AGNĖ PUKEVIČIŪTĖ

Institute of the Lithuanian Language

Research fields: geolinguistics, dialectology, sociolinguistics, experimental phonetics.

EASTERN AUKŠTAITIAN OF PANEVĖŽYS: CLUSTER ANALYSIS IN SEARCH OF DIALECTICISM

Eastern Aukštaitian of Panevėžys: Cluster Analysis in Search of Dialectal Features

ANNOTATION

The article seeks to test the suitability of one of the statistical methods of cluster analysis for identifying similarities among the settlements of Eastern Aukštaitian of Panevėžys according to the frequency of use of distinctive dialectal features. The grouping patterns of the settlements that emerged during the analysis make it possible to develop a research model for measuring the dialecticity and peripheral status of the settlements under investigation.

For the study, 90-minute audio recordings of the Eastern Aukštaitian of Panevėžys subdialect were selected at random, collected from two central dialect sites—Pašvalys (112), Kapčiūnai (138)—and two border dialect sites—Bučiūnai (52) and Stumbriškis (171). While listening to the recordings, the frequencies of occurrence of 13 distinctive Panevėžys dialectal features were recorded and subsequently entered into the SPSS Statistics computer program, and a cluster analysis was performed.

The results obtained suggest that the cluster analysis model applied in the study is suitable for revealing patterns in the grouping of four Panevėžys LKA sites according to the variables examined. The analysis shows that LKA sites located on the periphery of the dialect area are not necessarily peripheral—in them, dialectal features are very well preserved (at the LKA sites of Bučiūnai, Pašvalys, and Kapčiūnai, strong features are better preserved, whereas at Stumbriškis, weak features are better preserved).

The study shows that when dialects are examined using the cluster analysis method, the selection of research material is very important.

KEYWORDS: Eastern Aukštaitian of Panevėžys, dialecticity of LKA sites, peripherality of LKA sites, cluster analysis.

Straipsniai / Articles

175

AGNĖ PUKEVIČIŪTĖ

ANNOTATION

This paper aims to test the appropriateness of one of the statistical methods – cluster analysis – to determine the similarity of the populated localities of the Eastern Aukštaitian of Panevėžys in accordance with the frequency of usage of distinctive dialectal features. Grouping peculiarities that emerged from the analysis of the research areas allow us to create a model to measure the dialecticism and the peripherality of the research areas. 90 minute long recordings of the Eastern Aukštaitian of Panevėžys collected from two central – Pasvalys (112), Kapčiūnai (138) – and two marginal – Bučiūnai (52), Stum̃briškis (171) – dialectal points were randomly selected for this study. While listening to the recordings, the frequencies of usage of 13 dialectal features of the Eastern Aukštaitian of Panevėžys were identified, which were then entered into the SPSS Statistics computer program and the cluster analysis was performed.

The results suggest that the cluster analysis model applied in the study is suitable for revealing the grouping patterns of the points of the Eastern Aukštaitian of Panevėžys LLA according to the research dimensions. The analysis indicates that the marginal LLA points are not necessarily peripheral – dialectal characteristics are perfectly maintained in them (in Bučiūnai, Pasvalys and Kapčiūnai points of LLA the strong features are better maintained, while the weak ones are better preserved in Stum̃briškis).

The study indicates that the selection of the research material is very important for the examination of dialects by the cluster analysis method.

KEYWORDS: Eastern Aukštaitian of Panevėžys, dialecticism of LLA points, peripherality of LLA points, cluster analysis.

1. OVERVIEW OF RESEARCH ON THE PANEVĖŽYS SUBDIALECTS

Interest in Lithuanian dialects began in the late nineteenth century. The Eastern Aukštaitian Panevėžys subdialects are mentioned in the very earliest works on Lithuanian dialectology. Leopoldas Geitleris (Geitler 1875)¹, Eduardas Volteris (Вольтеръ 1886; 1904: 325–335, 338–340, 345–354), Kazimieras Jaunius (Jaunius 1898)²: 114–161)), and Antanas Baranauskas (Барановский 1898) wrote about the distinctive features of the Panevėžys subdialects. Work using the same descriptive method continued in the early twentieth century—important contributions were made by Kazimieras Būga (1924)³, Antanas Salys (1930, 1946)⁴, Jurgis Gerulis (1930: 35–46, 54–65, 66–76), and Petras Būtėnas (1932: 168–193).

¹ Remiamasi Danguole Mikulėniene (2011: 19–27).

² See Jaunius 1970.

³ See Būga 1961: 82–83.

⁴ See Salys 1992: 114–119, 122–131.

176

Acta Linguistica Lithuanica LXXII

Rytų aukštaičiai panevėžiškiai:

klasterinė analizė ieškant tarmiškumo

works. The research tradition inherited from the Neogrammarians remained popular in the postwar period.

Later studies began examining the salient features of the Panevėžys subdialects, their distribution and similarity to other subdialects (Zinkevičius 1966; Girdenis, Zinkevičius 1966: 139–147; LKT 1970: 14, 32–36; LKA 1977, 1982, 1991; Ručinskaitė 1984: 72–76; Kačiūskienė 1984: 42–53; Kačiūskienė, Girdenis 1997: 31–36)⁵, as well as the reasons for their emergence (Kazlauskas 1968: 10–20; Zinkevičius 1974: 148; Garšva 1977: 76–87; Kačiūskienė, Girdenis 1982: 189–191), using experimental auditory, acoustic, and spectrographic methods.

By the end of the twentieth century, sociolinguistic features had begun to appear in research on the Panevėžys subdialects: researchers started taking speakers’ ages into account when evaluating the data. This is evident both in descriptive studies (Zinkevičius 1975: 85; Zinkevičius 1979: 92–93) and in experimental spectrographic and auditory studies (Kačiūskienė 1982: 39–45; Garšva 1982a: 68–70; Kačiūskienė 1983: 24–38). In this way, the subdialect use of Panevėžys speakers from different generations was compared.

