JOLITA URBANAVIČIENĖ

Institute of the Lithuanian Language

ORCID id: orcid.org/0000-0002-5514-2263

Research areas: experimental phonetics, speech acoustics, phonology, lexicography, geolinguistics, sociolinguistics.

JANA TAPERTE

Institute of the Latvian Language, University of Latvia

ORCID id: orcid.org/0009-0009-1723-0666

Research areas: experimental phonetics, speech acoustics, phonology.

DOI: doi.org/10.35321/all88-05

FORMANTS OF THE TRILLS IN LITHUANIAN AND LATVIAN IN DIFFERENT POSITIONS: AN ACOUSTIC AND STATISTICAL STUDY

Lithuanian and Latvian Trills: An Acoustic and Statistical Study of Formant Structure

ABSTRACT

The article presents a formant study of the trilled sonorants of the contemporary Baltic languages—Lith. [r], [rʲ], and Latv. [r]—aimed at comparing the formants F1, F2, F3, and F4 of sonorants pronounced in prevocalic (<R>VR), intervocalic (V<R>V), and postvocalic (RV<R>) positions, examining their variation in the context of different vowels, highlighting differences between the formants of [r] and [rʲ] in Lithuanian and Latvian, and determining the statistical significance of these differences. The study shows that, according to the combined values of the first two formants, Latv. [r] occupies an intermediate position between unpalatalized Lith. [r] and palatalized Lith. [rʲ]. Syllable position has a statistically significant effect on the first formant of Lithuanian trilled sonorants, whereas in Latvian it affects all four formants. It was established that the context of adjacent vowels has a similar effect on the formants of the trilled sonorants of the contemporary Baltic languages: the first formant is associated with the height of the adjacent vowel, while the second and subsequent formants depend on the backness of the adjacent vowel.

100

Acta Linguistica Lithuanica LXXXVII

Lietuvių ir latvių kalbų virpamųjų sonantų formantės skirtingose pozicijose: akustinis ir statistinis tyrimas

the formants of the sonants have a similar effect: the first formant is associated with the height of the adjacent vowel, while the second and subsequent formants depend on the backness of the adjacent vowel.

KEYWORDS: Baltic languages, trill, formants, statistical significance, syllable position, velarization, palatalization.

ANNOTATION

The paper deals with the formant structure of trills in the contemporary Baltic languages – Lith. [r], [rʲ] and Lat. [r]. The aim of the paper is to compare the formant structure of trills (F1, F2, F3 and F4) in prevocalic (<R>VR), intervocalic (V<R>V) and postvocalic (RV<R>) positions and its variation in the context of different vowels, to highlight the trends of formant dynamics in Lithuanian and Latvian and to determine the statistical significance of the differences. The study shows that according to the average values of the first two formants, Lat. [r] occupies an intermediate position between non-palatalized Lith. [r] and palatalized Lith. [rʲ]. Syllable position has a statistically significant influence on the first formant of trills in Lithuanian and on all four formants in Latvian. According to the results, the context of adjacent vowels has a similar influence on the formants of trills in both languages: the first formant is related to the height of the adjacent vowel, and the second and subsequent formants depend on the backness of the adjacent vowel.

KEYWORDS: Baltic languages, trills, formants, statistical significance, syllable position, velarization, palatalization.

1. OVERVIEW OF RESEARCH ON THE TRILLS OF THE MODERN BALTIC LANGUAGES

Compared with other groups of sonants, trills have been studied least of all in Lithuanian and Latvian experimental phonetics. General information about the articulatory and spectral properties of the trills [r], [rʲ] can be found in the principal works on phonetics, including The Grammar of the Lithuanian Language (Ul vydas et al. 1965), as well as the works of Elzė Mikalauskaitė (1975), Valerija Vaitkevičiūtė (2001), Antanas Pakerys (2003), and Asta Kazlauskienė (2018). The quantity and spectral features of all consonants of Standard Lithuanian

^¹ The article uses characters from the International Phonetic Alphabet (IPA). The selection of characters was based on Protocol Resolution No. PN-14 (k–25) of the State Commission of the Lithuanian Language, dated 8 July 2021, which contains recommendations for adapting the IPA to designate the sound units of Standard Lithuanian. For more details, see: https://vlkk.lt/vlkk-nutarimai/protokoliniai-nutarimai/rekomendacija-del-tarptautines-fonetines-abėceles-pritaikymo-bendrinės-lietuviu-kalbos-garsiniams-vienetams-zymeti.

Straipsniai / Articles

101

JOLITA URBANAVIČIENĖ, JANA TAPERTE

were instrumentally investigated in Sigita Dereškevičiūtė’s dissertation Acoustic Properties of Lithuanian Consonants (2013). She established that the spectrum of the trills [r], [rʲ] is diffuse, i.e., the areas of energy concentration are distributed across a broad frequency band, with energy intensified in the mid-frequency range (2000–4000 Hz) (Dereškevičiūtė 2013: 136–137). It was also established that the palatalized trill [rʲ] is characterized by higher frequencies than the unpalatalized [r], cf.: [r] = 1300–1700 Hz, [rʲ] = 1700–2000 Hz. S. Dereškevičiūtė’s duration-analysis results showed that, of all Lithuanian consonants, trills are the shortest² (Dereškevičiūtė 2013: 138). In their scholarly monograph Acoustic Properties of Consonants: Palatalization and Voicing (Ambrazēvičius, Leskauskaitė 2014), Rytis Ambrazēvičius and Asta Leskauskaitė investigated the acoustic properties of trills alongside other consonants and established that one of the most important indicators of palatalization is F2 loci, while the voicing of consonants should be interpreted comprehensively—there is no single reliably distinctive feature of voiceless and voiced consonants in all cases. In the collective monographs of the series Corpus of the Baltic Languages at the Beginning of the 21st Century published in 2019 by researchers from the Institute of the Lithuanian Language and the LU Institute of the Latvian Language, an attempt was made for the first time to compare Lithuanian and Latvian trills: the monograph An Instrumental Study of Vowel and Sound Interactions contains examples of dynamic spectrograms of trills (see Jaroslavienė, Grigorjevs, Urbanavičienė, Indričānė 2019: 52–53, Figs. 6.41–6.48), while the monograph An Instrumental Study of Consonants presents the formant values of trilled consonants in the context of different vowels in prevocalic CVC position (Urbanavičienė, Indričānė, Jaroslavienė, Grigorjevs 2019: 206–220).

