowel quality and quantity, as well as in certain consonantal features. In the Adutiškis area (spoken in the town of Adutiškis, Švenčionys district), the long reflexes of the Baltic ā, ē vowels with low rise are retained in stressed position, and half-long reflexes in unstressed position, e.g.: ká·še ‘porridge (verb)’; ka.štuvù ‘with a strainer’; knĩ·ga.s ‘books’; ne·rà ‘is not’. In the Svirkos area (which covers most of the Adutiškis parish, Švenčionys district), these reflexes are diphthongized—pronounced [ieː/ᵉ·], [uː·], e.g.: kᵘ·š́·še ‘porridge (verb)’; žĩr·vᵘ·ž́·i ‘in the field’; kᵘ·š́·minᵘ ~ kòmino ‘chimney’; nie·rà ‘is not’. In stressed position 2 Here, in unstressed position, the vowels ę· and ie· are neutralized; cf.: piė·nas: pe.nė.li.s. 66 Rytų aukštaičių vilniškių ilgųjų balsių priegaidžių audicinis tyrimas Fig. 1. Northern Vilnius-area dialect points: Adutiškis, Svirkos, Vidutinė, and Linkmenys. used interchangeably are half-long [i.], [u.] and short [i], [u] vowels. Speakers from the Vidutinė area (Strūnaitis parish, Švenčionys district) pronounce a less tense and less labialized than in standard Lithuanian [ɔ·] and a less close [e·]. In unstressed position, a labialized [ɔ·] and a raised [e·] also occur here, e.g.: kɔːšė ‘porridge (noun)’, knĩːgɔˌi ‘in the book’, kɔˌštuvù ‘with a strainer’. In this area, short [i], [u] are also used in stressed position. A significant phonetic feature is also a hardened consonant r before back vowels, e.g.: škĩːrus. In the Linkmenys area (Ignalina district), which lies farther from the aforementioned points and the territory of Belarus, the vowels [ɔː] and [e·] occur in stressed position, while half-long [a.] and [e.] occur in unstressed position, e.g.: mɔːjus ‘having tried’, subatɔˌi ‘on Saturday’, tėːvas ‘father’, but knĩːga.i ‘in the book’, kɔːmina ‘of the chimney’, te.vėːli.s. In addition, in the Linkmenys area, dzūkavimas occurs only according to the first feature, i.e., before i-type vowels the consonants [t], [d] occur, rather than [c], [ʒ], e.g.: kiėːti.s ‘kietis (a plant)’, dĩr.va.i ‘in the field’, gɔːrdi ‘tasty’. In the other three areas under study, dzūkavimas occurs according to both features. 67 JOLITA URBANAVIČIENĖ, VYTAUTAS KARDELIS 3. MATERIAL AND METHODS OF THE AUDITORY EXPERIMENT The tones of monophthongs can be classified as highly subtle phonological phenomena. Auditory experiments are conducted to investigate such unstable elements of speech, whose existence not all speakers perceive and not all speakers realize3. This establishes whether dialect speakers themselves (in this case, the listeners) can distinguish a particular phenomenon by ear. This study took account of the distinctive character of the Vilnius-area subdialect: material for the auditory experiments was recorded separately from the informant at each point. The processed data, prepared for listening, were presented to different groups of listeners representing the four points under study. Four minimal pairs containing front and back sounds of monophthongal and polyphthongal origin were selected for the study of tones. Universal pairs suitable for all four points were used. Only the pair [kó·še.: kɔ̃·še.] had to be replaced for the Linkmenys point with the pair [mó·jus: mɔ̃·jus], after it was observed that both members of the first pair had a circumflex tone. The test words were inserted into short, clear, three-word dialect sentences, where phonetic differences are also clearly visible; cf.: Adutiškis point píe·nas rú·ksta vá.saru.: dú·mai rú·ksta iš kɔ·mina. rú·bais skī·rɔus iš_kitu.: knī·ga.s skī·rɔus ne·rà píe·nu. ká·še. ká·štuva: bú·lbu. ká·še. gar̃ži žĩr.va.i kiẽ·cis á.uga: gelažìe·s kiẽ·cis neišpasakì·tas 3 The pioneer of experimental commutation (lit. commutation test) was L. Hjemslevas (Hjelmslev 1935). M. K. Jensenas (Jensen 1961) refined the methodology for evaluating the data obtained. V. Labovas (Labov 1972) applied the commutation method to the study of psychophonetic phenomena. For formulas used in analogous biological and medical research, see Urbach (1975). In Lithuania, the methodology of auditory experiments was adapted for phonetic research by the Experimental Phonetics Laboratory of Vilnius University, headed by A. Girdenis. This methodology is often used to analyze various phenomena of Lithuanian dialects—for example, the tones of Kupiškėnai monophthongs (see Eidukaitienė 1977), the opposition of tones in diphthongal endings in Utena dialect (see Kosienė 1979), the murmured vowels of Northern Panevėžys dialect (see Kačiuškienė 1983), and the phonological secondary stress of Utena dialect (see Kosienė, Girdenis 1979 = Girdenis 2000b: 141–152). Among the most recent studies, auditory experiments on tones in Southern Aukštaitian (Leskauskaitė 2004: 192–194) and Western Aukštaitian Šiauliai dialect (Kazlauskaitė 2002: 144–145) are noteworthy. 