Labialization began to be discussed in Lithuanian linguistics as early as the middle of the 20th century. It was already noted then that vowels occurring next to labial consonants or vowels have different acoustic and articulatory properties. However, no comprehensive experiments had been conducted to date; observations and conclusions were based solely on auditory perception or small-scale exploratory studies. This work is one of the first attempts to establish experimentally the effect of labialization on long and short vowels in Standard Lithuanian. The qualitative characteristics of labialized allophones are analyzed by comparing vowel pronunciation data from natural and nonce words with isolated sounds. The experimental results show that the most pronounced differences are found in the third formant (F3) data of labialized allophones. Changes in the values of the second formant (F2) are also highly systematic, although the effect of labialization differs for back and front vowels. Labialized and non-labialized allophones differ least in the position of the first formant (F1) in the spectrum. Evaluation of the F1, F2 ir F3 formant values revealed that, in terms of compactness (diffuseness), gravisness, tonality, and tenseness, allophones in the CpV Cp position also differ qualitatively from the variants of vowels pronounced in isolation.
INTRODUCTION
Studies of Lithuanian phonetics have devoted considerable attention to experimental acoustic and articulatory investigations of vowels in Standard Lithuanian1 and dialectal vowels (Pakerys 1982; Kaukėnienė 2002: 41–59; 2003: 35–47; 2004: 199–211; Girdenis 2003: 222; Jaroslavienė 2014: 68–84; 2015; Jaroslavienė et al. 2019, etc.) and consonants (Vaitkevičiūtė 1965: 72–90; Mikalauskaitė 1975; Kliukienė 1994: 43–52; 1995: 58–68; 2011; Dereškevičiūtė 2008: 15–19; Dereškevičiūtė, Kazlauskienė 2009: 98–111; Ambrazevičius 2010: 5–10; 2012: 13–18; Urbanavičienė, Indričianė 2016a: 46–77; 2016b: 46–79; Urbanavičienė et al. 2019, etc.), but no comprehensive study of vowel–consonant interaction—
1 In this article, Standard Lithuanian is understood as the correct written or spoken variety of Lithuanian, based on the norms of vocabulary, grammar, spelling, and punctuation, and intended for the needs of public life (VKĮPI 2011).
BENDRINĖ KALBA 93 (2020)
http://journals.lki.lt/bendrinekalba
ISSN 2351-7204
Page 2
EDITA LEDICHOVA. Labializuotų bendrinės lietuvių kalbos balsių eksperimentinis tyrimas | 2
doi.org/10.35321/bkalba.2020.93.03
Research into coarticulation2 phenomena is still lacking. Only some cases of sound interaction have been investigated and described (Vaitkevičiūtė 1960: 207; Girdenis, Kubiliūtė-Kliukienė 1982: 30–38; Kliukienė 2002: 33–49; Jaroslavienė 2010: 133–139; 152–158; 2013: 105–125; Ledichova 2012: 76–85; Kliukienė, Raudžiūtė 2013: 111–119; Ambrazevičius, Leskauskaitė 2014; Jaroslavienė et al. 2019, among others).
Two cases of coarticulation in the speech stream are generally distinguished: assimilation and accommodation (Pakerys 2003: 178–179; Thomas 2010: 174, 293–294). Assimilation is the complete or partial resemblance of a sound to other sounds in the same or a neighboring word, which changes the phonological value of that sound (LKE 2008: 36). For example, the consonant [p] may be replaced by the consonant [b]: lĭpa [ˈlʲipʲə]³: lĭpdavo [ˈlʲibdɛʋoː]. Accommodation, on the other hand, is a type of sound adjustment that usually does not cause a change in phonemes, but only minor sound changes; the result of this process is variants, or allophones, of the same phonemes. Such sound changes are usually not reflected in writing (Pakerys 2003: 189).
This article examines one type of accommodation—labialization. Labialization (Latin labialis ‘labial’), or lip-rounding, is defined as the conversion of a non-labial sound into a labial one. It is caused by pronouncing a sound with the lips additionally protruded forward (Ladefoged, Maddieson 2008: 356; Ladefoged, Johnson 2014: 245), which lengthens the oral cavity and lowers the sound’s timbre (LKE 2008: 296); however, as in all cases of accommodation, this is only partial sound adjustment and does not alter the sound’s distinctive features or the phonemic inventory of the language (or dialect).
Although the term labialization is mentioned quite frequently in the work of Lithuanian linguists, no comprehensive studies of labialized sounds have been conducted. Elžbieta Mikalauskaitė (1975: 12) and Antanas Pakerys (2003: 192) do mention in their works on phonetics the possible effect of labial consonants on vowels, but they do not investigate this in greater detail or confirm these observations through experimental research; other works on Lithuanian phonetics likewise describe only the most general features of labialization4.
2 Coarticulation (English coarticulation) is defined as the overlap of adjacent sounds, resulting from the vocal tract continually changing shape in preparation for the articulation of the next sound (Ladefoged, Johnson 2014: 75).
3 Works on the phonetics of Standard Lithuanian and its dialects usually use the characters of the Lithuanian phonetic alphabet. However, to make the findings more accessible to researchers of other languages as well, this article uses the standardized universal International Phonetic Alphabet (IPA), created by the International Phonetic Association (for more on IPA symbols, see Ball et al. 2018: 155–164; on the applicability of IPA symbols to representing the sounds of Standard Lithuanian, see Jaroslavienė 2017: 196–218; Kazlauskienė 2018: 18–19, 54–55, 57; on their use for representing dialect sounds, see Bakšienė, Čepaitienė 2017: 105–135). The International Phonetic Alphabet does not use the usual stress marks; only the symbol “ˈ” marks primary stress. However, Lithuanian requires falling and rising-falling accented syllables to be marked differently, so this study follows the intonation-marking system proposed by Asta Kazlauskienė: “¹” marks the falling accent, and “²” the rising-falling accent (Kazlauskienė 2018: 19).
4 The term labialization is frequently used in studies of dialects; it is characteristic of the Anykščiai and Kupiškis dialects (Salys 1992: 119–120; Zinkevičius 1994: 68–69; 2006: 86–89; LKTCh 2004: 142, 152; Markevičienė 2011: 305; Balčiūnienė et al. 2017: 47), as well as some Samogitian subdialects, for example, those of Klaipėda, Kretinga, and Plungė (Girdenis, Riaubiškytė 1981: 92; Zinkevičius 1994: 122; LKTCh 2004: 272; LKE 2008: 296). However, this is not a change in sounds caused by coarticulation and therefore falls outside the scope of this article’s research.
BENDRINĖ KALBA 93 (2020)
http://journals.lki.lt/bendrinekalba
ISSN 2351-7204
Page 3
EDITA LEDICHOVA. Labializuotų bendrinės lietuvių kalbos balsių eksperimentinis tyrimas | 3
doi.org/10.35321/bkalba.2020.93.03
The fundamental principles of labialization have already been studied and described in considerable detail in foreign-language linguistics. Changes in the values of the second (F2) and third (F3) formants are considered the most important qualitative features of labialized vowels. A pronounced decrease in F2 values has been observed for front vowels adjacent to labial consonants, whereas back vowels, under the effect of labialization, more often tend toward a fronter articulation; that is, their F2 more often resonates at higher frequencies. In addition, the F3 values of vowels between labial consonants are at lower frequencies (Stevens, House 1963: 111–128; Ohde, Sharf 1975: 923–927; Strange et al. 1976: 213–224; Hillenbrand et al. 2001: 748–763; Flemming 2002: 22; Steinlen 2005: 87–91, 101–106, 118–122; Thomas 2011: 49; Fletcher, Bucher 2014: 98; Renwick 2014: 186; Sylak-Glassman 2014: 17–38; Beeley 2015, among others). Experimental studies have not documented any more pronounced qualitative change in the first formant associated with labialization.
The aim of this study is to systematically and comprehensively investigate and describe, through experimental research, the effect of labialization5 on the quality features of long6 [ɑː, oː, uː, æː, eː, iː] and short [ɐ, ɔ, ʊ, ɪ, ɛ] vowels in Standard Lithuanian. The objectives are: 1) to compile a database of target words; 2) to find suitable speakers and record the target words; 3) to conduct experimental studies of allophone quality using specialized software; 4) to compare and describe the acoustic features of labialized and isolated vowel variants.
1. RESEARCH METHODOLOGY
The study consists of two parts. First, the vowels of natural Lithuanian words7, mostly disyllabic, were examined: [ɑː] – bąbą [²ˈbɑːbɑː], pãpadė [²ˈpɑːpɐdæː]; [oː] – bóbos [¹ˈboːboːs], põpo [²ˈpoːpoː]; [uː] – būbų [²ˈbuːbuː], pūpso [²ˈpuːpsoː]; [æː] – bėbė [²ˈbʲæːbʲæː], pėplė [²ˈpʲæːpʲlæː]; [eː] – bėbė [²ˈbʲeːbʲeː], pėpė [²ˈpʲeːpʲeː]; [iː] – riebýbė [rʲɪe¹bʲiːbʲeː], pỹpė [²ˈpʲiːpʲeː]; [ɛ] – bàbsi [ˈbæpsʲi], pàpsi
5 The effect of labialization on vowels occurring between the labial consonants [p], [b] was investigated (in graphical notation, the target position is denoted CₚVCₚ).
6 Although contemporary Lithuanian linguistics holds that /ie/ and /uo/ function as long vowels in speech (Bacevičiūtė 2002: 5–17; Girdenis 2003: 102–104; 2009: 213–242), the methodology for studying and describing these sounds with changing height differs from that for sounds with constant articulation; therefore, they were not included in the experimental study.
7 In compiling the database of natural Lithuanian words to be examined and seeking the most accurate experimental results, an effort was made to ensure that as many of the target words as possible were disyllabic and, wherever possible, that the target vowels occurred between consonants with the same place of articulation and bore the same (falling) accent. However, ideal phonetic conditions are often difficult to create, so one or two trisyllabic words or words bearing a rising accent had to be included in the study.
