Prienų šnektos nosinių sonantų spektrinė charakteristika

JURGITA JAROSLAVIENE Institute of the Lithuanian Language Research interests: dialectology, phonology, experimental phonetics. SPECTRAL CHARACTERISTICS OF NASAL SONANTS IN THE PRIENAI DIALECT The spectral characteristics of the nasals in the subdialect of Prienai A B S T R A C T Based on instrumental research data, the article analyzes the acoustic distinctive features of nasal sonants in the Prienai dialect of the Eastern Kauniškiai subdialect. In describing the spectral characteristics of the sounds under study, the values of consonant formants and the distribution of energy are taken into account (the qualitative features of nasal sonants are determined by formants, although these are less prominent than those of vowels; moreover, the number of formants in the spectrum of nasal sounds is often much greater because the nasal cavity resonator is additionally involved in their articulation), as well as the quality of adjacent vowels, the values and trajectories of their formants (especially F2), and their structure. In differentiating the spectral characteristics of [m], [m̂], [n], [n̂], and nasal velar allophones, the values of the anti-formants occurring in their spectra are also reviewed. A B S T R A C T The main distinctive acoustic features of the nasals in the Eastern Kaunas Prienai subdialect have been investigated in the experimental study described in the present article. The obtained data suggest that the nasals are distinguished from the other sounds of the subdialect of Prienai on the basis of their formant structure (they are like vowels in that they can be characterizes largely in terms of their formant frequences, but they differ in that the formants are not as loud as they are in vowels; besides, the nasals are made by blocking the sound from coming out of the mouth while allowing it to come out through the nose, and this affects the relative amplitude of the formants) and transitions between it and the context vowel. Since one of the main distinction between nasals during the nasal murmur is provided by the oral anti-resonance, the nasal sonorants [m], [m̂], [n], [n̂] as well as nasal-velar allophones are characterized by the frequency values of the oral zero. K E Y W O R D S: nasal sonants, formant, anti-formant, spectral characteristics, acoustic properties. K E Y W O R D S: nasals, formant, oral zero, spectral characteristics, acoustic features. Straipsniai / Articles 105 JURGITA JAROSLAVIEnĖ 0. INTRODUCTION The spectral characteristics of the nasal sonorants of the Prienai subdialect have been described in general terms in an instrumental study of its sound system (Jaroslavienė 2010: 264–267)1. However, a more detailed analysis of their distinguishing acoustic features has not yet been presented; the second-formant transition between vowels (C→V) and the location and significance of the anti-formants occurring in the spectrum have not been measured. This prompted an additional study of the nasal sonorants of the subdialect under investigation, the results of which are presented in this article. 0.1. Aim of the Article Nasal sonorants, like other sonorous consonants, also have vowel-like properties: when they are articulated, the fundamental tone predominates (therefore, sonorous consonants usually have a clear formant structure); their qualitative features, like those of vowels, are determined by formants (especially the first two – F1, F2), although in sonorants these are less prominent than in vowels (cf. Girdenis 2003: 221 and the literature cited therein; 224; Ladefoged, Maddieson 2002: 102–136, 182–245; Ladefoged 2005: 49–62, 84, etc.). Moreover, because of the influence of an additional (nasal) resonator, the number of formants in the spectrum of nasal sounds is often greater (see Bacevičiūtė 2008: 10; Butkauskaitė 2010: 6, 8; Grigorjevs 2012: 267–272 and the literature cited therein). Analysis of the acoustic properties of the segments in question also takes into account the formant structure and dynamics of adjacent vowels2, since, as is assumed, 1 The consonant system of Lithuanian dialects has been little studied instrumentally. Apart from the consonants of the Prienai subdialect (2010: 238–267; 2010a: 307–324), only the spectral (and auditory) features of the consonants of the Northern Samogitian dialect and the regressive effect of consonant palatalization on the vowel spectrum have been described fairly comprehensively and systematically (see Girdenis 1967; 2000: 69–78; 258–267; Kliukienė 1992; 1994: 43–52; 1995: 58–68; 1998: 79–85; 2011, and elsewhere). Other instrumental studies focus mainly on individual cases of consonantism in Standard Lithuanian: for example, using the locus-measurement method (English loci), Rytis Ambrazevičius (2010: 5–10) described the differential acoustic features of soft and hard prevocalic Lithuanian sonorants. It was established that the most important features for distinguishing soft and hard sonorants are the F2 loci. The researcher also presents measurements of the F2 and F3 loci of the consonants under study and discusses the effects of place of articulation and regressive coarticulation on the loci (see also Bacevičiūtė 2008: 7–19; Dereškevičiūtė 2008: 15–19; Dereškevičiūtė, Kazlauskienė 2009: 98–111; Ambrazevičius 2012a: 13–18, etc.). 