The first instrumental studies of Latvian trills are associated with Elmārs Liepa, who measured the duration of sonants (Liepa 1970) and also analyzed the phonetic properties of the trill [r] in the context of voiceless consonants (Liepa 1957). Dace Markus and Juris Grigorjevs investigated the formant structure, the duration of the closure and opening phases, and formant transitions of Latvian trills in the book Methods of Phonetic Research and Visualization I (Markus, Grigorjevs 2002). J. Grigorjevs, drawing on data from five informants, analyzed the acoustic properties of the Latvian sonant [r] pronounced in symmetrical CVC sequences (Grigorjevs 2012). From 2012 to 2016, a project entitled “Acoustic characteristics of the sound system of Standard

² During the study, S. Dereškevičiūtė established that, compared with other consonant classes, sonants are the shortest—their duration ratios, in descending order, are distributed as follows: affricates: plosives: fricatives: sonants = 2: 1.6: 1.5: 1. According to the author, plosives and affricates have several articulation phases, the duration of friction is important for the articulation of fricatives, while during the articulation of sonants the position of the speech organs changes rapidly and in diverse ways (Dereškevičiūtė 2013: 138–139).

102

Acta Linguistica Lithuanica LXXXVIII

Lietuvių ir latvių kalbų virpamųjų sonantų formantės skirtingose pozicijose: akustinis ir statistinis tyrimas

Latvian by age groups” (headed by J. Grigorjevs) was carried out at the LU Institute of the Latvian Language. During the project, sonants [l, ʎ, m, n, ɲ, r] produced by informants from different age groups (5–15, 16–39, 40–59, and 60–80 years) were compared, as were their spectral characteristics in RVR (sonant–vowel–sonant) and VRV (vowel–sonant–vowel) sequences, in prevocalic and postvocalic positions (see Taperte 2013; 2014; 2014a; 2015; Čeirane, Indričāne, Taperte 2014).

A review of the research conducted on the trills of the modern Baltic languages shows that, on the one hand, the trills of the Baltic languages are studied continuously and from various perspectives; on the other hand, Lithuanian and Latvian trills are usually studied alongside other sonants, without taking into account their differing articulatory and acoustic properties, without applying different methodologies to individual sonant classes, and comparative studies of Latvian and Lithuanian trills are virtually nonexistent; moreover, the acoustic properties of trills pronounced in different positions have not yet been analyzed^³

The object of this study is the trilled sonorants of the contemporary Baltic languages, pronounced in prevocalic (<R>VR), intervocalic (V<R>V), and postvocalic (RV<R>) positions. The aim of the article is to analyze the F1, F2, F3, and F4 formant values of trilled sonorants, compare their differences in the context of different vowels, determine the statistical significance of the differences, and highlight tendencies in formant dynamics in Lithuanian and Latvian. The study will address the following tasks: 1) discuss the articulatory and acoustic properties of the trilled sonorants of the Baltic languages and their phonological status; 2) determine the F1, F2, F3, and F4 formant values of the trilled sonorants of the contemporary Baltic languages—Lith. [r, ř] and Latv. [r]—in prevocalic, intervocalic, and postvocalic positions; 3) compare the formant values of the trilled sonorants in the context of different vowels; 4) investigate the statistical significance of differences between formants by conducting a one-way analysis of variance and a Bonferroni test; 5) identify the factors determining changes in the formants. The study is based on data from male informants.

2. EMPIRICAL MATERIAL

The study analyzed Lithuanian and Latvian trills pronounced in isolated RVR-type syllables: here R is Lith. [r, ř], Latv. [r], and V is the short

^³ To compensate for the lack of systematic comparative studies of Baltic sonants, a research project entitled “Corpus of the Modern Baltic Languages at the Beginning of the 21st Century: A Comparative Acoustic and Perceptual Study of Sonants” (BaltSon, reg. No. S-LIP-21-7) was conducted at the Institute of the Lithuanian Language in 2021–2022. The project was supported by the Research Council of Lithuania. The project leader was Dr. Jolita Urbanavičienė.

Straipsniai / Articles

103

JOLITA URBANAVIČIENĖ, JANA TAPERTE

or long vowel: Lith. [i, iː, eː, ε, æː, e, αː, ɔ, ɒ, o, oː, ʊ, ʊ, uː uː] and Latv. [i(ː); e(ː); æ(ː); a(ː); ɔ(ː); u(ː)]^⁴. Thirty Lithuanian sequences and 24 Latvian sequences were analyzed:

Lith.: rir [rʲir], iri [irʲi]; rer [rʲe r], ere [erʲe]; rar [rer], ara [ærə]; ror [rɔr], oro [ɔrɔ]; rur [rur], uru [ʊrʊ]; ryr [rʲirʲ], yry [irʲiʲ]; rēr [rʲeːr], ēre [erʲeː]; rer [rʲæːr], ere [æːrʲæː]; rar [raːr], ara [αrαː]; ror [roːr], oro [ɔrɔː]; rūr [ruːr], ūrū [uːruː]; rior [rʲɔr], iorio [ɔrʲɔ]; rior [rʲoːr], iorio [oːrʲoː]; riur [rʲʊr], iuriu [ʊrʲʊ]; riūr [rʲuːr], iūriu [uːrʲuː];

Latv.: rir [rir], iri [iri]; rer [rer], ere [ere]; rer [rær], era [ærə]; rar [rar], ara [ɑrɑ]; ror [rɔr], oro [ɔrɔ]; rur [rur], uru [uru]; rīr [rīr], īrī [īri]; rēr [reːr], ēre [eːre]; rēr [ræðr],ērā [æːrɑ]; rār [rɑːr], ārā [ɑːrɑː]; ror [rɔːr], oro [ɔrɔː]; rūr [ruːr], ūru [uːruː].

Three positions were examined: prevocalic (<R>VR), intervocalic (V<R>V), and postvocalic (RV<R>). The test sequences were pronounced by 12 male informants^⁵: 6 native Lithuanian speakers and 6 native Latvian speakers (aged 21–42). The informants came from different regions of Lithuania and Latvia and spoke the standard language. If their native dialect influenced the informants’ speech, the dialectal features were neutralized when the data from all informants were combined. Each informant pronounced each sequence three times with neutral intonation and at a normal speaking rate (speech rate and intonation are among the factors determining the occurrence of [r] allophones; see Ladefoged 2003: 149). In total, approximately 1,458 RVR sequences were examined: 810 in Lithuanian and 648 in Latvian. Formant values were measured using the Praat program (Boersma, Weenink 2018); the data obtained were processed using Microsoft Excel (v. 13, Microsoft Corporation), and statistical calculations were performed using SPSS (v. 22, IBM Corporation).

3. FORMANT MEASUREMENT METHODOLOGY

In Praat, after opening the RVR segment under investigation, the command View&Edit is selected under formants; when it is clicked, the waveform of the selected segment appears on the screen above and the spectrogram below. Based on these displays

⁴ It should be noted that, according to the principles of the International Phonetic Alphabet, different symbols are used when transcribing Lithuanian and Latvian vowels; for more details, see Jaroslavienė, Grigorjevs, Urbanavičienė, Indričānė 2019: 17.

