Dabartinių baltų kalbų pučiamieji priebalsiai ir afrikatos: akustinių požymių lyginamoji analizė

JOLITA URBANAVIČIENĖ Institute of the Lithuanian Language Research areas: experimental phonetics, speech acoustics, phonology, dialectology, lexicography, sociolinguistics. INESE INDRIČĀNE LU Institute of the Latvian Language Research areas: experimental phonetics, speech acoustics, phonology, dialectology. FRICATIVES AND AFFRICATES IN CONTEMPORARY BALTIC LANGUAGES: A COMPARATIVE ANALYSIS OF ACOUSTIC FEATURES Fricatives and Affricates of Contemporary Baltic Languages: Comparative Analysis of Acoustic/ Spectral Cues ABSTRACT The subject of this article is the fricatives and affricates of Standard Lithuanian and Standard Latvian, compared according to two acoustic features: spectral peak frequencies and relative intensity. To obtain objective results, the fricatives and affricates of both languages were analyzed in Lithuania and Latvia using the same methodology. The aim of the study is to compare the spectral characteristics of fricatives and affricates in contemporary Baltic languages and to determine how palatalization, voicing, consonant manner and place of articulation, and the informant’s sex affect the spectral peak frequencies and relative intensity of the consonants under study. It was found that in both Standard Lithuanian and Standard Latvian, the acoustic features of consonants correlate with the same criteria: spectral peak frequency correlates with consonant articulation (especially sibilance), palatalization (i-type; only in Lithuanian), and informant sex; relative intensity correlates with consonant voicing and palatalization (i-type; only in Lithuanian). 144 Acta Linguistica Lithuanica LXXV Fricative Consonants and Affricates in Modern Baltic Languages: A Comparative Analysis of Acoustic Features The differing acoustic-feature data obtained allow us to conclude that Latvian fricative consonants and affricates are closer to palatalized Lithuanian consonants in terms of spectral peak values; accordingly, Latvian consonants are “softer” than non-palatalized Lithuanian consonants. It was also established that Latvian fricatives and affricates have greater relative intensity than the corresponding Lithuanian palatalized and non-palatalized consonants (matched for voicing, manner of articulation, and place of articulation). These research results may have been due to differences in the informants’ voice loudness (whether inherent or related to fatigue), as well as to variations in sound-signal strength during recording. KEYWORDS: Standard Lithuanian, Standard Latvian, fricative consonants, affricates, spectral peak frequencies, relative intensity. ANNOTATION This current paper deals with the fricatives and affricates of standard Lithuanian and Latvian. In the framework of this research two acoustic cues of these consonants – spectral peak’s frequency and relative intensity – have been examined. The analysis of fricatives and affricates is performed according to one and the same approach both for standard Lithuanian and Latvian. The aim of this study is to compare the fricatives and affricates of contemporary Baltic languages in order to determine the essential spectral differences and the effect of palatalization, voicing, manner and place of articulation, as well as the effect of speaker’s gender displayed by spectral peak’s frequency and relative intensity. The results indicate that the acoustic cues examined in this research reveal similar effects both in standard Lithuanian and Latvian: spectral peak’s frequency is affected by palatalization (in Standard Lithuanian), manner and place of articulation, as well as by speaker’s gender, whereas relative intensity displays the effect of palatalization (in Standard Lithuanian) and voicing. The data obtained in this study show that the spectral peak’s frequency values calculated for the fricatives and affricates of Standard Latvian are more similar to the data of palatalized rather than non-palatalized consonants of corresponding voicing, manner and place of articulation in Standard Lithuanian, i.e., the fricatives and affricates in Standard Latvian seem to be “softer” than the non-palatalized fricatives and affricates in Standard Lithuanian. The relative intensity values for the fricatives and affricates of Standard Latvian are higher in comparison to both the palatalized and non-palatalized consonants of the same voicing, manner and place of articulation in Standard Lithuanian. This could be explained either by a more energetic articulation of these speech sounds for speakers of Standard Latvian or by some differences related to the recording level. KEYWORDS: Standard Lithuanian, Standard Latvian, fricatives, affricates, spectral peak’s frequency, relative intensity. Straipsniai / Articles 145 JOLITA URBANAVIČIENĖ, INESE INDRIČĀNE 1. INTRODUCTION In recent years, joint instrumental studies of the sound systems of Lithuanian and Latvian have intensified, conducted in the territories of those countries using the same methodology1. Worth mentioning are the articles by Juris Grigorjevs and Jurgita Jaroslavienė on the vocalism of modern Baltic languages: the spectral structure and duration of monophthongs (Grigorjevs, Jaroslavienė 2015a: 68–92), as well as monophthong quality (Grigorjevs, Jaroslavienė 2015b: 57–89). The consonant systems of Lithuanian and Latvian were studied by the authors of this article, Inese Indričānė and Jolita Urbanavičienė: the spectra of plosive consonants were analyzed (Indričānė, Urbanavičienė 2015: 51–87), as were the loci of obstruents using the locus-equation method (Urbanavičienė, Indričānė 2015: 261–293). The two modern Baltic languages share a number of features: the phonological opposition between short and long vowels, a large number of diphthongs, a pitch-accent system, and so on. Over many centuries, however, Lithuanian and Latvian have also developed fundamental differences that account for the distinct phoneme inventories and phonological classifications of the modern Baltic languages. For example, Latvian has fixed stress and a rich subsystem of palatal consonants, whereas Lithuanian has free stress, secondary (i-type) palatalization, and a phonological opposition between palatalized vs. non-palatalized consonants, among other features. Such synchronic comparative studies not only help highlight the sound-specific features of different languages, nuances in their phonological classifications, and the differing functional weight of the same phonological elements, but also reveal language-specific characteristics from a typological perspective. The objectof this study is the fricative consonants and affricates of Lithuanian and Latvian. Fricatives are consonants articulated with two speech organs brought so close together that the airflow passing between them creates friction noise, or frication (Crystal 2008: 199). It is precisely this relatively long interval of aperiodic energy that makes it possible to assign fricatives to a distinct sound class (Kent, Read 1992: 121). Affricates consist of plosive and fricative consonants that share the same place of articulation (Reetz, Jongman 2009: 16). Affricates are interpreted as sounds intermediate between a plosive and a fricative articulated at the same place (Ladefoged, Maddieson 1998: 90). Like plosives, affricates begin with a closure; when it is released, air continues to flow through the narrow passage between the articulators, producing frication. Therefore, not only the spectra of fricative consonants but also those of affricates exhibit 1 These studies were inspired by the 2013–2015 research-group project carried out at the Institute of the Lithuanian Language, “Modern Baltic Languages: Spectral Characteristics of Sounds (An Instrumental Study)” (No. MIP-081/2013, led by Dr. Jurgita Jaroslavienė). 