Cues for Consonant Voicing: Its Assimilation in Word Juncture

172 acta Linguistica Lithuanica LXXI keyWords: voiced consonant, voiceless consonant, assimilation, word juncture. esMiniai ŽodŽiai: skardusis priebalsis, duslusis priebalsis, asimiliacija, žodžių sandūra. rytis aMbrazeViČius kaunas university of technology fields of research: musical and speech acoustics, ethnomusicology, psychology of music. asta Leskauskaitė institute of the Lithuanian Language fields of research: synchronic and diachronic phonetics, phonology, morphology, geolinguistics, dialectology of the baltic languages. cues for consonant Voicing: its assiMiLation in Word juncture Priebalsių balsingumo požymiai. asimiliacija žodžių sandūroje annotation the paper consists of two parts. the first one considers the general acoustic cues for voicing of obstruents (i.e. cases of the consonants in word juncture are not detailed). acoustic peculiarities of the lenis/fortis dichotomy are mentioned. relatively simple visual cues for voicing (acoustical expressions of a component of phonation) identifiable from oscillograms and spectrograms are discussed. Problems of the techniques of such identification are pointed out. the quantitative indexes for voicing of fricatives and stops are suggested and applied for analysis of the consonant voicing in standard Lithuanian. the second part is focused on the close assimilation of the consonants in word juncture taking place in the southern aukštaitians. drawing on the above mentioned techniques examples of spontaneous oral speech are analyzed. anotacija straipsnį sudaro dvi dalys. Pirmojoje nagrinėjami bendrieji akustiniai triukšminių priebalsių (t. y. neišskiriant žodžių sandūros priebalsių) balsingumo požymiai. aptariami lenis / fortis cues for consonant Voicing: its assimilation in Word juncture 173straipsniai / articles dichotomijos akustiniai savitumai, sąlygiškai paprasti balsingumo požymiai (fonacijos komponento akustinės išraiškos), identifikuojami iš oscilogramų ir spektrogramų, pažymimos tokio identifikavimo metodikos problemos. toliau siūlomi kiekybiniai pučiamųjų ir sprogstamųjų priebalsių balsingumo rodikliai. naudojant juos atliekama lietuvių bendrinės kalbos priebalsių balsingumo analizė. antrojoje straipsnio dalyje dėmesys sutelkiamas į pietų aukštaičių šnektose vykstančią žodžių sandūros dalinę priebalsių asimiliaciją. remiantis anksčiau aptarta metodika, praktiškai analizuojami sakytinės spontaninės kalbos pavyzdžiai. In standard Lithuanian and dialects, consonants adjust according to the activities of the vocal folds, palatalism, and the place of articulation. assimilation is established taking into account several characteristics: the distance between the adjusting sounds (close, distant), direction of adjustment (regressive, progressive), and degree (full, partial). this phonetic phenomenon can be complex. in other words, consonants can be assimilated both according to one of the features and by several features simultaneously. additionally, degemination is often inseparable from this phenomenon (Lkg i 93–94; Pakerys 2003: 178–189). usually linguists have no doubts about the consonant assimilation in words according to the activities of the vocal folds. the adjustment of voiced and voiceless consonants in word junctures is more complicated; therefore it is important to ascertain whether and in what cases this type of coarticulation is systemic for the position mentioned, as well as the degree of the change. in standard language, when a dependent word is next to an independent one, the laws for word junctures are the same as those for the combination inside a word; however, according to the mentioned feature, consonant assimilation in junctures of two independent words (like degemination) is facultative (Pakerys 2003: 180). the regularities of these phenomena in the Lithuanian dialects can be revealed from transcribed dialectal texts and sound recordings, as well as from notes of various dialectologists, but no attempt has been made to verify the regularities experimentally. the main goal of this article is to discuss partial close regressive consonant assimilation in word junctures according to the action of vocal folds, drawing on the results of experimental research. the research goals are: 1. to discuss general experimental methods of research of the consonant voicing in word juncture and its identification issues; 2. to establish some regularities of the close regressive consonant assimilation characteristic of word junctures in the southern aukš tai tians. the types and cases of the assimilation of the consonants of the southern aukštai tians have been