BENDRINĖ KALBA 93 (2020)
http://journals.lki.lt/bendrinekalba
ISSN 2351-7204
Page 4
EDITA LEDICHOVA. Labializuotų bendrinės lietuvių kalbos balsių eksperimentinis tyrimas | 4
doi.org/10.35321/bkalba.2020.93.03
[ˈpʲepʲsʲ]; [ɔ] – bòbo [ˈbɔbɔ:], pòpo [ˈpɔpɔ:]; [ɔ] – bùbų [ˈbɔbų:], pìpų [ˈpɔpų:]; [ɪ] – bìblijà, pìpkì [ˈpʲipʲkʲi:]; [ɛ] – bèbyti [ˈbʲɛbʲi:tʲi], pèpsi [ˈpʲɛpʲsʲi]. The second part of the study consists of vowels in artificial words8, that is, words artificially formed according to the structural features of Lithuanian words but not found in Lithuanian: [ɑ:] – baba9 [bɑ:bɑ:], papa [pɑ:pɑ:]; [ɔ:] – bobo [bɔ:bɔ:], popo [pɔ:pɔ:]; [u:] – būbū [bu:bu:], pūpū [pu:pu:]; [æ:] – bebe [bæ:bæ:], pepe [pæ:pæ:]; [e:] – bėbė [bʲe:bʲe:], pėpė [pʲe:pʲe:]; [i:] – byby [bʲi:bʲi:], pypy [pʲi:pʲi:]; [ɛ] – baba [bɛbɛ], papa [pɛpɛ]; [ɔ] – bobo [bɔbɔ], popo [pɔpɔ]; [ʊ] – bubū [bʊbʊ], pupu [pʊpʊ]; [ɪ] – bibi [bʲiˈbʲi], pipi [pʲiˈpʲi]; [ɛ] – bebe [bʲɛbʲɛ], pepe [pʲɛpʲɛ]. These words are ideal examples for experimental study because they allow vowels occurring between consonants with the same place of articulation to be examined comparatively.
The vowel allophones in natural and artificial words were compared with the quality features of vowels pronounced in isolation10. Sounds pronounced in isolation, that is, without any context, reflect natural articulation. When pronouncing them, one is freed from the influence of adjacent sounds and other factors that affect changes in quality parameters; therefore, such a vowel is the closest approximation to an acoustic reference. In addition, comparing vowels pronounced in isolation with those pronounced in connected speech makes it possible to identify differences in the qualitative features of allophones. The isolated vowel was presented in a short declarative sentence, for example, I am saying the long vowel [ɑ:], I am saying the long vowel [i:], and so on. Speakers were asked to read the sentences at a normal pace, without stressing or lengthening anything deliberately, and to pause briefly before pronouncing the target vowel.
In many studies of Lithuanian experimental phonetics, vowel formant values are calculated by taking the mean of the formant values across the entire sound (see Atkočaitytė 2002; Kazlauskaitė 2002a; Bacevičiūtė 2004; Kaukėnienė 2004; Leskauskaitė 2004; Trumpa 2005; Murinienė 2007). However, this approach aggregates not only the formant measurements of the stationary portion of the sound but also those of the transition segments—the sound's onset and offset—which are affected most by coarticulation with adjacent sounds. Therefore, in this study, sound formant values were calculated differently: the allophone was divided into three parts, and values were calculated for the beginning of the segment (position I), the stationary middle portion (position II), and the end
8 Many experimental studies of sound interaction use this methodology abroad (e.g., Whalen 1990: 149–176; Marchal, Hardcastle 1993: 137–153; Boyce, Epsy-Wilson 1997: 3741–3753; Herrick 2007: 231–254; Strange et al. 2007: 1111–1129; Maxwell et al. 2015; Zellou, Pacha 2018: 1–24), and some have also been conducted using Lithuanian phonetic data (Ledichova 2012: 76–85; Grigorjevs, Jaroslavienė 2014: 35–49).
9 Artificial words have no stress or accent, so the syllable that informants were asked to emphasize was underlined.
10 Vowels pronounced in isolation should not be confused with Jones's cardinal vowels—the universally recognized reference vowels for vowel quality—which are used as theoretical constructs to describe the vowel systems of particular languages or dialects but do not correspond to the actual vowels of any specific language (for a fuller discussion, see Jaroslavienė, Kaukėnienė 2004: 25–38; for comparisons of the acoustic and articulatory characteristics of vowels pronounced in isolation in individual Lithuanian dialects with Jones's cardinal vowels, see Murinienė 1998: 91–105; Bacevičiūtė 2000: 13–17; Leskauskaitė 2000: 83–94; Kazlauskaitė 2002b: 61–78; Urbanavičienė 2004: 65–80, etc.).
BENDRINĖ KALBA 93 (2020)
http://journals.lki.lt/bendrinekalba
ISSN 2351-7204
Page 5
EDITA LEDICHOVA. Labializuotų bendrinės lietuvių kalbos balsių eksperimentinis tyrimas | 5
doi.org/10.35321/bkalba.2020.93.03
(position III), i.e., the formant values for 1/3 of the sound segment. A total of 1,056 realizations of labialized allophones were examined.
The material for the experiment was collected in Vilnius. The examples for the study were recorded by two male speakers: A. S. (23 years old) and K. B. (22 years old), who came from the Western Aukštaitian dialect area and had good Lithuanian Standard pronunciation, which they actively used in everyday communication. Representatives of this dialect area preserve the quantity of long vowels particularly well, while the phonetic system and grammatical structure of the dialect itself are almost identical to those of Standard Lithuanian (LKE 2008: 310). At the time the research material was recorded, both speakers were third-year students in the acting program at the Music Academy. Their articulation had no noticeable deficiencies, and their pronunciation met the norms of Standard Lithuanian.
To prevent the speakers from knowing the purpose of the study and influencing the experimental results, the test words were presented in random order; in addition, words of analogous structure that were irrelevant to the study were inserted. The material was read at a customary, natural speaking rate, with a consistent, emotionally neutral declarative intonation and without lowering the voice at the end of a word. Each informant read the words three times.
For segmenting the recordings and conducting spectral analysis, the study used PRAAT, a sound-processing and analysis program developed by University of Amsterdam researchers Paul Boersma and David Weenink. Using it, the values of the first four formants (F1, F2, F3, F4) were calculated in hertz (Hz); formant values in the study were taken to be those at the beginning, in the steady-state (middle) portion, and at the end of the sound, rounded to the nearest 10 hertz.
To characterize more precisely the qualitative relationships among the allophones analyzed, the following acoustic parameters were calculated from the values (Hz) of the first, second, and third formants at the initial, steady-state, and final articulation positions: compactness (C), tonality (T), bemolness (b), diffuseness (df), and tenseness (it). The arithmetic operations were performed using the Excel spreadsheet in the Microsoft Office suite, which calculated indices for the segments examined according to the research method developed by Raimund G. Piotrovsky (Пиотровский 1960: 24–38). The results obtained help assess how allophone quality changes when it is pronounced in isolation and together with consonants, and which part of the segment under study—the beginning, steady-state portion (middle), or end—is most affected by coarticulation.
Pagal Bladono ir Fanto formulę (Bladon, Fant 1978: 2–3) papildomai skaičiuota vadinamosios efektyviosios antrosios formantės (F2') reikšmė.
BENDRINĖ KALBA 93 (2020) http://journals.lki.lt/bendrinekalba ISSN 2351-7204
Page 6
EDITA LEDICHOVA. Labializuotų bendrinės lietuvių kalbos balsių eksperimentinis tyrimas | 6
doi.org/10.35321/bkalba.2020.93.03
c is calculated on the assumption that B2 = 67 Hz, and K(f) = 12 · F2/1400, according to the following formula:
Į barkus rezultatai paversti pagal Traunmüllerio formulę (Traunmüller 1988: 97).
Bark = 26,81/ (1 + (1960/f)) − 0,53. (3)
The relationships among the vowel variants being compared are represented graphically in psychoacoustic F2/ F1 space. The allophones examined are marked there by circles with a diameter of 1 Bark (z), while the distances between the circles indicate the degree of difference in perceived quality: the farther apart the circles are, the greater the qualitative difference between the sounds (Jaroslavienė 2015: 5). It is argued that this representation of vowel relationships conveys auditory properties more accurately because it takes account of the logarithmic nature of perception (Grigorjevs 2013: 303).
2. RESULTS OF THE STUDY OF THE QUALITY OF LABIALIZED LONG AND SHORT VOWELS
Labialized vowels are assessed by comparing the values of their first three formants (F1, F2, and F3) at the beginning, in the stationary (middle) portion, and at the end with the corresponding formant values of vowels pronounced in isolation. This establishes which part of the allophone under study is most affected by the labial consonant. The scholarly literature (e.g., Ladefoged 2008: 440; Krug, Schlüter 2013: 234; Carignan et al. 2017: 363; Wayland 2019: 170) states that the first formant F1 is inversely proportional to tongue height: the lower this formant, the higher the tongue is raised and the closer the sound; the second formant F2 is associated with horizontal tongue movement: the higher this formant, the more front the sound. It is generally considered sufficient to use F1 and F2 values when describing the acoustic properties of vowels; other values most often convey only expressive information or individual voice characteristics (cf.
BENDRINĖ KALBA 93 (2020)
http://journals.lki.lt/bendrinekalba
ISSN 2351-7204
Page 7
EDITA LEDICHOVA. Labializuotų bendrinės lietuvių kalbos balsių eksperimentinis tyrimas | 7
doi.org/10.35321/bkalba.2020.93.03
Girdenis 2003: 221–222; Ladefoged 2008: 48). Tačiau, kaip atskleis šio tyrimo rezultatai, kai kuriais atvejais trečiosios formantės F₃ duomenys gali būti ypač svarbūs diferencijuojant balsinius alofonus.
2.1. Long vowels in natural words
First, the spectral characteristics of long labialized allophones pronounced in natural words are analyzed, and the results are then compared with data from the study of nonce words. The characteristics of short labialized vowels and vowels pronounced in isolation are described in the same way.