68 Rytų aukštaičių vilniškių ilgųjų balsių priegaidžių audicinis tyrimas Svirkos locality bread is baking in summer.: smoke rises from embers. clothes I separate from others.: books I separate from the sea of milk. a trough. a trough: bulb. a trough. iron ore, iron ore grows: an indescribable amount of iron Vidutinė locality The milk curdles in summer.: The smoke rises from the chimney. Clothes differ from other things.: The books differ, and they do not need milk. A shoe. A shoe with a taste: Ibu. A shoe. Bitter peas grow: an indescribable amount of iron Linkmenys locality The milk curdles in summer.: The smoke rises from the chimney. Clothes differ from other things.: The books differ, they do not need to be peeled; turnips on Saturday were peeled. In the fields hard things grow: an indescribable amount of iron A representative of each locality (in this case, the reader) had to repeat the specified sentences five times: in the dialect, with neutral intonation, at a normal pace, and with phrase stress emphasizing the middle word. The readers chosen were: (1) older-generation local residents (born 1933–1936), whose dialect was as little influenced by the standard language as possible and whose articulation was quite clear; (2) men, since the ear distinguishes low-pitched voices better (male voices are also more suitable for spectral analysis—they have more harmonics than female voices; see Girdenis 2003: 62, note 15); (3) people who had lived sedentary lives in the same village all their lives. The Adutiškis material was read by4 B. G. (born 1935 in Adutiškis, Švenčionys district); the Svirkos reader was L. M. (born 1935 in Kackoniai village, Švenčionys district); the sentences in Vidutinė were read by J. V. (born 1936 in Kaŗklinė village, Švenčionys district); and in Linkmenys by P. K. (born 1933 in Linkmenys village, Ignalina district). The material was recorded on a SONY tape recorder with a directional microphone and then transferred from there to a computer. 4 The sentences were recorded using a repetition method: the reader was told the required sentence in their native dialect and repeated the phrase five times. Repeating the sentence avoids the unnatural intonation that arises when reading text. Moreover, older-generation readers find it more difficult to read text than to repeat it. 69 JOLITA URBANAVIČIENĖ, VYTAUTAS KARDELIS Using PRAAT4.1.27, the central words were cut out from the phrases. Ten pairs were formed with them, with the words arranged according to a random-number table (see Girdenis 2003: 378–381). If an odd number was found in the table, the acute-accented form was written first (e.g.: rūksta); if the number was even, the circumflex-accented form was placed first in the pair (e.g.: rūksta). E.g.: rūksta “rūksta” – rūksta “rūgsta”, rūksta “rūksta” – rūksta “rūgsta”, rūksta “rūksta” – rūksta “rūgsta”, rūksta “rūksta” – rūksta “rūgsta”, rūksta “rūksta” – rūksta “rūgsta”, rūksta “rūksta” – rūksta “rūgsta”, rūksta “rūksta” – rūksta “rūgsta”, rūksta “rūksta” – rūksta “rūgsta”, rūksta “rūksta”. Before the “blocks” of word pairs, sentences from which the pairs had been cut were inserted. The words of each speaker were arranged separately. The material prepared in this way was kept on the computer, since transferring it to a tape recorder again would have distorted the sound quality5. The auditory experiments were conducted using a laptop6. The work then proceeded with groups of listeners, each consisting of seven people. The listeners had to hear the word pairs presented and determine which sentence the word they heard belonged to. They were given sheets containing sentences written in semi-phonetic spelling (numbered 1 or 2). First, they were allowed to listen to those sentences; afterwards, only the words were presented in pairs. The listeners had to determine which sentence the first word in each pair came from and which sentence the second came from (they indicated this with numbers). Because of the complexity of the experiment, younger and middle-aged people were selected as listeners. Their educational background (especially in philology), experience in playing music or singing, and relationship to their native dialect were also taken into account. These factors are thought to be capable of influencing the results of the experiment. 5 K. Garšva proposes a different methodology for auditory research: supposedly it is enough to pronounce one word naturally and let the listeners choose which member of a minimal pair it resembles more. In the author’s view, when homonyms with different meanings occur next to each other, there is a tendency to distinguish them phonetically as well (see Garšva 2004: 25). 