determined drawing on the sound recordings from the dialects archive of the research centre of geolinguistics at the institute of the Lithuanian 174 rytis aMbrazeViČius, asta Leskauskaitė acta Linguistica Lithuanica LXXI Language as well as on the published texts (Markevičienė 1999; Leskauskaitė 2004; MŠt). five recordings (the total duration of them exceeds ten hours) of spontaneous spoken language of southern aukštaitians1 were employed for sound analysis; Praat software was applied for this purpose. for the analysis of voicing in standard Lithuanian, utterances recorded from a single Lithuanian male speaker were applied2; the obtained values of acoustic parameters were employed as certain reference points in the succeeding study on dialectal speech. nuances in terMinoLogy the dichotomy of terms voiced and voiceless (unvoiced) points to the presence or absence of phonation as the main cue for discrimination. this is a well-known paradigm. however, to be precise, before the proceeding to further examination of consonant voicing, we should be aware of certain problems with such a simple description. We should remember that the full-scale vocal component is not always obligatory for the discrimination discussed. Moreover, some languages (for instance, some finno-ugric languages) do not employ phonation at all in their ‘voiced’ consonants. the main cue for the discrimination in the analogous contrasting consonant pairs is considered to be the strength of articulatory efforts which is less for the ‘voiced’ consonants and bigger for the ‘voiceless’ consonants. james M. Pickett (1999: 125) notes the conventional attitude that “the constrictions of unvoiced consonants are articulated with more force or tension than for the voiced consonants”. therefore it is common to use a more general description of the dichotomy of voiced/voiceless consonants and to consider the dichotomy of lenis/fortis instead. in the subsequent discussion, we will mean mostly the ‘real’ voicing/not voicing typical for the Lithuanian language. however, sometimes we will refer to the issues just mentioned. 1 ČŠ, Vainežeris, Lazdijai district, sound record LM222806; Vb, Miškiniai, Lazdijai district, sound record Lk061101, Lk061102; Pd, dainava, Varėna district, sound record LM230505; jP, Margionys, Varėna district, sound record Lk161501, LM222906; VM, kapčiamiestis, sound record nr.693. Viktoras. Miciulevičius.WaV. all recordings are kept at the dialects archive of the research centre of geolinguistics at the institute of the Lithuanian Language. 2 an actor juozas jaruševičius; the sound recordings provided in the cd as supplement for the book by Pakerys and Pupkis (2004). for visualization of the general cues for voicing (isolated sounds and nonsense VcV-patterns; see figs. 1–4 and 7), additional recordings were prepared by one of the authors, male speaker rytis ambrazevičius. cues for consonant Voicing: its assimilation in Word juncture 175straipsniai / articles siMPLe VisuaL cues the component of phonation appears in acoustical graphs in several ways. in the oscillograms of frication phases (cases of fricatives and affricates), it is identified from the partial periodicity of a waveform (fig. 2, top; compare to fig. 1, top).3 figure 1. intervocal voiceless sibilant š with small portions of adjacent vowels (aša): oscillogram, narrowand wide-band spectrograms4 figure 2. intervocal voiced sibilant ž with small portions of adjacent vowels (aža): oscillogram, narrowand wide-band spectrograms similarly, in the oscillograms of gap phases (cases of stops and affricates), the periodicity of a waveform, as opposed to the complete silence or more realistic weak noise, means the additional component of phonation (fig. 4, top; compare to fig. 3, top)5. this component can be also identified from spectrograms. in the narrow-band spectrograms, it appears as the corresponding structure of harmonics (figs. 2, 4; mids). in the wide-band spectrograms, pulsing is the marker of phonation (figs. 2, 4; bottoms). thus, briefly, “the presence or absence of voicing energy is the dominant cue for the perception of the voicing contrast in fricatives” (kent, read 2002: 161); and this presence or absence is clearly depicted in certain acoustical graphs. therefore the 3 for all acoustical measurements and the graphs, Praat software was employed. Ms excel provided basis for the statistical generalizations and the graphic representation of the results. 4 in figs. 1–4, isolated VcV-patterns are presented. both vowels are accented. 5 affricates will not be considered further in this chapter. in prospect, a mixture of the methods used for fricatives and stops could be applied in the study of voicing of affricates. 