TABLE 1. Quality features of the long vowels [ɑː], [oː], [uː], [æː], [eː], [iː] in labialized contexts in natural words and when pronounced in isolation
Notes: I – beginning of the vowel, II – middle, III – end; square brackets indicate the formant values and their assessment for vowels pronounced in isolation.
After analyzing trends in the formant values of long labialized vowels in natural words, assessing their qualitative parameters (see Table 1), and comparing the results obtained with
BENDRINĖ KALBA 93 (2020)
http://journals.lki.lt/bendrinekalba
ISSN 2351-7204
Page 8
EDITA LEDICHOVA. Labializuotų bendrinės lietuvių kalbos balsių eksperimentinis tyrimas | 8
doi.org/10.35321/bkalba.2020.93.03
Based on the data for vowels pronounced in isolation, several preliminary conclusions can be drawn (illustrated in Fig. 1). More pronounced differences in the first-formant (F₁) values of labialized vowels were recorded only for the mid vowels [oː] and [eː]—the F₁ of allophones occurring between labial consonants is slightly higher than that of the corresponding vowels pronounced in isolation, so they should be slightly more open. A stronger effect of coarticulation on the F₁ values of labialized [oː] was established when the stationary and final articulatory positions were assessed (cf. Cₚ[oː]:Cₚ F₁ = 550 Hz, 560 Hz and [oː] F₁ = 470 Hz; also cf. 500 Hz and 470 Hz), whereas all three positions of labialized [eː] are affected by adjacent consonants to a very similar degree (cf. Cₚ[eː]:Cₚ F₁ = 430 Hz, 440 Hz, 450 Hz and [eː] F₁ = 390 Hz). The results for the other vowel allophones studied show no significant qualitative change in the position of the first formant in the spectrum.
Differences between the second-formant values of labialized and non-labialized allophones are much more pronounced. As can be observed, they are related to the backness of the vowels. The F₂ values of the long labialized back vowels [ɑː], [oː], [uː] are at higher frequencies, so these allophones should be regarded as having a more front articulation than vowels pronounced in isolation. This is indeed highly likely, since the consonants [p], [b], between which the labialized vowels occurred, are also articulated in the front part of the mouth—the closure is formed with both lips pressed firmly together. When these consonants are pronounced, the tongue also shifts slightly forward. It is therefore natural that, in pronouncing a vowel immediately afterward, the speech organs do not have time to return to a back position, and back vowels are pronounced somewhat further forward. The F₂ values of the labialized front-vowel allophones [æː], [eː] and [iː] are lower than those of vowels pronounced in isolation. From an articulatory perspective, this relationship should be interpreted as follows: the tongue retracts toward the hard palate and the vowel becomes more back. It was established that coarticulation has a stronger effect on
BENDRINĖ KALBA 93 (2020) http://journals.lki.lt/bendrinekalba ISSN 2351-7204
Page 9
EDITA LEDICHOVA. Labializuotų bendrinės lietuvių kalbos balsių eksperimentinis tyrimas | 9
doi.org/10.35321/bkalba.2020.93.03
the initial and/or final part of the segment under study; however, the entire allophone affected by the labial consonants [p], [b] also differs markedly in quality from the vowel pronounced in isolation.
The direction of change in third-formant values is also highly consistent—in every case, and in every articulatory position of the labialized vowel under study, F₃ is at considerably lower frequencies.
FIG. 1. Spectra of labialized (---) and isolated (....) long vowels in natural words
Preliminary results of the experimental study
patvirtina užsienio fonetikos tyrėjų (Stevens, House 1963: 111–128; Ohde, Sharf 1975: 923–927; Hillenbrand ir kt. 2001: 748–763; Cole ir kt. 2010: 167–184; Paillereau 2016) darytas išvadas apie antrosios formantės reikšmių kitimo tendencijas, taip pat paremia Edwardo Flemmingo (2002: 22), Eriko R. Thomaso (2011: 49), Helenos Beeley (2015) ir Andrew Bucherio, Janetos Fletcher (2014: 98) mintį, kad labializacija labiausiai susijusi su nuosekliu trečiosios formantės reikšmių mažėjimu.
It should be noted that the change in first-formant values is very slight and, as may be surmised, is only weakly related to labialization. Foreign phoneticians, too, do not report more pronounced F₁ differences.
The results of assessing formant values also allow some preliminary conclusions to be drawn; however, the trends identified are likewise closely related to the inherent acoustic properties of vowels. Low vowels are inherently compact, so changes in the acoustic properties of their allophones are best reflected by compactness indices. The low vowel [ɑ:] between the labial consonants [p], [b] is less compact than isolated [ɑ:] (cf. Cₚ[ɑ:]Cₚ C = 908, 915 and [ɑ:] C = 923), whereas the compactness-index values of the labialized allophone of low [æ:] are higher in all three articulatory positions (cf. Cₚ[æ:]Cₚ C = 869, 874, 876 and [æ:] C = 866). The compactness-index values are also higher for the labialized mid vowel [e:] (cf. Cₚ[e:]Cₚ C = 784, 786, 791 and [e:] C = 763). For high vowels, the diffuseness index is more informative: allophones in CₚVCₚ position, in the initial, stationary, and final articulatory positions, are less diffuse than vowel variants pronounced in isolation (cf. Cₚ[u:]Cₚ df = 860, 930 and [u:] df = 990; Cₚ[i:]Cₚ df = 1950, 2000, 2010 and [i:] df = 2100). All three articulatory positions of labialized mid [o:] also have lower diffuseness-index values (cf. Cₚ[o:]Cₚ df = 640, 610 and [o:] df = 780). Differences in flatness between labialized and isolated vowel variants were recorded only for [æ:], [i:] and [u:]. The front vowels [æ:] and [i:] occurring between labial consonants are flatter (cf. Cₚ[æ:]Cₚ b = 106 and [æ:] b = 105; Cₚ[i:]Cₚ b = 108 and [i:] b = 107)—when articulated, lip activity should be greater than when a vowel is pronounced in isolation. It was expected that the flatness indices would also reveal the degree of labialization of the rounded vowel [u:], but the labial consonants do not enhance the characteristic
BENDRINĖ KALBA 93 (2020)
http://journals.lki.lt/bendrinekalba
ISSN 2351-7204
Page 10
EDITA LEDICHOVA. Labializuotų bendrinės lietuvių kalbos balsių eksperimentinis tyrimas | 10
doi.org/10.35321/bkalba.2020.93.03
the feature of lip rounding, even results in a lower degree of its bemolization (cf. Cₚ[u:]Cₚ b = 113 and [u:] b = 114). The changes in the second formant clearly showed that labialized back vowels are articulated farther forward, while front vowels are articulated farther back, than isolated vowels; the same tendency is characteristic of the tonality parameter values: for the [ɑ:], [o:], and [u:] allophones between labial consonants, higher tonality index values were found in all three articulation positions studied, so they should be regarded as having a higher timbre (cf. Cₚ[ɑ:]Cₚ T = −40, −65, −69 and [ɑ:] T = −141; Cₚ[o:]Cₚ T = −250, −310, −320 and [o:] T = −350; Cₚ[u:]Cₚ T = −246, −303, −238 and [u:] T = −336); the tonality indices of the labialized [æ:], [e:], and [i:] variants are lower than those of the isolated vowels, and their timbre is therefore lower as well (cf. Cₚ[æ:]Cₚ T = 182, 173, 190 and [æ:] T = 223; Cₚ[e:]Cₚ T = 579, 581, 554 and [e:] T = 664; Cₚ[i:]Cₚ T = 768, 805, 803 and [i:] T = 817). A general quality feature distinguishing labialized vowel variants from those pronounced in isolation is tenseness: the articulation of allophones in the CₚVCₚ position is characterized by lower tension of the speech organs.
FIG. 2. Relationships between labialized long vowels in natural words (green circles) and long vowels pronounced in isolation (yellow circles) in psychoacoustic space
The relationships between vowel variants affected by coarticulation are represented graphically in psychoacoustic F₂′/F₁ space in barks (z). It is argued that this representation of vowel relationships more accurately captures auditory characteristics, because it takes into account the logarithmic nature of perception (Grigorjevs 2013: 303).
As can be observed, most of the long labialized and isolated vowel variants studied differ markedly in quality (see Fig. 2)11, particularly [o:];
11 Fig. 2 and Figures 4 and 7–8 present the formant values, in barks (z), for labialized vowels in the initial pronunciation position (position I), which is often affected somewhat more strongly by labialization, and for long vowels pronounced in isolation.
BENDRINĖ KALBA 93 (2020)
http://journals.lki.lt/bendrinekalba
ISSN 2351-7204
Page 11
EDITA LEDICHOVA. Labializuotų bendrinės lietuvių kalbos balsių eksperimentinis tyrimas | 11
doi.org/10.35321/bkalba.2020.93.03
the [uː], [æː], and [eː] allophones—their comparative circles, with a diameter of 1 bark, are clearly far apart from one another in psychoacoustic space; thus, the vowel variants studied should be perceived as sounds of different quality. The [ɑː] allophones can also be considered sounds of different quality, since the comparative circles touch only slightly, but they differ in quality much less than the [oː], [uː], [æː], or [eː] variants discussed above. The [iː] comparative circles overlap most in psychoacoustic space—they almost coincide—so no perceptible qualitative difference between these allophones can be claimed. It should be noted that all the labialized vowel variants are shifted closer to the center of psychoacoustic space, whereas the variants pronounced in isolation occupy a peripheral position relative to the center.
2.2 Long vowels in nonce words
Qualitative spectral changes caused by labialization were also analyzed using nonce words, which, as already stated, provide an ideal research setting. The experiment showed that the quality features of labialized and non-labialized allophones in both natural words and nonce words exhibit the same tendencies.