2 When [m]- and [n]-type sounds are articulated, a so-called closure is formed in the oral cavity, but the airflow passes smoothly through the nose, and it is precisely this nasal resonance that determines certain spectral features of adjacent sounds (for an experimental study of the duration and quality of some nasalized vowels in Standard Lithuanian, see Ledichova 2012: 76–85; cf. also Dereškevičiūtė 2008: 15–19). When, for example, plosive consonants are articulated, a so-called complete obstruction (closure) is also formed in the oral cavity for a certain period, and the air bursts out through the speech organs as they suddenly separate; therefore, the quality of plosive consonants is more accurately indicated by the beginning or end of the formants of adjacent vowels: some consonants shift them in one direction, others in another (for a fuller account, see Girdenis 2003: 224; 1967; Ladefoged 2005: 50; Jaroslavienė 2010: 106 Acta Linguistica Lithuanica LXVIII Prienų šnektos nosinių sonantų spektrinė charakteristika nasal consonants are distinguished reliably only by certain spectral features of adjacent vowels (see Girdenis 2003: 224 and the literature cited therein; Kliukienė 1992: 78 and the literature cited therein; 1998: 80; cf. also Bacevičiūtė 2008: 7–19; Jaroslavienė 2010: 261–262; Grigorjevs 2012: 75–80). However, when distinguishing one nasal sonorant from another—as also confirmed by recent experimental studies of sonorants—it is likewise important to measure the values of the anti-formants occurring in the spectra of the consonants under investigation (cf. Grigorjevs 2012: 275–280; 288–290; 2012a: 22–23; Butkauskaitė 2010: 6 and the literature cited therein; see also Ladefoged, Maddieson 2002: 117–118)3. The aim of this article is, on the basis of instrumental study data, to analyze and describe the principal spectral characteristics of unstressed nasal sonorants in the Prienai subdialect and to determine which acoustic features distinguish the sounds under investigation from one another. 0.2. Research Material and Methodology For the spectral study of the allophones of the consonant phonemes /m/, /m̃/, /n/, /ñ/, the following words used in the dialect under investigation were selected: mã.ma. ~ m̃ãma, m̃ã.na. ~ mãna, nã.ma. ~ nãma, nã.ta. ~ nãta, mã.da. ~ mãda, mã.ta. ~ Mãta, dã.ma. ~ dãma, dã.na. ~ dãna, ñĩ.ko. ~ ñýko, mĩ.ñe. ~ mýñe, pĩ.ñe. ~ pýñe, mẽ.ne. ~ mẽñe, mẽ.te. ~ mẽtę, nẽ.še. ~ nẽšę, mã.-na. ~ mẽna, pã.na. ~ pẽna, mó.ko. ~ móko, nó.ko. ~ nóko, and so on. In analyzing the spectral characteristics, the formant structure and trajectories of adjacent vowels were also taken into account. For comparison, the spectral distribution properties of other sounds were additionally examined in the words tã.ka. ~ tãką, kã.pa. ~ kãpą, lã.pa. ~ lãpą, bã.da. ~ bãdą, pã.-da. ~ pãdą, vã.da. ~ vãdą, pĩ.ko. ~ pỹko, pĩ.lę. ~ pỹlę, sẽ.ne. ~ sẽnę, tẽ.ke. ~ tẽkę, kẽ.pe. ~ kẽpę, and others. Given that very different vowel allophones are usually found next to hard and soft consonants, and also in order to measure as objectively as possible the antiformants of hard and soft sounds (including the nasal palatal sonant), the study also used examples, mostly meaningless, containing short low front and back vowels: [mama], [m̃em̃e], [nana], [ñeñe], [nan-ka], [ñeñke], and so on. Only the first, unstressed, sonant in these words4—[m], [m̃], [n], and [ñ]—was examined; the unstressed velar allophones [ŋ] and [ŋ̃] were measured word-medially. 3 For the differential acoustic features of soft and hard prevocalic Lithuanian sonants, see Ambrazevičius 2010: 5–10 (see also note 1). 4 In pronouncing them, the final short vowel was stressed. Straipsniai / Articles 107 JURGITA JAROSLAVIENĖ The research material was read directly into a computer three times, with declarative intonation (i.e., in isolation, as separate sentences), by three middle-aged men and one woman from Prienai, all without speech defects. Examples pronounced unnaturally were excluded from the study. The material was saved as *.wav files. The segments under study were cut using PRAAT (versions 5.0.34 and 5.3.23), a computer sound-analysis program developed by University of Amsterdam researchers Paul Boersma and David Weenink. For the spectral analysis, first a natural grayscale sound or word spectrogram of the Visible Speech type was generated from the extracted segments (words or consonants and adjacent vowels), with the maximum frequency set at 5000 Hz. Formant trajectories were then measured automatically and, where necessary, plotted, also with a maximum frequency of 5000 Hz. Natural spectrograms and plots of formant dynamics were often combined into a single image, precisely aligned in terms of frequency and duration. Spectral slices of nasal sonants were also plotted: when measuring sound antiformants (or zero formants, English oral zeros), the so-called FFT and LPC curves were combined (the method explained by Latvian phonetician Juris Grigorjevs was used to determine and measure FFT and LPC and combine them in a single plot5). 1. OVERVIEW OF THE SPECTRA OF NASAL CONSONANTS The results of the instrumental study of nasal sonants in the Prienai dialect are discussed and analyzed below. 1.1. The formant structure and energy distribution of nasal sonorants in the Prienai dialect A comparison of the spectral characteristics of nasal and non-nasal consonants in the Prienai dialect shows (cf. Figs. 1–24; see also Note 3) that the spectral energy, due to the specific nature 5 To plot the spectral slices of sonorants, first open the spectrogram of the target word or segment in the right-hand panel of PRAAT, then select the spectrum of the nasal sonorant (its stationary portion) and choose Spectrum: View spectral slice. The name of the extracted FFT spectrum appears in the left-hand panel of the program. Next, select LPC smoothing (in the same—left-hand—panel) and enter the required coefficients in the pop-up window Spectrum: LPC smoothing. For example, the number selected for Number of peaks in this study is 26 (since the material under investigation was recorded at a sampling rate of 44.1 kHz, half of which is 22 + 4 (or 5) = 26); Pre-emphasis from (Hz) is set to 50.0, and, after clicking OK, an LPC plot of the spectrum slice (the formant trajectory) is automatically generated from the spectrogram. Later, in the right-hand panel of PRAAT, the FFT and LPC plots of the sound slices obtained are combined into one (when combining the plots, it is again important to select the frequencies and other required parameters correctly). 