176 rytis aMbrazeViČius, asta Leskauskaitė acta Linguistica Lithuanica LXXI identification of voicing seems to be quite simple task based on the simple observations of the acoustical graphs. however, here we come back to the issues mentioned in the beginning of the paper. for instance, in certain cases, a fricative is perceived as voiced even if it contains a segment of devoicing (fig. 5). then the question arises where the boundary between the voiced and voiceless fricatives is, i.e. what duration of the devoiced segment (or other characteristics) corresponds to the boundary between the voiced and voiceless counterparts. Moreover, this boundary is expected to be dependent on a certain language and even on some specific nuances of speech flow. figure 3. intervocal voiceless stop p with small portions of adjacent vowels (apa): oscillogram, narrowand wide-band spectrograms figure 4. intervocal voiced stop b with small portions of adjacent vowels (aba): oscillogram, narrowand wide-band spectrograms figure 5. spectrogram of the word lá.užu ‘fire (sing. instr.)’. the segment of devoicing is marked cues for consonant Voicing: its assimilation in Word juncture 177straipsniai / articles a good example is Vot (Voice onset time) used frequently as the differential cue for the english stops. the boundary between the voiced and voiceless english stops is said to be at approximately 30 ms (fig. 6).6 on the contrary, our preliminary observations showed that, for the Lithuanian voiced stops, Vot is always negative. if the stop is preceded by a pause, an epenthetic vowel is formed. if the stop is intervocal, voicing is not interrupted, i.e. it fills the entire gap phase. in some cases, seemingly, some very short devoiced segments are formed somewhere around the noise bursts, yet these are hardly identifiable from the acoustic graphs. all this shows that certain quantitative indexes for the differentiation of voiced and voiceless consonants are highly desirable. the characteristic values of the indexes (and especially the boundary values) could be obtained from the acoustical measurements of the real utterances of clear instances of the voiced/voiceless counterparts, as well as from the perceptual evaluations of the continua of synthesized sounds. QuantitatiVe cues: fricatiVes several differential indexes and algorithms for the evaluation of voicing of fricatives were proposed by different authors. one should mention, for instance, HNR (harmonic-to-noise ratio; boersma 1993; hamann, boersma, Ćavar 2010) and intensity and temporal parameters of the first harmonic (Pirello, blumstein, kurowski 1997; stevens et al 1992). based on generally appropriate assumptions, these methods have certain shortcomings which make their applicability for the study more or less problematic. on the one hand, HNR shows illogical boundary values (http://uk.groups.yahoo.com/group/praat-users/message/2533). on the other hand, the evaluations of dynamics of fundamental including measurements of durations, extraction of fundamental, measurements of its SPL at several time 6 incidentally, refer to the quote of kent and read: while stating on “presence or absence of voicing energy” as “the dominant cue for the perception of the voicing contrast” they mean fricatives but not stops. figure 6. discrimination of voiced and voiceless stops by Vot (kent, read 2002: 151) 178 rytis aMbrazeViČius, asta Leskauskaitė acta Linguistica Lithuanica LXXI points, and the successive comparisons of SPLs take considerable time resources and therefore are probably suitable in the cases of single instances, but hardly practical in the cases of large samples. a set of the algorithms employs quite distinct separation of the voice/noise frequency bands (childers 2000; balbonas, daunys 2005; etc.). on the one hand, structure of low harmonics up to 1–2 khz (with the most intense band up to approximately 0.5 khz) is characteristic of the spectra of voiced fricatives. on the other hand, the noise of frication is characteristic of significantly higher frequencies (fig. 7). for instance, for english voiceless sibilants, noise is significantly more intense for > 2.5 khz. the most intense spectral peaks are at about 3 khz (š) and about 4 khz or higher (s) (Pickett 1999: 119, 139; kent, read 2002: 163, 168, etc.). based on this difference, a simple LHR method was proposed (Low-high-ratio index; i.e. ratio of intensities of low and high frequency bands; ambrazevičius 2013). LHR was defined by the following expression: LHR (db) = 10lg (IL/IH) = –10lg ; here IL, IH, L, and LL stand for, correspondingly, the intensities