TABLE 2. Quality features of labialized and isolated long vowels [ɑː], [ɔː], [uː], [æː], [eː], [iː] in nonce words
Iš dirbtinių žodžių ilgųjų labializuotų ir izoliuotai ištartų [ɑː], [oː], [uː], [æː], [eː], [iː] variantų kokybės tyrimo rezultatų (žr. 2 lent. ir 3 pav.) aiškiai matyti, kad labializuoto vidutinio pakilimo balsio [oː] ir ypač [eː] pirmoji formantė linkusi būti aukštesnių dažnių nei izoliuotai ištartų variantų formantė. Ryškesnis abiejų CpVCp pozicijos alofonų F1 skirtumas užfiksuotas tiriant pradines artikuliacijos pozicijas (plg. Cp[oː]Cp F1 = 510 Hz, 530 Hz ir [oː] F1 = 470 Hz; Cp[eː]Cp F1 = 580 Hz, 610 Hz ir [eː] F1 = 390 Hz). Stebint tokią pačią tendenciją, kokia išryškėjo tiriant natūralių žodžių tyrimo duomenis, galima patikimiau teigti, jog labializuotų ir nelabializuotų alofonų pirmosios formantės padėtis spektre susijusi su tiriamųjų balsių pakilimu – tarp lūpinių priebalsių [p], [b] esantys vidutinio pakilimo [oː] ir [eː] turėtų būti kiek atviresni bei žemesnio pakilimo nei izoliuotai ištarti jų variantai. Aukštutinio ir žemutinio pakilimo balsių tyrimo rezultatai jokio aiškesnio pirmosios formantės reikšmių kitimo neatskleidė. Kai kurie užsienio fonetikos specialistai irgi yra nurodę, kad F1 skirtumas analizuojant labializuotus ir nelabializuotus balsių variantus dažnai būna visai
and the spectra of isolated (...) long vowels
BENDRINĖ KALBA 93 (2020)
http://journals.lki.lt/bendrinekalba
ISSN 2351-7204
Page 13
EDITA LEDICHOVA. Labializuotų bendrinės lietuvių kalbos balsių eksperimentinis tyrimas | 13
doi.org/10.35321/bkalba.2020.93.03
mažas (plačiau žr. Steinlen 2005: 87–91, 101–106, 118–122; Paillereau 2016).
The analysis of second-formant values also confirmed the tendency identified in the previous stage of the experimental study: the spectral position of F₂ in labialized allophones is related to place of articulation. F₂ is always higher in back [ɑː], [oː], [uː] vowels between labial consonants than the corresponding formant in isolated vowels; therefore, labialized allophones should tend toward more front articulation. In front [æː], [eː], [iː], F₂ is lower when labialized, meaning that during their articulation the tongue retracts toward the hard palate and the allophones become backed. Foreign phoneticians have likewise noted and confirmed through experimental research that the F₂ of labialized vowels resonates in a higher or lower frequency range depending on place of articulation: front variants are characterized by decreasing F₂ values, while back variants show an increase (Stevens, House 1963: 111–128; Strange ir kt. 1976: 213–224; Hillenbrand ir kt. 2001: 748–763; Cole ir kt. 2010: 167–184; Renwick 2014: 186; Paillereau 2016).
The third-formant data are also very clear. As already stated in the analysis of the qualitative characteristics of allophones in natural words, and as also demonstrated by experimental results obtained by other phoneticians (see Flemming 2002: 22; Thomas 2011: 49; Bucher, Fletcher 2014: 98; Beeley 2015 for more detail), the F₃ values of labialized allophones are lower in every case than the F₃ values of their isolatedly pronounced counterparts. In view of this tendency, it can be concluded that the third formant is the one most affected by labialization.
The results of evaluating quality values almost coincide with the data from the natural-word study (cf. Table 1), though one or two differences emerged. The low allophone [ɑː] in the CₚVCₚ position is less compact than the isolatedly pronounced variant (cf. Cₚ[ɑː]Cₚ C = 914, 918, 915 and [ɑː] C = 923), whereas the low [æː] is more compact (cf. Cₚ[æː]Cₚ C = 869, 868 and [æː] C = 866), although its compactness index in the initial articulation position is lower. The mid-height [eː] between labial consonants also appears more compact (cf. Cₚ[eː]Cₚ C = 830, 835, 836 and [eː] C = 763). The trends in the diffuseness index for the high [uː], [iː] and mid-height [oː] are the same as for natural words: labialized [uː], [iː], [oː] allophones have lower diffuseness in the initial, steady-state and final phases of articulation than their isolatedly pronounced counterparts (cf. Cₚ[uː]Cₚ df = 850, 960, 910 and [uː] df = 990; Cₚ[iː]Cₚ df = 1950,
BENDRINĖ KALBA 93 (2020)
http://journals.lki.lt/bendrinekalba
ISSN 2351-7204
Page 14
EDITA LEDICHOVA. Labializuotų bendrinės lietuvių kalbos balsių eksperimentinis tyrimas | 14
doi.org/10.35321/bkalba.2020.93.03
1980, 1990 and [i:] df = 2100; Cp[o:]Cp df = 670, 660 and [o:] df = 780). The values of the rounding index are consistent with the results of the natural-word study—differences between labialized and isolatedly pronounced vowel variants were observed only for [æ:], [i:] and [u:]: the rounding values of [æ:], [i:] allophones between the labial consonants [p], [b] are higher (cf. Cp[æ:]Cp b = 106 and [æ:] b = 105; Cp[i:]Cp b = 108 and [i:] b = 107). Thus, lip activity is presumably somewhat greater during their articulation. Conversely, the isolatedly pronounced variant of vowel [u:] had higher rounding values (cf. Cp[u:]Cp b = 112, 113 and [u:] b = 114). The rounding indices of the other allophones studied do not differ. As in the previous research stage, two vowel groups can be distinguished by tonality: labialized back [ɑ:], [o:], [u:] allophones, which have a higher degree of tonality and thus a higher pitch (cf. Cp[ɑ:]Cp T = −70, −90, −84 and [ɑ:] T = −141; Cp[o:]Cp T = −280, −305, −316 and [o:] T = −350; Cp[u:]Cp T = −230, −311, −322 and [u:] T = −336), and labialized front [e:] and [i:], whose tonality-index values are lower and thus whose pitch is lower than that of the isolatedly pronounced variants (cf. Cp[e:]Cp T = 424, 415, 403 and [e:] T = 664; Cp[i:]Cp T = 757, 774, 781 and [i:] T = 817). Clearly, changes in tonality correspond directly to the second-formant data. It should be noted that the artificial-word study also revealed a different relationship for labialized [æ:], which has a lower timbre, as well as for the back-articulated vowels: although this vowel also has somewhat lower F2 values, its tonality coefficients are higher (cf. Cp[æ:]Cp T = 280, 246, 228 and [æ:] T = 223). This relationship can probably be associated with the changing articulation of [æ:]: at the beginning of its production the vowel is close and front, while its final part is quite open, resembling [ɑ:]. It may be that in pronouncing artificial words, the final part of the vowel was emphasized more, and it was pronounced very openly; consequently, its allophonic relationships are closer to the pronunciation tendencies of [ɑ:] than of [æ:]. Tension indices, as with vowels in natural words, indicate that the articulation of labialized [ɑ:], [o:], [u:], [æ:], [e:], [i:] allophones involves lower articulatory tension, except in the steady-state portion of [æ:].
BENDRINĖ KALBA 93 (2020)
http://journals.lki.lt/bendrinekalba
ISSN 2351-7204
Page 15
EDITA LEDICHOVA. Labializuotų bendrinės lietuvių kalbos balsių eksperimentinis tyrimas | 15
doi.org/10.35321/bkalba.2020.93.03
FIG. 4. Relationships between labialized (green circles) and isolated (yellow circles) long vowels in artificial words in psychoacoustic space
A comparison of the relationships between the vowel variants examined in psychoacoustic space (see Fig. 4) clearly shows that the labialized and isolated long-vowel variants in artificial words exhibit trends similar to those of allophones in natural words (cf. Fig. 2): for all vowels except [ɑː] and [iː], the circles being compared are clearly separated or at least do not touch at their edges; thus, the allophones should be perceived as sounds of different quality. The greatest distance in the psychoacoustic-space diagram was found between the variants of [eː]—the difference amounts to more than 2 Bark. As the study of natural words also shows, the qualitative difference between the labialized and unlabialized allophones of [ɑː] is small: the circles overlap only slightly, which means that the vowel variants are perceived as having very similar, though not identical, quality. The same can be said of [iː]. Although the circles for the labialized and unlabialized variants of [iː] touch in psychoacoustic space, the artificial-word data indicate a greater qualitative difference between the allophones being compared than do the results of natural-word pronunciation (cf. Fig. 4). The recurring position of the allophones in the psychoacoustic-space diagram should not be overlooked: all labialized vowels exhibit a greater or lesser shift toward the center of psychoacoustic space, whereas their isolated variants tend to occupy a more peripheral position, i.e. they are distributed along the margins of the quadrants of the coordinate system.
2.3. Short vowels in natural and artificial words
BENDRINĖ KALBA 93 (2020)
http://journals.lki.lt/bendrinekalba
ISSN 2351-7204
Page 16
EDITA LEDICHOVA. Labializuotų bendrinės lietuvių kalbos balsių eksperimentinis tyrimas | 16
doi.org/10.35321/bkalba.2020.93.03
The results of the study of short vowels in both natural and artificial words reveal the same trends in changes to qualitative parameters. Therefore, the data on short vowels are discussed together below, highlighting the most general patterns and discussing some exceptional cases.