146 Acta Linguistica Lithuanica LXXV Dabartinių baltų kalbų pučiamieji priebalsiai ir afrikatos: akustinių požymių lyginamoji analizė The frication interval is similar, only shorter (in the case of fricatives, frication in fact occupies the entire duration of the fricative). Although the spectral characteristics of affricates are the same as those of the corresponding plosives and fricatives in terms of voicing and place of articulation (Kodzasov, Krivnova 2001: 177), “all the methods used to study fricatives can be applied to the differentiation […] of affricates” (Ambrazevičius, Leskauskaitė 2014: 131). Therefore, this article examines two closely related consonant classes—fricatives and affricates—using the same acoustic features. The theoretical literature indicates that fricatives can be characterized by the duration of frication, their spectral properties and intensity, and the properties of spectral formant transitions (locus equations) (Reetz, Jongman 2009: 189). The aim of this article is to investigate the spectral properties of fricatives and affricates in the modern Baltic languages and to identify their differential distinctions in Lithuanian and Latvian. This aim can be divided into the following objectives: 1) to determine the spectral peak frequency values of fricatives and affricates in Lithuanian and Latvian; 2) to experimentally analyze the relative intensity of fricatives and affricates in both languages; 3) to compare the effects of consonant palatality, voicing, and manner and place of articulation on the spectral features of fricatives and affricates. The consonant data from male and female speakers were analyzed separately to determine the effect of sex on the spectral properties of consonants. The study is new and relevant because no comparative studies of the acoustic features of Lithuanian and Latvian consonants conducted using the same methodology have been published to date. Furthermore, the methods for studying consonant acoustic features described in this article have not previously been used in Lithuanian acoustic phonetics. 2. ARTICULATORY CLASSIFICATION OF FRICATIVES AND AFFRICATES IN LITHUANIAN AND LATVIAN The article uses the International Phonetic Alphabet, widely used around the world, to transcribe the consonants examined in the modern Baltic languages2 (TFA or IPA – International Phonetic Alphabet), which was developed by the International Phonetic Association 2 This study uses the IPA symbols adapted for Lithuanian and Latvian consonants during the project “Modern Baltic Languages: Spectral Characteristics of Sounds (An Instrumental Study)” (No. MIP-081/2013). Straipsniai / Articles 147 JOLITA URBANAVIČIENĖ, INESE INDRIČĀNE (for more details, see IPA 2015). The table below (see Table 1) presents those fricatives and affricates whose IPA transcription differs from that of the national phonetic alphabets. 1 TABLE. Symbols for fricative and affricate phonemes in Lithuanian and Latvian according to the national (LTFA and LVFA) and international (IPA) phonetic alphabets LT graphemeLTFA phonemeIPA phonemeLV graphemeLVFA phonemeIPA phonemevvʋ1vvvjjjjjjchxxhxxh2hɣ–––ššʃššʃžžʒžžʒcctscctsččtɕččtɕdzʒdʑdzʒ3dʑdžʒdʒdžʒ4dʒ This article analyzes 34 consonant phonemes of the modern Baltic languages, classified by manner of articulation into: 1) fricatives – 22 phonemes: a) Lith. /f, fʲ, s, sʲ, z, zʲ, ʃ, ʃʲ, ʒ, ʒʲ, x, xʲ, ɣ, ɣʲ/ – 14 phonemes; b) Latv. /f, v, s, z, ʃ, ʒ, ʝ, x/ – 8 phonemes; 3 For the different phonological interpretations of the sonorants v and j in Lithuanian and Latvian, see Section 3. 4 The Latvian consonant system has no voiced velar consonant h, but the grapheme h is used in the following cases: 1) to denote the primary variant of the phoneme /x/, the velar fricative [x]; 2) to denote an allophone of the phoneme /x/, the pharyngeal fricative [ħ]; 3) to denote an allophone of the phoneme /x/, the palatal fricative [ç] (LVG 2013: 70). The allophones of the phoneme /x/—the pharyngeal [ħ] and palatal [ç]—are most often pronounced under the influence of English and German. 5 In the Latvian national transcription (LVFA), the affricate dz may be represented as either dz [d͡z̥sma] or ʒ [ʒi̯esma] (LVG 2013: 26). 6 In the Latvian national transcription (LVFA), the affricate dž may be represented as either dž [d͡ʒems] or ǯ [ǯems] (LVG 2013: 26). 7 In the International Phonetic Alphabet, the superscript modifier “j” written after a consonant denotes a palatalized consonant, for example, [tʲ], [dʲ]. 148 Acta Linguistica Lithuanica LXXV Dabartinių baltų kalbų pučiamieji priebalsiai ir afrikatos: akustinių požymių lyginamoji analizė 2) affricates – 12 phonemes: a) Lith. /t͡s, t͡sʲ, dz, dzʲ, t͡ʃ, t͡ʃʲ, d͡ʒ, d͡ʒʲ/ – 8 phonemes; b) Lat. /t͡s, dz, t͡ʃ, d͡ʒ/ – 4 phonemes. According to place of articulation, the Baltic consonants under study are divided into the following classes (see Urbanavičienė, Indričānė 2015: 265–267; cf. DLKG 2005: 33–34; LVG 2013: 58): 1) labiodental: Lith. /f, fʲ/ and Lat. /f, v/; 2) dental: Lith. /s, sʲ, z, zʲ, t͡s, t͡sʲ, dz, dzʲ/ and Lat. /s, z, t͡s, dz/; 3) alveolar: Lith. /ʃ, ʃʲ, ʒ, ʒʲ, t͡ʃ, t͡ʃʲ, d͡ʒ, d͡ʒʲ/ and Lat. /ʃ, ʒ, t͡ʃ, d͡ʒ/; 4) palatal: Lat. /j/; 5) palatalized velar (palatovelar): Lith. /xʲ, ɣʲ/; 6) velar: Lith. /x, ɣ/ and Lat. /x/. The number of Lithuanian consonant phonemes under study (22 phonemes) is almost twice that of Latvian (12 phonemes), because Lithuanian has consonant oppositions based on voicing (voiceless – voiced) and palatalization (non-palatalized – palatalized), whereas Latvian has oppositions based only on voicing. 3. THE PHONOLOGICAL STATUS OF CONSONANTS /v/ AND /j/ IN LITHUANIAN AND LATVIAN By manner of articulation, Lithuanian /ʋ, ʋʲ, j/ are classified as approximants (and are therefore excluded from the scope of this study), whereas Latvian /v, j/ are classified as fricative consonants. Instrumental research has established that the voiced labiodental /v/ and palatal /j/ in Latvian have several pronunciation variants (Čेirane 2011). When the gap between the speech organs is narrow in the articulation of these consonants, fricatives [v] or [j] are produced—the airstream flowing through the constriction creates turbulence and gives the constriction consonant a noise-frication component (see LVG 2013: 66). When articulating /v/ and /j/, if the gap between the speech organs is more open, the consonants are produced without frication, i.e. as non-fricatives, or approximants [ʋ] and [j]⁸ (see LVG 2013: 66). In Latvian, approximants [ʋ, j] may be produced instead of fricatives [v, j] when speech is subdued or less energetic (LVG 2013: 68; 71). In tautosyllabic position after a short or long vowel, Latvian consonant phonemes [v, j] undergo complete or partial vocalization, e.g.: [tɐus] tavs 8 Both realizations of the phonemes /v/ and /j/ (fricatives [v], [j] and approximants [ʋ] and [j]) are described in Latvian-language research by the term “noisy constriction consonant” (Latvian spraudzīens troksnenis) (cf. Čेirane 2010; Čेirane 2011; Čेirane, Indričāne 2012; Markus, Grigorjevs 2002). Straipsniai / Articles 149 JOLITA URBANAVIČIENĖ, INESE INDRIČĀNE ‘yours’, [sɑus] sɑvs ‘one’s own’, [dzeiniek͡s] dzeinieks ‘poet’, [zveiniek͡s] zveinieks ‘fisherman’; [pɑːrdevɛːis] pārdevējs ‘seller’, [teːus] tēvs ‘father’, [pɑːrmī̯nieks] pārmi̯nieks ‘hearth-keeper’; [tuːs] tuos ‘near’ (LVG 2013: 86). In traditional Lithuanian phonetics9 (Pakerys 2003) and in the academic grammar (DLKG 2005: 33), [j], [ʋ], [ʋʲ] are classified as approximants (sonorants). This interpretation of [ʋ, ʋʲ, j] in Lithuanian was determined not by articulation (the nature of the constriction is the same as for other fricatives10), but by the functional properties of these consonants: at the end of a word, [ʋ, j] lose their friction, become even more voiced, and turn into nonsyllabic [u̯, i̯], forming secondary diphthongs with the preceding vowels, e.g.: [jūːro̯i] ‘in the sea’, [kėːdėːi] ‘on the chair’, [sudie̯u] ‘goodbye’. Of the two positional variants—fricatives [v, j] and approximants [ʋ, j]—the fricatives /v, j/ are treated as phonemes in Latvian (see LVG 2013: 76–77), while approximants /ʋ, ʋʲ, j/ are treated as phonemes in Lithuanian. The distinction between approximants [ʋ ʋʲ] and fricatives [v, vʲ] is difficult to perceive, because voiced fricatives, owing to their specific articulation (when voiced fricatives are produced, frication noise is generated in the vocal tract while the vocal folds vibrate, impeding airflow), tend to become approximants11. The problem of identifying consonants /ʋ, ʋʲ, j/ in Lithuanian has not been fully resolved and requires detailed instrumental evidence (based on an objective study of the articulatory properties of speech sounds). This article follows the traditional interpretation; therefore, Lithuanian sonorants /ʋ, ʋʲ/ (so Lithuanian labiodental /f/ and /fʲ/ have no voiced counterpart) and /j/ are excluded from the study of fricatives, whereas voiced fricatives /v, j/ are included in the Latvian data (where the /f/ and /v/ opposition exists). 4. FRICTION SPECTRUM AND INTENSITY From the perspective of voicing, otherwise identical fricative consonants differ in the distribution of spectral energy across the spectrum and in intensity (Kent, Read 1992: 122). These acoustic 9 In recent classifications of Lithuanian consonants (see Abrazevičius, Leskauskaitė 2014: 167; Pakerys 2014: 91), consonants /v, vʲ/ are considered not sonorants but voiced labiodental fricatives, which contrast in voicing with their corresponding voiceless counterparts /f, fʲ/. 