Studies from the early twenty-first century began drawing on statistical, psychoacoustic, spectral, and oscillographic methods (used both in comprehensive works (Markevičienė 2001: 29–32; LKTCh 2004: 30–34, 184–194⁶; Zinkevičius 2006: 74–82) and in individual researchers’ studies (Kačiūskienė 2000: 25–31; Garšva 2002: 9–18; Kačiūskienė 2003: 88–92; Kačiūskienė 2005a: 49–53; Kačiūskienė 2005b: 499–503; Kačiūskienė 2009: 61–67; Kačiūskienė 2012: 53–62)). During this period, studies with sociolinguistic features also became more numerous: questionnaire methods were applied (Kačiūskienė 2005c), age (Kačiūskienė 2006; Meiliūnaitė 2009) and education (Kačiūskienė 2013a: 8) variables were introduced, and linguistic self-awareness was investigated (Kačiūskienė 2013b: 22).

The increasingly frequent incorporation of sociolinguistics into research on the Panevėžys subdialects shows that, amid the intensive process of dialect change taking place at the beginning of the twenty-first century, conventional methods are no longer sufficient to identify and explain this phenomenon.

2. METHODOLOGY OF DIALECTALITY RESEARCH

As dialects change, so does the geolinguistics that studies them—it has become interdisciplinary: the usual aims of dialectologists are supplemented by sociolinguistics (for example,

⁵ The results of this period were also published in other publications. Worth mentioning are Kazimieras Garšva’s Texts of the Joniškėlis-area dialect (1982b).

⁶ The dialectal speech was recorded by Aldona Jonaitytė, Stasys Tumėnas, D. Kazilionytė, S. Liolytė, Kazys Morkūnas, Aloyzas Vidugiris, and Edmundas Trumpa, and transcribed by Aloyzas Vidugiris, Asta Leskauskaitė, and Rima Bacevičiūtė.

Straipsniai / Articles

177

AGNĖ PUKEVIČIŪTĖ

sociolinguistic variables (age, gender, education, etc.) are taken into account. As geolinguistics acquired new features, the theoretical approaches and methodological principles of dialectology at the beginning of the 21st century began to be described by the term sociogeolinguistics (Aliūkaitė 2011: 257; Mikulėnienė, Meiliūnaitė 2014).

In sociogeolinguistic research, linguistic phenomena can be examined not only using the instruments of these two disciplines, but also using methods from other fields of science. The aim of this article is to test the suitability of one cluster analysis method used in statistics for determining similarities among the settlements of the Eastern Aukštaitian Panevėžiškiai dialect area based on the frequency of use of distinctive dialect features⁷. The subject of the research is the characteristic features of the Panevėžiškiai subdialect.

The statistical cluster analysis performed will make it possible to verify how the locations under study cluster according to the frequency of use of distinctive dialect features. In this way, a research model is created to measure the dialectality and peripherality of the speech of the locations studied: by selecting four LKA points (two central and two peripheral), the study determines how well dialect features have been preserved there and whether the model used is suitable for doing so.

In this article, the terms central and peripheral LKA point are used in a geographical sense: a location geographically closer to the center of the Eastern Aukštaitian Panevėžiškiai area is called central, while a settlement near the boundaries of other subdialects is called peripheral. Consequently, from a linguistic perspective it may have become peripheral (transitional, i.e., it may have adopted features of other languages or distinctive dialect features of other subdialects)⁸.

The research material consists of randomly selected 90-minute audio recordings of the Eastern Aukštaitian Panevėžiškiai subdialect. They were collected in 1992, 2004, and 2012 in locations that had already been studied 60 years earlier (see LKA 1977: 21–23, LKA 1982: 9–11, LKA 1991: 9–11)—Bučiūnai (52) (recording No. LK259701, compact disc registration No. CD 1496 RAP 116), Pąsvalys (112) (recording No. LK204602, compact disc registration No. CD 1508 RAP 122), Kapčiūnai (138) (recording No. LK029101, compact disc registration No. CD 1156 RAP 79), and Stūm̃briškis⁹ (171)¹⁰.

⁷ The Eastern Aukštaitian Panevėžiškiai are in the western part of the Eastern Aukštaitian region. They border the Western Aukštaitian Šiauliškiai, Eastern Aukštaitian Kupiškėnai, Anykštėnai, and Širvintiškiai (see Figure 1).

⁸ Apie centrinio ir periferinio LKA punkto terminus plačiau žr.: Morkūnas 1993: 4–7; Kardelis 2014; Kardelis, Stakutytė 2015).

⁹ A registration number has not yet been assigned to this recording; it is stored on the server of the Institute of the Lithuanian Language together with other data collected in 2012.

¹⁰ The recordings were gathered in 2007–2013 as part of the project Contemporary Geolinguistic Research in Lithuania: Optimization of the Point Network and Interactive Dissemination of Dialect Information (see Lithuanian Dialects).

178

Acta Linguistica Lithuanica LXXII

Rytų aukštaičiai panevėžiškiai: klasterinė analizė ieškant tarmiškumo

The dialectal features used in the similarly paced speech of three women (D. N., born 1926, living in Bučiūnai; J. L., born 1934, living in Pasvalys¹¹; V. M., born 1909, living in Klovainiai¹²) and one man (B. N., born 1946, living in Stumbriškis) are examined experimentally¹³.

As shown in Figure 1, the localities under study lie within the Aukštaitian area, and therefore share the distinguishing feature of all Aukštaitian dialects: stressed diphthongs ie, uo are retained intact, as in Standard Lithuanian, unlike in the Samogitian area (Zinkevičius 1966: 13–14; LKT 1970: 15; LKTCh 2004: 29). The LKA survey points belong to the Eastern Aukštaitian area, since in this subdialect the diphthongs an, am, en, em and former nasal vowels ą, ę are narrowed and pronounced as ɔn, ɔm, ɛn, ɛm and o, e (Zinkevičius 1966: 15, 83–84, 99–101; LKT 1970: 32–33; LKTCh 2004: 30, 173–175)¹⁴.

The main distinguishing feature of the Panevėžys subdialect is the merger of short ending vowels, known as vowel reduction. In terms of the degree of reduction, the LKA survey points under study belong to the Northern Panevėžys subdialect, whose short ending vowels have become indistinct murmured sounds, unlike those of the southern area, where short a, u in endings merge into a single sound similar to u, but unrounded, transcribed as v (Zinkevičius 1966: 92, 119–121; LKT 1970: 35; LKTCh 2004: 175–176).

In this study, the degree of dialectalness of the localities under investigation is measured by examining the productivity of six vowels (o, ė, i, u, y, ū) and two diphthongs (ie, uo), which differ in stress or occur in different parts of words. The frequency of stress retraction is also recorded. Two features are used distinctively only in the Northern Panevėžys area.