6 Before personal computers became available, the speaker also participated directly in auditory experiments: they pronounced the phrases and pairs written on sheets, while the listeners marked which phrase each word belonged to. Modern technology makes it possible to avoid direct contact between the speaker and the groups of listeners; moreover, the speaker can also participate in the experiment as a listener (cf. Girdenis 2003: 62–63). 70 Rytų aukštaičių vilniškių ilgųjų balsių priegaidžių audicinis tyrimas The recorded pairs were listened to in Adutiškis by local residents (or residents originating from Adutiškis) who spoke the dialect: R. G. (born 1939), A. G. (born 1964), D. G. (born 1969), I. G. (born 1972), D. G. (born 1973), S. G. (born 1988), and I. G. (born 1994). The group of listeners from Adutiškis was exceptionally musical—five of the experiment participants had experience playing music or singing. Two listeners had higher education, one of them being a qualified philologist. The Svirkos group consisted of dialect speakers born, raised, and living in the Svirkos area (or having moved away from it) who spoke the dialect well. They were V. M. (born 1935), living in Svirkos; V. G. (born 1968), from Kackonys village, living in Vilnius; T. R. (born 1969), from Kackonys village, living in Švenčionys; R. S. (born 1969), from Kackonys village, living in Vilnius; R. P. (born 1972), living in Svirkos; G. J. (born 1972), living in Svirkos; and V. N. (born 1973), from Kackonys village, living in Vilnius. Five listeners spoke the dialect well and used it regularly, although three of them currently lived in a city. Two other listeners were permanent residents of Svirkos village, but their dialect had been considerably influenced by the standard language. Three experiment participants had completed higher education, one of them being a Lithuanian-language specialist. One listener reported having a well-trained musical ear. Four residents of Vidutinė participated in the auditory experiment in Vidutinė—E. P. (born 1940), J. M. (born 1940), J. B. (born 1935), and B. J. (born 1939)—as well as three people from the neighboring village of Karklinė: R. M. (born 1974), A. J. (born 1961, currently living in Vilnius), and I. C. (born 1945). All experiment participants spoke the dialect. Four listeners had higher-education degrees, one of them being a Lithuanian-language specialist. One of the listeners noted that she had a musical ear. The Linkmenys group comprised listeners born and raised there: V. P. (born 1972), D. A. (born 1972), N. A. (born 1973), D. P. (born 1974), G. K. (born 1976), J. K. (born 1978), and D. K. (born 1981). All spoke the dialect well. The four experiment participants living in Linkmenys used the dialect constantly. Three listeners now lived in Visaginas, Utena, and Vilnius, but had retained the dialect well and spoke it often. Two listeners had completed higher education (one of them a Lithuanian-language specialist); three had secondary vocational education, and two had post-secondary education. The listeners reported that they did not have a well-trained ear. 71 Jolita Urbanavičienė, Vytautas Kardelis 4. EMPIRICAL EXPERIMENT RESULTS All participants in the experiment perceived that the words under study differed in pitch accent, although this was not emphasized7. At the beginning of the experiment, the listeners were allowed to listen only to the sentences so that they could identify the differences between the pairs under investigation. After listening to the sentences, the Adutiškis listeners indicated that the words [kiėˑcis, skiˑr̄us, kȧˑšė, rụˑksta] were pronounced more abruptly, “sharply,” and more briefly, while their circumflex counterparts were longer and more drawn out. After listening to the pair [káˑšę.: kȧˑšę.], several listeners heard a difference in the quality of the stressed vowel: by ear, the acute vowel sounded like a “pure” [ɑˑ], whereas the circumflex vowel was slightly labialized – [åˑ]. Thus, listeners living in Adutiškis emphasized differences in duration and quality. Listeners in the Svirkos group said that the vowel [iˑ] in the word [skīˑr̄us] sounded clearer to them than in the word [skīˑr̄us], and that in the word [kiėˑcis] [ieˑ] was more diphthongized than in the word [kiėˑcis]. Some listeners heard that the circumflex vowels were longer. Like the Adutiškis listeners, representatives of the Svirkos point noted subtle differences in duration and quality (clearer articulation and diphthongization), but