of low (0–0.5 khz) and high (>0.5 khz) frequency bands, the total SPL, and SPL of the low band. figure 7. spectrum of a prolonged isolated sibilant ž cues for consonant Voicing: its assimilation in Word juncture 179straipsniai / articles figure 8. LHR for the Lithuanian sibilants7 the method was applied for the evaluation of voicing of 201 sibilant in standard Lithuanian8; the utterances of male speaker presented in the recordings of Pakerys and Pupkis 2004 were examined. it was estimated that the boundary between the voiced and voiceless sibilants corresponds roughly to LHR = –10 db (fig. 8). QuantitatiVe cues: stoPs the logic of evaluation of low frequency bands could be also applied in the case of stops. We suggest the relative SPL of the low-pass filtered intervocal gap (fig. 9) as the differential index: ΔLL (db) = LLC – LLV ; 7 here and further: medians – black diamonds, interquartiles – white diamonds, extremes – dashes. Values less than –30 db not depicted. the results for z are somewhat preliminary because of quite a small sample size. 8 therefore the results differ slightly from those presented in ambrazevičius 2013 which included 188 sibilants. 180 rytis aMbrazeViČius, asta Leskauskaitė acta Linguistica Lithuanica LXXI here LLC and LLV are, correspondingly, SPLs of low (0–0.5 khz) frequency bands of a consonant gap and the adjacent vowel or vowels. SPLs of the preceding vowel, succeeding vowel, or their average, can be used depending on whether the consonant is postvocal, prevocal, or intervocal (see also the discussion below). figure 9. utterance preba… (from prebalsis): sPL track and narrow-band spectrogram of low-pass (0–0.5 khz) filtered recording figure 10. ΔLL for the Lithuanian stops cues for consonant Voicing: its assimilation in Word juncture 187straipsniai / articles figure 12. utterance lkis bùvɔ. ‘had remained’. top: oscillogram, bottom: sPL track and wide-band spectrogram of low-pass (0–0.5 khz) filtered recording figure 13. utterance vi.šini.kαs bùvɔ. ‘the chief was’. top: oscillogram, bottom: sPL track and wide-band spectrogram of low-pass (0–0.5 khz) filtered recording both generalized LHR values and the values for separate occurences of sibilants s, š in word junctures of spontaneous language show that these sibilants most often become voiced z, ž, therefore regressive assimilation is obvious. however, this phenomenon is not intense in some cases, therefore the sibilants s, š are identified as only partially voiced. it is not always easy to estimate the degree of voicing by ear. 188 rytis aMbrazeViČius, asta Leskauskaitė acta Linguistica Lithuanica LXXI the consonants s, ž used before voiced consonants in word junctures can remain almost completely voiceless. for example, in the phrase lkis bùvɔ, the sibilant s has a naturally weak shade of z – its LHR equals –2 db (fig. 12)11. this sound did not undergo regressive assimilation since, because of the dispersive influence of the context, it was pronounced as rather short (about 60 ms), while reverberation was stronger. if a facultative vowel is inserted in word juncture, the sibilant consonant before a further voiced stop does not undergo regressive assimilation. for example, LHR value for the consonant s in figure 13 is –10.4 db. it corresponds approximately to the median of LHR for all cases of the voiceless s (cf. fig. 11). therefore the fricative is really voiceless in this case. Types T←Ts and T←Td. the tendency of interaction of the voiceless stops k, p, t and the following voiced stops g, b, d in word junctures is similar to that of fricatives (fig. 14). in this case, the degree of voicing is estimated basing on the ΔLL index discussed earlier. fig. 14 shows that the ranges of the ΔLL values of consonants k, p, t influenced by the voiced consonants g, b, d and those which did not experience partial close regressive assimilation intersect. therefore it is impossible to indicate the boundary of separation of voicing and unvoicing between them. it can be surmised that, as in case of fricatives, assimilation of these consonants in word junctures can be of different degree, with similar reasons like those characteristic of fricatives discussed above. since the ΔLL index did not show a very obvious difference between assimilated and nonassimilated stops, additionally the low frequency energy values of these sounds were compared (i.e. not taking into account vowel context; fig. 15). a clear separation is seen only for the occurrences of labial consonant p preceding voiced and voiceless stops – the corresponding ranges of the LL values do not intersect. obviously, in the position discussed, the voicing/unvoicing