10 E. Mikalauskaitė (1975: 41–46) lietuvių /ʋ/, /ʋʲ/ ir /j/ priskyrė pučiamiesiems, o akademinė Lietuvių kalbos gramatika (LKG 1965: 70) šiuos priebalsius, kaip turinčius didelį kiekį virpesių, neturinčius tono moduliacijos, nesukeliančius priebalsių skardėjimo, laiko tarpiniais tarp pučiamųjų priebalsių ir sonantų. 11 Voiced fricative consonants are a relatively rare phenomenon in the world’s languages (Johnson 2003: 122). 150 Acta Linguistica Lithuanica LXXV Dabartinių baltų kalbų pučiamieji priebalsiai ir afrikatos: akustinių požymių lyginamoji analizė On the basis of these features, theoretical literature distinguishes two groups of fricative consonants: sibilants (sibilant) (for example, English alveolar fricatives /s/, /z/ and postalveolar fricatives /ʃ/, /ʒ/) and non-sibilants (non-sibilant) (for example, English labiodental fricatives /f/, /v/, interdental fricatives /θ/, /ð/, and the pharyngeal fricative /h/). Sibilants differ from non-sibilants in having greater spectral energy and intensity (Kent, Read 1992: 122): sibilants are also called “bright,” whereas non-sibilants are called “dull” consonants (Ambrazevičius 2011: 37). Earlier studies of Lithuanian and Latvian phonetics did not generally classify consonants by sibilance, i.e. according to the direction of the airstream toward the teeth. In the most recent study of the acoustics of Lithuanian consonants (see Ambrazevičius, Leskauskaitė 2014: 301), the Lithuanian consonants /s, sʲ, z, zʲ, ʃ, ʒ, ʒʲ/ and affricates /ts, tsʲ, dz, dzʲ, tʃ, tʃʲ, dʒ, dʒʲ/ are classified as sibilants. In Latvian, the fricatives /s, z, ʃ, ʒ/ and affricates (stop consonants) /ts, dz, tʃ, dʒ/ are likewise regarded as sibilants (cf. I. Indričāne’s studies of sibilants and non-sibilants among Latvian voiceless obstruents; see Indričāne 2012; 2013; 2014). In this article, the terms “sibilant” and “non-sibilant” will also be used where appropriate. An examination of the spectral-peak intensity of Latvian prevocalic voiced fricatives in CVC sequences found that, in this group, higher mean intensity values are characteristic of the dental [z] and alveolar [ʒ], while lower values are characteristic of the labiodental [v] and palatal [j]. The labiodental [v] has the lowest mean intensity value, and the alveolar [ʒ] the highest, although the latter differs little from the intensity value of the dental [z]. When standard deviation values are taken into account, the dental [z] and alveolar [ʒ] can be distinguished from the labiodental [v], whereas the palatal [j] has relatively large standard deviations, so its range of values overlaps with those of both the labiodental [v] and the dental [z] (Čeirane 2011: 117–119). The spectral-peak intensity, mean intensity, and relative intensity of Latvian prevocalic voiceless fricatives in CVC sequences were also examined. The various methods yielded consistent results: higher mean intensity values are characteristic of the dental [s] and alveolar [ʃ], while lower values are characteristic of the labiodental [f] and velar [x]. In all cases, the labiodental [f] has the lowest mean intensity value, while the highest belongs to the alveolar [ʃ] (for spectral-peak intensity) or the dental [s] (for mean and relative intensity). Within the group of voiceless consonants, the intensity values of the dental [s] and alveolar [ʃ] are also very similar. The labiodental [f] stands out from the other voiceless fricatives with its lower intensity value, while the intensity value of the velar [x] is close to those of the dental [s] and alveolar [ʃ]. A comparison of standard deviations shows that the intensity of voiceless fricatives Straipsniai / Articles 151 JOLITA URBANAVIČIENĖ, INESE INDRIČĀNE verčių sritys visais atvejais persidengia (žr. Indričāne 2014: 112, 114–115; 1, 4 ir 7 grafikai). Studies of Latvian prevocalic voiced fricatives (see Čeirane 2011; Indričāne 2014: 101–119) have identified similar trends: in both the voiced and voiceless fricative groups, the labial fricative (a non-sibilant) has the lowest intensity, while a dental or alveolar consonant (a sibilant) has the highest. Moreover, mean intensity values show that the intensity of labiodental, palatal, and velar fricatives (non-sibilants) is lower than that of dental and alveolar fricatives (sibilants). Thus, the results of these studies accord with observations in the theoretical literature, for example, regarding intensity differences between sibilants and non-sibilants in English12. Very few studies of the intensity of Lithuanian consonants have been conducted to date. Sigita Dereškevičiūtė’s Sigita Dereškevičiūtė’s dissertation Acoustic Properties of Lithuanian Consonants (2013: 138) presents data on the spectral-peak frequencies and regions of energy concentration of Lithuanian fricatives: more intense spectral peaks are characteristic of palatalized consonants (especially [xʲ], [ʋʲ]), while higher characteristic frequencies and greater intensity are found in voiceless consonants. The author’s research also shows an increase in spectral energy for sibilants (especially [ʃ, ʃʲ, ʒ, ʒʲ, ʤ, ʤʲ]), associated with two-place articulation, and a lowering of the regions of energy concentration for labial consonants, associated with labial articulation. The intensity of fricatives in the modern Baltic languages is investigated by determining the intensity value of the dominant spectral peak (peak intensity/ amplitude), the mean intensity of the fricative (overall noise intensity/ amplitude), or relative intensity (relative intensity/ amplitude). The latter measure is considered particularly promising in studies of consonant acoustics because, for example, it makes it possible to classify all English fricatives by place of articulation (see Jongman et al. 2000: 1252, 1259–1262; Reetz, Jongman 2009: 192; Indričāne 2014: 103). Theoretical literature notes that voiced fricatives have lower intensity than their voiceless counterparts, because in articulating a voiced fricative more 12 Studies of English consonants show, for example, that the intensity of voiceless sibilants /s/, /ʃ/ is 10–15 dB greater than that of voiceless non-sibilants /f/, /θ/ (Reetz, Jongman 2009: 192). The intensity of the voiced alveolar fricative /z/ is 6 dB greater than that of the voiceless alveolar /s/, while the intensity of the voiced labiodental fricative /v/ is 13 dB greater than that of the voiceless labiodental fricative /f/. It is also evident that, among fricatives matched for voicing, sibilants have greater intensity than non-sibilants: the intensity of the voiced sibilant /z/ (70 dB) is greater than that of the voiced non-sibilants /v/, /ð/ (66 dB), while the intensity of the voiceless sibilants /s/, /ʃ/ (64–65 dB) is greater than that of the voiceless non-sibilants /f/, /θ/ (53–54 dB) (Jongman 1989: 1724; Kent, Read 1992: 122). 152 Acta Linguistica Lithuanica LXXV Dabartinių baltų kalbų pučiamieji priebalsiai ir afrikatos: akustinių požymių lyginamoji analizė energy is expended when the airstream flows through vibrating vocal folds (Kent, Read 1992: 122; Reetz, Jongman 2009: 192; Ambrazevičius 2011: 33). 5. RESEARCH MATERIAL AND INFORMANTS The research material for the article consists of the affricates and fricative consonants of both contemporary Baltic languages, pronounced in isolated CVC sequences in prevocalic position (both before short and before long vowels), e.g.: – 1ie. sis [sˈis], sès [sˈɛs], sàs [sɐs], sòs [sɔs], sùs [sʊs], sýs [sˈiːs], sẽs [sˈeːs], sės [sˈæːs], sãs [sɐːs], sõs [sɔːs], sǖs [suːs], siòs [sˈɔs], siõs [sˈoːs], siùs [sˈʊs], siũs [sˈuːs]; – 1a. sis [sis], ses [ses], ses [sæs], sas [sas], sos [sɔs], sus [sus], sīs [siːs], sēs [seːs], sẽs [sæːs], sās [sɑːs], sōs [sɔːs], sūs [suːs]. The influence of pitch accent was eliminated from this study on the assumption that pitch accents affect the end of the vocalic segment and the postvocalic consonant more than the prevocalic position. The material for the instrumental study was recorded in Lithuania and Latvia, using the same informant selection criteria. The same equipment was used for the recordings, and its technical parameters were standardized. The target sequences were read by 12 Lithuanian and 12 Latvian informants: 6 men and 6 women in each group (aged 20–48). All informants are native speakers of Lithuanian or Latvian. Their pronunciation is free of noticeable defects and conforms to the norms of standard Lithuanian or Latvian. The informants pronounced the target CVC sequences 3–5 times, with neutral intonation and at as uniform a pace as possible. The research material was recorded using a portable high-resolution digital recorder Tascam DR-100MK II and a vocal microphone AKG C 520. The audio signal was digitized at 44,100 Hz, with a quantization level of 16 bits. Data for men and women were analyzed separately. The instrumental study of fricative consonants and