In unstressed position, unlike speakers of the southern area, Northern Panevėžys speakers realize diphthongs ie, uo as e, o (Zinkevičius 1966: 16, 87; LKTCh 2004: 176). Stress is also retracted exclusively in the localities under study: in the Northern Panevėžys area, the general stress-retraction rule applies, whereby stress is retracted not only from a short but also from a long circumflex ending, and not only to a long but also to a short syllable; intensive conditional stress retraction also applies, whereby stress is retracted from

¹¹ This recording captures the speech of several women. However, one informant speaks for most of the recording, so the study treats the speech examined as that of this one woman.

¹² This informant lives in Klovainiai, but the recording is assigned to the Kapčiūnai (138) survey point in the database.

¹³ Women make up a larger percentage of informants than men, and accordingly, this study examines more recordings of women’s than of men’s speech.

¹⁴ These and other features distinguishing the Panevėžys subdialect from others were noted as early as the late nineteenth century: Volteris wrote about pronunciation differences across various parts of the country involving stressed stem ie, uo, diphthongs an, en, short vowels i, u, stressed stem o, and other features (Вольтеръ 1886: 1–12).

Straipsniai / Articles

179

AGNĖ PUKEVIČIŪTĖ

trumpos galūnės į priešpaskutinį bet kokio ilgumo skiemenį (Zinkevičius 1966: 37–40, LKT 1970: 34–35, LKTCh 2004: 30–31)¹⁵.

Other features under investigation are more universal. These include the aforementioned stressed stem ie, uo, whose quality in the area under study does not differ from that of other Highlanders or the standard language.

Like all Eastern Highlanders, Northern Panevėžys speakers lengthen stressed (not stress-retracted) short u, i outside the word ending and pronounce them very close (Zinkevičius 1966: 107; LKT 1970: 34; LKTCh 2004: 177). In pronouncing unstressed stem y, ū, the reverse occurs: they are converted into the corresponding short i, u (Zinkevičius 1966: 110; LKT 1970: 34; LKTCh 2004: 176).

However, unlike in many Highland dialects, the stressed stem o in the Panevėžys area is pronounced without an a coloring, but in an unstressed ending it becomes a (Zinkevičius 1966: 70). Stressed stem ė is generally retained (Zinkevičius 1966: 74), whereas unstressed final ė is pronounced e (Zinkevičius 1966: 75; LKTCh 2004: 176–178).

Thus, the study analyzes the frequency of seven strong or salient (stressed stem ie, uo, o, ė, i, u, stress retraction) and six weak or inconspicuous (unstressed stem y, ū, ie, uo, unstressed word-final ė, o) distinctive dialectal features of Panevėžys speech (see Table 1)¹⁶. These features are frequent and, as can be seen from the table of their frequency of use, occur many times in Panevėžys speech. This is also demonstrated by recent studies of this subdialect (see Aliūkaitė et al. 2014: 188–191).

The material collected from the dialect localities under study was processed in several stages: 1) audio recordings collected at various times from the LKA survey points under study were listened to; 2) the recordings were analyzed: while listening, the frequency of occurrence of 13 distinctive Panevėžys dialect features was recorded; 3) the frequencies were entered into the SPSS Statistics software, and cluster analysis was performed¹⁷.

Applying the cluster analysis method, the similarity of the selected four LKA survey points in the Eastern Highland Panevėžys subdialect was established, and they were grouped

¹⁵ As a rule, stress is not retracted from the ending in comparative adverbs, oxytone pronouns, and third-person future-tense forms (Zinkevičius 1966: 38); in polysyllabic words, stress on the ending is retracted to the nearest syllable (Zinkevičius 1966: 39), but not to the prefix (LKTCh 2004: 180).

¹⁶ The main criteria for distinguishing strong (or salient) and weak (or inconspicuous) features are their area of distribution, recognizability, and leveling: strong features are characteristic of a relatively small area of a given subdialect, are quickly noticed, and are easily leveled, whereas weak features are characteristic of many related subdialects and are harder to notice, which is why speakers are less inclined to modify them (Aliūkaitė, Mikulėnienė 2014: 46–47).

¹⁷ Apie programą plačiau žr.: IBM SPSS Statistics; Vaitkevičius, Saudargienė 2006.

180

Acta Linguistica Lithuanica LXXII

Eastern Aukštaitians of Panevėžys: Cluster Analysis in Search of Dialectal Features

1 PAV. Tiriamosios rytų aukštaičių panevėžiškių vietovės (Mikulėnienė, Meiliūnaitė 2014: 189)

Straipsniai / Articles

181

AGNĖ PUKEVIČIŪTĖ

into groups of similar objects—clusters. The data under study were grouped so that the differences within a cluster formed from the localities studied would be as small as possible, and those between clusters as large as possible (Čekanavičius, Murauskas 2002: 195; Vaitkevičius, Saudargienė 2006).

TABLE 1.

Frequency of occurrence of the distinguishing dialectal features of the Panevėžys subdialect in the localities studied¹⁸

STRENGTH

FEATURE

BUČIŪNAI Recording duration 43 min. 8 sec.

PASVALYS Recording duration 46 min. 42 sec.

KAPČIŪNAI Recording duration 28 min. 28 sec.

STUMBriškis Recording duration 62 min. 11 sec.

Strong

Stressed stem ie is retained

~ 79

~ 55

~ 28

~ 66

Strong

Stressed stem uo is retained

~ 42

~ 16

~ 21

~ 24

Strong

Stressed stem ė is retained

~ 59

~ 67

~ 50

~ 70

Strong

The stressed stem vowel o is retained

~ 65

~ 87

~ 40

~ 63

Strong

The stressed stem vowel u is lengthened

~ 70

~ 50

~ 47

~ 87

Strong

The stressed stem vowel i is lengthened

~ 63

~ 70

~ 50

~ 83

Stress retraction

Stress retraction

~ 88

~ 90

~ 68

~ 137

Weak

Unstressed stem uo becomes o

~ 2

~ 3

~ 1

~ 4

Weak

Unstressed stem ie becomes ę

~ 7

~ 13

~ 4

~ 12

Weak

Unstressed stem y becomes i

~ 15

~ 12

~ 10

~ 25

Weak

Unstressed stem ū becomes u

~ 4

~ 1

0

~ 7

Weak

In an unstressed ending, ė is rendered as e

~ 40

~ 28

~ 8

~ 41

Weak

In an unstressed ending, o is rendered as a

~ 121

~ 78

~ 86

~ 157

¹⁸ The recordings were listened to by a researcher who was a native speaker of the dialect. Since the frequency was recorded by listening to audio recordings of spontaneous spoken language, the data presented may contain errors due to human factors.