of the consonants k and t was greatly influenced by the context and other factors. it was allready mentioned that, because of scanty data hardly sufficient for reliable statistical evaluation, other types of assimilation in word junctures have not been discussed. however, the discussed phenomena seems to be analogical in the southern aukštaitians’ subdialects as well as the Lithuanian language. therefore possibly similar general tendencies could be identified. it should be noted that the sound recordings of spontaneous southern aukštaitians’ speech contained fewer cases of the types t←ts and t←td compared to s←ts: S←td, therefore the samples of their usage are not large. this circumstance can influence the credibility of the results obtained. 11 the Lhr of the fricative consonant s of the word junctures in pictures 12 and 13 have not been included in the general results of measurements algorythms calculation. cues for consonant Voicing: its assimilation in Word juncture 189straipsniai / articles figure 14. ΔLL for the southern aukštaitian stops in word junctures12 figure 15. LL; sound pressure level for the low-pass (0–0.5 khz) filtered southern aukštaitians’ stops in word junctures 12 here the type t←ts seemingly lacks the third quartile as it is blended with median. it may linked to the scanty occurrences of the position researched. 190 rytis aMbrazeViČius, asta Leskauskaitė acta Linguistica Lithuanica LXXI concLusions it is well known that “the presence or absence of voicing energy is the dominant cue for the perception of the voicing contrast in consonants” (rehashed from kent & read 2002: 161). however, one encounters not simply two dichotomic cases “a consonant with voicing energy” and “a consonant without voicing energy”, but rather entire continuum of fractions of voicing energy in the whole spectra of the consonants considered. the boundaries between fortis and lenis categories are individual for different languages. Moreover, the perception of voicing is complicated as includes not only spectral but also temporal cues. all this makes the procedure of the objectivized evaluation of voicing not trivial. the paper introduces two relatively simple techniques of identification of voicing: the indices LHR (case of fricatives) and ΔLL (case of stops), both based on the evaluation of energy of low frequencies in the consonants (or their temporal segments). referring to the abovementioned complications of the objectivized evaluation of voicing, these indices work generally quite well as differential cues for voicing. nevertheless, because of the complications, a certain caution should be noted again when employing these techniques: the boundaries between the voiced and voiceless counterparts are individual for the discussed cases. additionally, the results could be distorted by some extrinsic technical causes, such as intense background noise and room reverberation. this should be taken into account both when choosing the sound recordings for analysis and when applying the techniques of acoustical measurements. the indicated techniques of analysis could be employed in the study of phenomena of consonant assimilation in the Lithuanian dialects too. the research carried out into the examples of the spontaneous southern aukštaitians’ speech (types s←ts : S←td ir t←ts : t←td) essentially gives proof for the statements presented here. it is obvious that, from the point of view of acoustics, consonant assimilation in word juncture is quite complex phenomenon and its study requires precision. references am brazev ičius rytis 2013: akustiniai pučiamųjų priebalsių balsingumo požymiai. – Žmogus ir žodis 15(1), 7–14. balbonas dainius, daunys gintautas 2005: fonemų klasifikavimas panaudojant garso ir vaizdo informaciją. – Elektronika ir elektrotechnika 5(61), 74–77. bo er sma Paul 1993: accurate short-term analysis of the fundamental frequency and the harmonics-to-noise ratio of a sampled sound. – IFA Proceedings 17, 97–110. childers g. donald 2000: Speech Processing and Synthesis Toolboxes. new york: john Wiley & sons. Dat – Dūkšto apylinkių tekstai, sudarė Vytautas kardelis. Vilnius: Lietuvių kalbos institutas, 2010. cues for consonant Voicing: its assimilation in Word juncture 191straipsniai / articles Duat – Dusetų apylinkių tekstai, sudarė regina rinkauskienė. Vilnius: Lietuvių kalbos institutas, 2013. gat – Griškabūdžio apylinkių tekstai, sudarė rima bacevičiūtė, Vilija sakalauskienė. Vilnius: Lietuvių kalbos institutas, 2008. hamann silke, boersma Paul, Ćavar Małgorzata 2010: Language-specific differences in the weighting of perceptual cues for