affricates was conducted using the computer programs PRAAT (version 5.3.60, authored by Paul Boersma and David Weenink), MS EXCEL (version 13, Microsoft Corporation) and SPSS (version 19, IBM Corporation). Straipsniai / Articles 153 JOLITA URBANAVIČIENĖ, INESĖ INDRIČANĖ 6. RESEARCH METHODOLOGY AND RESULTS 6.1. Spectral peak frequency 6.1.1. Determining the spectral peak frequency value To determine the spectral peak frequency value (Hz), FFT spectra of all the consonants under analysis are first obtained using the PRAAT sound analysis program13. Position the computer mouse in the middle of the noise interval of the fricative / affricate, click Select, choose the option Move start of selection by, select a distance of 0.01 s, click the options Spectrum – View spectral slice – View – Zoom, and set the range from 0.0 to 8000.0 Hz14. In order for the number of spectral elements to be approximately equal and for the FFT spectra to be comparable, spectra of all the consonants studied were obtained from intervals of the same duration—0.01 s, or 10 ms (see Figs. 1–4). In most cases, the segment from which this interval was taken exceeds 10 ms in duration, so the onset of the vowel, which could affect the energy distribution in the FFT spectrum, is not included in the interval under study. The 10 ms interval for fricatives was extracted approximately from their middle portion (Figs. 1–2), and for affricates, approximately from the middle of the release phase (Figs. 3–4). 0.01 s, or 10 ms, intervals FIG. 1. Voiced fricative [z] FIG. 2. Voiceless fricative [s] 13These are so-called unsmoothed spectrograms, or the fast Fourier transform, Eng. Fast Fourier Transform, abbreviated as FFT spectra. 14 Cf.: H. Reetz and A. Jongman argue that the energy characterizing the sound wave is distributed below 8000 Hz in the spectrum, while energy at higher frequencies is considered background noise (Reetz, Jongman 2009: 160). 154 Acta Linguistica Lithuanica LXXV Dabartinių baltų kalbų pučiamieji priebalsiai ir afrikatos: akustinių požymių lyginamoji analizė FIG. 3. Voiced affricates [dz] FIG. 4. Voiceless affricates [ts] The frequency value was measured at the dominant, i.e., highest, peak in the spectrum (Figs. 5–6). Determining the frequency value (Hz) of the spectrum peak FIG. 5. FFT spectrum of the voiced fricative [z] FIG. 6. FFT spectrum of the voiceless fricative [s] The frequency values (Hz) of the spectrum peaks determined using PRAAT were transferred to EXCEL and subsequently processed with the statistical analysis program SPSS, which calculated the mean statistical values (MSV) of the spectrum peak frequencies. 6.1.2. Spectrum peak frequency of fricatives and affricates in Lithuanian and Latvian The MSVs of the spectrum peak frequency of fricatives and affricates in Lithuanian and Latvian obtained in the study are presented in Table 2. Straipsniai / Articles 155 JOLITA URBANAVIČIENĖ, INESE INDRIČĀNE TABLE 2. Fricatives and affricates in Lithuanian and Latvian: spectral peak frequency SPF (Hz) C | Male data | Female data | Lith. | Lat. | Lith. | Lat. | Fricatives | [ʋ] | — | 2414 | — | 3056 | [f] | 1686 | 4466 | 3180 | 6770 | [fʲ] | 3043 | — | 3183 | — | [s] | 5924 | 5715 | 5601 | 6731 | [sʲ] | 5805 | — | 5654 | — | [z] | 5230 | 5298 | 5288 | 6740 | [zʲ] | 5714 | — | 5749 | — | [ʃ] | 2436 | 3055 | 4451 | 4326 | [ʃʲ] | 3031 | — | 4523 | — | [ʒ] | 2640 | 3189 | 3677 | 4443 | [ʒʲ] | 3166 | — | 4256 | — | [j] | — | 3935 | — | 4224 | [x] | 946 | 1797 | 1023 | 2105 | [xʲ] | 2896 | — | 3704 | — | [ɣ] | 1010 | — | 1053 | — | [ɣʲ] | 2480 | — | 3678 | — | Affricates | [t͡s] | 4402 | 5763 | 5326 | 6560 | [t͡sʲ] | 5911 | — | 6030 | — | [dz] | 5497 | 5768 | 5188 | 6775 | [dzʲ] | 6040 | — | 5718 | — | [t͡ʃ] | 3241 | 2925 | 4601 | 4379 | [t͡ʃʲ] | 3079 | — | 4331 | — | [dʒ] | 2991 | 3129 | 4540 | 5040 | [dʒʲ] | 3452 | — | 4747 | — 156 Acta Linguistica Lithuanica LXXV Dabartinių baltų kalbų pučiamieji priebalsiai ir afrikatos: akustinių požymių lyginamoji analizė The consonants of Lithuanian and Latvian, arranged in ascending order of spectral peak frequency, occur in the following sequence: 1) male data: 1ith. [x]<[ʋ]<[f]<[ʃ]<[ʋʲ]<[ʒ]<[xʲ]<[dʒ]<[ʃʲ]<[fʲ]<[tʃʲ]<[ʒʲ]<[tʃ]<[dʒʲ]<[ts]<[z]<[dz]<[zʲ]<[sj]<[tsj]<[s]<[dzj], 1at. [x]<[v]<[tʃ]<[ʃ]<[dʒ]<[ʒ]<[j]<[f]<[z]<[s]<[ts]<[dz]; 2) female data: 1ith. [x]<[ʋ]<[f]<[fʲ]<[ʒ]<[ɣʲ]<[xʲ]<[ʒʲ]<[tʃʲ]<[ʃ]<[ʃʲ]<[dʒ]<[tʃ]<[dʒʲ]<[dz]<[z]<[ts]<[s]<[sj]<[dzj]<[zj]<[tsj], 1at. [x]<[v]<[j]<[ʃ]<[tʃ]<[ʒ]<[dʒ]<[ts]<[s]<[z]<[f]<[dz]. 6.1.3. Comparison of palatalized and non-palatalized consonants in Lithuanian with consonant data from Latvian The spectral peak frequencies (SPF) of palatalized and non-palatalized Lithuanian consonants are compared with Latvian data in Figures 7–8. FIG. 7. Comparison of the mean statistical values of the spectral peak frequencies (Hz) of palatalized and non-palatalized Lithuanian consonants with Latvian results: male data Latvian consonants | Lithuanian consonants | palatalized | non-palatalized | ● | ○ | ● Straipsniai / Articles 157 JOLITA URBANAVIČIENĖ, INESE INDRIČANĖ FIG. 8. Comparison of mean statistical values (Hz) of the spectral peak frequencies of palatalized and non-palatalized Lithuanian consonants with the Latvian results: female data Latvian consonants | Lithuanian consonants | palatalized | non-palatalized | ▲ | △ | ▲ The study results show that the spectral peak frequencies (VSV) of palatalized Lithuanian consonants are generally higher than those of corresponding non-palatalized Lithuanian and Latvian consonants (in terms of voicing, place of articulation, and manner of articulation). This pattern is observed among velar consonants (data for men and women), alveolar and dental consonants, and affricates (data for men only). However, the opposite tendency was also observed: the spectral peak frequencies (VSV) of palatalized Lithuanian consonants are lower than those of the corresponding Latvian consonants; compare: in the male data—Lith. [fʲ] < Latv. [f], Lith. [ʒʲ] < Latv. [ʒ], Lith. [ʃʲ] < Latv. [ʃ]; in the female data—Lith. [fʲ] < Latv. [f], Lith. [zʲ] < Latv. [z], Lith. [sʲ] < Latv. [s], Lith. [džʲ] < Latv. [dž], Lith. [t͡sʲ] < Latv. [t͡s], Lith. [ʒʲ] < Latv. [ʒ], Lith. [dʒʲ] < Latv. [dʒ], Lith. [tʲ] < Latv. [t]. Moreover, in many cases, the spectral peak frequencies of palatalized Lithuanian alveolar consonants differ from the values for corresponding non-palatalized Lithuanian consonants and are closer to the spectral peaks of Latvian consonants. This can be explained by the fact that the pronunciation of Latvian alveolar fricatives and affricates involves an additional articulation—palatalization, i.e., raising the middle part of the tongue toward the hard palate (cf. LVG 2013: 64; 158 Acta Linguistica Lithuanica LXXV Dabartinių baltų kalbų pučiamieji priebalsiai ir afrikatos: akustinių požymių lyginamoji analizė 66); however, in Latvian, unlike in Lithuanian, this additional palatalization of consonants has no phonological significance. Among labial consonants, the MSVs of the spectral peak frequencies of palatalized Lithuanian consonants (in both the male and female data) are lower than those of the corresponding non-palatalized consonants, but higher than the spectral peak frequencies of Latvian labial consonants. The mean statistical values of the spectral peak frequencies of voiceless labiodental consonants—Lith. [f]/[fʲ] and Latv. [f]—differ from one another more than those of other consonants (see Figs. 7–8): the palatalized Lithuanian [fʲ] has a considerably lower mean statistical value for its spectral peak frequency than the Latvian [f]; consequently, in Latvian the spectral peak of this consonant is recorded in the high-frequency range, whereas in Lithuanian it is in the lower-frequency range. Larger differences in spectral peak frequencies were also recorded in the subgroups of velar and palatalized velar (palatovelar) consonants. Since the MSVs of the spectral peak frequencies for the Latvian consonant [x] were calculated together with palatovelar/palatal variants15, the spectral peak frequency value recorded for the non-palatalized Latvian velar [x] is higher than those of the non-palatalized Lithuanian velars [x], [ɣ], but lower than those of the palatalized Lithuanian velars and the Latvian palatal consonant (see Figs. 7–8). In ascending order of spectral peak frequency MSVs, the Lithuanian and Latvian velars, palatalized Lithuanian velars, and Latvian palatal consonant are arranged as follows: Lith. velars [x], [ɣ] < Latv. velar [x] < Lith. palatalized velars [xʲ], [ɣʲ] < Latv. palatal [j] Although the patterns of change in the MSVs of spectral peak frequencies for non-palatalized and palatalized Lithuanian consonants have many exceptions, in most cases the spectral peaks of palatalized consonants are recorded in higher frequency ranges than those of the corresponding non-palatalized consonants, while the spectral peaks of Latvian palatal consonants are in higher frequency ranges than those of palatalized Lithuanian consonants. 