did not hear differences in loudness or pitch. After listening to the words spoken in sentences, listeners from Vidutinė and Karklinė heard that in the words [rūˑksta, skīˑr̄us] the stressed vowel was longer and more sustained, whereas in [rūˑksta, skīˑr̄us] it was shorter and more abrupt. However, after listening to the pair [káˑšę.: kȧˑšę.], local residents heard that the circumflex-stressed sound was sustained longer, although its onset was said to be more strongly emphasized. Some participants in the experiment heard that the words [skīˑr̄us: skīˑr̄us] differed in the quality of [r] – the noun’s consonant was harder than the verb’s. All listeners acknowledged that the pairs [rūˑksta: rūˑksta], [kóˑšę.: kȧˑšę.] were more clearly distinguished than [skīˑr̄us: skīˑr̄us], [kiėˑcis: kiėˑcis] (members of the Adutiškis group said much the same). All the Linkmenys listeners, after hearing the pairs of words under investigation, indicated differences in duration. In their view, the stressed vowels in acute-accented words [kiéˑti.s, skīˑr̄us, mȯˑjus, rūˑksta] are more abrupt and shorter, while the corresponding vowels in the circumflex pairs are longer and more sustained. Some listener groups particularly emphasized the length of the circumflex vowels. Several informants stated that in the acute form of the word [kiėˑti.s] 7 This, of course, was self-evident only to experiment participants with a background in Lithuanian studies. 72 Rytų aukštaičių vilniškių ilgųjų balsių priegaidžių audicinis tyrimas the vowel onset is clearer and stronger. The listeners heard no qualitative differences. 5. STATISTICAL EXPERIMENT RESULTS The results were evaluated using the UKR.PAS program, which calculates the so-called u criterion. After the total number of responses and the number of correct responses are entered, the program calculates a 95% confidence interval, the u criterion, and its critical values. If the value of the u criterion is lower than the critical value, the number of responses does not differ significantly from the number expected by chance, that is, the difference between the compared words is insufficient and the features under investigation are irrelevant (or were not brought out by the experiment). If the value of the u criterion is greater than u0,05 (u > 1,96), the words under investigation are considered to differ in a phonologically significant feature, since they exceed the 50% chance-guessing threshold. When a critical threshold of 67%, or 2/3 correct responses, is selected, the experimental results are considered highly reliable (for more detail, see Girdenis 2003: 63–64)8. The results of the auditory experiments at each point studied are discussed below. 6. ADUTIŠKIS POINT Adutiškis speakers distinguished the pairs with back vowels best, [rūˈksta: rūˈksto, káːše.: kàːše.], and words with front stressed vowels somewhat less well, [skíːrǫus: skìːrǫus, kiéːcis: kièːcis]. 8There are also other views on the evaluation of auditory-experiment results. For example, Jensen (Jensen 1961: 159–161) is convinced that 85–65% correct responses indicate a weak (“semi-phonological”) opposition. In his view, 100–85% correct responses would indicate a clear opposition (Garšva (2004: 25) also holds this view). Still other researchers (e.g., Harris (1963: 32–33)) discuss only results reaching the thresholds of 100% (unambiguous difference) and 50% (guessing). E. V. Eidukaitienė considers responses significant when they exceed the 50% chance-guessing threshold: “if the lower bound of the interval reaches 50%, there is no phonological opposition between the phenomena under investigation; if the lower bound of the interval is substantially higher than 50%, the results are significant and there is a phonological opposition” (Eidukaitienė 1977: 21). Most often, the number of correct responses exceeds the 50% threshold but does not reach 85%. Results of 100% can be considered unrealistic, since they could be achieved only with ideal speakers and listeners who have well-trained hearing, work under excellent acoustic conditions, and are able to concentrate fully during the experiment. 