108 Acta Linguistica Lithuanica LXVIII Prienų šnektos nosinių sonantų spektrinė charakteristika [m], [m̃], [n], [ñ]6 are articulated over a wider frequency range—this is one of the acoustic distinguishing characteristics of nasal consonants (Girdenis 1967: 249; 2003: 225; Kliukienė 1992: 134; 1998: 80). As noted, because of the influence of an additional resonator, the number of formants in the spectrum of nasal sonorants is often greater; moreover, nasal resonance also affects the spectrum of adjacent sounds: when these are articulated, the soft palate does not completely close the opening to the nasal cavity, and some of the air escapes through the nose (cf. Pakerys 2003: 191; Kliukienė 1998: 80; Girdenis 2003: 224; Murinienė 2007: 62, etc.). Therefore, the next section will also review the characteristics of the formant structures of adjacent sounds. The feature distinguishing the spectral characteristics of consonants [m], [m̃], [n], [ñ], [ŋ] and [ŋ̃]—which has not previously been investigated instrumentally in detail for Lithuanian—namely, the antiformants formed in the spectrum of nasal sonorants and their importance in distinguishing members of the class under discussion from one another, is examined in greater depth in the second part of this article. In this section, on the basis of the results obtained and with reference to the spectrograms and formant-trajectory plots, the qualitative features of [m], [m̃], [n], [ñ] are first reviewed; these features are determined by the first two formants, F₁ and F₂. They depend on the resonant cavities formed in the mouth and, in part, in the nasal cavity, as well as on the relationships between them (Girdenis 2003: 221). The first formant of nasal consonants (Eng. nasal formant, less often nasal zero7) occurs in the very low-frequency range (see Table 1, cf. Figs. 1–24), whereas the second is closely associated with the quality of the adjacent vowel (an important consideration when distinguishing hard and soft members of this class; cf. also Kliukienė 1992: 56–57; 78–79; Bacevičiūtė 2008: 10; Ambrazevičius 2010: 8–9; Grigorjevs 2012: 275–280)8. According to the data from the study of the Prienai dialect, the F₁ values of unstressed [m], [m̃], [n], [ñ] in words pronounced in isolation (the stationary portion of the sonorants was measured) range from approximately 250 to 290 Hz (see Table 1): in summary calculations, before the vowel [a.], the male speaker's [m] has F₁ = 253 Hz; [n] has F₁ = 276 Hz; the female speaker's [m] has F₁ = 266 Hz; [n] has F₁ = 288 Hz; before the vowel [i], the male speaker's [m̃] has F₁ = 269 Hz; [ñ] has F₁ = 271 Hz; the female speaker's [m̃] has F₁ = 268 Hz; [ñ] has F₁ = 275 Hz, and so on. As can be seen, the value tends to be somewhat higher 6 The consonants under investigation consist primarily of tonal elements, so their formant bands are quite clearly visible. When they are articulated, the soft palate lowers and draws away from the posterior wall of the pharynx, allowing air to pass relatively freely through the nose (i.e., there is no burst, since no substantial air pressure builds up in the mouth). For the effects of the place of articulation and phonetic context on the loci of prevocalic Lithuanian sonorants, see Ambrazevičius 2010: 8–9. For the acoustic distinguishing characteristics of Latvian sonorants, see Grigorjevs 2012: 267–292. 7 It should be noted that some researchers use this English term (nasal zero) to mean an antiformant (or zero formant), while calling F₁ in nasal consonants a nasal formant (nasal formant) (see Ladefoged, Maddieson 2002: 117, Table 4.6). 8 Foreign researchers report (based on data from some of the world's languages) that the F₁ frequency range for [m], [n] extends from 200, 250, or 300 Hz to 300, 400, or 450 Hz (see Butkauskaitė 2010: 6 and references cited therein; Grigorjevs 2012: 270 and references cited therein, etc.). For example, Peter Ladefoged and Ian Maddieson (2002: 117) give the first formants of Catalan nasal consonants as follows: [m] F₁ = 250 Hz; [n] F₁ = 280 Hz. Straipsniai / Articles 109 JURGITA JAROSLAVIENĖ FIG. 1. Spectrogram of the word mą.ma. ~ mąmą (1st male voice) FIG. 2. Spectrogram of the word mą.na. ~ mąną (1st male voice) FIG. 3. Spectrogram of the word ná.ma. ~ namą (1st male voice) FIG. 4. Spectrogram of the word mą.ma. ~ mąmą (female voice) FIG. 5. Spectrogram of the word ná.ma. ~ namą (1st male voice) FIG. 6. Spectrogram of the word sná.pa. ~ snąpą (female voice) 110 Acta Linguistica Lithuanica LXVIII Spectral characteristics of nasal sonorants in the Prienai dialect FIG. 7. Spectrogram of the word ná.ma. ~ námą (female voice) FIG. 8. Spectrogram of the word má.na. ~ mėną (female voice) FIG. 9. Spectrogram of the word mė.ne. ~ mėnę (female voice) FIG. 10. Spectrogram of the word mā.na. ~ māną (1st male voice) FIG. 11. Spectrogram of the word mā.na. ~ mėną (1st male voice) FIG. 12. Spectrogram of the word mé.ne. ~ mėnę (1st male voice) Straipsniai / Articles 111 JURGITA JAROSLAVIENĖ FIG. 13. Spectrogram of the word nìko ~ nyko (voice of the second male speaker) FIG. 14. Spectrogram of the word mìnė ~ myné (voice of the second male speaker) FIG. 15. Spectrogram of the word móko ~ móko (voice of the second male speaker) FIG. 16. Spectrogram of the word nóko ~ nóko (voice of the second male speaker) The first formant of the hard dental [n] is lower than that of the corresponding bilabial [m], whereas the F1 values of the palatalized sonorants are very similar; however, when comparing corresponding hard and palatalized members of this class, the F1 values of [m̃] and [ñ] tend more often to be somewhat higher. The first formant of the sonorants [m] and [n] of the dialect under study, pronounced in connected speech, does not differ, but occurs in a somewhat higher frequency range (in the aggregated data, F1 = 360 Hz for [m] and [n] produced by male speakers) than in the corresponding words pronounced in isolation9. 