6.1.4. Comparison of voiceless and voiced consonants In neither Lithuanian nor Latvian does consonant voicing affect the frequency values of spectral peaks—the study results show that the spectral-peak frequency values (SPF) of voiceless consonants and their voiced counterparts vary unsystematically, with no discernible patterns. 15 In Latvian, before the front vowels [i], [iː], [e], [eː] and diphthongs beginning with the components [i] and [e], the fricative phoneme /x/ can be realized in two ways: as a palatovelar allophone [xʲ] and as a palatal allophone [ç] (LVG 2013: 69). Straipsniai / Articles 159 JOLITA URBANAVIČIENĖ, INESE INDRIČĀNE 6.1.5. Comparison of data for men and women Data for both Lithuanian and Latvian show that consonants produced by men have lower spectral-peak frequency values than the corresponding consonants produced by women. Thus, spectral-peak frequency values are associated with the speaker’s sex. The Lithuanian palatalized consonants [sʲ], [dʲ] and non-palatalized consonants [s], [d]16 do not follow this trend: these consonants produced by men have higher spectral-peak frequency values (SPF) than the corresponding consonants produced by women. 6.1.6. The influence of consonant manner/place of articulation on consonant acoustic properties The concentration of spectral energy in particular frequency bands provides information about the articulatory characteristics of sounds. If a consonant’s energy is concentrated in the 5000–6000 Hz region, a high-pitched s-type fricative is heard; if the energy is concentrated in the 3000–4000 Hz range, a low-pitched, hushing sh-type sound with intense noise is heard (Kodzasov, Krivnova 2001: 565–566). On the basis of these criteria, Lithuanian and Latvian consonants can first be divided into non-sibilants (whose energy is concentrated in the ~1000–3000 Hz region), namely Lithuanian [x], [xʲ], [ɣ], [ɣʲ] and Latvian [v], [j], [x], and sibilants (whose spectral-peak frequencies are concentrated in the 3000–6000 Hz range). Sibilants can be further subdivided into alveolar consonants (with spectral peaks located in the 3000–4000 Hz region), namely Lithuanian [ʃ], [ʃʲ], [ʒ], [ʒʲ], [tʃ], [tʃʲ], [dʒ], [dʒʲ] and Latvian [ʃ], [ʒ], [tʃ], [dʒ], and dental consonants (whose greatest spectral energy is concentrated in the 5000–6000 Hz range), namely Lithuanian [s], [sʲ], [z], [zʲ], [ts], [tsʲ], [dz], [dzʲ] and Latvian [s], [z], [ts], [dz]. Among the non-sibilants, the velars Lithuanian [x], [xʲ], [ɣ], [ɣʲ] and Latvian [x] stand out (with Latvian labiodental [v] also grouping with them), as their spectral peaks are concentrated in the ~1000–3000 Hz range. Nevertheless, the classification of consonants as sibilants and non-sibilants is related primarily not to spectral energy but to intensity. Sibilants are characterized by greater intensity because, in producing them, the airstream must pass through an obstruction—the teeth. There is no direct correlation between sibilance and spectral energy; compare: non-sibilants 16 The specific characteristics of the consonants [s] and [dz] may also have led to data error. Usually, only one spectral peak for consonants [s] and [dz] is visible in the 0–8000 Hz range. When a wider range is examined (for example, 0–10,000 Hz), two spectral peaks for these consonants may be visible (as with [ʃ]). Since men’s voices are characterized by lower frequencies, the second (higher) spectral peak may also have appeared within the narrower 0–8000 Hz range. Women’s produced sounds, however, are characterized by higher frequencies, so the value of the first peak was always assigned to the consonants under study (since the second peak was not visible in the 0–8000 Hz range). 160 Acta Linguistica Lithuanica LXXV Dabartinių baltų kalbų pučiamieji priebalsiai ir afrikatos: akustinių požymių lyginamoji analizė The spectral peaks of labiodental Lith. [f], [fʲ] are concentrated in the 2000–3200 Hz range; Latv. [f] in the 4500–7000 Hz range; and palatal [ɟ] in the ~4000 Hz range. In summary, the regions in which spectral peaks are concentrated allow fricatives and affricates to be differentiated into the following articulatory groups: alveolar sibilants (their spectral energy is concentrated in the mid-frequency range), dental sibilants (their energy is concentrated in the high-frequency range), and velar non-sibilants (characterized by the lowest spectral peak values, VSV). Other consonants (non-sibilants) cannot be differentiated by spectral peak frequencies. 6.1.7. Summary of the spectral peak frequency study The study compared spectral peak frequency values for fricatives and affricates in Lithuanian and Latvian with respect to voicing, palatalization, manner and place of articulation, and informants’ sex. It was found that consonant voicing does not affect spectral peak frequency values in either Lithuanian or Latvian (cf. S. Dereškevičiūtė’s (2013: 138) findings, which show that the energy of voiced Lithuanian consonants is concentrated in the low-frequency range compared with that of their voiceless counterparts). Due to palatalization, the spectral peak frequencies of Lithuanian consonants increase, i.e., they are higher than those of their non-palatalized counterparts. Lithuanian fricatives and affricates are arranged in the following order of increasing spectral peak frequency: velars < labiodental < alveolars/palatalized velars < palatalized labiodental/palatalized alveolars < dentals < palatalized dentals. The study and comparison of spectral peak frequencies for non-palatalized, palatalized, and palatal fricatives and affricates in Lithuanian and Latvian revealed the following trends: 1) Latv. [f] has higher spectral peak frequencies than Lith. [f], [fʲ] (in both male and female data); 2) according to the male data, the spectral peak frequencies of Latvian dental consonants are similar in most cases to those of non-palatalized Lithuanian dentals (except [z]); according to the female data, Latvian dentals are characterized by higher spectral peaks than non-palatalized and palatalized Lithuanian dentals; 3) the spectral peak frequencies of Latvian alveolar consonants (in both male and female data) are close to those of palatalized Lithuanian alveolars (since Latvian alveolars are also characterized by palatalization, although it is not phonological because there are no non-palatalized counterparts; see LVG 2013, 64); Straipsniai / Articles 161 JOLITA URBANAVIČIENĖ, INESE INDRIČANĖ 4) Latvian [x] occupies an intermediate position in terms of spectral peak frequency values between Lithuanian [x] and Lithuanian [x̟], because the quality of Latvian [x] depends on the following vowel: its articulation can range from velar to palatalized velar/palatal; 5) The closest Lithuanian consonants to Latvian [v] and [j]—the Lithuanian sonorants [u], [u̯], and [j]—were not investigated. However, it was observed that the spectral peak frequency value of Latvian [v] is closer to the values for Lithuanian [f] and [fʲ] than to that for Latvian [f]. Conversely, the spectral peak frequency value of the Latvian palatal consonant [j] is close to the values for Latvian [ʃ], [ʒ], [tʃ], and [dʒ] (i.e., consonants that have a non-phonological palatalization feature in Latvian). According to the frequency values of the spectral peak, the following articulatory consonant groups of Lithuanian and Latvian are distinguished: a) alveolar (sibilants): Lithuanian [ʃ], [ʃʲ], [ʒ], [ʒʲ], [tʃ], [tʃʲ], [dʒ], [dʒʲ], and Latvian [ʃ], [ʒ], [tʃ], [dʒ]; b) dental (sibilants): Lithuanian [s], [sʲ], [z], [zʲ], [ts], [tsʲ], [dz], [dzʲ], and Latvian [s], [z], [ts], [dz]. c) velar (non-sibilants): Lithuanian [x], [xʲ], [ɣ], [ɣʲ], and Latvian [x] d) labiodental / palatal (non-sibilants): Lithuanian [f], [fʲ], and Latvian [f], [v], [j]. Informants’ gender affects spectral peak frequencies: consonants produced by men have lower spectral peaks than the corresponding consonants produced by women. Since the spectral characteristics of consonants are related to the size of the oral front cavity, which is affected by vocal tract size (the larger the front cavity, the more acoustic energy is concentrated in the lower frequency range), the energy of consonants produced by men is expected to be concentrated at lower frequencies than that of women (cf. the male vocal tract is ~2.8 cm longer than the female vocal tract). 