182

Acta Linguistica Lithuanica LXXII

Rytų aukštaičiai panevėžiškiai: klasterinė analizė ieškant tarmiškumo

The study experimentally tested two clustering methods—hierarchical and non-hierarchical (two-step). They are distinguished by how their similarity measures and criteria for determining the distance between clusters are selected, as well as by their clustering strategies. The main difference is that when the research data are grouped using the two-step cluster analysis method, the number of clusters is known (selected) in advance (Čekanavičius, Murauskas 2002: 202).

When analyzing the results of the cluster analysis, it should be taken into account that the study examines informants’ spontaneous speech rather than uniform texts, that speaking conditions differ, and that the sample of research data is very small; therefore, the cluster analysis results may not be entirely accurate. However, given the research material, this study aims only to test whether this method can reveal any patterns in how localities cluster according to the frequency of dialectal features used. More specifically, it tests whether cluster analysis can identify localities tending toward the periphery on the basis of the frequency of dialectal features used, and show the similarities among localities that have best preserved dialectal features, taking into account various factors, such as geographical distance, the peripheral versus central position of localities within the boundaries of the Eastern Aukštaitian Panevėžys subdialect, and sociogeolinguistic differences.

3.

CLUSTER ANALYSIS OF THE DISTINCTIVE FEATURES OF THE PANEVEZYS SUBDIALECT

To test how the study localities cluster according to the frequency with which 13 main distinctive features of the Panevėžys subdialect are used (see Table 1), one hierarchical and one non-hierarchical method were employed: hierarchical (agglomerative) and two-step cluster analyses were performed using the SPSS Statistics software. SPSS Statistics

Because of the small amount of data, the minimum number of clusters—2—was selected in both cases. The hierarchical cluster analysis results show that the frequency of use of distinctive dialectal features is most similar in Bučiūnai, Pasvalys, and Kapčiūnai; these LKA survey points therefore formed one cluster. The other group of similar objects consists of Stum̃briškis (see Table 2).

TABLE 2.

Results of hierarchical cluster analysis by frequency of use of 13 distinctive features of the Eastern Aukštaitian Panevėžys subdialect

Cluster No.

LKA survey point included in the cluster

1

Bučiūnai, Pasvalys, Kapčiūnai

2

Stum̃briškis

Straipsniai / Articles

183

AGNĖ PUKEVIČIŪTĖ

Since the first cluster consists of one peripheral (Bučiūnai) and two central (Pasvalỹs, Kapčiūnai) LKA sites, it can be argued that the distinguishing dialectal features may be preserved equally well in the research locations, regardless of their geographical position. On the other hand, data from the settlement of Stum̃briškis, located at the edge of the Panevėžys subdialect area and least similar to the other three research LKA sites, show that characteristic Panevėžys subdialect features may be less well preserved at a peripheral LKA site.

The dendrogram visually shows the sequence in which the clusters were joined (see Figure 2). It shows that the peripheral and central LKA sites are most similar: in the residential areas of Bučiūnai and Pãsvalis, the productivity of use of the distinguishing Panevėžys subdialect features is similar. Fewer of these features occurred in the speech of the informant living near Kapčiūnai. The peripheral LKA site of Stum̃briškis stands out most: the level of dialecticity in this settlement is least similar to that of the other three research locations.

² FIG. Dendrogram of cluster joining (1 Bučiūnai, 2 Pasvalỹs, 3 Kapčiūnai, 4 Stum̃briškis)

These hierarchical cluster analysis results may have been determined by various sociogeolinguistic factors associated with the LKA sites: unlike the other three, the village of Stum̃briškis is closest to the major county city of Panevėžys (the main center of urbanization), so, compared with the other LKA sites, the standard language may influence the subdialect, resulting in a lower frequency of use of dialectal features. The results may also have been affected by the fact that Stum̃briškis is a peripheral LKA site—it borders the Kupiškėnai area, and some residents of this village are non-local and speak a semi-dialectal form of speech¹⁹.

¹⁹ The possibility that the results were affected by the longest audio recording from the Stum̃briškis LKA site is ruled out, since comparison of data from this LKA site with the frequency of use of dialectal features recorded in the other locations yields fairly similar results. In addition, the recording was made with an informant who was not alone—researchers were continually asking him questions.

184

Acta Linguistica Lithuanica LXXII

Rytų aukštaičiai panevėžiškiai: klasterinė analizė ieškant tarmiškumo

These arguments suggest that the Stum̃briškis LKA site may be peripheral: alongside the characteristic features of the Eastern Aukštaitian Panevėžys subdialect, elements of other subdialects (Utena and Kupiškėnai) or of the standard language may also be used in the settlement under study, which accounts for its formation of a separate group. Figure 1 shows that Bučiūnai is also a peripheral LKA site—it borders the Šiauliai area. However, as the hierarchical cluster analysis results show, there is no doubt about the vitality of its distinguishing Panevėžys subdialect features: according to the research data, this village is most similar to Pãšvalis, whose subdialect, according to the latest research, is a strong Western Eastern Aukštaitian geolect (Aliūkaitė, Mikulėnienė 2014: 261)²⁰. Thus, dialectal features are preserved equally well in Bučiūnai, located at the edge of the Panevėžys subdialect area, and in a settlement in the center of the dialect area. Consequently, unlike Stum̃briškis, there is no basis for calling the peripheral Bučiūnai LKA site peripheral in the linguistic sense.

The two-step cluster analysis yielded somewhat different results: the frequency of use of distinguishing dialectal features is most similar in Bučiūnai, Pasvalys, and Stum̃briškis; these LKA sites form one group of similar objects, while Kapčiūnai forms another (see Table 3).

TABLE 3.

Results of the two-step cluster analysis according to the frequency of use of 13 distinguishing features of the Eastern Aukštaitian Panevėžys subdialect

Cluster No.