labiodentals. – http://user.phil-fak.uni-duesseldorf.de/~hamann/hamannboersmacavar2010.pdf. Jat – Jurbarko apylinkių tekstai, sudarė Vilija sakalauskienė. Vilnius: Lietuvių kalbos institutas, 2013. kent d. raymond, read charles 2002: The Acoustic Analysis of Speech. albany: thomson Learning. kruopienė irena, kačiuškienė genovaitė 2012: keletas pastabų dėl išorės sandžio šiaurės panevėžiškių tarmėje. – Baltistica 47(2), 343–348. Lkg i – Lietuvių kalbos gramatika 1, vyr. red. kazys ulvydas. Vilnius: Mintis, 1965. LzŠt – Lazūnų šnektos tekstai, sudarė jurgita jaroslavienė, nijolė tuomienė. Vilnius: Lietuvių kalbos institutas, 2013. Markevičienė Žaneta 1999: Aukštaičių tarmių tekstai 1, Vilnius: Vilniaus universiteto leidykla. Mat – Mielagėnų apylinkių tekstai, sudarė Vytautas kardelis. Vilnius: Lietuvių kalbos institutas, 2006. MŠt – Marcinkonių šnektos tekstai, sudarė asta Leskauskaitė. Vilnius: Lietuvių kalbos institutas, 2009. Leskauskaitė asta 2006: Kučiūnų šnektos tekstai. Vilnius: Lietuvių kalbos instituto leidykla. Pakerys antanas 2003: Lietuvių bendrinės kalbos fonetika. Vilnius: enciklopedija. Pakerys antanas, Pupkis aldonas 2004: Lietuvių kalbos bendrinė tartis. Kompaktinės plokštelės tekstai. Vilnius: gimtasis žodis. Pickett M. james 1999: The Acoustics of Speech Communication. Fundamentals, Speech Perception Theory, and Technology. needham heights: allyn & bacon. Pirello karen, blumstein e. sheila, kurowski kathleen 1997: the characteristics of voicing in syllable-initial fricatives in american english. – Journal of the Acoustical Society of America 101(6), 3754–3765. stevens n. kenneth, blumstein e. sheila, glicksman Laura, burton Marth a, kurowski kat h leen 1992: acoustic and perceptual characteristics of voicing in fricatives and fricative clusters. – Journal of the Acoustical Society of America 91(5), 2979–3000. ŠŠt – Šakių šnektos tekstai, parengė rima bacevičiūtė. Vilnius: Lietuvių kalbos institutas, 2006. 192 rytis aMbrazeViČius, asta Leskauskaitė acta Linguistica Lithuanica LXXI Priebalsių balsingumo požymiai. jų asimiliacija žodžių sandūroje santrauka Paprastai ir tradiciškai sutariama, kad skardžiuosius ir dusliuosius priebalsius skiria fonacijos (balso klosčių veiklos), t. y. balso, balsingojo komponento buvimas ar nebuvimas. tai ne visai tikslu. kai kurių kalbų (pavyzdžiui, kai kurių ugrofinų kalbų) priebalsiai, sudarantys analogiškas kontrastines poras, skiriasi ne arba ne tiek balsingumo požymiais, o tam tikrais (mūsų požiūriu – iš lietuvių kalbos fonetikos pozicijų) antriniais požymiais. atkreipus dėmesį į skirtingas artikuliacines pastangas tariant skardžiuosius ir dusliuosius priebalsius (ar jų dichotominius analogus), pirmuosius kartais linkstama vadinti neįtemptaisiais (lenis), antruosius – įtemptaisiais (fortis). straipsnyje aptariami pavyzdžiai, kur diferencijuojant fortis–lenis poras balsingumo požymis yra bent kiek nors reikšmingas. Pagrindinis akustinis skardžiuosius priebalsius nuo dusliųjų skiriantis požymis – fonaciją rodantis komponentas: kvaziperiodiniai virpesiai oscilogramose, papildomos balso harmonikos spektrogramose. tačiau net ir nutrūkus fonacijai pučiamasis priebalsis išlieka skardžiuoju. fonacijos požiūriu labai skiriasi, pavyzdžiui, lietuvių ir anglų kalbų sprogstamieji priebalsiai. tokiais pavyzdžiais iliustruojami kalbiniai fortis–lenis skirtumai ir argumentuojami kiekybinių metodų privalumai. siūlomi kiekybiniai balsingumo rodikliai. skardžiųjų pučiamųjų priebalsių spektre išryškėja žemadažnės harmonikos. tuo pasinaudota apibrėžiant pučiamųjų priebalsių LHR (Low-HighRatio) rodiklį, lygų priebalsio žemadažnės (iki 500 hz) ir aukštadažnės (per 500 hz) energijos santykiui. kiek kitaip, bet pasiremiant panašiu reiškiniu (žemadažnės energijos vyravimas skardžiųjų priebalsių spragos fazėje), konstruojamas sprogstamųjų priebalsių ΔLL rodiklis – santykinis (priebalsio artimame ar tolimame kontekste) žemadažnio komponento garso lygis. Minėta metodika panaudota ir analizuojant pietų aukštaičių dusliųjų ir skardžiųjų priebalsių sąveiką žodžių sandūroje. s←ts : s←td ir t←ts : t←td tipų analizė parodė, kad dalinė artimoji regresyvinė asimiliacija aptariamojoje pozicijoje nėra vienoda, o kai kuriais atvejais ir apskritai nevyksta. be to, norint atlikti patikimą spontaniškos kalbos priebalsių asimiliacijos analizę svarbu atkreipti dėmesį į garso įrašo techninę kokybę, įrašymo patalpos reverberaciją ir kitus veiksnius. handed in on 4 september 2014. rytis aMbrazeViČius Kaunas University of Technology Adomo Mickevičiaus g. 37, LT-44244 Kaunas, Lithuania [email protected] asta Leskauskaitė Institute of the Lithuanian Language Petro Vileišio g. 5, LT-10308 Vilnius, Lithuania [email protected]