6.2. Relative intensity 6.2.1. Determining relative intensity Relative intensity indicates the ratio between the mean intensity of a fricative or affricate and the mean intensity of the vowel in its phonetic environment over a specified time interval. Measurements were made using the PRAAT sound analysis program. Mean intensity values (dB (μE)) for fricatives and affricates were obtained from an interval lasting 0.01 s, or 10 ms (see Figs. 9 and 11), while vowel intensity values were obtained from an interval lasting 0.05 s, or 50 ms, located approximately at the midpoint of the vowel (Figs. 10 and 12). 162 Acta Linguistica Lithuanica LXXV Dabartinių baltų kalbų pučiamieji priebalsiai ir afrikatos: akustinių požymių lyginamoji analizė determining mean intensity (μE) FIG. 9. Voiceless prevocalic fricative [s] in the syllable [sas]: interval lasting 0.01 s, or 10 ms FIG. 10. Vowel [ɑ] in the phonetic environment of the voiceless prevocalic fricative [s] (in the syllable [sas]): interval lasting 0.05 s, or 50 ms FIG. 11. Voiceless prevocalic affricate [t͡s] in the syllable [t͡sats]: interval lasting 0.01 s, or 10 ms FIG. 12. Vowel [ɑ] in the phonetic environment of the voiceless prevocalic affricate [t͡s] (in the syllable [t͡sats]): interval lasting 0.05 s, or 50 ms Relative intensity was calculated by dividing the mean intensity value of a fricative or affricate by the mean intensity value of the vowel in its phonetic environment. For example, if the mean intensity of the prevocalic voiceless fricative [s] in the syllable sas [sas] over a 10 ms interval is 55.06 dB (μE) (see Fig. 9), and the mean intensity of the vowel [ɑ] over a 50 ms interval is 73.41 dB (μE) (see Fig. 10), relative intensity is calculated as follows: 55.06 dB (μE): 73.41 dB (μE) = 0.750 dB (μE). The resulting relative intensity values were processed using the SPSS statistical data analysis program: mean statistical values for relative intensity (dB), used throughout this article, were calculated in this program. Straipsniai / Articles 163 JOLITA URBANAVIČIENĖ, INESE INDRIČĀNE 6.2.2. Comparison of palatalized and non-palatalized Lithuanian fricative consonants and affricates with Latvian data The mean statistical values (MSV) of the relative intensity of fricative consonants and affricates in Lithuanian and Latvian are presented in Table 3. TABLE 3. Lithuanian and Latvian fricative consonants and affricates: MSV of relative intensity (dB) C Male data Female data Lith. Lat. Lith. Lat. Fricative consonants [v] 0,788 0,819 [f] 0,648 0,648 0,619 0,608 [fʲ] 0,653 0,630 [s] 0,661 0,779 0,759 0,782 [sʲ] 0,786 0,738 [z] 0,761 0,815 0,783 0,863 [zʲ] 0,766 0,793 [ʃ] 0,699 0,751 0,704 0,752 [ʃʲ] 0,734 0,747 [ʒ] 0,799 0,806 0,816 0,858 [ʒʲ] 0,788 0,810 [j] 0,818 0,852 [x] 0,686 0,712 0,679 0,696 [xʲ] 0,675 0,639 [ɣ] 0,770 0,744 [ɣʲ] 0,802 0,721 Affricates [ʦ] 0,701 0,796 0,728 0,818 [ʦʲ] 0,722 0,746 [ʣ] 0,751 0,788 0,765 0,833 [ʣʲ] 0,759 0,774 [ʧ] 0,723 0,785 1,003 0,795 [ʧʲ] 0,765 0,750 [ʤ] 0,766 0,801 0,807 0,837 [ʤʲ] 0,774 0,778 164 Acta Linguistica Lithuanica LXXV Dabartinių baltų kalbų pučiamieji priebalsiai ir afrikatos: akustinių požymių lyginamoji analizė The mean statistical values of the relative intensity of Lithuanian and Latvian fricative consonants and affricates increase in the following order: 1) male data: Lith. [f]<[fʲ]<[s]<[xʲ]<[x]<[ʃ]<[t͡s]<[t͡sʲ]<[t͡ɕ]<[d͡z]<[d͡zʲ]<[z]<[t͡ɕʲ]<[zʲ]<[d͡ʒ]<[ɣ]<[d͡ʒʲ]<[sʲ]<[ʒʲ]<[ʒ]<[ɣʲ], Lat. [f]<[x]<[ʃ]<[s]<[t͡ɕ]<[v]=[d͡z]<[t͡s]<[d͡ʒ]<[ʒ]<[z]<[j]; 1) female data: Lith. [f]<[fʲ]<[xʲ]<[x]<[ʃ]<[ɣʲ]<[t͡s]<[sʲ]<[ɣ]<[t͡sʲ]<[ʃ]<[t͡ɕ]<[s]<[d͡z]<[d͡zʲ]<[z]<[zʲ]<[d͡ʒ]<[ʒʲ]<[ʒ]<[t͡ɕʲ], Lat. [f]<[x]<[ʃ]<[s]<[t͡ɕ]<[t͡s]<[v]<[d͡z]<[d͡ʒ]<[j]<[ʒ]<[z]. A comparison of the MSV of relative intensity for palatalized and non-palatalized Lithuanian fricative consonants and affricates with Latvian data is shown in Figs. 13–14. FIG. 13. Comparison of the MSV of relative intensity (dB) for palatalized and non-palatalized Lithuanian fricative consonants and affricates with Latvian results: male data Latvian consonants Lithuanian consonants palatalized non-palatalized ● ○ ● Straipsniai / Articles 165 JOLITA URBANAVIČIENĖ, INESE INDRIČĀNE Lithuanian fricatives and affricates are characterized by slightly lower relative intensity values than their Latvian counterparts with respect to both voicing and place of articulation (based on data from both men and women). Several consonants do not support this trend: in the men's data, Lith. [fʲ] > Latv. [f], but Lith. [f] = Latv. [f]; in the women's data, Lith. [fʲ], [f] > Latv. [f]; Lith. [t͡ɕ] > Latv. [t͡ɕ] > Lith. [t͡ɕʲ] (the Lithuanian alveolar affricate [t͡ɕ] produced by the female speaker stands out for its exceptionally high relative intensity values; see Fig. 14). A comparison of palatalized and non-palatalized Lithuanian consonants shows that the relative intensity values for palatalized consonants are more often higher. However, there are exceptions, cf.: men's data – [xʲ] < [x], [ʒʲ] < [ʒ]; women's data – [xʲ] < [x], [ɣʲ] < [ɣ], [sʲ] < [s], [t͡ɕʲ] < [t͡ɕ], [d͡ʑʲ] < [d͡ʑ], [ʒʲ] < [ʒ]. Latvian palatal [j] also has one of the highest intensity values (according to the women's data), or the highest (according to the men's data). It is likely that, with more data or the use of other analysis programs, the correlation between palatalization and relative intensity would be even clearer. FIG. 14. Comparison of the relative intensity values (dB) of palatalized and non-palatalized Lithuanian fricatives and affricates with the Latvian results: women's data Latvian consonants | Lithuanian consonants palatalized | non-palatalized △ | △ | ▲ 166 Acta Linguistica Lithuanica LXXV Dabartinių baltų kalbų pučiamieji priebalsiai ir afrikatos: akustinių požymių lyginamoji analizė 6.2.3. Comparison of voiced and voiceless fricatives and affricates The study results show (see Figs. 13–14, Table 3) that, in both Lithuanian and Latvian, the relative intensity values of voiced fricatives and affricates are generally higher than those of their voiceless equivalents. Thus, consonant voicing affects the relative intensity measure. Among the Lithuanian study materials, this trend is not supported by the voiced palatalized dental consonant [z̪] produced by a male speaker, which has a lower mean value (0.766 dB)) than the voiceless dental [s̪] (0.786 dB)). In the subgroup of non-palatalized consonants, the exception is the voiced alveolar affricate [dʒ] produced by a female speaker, whose mean relative intensity value (0.807 dB)) is lower than that of the voiceless affricate [tʃ] (1.003 dB)). Voiced Latvian consonants also tend to have higher relative intensity values than their voiceless correlates. The exception is the voiced dental affricate [dʒ] produced by a male speaker, which has a lower mean value (0.788 dB)) than the corresponding voiceless affricate [tʃ] (0.796 dB)). 6.2.4. Comparison of men's and women's data Differences in the physiology of the sexes, which result in a longer vocal tract in men (cf. vocal tract length: ~16.9 cm in men, ~14.1 cm in women), lead us to expect differences in relative intensity results as well. The results of the study of Latvian consonants confirm the expected trends: higher relative intensity values are more often associated with consonants produced by women. However, there are exceptions, cf.: the consonants [f], [x] produced by men have higher relative intensity values than the corresponding consonants produced by women. A comparison of male and female data for the same Lithuanian fricative or affricate revealed no systematic differences. 6.2.5. The Influence of Consonant Manner/Place of Articulation on the Acoustic Features of Consonants In neither Lithuanian nor Latvian do relative intensity data allow fricatives and affricates to be differentiated by manner and place of articulation. It is likely that relative intensity data could distinguish sibilants, i.e., “brighter” consonants (with greater intensity), from non-sibilants, i.e., “duller” consonants (with lower intensity). The theory of speech acoustics holds that the lowest intensity is characteristic of voiceless non-sibilants (e.g., −15 dB), while the highest intensity is characteristic of voiced sibilants Straipsniai / Articles 167 JOLITA URBANAVIČIENĖ, INESE INDRIČĀNE (0 dB); the intensity of voiced non-sibilants and voiceless sibilants is similar—~5 dB (see Ambrazevičius 2011: 38). The results show that, in Lithuanian and Latvian, on the basis of the mean statistical values for relative intensity, sibilants cannot be distinguished from non-sibilants within the groups of voiced fricatives and affricates produced by men; that is, [ɣ] does not stand apart in the group of non-palatalized Lithuanian fricatives and affricates [z], [ʒ], [dʑ], [dʒ]; [ɣʲ] does not stand apart in the group of palatalized Lithuanian consonants [zʲ], [ʒʲ], [dʑʲ], [dʒʲ]; and [v], [j] do not stand apart