TABLE 3. Quality features of labialized and isolated short vowels [ɐ], [ɔ], [ʊ], [ɪ], [ɛ] in natural words
Position Sound Formants (Hz) F1 F2 F3 Indices C b T df it I Cₚ[ɐ]Cₚ 670 1290 2420 891 107 11 620 460 II 650 1330 2420 883 107 51 680 400 III 640 1350 2440 878 107 67 710 350 [ɐ] 740 1290 2690 904 106 −78 550 640 I Cₚ[ɔ]Cₚ 600 920 2590 919 109 −264 480 770 II 590 920 2590 916 109 −257 490 760 III 580 920 2580 913 109 −247 500 740 [ɔ] 580 890 2720 918 109 −299 530 910 I Cₚ[ʊ]Cₚ 410 900 2520 860 112 −106 690 710 II 410 870 2550 865 112 −140 720 770 III 410 870 2570 865 112 −144 720 790 [ʊ] 420 810 2620 878 112 −225 770 890 I Cₚ[ɪ]Cₚ 390 1990 2600 775 107 592 1600 700 II 380 2010 2610 771 107 610 1630 740 III 380 1990 2560 772 107 610 1610 670 [ɪ] 400 2090 2790 772 107 593 1690 980 I Cₚ[ɛ]Cₚ 500 1740 2570 818 107 372 1240 310 II 510 1750 2310 823 107 415 1240 450 III 500 1750 2580 818 107 375 1250 330 [ɛ] 650 1790 2760 849 105 252 1140 700
TABLE 4. Quality features of labialized and isolated short vowels [ɛ], [ɔ], [ʊ], [ɪ], [ɛ] in artificial words
Position Sound Formants (Hz) F1 F2 F3 Indices C b T df it I Cₚ[ɛ]Cₚ 640 1310 2550 881 107 22 670 380 II 690 1320 2530 892 107 −1 630 400 III 670 1290 2590 890 107 −18 620 470 [ɛ] 740 1290 2690 904 106 −78 550 640 I Cₚ[ɔ]Cₚ 570 930 2520 909 110 −220 500 660 II 570 910 2580 912 110 −249 520 740 III 560 870 2600 916 110 −284 570 790
BENDRINĖ KALBA 93 (2020) http://journals.lki.lt/bendrinekalba ISSN 2351-7204
Page 17
EDITA LEDICHOVA. Labializuotų bendrinės lietuvių kalbos balsių eksperimentinis tyrimas | 17
Analysis of the pronunciation results for both natural (see Table 3 and Fig. 5) and artificial words (see Table 4 and Fig. 6) showed that the lower vowels [ɛ], [ɛ] exhibit a more pronounced difference in the values of the first formant: the F₁ values in the initial, steady-state, and final articulatory positions of [ɛ], [ɛ] allophones occurring between the labial consonants [p], [b] are at lower frequencies than those of the vowels pronounced in isolation (cf. Cₚ[ɛ]Cₚ F₁ = 670 Hz, 650 Hz, 640 Hz and [ɛ] F₁ = 740 Hz; Cₚ[ɛ]Cₚ F₁ = 500 Hz, 510 Hz and [ɛ] F₁ = 650 Hz (natural-word data); Cₚ[ɛ]Cₚ F₁ = 640 Hz, 690 Hz, 670 Hz and [ɛ] F₁ = 740 Hz; Cₚ[ɛ]Cₚ F₁ = 580 Hz, 610 Hz, 600 Hz and [ɛ] F₁ = 650 Hz (artificial-word data)). It may therefore be assumed that, in articulating the labialized allophones, the degree of tongue raising is greater and they are more closed. The F₁ values of the labialized variants of [ɔ], [ʊ], and [ɪ], in studies of either natural or artificial words, do not reveal any clearer qualitative difference—the 10–20 Hz difference between the compared positions can be considered entirely insignificant.
The values of the second formant vary quite consistently, and the pattern is analogous to that found in the previously discussed results for long vowels in natural words (cf. Tables 1 and 2): the F₂ values of the back-vowel labialized allophones [ɛ], [ɔ], [ʊ] are higher, so they should all be regarded as articulated further forward than the vowels pronounced in isolation, whereas F₂ for the labialized front-vowel allophones [ɪ], [ɛ] is lower, suggesting that, in articulating them, the tongue shifts somewhat farther back in the oral cavity. This feature is characteristic of the results for both natural and artificial words. It should be noted that the data on the formant values of short vowels
BENDRINĖ KALBA 93 (2020)
http://journals.lki.lt/bendrinekalba
ISSN 2351-7204
Page 18
EDITA LEDICHOVA. Labializuotų bendrinės lietuvių kalbos balsių eksperimentinis tyrimas | 18
doi.org/10.35321/bkalba.2020.93.03
reveal a less pronounced qualitative difference than is found for long vowels; however, the variation in F₂ is highly consistent and supports the conclusion that this is not coincidental, but results from the definite influence of the labial consonants [p], [b].
FIG. 5. Spectra of short vowels in natural words: labialized (---) and isolated (....)
The most consistent qualitative results, common to the entire group of allophones under study, are shown by the third-formant data: the F₃ values in the three articulatory positions of the labialized allophones [ɛ], [ɔ], [ʊ], [ɪ], [ɛ] are lower than those of the vowel variants pronounced in isolation.
The trend evident throughout the experimental study, also reported by many foreign phoneticians (Flemming 2002: 22; Thomas 2011: 49; Fletcher, Bucher 2014: 98; Beeley 2015), permits the confident conclusion that the marked differences in the F₃ results are associated with the effect of labialization.
It should be noted that no greater difference in formant values was recorded between the initial, steady-state, and final articulatory positions: a similar effect of labial consonants on the quality results is observed in all three phases of allophone pronunciation. No greater difference in F₁, F₂, and F₃ values was found between the results for natural and artificial words; the effect of labialization on the short-vowel variants studied was very similar in both cases.
FIG. 6. Spectra of short vowels in artificial words: labialized (---) and isolated (....)
Evaluation of the formant values allows a more precise description of the features of short labialized allophones and those pronounced in isolation. The compactness of the low [ɐ], [ɛ] variants in the three articulatory positions studied, occurring between labial consonants, is lower (cf. Cₚ[ɛ]Cₚ C = 891, 883, 878 and [ɛ] C = 904; Cₚ[ɛ]Cₚ C = 818, 823, 818 and [ɛ] C = 849 (natural-word data), and Cₚ[ɛ]Cₚ C = 881, 892, 890 and [ɐ] C = 904; Cₚ[ɛ]Cₚ C = 839, 844 and [ɛ] C = 849 (artificial-word data)). It is necessary
BENDRINĖ KALBA 93 (2020)
http://journals.lki.lt/bendrinekalba
ISSN 2351-7204
Page 19
EDITA LEDICHOVA. Labializuotų bendrinės lietuvių kalbos balsių eksperimentinis tyrimas | 19
doi.org/10.35321/bkalba.2020.93.03
It should be noted that, in the study of long vowels, the compactness values established for the labialized low [æː] were higher than those established for the isolated vowel variant, whereas the previously discussed labialized short vowel [ɛ] variant is characterized by lower compactness. The diffuseness values for the high [ʊ], [ɪ] and mid [ɔ] vowels vary slightly, but certain general tendencies can be discerned. The labialized [ʊ], [ɪ], [ɔ] allophones in natural words, in the initial, steady-state, and final articulation positions, are less diffuse than their isolated counterparts (cf. Cₚ[ʊ]Cₚ df = 690, 720 and [ʊ] df = 770; Cₚ[ɪ]Cₚ df = 1600, 1630, 1610 and [ɪ] df = 1690; Cₚ[ɔ]Cₚ df = 480, 490, 500 and [ɔ] df = 530). Results from the artificial-word study show that among labial consonants
the [ʊ] in all three articulation positions studied (cf. Cₚ[ɔ]Cₚ df = 730, 720, 760 and [ɔ] df = 770), [ɪ] in the steady-state and final positions (cf. Cₚ[ɪ]Cₚ df = 1680, 1670 and [ɪ] df = 1690), and [ɔ] in the initial and steady-state positions (cf. Cₚ[ɔ]Cₚ df = 500, 520 and [ɔ] df = 530) are less diffuse than the isolated pronunciations; in the unaddressed phases of articulation, the labialized variant has a higher degree of diffuseness. The flatness index indicates that the labialized [ɛ] and [ɜ] allophones involve more active lip movement than their isolated counterparts (cf. Cₚ[ɛ]Cₚ b = 107 and [ɛ] b = 106; Cₚ[ɛ]Cₚ b = 108 and [ɛ] b = 107 (natural-word data); Cₚ[ɛ]Cₚ b = 107 and [ɛ] b = 106; Cₚ[ɛ]Cₚ b = 106 and [ɛ] b = 105 (artificial-word data)). It should be noted, however, that no difference in the flatness-index values was recorded between the long labialized and isolated [ɑː]. It is noteworthy that only the artificial-word [ɔ] pronounced between labial consonants has a higher flatness index (cf. Cₚ[ɔ]Cₚ b = 110 and [ɔ] b 109); no change in the flatness-index values was recorded for natural-word [ɔ]. Recall that the second-formant values clearly distinguished two vowel groups: labialized back vowels, whose F₂ values are higher, and front vowels, whose F₂ values are lower. However, analysis of the tonality parameters revealed a different relationship: all back and front
BENDRINĖ KALBA 93 (2020)
http://journals.lki.lt/bendrinekalba
ISSN 2351-7204
Page 20
EDITA LEDICHOVA. Labializuotų bendrinės lietuvių kalbos balsių eksperimentinis tyrimas | 20
doi.org/10.35321/bkalba.2020.93.03
short [ɛ], [ɔ], [ʊ], [ɪ]12 in [ɛ] vowels between labial consonants, studied in three articulation positions, have higher tonality values than the tonality values of these vowels pronounced in isolation, so the labialized allophones should be regarded as having a higher timbre (cf. Cp[ɛ]Cp T = 11, 51, 67 and [ɛ] T = −78; Cp[ɔ]Cp T −264, −257, −247 and [ɔ] T = −299; Cp[ʊ]Cp T = −106, −140, −144 and [ʊ] T = −225; Cp[ɪ]Cp T = 610 and [ɪ] T = 593; Cp[ɛ]Cp T = 372, 415, 375 and [ɛ] T = 252 (natural-word data); Cp[ɔ]Cp T = −22, −1, −18 and [ɛ] T = −78; Cp[ɔ]Cp T = −220, −249, −284 and [ɔ] T = −299; Cp[ʊ]Cp T = −195, −190, −211 and [ʊ] T = −225; Cp[ɪ]Cp T = 616, 632, 614 and [ɪ] T = 593; Cp[ɛ]Cp T = 330, 325, 322 and [ɛ] T = 252 (artificial-word data). It was observed that the most consistently changing quality parameter is tenseness—all labialized short-vowel allophones in both natural and artificial words are pronounced with less tense articulators than their isolated counterparts. The fact that the allophones between labial consonants are less tense was observed in all vowel groups.
