1
The phonetics and phonology of pretonic prominence in Aŭciuki
Belarusian
Lena Borise
Abstract
In the Aŭciuki dialect of Belarusian, pretonic vowels in certain contexts are pronounced with greater
prominence than stressed ones. This phenomenon, pretonic prominence, has been analysed as a
retraction of stress or H tone/ F0 peak associated with stress to the pretonic syllable. Novel instrumental
data shows that the pretonic vowel in pretonic prominence contexts receives an increase in duration,
but its F0 and intensity are parallel to those found on pretonic vowels in other contexts, which refutes
an analysis of pretonic prominence as an F0 peak retraction. Vowel neutralization facts and intonational
pitch accent distribution show that neither is pretonic prominence a retraction of stress. Building on
existing analyses of dissimilative vowel neutralization in Slavic, I propose that pretonic prominence is
a phenomenon similar in spirit though independent in nature from vowel neutralization, and results from
redistribution of stress-induced acoustic prominence over two syllables, pretonic and stressed.
Keywords: stress, pretonic prominence, vowel duration, F0, intensity, Belarusian, East Slavic
***
1. Introduction
The object of investigation in this paper is an unusual stress-related property found in the Aŭciuki
dialect of Belarusian, a south-eastern variety spoken in the Kalinkavichy district of the Homel province,
predominantly in the villages of Malyja Aŭciuki and Vialikija Aŭciuki. In the Aŭciuki dialect, the
pretonic vowel may receive acoustic prominence comparable to or exceeding that of the stressed vowel,
depending on the height of the pretonic and stressed vowels (Kryvicki 1959; Vajtovich 1968; Bethin
2005; 2006a; 2006b, a.o.). In this paper, I will refer to this phenomenon as pretonic prominence. The
theoretical interest of pretonic prominence lies in the fact that V1 in such environments is described as
being equally or more prominent than V2, which raises the question about the location and acoustic
nature of stress.
The conditioning for pretonic prominence in Aŭciuki is the following: it is found in those cases
where the stressed vowel (V2) is high or mid-high (i/ɨ, u, e, o), and the pretonic vowel (V1) is mid-low
2
or low (ɛ, ɔ, a), as shown in (1).1,2 In the remainder of the paper, building on Bethin (2005), I will refer
to the environment for pretonic prominence as CaCi, where ‘C’ stands for consonant(s), ‘a’ for the non-
high pretonic vowel, and ‘i’ for the non-low stressed vowel.
(1) a. sestru ‘sister.ACC’ [sjɛːˈstru]
b. dvorɨ ‘courtyards’ [dvɔːˈrɨ]
c. nasi ‘carry.IMP’ [naːˈsi]
In contrast with CaCi environments, pretonic prominence does not apply to contexts in which both
V1 and V2 are non-high, as shown in (2), or those in which both vowels are non-low, as illustrated in
(3). These environments will be referred to as CaCa and CiCi, respectively.
(2) a. sestra ‘sister.NOM’ [sjɛˈstra]
b. nazad ‘backwards’ [naˈzat]
c. balota ‘bog’ [baˈlɔtǝ]
(3) a. krušɨna ‘buckthorn’ [kruˈšɨna]
b. idu ‘come.1SG’ [iˈdu]
c. žɨvu ‘live.1SG’ [ʒɨˈvu]
Recent borrowings into the dialect are subject to pretonic prominence too, as shown in (4), which
means that it is fully productive in the speech of the dialect users (note, however, that the Aŭciuki
dialect itself is critically endangered; more on this in Section 2).
(4) z brɨhadziram ‘with crew chief’ [z brɨɣaːˈdziram]
scienakardzija ‘stenocardia’ [scjenakaːˈrdzija]
izasarbid ‘isosorbide’ [izasaːˈrbit]
Some other East Slavic dialects have been reported to exhibit phenomena similar to pretonic
prominence of the Aŭciuki dialect. In the traditional literature, based on fieldworkers’ reports, they are
usually described as having a special ‘musical’ or tonal contour on the pretonic syllable, or even a shift
of stress one syllable to the left. In Russia, this has been noted for the dialects of Mosalsk (Broch 1916),
Pereslavl-Zalessky (Avanesov 1927: 68), and dialects of the Vladimir-Volga basin more generally
(Durnovo 1914: 373; Zakharova 1970: 357; Vysotskij 1973: 35; Almukhamedova & Kulsharipova
1980; Kasatkin 1989: 35), the dialect of the Gnilovka village in the Tver’ dialect area (Nikolaev 2009),
and some north-Russian dialects (Kolesov 1964; Burova & Kasatkin 1977). The so-called Old Moscow
pronunciation, in which the pretonic syllable receives a lengthening and a particular tonal contour, is
1 The Aŭciuki dialect has been described as having a seven vowel system, /i~ɨ, u, e, o, ɛ, ɔ, a/, including a contrast mid-
high/tense vowels /e, o/ and mid-low/lax vowels /ɛ, ɔ/ (Kryvicki 1959; Vojtovich 1972a). The contrast in question is not found
in standard East Slavic languages, but obtains in a number of dialects (Bernshtein 2005: 280; Pozharitskaja 2005: 35). In all
varieties, categorical phonological distinctions between the two mid-level vowels barely exist: there are no minimal pairs for
[ɛ] vs. [e], while [ɔ] and [o] occur in the identical environment only in a handful of grammatical forms, such as different case
forms in adjectival paradigms. Historically, the two tense vowels have different sources: [o] is found under stress in (some)
closed syllables ([ɔ] also can occur in the same environment), while [e] is found under stress as a reflex of Proto-Slavic *ĕ –
regardless of syllable type (Kryvicki 1959:98-99).