⁵ The sound recordings from the project “Corpus of the Modern Baltic Languages at the Beginning of the 21st Century: A Comparative Acoustic and Perceptual Study of Sonants” (BaltSon, reg. No. S-LIP-21-7) were used for the study described in the article. Recordings in both languages were made using the portable high-resolution digital recorder Tascam DR-100MK II and the head-mounted microphone AKG C 520. The sound signal was digitized at 44.1 kHz with a quantization level of 16 bits.

104

Acta Linguistica Lithuanica LXXXVII

Lietuvių ir latvių kalbų virpamųjų sonantų formantės skirtingose pozicijose: akustinis ir statistinis tyrimas

the stable portion of the trill is marked (see Fig. 1), and the command Formant – Formant listing is selected, which provides digital values for the F1–F4 formants of the marked sound interval.

FIG. 1. Waveform (bottom) and spectrogram (top) of Latv. [rɪr] segment in the Praat window

From the generated list, the formants at the midpoint of the sound are selected: the midpoint formant value is taken, rather than the average of the formants of the stationary portion of the sound. All midpoint formant values found, rounded to the nearest 10 Hz, are transferred to Excel. In this program, the following are calculated: 1) the mean of all realizations by the same informant; 2) the mean of the sonant produced in the same position by all informants; 3) the mean of the sonant produced in the same vowel environment by all informants. Based on the data obtained, graphs were drawn and statistical analyses were performed in SPSS: one-way analysis of variance ANOVA, the Bonferroni test, and post hoc calculations.

One-way analysis of variance ANOVA is applied when it is necessary to evaluate differences between sample means and the overall mean and to determine whether a statistically significant difference exists among them (Čekanavičius, Murauskas 2004). Once it has been established that the means of the samples being compared differ significantly, additional tests must be performed to identify the specific samples that differ. For this purpose, this article used the Bonferroni test, selected from the post hoc criteria available in SPSS. In the Bonferroni test, two independent samples (e.g., prevocalic and intervocalic, prevocalic and postvocalic, intervocalic and postvocalic

⁶ The Bonferroni criterion is not applicable when there are many samples, because in that case a statistically significant difference between the means is not detected even though the true sample means differ (Čekanavičius, Murauskas 2002).

Straipsniai / Articles

105

JOLITA URBANAVIČIENĖ, JANA TAPERTE

positions) are compared using Student’s t-test. The pairs being compared differ significantly if the significance level is p < 0.05; the pairs do not differ statistically significantly if p ≥ 0.05 (for a detailed discussion of the statistical results, see the “Results” section).

4. ACOUSTIC AND ARTICULATORY PROPERTIES OF TRILLS

Works on phonetics emphasize that trills (also called trilled sonants) are characterized by vibration of one active speech organ against another—the passive organ—determined by aerodynamic conditions (Ladefoged, Maddieson 1996: 217). The obstruction in the articulation of trills is variable: “the tip of the tongue, vibrating, strikes the alveoli [...] and then moves away from them again [...]” (Pakerys 2003: 72). When the tongue tip touches the alveoli, a closure phase is formed, and when it moves away, an opening phase occurs (Ladefoged, Maddieson 1996: 218). In Standard Lithuanian, there may be one or two such vibrations, for example (Girdenis 2003: 25), in Standard Latvian two or three (Laua 1997: 48), and in Lithuanian dialects up to five^⁷ (Mikalauskaitė 1975: 53). According to the number of vibrations (or taps), the International Phonetic Alphabet distinguishes between single-tap trills (English Tap or Flap) and multiple-tap trills (English Trill) (IPA 2015)^⁸. Lithuanian and Latvian trills are considered multiple-tap trills; however, the number of tongue vibrations is not high and does not exceed 2–3 vibrations (Švageris 2022: 59)^⁹.

The first vibration is somewhat longer than the subsequent vibrations (Ladefoged, Maddieson 1996: 218). The number of vibrations depends on several factors: intonation (emphatic tone increases the number of vibrations), adjacent vowels ([r] next to rounded vowels has more vibrations), tone accent (a falling tone [r] is pronounced with more vibrations^¹⁰), and dialect (an [r] pronounced by Eastern Highlanders will have more vibrations than, for example, the [r] of Western Highlanders) (Mikalauskaitė 1975: 53). It has been observed that there are more taps of the tongue tip at the end of a word than at the beginning of a word

^⁷ However, compared with other languages (Polish, Russian), Lithuanian [r], [r̥] are weaker (Mikalauskaitė 1975: 53).

^⁸ Other languages also have the retroflex [ɽ] (e.g., in English) and the uvular [ʀ] (e.g., in French) (IPA 2015).

^⁹ According to E. Švageris, “the small number of taps is probably determined by phonological reasons [...]. Under this parameter, no other sound opposes the Lithuanian trill, so [...] the interpretation of the sound does not change whether it is articulated with 1–2 or 4–5 taps” (Švageris 2022: 59).

^¹⁰ In stressed falling-tone diphthongs, sonants are lengthened to half-long, e.g.: pirštas [²pʲirʲt͡s], viršti [²u̯irʲsʲti] (DLKG 2005: 29).

106

Acta Linguistica Lithuanica LXXXVIII

Lietuvių ir latvių kalbų virpamųjų sonantų formantės skirtingose pozicijose: akustinis ir statistinis tyrimas

at the beginning (Vaitkevičiūtė 2001: 114). When pronouncing the unpalatalized [r], the tongue tip also vibrates more strongly than when pronouncing [rʲ]. However, objective Baltic-language data are still lacking to confirm most of these claims—no instrumental studies have been conducted, or they were conducted using technologies that are now outdated¹¹.

² FIG. Trill [r] spectrograms: spectrum of four vibrations (left) and spectrum of one vibration (right)

The articulation of palatalized [rʲ] requires additional aerodynamic conditions: not only is the tongue tip brought close to the alveoli (to produce vibration), but its middle part is also raised upward, and the sides of the middle part of the tongue make more contact with the edges of the hard palate than when [r] is pronounced (Mikalauskaitė 1975: 54; Ladefoged, Maddieson 1996: 221; Vaitkevičiūtė 2001: 114). The more complex articulation of palatalized [rʲ] results in fewer vibrations than in the pronunciation of unpalatalized [r]. V. Vaitkevičiūtė has also noted a lower degree of voicing in the consonant [rʲ], which can be explained by the articulatory properties of [rʲ]: “when the tongue touches the palate, it creates an obstruction which, as the air exhaled from the lungs overcomes it, produces more noise than when [r] is pronounced” (Vaitkevičiūtė 2001: 114).

In the language of Western Courland and among Latvians abroad, the palatalized Latv. [rʲ] has survived as a facultative variant of the /r/ phoneme (see LVG 2013: 68; Markus,

¹¹ Certain experimental data on Baltic trills are provided, for example, by Elzbieta Mikalauskaitė’s (1975: 157–158, 176) and Alise Laua’s (1997: 48–49) radiographs and palatograms. Unfortunately, they were made around the middle of the 20th century.