12 Only the tonality parameter of the studied labialized [ɪ] in the initial articulation position is somewhat lower (cf. Cp[ɪ]Cp T = 592 and [ɪ] T = 593).
BENDRINĖ KALBA 93 (2020)
http://journals.lki.lt/bendrinekalba
ISSN 2351-7204
Page 21
EDITA LEDICHOVA. Labializuotų bendrinės lietuvių kalbos balsių eksperimentinis tyrimas | 21
doi.org/10.35321/bkalba.2020.93.03
FIG. 7. Relationships between labialized (green circles) and isolated (yellow circles) short vowels in natural words in psychoacoustic space
FIG. 8. Relationships between labialized (green circles) and isolated (yellow circles) short vowels in artificial words in psychoacoustic space
Analysis of vowel relationships (see Figs. 7 and 8) revealed that, in both natural and artificial words, the low short labialized and isolated [ɛ] and [ɛ] variants show the greatest difference in psychoacoustic space—their comparative circles are clearly far apart. The distances unequivocally indicate that the allophones are perceived as sounds that differ substantially in quality. The [ɪ] variants studied should also be perceived as differing in quality, although their comparative circles are somewhat closer together than those of the [ɔ], [ɛ] allophones. The [ʊ] variants may be perceived similarly, but it should be noted that the qualitative difference between the natural-word vowel variants is greater (the circles do not touch) than between the artificial-word vowel variants (the circles touch slightly at their edges). The labialized and isolated [ɔ] variants differ least in psychoacoustic space—the circles overlap by a third or more, meaning that the allophones should be perceived as very similar, but not identical, in quality. All labialized short vowels, except natural-word [ɔ], are shifted closer to the center of psychoacoustic space, while the isolated variants of these vowels occupy a more peripheral position. The same tendency was found in analyzing differences in the positions of long-vowel variants in psychoacoustic space.
BENDRINĖ KALBA 93 (2020) http://journals.lki.lt/bendrinekalba ISSN 2351-7204
Page 22
EDITA LEDICHOVA. Labializuotų bendrinės lietuvių kalbos balsių eksperimentinis tyrimas | 22
doi.org/10.35321/bkalba.2020.93.03
CONCLUSIONS
A detailed study of the quality parameters of labialized long and short allophones in natural and artificial words, and comparison with data for vowels pronounced in isolation, established the following characteristics of vowel variants:
1. The third-formant F3 data differ most markedly and consistently across all vowel groups: for labialized long and short vowel allophones in natural and artificial words, F3 values in all articulation positions studied—that is, initial, steady-state, and final—are at lower frequencies. Thus, the clearest acoustic indications of sound labialization in the spectrum are likely associated above all with lowering of the third formant.
2. Changes in second-formant values are also very marked and systematic, but are related to the vowels’ inherent backness. For labialized long and short back vowels [ɑː], [ɐ], [oː], [ɔ], [uː], [ʊ] in natural and artificial words, F2 values are at higher frequencies than for vowels pronounced in isolation; thus, all of them should be articulated somewhat farther forward. The F2 values of labialized front-vowel [æː], [ɛ], [eː], [iː], [ɪ] variants resonate in a lower-frequency range, so they are likely characterized by more retracted articulation. A somewhat more marked difference in second-formant F2 values was recorded for back vowels, suggesting that these vowels are more strongly affected by the consonants [p], [b]; the direction of change for front vowels is somewhat less pronounced, but the overall tendency is maintained across all vowel groups studied.
3. The effect of labialization on first-formant F1 values is related to the height of the vowels studied: only the long mid vowels [oː] and [eː] and the short low vowels [ɐ] and [ɛ] show marked F1 shifts in the spectrum, but their qualitative properties differ. The F1 values of [oː] and [eː] allophones between labial consonants are higher, so they should be more open than their isolated counterparts, whereas the first formant of labialized [ɐ] and [ɛ] is lower than that of the isolated vowels, so the allophones should be regarded as more close. The other long and short vowels show no marked change in first-formant position in the spectrum.
4. The results of the formant-value assessment can be summarized as follows:
4.1. The compactness of low [ɑː], [ɐ], [ɛ] and mid [eː] allophones in CpVCp position is lower than that of their isolated counterparts, while the compactness-index values for the labialized low [æː] variant are higher in all three articulation positions; high
BENDRINĖ KALBA 93 (2020)
http://journals.lki.lt/bendrinekalba
ISSN 2351-7204
Page 23
EDITA LEDICHOVA. Labializuotų bendrinės lietuvių kalbos balsių eksperimentinis tyrimas | 23
doi.org/10.35321/bkalba.2020.93.03
The high vowels [uː], [o], [iː], [ɪ] and the mid vowels [oː], [ɔ] reveal a somewhat different relationship: in the initial, steady-state, and final articulatory positions, the labialized variants of the vowels under study tend to be less diffuse than the variants pronounced in isolation.
4.2. With regard to tonality, the following groups of vowels under study can be distinguished: long labialized back vowels [ɑː], [oː], [uː], which have a higher degree of tonality and thus a higher pitch, and long labialized front vowels [æː], [eː], [iː], whose degree of tonality in the CpVCp position is lower, and thus whose pitch is lower; the short labialized back and front vowels [ɐ], [ɔ], [ʊ], [ɛ], [ɪ] should be regarded as having a higher pitch.
4.3. Differences in bemollency between labialized and isolated allophones were recorded only in the study of [ɐ], [æː], [ɛ], [iː], and [uː]: in the CpVCp position, the front vowels [æː], [ɛ], [iː] and the back vowel [ɐ] are more bemolized—the lips are likely to be more active in their articulation than in pronouncing isolated variants of these vowels; the back vowel [uː] allophone between labial consonants is characterized by lower bemollency values.
4.4. Tenseness is a general qualitative feature distinguishing labialized from isolated vowel variants—all allophones in the CpVCp position are articulated with less tense speech organs.
5. Labialization affects the initial, or the initial and final, articulatory positions of long allophones somewhat more strongly. In the short-vowel group, no greater difference was recorded among the initial, steady-state, and final positions compared: a similar influence of labial consonants on the quality results was found in all parts of the short allophone.
6. No greater difference in qualitative values was found between natural and artificial words: labialization affects the characteristics of words used in natural speech flow and of artificially constructed words in a very similar way; that is, in all cases, the same direction of change in quality values is maintained.
7. Most of the labialized and isolated vowels studied show clear differences in psychoacoustic space. This is particularly true of the long vowels [oː], [uː], [æː], [eː] and the short vowels [ɐ], [ɛ]: the distances between the circles representing labialized and isolated allophones in psychoacoustic space clearly indicate that these vowel variants should be perceived as qualitatively different sounds. The comparative circles for the long [iː] and short [ɪ] variants overlapped most, so the qualitative difference between their allophones is small.
BENDRINĖ KALBA 93 (2020)
http://journals.lki.lt/bendrinekalba
ISSN 2351-7204
Page 24
EDITA LEDICHOVA. Labializuotų bendrinės lietuvių kalbos balsių eksperimentinis tyrimas | 24
doi.org/10.35321/bkalba.2020.93.03
8. Labialized vowels exhibit a greater or smaller shift toward the center of psychoacoustic space, whereas isolated variants tend to occupy a more peripheral position, that is, they are located at the edges of the quadrants of the coordinate system.
LITERATŪRA
A m b r a z e v i č i u s R y t i s 2010: Differential Acoustical Cues for Palatalized vs Nonpalatalized Prevocalic Sonants in Lithuanian. – Žmogus ir žodis 12(1), 5–10.
A m b r a z e v i č i u s R y t i s 2012: Loci of Palatalized vs Nonpalatalized Prevocalic Plosives in Lithuanian. – Žmogus ir žodis 14(1), 13–18.
A m b r a z e v i č i u s R y t i s, L e s k a u s k a i t ė A s t a 2014: Priebalsių akustinės ypatybės: palatalizacija ir balsingumas, Kaunas: Technologija.
A t k o č a i t y t ė D a i v a 2002: Pietų žemaičių raseiniškių prozodija ir vokalizmas, Vilnius: Lietuvių kalbos instituto leidykla.
B a c e v i č i ū t ė R i m a 2000: Lukšių šnektos izoliuotų balsių tyrimas. – Kalbotyra 48(1)–49(1), 5–18.
B a c e v i č i ū t ė R i m a 2002: Lietuvių kalbos kintamųjų garsų [ie], [uo] eksperimentinė analizė. – Kalbotyra 51(1), 5–17.
B a c e v i č i ū t ė R i m a 2004: Šakių šnektos prozodija ir vokalizmas, Vilnius: Lietuvių kalbos instituto leidykla.
B a k š i e n ė R i m a, Č e p a i t i e n ė A g n ė 2017: Tarptautinės fonetinės abėcėlės taikymo lietuvių tarmių garsams galimybės. – Baltistica 52(1), 105–135. DOI: 10.15388/Baltistica.52.1.2303.
B a l č i ū n i e n ė A s t a, R i n k a u s k i e n ė R e g i n a, M e i l i ū n a i t ė V i o l e t a 2017: Kupiškėnų ir uteniškių paribio šnektų fonetinė kaita XXI a.: Vabalninko, Darsiškių ir Salamiesčio punktų atvejis. – Lituanistica 63, 38–60.
B a l l M a r t i n J., H o w a r d S a r a J., M i l l e r K i r k 2018: Revisions to the extIPA Chart. – Journal of the International Phonetic Association 48(2), 155–164. DOI: 10.1017/S0025100317000147.
B e e l e y H e l e n a 2015: British English [kw], [k], and [w] Distinction in Back Round Vowel Contexts. – ICPhS. Prieiga internete:
BENDRINĖ KALBA 93 (2020)
http://journals.lki.lt/bendrinekalba
ISSN 2351-7204
Page 25
EDITA LEDICHOVA. Labializuotų bendrinės lietuvių kalbos balsių eksperimentinis tyrimas | 25
Bladon Anthony, Fant Gunnar 1978: A Two-formant Model and the Cardinal Vowels. – Speech Transmission Laboratory – Quarterly Progress and Status Report 1, Stockholm: KTH, 1–8.