73 Jolita Urbanavičienė, Vytautas Kardelis Table 1.Evaluation of the auditory-experiment data from the Adutiškis point using the u criterion9 Word pair n₁ n p (%) 95% conf. int. u ≥ uₐ P Difference kiė:cis: kiė:cis 44 70 62,9 50,4 ÷ 74,2 2,18 > 1,96 P < 0,05 sufficient škī: rū̃s: škī: rū̃s 47 70 67,1 54,8 ÷ 78,0 2,93 > 2,58 P < 0,01 fairly large kã:še.: kã:še. 51 70 72,9 60,8 ÷ 82,9 3,97 > 3,29 P < 0,001 large rū̃:ksta: rū̃:ksta 59 70 84,3 73,5 ÷ 92,0 3,48 > 3,29 P < 0,001 very large Total 201 280 71,8 66,1 ÷ 77,0 7,54 > 3,29 P < 0,001 large The results show that the value of the u criterion exceeds the critical value in every case. Even the least successfully identified pair, kiė:cis: kiė:cis, has a high percentage of correct responses (62.9%). Correct responses for the other pairs exceed the 67% threshold. This recognition rate indicates that more than 2/3 of responses were correct. For the pair [škī:r̃us: škī:r̃us], the u criterion indicates a fairly large difference (P < 0,01). The pair [kã:še.: kã:še.] was correctly identified by 72.9% of listeners, while the u criterion registered a statistically large difference (P < 0,001). Listeners distinguished the pair [rū̃:ksta: rū̃:ksta] best: 84.3% of responses were correct, the confidence interval's upper bound reached as high as 92%, and the value of the u criterion exceeded the critical value (P < 0,001); the probability that a difference exists is very high. The overall results for Adutiškis listeners reached 71.8%: of the 280 pairs presented, 201 were correctly identified. Exceeding the 67% threshold provides evidence that leaves little room for doubt (Girdenis 2003: 63). 7. SVIRKOS SITE Unlike the Adutiškis listeners, the Svirkos listeners were least successful at identifying the accentual patterns of the words [rū̃:ksta] and [kʰu:šˈe.]—correct answers accounted for 61.4% (see Table 2). The value of the u criterion does not exceed the critical value (P > 0.05), but in this case too, the probability of a difference is at least 90%. 9 Symbol meanings: n1 – number of correct answers, n – total number of answers, p – percentage of correct answers, 95% p. int. – 95% confidence interval (minimum and maximum values in percent), ua – critical value of the u criterion, P – probability of exceeding the 50% threshold. 74 Rytų aukštaičių vilniškių ilgųjų balsių priegaidžių audicinis tyrimas Table 2. Evaluation of the auditory experiment data for the Svirkos site using the u criterion Word pair n₁ n p (%) 95% confidence interval u ≥ uₐ P Difference rū-ksta: rū-ksta 43 70 61.4 49.0 ÷ 72.9 1.93 < 1.96 P > 0.05 insufficient kʊ̃š.ė: kʊ̃š.ė 43 70 61.4 49.0 ÷ 72.9 1.93 < 1.96 P > 0.05 insufficient škíːr̥us: škíːr̥us 47 70 67.1 54.8 ÷ 78.0 2.93 > 2.58 P < 0.01 rather large kíeːcis: kíeːcis 55 70 78.6 67.0 ÷ 87.6 5.09 > 3.29 P < 0.001 large Total 188 280 67.1 61.3 ÷ 72.6 5.86 > 3.29 P < 0.001 large In the other cases, correct answers account for more than 2/3 of all responses, i.e. exceed the 67% threshold. The listeners correctly identified 47 pairs of the word [škíːr̥us – škíːr̥us] (67.1%) and pairs of the word [kíeːcis – kíeːcis] (78.6%). In the first case, the u criterion indicates a fairly large difference (P < 0.01); in the second, an unequivocally large difference was recorded (P < 0.001). Summing the results showed that 145 of the 210 words presented were correctly recognized (69.1%). The recognition percentage of the Svirkos listeners was slightly lower than that of the Adutiškis group (71.8%), but in both cases the probability that a difference exists is very high (P < 0.001). It is likely that the accentual patterns of the polyphthong [ieː] helped listeners identify differences in the degree of diphthongization—the circumflex [iẽː] is more pronounced than [íeː]. Svirkos listeners also distinguished pairs with the front vowels [i], [ie] better than words with the back vowels [uː], [uɔː]]. The Adutiškis listeners, however, showed the opposite pattern: they distinguished the accentual patterns of high-frequency sounds better than those of vowels in the low-frequency range. Thus, these experiments do not indicate that the distinctiveness of accent contrasts is related to the syllable's phonetic basis. 8. VIDUTINĖS SITE Like the residents of Adutiškis, the representatives of this site were least successful at distinguishing pairs with stressed front vocalism [škíːr̥us, kíeːcis], and best identified the accentual patterns of stressed back vowels [kɔːše, ruːksta] (see Table 3). 75 JOLITA URBANAVIČIENĖ, VYTAUTAS KARDELIS Table 3. Evaluation of the auditory experiment data for the Vidutinė locality using the u criterion Word pair | n1 | n | p (%) | 95% confidence interval | u≥uα | P | Difference ški·rруs: ski·rруs | 40 | 70 | 57,1 | 44,7 – 69,0 | 1,20 < 1,96 | P > 0,05 | insufficient kiė·cis: kiė·cis | 42 | 70 | 60,0 | 47,6 – 71,6 | 1,68 < 1,96 | P > 0,05 | insufficient kɔ́ šė.: kɔ̌ šė. | 48 | 70 | 68,6 | 56,3 ÷ 79,3 | 3,18 > 2,58 | P < 0,01 | rather large rū·ksta: rū·ksta | 52 | 70 | 74,3 | 62,3 ÷ 84,1 | 4,24 > 3,29 | P < 0,001 | large Total | 182 | 280 | 65,0 | 59,1 ÷ 70,6 | 5,10 > 3,29 | P < 0,001 | large After listening to the pairs [ški·rруs: ski·rруs], the listeners correctly identified 57.1% of cases, while for the pairs [kié·cis: kiė·cis] they recognized 60.0% of cases. In both cases, the u criterion does not exceed the critical value, and statistically the difference between the pairs is not significant (P > 0,05). The pairs [kɔ́ šė.: kɔ̌ šė.] and [rū·ksta: rū·ksta] were distinguished more reliably: 68.6% and 74.3% of responses were correct, respectively. The lower limit of the confidence interval exceeds 50%, while the upper limit reaches 79–84%. For the pair [kɔ́ šė.: kɔ̌ šė.], a fairly large difference was recorded (P < 0,01), while the auditory difference between the words [rū·ksta: rū·ksta] is unequivocally large (P < 0,001). The overall results of the auditory experiment in Vidutinė are close to the data from the Svirkos listeners. However, in terms of recognition of specific words, the responses from Vidutinė participants match those of the Adutiškis participants; that is, words with stressed [u·], [ɔ·] are recognized best. 9. LINKMENYS LOCALITY The aggregation of the results from listeners in Linkmenys revealed that recognition of pairs containing stressed back vowels varies considerably (see Table 4). After listening to the pairs [rū·ksta: rū·ksta], 80.0% of cases were correctly identified. By contrast, analysis of the word pairs [mɔ́·jus: mɔ̄·jus] yielded only 55.7% correct responses, indicating that the difference between random guessing and actual recognition is insufficient. The distinction between the tones of stressed front vowels ranges from 65.7% to 67.1%. For both [ški·rруs: ski·rруs] and [kiē·ti.s: kiē·ti.s], the value of the u criterion exceeds the critical value, and the recorded difference is fairly large (P < 0,01). 76 Rytų aukštaičių vilniškių ilgųjų balsių priegaidžių audicinis tyrimas Table 4. Evaluation of the auditory experiment data for the Linkmenys locality using the u criterion Word pair | n₁ | n | p (%) | 95% confidence interval | u ≷ uα | P | Difference mɔ̄jus: mɔ̄jus | 39 | 70 | 55,7 | 43,3 ÷ 67,7 | 0,96 < 1,96 | P > 0,05 | insufficient skī̄rɔus: skī̄rɔus | 46 | 70 | 65,7 | 53,3 ÷ 76,8 | 2,67 > 2,58 | P < 0,01 | rather large kiɛ·tɪ.s: kiɛ·tɪ.s | 47 | 70 | 67,1 | 54,8 ÷ 78,0 | 2,93 > 2,58 | P < 0,01 | rather large rū̄ksta: rū̄ksta | 56 | 70 | 80,0 | 68,6 ÷ 88,7 | 5,38 > 3,29 | P < 0,001 | large Total | 188 | 280 | 67,1 | 61,3 ÷ 72,6 | 5,86 > 3,29 | P < 0,001 | large Despite these variations, the overall results for Linkmenys listeners are better than the data for the Vidutinė area and coincide with the overall results for the Svirkos listeners. Correct responses make up 67.1% of all responses, so random guessing can be ruled out. 10. GENERAL RESULTS Analysis of the data from all locations yielded no unequivocal answers. Even within a single location, listeners’ responses range from an insufficient distinction, hovering on the borderline of guesswork, to a result that leaves no room for doubt. Nevertheless, certain patterns can be discerned. Table 5. Assessment of auditory experiment data by location according to the u criterion Location | n₁ | n | p (%) | 95% confidence interval | u ≷ uα | P | Adutiškis contrast | 201 | 280 | 71,8 | 66,1 – 77,0 | 7,54 > 3,29 | P < 0,001 | large Svirkai | 188 | 280 | 67,1 | 61,3 ÷ 72,6 | 5,86 > 3,29 | P < 0,001 | large Linkmenys | 188 | 280 | 67,1 | 61,3 ÷ 72,6 | 5,86 > 3,29 | P < 0,001 | large Vidutinė | 182 | 280 | 65,0 | 59,1 ÷ 70,6 | 5,10 > 3,29 | P < 0,001 | large The representatives of Adutiškis, the location farthest east in the dialect area, distinguished the pitch accents best by ear (71.8% correct answers). The listeners from Vidutinė, in an area closer to the Southern Aukštaitian dialect, identified the presented pairs least successfully (65.0% correct answers). Slightly less successfully than the people of Adutiškis, 77 Jolita Urbanavičienė, Vytautas Kardelis but better than the representatives of Vidutinė, the listeners from the Svirkai and Linkmenys groups recognized the pitch accents. Interestingly, the overall percentage of positive responses (67.1%) is close to the results of a similar experiment among the Utena dialect speakers—68.2% (see Kosienė 1982: 68). Thus, the hypothesis put forward at the beginning of the article—that locations closer to Utena and along the boundary between the Utena and Vilnius dialect areas might level pitch accents to a greater extent—is unfounded. It can only be surmised that the auditory salience of pitch accents may be lower in the southern part of the Vilnius dialect area, closer to the Southern Aukštaitian dialect. Southern Aukštaitian speakers recognized only 50–56% of the presented pairs by ear (see Leskauskaitė 2004: 192–194); thus, one can speak of a tendency toward pitch-accent leveling and the absence of a phonologically significant distinction in the Southern Aukštaitian area10. 