Straipsniai / Articles

107

JOLITA URBANAVIČIENĖ, JANA TAPERTE

Auzina 2013: 19), which differs from Standard Latvian [r] in terms of formant values: the first formant of palatalized [rʲ] is lower, while the other formants are higher than those of unpalatalized [r], cf.: F1 [r] – ~500 Hz, [rʲ] – ~350–400 Hz; F2 [r] – ~1350 Hz, [rʲ] – ~1750–1800 Hz; F3 [r] – ~2000–2500 Hz, [rʲ] – ~2500–3000 Hz (LVG 2013: 68).

According to the active speech organ (see Table 1), Lithuanian and Latvian trills should be considered apical, and according to the passive organ, alveolar (Mikalauskaitė 1975: 53–54; Pakerys 2003: 73; Bacevičiūtė 2008: 59; Kazlauskienė 2018: 57; Laua 1997: 48; LVG 2013: 68). In the context of the world’s languages, the Baltic trills can be classified as alveolar apicals¹² (in whose articulation the tip of the tongue participates). Latvian has only one trill, while in Lithuanian the trills differing with respect to palatalization have independent phonemic status—/r/ and /rʲ/ (for more details, see “Phonological Interpretation of Trills”).

TABLE 1. Classification of the Trills of the Modern Baltic Languages

Both Lithuanian trills [r] and [rʲ] are classified as alveolar according to place of articulation. Their main articulatory difference is that when palatalized [rʲ] is pronounced, the middle part of the tongue rises slightly upward, toward the palate, thereby changing the shape of the oral cavity (see Figs. 3–4, cf. Pakerys 2003: 74).

¹² R. Ambrazēvičius and A. Leskauskaitė (2014: 165) classify unpalatalized Lithuanian [r] as an apical alveolar or postalveolar trill, and palatalized [rʲ] as an apical or laminal alveolar or postalveolar trill. A. Laua (1997: 48) classifies the Latvian trill [r] as apical, whereas “The Grammar of the Latvian Language” classifies it as laminal, i.e., one in whose pronunciation the anterior part of the tongue is active (LVG 2013: 62).

108

Acta Linguistica Lithuanica LXXXVII

Lietuvių ir latvių kalbų virpamųjų sonantų formantės skirtingose pozicijose: akustinis ir statistinis tyrimas

FIG. 3. Activity of the speech organs in pronouncing the unpalatalized Lithuanian trill [r]^¹³

FIG. 4. Activity of the speech organs in pronouncing the palatalized Lithuanian trill [r^ʲ]

The relatively precise pronunciation of unpalatalized [r] and palatalized [r^ʲ] poses difficulties for both first-language (L1) and second-language (L2) users. Children often tend to replace trills with more easily articulated lateral sonants (Borden et al. 1994: 116–117) or pronounce the unpalatalized [r] in all positions. For example, Maria-Josep Solé’s (2002) research shows that trills (together with sibilants) are the last group of sounds that children learn to pronounce, and some native speakers never learn to articulate them correctly because of the tongue-tip vibration, which requires an appropriate position and tension of the speech organs and suitable aerodynamic conditions (Solé 2002: 656). It is likely that precisely this complexity of the vibration mechanism created the conditions for the emergence of various types of trills, e.g., a trill is produced not by vibrating the tongue but by touching it against the alveoli only once, assibilation occurs, and so on (Scobbie 2006; Solé 2002; Vásquez Carranza 2006).

The closure and opening phases of trills are visible in spectrograms (see Figs. 5–8): light areas mark the closure phase, which resembles the closure of plosives, when the formants are barely discernible, while dark areas mark the opening phase, when the spectrogram displays a formant structure and concentrated energy (Ladefoged, Maddieson 1996: 217; Ladefoged 2003: 150; Dereškevičiūtė 2013: 50–51). In the opening phase, the formants of trills resemble the formant structure of the high front vowel [i] (Švageris 2022: 61). Studies of other languages show that both the closure and opening intervals last approximately 25 ms (Ladefoged, Maddieson 1996: 218).

^¹³ For an animated view of speech-organ activity, see TARTIS (available online at: http://tartis.vdu.lt/fonetika-ir-tartis/igudziu-tobulinimas/garsu-ypatybes/).

Straipsniai / Articles

109

JOLITA URBANAVIČIENĖ, JANA TAPERTE

FIG. 5. Isolatedly pronounced Lith. [riːr] (male voice^¹⁴)

FIG. 6. Isolatedly pronounced Lith. [rɑːr] (male voice)

Studies have recorded the influence of adjacent vowels on the formant dynamics of trills (Grigorjevs 2012: 286–287, see Fig. 9). It was established that in Latvian the value of the first formant of [r] next to the high vowels [i, u] is the lowest, in the context of mid vowels [e, ɔ] it is intermediate, while next to the low vowels [æ, a] the F1 formant of the trill [r] is the highest. The lowest value of the second formant is recorded next to back vowels [ɔ, u] (because labialization lowers the F2 frequency), and the highest next to front vowels [i, e]. No clear correlation was found between the frequency of the F3 formant and the quality of adjacent vowels, while the F4 formant changes similarly to F2—its value regularly decreases in the context of rounded vowels.

^¹⁴ The spectrograms were drawn in the PRAAT program using sound recordings from the project “Corpus of the Modern Baltic Languages at the Beginning of the 21st Century: A Comparative Acoustic and Perceptual Study of Sonants” (BaltSon, reg. No. S-LIP-21-7).

110

Acta Linguistica Lithuanica LXXXVIII

Lietuvių ir latvių kalbų virpamųjų sonantų formantės skirtingose pozicijose: akustinis ir statistinis tyrimas

FIG. 7. Isolatedly pronounced Latv. [rir] (male voice)

FIG. 8. Isolatedly pronounced Latv. [rar] (male voice)

The influence of adjacent vowels on Lithuanian trills has not yet been analyzed; this is one of the aims of the present study.

Straipsniai / Articles

111

JOLITA URBANAVIČIENĖ, JANA TAPERTE

FIG. 9. LPC spectrum of the F1–F4 formants of the trill [r]^¹⁵ in different vowel contexts: aggregated Latvian data (Taperte 2016)

5. PHONOLOGICAL INTERPRETATION OF TRILLS

The unpalatalized and palatalized rhotic sonants of Lithuanian contrast as separate phonemes in the position before back vowels (see Table 2), e.g.: trūko [ˈtrʊkɔ]: triūko [ˈtʲrʲʊkɔ]; tvarūs [tʊəˈrʊs]: tvariūs [tʊəˈrʲʊs]. In other positions, the opposition between unpalatalized and palatalized rhotic sonants is neutralized:

before front vowels and before other palatalized consonants or /j/, the phoneme /rʲ/ is realized, e.g.: rýtas [ˈrʲiːtɛs], bérti [ˈbʲæːrʲtʲ];

before unpalatalized consonants and word-finally, the allophone of the phoneme /r/ is pronounced, e.g.: rãtas [ˈrɑːtɛs], dár [ˈdɑr].