Acoustically, there is some preliminary evidence that [e] in the Aŭciuki dialect is realised closer to the front of the mouth
than [ɛ], based on mean F2 values, while no such contrast – or any other formant-based contrast – has been detected for [ɔ]
and [o] (Borise 2018). The matter requires further investigation. Examples with /ɛ, ɔ/ in the pretonic position, which are subject
to pretonic prominence, are included into the dataset used for the current study. At the same time, if a stressed mid-high vowel
causes pretonic prominence, it is taken to be /e/ or /o/. 2 In the examples throughout the paper, the stressed vowel is boldfaced, and, if affected by pretonic prominence, the pretonic
vowel is underscored. Unless otherwise noted, the cited examples were collected during the author’s fieldwork in Malyja
Aŭciuki in 2015.
3
often mentioned as a subtype of the same phenomenon (Vysotskij 1973; Kasatkina 2005). Finally,
similar claims have been made for certain Chernihiv dialects in Ukraine, known as the Upper Snov
dialects, which are adjacent to the Aŭciuki dialectal area (Sinjavskyj 1934; Zhylko 1953; Bila 1970). In
the Gnilovka, Aŭciuki and Upper Snov dialects pretonic prominence is conditioned by vowel height; in
the other reported cases, acoustic prominence on the pretonic syllable is found across the board. Most
of the available accounts are based on impressionistic observations; the available instrumental results
are summarised in Section 3.
Outside of Slavic languages, phenomena similar to the Aŭciuki pretonic prominence have been
described for Tiberian Hebrew (McCarthy 1981), Canadian French (Walker 1984), and Córdoba
Argentinian Spanish (Lang-Rigal 2014 and references therein). In particular, in Tiberian Hebrew, /a, e/
and, in some instances, /o/ receive a degree of lengthening if found in an open pretonic syllable.
McCarthy (1981) shows that the domain of application of pretonic lengthening is the prosodic word,
but does not provide a theoretical account of pretonic lengthening as such. In Canadian French,
according to Walker (1984: 46), intrinsically long vowels (/ø o ɑ/ and nasalised ones) and those
lengthened by a particular class of coda consonants (/v z ʒ r/) can retain their length if found in the root
and followed by a (stressed) affix. At the same time, the intrinsically long vowels as well as /e/ may
optionally be lengthened if they are found in an open pretonic syllable, regardless of the morphological
make-up of a word. No formal account of pretonic lengthening in Canadian French is offered in Walker
(1984). Finally, pretonic lengthening in Córdoba Spanish has been described in similar terms to the
Aŭciuki facts. In particular, Malmberg (1950: 219) hypothesises that it is accompanied by a particular
tonal contour and may constitute a shift of stress, and Vidal de Battini (1964) labels it a ‘musical accent’.
The distribution of pretonic lengthening in Córdoba Spanish, however, differs from that found in the
languages discussed so far. In particular, pretonic lengthening in Córdoba Spanish is described as a
phrase-level – as opposed to word-level – phenomenon, in that it targets the vowel that precedes the
last accent of an intonational phrase (Fontanella de Weinberg 1971; Yorio 1973). The analysis provided
in Lang-Rigal (2014) targets the perceptual distinctiveness and sociolinguistic import of pretonic
lengthening in Córdoba Spanish, and does not extend into the formal domain either.
As this short overview of the existing research into pretonic prominence shows, the Aŭciuki
phenomenon is by no means a single occurrence, within East Slavic or beyond. Moreover, a certain
profile of pretonic prominence emerges from the available descriptions: it seems to preferentially target
non-high vowels in open pretonic syllables. At the same time, few formal analyses of pretonic
prominence are available, and even those are, for Aŭciuki, are not supported by the new instrumental
data presented in this paper. In particular, as Section 4 shows, there is no evidence for treating pretonic
prominence as a retraction of stress or an F0 peak, the processes that have been hypothesised to give
rise to pretonic prominence.