11. CONCLUSION A discussion of the experiment’s results allows several broader conclusions to be drawn. The results of auditory experiments depend on both the speakers and the listeners. Since speakers who were equivalent in all respects as dialect users were selected, it is likely that the listeners were chiefly responsible for determining the experimental results; the most important factors characterizing them were age, education, dialect use, and musical hearing. The influence of musical hearing on the study of pitch accents is evident, since a well-trained ear helps one perceive pitch modulations more acutely. Age and education also affected the experimental results: younger listeners (apparently because of more sensitive hearing) registered subtle pitch-accent differences more accurately than older participants. Listeners with higher levels of education are better able to concentrate and describe differences more precisely. But it was clearest that listeners who actively use the dialect and produce the phenomena under study in their own speech (regardless of age and education) perceive differences more keenly than those whose speech has been influenced by the standard language (the fact that listeners are best at distinguishing the features of their own speech has been noted by Clark, Yallop 2002: 232). Thus, all the factors listed above influenced the results of auditory experiments to a greater or lesser extent. 10 Compare also data from other dialects: Western Aukštaitian speakers identified two of the three pairs studied very well (69–76% correct answers, u > u0,001, see Kazlauskaitė 2002: 145). Eastern Aukštaitian speakers from Kupiškis gave 64–65% correct answers (see Eidukaitienė 1977: 23). 78 Rytų aukštaičių vilniškių ilgųjų balsių priegaidžių audicinis tyrimas When comparing the tones of syllables whose nucleus consists of monophthongs and polyphthongs, no evidence was found that the distinctiveness of the tones depended on the syllable’s phonetic nucleus. Auditory examination of the tones in the Adutiškis and Vidutinė areas gave the impression that tones on low-frequency sounds were distinguished better than those on vowels in the high-frequency range. However, data from Svirkos and Linkmenys did not confirm this assumption: speakers from Svirkos identified words with front vowels [skiːr̃us, kieːcis] better, whereas speakers from Linkmenys did so with [uː] and [ieː]. Quite different results would probably be obtained by examining the tones of diphthongs and mixed diphthongs (cf. Pakerys 1982: 156). The characteristics of the phonological system may also affect the experimental results. For example, speakers from Svirkos distinguished the pair [kiéːcis: kiẽːcis] better because the degree of diphthongization of the polyphthong [ieː] differs. Speakers from Adutiškis differentiated the pair [káːše.: kãːše.] by its differing labialization—the circumflex vowel is more strongly rounded. However, this factor—phonological characteristics—is not absolute either, since speakers from Svirkos did not distinguish the pair [kúːšė.: kũːšė.] on the basis of diphthongization, while representatives of the Vidutinė area did not recognize the pair [škiːr̃us: škiːr̃us] by the differing depalatalization of [r]. Yet representatives of three areas (excluding Svirkos) distinguished the pair [rúːksta: rũːksta] best, although it has no distinctive phonetic features. Having discussed the emerging trends, it should be emphasized once again that the article describes a subjective experiment whose data were analyzed using objective—statistical—methods. The possibility cannot be ruled out that the results might differ somewhat with other groups of listeners. Moreover, the samples in the experiments conducted are not large. The aim was less to investigate a single locality in depth than to compare data from various areas. Therefore, the conclusions of this experiment should also be regarded as hypotheses for further research: (1) No complete leveling of tones was recorded in the peripheral northeastern Aukštaitian Vilniškiai localities studied. Tone convergence can be discussed only in individual cases. (2) The results of the auditory experiment are determined by the combined effect of several listener-related factors: 1) use of the dialect; 2) age; 3) education; 4) musical hearing. (3) Tones are distinguished better by listeners who know the dialect well and themselves realize the tonal opposition in speech. The distinctive features of the areas 79 Jolita Urbanavičienė, Vytautas Kardelis have no effect on tone identification. The salience of long-vowel tones is also independent of the phonetic characteristics of the syllable nucleus11. LITERATŪRA Clark, J., Yallop, C. 2002: An Introduction to Phonetics and Phonology. 