TABLE 2. Distribution and phonological interpretation of Lithuanian rhotic sonants

Lithuanian rhotic sonants | Positions | Phonemes [r] | [−Vᵘ] + | [−Vⁱ] | [−C] + | [−Cʲ] | [−#] + | /r/ [rʲ] | [−Vᵘ] + | [−Vⁱ] + | [−C] | [−Cʲ] + | [−#] | /rʲ/

^¹⁵ The LPC (English Linear Predictive Coding) spectral slice plot highlights characteristic frequencies of sounds (formants, if the sounds are periodic) or the overall distribution of energy in the spectrum (if the sounds are non-periodic) (Johnson 2003: 100).

112

Acta Linguistica Lithuanica LXXXVIII

Lietuvių ir latvių kalbų virpamųjų sonantų formantės skirtingose pozicijose: akustinis ir statistinis tyrimas

In Latvian, rhotic sonants are represented by a single phoneme /r/, which has no palatalized counterpart. This is because “first, the consonants of Latvian are not affected by the vowel following them—the consonants are pronounced identically before both front and back vowels; second, Latvian has no opposition between hard and soft consonants before back vowels” (Kabelka 1987: 68); thus, Latvian has no opposition between palatalized and unpalatalized consonants (cf. also LVG 2013: 75–79). Latvian /rʲ/ retained in dialects is discussed in the section “Acoustic and articulatory features of rhotic sonants.”

6. RESULTS

6.1. Formants of Baltic rhotic sonants in different RVR / VRV positions

During the study, changes in the F1–F4 formant values of rhotic sonants were compared in different positions: Lith. and Lat. [r] in prevocalic, intervocalic, and postvocalic positions, and Lith. [rʲ] in prevocalic and intervocalic positions. The table (see Table 3) and figures (see Figs. 10–12) present summarized informant data, rounded to the nearest 10 Hz.

TABLE 3. Summarized F1–F4 formant values of Baltic [r] / [rʲ] rhotic sonants in prevocalic (PRE), intervocalic (INTER), and postvocalic (POST) positions: male data¹⁶

¹⁶ The study data were not normalized because data from only one sex—men—were compared; moreover, the main focus was not on specific formant values but on trends of change, which are clearly visible when examining non-normalized data.

Straipsniai / Articles

113

JOLITA URBANAVIČIENĖ, JANA TAPERTE

FIGS. 10–12. Mean F1–F4 formant values of Baltic [r] / [rʲ] rhotic sonants in different positions: male data

The F1 formants of Lithuanian unpalatalized [r] and palatalized [rʲ] differ by about 100 Hz in prevocalic and intervocalic positions—in both cases, the first formant of [r] is higher than that of [rʲ]. Even greater differences between the two rhotic sonants emerge most clearly when comparing F2–F4 formants. The F2 of the palatalized sonant [rʲ] is generally higher than that of its unpalatalized counterpart—the difference increases particularly in intervocalic position (Fig. 11), when [rʲ] occurs between two front vowels, which reinforce secondary palatalization. Differences between [r] and [rʲ] are also recorded in the F3–F4 ranges—in both cases, higher formants are characteristic of palatalized [rʲ]. The values of F3, F4, and higher formant frequencies are determined less by the phonetic features or distinctive features of the sounds (as in the case of F1 and F2) than by intonation, individual voice characteristics, and adjacent sounds (Zinder 1979: 176; Girdenis 2003: 222); therefore, in the next section, the formants of Baltic rhotic sonants will be compared in different vowel contexts.

A comparison of the formant values of Lith. [r] and Lat. [r] shows that the Latvian sonant is characterized by a higher F1 value and, consequently, greater velarization. According to the F2 values, Lat. [r] occupies an intermediate position between unpalatalized Lith. [r] and

114

Acta Linguistica Lithuanica LXXXVII

Lietuvių ir latvių kalbų virpamųjų sonantų formantės skirtingose pozicijose: akustinis ir statistinis tyrimas

palatalized [r^ʲ], cf.: in prevocalic position, Lith. [r] F2 = 1240 Hz, Lat. [r] F2 = 1450 Hz, Lith. [r^ʲ] F2 = 1750 Hz. It can be stated that, in terms of palatalization, Lat. [r] is pronounced “more softly” than Lith. [r], but “more firmly” than Lith. [r^ʲ].

To determine whether the F1–F4 formants of Baltic rhotic sonants differ in different positions (prevocalic, intervocalic, and postvocalic) and whether these differences are statistically significant, a one-way analysis of variance ANOVA was performed in SPSS (see Table 4). The results of the study of the Lithuanian unpalatalized rhotic sonant [r] showed that F1 and F4 values depend particularly strongly on position—their differences are statistically significant (the significance levels of the differences are p = 0,006 and p < 0,001, respectively). The F1 and F3 formants of the Lithuanian palatalized rhotic sonant [rʲ] also change statistically significantly in different positions (the p value is, respectively, p < 0,001 and p = 0,041). Examination of the Latvian rhotic sonant [r] formants in different positions established that the values of all formants depend statistically significantly on position—the F2 values differ least significantly (see Table 4).

TABLE 4.

Results of the one-way analysis of variance of F1–F4 formants of contemporary Baltic rhotic sonants (formants for which statistically significant differences were recorded in the comparisons are marked in gray)

Formant | Lithuanian | Latvian [r] | [rʲ] | [r] p value | p value | p value F1 | 0,006 | < 0,001 | < 0,001 F2 | 0,095 | 0,358 | 0,022 F3 | 0,189 | 0,041 | < 0,001 F4 | < 0,001 | 0,345 | < 0,001

Using the Bonferroni test (cf. Čekanavičius, Murauskas 2004), a post hoc comparison of the F1–F4 formants of Lith. [r] and Lat. [r] was performed within the specific groups being compared, namely, prevocalic, intervocalic, and postvocalic positions. The test results showed that the F1 values of Lithuanian unpalatalized [r] formants differ statistically significantly between prevocalic and postvocalic positions (p = 0,012), as well as between intervocalic and postvocalic positions (p = 0,020), whereas the F1 differences between prevocalic and intervocalic positions are not significant. No significant difference was established among the F2 and F3 values of Lith. [r] formants, whereas F4 differs significantly in all positions (significance level p < 0,001), except between intervocalic and postvocalic positions.