LKA site assigned to the cluster

1

Bučiūnai, Pasvalỹs, Stum̃briškis

2

Kapčiūnai

Unlike the previously obtained data, however, this time one central (Pasvalys) and two peripheral (Bučiūnai, Stum̃briškis) LKA sites cluster together, while the second cluster consists not of a peripheral but of a central settlement (Kapčiūnai). The results raise doubts: they indicate that strong and weak dialectal features may be better preserved at peripheral LKA sites than at a central one.

Table 4 shows that the means for all features used at the LKA sites forming the first cluster are higher than those for the second. This would mean that dialecticity is better preserved at the LKA sites in the first cluster than in the second—that the frequency of use of all the features investigated is higher in Bučiūnai, Pasvalys, and Stum̃briškis than in Kapčiūnai. One possible cause for doubt is that in the first cluster almost all the data are highly dispersed (the standard deviation ranges from

²⁰ A geolect is defined as “[…] a non-homogeneous (heterogeneous) regional language variety whose area has no very clear geographical boundaries. From a linguistic perspective, it is a continuation of the traditional dialect, retaining a certain set of old dialectal features” (Aliūkaitė, Mikulėnienė 2014: 41).

Straipsniai / Articles

185

AGNĖ PUKEVIČIŪTĖ

TABLE 4: Cluster centers according to 13 distinguishing features

Cluster | ie Mean 66.67; Std. dev. 12.014 | 28.00; · | 57.00; 21.679 | uo Mean 27.33; Std. dev. 13.317 | 21.00; · | 25.75; 11.325 | o Mean 71.667; Std. dev. 13.317 | 40.000; · | 63.750; 19.207 | ė Mean 65.333; Std. dev. 5.686 | 50.000; · | 61.500; 8.963 | i Mean 72.000; Std. dev. 10.149 | 50.000; · | 66.500; 13.772 | u Mean 69.000; Std. dev. 18.520 | 47.000; · | 63.500; 18.699 | Retracted accent Mean 105.00; Std. dev. 27.731 | 68.00; · | 95.75; 29.239 | Unstressed endings | Cluster | ie Mean 10.67; Std. dev. 3.215 | 4.00; · | 9.00; 4.243 | uo Mean 3.000; Std. dev. 1.000 | 1.000; · | 2.500; 1.291 | y Mean 17.333; Std. dev. 6.807 | 10.000; · | 15.500; 6.658 | ū Mean 4.000; Std. dev. 3.000 |,000; · | 3.000; 3.162 | ė Mean 36.33; Std. dev. 7.234 | 8.00; · | 29.25; 15.349 | o Mean 118.667; Std. dev. 39.552 | 86.000; · | 110.500; 36.189 | 1 2 Overall 1 2 Overall

186

Acta Linguistica Lithuanica LXXII

Rytų aukštaičiai panevėžiškiai: klasterinė analizė ieškant tarmiškumo

from low (1) to very high—even 39.552). Perhaps this can be explained by the fact that the values examined in the study are quite large. And the larger the mean values, the larger the standard deviations.

The lowest frequencies of features used in the locality of Kapčiūnai (see Table 1), which, of course, correspond to the means presented (unstressed stem ū—0 occurrences, unstressed stem uo—1, unstressed stem ie—4), show that the aforementioned weak features are unproductive at this LKA site and are poorly retained in the speech of the informants. One might think that for this reason they are likely to disappear and be replaced by other variants. However, these data differ little from the mean frequencies of the features used at the three LKA sites that make up the other cluster (4, 3, and 10.67, respectively). Therefore, it is assumed that unstressed stem ū, uo, ie occur rarely in the speech of informants living in all four localities, but have been retained.

The study does not rule out the possibility that applying different cluster analysis methods to the same data may yield different results (Čekanavičius, Murauskas 2002: 196). However, it is assumed that such results may have been due to differences in the duration of the sound recordings examined—for this reason, fewer of the features under analysis occurred in the shortest recording from the Kapčiūnai LKA site than in the other three. This may also have affected the low mean frequencies of use of the distinguishing features obtained for this settlement, indicating a low level of dialecticism in this village.

To substantiate the assumption, the frequency of features used in Kapčiūnai was artificially normalized (in the SPSS Statistics program, the frequency of distinguishing dialect features used at this LKA site was replaced with the mean of all frequencies at the other three LKA sites), and the calculations were repeated. Once this had been done, the results matched the data obtained through hierarchical analysis (see Table 5).

TABLE 5.

Results of two-step cluster analysis based on the frequency of use of 13 distinguishing features of the Eastern Aukštaitian Panevėžys subdialect, after artificial normalization of the frequency of features used in Kapčiūnai

Cluster No.

LKA site included in the cluster

1

Bučiūnai, Pasvalys, Kapčiūnai

2

Stumbrškis

Comparing the results of grouping the LKA sites under study obtained by the two methods reveals a tendency: on the basis of similarity in the frequency of dialect feature use, a cluster is never made up exclusively of central LKA sites while another consists exclusively of peripheral ones. This confirms the view that in the Eastern Aukštaitian Panevėžys area

Straipsniai / Articles

187

AGNĖ PUKEVIČIŪTĖ

Cluster centers according to 13 distinguishing features, features of the Panevėžys subdialect, after artificially normalizing the frequency of features used in Kapsiūnai

Cluster ie Mean Std. dev. n. 67,00 12,000 66,00 · 66,75 9,811 uo Mean Std. dev. n. 28,33 13,051 24,00 · 27,25 10,874 o Mean Std. dev. n. 74,67 11,240 63,00 · 71,75 10,874 ė Mean Std. dev. n. 63,67 4,163 70,00 · 65,25 4,646 i Mean Std. dev. n. 68,33 4,726 83,00 · 72,00 8,287 u Mean Std. dev. n. 63,00 11,269 87,00 · 69,00 15,122 Stress retraction Mean Std. dev. n. 94,33 9,292 137,00 · 105,00 22,642 1 2 Total Unstressed stem ie Mean Std. dev. n. 10,33 3,055 12,00 · 10,75 2,630 uo Mean Std. dev. n. 2,67,577 4,00 · 3,00,816 y Mean Std. dev. n. 14,67 2,517 25,00 · 17,25 5,560 ū Mean Std. dev. n. 3,00 1,732 7,00 · 4,000 2,449 ė Mean Std. dev. n. 34,67 6,110 41,00 · 36,25 5,909 o Mean Std. dev. n. 106,00 24,269 157,00 · 118,75 32,294 1 2 Total

188

Acta Linguistica Lithuanica LXXII

Rytų aukštaičiai panevėžiškiai: klasterinė analizė ieškant tarmiškumo

Dialecticism can be preserved equally well in settlements in the center and on the periphery of the dialect area (in this case, Bučiūnai).