Boyce Suzanne, Espy-Wilson Carol Y. 1997: Coarticulatory Stability in American English /r/. – Journal of the Acoustical Society of America 101(6), 3741–3753.
Butcher Andrew, Fletcher Janet 2014: Sound Patterns of Australian Languages. – The Languages and Linguistics of Australia. The World of Linguistics 3, De Gruyter Mouton, Berlin, Germany, 89–136.
Carignan Christopher, Ximenes Arwen B., Shaw Jason A. 2017: A Comparison of Acoustic and Articulatory Methods for Analysing Vowel Differences across Dialects: Data from American and Australian English. – Journal of the Acoustical Society of America 142, 363–377.
Cole Jennifer S., Linebaugh Gary, Munson Cheyenne, McMurray Bob 2010: Unmasking the Acoustic Effects of Vowel-to-vowel Coarticulation: A Statistical Modeling Approach. – Journal of Phonetics 38(2), 167–184.
Dereškevičiūtė Sigita 2008: Dėl lietuvių kalbos sklandžiųjų priebalsių kiekybės. – Žmogus ir žodis 10(1), 15–19.
Dereškevičiūtė Sigita, Kazlauskienė Asta 2009: Dusliųjų sprogstamųjų priebalsių spektrinė analizė ir jų sprogimo trukmė. – Garsas ir jo tyrimo aspektai: metodologija ir praktika, Vilnius: Lietuvių kalbos institutas, 98–111.
Flemming Edward 2002: Auditory Representations in Phonology, Routledge, New York.
Girdenis Aleksas 2003: Teoriniai lietuvių fonologijos pagrindai, Vilnius: Mokslo ir enciklopedijų leidybos institutas.
Girdenis Aleksas 2009: Vadinamųjų sutaptinių dvibalsių [ie uo] garsinė ir fonologinė sudėtis. – Baltistica 44(2), 213–242.
Girdenis Aleksas, Kubiliūtė-Kliukienė Regina 1982: Regresyvinis priebalsių palatalizacijos poveikis balsių spektrui šiaurės žemaičių tarmėje. – Kalbotyra 33(1), 30–38.
Girdenis Aleksas, Riaubiškytė Svajonė 1981: Viena šiaurės žemaičių ir prūsų fonetikos paralelė. – Baltistica 17(1), 92–95.
BENDRINĖ KALBA 93 (2020)
http://journals.lki.lt/bendrinekalba
ISSN 2351-7204
Page 26
EDITA LEDICHOVA. Labializuotų bendrinės lietuvių kalbos balsių eksperimentinis tyrimas | 26
doi.org/10.35321/bkalba.2020.93.03
Grigorjevs Juris 2013: Problems Using the Traditional Acoustic Cues for the Phonological Interpretation of Vowels. – Baltistica 48(2), 301–312.
Grigorjevs Juris, Jaroslavienė Jurgita 2014: Dabartinių baltų bendrinių kalbų nosinių sonantų spektrinė analizė. – Žmogus ir žodis. Didaktinė lingvistika 16(1), 35–49.
Herrick Dylan 2007: An Acoustic Description of Central Catalan Vowels Based on Real and Nonsense Word Data. – Catalan Review 21, 231–256.
Hillenbrand Jame M., Clark Michael J., Nearey Terrance M. 2001: Effects of Consonant Environment on Vowel Formant Patterns. – Journal of the Acoustical Society of America 109(2), 748–763. Prieiga internete: https://etalpykla.lituanistikadb.lt/fedora/objects/LT-LDB-0001:J.04~2013~1383055005364/datastreams/DS.002.0.01.ARTIC/content.
Jaroslavienė Jurgita 2010: Rytinių kauniškių Prienų šnektos fonologija: humanitarinių mokslų daktaro disertacija, Vilnius: Vilniaus universitetas.
Jaroslavienė Jurgita 2013: Prienų šnektos nosinių sonantų spektrinė charakteristika. – Acta Linguistica Lithuanica 68,105–125.
Jaroslavienė Jurgita 2014: Spectral Characteristics of the Lithuanian Vowels: Some Preliminary Results of a New Experimental Research. – Linguistica Lettica 22, 68–84.
Jaroslavienė Jurgita 2015: Lietuvių kalbos trumpųjų ir ilgųjų balsių kiekybės ir kokybės etalonai. – Bendrinė kalba 88. Prieiga internete: http://www.bendrinekalba.lt/?88.
Jaroslavienė Jurgita 2017: Different IPA Equivalents for Some Lithuanian and Latvian vowels. – Acta Linguistica Lithuanica 77, 196–218.
Jaroslavienė Jurgita, Grigorjevs Juris, Urbanavičienė Jolita, Indričāne Inese 2019: Baltų kalbų garsynas XXI a. pradžioje: Balsių ir garsų sąveikos instrumentinis tyrimas: mokslo monografija su garsinėmis iliustracijomis, Vilnius: Lietuvių kalbos institutas. Prieiga internete: http://lki.lt/wp-content/uploads/2020/03/Balt%C5%B3_kalb%C5%B3_garsynas_XXI_a_prad%C5%BEioje_1_knyga_2019.pdf.
Kazlauskienė Asta 2018: Bendrinės Lietuvių kalbos fonetikos ir fonologijos pagrindai, Kaunas: Vytauto Didžiojo universitetas. DOI: 10.7220/9786094673290.
Kliukienė Regina 1994: Šiaurės žemaičių priebalsių spektrai ir akustinės priebalsių klasės. – Kalbotyra 43(1), 43–52.
Kliukienė Regina 1995: Šiaurės žemaičių intervokalinių priebalsių trukmė. – Kalbotyra 44(1), 58–68.
Kliukienė Regina 2002: Regresyvinis minkštųjų priebalsių poveikis trumpųjų aukštutinių ir žemutinių balsių spektrui lietuvių bendrinėje kalboje. – Kalbotyra 51, 33–49.
Kliukienė Regina 2011: Šiaurės žemaičių tarmės konsonantizmas: akustinės ir audiacinės ypatybės, Vilnius: Vilniaus universitetas.
Kliukienė Regina, Raudžiūtė Ieva 2013: Minkštųjų priebalsių poveikis lietuvių bendrinės kalbos ilgųjų kirčiuotų balsių spektrui. – Žmogus ir žodis 15(1), 111–119.
Krug Manfred, Schlüter Julia 2013: Research Methods in Language Variation and Change, Cambridge: Cambridge University Press.
Ladefoged Peter 2008: Vowels and Consonants: An Introduction to the Sounds of Languages, Oxford: Blackwell Publishers.
Ladefoged Peter, Johnson Keith 2014: A Course in Phonetics. Cengage Learning, Inc.
Ladefoged Pete, Maddieson Ian 2008: The Sounds of the World's Languages, Cambridge: Cambridge University Press.
Ledichova Edita 2012: Lietuvių bendrinės kalbos nazalizuotųjų balsių [aː], [oː], [uː] trukmės ir kokybės ypatybių eksperimentinis tyrimas. – Žmogus ir žodis 14(1), 79–85.
BENDRINĖ KALBA 93 (2020)
http://journals.lki.lt/bendrinekalba
ISSN 2351-7204
Page 28
EDITA LEDICHOVA. Labializuotų bendrinės lietuvių kalbos balsių eksperimentinis tyrimas | 28
Leskauskaitė Asta 2004: Pietų aukštaičių vokalizmo ir prozodijos bruožai, Vilnius: Lietuvių kalbos instituto leidykla.
LKE 2008: Lietuvių kalbos enciklopedija, red. Vytautas Ambrazas, Vilnius: Mokslo ir enciklopedijų leidykla.
LKTCh 2004: Lietuvių kalbos tarmių chrestomatija. Sud. Bacevičiūtė Rima, Ivanauskienė Audra, Leskauskaitė Asta, Trumpa Edmundas, Vilnius: Lietuvių kalbos instituto leidykla.
Marchal Alain, Hardcastle William J. 1993: Instrumentation and Database for the Cross-language Study of Coarticulation. – Language and Speech 36, 137–153.
Maxwell Olga, Baker Brett, Bundgaard-Nielsen Rikke, Fletcher Janet 2015: A Comparison of the Acoustics of Nonsense and Real Word Stimuli: Coronal Stops in Bengali. – International Congress of Phonetic Sciences (ICPhS). Prieiga internete: https://www.internationalphoneticassociation.org/icphs-proceedings/ICPhS2015/Papers/ICPhS0900.pdf.
Mikalauskaitė Elžbieta 1975: Lietuvių kalbos fonetikos darbai, Vilnius: Mokslas.
Murinienė Lina 2007: Rytinių šiaurės žemaičių fonologija: vokalizmas ir prozodija, Vilnius: Lietuvių kalbos institutas.
Ohde Ralph N., Sharf Donald J. 1975: Coarticulatory Effects of Voiced Stops on the Reduction of Acoustic Vowel Targets. – Journal of the Acoustical Society of America 58(4), 923–927.
Paillereau Nikola 2016: Do Isolated Vowels Represent Vowel Targets in French? An Acoustic Study on Coarticulation. – Congrès Mondial de Linguistique Française. DOI: 10.1051/SHS 27 shsconf/20162709003.
Pakerys Antanas 1982: Lietuvių bendrinės kalbos prozodija, Vilnius: Mokslas.
Pakerys Antanas 2003: Lietuvių bendrinės kalbos fonetika, Vilnius: Enciklopedija.
Renwick Margaret E. L. 2014: The Phonetics and Phonology of Contrast: The Case of the Romanian Vowel System, Berlin: de Gruyter Mouton.
BENDRINĖ KALBA 93 (2020)
http://journals.lki.lt/bendrinekalba
ISSN 2351-7204
Page 29
EDITA LEDICHOVA. Labializuotų bendrinės lietuvių kalbos balsių eksperimentinis tyrimas | 29
doi.org/10.35321/bkalba.2020.93.03
Salys Antanas 1992: Lietuvių kalbos tarmės. – Raštai 4, Roma: Lietuvių Katalikų Mokslo Akademija.