9 Thus, when comparing data, it is indeed important to consider not only the phonetic environment and suprasegmental features of the sounds under study (for example, the spectrum of stressed consonants is often more intense, and sounds are generally much longer than their unstressed counterparts, etc.), but also other factors (speech rate, intonation, etc.). It is true that the duration of nasal sonorants was not specifically investigated in this study; it was merely noted that when minimal pairs are compared (i.e., not in incidental cases), the unstressed consonant [m] is longer than the corresponding unstressed [n]. 112 Acta Linguistica Lithuanica LXVIII Prienų šnektos nosinių sonantų spektrinė charakteristika In the Northern Samogitian dialect, the spectra of both palatalized and hard nasal sonorants are almost indistinguishable. It is assumed that consonant nasalization apparently suppresses the feature of hardness and palatalization, which is perceived only from adjacent vowels (see Kliukienė 1992: 57, note 44, 63, Table 13). Thus, it is assumed that the [m]: [m̑], [n]: [n̑] opposition in this dialect is maintained only by vowels (ibid., 57, 78–79). 1 TABLE. Aggregated first-formant values of nasal sonorants in the Prienai dialect in words pronounced in isolation10 Sound | [<m>a.] | [<m̂>e.] | [<n>a.] | [<n̂>e.] | [<m>oˑ] | [<m̂>iˑ] | [<n>oˑ] | [<n̂>iˑ] Informant | F₁ (Hz) | F₁ (Hz) | F₁ (Hz) | F₁ (Hz) | F₁ (Hz) | F₁ (Hz) | F₁ (Hz) | F₁ (Hz) Male voice | 253 | 285 | 276 | 287 | 250 | 269 | 264 | 271 Female voice | 266 | 287 | 288 | 289 | 264 | 268 | 268 | 275 The second formant (F₂) of [m], [m̂], [n], and [n̂] in the Prienai dialect indicates that the bilabial sounds [m] and [m̂] have a lower timbre (see Table 2); however, when comparing only the palatalized members of the class, [m̂] and [n̂], the frequency ranges of the second formant can also be quite similar. For example, before the long high vowel [iˑ], the F₂ values of the sonorants [m̂] and [n̂] differ very little. The F₂ values of [m̂] and [n̂] pronounced before the low vowel of positional length [e.] also differ only slightly; however, a tendency was observed for the palatalized non-bilabial sound to be pronounced with a somewhat higher timbre more often (see Table 2). 2 TABLE. Aggregated second-formant values of nasal sonorants in the Prienai dialect in words pronounced in isolation11 Sound | [<m>a.] | [<m̂>e.] | [<n>a.] | [<n̂>e.] | [<m>oˑ] | [<m̂>iˑ] | [<n>oˑ] | [<n̂>iˑ] Informant | F₂ (Hz) | F₂ (Hz) | F₂ (Hz) | F₂ (Hz) | F₂ (Hz) | F₂ (Hz) | F₂ (Hz) | F₂ (Hz) Male voice | 1041 | 1623(?) | 1378 | 1704(?) | 800 | 2327(?) | 1051 | 2320(?) Female voice | 1126 | 1764(?) | 1309 | 1822(?) | 729 | 2463(?) | 1110 | 2499(?) 10 In Tables 1 and 2, the vowel following the nasal sonorant is shown in square brackets; the sample of sonorants measured was n = 36 for male voices and n = 12–24 for the female voice. 11 The question marks in the table draw attention to the difficulty of determining the formant of the stationary portion, since the spectra of palatalized nasal sonorants are considerably more often entirely faded than those of the corresponding hard consonants; therefore, the calculated F₂ values cannot be considered reliable for differentiating the spectral characteristics of palatalized [m̂] and [n̂]. It is much more reliable to measure the F₂ location in CV sequences, i.e. at the end of the consonant and the beginning of the vowel, or, in other words, the vowel onset; it is also important to take into account vowel F₂ dynamics (see Table 3; cf. also Ambrazevičius 2010: 9, Fig. 4; Grigorjevs 2012: 279, Table 2). Straipsniai / Articles 113 JURGITA JAROSLAVIENE The second formant of the Prienai dialect sonants [m], [n] pronounced in sentences (when the adjacent vowel is of the [a.] type) is somewhat higher than in the corresponding words pronounced in isolation: in the aggregated data, F₂ of [m] in connected speech is = 1190–1240 Hz, and F₂ of [n] is = 1310–1390 Hz. Comparing the palatalized and non-palatalized allophones, it may be noted that the spectra of the palatalized sonants under investigation are more often attenuated, especially F₂; therefore, the transition and dynamics of the consonant and adjacent vowel formants should be regarded as more reliable acoustic cues for distinguishing non-palatalized and palatalized sonants (for more, see the next section of this article; see also Ladefoged, Maddieson 2002: 116–117; Ambrazevičius 2010; 2012). The third formant of nasal sonants (cf. especially Figs. 2, 5–7, 10, 15, 19, etc.) is sometimes completely attenuated due to the influence of an additional resonator. Statistical calculations showed that F₃ is higher more often for palatalized sonants than for the corresponding non-palatalized ones. The spectral features of nasal sonants and adjacent vowels (formant structure and dynamics, as well as the C→V F₂ transition) are reviewed below. 1.2. Spectral features of nasal sonants and adjacent vowels Analysis of more than 500 spectrograms (see Figs. 1–24 for examples) makes it immediately clear that the formant structure and dynamics of vowels adjacent to nasal sonants differ somewhat from those of vowels adjacent to oral consonants. For example, because of nasal resonance, the second formant of [m] and [n] is lowered when pronounced before [a.], whereas the third is somewhat higher and usually attenuated (cf. also Ladefoged, Maddieson 2002: 299–300). The vowel’s first formant also tends to be fainter and somewhat higher because of the influence of the nasal formant (cf. ibid. 117, 299–300; Girdenis 2003: 225; Ladefoged 2005: 55–56, 84–85, etc.). It therefore appears that the two first formants of nasalized [a.] lie very close to one another (at approximately the 1000 Hz boundary). The dynamics of the vowel’s second formant are particularly important: for example, between non-palatalized nasal consonants, the F₂ of back vowels (especially close vowels) is often lower throughout the entire duration of articulation, whereas next to a single nasal consonant, the F₂ of vowels of the same quality and of plosive consonants generally lowers only for a certain period (see Jaroslavienė 2010: 261–264). The results of this study show that, under the influence of nasal sonants, adjacent back vowels are generally lower in timbre and articulated farther back than those adjacent to oral consonants (see also Ledichova 2012: 79–85). Next to palatalized sonants, the first and third formants of front vowels are more strongly attenuated than next to palatalized oral sounds (cf. Figs. 11, 12–14, etc.); however, no marked lowering of the second formant was observed, as in the case of the vowel [a.]