The results of hierarchical and two-step cluster analyses (after data normalization) also suggest that an LKA site on the periphery is not always peripheral: perhaps the LKA dialect site of Stumbríškis could be called peripheral, since its results differ most from the data for the other three localities. However, despite its geographic position within the Panevėžys dialect area, Bučiūnai is not a peripheral LKA site—the dialect features there are preserved just as well as at the central LKA sites of Pãsvalis and Kapčiūnai.

As the results changed due to the normalization process, the mean frequencies of features used at the LKA sites also changed—they decreased slightly in the first group of similar objects and increased in the second cluster (see Table 6).

The data obtained show that all weak features and most strong features are better preserved in the settlement of Stumbríškis than in Bučiūnai, Pãsvalis, and Kapčiūnai. This casts doubt on the assumption made earlier in the study that the level of dialecticism in Stumbríš kis is the lowest for various sociogeolinguistic reasons. On the contrary, the LKA site of Stumbríškis stands apart from the other three in its productive use of stressed stem ė, i, u, stress retraction, unstressed stem ie, uo, y, ū, and unstressed word-final ė, o—these features are numerous and tend to persist in the speech of informants in the settlement.

These dialect features are also fairly well preserved in Bučiūnai, Kapčiūnai, and Pasvalys, although the highly dispersed data in the first cluster do not allow such a firm assertion (the standard deviation ranges from 0.577 for unstressed stem uo to 24.269 for unstressed word-final o).

These results explain why the LKA site of Stumbríškis formed a separate cluster according to the results of hierarchical and two-step cluster analyses (after data normalization). They also disprove the assumption made in the study that the frequency of dialect features used in Stumbríš kis differs from that of the other three LKA sites because it is peripheral. The data obtained from the two-step cluster analysis indicate that, given the predominant productivity of the dialect features mentioned above, there is no reason to call this settlement peripheral, even though it is on the periphery.

The analysis data also confirm the view that, in terms of dialecticism, the LKA site of Bučiūnai is not peripheral—the mean frequencies of use of its distinguishing features are similar to those at the other two central LKA sites, Kapčiūnai and Pãsvalis. Thus, it can be seen that the dialecticism of the LKA sites of Bučiūnai and Stumbríškis, located on the periphery of the Panevėžys dialect area, is affected neither by the strong but rapidly declining Kupiškėnai subdialect nor by the Šiauliai dialect speakers, whose dialect features are less pronounced.

Straipsniai / Articles

189

AGNĖ PUKEVIČIŪTĖ

In the residential areas of Bučiūnai, Kapčiūnai, and Pašvalys, only the mean frequencies of three strong features are higher than in Stumbr iškis: stressed stem ie, uo, o. Therefore, it can be concluded that the dialectal character of the three LKA points mentioned is maintained by the dominance of strong features in the informants’ speech, whereas in Stumbr iškis it is maintained more by weak than by strong features. Table 6 also shows that unstressed stem ie, uo, ū are relatively rare distinguishing features in all the localities studied. However, it should not be claimed that these features are tending to disappear; it is thought that, in general, informants use few words containing them (for example, piemuõ, puodynė, pūgà).

Looking at the frequency of individual features (see Tables 1, 4, and 5), it can be seen that, in the four LKA locations studied, the features most likely to persist are stressed stem ie, o, ė, i, u, stress retraction, and the weak feature of unstressed ending o. These are the features most frequently used in the localities examined.

The features least frequently found in the informants’ speech are the weak features—unstressed stem ie, uo, y, ū, ē—and one strong feature, stressed stem uo. Whether they are tending to disappear is a question for other, more comprehensive studies.

4. CONCLUSIONS

The experimental application of the cluster analysis method to determine the similarity, dialectal character, and peripherality of four LKA points in the Eastern Aukštaitian Panevėžys subdialect allows the following preliminary conclusions to be drawn.

The cluster analysis model applied in the study is suitable: this method revealed patterns in the grouping of four LKA Panevėžys subdialect points according to the variables under investigation.

The analysis shows that borderland LKA points are not necessarily peripheral in terms of dialectal character: in the localities of Bučiūnai and Stumbr iškis, situated on the edge of the Eastern Aukštaitian Panevėžys subdialect area, the dialectal character has been well preserved, and these LKA points have not adopted clear dialectal features from the neighboring subdialects.

At the LKA points of Bučiūnai, Pašvalys, and Kapčiūnai, dialectal character is best preserved through the dominance of strong features (stressed stem ie, uo, o) in informants’ speech, whereas in Stumbr iškis it is preserved more through weak (unstressed stem ie, uo, y, ū and unstressed ending ė, o) than strong features (stressed stem ė, i, u and stress retraction). The other features examined in the study show no clear patterns.

When dialects are investigated using the cluster analysis method, the selection of research material is very important: recordings of similar duration and a larger number of them should be selected. Their varying dates of collection and the informants’ speaking rate may also affect the results. To avoid inaccuracies, the frequency of feature use should

190

Acta Linguistica Lithuanica LXXII

Rytų aukštaičiai panevėžiškiai: klasterinė analizė ieškant tarmiškumo

The assessment should be made by a specialist with an excellent understanding of dialectal features. Transcribed texts could be used for the study.

LITERATŪRA

Aliūkaitė Daiva 2011: Projektas „Šiuolaikiniai geolingvistikos tyrimai Lietuvoje: punktų tinklo optimizacija ir interaktyvioji tarminės informacijos sklaida“. – Respectus philologicus 20(25), 257–260.

Aliūkaitė Daiva, Bakšienė Rima, Jaroslavienė Jurgita, Judžentyte Gintarė ir kt. 2014: Punktų tinklo tankis geolingvistinės skirties aukštaičiai: žemaičiai požiūriu. – XXI a. pradžios lietuvių tarmės: geolingvistinis ir sociolingvistinis tyrimas (žemėlapiai ir jų komentarai), 129–196.