in the group of Latvian fricatives and affricates [z], [ʒ], [dʑ], [dʒ]. The female data show that, in the voiced consonant subsystem of Lithuanian, the voiced non-sibilants [ɣʲ] and [ɣ] can be distinguished from the voiced sibilants [z], [zʲ], [ʒ], [ʒʲ], [dʑ], [dʑʲ], [dʒ], [dʒʲ]: the non-sibilants [ɣʲ] and [ɣ] have lower relative-intensity means than the sibilants. In the voiced consonant subsystem of Latvian, the non-sibilant [v] is distinguished from the sibilants [z], [ʒ], [dʑ], [dʒ]; it is characterized by slightly lower relative-intensity means (also based only on the female data). In the groups of voiceless fricatives and affricates in both Lithuanian and Latvian, relative intensity values help distinguish sibilants (with higher means) from non-sibilants (with lower means). Cf.: in the group of non-palatalized Lithuanian consonants, the sibilants [s], [ʃ], [ts], [tʃ] differ from [f], [x] (only in the female data; the values for [s] and [x] produced by men overlap); in the group of palatalized Lithuanian fricatives and affricates, the sibilants [sʲ], [ʃʲ], [tsʲ], [tʃʲ] are distinguished from the non-sibilants [fʲ], [xʲ] (based on data from both sexes); in Latvian, in the group of fricatives and affricates, the sibilants [s], [ʃ], [ts], [tʃ] are distinguished from the non-sibilants [f], [x] by their relative intensity values (based on data from informants of both sexes). 6.2.6. Summary of the Relative Intensity Study Lithuanian fricatives and affricates are characterized by slightly lower relative-intensity means than Latvian consonants that are otherwise similar in voicing and place of articulation. These results may have been due to the sound recordings: although the technical recording parameters were standardized in Lithuania and Latvia, during recording the greatest attention was paid to recording quality (since the recordings were used to study various acoustic features of sounds17), rather than to the consistency of the sound signal, which may have fluctuated because of the informants’ individual voice characteristics or fatigue resulting from speaking into a voice recorder for a long time. This sound 17 The recordings were made as part of the project “Contemporary Baltic Languages: Spectral Characteristics of Sounds (An Instrumental Study)” (No. MIP-081/2013). 168 Acta Linguistica Lithuanica LXXV Dabartinių baltų kalbų pučiamieji priebalsiai ir afrikatos: akustinių požymių lyginamoji analizė Signal irregularity may have affected the results of the intensity analysis. Thus, it can be stated provisionally that the study detected greater sound strength, or intensity, in Latvian consonants than in Lithuanian consonants; however, more extensive studies should be conducted in the future to fully eliminate the influence of technical parameters. The results for the relative intensity of palatalized and non-palatalized Lithuanian consonants indicate a certain correlation between consonant palatality and relative intensity, i.e. palatalization increases consonant intensity. This is also supported by the Latvian data: the intensity of the palatal consonant [j] is the highest (according to the women’s data, one of the highest) among all the consonants studied. The association between relative intensity and voicing is beyond doubt: in both Lithuanian and Latvian, the relative intensity means (VSV) of voiced fricatives and affricates are higher than those of their voiceless equivalents. The study did not confirm a direct relationship between informants’ sex and relative intensity: physiological differences do not have a significant effect on relative intensity values. In neither Lithuanian nor Latvian do relative intensity data allow fricatives and affricates to be differentiated by manner or place of articulation. 7. SUMMARY The study compared the spectral peak frequencies and relative intensity values of Lithuanian and Latvian fricatives and affricates with respect to voicing, palatalization, manner and place of articulation, and informants’ sex. 7.1. Spectral peak frequency. No correlation was found between spectral peak frequencies and the voicing of Lithuanian and Latvian fricatives and affricates. Due to palatalization, the spectral peak frequencies of Lithuanian consonants increase; that is, they are higher than the spectral peak frequencies of their non-palatalized equivalents. The following Lithuanian and Latvian consonant groups can be distinguished by their spectral peak frequency values: alveolar and dental (sibilants), and velar and labiodental / palatal (non-sibilants), according to their articulatory characteristics. Informants’ sex affects spectral peak frequencies: the spectral peak frequencies of fricatives and affricates produced by men are lower than Straipsniai / Articles 169 JOLITA URBANAVIČIENĖ, INESE INDRIČANĖ the frequencies of the corresponding consonants produced by women. The lowering of the spectral peak frequency is associated with the larger size of the anterior oral cavity in men. 7.2. Relative intensity. The study revealed a relationship between relative intensity and voicing in contemporary Baltic languages: the relative intensity of voiced fricatives and affricates is greater than that of the corresponding voiceless correlates. A correlation was observed between consonant palatalization and relative intensity: Lithuanian palatalized fricatives and affricates have higher relative intensity values than their non-palatalized equivalents. The relative intensity of the Latvian palatal consonant [j] is also higher than that of other Latvian consonants. In neither Lithuanian nor Latvian do the relative-intensity data allow fricatives and affricates to be differentiated by manner and place of articulation. The informant’s sex has no notable effect on relative intensity values: the observed differences are not systematic. TABLE 4. Correlation of acoustic properties of Lithuanian and Latvian fricatives and affricates with voicing, palatalization, articulation, and informant sex Acoustic properties → Spectral peak frequency (Hz) Lie. La. Voicing No correlation found No correlation found Palatalization Cį is greater than C: velars: 1470–2680 Hz; alveolars: 70–600 Hz; dentals: 50–150 Hz; labiodentals: 135 Hz. – Manner/place of articulation Sibilants: alveolar: 2440–4600 Hz; dental: 4400–6040 Hz. Non-sibilants: velar: 950–3700 Hz; labiodental: 1960–3180 Hz. Sibilants: alveolar: 2930–5040 Hz; dental: 5190–6780 Hz. Non-sibilants: velar: 1800–2100 Hz; labiodental: 2410–6770 Hz; palatal: 3940–4220 Hz. Gender Men: 950–6040 Hz; women: 1020–6030 Hz. Men: 1800–5770 Hz; women: 2110–6780 Hz. 170 Acta Linguistica Lithuanica LXXV Fricative consonants and affricates in the modern Baltic languages: a comparative analysis of acoustic features Acoustic features → Relative intensity (dB) Lith. Lat. Voicing Voiceless: 0.620–0.790 (dB). Voiced: 0.720–0.820 (dB). Voiceless: 0.610–0.820 (dB). Voiced: 0.790–0.860 (dB). Palatalization Cᶦ greater than C: velars: 0.032 (dB); labiodentals: 0.005–0.011 (dB); dentals/alveolars: 0.005–0.125 (dB). – Manner/place of articulation No correlation found No correlation found Gender No correlation found No correlation found 7.3. The results of the study of acoustic features of fricative consonants and affricates in the modern Baltic languages are summarized in Table 4. It can be seen that the acoustic features of the consonants studied (spectral peak frequency and relative intensity) correlate with the same criteria in both languages: spectral peak frequency correlates with consonant articulation (especially sibilance), palatalization (only in Lithuanian), and informants’ gender; relative intensity correlates with consonant voicing and palatalization (in Lithuanian). However, the different acoustic-feature data obtained for Lithuanian and Latvian (see Table 4) allow conclusions to be drawn about differential differences in the consonant systems of the Baltic languages. 7.4. The study revealed the following differential Lithuanian and Latvian consonant differences: 7.4.1. Based on spectral peak values, palatalized Lithuanian consonants approach the corresponding non-palatalized consonants in Latvian, while the spectral peak frequencies of Latvian palatal consonants are among the highest (only the spectral peaks of dental consonants in both languages are higher). It can be argued that Latvian consonants are “softer” than non-palatalized Lithuanian consonants, while Latvian palatal consonants are “softer” than palatalized Lithuanian consonants. 