. Specific examples are reviewed below, namely the spectral characteristics of vowels between or adjacent to palatalized [m̃], [ñ] and non-palatalized [m], [n], as well as the effect of nasalization on the distribution and dynamics of energy in vowel formants. For example, in the words mã.da. ~ mądą (cf. Fig. 17), mã.ta. ~ Mątą (cf. Fig. 19), ná.ta. ~ natą (cf. Fig. 23), sná.pa. ~ snąpą (cf. Fig. 6), the second formant of the first [a.] for- 114 Acta Linguistica Lithuanica LXVIII Spectral characteristics of nasal sonorants in the Prienai dialect FIG. 17. Spectrogram of the word mã.da. ~ mãdą (voice of the third male speaker) FIG. 18. Spectrogram of the word bã.da. ~ bãdą (voice of the third male speaker) FIG. 19. Spectrogram of the word mã.ta. ~ Mãtą (voice of the third male speaker) FIG. 20. Spectrogram of the word bã.ta. ~ bãtą (voice of the third male speaker) FIG. 21. Spectrogram of the word kã.pa. ~ kãpą (voice of the third male speaker) FIG. 22. Spectrogram of the word tã.ka. ~ tãką (voice of the third male speaker) Straipsniai / Articles 115 JURGITA JAROSLAVIENĖ FIG. 23. Spectrogram of the word ná.ta. ~ nãta (voice of the third male speaker) FIG. 24. Spectrogram of the word dã.ta. ~ dãta (voice of the third male speaker) The formant has lowered over the entire duration of the sound or, more often, part of it (mostly in the middle of the sound), while the third formant is more often higher and fainter than in the corresponding words between two non-nasal consonants: bá.da. ~ bãdą (cf. Fig. 18), bá.ta. ~ bãtą (cf. Fig. 20), tá.ka. ~ tãką (cf. Fig. 22), dá.ta. ~ dãtą (cf. Fig. 24), ká.pa. ~ kãpą (cf. Fig. 21). The vowel [a.] is most strongly nasalized between two nasal sonorants, i.e. in the words má.ma. ~ mãmą (cf. Figs. 1, 4), ná.ma. ~ nãmą (cf. Figs. 3, 5, 7), má.na. ~ mãną (cf. Figs. 2, 10), and so on. The degree of nasalization is indicated not only by the first two formants of the vowel [a.] being close together (F₂ is lower throughout the utterance, while F₁ is higher), but also by their energy intensity: the stronger the nasalization, the fainter (and higher) the third formant of the nasalized vowel. Examination of the spectra of other long and short vowels adjacent to nasal and non-nasal consonants also showed that vowels are most nasalized between two nasal sonorants: the third vowel formant is more often higher, and the intensity of the first formant is somewhat fainter¹². Many researchers have observed that nasal consonants tend to nasalize all the corresponding vowels in words (Kliukienė 1998: 80; Pakerys 2003: 191–192; Murinienė 2007: 62, etc.; see also Ledichova 2012: 79–85). To summarize, the parameters of formant dynamics for nasalized [a.], [e.], [o.], [i̇] (i.e. vowels adjacent to or between nasal consonants; the measured values are F₂ at the very beginning of the sound, or at vowel onset, and in the middle) are presented in Table 3. As can be seen (see Table 3), the second-formant trajectories of the sounds produced by the female and male speakers differ somewhat: the F₂ trajectories of the vowels pronounced by the male speaker are smoother than those of the female speaker (in addition, the female speaker’s formant values are generally higher than the male speaker’s). It should be recalled that the location of the onset of F₂ (i.e. the end of the consonant and the beginning of the vowel) is closely related to the place of articulation of the nasal sonorant and the adjacent vowel. ¹² See also the corresponding spectral characteristics observed in studies of nasalized vowels in some of the world’s languages (Ladefoged, Maddieson 2002: 298–300). 116 Acta Linguistica Lithuanica LXVIII Prienų šnektos nosinių sonantų spektrinė charakteristika TABLE 3. Formant frequencies (in hertz) at the beginning (F₂_pr.) and midpoint (F₂_mid.) of certain vowels in the Prienai dialect¹³ Following vowel [a.] [e.] [o˙] [i˙] Sonant F₂_pr. F₂_mid. F₂_pr. F₂_mid. F₂_pr. F₂_mid. F₂_pr. F₂_mid. [m] male 1095 1267 860 724 female 1191 1334 928 711 [m̃] male 1717 1828 2371 2563 female 1770 1850 2569 2679 [n] male 1339 1304 904 762 female 1452 1293 964 773 [ñ] male 1764 1807 2455 2500 female 1840 1723 2600 2645 By comparing the spectral properties of the allophones of hard and soft nasal sonants, the structure and quality of their formants, and the energy distribution and trajectories of the formants of neighboring sounds (especially F₂), several general conclusions can be drawn: • In the spectrum of nasal sonants, the formant bands that determine sound quality are quite clearly visible; the number of formants in the spectrum of nasal sounds is often greater; nasal sonants are characterized by spectral energy being spread across a wider frequency range due to the influence of an additional resonator; • Nasal sonants are characterized by an exceptionally low first, or nasal, formant (according to aggregated data, in the isolated words [m], [n], [m̃], [ñ] pronounced in isolation, F₁ = 250–290 Hz); • The location of the second formant of nasal sonants is related to the quality of the adjacent vowel ([m̃], [ñ] are much higher-pitched sounds than the corresponding hard [m], [n]); • The energy of the formants of nasalized vowels (especially F₃) is often attenuated, and the F₂ of close vowels is lowered; • As a result of the influence of nasal sonants, adjacent vowels—especially back vowels—are generally lower-pitched and articulated farther back than those adjacent to corresponding nonnasal consonants; however, this effect is not uniform: vowels are more nasalized between nasal sonants than when adjacent to just one nasal sonant; adjacent to, and especially between, two nasal sonants, the third vowel formant is usually higher, while the first is somewhat attenuated and tends to be higher (i.e., closer to F₂ due to the influence of the nasal formant). 