Aliūkaitė Daiva, Mikulėnienė Danguolė 2014: Geolingvistika: ideologija, teorija ir metodai. Pagrindinės sąvokos. – XXI a. pradžios lietuvių tarmės: geolingvistinis ir sociolingvistinis tyrimas (žemėlapiai ir jų komentarai), 29–47.

Būga Kazimieras 1961: Rinktiniai raštai 3. Vilnius: Valstybinė politinės ir mokslinės literatūros leidykla.

Būtenas Petras 1932: Aukštaičių tarmės okuojančiosios pašnektės sienos. – Archivum Philologicum 3, 168–209.

Čekanavičius Vydas, Murauskas Gediminas 2002: Statistika ir jos taikymai 2. Vilnius: TEV.

Garšva Kazimieras 1977: Akcentuacijos ir vokalizmo sąryšis šiaurės vakarų panevėžiškių tarmėje. – Lietuvių kalbotyros klausimai 17, 76–87.

Garšva Kazimieras 1982a: Svarbesnės šiaurės vakarų panevėžiškių fonologijos ypatybės. – Baltistica 18(1), 65–74.

Garšva Kazimieras 1982b: Joniškėlio apylinkių šnektos tekstai. Vilnius: Vilniaus universiteto leidykla.

Garšva Kazimieras 2002: Panevėžiškių fonetikos raida abipus Latvijos sienos. – Acta Linguistica Lithuanica 47, 9–18.

Gerullis Georg 1930: Litauische Dialektstudien. Leipzig: Markert & Petters Verlag.

Girdenis Aleksas, Zinkevičius Zigmas 1966: Dėl lietuvių tarmių klasifikacijos. – Kalbotyra 14, 139–147.

IBM SPSS Statistics. Prieiga internete: http://www.insol.lt/statistics.

Straipsniai / Articles

191

AGNĖ PUKEVIČIŪTĖ

Jaunius Kazimieras 1898 (1970): Panevėžiškių tarmė. – Kalbininkas Kazimieras Jaunius, sud. Vincentas Drotvinas, Vladas Grinaveckis. Vilnius: Mintis.

Kačiuškienė Genovaitė 1982: Vakarinių šiaurės panevėžiškių balsių kiekybė ir jos fonologinė interpretacija. – Kalbotyra 33(1), 39–45.

Kačiuškienė Genovaitė 1983: Šiaurės panevėžiškių murm amųjų balsių fonetinės ypatybės, distribucija ir fonologinė interpretacija. – Kalbotyra 34(1), 24–38.

Kačiuškienė Genovaitė 1984: Šiaurės panevėžiškių tarmės balsių psichoakustinė klasifikacija. – Kalbotyra 35(1), 42–53.

Kačiuškienė Genovaitė 2000: Nekirčiuotų i/u ir ę/ǫ vartojimas Joniškėlio ir Pakruojo apylinkių rišli uose tekstuose. – Baltų filologija 9, 25–31.

Kačiuškienė Genovaitė 2003: Trečiojo amžiaus universitetas Šiauliuose: klausytojų požiūris į gimtąją tarmę. – Pedagogija 69, 88–92.

Kačiuškienė Genovaitė 2005a: Panevėžio tarmių aprašas – pirmoji mokslinė šiaurės panevėžiškių tarmės studija. – Kalbos istorijos ir dialektologijos problemos 1, 49–53.

Kačiuškienė Genovaitė 2005b: Antano Baranausko ir Kazimiero Jauniaus vieta šiaurės panevėžiškių tarmės tyrimo istorijoje. – Lietuvių katalikų mokslo akademijos suvažiavimo darbai 19, 499–503.

Kačiuškienė Genovaitė 2005c: Sodybų statinių pavadinimai šiaurės panevėžiškių tarmėje: anketinis tyrimas. – Žmogus kalbos erdvėje, 60–64.

Kačiuškienė Genovaitė 2006: Šiaurės panevėžiškių tarmės fonologijos bruožai. Vilnius: Vilniaus universiteto leidykla.

Kačiuškienė Genovaitė 2009: Sakytinės lietuvių kalbos tekstynas: tarminių formų vartosena. – Filologija 14, 61–67.

Kačiuškienė Genovaitė 2012: Tarmių fonetinių ypatybių atspindžiai „Sakytinės lietuvių kalbos tekstyne“. – Baltistica 8, 53–62.

Kačiuškienė Genovaitė 2013a: Stipriosios ir silpnosios šiaurės panevėžiškių tarmės ypatybės XXI a. pradžioje. – Tarptautinė tarpdaykinė mokslinė konferencija XXI amžiaus tarmių tyrimai: geolingvistinė ir sociolingvistinė perspektyvos. Vilnius: Lietuvių kalbos institutas, 8.

Kačiuškienė Genovaitė 2013b: Šiaurės panevėžiškių tarminės nuostatos: sociolingvistinis tyrimas. – Tarptautinė mokslinė konferencija Tarmės – Europos tautų kultūrų paveldas (pranešimų tezės). Vilnius: Mykolo Romerio universitetas, 22.

Kačiuškienė Genovaitė, Girdenis Aleksas 1982: „Žiemgališkoji“ anaptiksė šiaurės panevėžiškių tarmėje ir jos kilmė. – Baltistika 18(2), 189–191.

192

Acta Linguistica Lithuanica LXXII

Rytų aukštaičiai panevėžiškiai: klasterinė analizė ieškant tarmiškumo

Kačiūškienė Genovaitė, Girdenis Aleksas 1997: Rytų aukštaičių ir šiaurės žemaičių priegaidės: bendrybės ir skirtumai. – Kalbotyra 46(1), 31–36.

Kardelis Vytautas 2014: Lietuvių kalbos salos: probleminiai aspektai. – Lietuvių kalba 8. Prieiga internete: http://www.lietuviukalba.lt/index.php/lietuviu-kalba/article/view/56/60.

Kardelis Vytautas, Stakutytė Emilija 2015: Iš lietuvių dialektologijos raidos. – Lietuvių kalba 9. Prieiga internete: http://www.lietuviukalba.lt/index.php/lietuviu-kalba/article/view/161/131.

Kazlauskas Jonas 1968: Lietuvių kalbos istorinė gramatika. Vilnius: Mintis.

Labov William 1972: Sociolinguistic patterns. Philadelphia: University of Pennsylvania Press.

Lietuvių kalbos tarmės. Prieiga internete: www.tarmes.lt.