Steinlen Anja K. 2005: The Influence of Consonants on Native and Non-native Vowel Production: A Cross-linguistic Study, Tübingen: Gunter Narr.
Stevens Kenneth N., House Arthur S. 1976: Perturbation of Vowel Articulations by Consonantal Context: An Acoustical Study. – Journal of Speech Language and Hearing Research 6(2), 111–128.
Strange Winifred, Verbrugge Robert R., Shankweiler Donald 1976: Consonantal Environment Specifies Vowel Identity. – Journal of the Acoustical Society of America 60, 213–224.
Strange Winifred, Weber Andrea, Levy Erika S., Shafiro Valeriy, Hisagi Miwako, Nishi Kanae 2007: Acoustic Variability within and across German, French, and American English Vowels: Phonetic Context Effects. – Journal of the Acoustical Society of America 122(2), 111–129.
Sylak-Glassman John 2014: The Effects of Post-Velar Consonants on Vowels in Ditidaht. – Papers from the 49th Annual International Conference on Salish and Neighbouring Languages, 17–38.
Thomas Erik R. 2011: Sociophonetics. An Introduction, Basingstoke: Palgrave Macmillan.
Traunmüller Hartmut 1988: Analytical Expressions for the Tonotopic Sensory Scale. – PERILUS 8, Stockholm: University of Stockholm, 93–102.
Trumpa Edmundas 2005: Pelesos šnektos balsinės fonemos ir jų variantai (daktaro disertacija), Vilnius: Vytauto Didžiojo universitetas.
Urbanavičienė Jolita, Indričāne Inese 2016a: Dabartinių baltų kalbų pučiamieji priebalsiai ir afrikatos: akustinių požymių lyginamoji analizė. – Acta Linguistica Lithuanica 75, 46–77.
Urbanavičienė Jolita, Indričāne Inese 2016b: Lietuvių ir latvių kalbų uždarumos priebalsiai: sprogimo fazės trukmė ir FFT spektrai. – Žmogus ir žodis 18(1), 46–79. DOI: http://dx.doi.org/10.15823/zz.2016.3.
BENDRINĖ KALBA 93 (2020)
http://journals.lki.lt/bendrinekalba
ISSN 2351-7204
Page 30
EDITA LEDICHOVA. Labializuotų bendrinės lietuvių kalbos balsių eksperimentinis tyrimas | 30
doi.org/10.35321/bkalba.2020.93.03
Urbanavičienė Jolita, Indričāne Inese, Jaroslavienė Jurgita, Grigorjevs Juris 2019: Baltų kalbų garsynas XXI a. pradžioje: Priebalsių instrumentinis tyrimas (mokslo monografija su garsinėmis iliustracijomis), Vilnius: Lietuvių kalbos institutas. Prieiga internete: http://lki.lt/wp-content/uploads/2020/03/Balt%C5%B3_kalb%C5%B3_garsynas_XXI_a_prad%C5%BEioje_2_knyga_2019.pdf.
Vaitkevičiūtė Valerija 1960: Lietuvių kalbos balsių ir dvibalsių ilgumas, arba kiekybė. – Lietuvių kalbotyros klausimai 3, 207–217.
Vaitkevičiūtė Valerija 1965: Fonetika. – Lietuvių kalbos gramatika 1, Vilnius: Mintis, 72–90.
VKĮPĮ 2011: Lietuvos Respublikos valstybinės kalbos įstatymo pakeitimo įstatymas. Projektas XP-1320(3). Prieiga internete: https://e-seimas.lrs.lt/portal/legalAct/lt/TAP/TAIS.402981?jfwid=-wbkgqm837.
Wayland Ratree 2019: Phonetics – A Practical Introduction, Cambridge: Cambridge University Press.
Whalen Douglas H. 1990: Coarticulation is Largely Planned. – Haskins Laboratory Status Report on Speech Research 101/102, 149–176.
Zellou Georgia, Pycha Anne 2018: The Gradient Influence of Temporal Extent of Coarticulation on Vowel and Speaker Perception. – Journal of Laboratory Phonology 9(1), 1–24.
Zinkevičius Zigmas 1994: Lietuvių dialektologija, Vilnius: Mokslo ir enciklopedijų leidykla.
Zinkevičius Zigmas 2006: Lietuvių tarmių kilmė, Vilnius: Lietuvių kalbos institutas.
Пиотровский Раймонд Г. 1960: Еще раз о дифференциальных признаках фонемы. – Вопросы языкознания 6, 24–38.
Gauta 2020 04 16
Priimta 2020 06 19
BENDRINĖ KALBA 93 (2020)
http://journals.lki.lt/bendrinekalba
ISSN 2351-7204
Page 31
EDITA LEDICHOVA. Labializuotų bendrinės lietuvių kalbos balsių eksperimentinis tyrimas | 31
doi.org/10.35321/bkalba.2020.93.03
THE EXPERIMENTAL RESEARCH OF QUALITATIVE CHARACTERISTICS OF LABIALISED VOWELS IN STANDARD LITHUANIAN
Summary
In natural language, sounds usually do not exist in isolation, they are included in larger units – syllables, words, phrases, sentences, and their specific realisation depends mostly on the influence of adjacent sounds. Language would be very easy to study if it was realized through separate allophones. We would then see clearly, where one sound ends and where another begins, but the sounds merged in the speech stream become not only difficult to distinguish, but also acquire the sound of another phonetic unit or completely new features. Traditionally, there are two variants of vowels usually recognized in phonetics: positional and combinational. Positional variants depend on accent. This feature has been extensively investigated in the Lithuanian language. Combinational variants of vowels is another phenomenon, which occurs due to the influence of adjacent sounds.
Although the term labialization is mentioned quite often in the works of Lithuanian linguists, so far little attention has been paid to detailed and systematic studies of Lithuanian standard language vowels influenced by labialization, usually only the most general features of labialization are described. Therefore, the aim of this paper is to describe labialized vowels and compare the main qualitative characteristics of labialized and isolated (zero context) vowels found by experimental research.
All long [ɑː, oː, uː, æː, eː, iː] and short [ɛ, ɔ, ʊ, ɪ, ɛ] vowels of the Lithuanian standard language were selected for the study. They were produced in zero context and CpVCp sequences by 2 native male speakers with flawless articulation. To achieve the aim, two types of experiments were carried out using words that are actually used in speech and artificially constructed ones. The segmentation and analysis of the sounds was performed using free license sound processing and analysis software PRAAT. The allophones analysed in this study were divided into 3 parts and the meaning of formants in the beginning, the middle part and the end of the segment were calculated, i. e., 1/3 of the sound part. This was done to find out which part of the investigated allophone receives the strongest coarticulation influence. For the measurements of the isolative pronounced vowel, the average of the values forming the whole sound was taken, because this sound is not affected by the adjacent sounds, the dynamics of its formant values is quite even, the spectrum varies little. In order
BENDRINĖ KALBA 93 (2020)
http://journals.lki.lt/bendrinekalba
ISSN 2351-7204
Page 32
EDITA LEDICHOVA. Labializuotų bendrinės lietuvių kalbos balsių eksperimentinis tyrimas | 32
doi.org/10.35321/bkalba.2020.93.03
to describe better the qualitative ratios of the analysed allophones, the following acoustic parameters of the beginning, the middle part and the end of the segment articulation positions were calculated according to the first, second and third formant values (Hz): compactness (C), tonality (T), flatness (b), diffusivity (df) and tension (it). The relationships of the graphically compared vowel variants are represented in the psychoacoustic F2'/F1 space, in which the researched allophones are marked by circles of 1 bark size (z) in diameter, and the distances between them indicate the degree of difference in perceived quality: the farther the circles are from each other, the more qualitatively those sounds differ.
After a detailed study of the qualitative characteristics of labialized vowels of the Lithuanian standard language and their comparison with the data of isolated vowels, the following general tendencies were identified. Labialized allophones are most evidently and stably characterised by the data of the third formant F3 – in all articulation positions, i. e., in the beginning, the middle part, and the end of the segment, they are characterised by lower-frequency F3 values. Thus, it is considered that the most obvious signs of sound labialization in the spectrum are primarily related to the degradation of the third formant. The change in the values of the second formant F2 of labialized vowels is also greatly pronounced and tendentious, but its changes are related to the pronunciation of vowels. The labialized allophones of back vowels [ɑː], [ɐ], [oː], [ɔ], [uː], [ʊ] of all the studied groups more often have higher frequency F2 values, so they should be slightly more front articulation than pronounced in isolation. The corresponding variants of the front vowels [æː], [ɛ], [eː], [iː], [ɪ] have a resonant F2 in the lower frequency range, so they are likely to have a slightly more back articulation. The effect of labialization on the values of the first formant F1 is not systematic. The evaluation of labialized allophone formant values has shown that consistent trends were found by analysing the values of compactness/diffusion, tonality, flatness and stress indices, but systematic differences were not recorded in all groups of vowels.
The analysis of the relationships of vowel variants in psychoacoustic space has shown that in many cases, the allophones pronounced in different positions are perceived as sounds of unequal quality.
Attention was drawn to the fact that no greater qualitative difference was observed between the words that are actually used in speech and artificially constructed ones – labialization has a very similar effect on their acoustic characteristics.
BENDRINĖ KALBA 93 (2020) http://journals.lki.lt/bendrinekalba ISSN 2351-7204
Page 33
EDITA LEDICHOVA. Labializuotų bendrinės lietuvių kalbos balsių eksperimentinis tyrimas | 33
doi.org/10.35321/bkalba.2020.93.03
The strength of coarticulation influence on the quality characteristics of labialized vowels depends primarily on the length of the vowel. Long allophones are more likely to experience a stronger influence of adjacent consonants in the beginning or in the beginning and the end of the segment articulation positions. Short vowel variants experience a stronger effect of full-sound coarticulation because no more pronounced differences were found between the measured articulation positions – all allophonic phases receive a similar effect of adjacent consonants.