7.4.2. Fricative consonants and affricates in Latvian have greater relative intensity than corresponding Lithuanian consonants with respect to voicing and articulation. These results may have been caused, on the one hand, by the more energetic articulation of the Latvian informants and, on the other, by fluctuations in the audio signal during recording. To rule out the influence of technical parameters, more detailed studies of this kind should be conducted in the future. Straipsniai / Articles 171 JOLITA URBANAVIČIENĖ, INESE INDRIČĀNE LITERATŪRA Ambrazevičius Rytis 2011: Kalbos akustika glaustai. Kaunas: Technologija. Ambrazevičius Rytis, Asta Leskauskaitė 2014: Priebalsių akustinės ypatybės: palatalizacija ir balsingumas. Kaunas: Technologija. Crystal David 2008: Dictionary of Linguistics and Phonetics. 6th edition. Malden, Oxford, Victoria: Blackwell Publisher Ltd. Čეirane Solveiga 2010: Latviešu valodas balsīgo spraudzeņu troksneņu akustisks raksturojums. – Žmogus ir žodis 12(1), 14–19. Čეirane Solveiga 2011: Latviešu valodas balsīgo troksneņu akustisks raksturojums. Promocijas darbs filoloģijas doktora grāda iegūšanai. Rīga: Latvijas Universitāte. Čეirane Solveiga, Inese Indričāne 2012: Latviešu valodas troksneņu raksturojums pēc lokusa vienādojumiem. – Baltistica 47(1), 37–50. Dereškevičiūtė Sigita 2013: Lietuvių kalbos priebalsių akustinės ypatybės (daktaro disertacija). Kaunas: Vytauto Didžiojo universitetas, Lietuvių kalbos institutas. DLKG – Ambrazas Vytautas (red.) 2005: Dabartinės lietuvių kalbos gramatika. 4 pat. leid. Vilnius: Mokslo ir enciklopedijų leidybos institutas. Grigorjevs Juris, Jurgita Jaroslavienė 2015a: Spectral structure and duration of the monophthongs in both contemporary Baltic languages. – Linguistica Lettica 23, 68–92. Grigorjevs Juris, Jurgita Jaroslavienė 2015b: Comparative study of the qualitative features of the Lithuanian and Latvian monophthongs. – Baltistica 50(1), 57–89. Indričāne Inese 2012: Latviešu valodas nebalsīgo frikatīvo spraudzeņu raksturojums pēc FFT spektriem. – Linguistica Lettica 20, 111–120. Indričāne Inese 2013: Latviešu valodas nebalsīgo troksneņu akustisks un auditīvs raksturojums. Promocijas darbs filoloģijas doktora grāda iegūšanai. Rīga: Latvijas Universitāte. Indričāne Inese 2014: Latviešu valodas nebalsīgo frikatīvo spraudzeņu intensitātes raksturojums. – Baltistica 49(1), 101–119. Indričāne Inese, Jolita Urbanavičienė 2015: Lietuvių ir latviešu kalodas eksplozīvo slēdzeņu spektrālais raksturojums. – Baltu filoloģija 24(1), 51–87. IPA – International Phonetic Association. Prieiga per internetą: https://www.internationalphoneticassociation.org/content/full-ipa-chart [žiūrėta 2016 m. balandžio 18 d.]. Johnson Keith 2003: Acoustic & auditory phonetics. Oxford: Blackwell Publishers. 172 Acta Linguistica Lithuanica LXXV Dabartinių baltų kalbų pučiamieji priebalsiai ir afrikatos: akustinių požymių lyginamoji analizė Jongman Allard 1989: Duration of frication noise required for identification of English fricatives. – Journal of the Acoustical Society of America 85(4), 1718–1725. Jongman Allard, Ratree Wayland, Serena Wong 2000: Acoustic characteristics of English fricatives. – Journal of the Acoustical Society of America 108, 861–867. Kent Raymond D., Charles Read 1992: The acoustic analysis of speech. San Diego, California: Singular Publishing Group, Inc. Kodzasov, Krivnova – Кодзасов Сандро В., Ольга Ф. Кривнова 2001: Общая фонетика. Москва: Российский государственный гуманитарный университет. Ladefoged Peter, Ian Maddieson 1998: The Sounds of the World’s Languages. Oxford: Blackwell. LKG – Ulvydas Kazys (red.). Lietuvių kalbos gramatika 1. Fonetika ir morfologija (daiktavardis, būdvardis, skaitvardis, įvardis). Vilnius: Mintis, 1965. LVG – Auzina Ilze, Ieva Brenķe, Juris Grigorjevs, Inese Indričāne, Baiba Ivulāne, Andra Kalnača, Linda Lauze, Ilze Lokmane, Dace Markus, Daina Nītiņa, Gunta Smiltniece, Baiba Valkovska, Anna Vulāne. Latviešu valodas gramatika. Rīga: LU Latviešu valodas institūts, 2013. Markus Dace, Juris Grigorjevs 2002: Fonētikas pētīšanas metodes 1. Rīga: Rasa ABC. Mikalauskaitė Elzbieta 1975: Lietuvių kalbos fonetikos darbai. Vilnius: Mokslas. Pakerys Antanas 2003: Lietuvių bendrinės kalbos fonetika. Vilnius: Enciklopedija. Pakerys Jurgis 2014: Kalbos konstruktorius / Įvadinės paskaitos. Vilnius: Vilniaus universiteto leidykla. Reetz Henning, Allard Jongman 2009: Phonetics. Transcription, Production, Acoustics and Perception. Malden, Oxford: Wiley-Blackwell. Urbanavičienė Jolita, Inese Indričāne 2015: Lietuvių ir latvių kalbų trankieji priebalsiai: lokuso lygių rezultatai. – Baltistica 50(2), 261–293. Straipsniai / Articles 173 JOLITA URBANAVIČIENĖ, INESE INDRIČĀNE Fricatives and Affricates of Contemporary Baltic Languages: Comparative Analysis of Acoustic/ Spectral Cues SUMMARY In the current paper two acoustic cues – spectral peak’s frequency and relative intensity – obtained for the fricatives and affricates of contemporary Baltic languages have been studied. The research is conducted according to one and the same approach both for standard Lithuanian and Latvian. The aim of this paper is to compare the fricatives and affricates of the two Baltic languages in order to determine the essential spectral differences and the effect of palatalization, voicing, manner and place of articulation, as well as the effect of speaker’s gender displayed by the acoustic cues examined in the paper. The research object is 34 consonant phonemes of contemporary Baltic languages: 1) 22 fricatives – /f, fʲ, s, sʲ, z, zʲ, ʃ, ʃʲ, ʒ, ʒʲ, x, xʲ, ɣ, ɣʲ/ of Standard Lithuanian and /f, v, s, z, ʃ, ʒ, x/ of Standard Latvian; 2) 12 affricates – /t͡s, t͡sʲ, d͡z, d͡zʲ, t͡ʃ, t͡ʃʲ, d͡ʒ, d͡ʒʲ/ of Standard Lithuanian and /t͡s, d͡z, t͡ʃ, d͡ʒ/ of Standard Latvian. The recording material consists of isolated CVC syllables pronounced by 12 Lithuanian and 12 Latvian native speakers (at the ages of 20 to 48): 6 males and 6 females within each group. Data processing was performed using the software PRAAT, MS EXCEL, and SPSS. The results indicate that spectral peak’s frequency displays the effect of palatalization (in Standard Lithuanian), manner and place of articulation, as well as the effect of speaker’s gender. The palatalization causes higher spectral peak’s frequencies for palatalized vs. non-palatalized fricatives and affricates in Standard Lithuanian. According to spectral peak’s frequency, it is possible to divide the set of fricatives and affricates into sibilants (dentals and alveolars) vs. non-sibilants (labiodentals, palatals, and velars) both in standard Lithuanian and Latvian. The spectral peak’s frequency also displays gender-related differences, since the frequency values obtained for male speakers are lower in comparison to those of female speakers. The spectral peak’s frequency values do not reveal the effect of voicing. The results obtained in this research show the correlation between relative intensity and voicing: the values obtained for the voiced fricatives and affricates are higher in comparison to their voiceless counterparts. The highest relative intensity values are characteristic of the palatalized fricatives and affricates of Standard Lithuanian, i.e., palatalization increases the strength of the consonant. On the basis of relative intensity, it is impossible to classify the fricatives or affricates of Standard Lithuanian or Standard Latvian according to their manner and place of articulation. The results obtained in this research do not provide the effect of speaker’s gender. 174 Acta Linguistica Lithuanica LXXV Dabartinių baltų kalbų pučiamieji priebalsiai ir afrikatos: akustinių požymių lyginamoji analizė Similar correlations between the acoustic cues examined in this paper and the effect of voicing, manner and place of articulation, as well as the effect of speaker’s gender are observed both in standard Lithuanian and Latvian. However, some differences in the data of the two Baltic languages are also observed. The results show that the fricatives and affricates of Standard Latvian are more similar to the palatalized vs. non-palatalized fricatives and affricates of the same voicing, manner and place of articulation in Standard Lithuanian, i.e., the fricatives and affricates in Standard Latvian seem to be “softer” in comparison to the non-palatalized fricatives and affricates in Standard Lithuanian. It is also observed that the fricatives and affricates of Standard Latvian have a higher relative intensity in comparison to the corresponding fricatives and affricates of Standard Lithuanian. This can be explained either by more energetic articulation of these speech sounds for speakers of Standard Latvian or by some differences related to the recording level. Įteikta 2016 m. gegužės 19 d. JOLITA URBANAVIČIENĖ Lietuvių kalbos institutas Petro Vileišio g. 5 LT-10308 Vilnius Lithuania [email protected] INESE INDRIČĀNE LU Latviešu valodas institūts Akadēmijas laukums 1 LV-1050 Rīga Latvia [email protected] Straipsniai / Articles 175