Straipsniai / Articles

115

JOLITA URBANAVIČIENĖ, JANA TAPERTE

The results of the post hoc study of the Latvian rhotic sonant [r] show statistically significant F1 differences in all positions (the significance level is p < 0,001), except between intervocalic and postvocalic positions (p = 0,812). Significant differences in F2 were recorded only when comparing prevocalic and postvocalic positions (p = 0,021); in other cases, the p value exceeds the 0.05 threshold and does not indicate a significant difference. The F3 and F4 data also differ significantly in all positions (p < 0,001), except between prevocalic and intervocalic positions—the formant differences in these positions are p = 0,112 (for F3) and p = 0,350 (for F4).

Thus, summarizing the results of the one-way analysis of variance of the F1–F4 formants of Lithuanian and Latvian rhotic sonants, it can be stated that the F1 of the [r] rhotic sonants of both languages differs significantly in all positions—the F1 of Latvian [r] is higher and is associated with a greater degree of velarization than the F1 of Lithuanian [r]. Differences in F3 and F4 were also recorded; these are associated with the context of adjacent vowels and merit closer analysis.

6.2. Formants of Baltic rhotic sonants in different vowel contexts

The F1–F4 formant values of the Lithuanian unpalatalized rhotic sonant [r] were examined in the context of back vowels [ɒ, ɑː, ɔ, oʊ, ʊ, u], while the palatalized sonant [rʲ] was examined not only before front vowels but also before fronted back vowels [ɪ, ï, ɛ, e, æ, ɒ, ø, ʏ, y]. The Latvian rhotic sonant [r] was analyzed both before back vowels [ɑ(ː); ɔ(ː); u(ː)] and before front vowels [i(ː); e(ː); æ(ː)], since it is indifferent to palatalization.

The study results, presented in Figs. 13–15, show that the F1 value of Lith. [r] is lower before high vowels [ʊ, u] and higher before mid- and low-height vowels [ɔ, oː, ɛ, a]. The F1 curve of Lat. [r] changes similarly: F1 is lowest next to high vowels [i, u], intermediate in the context of mid-height vowels [e, ɔ], and the F1 formant of the rhotic [r] is highest next to low vowels [æ, a] (cf. the analogous results of J. Grigorjevs (2012: 286–287)). The F1 values of Lithuanian palatalized [rʲ] are lower than those of the unpalatalized equivalents—they do not reach 500 Hz. Lower F1 values of the rhotic sonant [rʲ] were also recorded before high (including fronted) vowels, whereas higher values occurred before low vowels [ɛ, æ]. It should be noted that in intervocalic position, the first formant of [rʲ] is lower than in prevocalic position—the surrounding front vowels lower F1 even further (see Fig. 14).

116

Acta Linguistica Lithuanica LXXXVIII

Lietuvių ir latvių kalbų virpamųjų sonantų formantės skirtingose pozicijose: akustinis ir statistinis tyrimas

FIG. 13. Formants of the Lithuanian unpalatalized rhotic sonant [r] in different vowel contexts: male data

FIG. 14. Formants of the Lithuanian palatalized rhotic sonant [rʲ] in different vowel contexts: male data

FIG. 15. Formants of the Latvian rhotic sonant [r] in different vowel contexts: male data

Straipsniai / Articles

117

JOLITA URBANAVIČIENĖ, JANA TAPERTE

The second formant of Lithuanian rhotic sonants [r] and [rʲ] differs significantly: that of unpalatalized [r] occurs in the 1240–1310 Hz range, whereas that of palatalized [rʲ] occurs in the 1750–1860 Hz range (see Table 3, Figs. 13–14). The F2 of unpalatalized [r] is lower before rounded vowels [o, u, ɔ, o̞] and higher before unrounded vowels [e, ɑ], showing that rounding lowers the F2 value. Even before fronted back vowels [ɤ, ɯ, ə, ɞ], the F2 value of rhotic sonant [rʲ] is lower than before front vowels [ɪ, i, e̞, ɛ, æ̞] (Fig. 14). In Latvian, as in Lithuanian [r], the lowest F2 value of rhotic sonant [r] is recorded next to back vowels [ɔ, u], and the highest next to front vowels [i, e] (see Fig. 15). It is noteworthy that, according to F2, Latvian rhotic sonant [r] occupies an intermediate position between unpalatalized Lith. [r] and palatalized Lith. [rʲ]—its second formant varies in the 1400–1450 Hz range.

The dynamics of the third formant in both Lithuanian and Latvian reproduces the trajectory of the second formant: before front vowels, F3 values are higher, whereas before back vowels they are lower (see Figs. 13–15). The fourth formant is characterized by greater variability among its values in different positions. The F4 value of Lith. [r] varies particularly strongly (see Fig. 13), cf.: 3270 Hz in prevocalic position, 3610 Hz in intervocalic position, and 3720 Hz in postvocalic position (see Table 3). It could be claimed that there is no obvious correlation between F4 frequency and the quality of rounded vowels. However, the F4 formant of Lat. [r] and Lithuanian palatalized [rʲ] varies more consistently—it regularly decreases in the context of rounded vowels (in the case of Latvian) and fronted rounded vowels (in the case of Lithuanian).

The study of the formants of Baltic rhotic sonants in different vowel contexts showed that the dynamics of the first formant is influenced by the height of the following vowel (F1 decreases before high vowels and increases before low vowels), whereas the trajectory of the second and subsequent formants depends on the backness of the following vowel (back vowels lower F2, while front vowels raise it). It is noteworthy that fronted back vowels raise F2 less than front vowels do.

The results of the study, showing changes in the formants of Lithuanian and Latvian rhotic sonants [r], [rʲ] in different positions and in different vowel contexts, are summarized in Table 5:

118

Acta Linguistica Lituanica LXXXVIII

Lietuvių ir latvių kalbų virpamųjų sonantų formantės skirtingose pozicijose: akustinis ir statistinis tyrimas

5. TABLE. F1–F4 formants of contemporary Baltic rhotic sonants in different positions

Straipsniai / Articles

119

JOLITA URBANAVIČIENĖ, JANA TAPERTE

CONCLUSIONS

1. A comparison of the summarized formant values (see Table 5) established that the highest F1 value is characteristic of Lith. [r] and is associated with velarization, whereas the lowest F1 value was found for Lith. [rʲ], which is not characterized by velarization. The highest F2 value was calculated for Lith. [rʲ] and can be explained by secondary palatalization, whereas the lowest F2 formant is characteristic of unpalatalized Lith. [r]. The study confirmed that the structure of the first two formants of Latvian [r] is close to that of the neutral [ə], cf.: F1 = 510 Hz, F2 = 1430 Hz (see LVG 2013: 68), whereas the formants of Lithuanian rhotic sonants are polarized because of palatalization in Lithuanian.

2. Position in the sequence ([R]VR – V[R]V – RV[R]) influences the formant structure of Baltic rhotic sonants. In Lithuanian, the first formant depends particularly strongly on position, while the second formant is the least sensitive. In Latvian, the prevocalic, intervocalic, and postvocalic positions of sonant [r] have a statistically significant effect on all four formants. The results of the study allow us to state that the formants of Lith. [r] and [rʲ] change similarly when comparing the [R]VR – V[R]V – RV[R] positions: F1 decreases, while F2–F4 increase; the formants of Latvian [r] are highest in prevocalic position and lowest in postvocalic position.