13 Formant values were measured in the same words pronounced in isolation as in Tables 1 and 2; see also note 10. It should be noted that segmentation of the consonant’s post-vocalic portion and the beginning of the vowel is quite problematic. In determining formant values, primary consideration was given to the auditory impression of the consonant and vowel, oscillograms, etc. For F₂ values and assessments of vowels in isolated words and connected speech in the Prienai dialect, see Jaroslavienė 2011: 65–86. Straipsniai / Articles 117 JURGITA JAROSLAVIENĖ 2. ACOUSTIC DISTINCTIVE FEATURES OF NASAL SONORANTS An overview of the spectra of nasal consonants showed that the sounds under study are distinguished from other sounds of the Prienai dialect primarily by the dispersion of spectral energy across a broader frequency range, characteristic of nasal sonorants, and by the spectral characteristics of the adjacent vowel: formant values and dynamics (especially those of the second formant), intensity, energy dispersion, and so on. It was established that the first formant of nasal sonorants occurs in the low-frequency region (see Table 1), while the second is strongly related to the quality of the adjacent vowel (see Tables 2 and 3). The distinctive acoustic features of the soft and hard sonorants described here can be identified not only by the spectral characteristics of the consonants and the formant transitions and dynamics in C→V (or, correspondingly, V→C¹⁴ (see Table 3; cf. also the figures presented above), but also by anti-formant values (see Table 4). In other words, to distinguish [m], [m̃], [n], [n̂], [ŋ], and [ŋ̂]¹⁵ more precisely, another very important feature should also be taken into account—anti-formants, which are difficult to identify from the spectrograms and formant-trajectory plots presented above. Therefore, combined FFT spectra and LPC trajectory plots were produced (see, for example, Figs. 25–30). The location of the anti-formants (Z₁) of hard and soft [m], [m̃], [n], [n̂], [ŋ], and [ŋ̂] is strongly related to their articulatory features (place of articulation), which also depend to some extent on the following vowels. The Z₁ values of sonorants produced by men and women may also differ, but the relationships between them remain constant (cf. Grigorjevs 2012: 267–272 and literature cited there, 276–278; cf. also Ladefoged, Maddieson 2002: 116–118; Butkauskaitė 2010: 6 and literature cited there, etc.). TABLE 4. Anti-formant (Z₁) values of nasal sonorants in the Prienai dialect, in hertz¹⁶ Sonorant [m] [m̃] [n] [n̂] [ŋ] [ŋ̂] Informant (aggregated data) Man 600–900 1000–1250 900–1300 1400–1800 1900–2300 3200–4000 Woman 600–800 900–1300 1000–1400 1600–2000 2100–2400 2800–3700 According to the aggregated data from the study of the consonants under investigation in the Prienai dialect, the anti-formant values of the same hard or soft nasal sonorants produced by male informants and a female informant differ— The formant transitions and dynamics of VC-type sequences are not analyzed in greater detail in this article. The front, dental [ŋ] ([ŋ̂]) of the tongue is pronounced as a back, velar [ŋ] (or, respectively, as a mid-tongue, palatal [ŋ̂]) before the back-of-the-tongue [k], [g] (and the mid-tongue [k̂], [ĝ]) (Pakerys 2003: 134). It should be recalled that the anti-formants of nasal sonorants were measured not only adjacent to long vowels but also adjacent to short vowels (see the section “Research Materials and Methodology” for more details). 118 Acta Linguistica Lithuanica LXVIII Prienų šnektos nosinių sonantų spektrinė charakteristika The formant (in hertz) ranges overlap considerably (the Z1 values for nasal sonorants produced by men are often lower, except for [m] and the velar [ŋ]); however, the tendencies are similar in all cases: the Z1 of the labial [m] occurs in a lower frequency range than that of the corresponding non-labial sonorant [n], whereas [ŋ] occurs at higher frequencies than [ɲ]. The anti-formants of the velar [ŋ] and [ɲ] are in the highest frequency range (see Table 4). That the anti-formants of [m], [ɲ], [n], [ɲ], [ŋ], and [ɲ] differ is clearly evident from the spectral cross-section plots of the speakers’ preceding vowels (i.e. the combined FFT spectrum and LPC trajectory graphs), in which the smaller or larger gap formed between the LPC trajectory and the FFT spectrum indicates the position and value of Z1 (cf. Figs. 25–30). FIG. 25. Spectral cross-section of the hard [m] (pronounced by a woman on the left, by a man on the right) FIG. 26. Spectral cross-section of the soft [m] (pronounced by a woman on the left, by a man on the right) As the examples in Figures 25 and 26 show, the anti-formant of the hard [m] occurs at approximately 700 Hz (pronounced by a woman) and 750 Hz (pronounced by a man), whereas the Z1 of the soft one begins at the 1100 Hz (pronounced by a woman) or 1000 Hz (pronounced by a man) boundary17. 