2nd edition. Oxford– Cambridge: Blackwell. DLKG – Dabartinės lietuvių kalbos gramatika. 5 leid., red. V. Ambrazas. Vilnius: Mokslo ir enciklopedijų leidybos institutas, 2005. Eidukaitiene, E. V. 1977: Kupiškėnų monoftongų priegaidės (audicinis tyrinėjimas). Kalbotyra 28 (1), 18–23. Garšva, K. 2004: Eksperimentinės komutacijos metodas dialektologijoje. In: Baltų ir kitų kalbų fonetikos ir akcentologijos problemos. Vilnius: VPU, 21–25. Gerullis, G. 1930: Litauische Dialektstudien. Leipzig: Markert & Peters Verlag. Girdenis, A. 1974: Prozodinės priegaidžių ypatybės šiaurės žemaičių tarmėje (trukmė, pagrindinis tonas, intensyvumas). In: Eksperimentinė ir praktinė fonetika. Vilnius: Vilniaus pedagoginis institutas, 160–198. [= Girdenis, A. 2000: Kalbotyros darbai 1, 284–307.] Girdenis, A. 1996: Energetinė šiaurės žemaičių tarmės priegaidžių fonetinės prigimties interpretacija. Baltistica 31 (1), 71–84. [= Girdenis, A. 2001: Kalbotyros darbai 3, 243–256.] Girdenis, A. 1998: Šiaurės žemaičių priegaidžių skiriamieji požymiai. Lietuvių kalba: Tyrimai ir tyrėjai. Kazimiero Jauniaus 150-osioms gimimo ir 90-osioms mirimo metinėms paminėti. Konferencijos pranešimų tezės, Vilnius: LKI; VPU, 37–38. [= Girdenis, A. 2001: Kalbotyros darbai 3, 403–404.] Girdenis, A. 2003: Teoriniai lietuvių fonologijos pagrindai. Vilnius: Mokslo ir enciklopedijų leidybos institutas. Harris, Z. S. 1963: Structural Linguistics. Chicago–London: Phoenix Books. 11 Straipsnio autoriai nuoširdžiai dėkoja dizaineriui Mantui Kelpšui už pagalbą rengiant spaudai žemėlapį. 80 Rytų aukštaičių vilniškių ilgųjų balsių priegaidžių audicinis tyrimas Hjelmslev, L. 1935: On the Principles of Phonematics. In: Proceedings of the 2nd International Congress of Phonetic Sciences. Cambridge: University Press, 49–54. Jensen, K. M. 1961: Tonemicity. Bergen–Oslo: Norwegian University Press. Kačiūskienė, G. 1983: Šiaurės panevėžiškių murmamųjų balsių fonetinės ypatybės, distribucija ir fonologinė interpretacija. Kalbotyra 34 (1), 14–23. Kardelytė, J. 1975: Gervėčių tarmė: Fonetika ir morfologija. Vilnius: Mintis. Kazlauskaitė, R. 2002: Pašūšvio šnektos fonologinė sistema: vokalizmas ir prozodija. Daktaro disertacija. Vilnius: VU. Kazlauskas, J. 2000: Rinktiniai raštai 1: Lietuvių kalbos istorinė gramatika. Vilnius: Mokslo ir enciklopedijų leidybos institutas. Kosienė, O. 1979: Priegaidžių opozicija dvigarsinėse uteniškių galūnėse. In: Ak-tualiosios kalbotyros problemos. Mokslinės konferencijos pakvietimas, programa ir tezės. Vilnius, 44–46. Kosienė, O. 1982: Rytų aukštaičių uteniškių monoftongų priegaidės. Kalbotyra 33 (1), 61–71. Kosienė, O., Girdenis, A. 1979: Fonologinis šalutinis kirtis rytų aukštaičių uteniškių tarmėje. Kalbotyra 30 (1), 48–56. [= Girdenis, A. 2000: Kalbotyros darbai 2, 141–152.] Kudirka, R. 2005: Lietuvių bendrinės kalbos monoftongų priegaidžių akustiniai požymiai. Acta Linguistica Lithuanica 52, 1–21. Labov, W. 1972: On the use of the present to explain the past. In: L. Heilmann, ed., Proceedings of the XII International Congress of Linguists, Bologna – Florence. Bologna: Il Mulino, 825–851. Leskauskaitė, A. 2004: Pietų aukštaičių vokalizmo ir prozodijos bruožai. Vilnius: Lietuvių kalbos institutas. LKA2 – Lietuvių kalbos atlasas 2: Fonetika, red. K. Morkūnas. Vilnius: Mokslas, 1982. 81 Jolita Urbanavičienė, Vytautas Kardelis LKG – Lietuvių kalbos gramatika 1: Fonetika ir morfologija, red. K. Ulvydas. Vilnius: Mintis, 1965. LKT – Lietuvių kalbos tarmės. Chrestomatija, sudarė E. Grinaveckienė, A. Jonait yte, K. Morkūnas, B. Vanagienė, A. Vidugiris. Vilnius: Mintis, 1970. Mikalauskaitė, E. 1975: Lietuvių kalbos fonetikos darbai. Vilnius: Mokslas. Mikulėnienė, D., Morkūnas, K. 1997: Dieveniškių šnektos tekstai. Vilnius: Mokslo ir enciklopedijų leidybos institutas. Pakerys, A. 1982: Lietuvių bendrinės kalbos prozodija. Vilnius: Mokslas. Urbach, V. J. 1975: Статистический анализ в биологических и медицинских исследованиях. Москва: Медицина. Urbanavičienė, J. 2005: Svirkų šnektos (rytų aukštaičių vilniškių) fonologinė sistema: vokalizmas ir prozodija. Daktaro disertacija. Vilnius: VDU, LKI. Zinkevičius, Z. 1966: Lietuvių dialektologija. Vilnius: Mintis. Zinkevičius, Z. 1974: Dėl akūto ir cirkumflekso skyrimo rytų Lietuvos tarmėse. Baltistica 10 (1), 93–94. Jolita Urbanavičienė Leksikografijos centras Lietuvių kalbos institutas P. Vileišio g. 5, LT-10308 Vilnius [email protected] Vytautas Kardelis Baltistikos ir bendrosios kalbotyros katedra Vilniaus universitetas Universiteto g. 5, LT-01513 Vilnius [email protected] 82