3. The context of adjacent vowels has a similar effect on changes in the formants of Lithuanian and Latvian rhotic sonants: the first formant is associated with the height of the adjacent vowel (F1 decreases before high vowels and increases before low vowels), whereas the second and subsequent formants depend on the backness of the adjacent vowel (back vowels lower F2, while front vowels raise it). Only the F4 data of Lith. [r] did not confirm these correlations. The study showed that fronted back vowels raise the F2 of Lith. [rʲ] less than adjacent front vowels do.

The results of the study show that Lat. [r], according to the summarized formant values, occupies an intermediate position between unpalatalized Lith. [r] and palatalized Lith. [rʲ], and this conclusion correlates with the results of studies of consonants in other Baltic languages¹⁷. The study established that different positions (prevocalic,

¹⁷ Cf. the conclusions of studies of consonant locus levels in the Baltic languages, which state that “palatalized Lithuanian obstruents are located at the top of the coordinate plane according to the values of the F2 loci, unpalatalized obstruents in the lower part, while Latvian obstruents lie right in the center. Such F2 locus values can be justified from the articulatory perspective: the articulation of Latvian consonants occupies central positions, whereas the articulation of Lithuanian consonants is characterized by a certain polarization according to the presence or absence of an additional articulatory feature” (Jaroslavienė, Grigorjevs, Urbanavičienė, Indričāne 2019: 190).

120

Acta Linguistica Lithuanica LXXXVIII

Lietuvių ir latvių kalbų virpamųjų sonantų formantės skirtingose pozicijose: akustinis ir statistinis tyrimas

intervocalic, postvocalic) affects the formants of Lithuanian and Latvian rhotic sonants differently: when comparing the [R]VR – V[R]V – RV[R] positions, the formants of Latvian [r] decrease, whereas those of Lithuanian [r] and [rʲ] increase (except F1). However, the context of adjacent vowels has a similar effect on the formant structure of Baltic rhotic sonants.

LITERATŪRA

Ambrazevičius Rytis, Leskauskaitė Asta 2014: Priebalsių akustinės ypatybės: palatalizacija ir balsingumas, Kaunas: Technologija.

Bacevičiūtė Rima 2008: Fonetikos terminai: žodynėlis ir mokomosios užduotys, Vilnius: Vilniaus pedagoginio universiteto leidykla.

Boersma Paul, Weenink David 2018: Praat: Doing phonetics by computer (version 6.0.43). Prieiga internete: http://www.praat.org/.

Borden Gloria J., Harris Katherine S., Raphael Lawrence J. 1994: Speech Science Primer. Physiology, Acoustics, and Perception of Speech, ³rd edition, Baltimore: Williams & Wilkins.

Čeirane Solveiga, Indričāne Inese, Taperte Jana 2014: Locus equations for Latvian Consonants. – Linguistica Lettica 22, 29–47. Prieiga internete: http://www.lu-lavi.lv/zurnals-linguistica-lettica.

Čekanavičius Vydas, Murauskas Gediminas 2004: Statistika ir jos taikymai 2, Vilnius: TEV.

Dereškevičiūtė Sigita 2013: Lietuvių kalbos priebalsių akustinės ypatybės: humanitarinių mokslų daktaro disertacija, Vytauto Didžiojo universitetas, Lietuvių kalbos institutas.

DLKG – Dabartinės lietuvių kalbos gramatika, red. V. Ambrazas, 4-as patais. leidimas, Vilnius: Mokslo ir enciklopedijų leidybos institutas, 2005.

Girdenis Aleksas 2003: Teoriniai lietuvių fonologijos pagrindai, Vilnius: Mokslo ir enciklopedijų leidybos institutas.

Grigorjevs Juris 2012: Acoustic characteristics of the Latvian sonorants. – Baltistica 47(2), 267–292. Prieiga internete: http://www.baltistica.lt/index.php/baltistica/issue/view/106.

Straipsniai / Articles

121

JOLITA URBANAVIČIENĖ, JANA TAPERTE

IPA – International Phonetic Association, 2015. Prieiga internete: https://www.internationalphoneticassociation.org/content/full-ipa-chart.

Jaroslavienė Jurgita, Grigorjevs Juris, Urbanavičienė Jolita, Indričānė Inese 2019: Baltų kalbų garsynas XXI a. pradžioje: Balsių ir garsų sąveikos instrumentinis tyrimas: mokslo monografija su garsinėmis iliustracijomis, Vilnius: Lietuvių kalbos institutas. Prieiga internete: http://lki.lt/e-publikacijos/.

Johnson Keith 2003: Acoustic & auditory phonetics, 2^nd edition, Oxford: Blackwell Publishers.

Kazlauskienė Asta 2018: Bendrinės lietuvių kalbos fonetikos ir fonologijos pagrindai, Kaunas: Vytauto Didžiojo universitetas.

Kodzasov Sandro V., Krivnova Olga F. 2001: Кодзасов, Сандро В., Кривнова, Ольга Ф. Общая фонетика [Obščaja fonetika], Москва: Российский государственный гуманитарный университет [Moskva: Rossijskij gosudarstvennyj gumanitarnyj universitet].

Ladefoged Peter 2003: Phonetic data analysis: An introduction to fieldwork and instrumental techniques, Malden, MA & Oxford: Blackwell.

Ladefoged Peter, Maddieson Ian 1996: The sounds of the world’s languages, Oxford: Blackwell Publishers.

Laua Alise 1997: Latviešu literārās valodas fonētika, Rīga: Zvaigzne ABC.

Liepa Elmārs 1957: Daži mūsdienu latviešu literārās valodas fonēmu pareizrunas jautājumi: disertācija filoloģijas zinātņu kandidāta grāda ieguvei, Rīga: Rīgas Pedagoģiskais institūts.

Liepa Elmārs 1970: Skaņu kvantitāte. – Latviešu valodas struktūras jautājumi, Rīga: LVU, 39–88.

LVG – Latviešu valodas gramatika, aut. I. Auziņa, I. Brenķe, J. Grigorjevs, I. Indričānė, B. Ivulāne, A. Kalnača, L. Lauze, I. Lokmane, D. Markus, D. Nītina, G. Smiltniece, B. Valkovska, A. Vulāne, Rīga: LU Latviešu valodas institūts, 2013.

Maddieson Ian 1984: Patterns of Sounds, Cambridge: Cambridge University Press.

Markus Dace, Auziņa Ilze 2013: Fonētika. – Latviešu valoda, red. A. Veisbergs, Rīga: LU Akadēmiskais apgāds.