17 In some languages, the anti-formant of the nasal sonorant [m] occurs in a similar frequency range (cf. Grigorjevs 2012: 270 and references cited therein, 278; Butkauskaitė 2010: 6 and references cited therein). Straipsniai / Articles 119 JURGITA JAROSLAVIĖNĖ 27 FIG. Spectral section of the hard [ŋ] (uttered by a woman on the left, by a man on the right) 28 FIG. Spectral section of the soft [ŋ̇] (uttered by a woman on the left, by a man on the right) Figures 27 and 28 show that the antiformant of the [ŋ] uttered by the man begins at approximately 1000 Hz, while that of the [ŋ] uttered by the woman begins at 1100 Hz, i.e. higher than when the corresponding labial sonant is uttered. The Z1 of the soft [ŋ̇] begins at approximately 1600 Hz—also higher than that of the corresponding [m]. The antiformants of the soft and hard allophones of nasal velars also do not coincide: in Figure 29, the Z1 of [ŋ] occurs at approximately 2000 Hz, whereas the antiformant of [ŋ̇] is higher, beginning above 3000 Hz (see Fig. 30). The research data show that the antiformant frequency range of the same (i.e. hard or soft) sonant may vary somewhat depending on the quality (and quantity) of the adjacent (following) vowel: [ṁ] and the Z1 of [ŋ̇] are somewhat higher before the tense, long, high-pitched vowel [i̇] than before the corresponding short [i], while before [a], the Z1 of sonants [m] and [ŋ] is often higher than before [u], and so on. Thus, when the non-labial [ŋ], [ŋ̇] and especially [ŋ̩], [ŋ̩́] are pronounced, their antiformants are higher than those of labial sonants—this depends particularly on the place of articulation: the size of the resonant cavities of the mouth (and partly the nose), and the position of the tongue in the mouth and its 120 Acta Linguistica Lithuanica LXVIII Prienų šnektos nosinių sonantų spektrinė charakteristika 29 FIG. Spectral section of the velar [ŋ] (uttered by a woman on the left, by a man on the right) 30 FIG. Spectral section of the velar [ŋ̃] (uttered by a woman on the left, by a man on the right) forward–backward displacement, which is also influenced to some extent by the articulation of the adjacent vowel (cf. also Grigorjevs 2012: 269 and the literature cited there, 276–279). The antiformant of a nasal sound is highest when velars [ŋ], [ŋ̃] are pronounced, i.e. when the resonant cavity of the mouth is shortest and the middle part of the tongue retracts toward the palate. Briefly summarizing the results of the study of acoustic distinctive features, it can be stated that: • in describing the acoustic distinctive features of nasal sonants in the Prienai dialect, it is necessary to consider both the formants that determine the qualitative properties of the sonants themselves and the spectral characteristics of adjacent sounds, as well as the second formant (its transition and dynamics) of CV. The transition and dynamics of the second formant of CV should be regarded as one of the most important distinctive features of the place of articulation of soft and hard nasal sonants. All of this differentiates the consonants under consideration and distinguishes them from other sounds; • in distinguishing [m], [m̃], [n], [ñ], [ŋ], [ŋ̃] from one another, the antiformant values are also very important, and these depend particularly on the place of articulation of the sound (the resonant cavity of the mouth Straipsniai / Articles 121 JURGITA JAROSLAVIENĖ the size and shape of the cavity, tongue displacement, and certain other articulatory features when producing a sonant): in all cases, the antiformant of the labial [m] is lower than that of the corresponding non-labial sonant ([n̪]) and especially the palatal allophone [ŋʲ]; to summarize these data, the Z₁ values of the man’s [m] and [n] differ the least; in the Prienai dialect, the antiformants of the palatal allophones are the highest among all the nasal consonants studied; comparison of palatalized and non-palatalized consonants shows that the Z₁ values of [m̃], [ñ], and [ŋ̃], respectively, are higher. 3. MAIN CONCLUSIONS The results of the instrumental study of nasal sonants in the Prienai dialect show that the consonants under investigation can be differentiated by their acoustic properties and by the spectral characteristics of the consonants and adjacent vowels: The qualitative features of sonants are determined particularly by the low-frequency first (or nasal) formant and the second formant at the onset (or vowel onset) of sonants and adjacent vowels, as well as by its dynamics. Also important are the formant structure, number, intensity, and spectral spread of the consonants under investigation and adjacent sounds. Nasal sonants are characterized by spectral energy spreading across a broader frequency range due to the influence of an additional resonator. The first formant of vowels adjacent to or occurring between nasal consonants is often somewhat weakened and higher (owing to the influence of the nasal formant); the third formant of adjacent vowels is also often higher. Vowels are more strongly nasalized between two nasal sonants than when adjacent to just one nasal sonant: for example, between hard nasal consonants, the second formant of back vowels (especially close vowels) is lower throughout their articulation, whereas next to a single nasal consonant the F₂ of consonants of the same quality and quantity usually lowers only for a certain period. In distinguishing the nasal sonants [m], [m̃], [n], [ñ], [ŋ], and [ŋ̃] from one another, the Z₁ antiformant that appears in their spectra is particularly important. To summarize these data, in all cases the antiformants of labial sonants are lower than those of the corresponding non-labial sonants. Comparison of palatalized and non-palatalized consonants showed that the antiformant values of the corresponding palatalized [m̃], [ñ], and [ŋ̃] are higher. The nasal palatal sonant has the highest antiformant in the Prienai dialect. LITERATŪRA Ambrazevičius Rytis 2010: Differential Acoustical Cues for Palatalized vs Nonpala- talized Prevocalic Sonants in Lithuanian. – Žmogus ir žodis 1, 5–10. 