LKA 1977: Lietuvių kalbos atlasas. 1. Leksika. Vilnius: Mokslas.

LKA 1982: Lietuvių kalbos atlasas. T. 2. Fonetika. Vilnius: Mokslas.

LKA 1991: Lietuvių kalbos atlasas. T. 3. Morfologija. Vilnius: Mokslas.

LKT 1970: Lietuvių kalbos tarmės. Chrestomatija, red. Elena Grinaveckienė, Kazimieras Morkūnas. Vilnius: Mintis.

LKTCh 2004: Lietuvių kalbos tarmių chrestomatija, sud. Rima Bacevičiūtė, Audra Ivanauskienė, Asta Leskauskaitė, Edmundas Trumpa. Vilnius: Lietuvių kalbos instituto leidykla.

Markevičienė Žaneta 2001: Aukštaičių tarmių tekstai 2. Vilnius: VU leidykla.

Meiliūnaitė Violeta 2009: Pietų panevėžiškių vokalizmas ir jo raidos bruožai. Kaunas: Vytauto Didžiojo universitetas, Lietuvių kalbos institutas.

Mikulėnienė Danguolė 2011: Leopold Geitler’s Contribution to Lithuanian dialectology: geolinguistic aspect. – Dialektologie a geolingvistika v současné střední Evropě, 19–27.

Morkūnas Kazys 1993: Dėl sąvokos „periferinės tarmės“. – Lietuvių kalbotyros klausimai 30, 4–7.

Ručinskaitė Irena 1984: „Murmamųjų“ balsių kokybinė opozicija šiaurės panevėžiškių tarmėje. – Kalbotyra 35(1), 72–76.

Salys Antanas 1992: Raštai 4. Lietuvių kalbos tarmės. Roma: Lietuvių katalikų mokslo akademija.

Straipsniai / Articles

193

AGNĖ PUKEVIČIŪTĖ

Spolsky Bernard 2011: Ferguson and fishman: Sociolinguistics and tge sociology of language. The SAGE handbook of sociolinguistics. London: SAGE Publications.

Vaitkevičius Raimundas, Saudargienė Aušra 2006: Statistika su SPSS psichologiniuose tyrimuose: mokomoji knyga. Kaunas: VDU leidykla.

Mikulėnienė Danguolė, Meiliūnaitė Violeta (sud.) 2014: XXI a. pradžios lietuvių tarmės: geolingvistinis ir sociolingvistinis tyrimas (žemėlapiai ir jų komentarai). Vilnius: Briedis.

Zinkevičius Zigmas 1966: Lietuvių dialektologija. Vilnius: Mintis.

Zinkevičius Zigmas 1974: Smulkmenos VI. – Baltistica 10(2), 148.

Zinkevičius Zigmas 1975: Apie č‘, dž‘ // t‘, d‘ kaitą šiaurės panevėžiškių vardažodžių paradigmoje. – Kalbotyra 26(1), 85.

Zinkevičius Zigmas 1979: Kirčio atitraukimas ir kirčiavimo paradigmos. – Kalbotyra 30(1), 90–93.

Zinkevičius Zigmas 2006: Lietuvių tarmių kilmė. Vilnius: Lietuvių kalbos instituto leidykla.

Барановскій Антоній 1898: Замѣтки о литовскомъ языкѣ и словарѣ. Санктпетербургъ: Типографія Императорской Академіи Наукъ.

Вольтеръ Эдуардъ 1886: Программа для указанія особенностей говоровъ Литвы и Жмуди. СанктПетербургъ: Типографія Императорской Академіи Наукъ.

Вольтеръ Эдуардъ 1904: Литовская хрестоматия 2. СанктПетербургъ: Типографія Императорской Академіи Наукъ.

Eastern Aukštaitian of Panevėžys: Cluster Analysis in Search of Dialecticism

SUMMARY

This paper aims to test the appropriateness of one of the statistical methods – cluster analysis (both hierarchical and two-steps) – to determine the similarity of the populated localities of Eastern Aukštaitian of Panevėžys in accordance with the frequency of usage of distinctive dialectal features. The grouping peculiarities that emerged from the analysis of the research areas allow us to create a model to measure the dialecticism and peripherality of the research areas.

194

Acta Linguistica Lithuanica LXXII

Rytų aukštaičiai panevėžiškiai: klasterinė analizė ieškant tarmiškumo

90 minute long recordings of the Eastern Aukštaitian of Panevėžys collected from two central – Pasvalys (112), Kapčiūnai (138) – and two marginal – Bučiūnai (52), Stumbriškis (171) – dialectal points were randomly selected for this study. While listening to the recordings, the frequencies of usage of 13 dialectal features of the Eastern Aukštaitian of Panevėžys were identified: 7 of the dialectal features are strong (the stressed ie, uo, o, ė, i, u in the stem and the retraction of accent), while 6 of them are weak (the unstressed y, ū, ie, uo in the stem and the unstressed ė, o in the ending). The frequencies were then entered into the SPSS Statistics computer program and the cluster analysis was performed.

The results suggest that the cluster analysis (both hierarchical and two-steps) model applied in the research is suitable for revealing the grouping patterns of the Eastern Aukštaitian of Panevėžys LLA points according to the research dimensions. The analysis indicates that the marginal LLA points are not necessarily peripheral – dialectal features are perfectly maintained in the marginal areas of Bučiūnai and Stumbriškis.

In Bučiūnai, Pasvalys and Kapčiūnai points of LLA dialecticism is better maintained due to the strong features (the stressed ie, uo, o in the stem), while in Stumbriškis – the weak ones (the unstressed ie, uo, y, ū in the stem and the unstressed ė, o in the ending) more than the strong ones (the stressed ė, i, u in the stem and the retraction of accent). Other analyzed features show no clear regularities.

The research indicates that the selection of the research material is very important for the examination of dialects by the cluster analysis method: recordings of similar duration and a higher number of them should be selected. The results may be influenced by a different collection period or the speech rate of informants. To avoid the inaccuracies, the frequency of the usage of dialectal features should be examined by the specialist of dialectal features. Transcribed texts, as well, could be used for the research.

Įteikta 2015 m. sausio 21 d.

AGNĖ PUKEVIČIŪTĖ

Lietuvių kalbos institutas

Petro Vileišio g. 5, LT-10308 Vilnius, Lietuva

[email protected]

Straipsniai / Articles

195