Markus Dace, Grigorjevs Juris 2002: Fonētikas pētīšanas un vizualizēšanas metodes, 1 grām., Rīga: Rasa ABC.

Mikalauskaitė Elzbieta 1975: Lietuvių kalbos fonetikos darbai, Vilnius: Mokslo.

Pakerys Antanas 2003: Lietuvių bendrinės kalbos fonetika, Vilnius: Enciklopedija.

122

Acta Linguistica Lithuanica LXXXVIII

Lietuvių ir latvių kalbų virpamųjų sonantų formantės skirtingose pozicijose: akustinis ir statistinis tyrimas

Scobbie James M. 2006: (R) as a variable. – Encyclopedia of Language & Linguistics, ^2nd edition, ed. K. Brown, vol. 10, Oxford: Elsevier, 337–344.

Solé Maria-Josep 2002: Aerodynamic characteristics of trills and phonological patterning. – Journal of Phonetics 30(4), 665–688.

Švageris Evaldas 2022: Eksperimentinė fonetika, Vilnius: Vilniaus universiteto leidykla.

Taperte Jana 2013: Latviešu valodas nazālo slēdzeņu akustiskās īpašības. – Valoda – 2013: valoda dažādu kultūru kontekstā 23, Daugavpils: DU akadēmiskais apgāds „Saule“, 281–289.

Taperte Jana 2014: Locus equations and the place of articulation for the Latvian sonorants. – Baltistica 49(1), 71–99. Prieiga internete: http://www.baltistica.lt/index.php/baltistica/issue/view/109.

Taperte Jana 2014a: Latviešu valodas laterālo spraudzeņu akustisks raksturojums informantu vecuma grupā no 16 līdz 39 gadiem. – Linguistica Lettica 22, 158–172.

Taperte Jana 2015: Spectral features of nasals in Standard Latvian. – Baltistica 50(1), 115–128. Prieiga internete: http://www.baltistica.lt/index.php/baltistica/issue/view/111.

Taperte Jana 2016: Acoustic properties of trill in Standard Latvian (data from speakers of different age groups). – International Scientific Conference dedicated to the 80th anniversary of birth of professor Aleksas Girdenis (1936–2011) „Transformations of contemporary Baltic languages and dialects“, 2016 October 19–21, Vilnius, Lithuania.

TARTIS – Mokomoji tarties ir kirčiavimo programa TARTIS. Prieiga internete: http://tartis.vdu.lt/fonetika-ir-tartis/igudziu-tobulinimas/garsu-ypatybes/.

Ulvydas et al. – Lietuvių kalbos gramatika 1: Fonetika ir morfologija (daiktavardis, būdvardis, skaitvardis, įvardis), red. K. Ulvydas, Vilnius: Mintis, 1965.

Urbanavičienė Jolita, Indričianė Inese, Jaroslavienė Jurgita, Grigorjevs Juris 2019: Baltų kalbų garsynas XXI a. pradžioje: Priebalsių instrumentinis tyrimas: mokslo monografija su garsinėmis iliustracijomis, Vilnius: Lietuvių kalbos institutas. Prieiga internete: https://doi.org/10.35321/e-pub.2.baltu-garsynas.

Vaitkevičiūtė Valerija 2001: Lietuvių kalbos tarties pagrindai ir žodynas, Vilnius: Pradai.

Vásquez Carranza, Luz Marina 2006: On the phonetic realization and distribution of Costa Rican rhotics. – Filología y Lingüística 32(2), 291–309.

Straipsniai / Articles

123

JOLITA URBANAVIČIENĖ, JANA TAPERTE

VLKK – Valstybinė lietuvių kalbos komisija. Prieiga: internete: https://vlkk.lt/vlkk-nutarimai/protokoliniai-nutarimai/rekomendacija-del-tarptautines-fonetines-abeceles-pritaikymo-bendrinės-lietuvių-kalbos-garsiniams-vienetams-žymėti [žiūrėta 2022–10-15].

Zinder Lev R. 1979: Зиндер, Лев Р. Общая фонетика [Obščaja fonetika], Москва: Высшая школа [Moskva: Vyššaja škola].

Lithuanian and Latvian Trills: An Acoustic and Statistical Study of Formant Structure

SUMMARY

The object of this study is the trills of contemporary Baltic languages produced in prevocalic (<R>VR), intervocalic (V<R>V) and postvocalic (RV<R>) positions. The aim of the paper is to analyze the structure of the first four formants, to compare their variation in the context of different vowels, to determine the statistical significance of the differences and to highlight the structure of formants in Lithuanian and Latvian. The data of 12 male speakers (6 native Lithuanian speakers and 6 native Latvian speakers aged 21–42) were analyzed; in total, 1458 realizations (810 Lithuanian and 648 Latvian) were studied. Formant values were measured using Praat, the obtained data were processed using Microsoft Excel, statistical analysis was performed using SPSS.

The following trends in the structure of F1–F4 have been determined:

1. According to the formant data, Lat. [r] occupies an intermediate position between non-palatalized Lith. [r] and palatalized Lith. [rʲ], and this conclusion correlates with the results of other studies of consonants in the Baltic languages.

2. Lith. [r] has the highest F1 value, which is associated with velarization, and the lowest F1 value is found in Lith. [rʲ]. Lith. [rʲ] has the highest F2 value, which can be explained by secondary palatalization, while the lowest F2 is characteristic of non-palatalized Lith. [r]. Lat. [r] occupies an intermediate position between non-palatalized Lith. [r] and palatalized Lith. [rʲ].

3. In Lithuanian, F1 values mainly depend on syllable position, and F2 values are the least variable. In Latvian, the syllable position of [r] has a statistically significant effect on all four formants.

124

Acta Linguistica Lithuanica LXXXVIII

Lietuvių ir latvių kalbų virpamųjų sonantų formantės skirtingose pozicijose: akustinis ir statistinis tyrimas

4. Vowel context has a similar effect on formant structure of trills both in Lithuanian and Latvian: F1 is related to the height of the adjacent vowel (F1 lowers before high vowels, rises before low ones), while F2, F3, F4 depend on the backness of the adjacent vowel (F2 is lower in the context of back vowels and is higher in the context of front vowels). These correlations were not confirmed only for the Lith. [r] F4 formant data.

5. It was confirmed that syllable position affects the formants of the Lithuanian and Latvian trills differently: when comparing the <R>VR – V<R>V – RV<R> positions, the formant frequencies of Lat. [r] descend, while the formant frequencies of Lith. [r] and [r]^ʲ], on the contrary, ascend (except F1).

Įteikta 2022 m. spalio 26 d.

JOLITA URBANAVIČIENĖ

Lietuvių kalbos institutas

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

jolita.urbanavič[email protected]

JANA TAPERTE

Latvijas universiteto Latvių kalbos institutas

Kalpaka bulvaris 4, LV-1050 Ryga, Latvija

[email protected]

Straipsniai / Articles

125