122 Acta Linguistica Lithuanica LXVIII Prienų šnektos nosinių sonantų spektrinė charakteristika Ambrazevičius Rytis 2012: Akustiniai priebalsių palatalizacijos požymiai. – Kalbų studijos: Lingvistika (21), 5–13. Prieiga internete: http://dx.doi.org/10.5755/j01.sal.0.21.2073. Ambrazevičius Rytis 2012a: Loci of Palatalized vs Nonpalatalized Prevocalic Plosives in Lithuanian. – Žmogus ir žodis 1, 13–18. Bacevičiūtė Rima 2008: Dėl lietuvių kalbos minkštųjų priebalsių junginių (eksperimentinis tyrimas). –Perspectives of Baltic Philology. Poznań, 7–19. Butkauskaitė Edita 2010: Nazalizacija: samprata ir tyrimai. – Lietuvių kalba 4 (prieiga per internetą: www.lietuviukalba.lt). Dereškevičiūtė Sigita 2008: Dėl lietuvių kalbos sklandžiųjų priebalsių kiekybės. – Žmogus ir žodis I, 15–19. Dereškevičiūtė Sigita, Kazlauskienė Asta 2009: Dusliųjų sprogstamųjų priebalsių spektrinė analizė ir jų sprogimo trukmė. – Garsas ir jo tyrimo aspektai: metodologija ir praktika. Sud. Leskauskaitė A., Meiliūnaitė V. Vilnius: Lietuvių kalbos institutas, 98–111. Fant Gunnar 1970: Sound, Features, and Perception. – Proc. of the 6th Intern. Congr. of Phonetic Sciences. Prague, 49–60. Girdenis Aleksas 1967: Mažeikių tarmės fonologinė sistema (Filol. m. kand. dis.). Vilnius: Vilniaus universitetas. Girdenis Aleksas 2000: Kalbotyros darbai 2. Vilnius: Mokslo ir enciklopedijų leidybos institutas. Girdenis Aleksas 2003: Teoriniai lietuvių fonologijos pagrindai. Vilnius: Mokslo ir enciklopedijų leidybos institutas. Grigorjevs Juris 2008: Latviešu valodas patskaņu sistēmas akustisks un auditīvs raksturojums. Rīga: LU Latviešu valodas institūts. Grigorjevs Juris 2012: Acoustic characteristics of the Latvian sonorants. – Baltistica XLVII (2), 267–292. Grigorjevs Juris 2012a: Relation of the locus equations to the place of articulation of the Latvian consonants. – Profesoriaus Alekso Girdenio (1936–2011) atminimo konferencija, Vilniaus universitetas, 2012 m. spalio 18–20 d., Pranešimų tezės. Vilnius: Vilniaus universiteto leidykla, 22–23. Jaroslavienė Jurgita 2010: Rytinių kauniškių Prienų šnektos fonologija (Daktaro disertacija). Vilnius: Vilniaus universitetas. Jaroslavienė Jurgita 2010a: Prienų šnektos sprogstamųjų priebalsių spektro sklaida ir kitos skiriamosios ypatybės. – Baltistica XLV (2), 307–324. Jaroslavienė Jurgita 2011: Prienų šnektos izoliuotieji ir rišliosios kalbos balsiai: akustiniai ir artikuliaciniai požymiai. – Acta Linguistica Lithuanica LXIV–LXV, 65–86. Straipsniai / Articles 123 JURGITA JAROSLAVIENĖ Jassem Wiktor 1973: Podstawy fonetyki akustycznej. Warszawa: Państwowe wydawnictwo naukowe. Kliukienė Regina 1992: Šiaurės žemaičių tarmės konsonantizmo akustinės ir audicinės ypatybės (Filol. m. kand. dis.). Vilnius. Kliukienė Regina 1994: Šiaurės žemaičių priebalsių spektrai ir akustinės priebalsių klasės. – Kalbotyra XLIII (1), 43–52. Kliukienė Regina 1995: Šiaurės žemaičių intervokalinių priebalsių trukmė. – Kalbotyra XLIV, 58–68. Kliukienė Regina 1998: Šiaurės žemaičių priebalsių diferenci­nių požymių sistema. – Kalbotyra XLVII, 79–85. Kliukienė Regina 2011: Šiaurės žemaičių tarmės konsonantizmas: akustinės ir audicinės ypatybės. Vilnius: Vilniaus universitetas. Ladefoged Peter 2005: Vowels and Consonants: (An Introduction to the sounds of Languages). 2nd Edition. University of California, Los Angeles: Blackwell Publishing. Ladefoged Peter, Ian Maddieson 2002: The Sounds of the World’s Languages. Malden: Blackwell Publishers. Ledichova Edita 2012: Lietuvių bendrinės kalbos nazalizuotųjų balsių [aⁿ], [oⁿ], [uⁿ] trukmės ir kokybės ypatybių eksperimentinis tyrimas. – Žmogus ir žodis I, 79–85. Murinienė Lina 2007: Rytinių šiaurės žemaičių fonologija: vokalizmas ir prozodija. Vilnius: Lietuvių kalbos institutas, Vilniaus pedagoginis universitetas. Pakerys Antanas 2003: Lietuvių bendrinės kalbos fonetika. Vilnius: Enciklopedija. The spectral characteristics of the nasals in the subdialect of Prienai SUMMARY The main distinctive spectral features of the nasals in the subdialect of Prienai have been investigated in the experimental study described in the present article. The obtained data suggest that the nasals can be distinguished from the other sounds of the subdialect of Prienai on the basis of their set of acoustic parameters (formant structure) and transitions between it and the context vowel. Nasals are like vowels in that they can be characterizes largely in terms of their formant frequencies (one of the main properties of nasals is low first formant), but they differ in that the formants are not as loud as they are in vowels; besides, the nasals are made by blocking the sound from coming out of the mouth while allowing it to come out through the nose, and 124 Acta Linguistica Lithuanica LXVIII Prienų šnektos nosinių sonantų spektrinė charakteristika this affects the relative amplitude of the formants (both for consonants and context vowels). For example, the first two formants of that context vowel [ɑː] are very close together with the center between them being at about 1000 Hz. Usually the first formant of that nasalized vowels are weaker than the second due to the incidence of a nasal zero, and the third formant moves up if the nasalization increases or in case the vowel is fully nasalized. Since the main distinction between nasals during the nasal murmur is provided by the oral anti-resonance (oral zeros or anti-formants whose frequencies are determined by place of articulation), the nasal sonorants [m], [m̃], [n], [ñ] as well as nasal-velar allophones are characterized by the frequency values of the oral zero. The region of the oral zero was searched taking into account the length of the oral rezonator as well as the tongue configuration affected by the context vowel. The results show that the frequency of the oral zero differs because of articulatory habits and gender differences of the informants. Įteikta 2012 m. lapkričio 19 d. JURGITA JAROSLAVIENĖ Lietuvių kalbos institutas Petro Vileišio g. 5, LT-10308 Vilnius, Lietuva [email protected] Straipsniai / Articles 125