While providing a unified analysis of the known instances of pretonic prominence falls outside the
scope of the current paper, I propose that the patterns found in the Slavic languages can be readily
accounted for in a unified way. In particular, the analysis offered in this paper capitalises on a
connection between pretonic prominence and patterns of dissimilative vowel neutralization in East
Slavic, which has been noted in the literature before (e.g. Vojtovich 1972b; Čekmonas 1987). Vowel
neutralization, which targets unstressed syllables in many East Slavic varieties, typically differentiates
between the immediately pretonic syllable and other unstressed syllables. Specifically, the pretonic
syllable is subject to a distinct pattern of neutralization, which, in the dissimilative varieties, may be
conditioned by the identity (height) of the stressed vowel. An insightful analysis of vowel neutralization
in pretonic syllables in Slavic is offered by Crosswhite (1999; 2000), who proposes that the two vowels,
stressed and pretonic, form an iambic foot within which stress-related prominence may be distributed.
4
In this paper, the same insight is adopted for the analysis of pretonic prominence in the Aŭciuki dialect
(the objections to such an analysis, raised in Bethin 2005, are also addressed). The current analysis,
therefore, creates a formal connection between the two phenomena, pretonic prominence and
dissimilative vowel neutralization, which have long been suspected to be diachronically related; more
on this in Section 5.
Beyond accounting for a typologically unusual stress-related phenomenon, pretonic prominence,
the results discussed in this paper have wider significance, both descriptively and theoretically. With
respect to the former, they help documenting a critically endangered variety of Belarusian. On the latter
front, they contribute to our understanding of the acoustic nature of stress and its association with
segmental material.
The paper is structured the following way. Section 2 lays out the basic properties of the Aŭciuki
dialect. Section 3 provides an overview of existing work on the Aŭciuki phenomenon. Section 4 presents
the current instrumental study, based on data collected in 2014 and 2015, and addresses the
methodology (4.1) and results (4.2), as well as providing a discussion (4.3). Section 5 offers an
Optimality Theory (OT) style analysis of the phenomenon, building on the connection between pretonic
prominence and dissimilative vowel neutralization. Before the current OT analysis is laid out in Section
5.3, some of the preliminary observations are laid out in Section 5.1, and existing formal approaches to
Slavic vowel neutralization are discussed in Section 5.2. A summary and conclusion are provided in
Section 6.
2. Belarusian and the Aŭciuki dialect
Belarusian is an East Slavic language spoken in Belarus and some neighbouring territories and a
national language of the Republic of Belarus. Standard Belarusian, as well as its dialects, has free stress
and no tonal contrast; neither is vowel length phonemic in the language. Stress in Belarusian may be
described as ‘phonologically active’, according to the terminology in Hyman (2012), in that other
morphophonological phenomena interact with stress placement: there are numerous minimal pairs
based on stress, including in accentual paradigms, and consistent patterns of vowel neutralization in
unstressed syllables.
According to grammars, stress in standard Belarusian is primarily signaled by greater duration of
the stressed vowel as compared to the neighboring ones, as well as lack of neutralization (Czekman &
Smułkowa 1988; Sussex & Cubberly 2006: 179; cf. Zlatoustova 1954; Bondarko 1966; 1977; Jones &
Ward 1969: 206; Matusevich 1976; Kuznetsov, Ott & Ventsov 1987; Gouskova 2010 for Russian). The
phonetic reality of this has not been verified; no contemporary instrumental investigations of stress in
standard Belarusian or its dialects have been reported. Intonational pitch accents, as expected, align
with stressed syllables; depending on the type of the pitch accent used (e.g., H*, L*), the stressed
syllable can have higher or lower F0 than vowels in the neighboring syllables.
Belarusian and its dialects exhibit different degrees of vowel neutralisation. In the standard
language, mid-low vowels /ɛ/ and /ɔ/ are neutralised to /a/ unless stressed (Czekman & Smułkowa 1988;
Mayo 1993: 891); in this, standard Belarusian is different from standard Russian, in which neutralisation
to /a/ occurs only in the immediately pretonic syllable, while non-high vowels in other unstressed
syllables (further pretonic and post-tonic) neutralise to /ǝ/. In Belarusian dialects, the degree of vowel
neutralisation decreases from north-east to south-west of the country (Vojtovich 1971). The Aŭciuki
dialect lies on the boundary between neutralising and non-neutralising dialects, and has irregular vowel
neutralisation (Zhylko 1953; Kryvicki 1959; Vojtovich 1972b). The pattern of vowel neutralization,
however, bears similarity to that of pretonic prominence. In particular, it has been noted that the
5
sensitivity to the height of V1 and V2 that is the Aŭciuki pretonic prominence is built on is similar in
spirit to the dissimilative pattern of vowel neutralization, according to which the quality of V1 ise
dependent on the quality of V2 (Belaja 1974; Vojtovich 1972a). The analysis offered in this paper builds
on this hypothesised connection. The pattern of vowel neutralisation in the Aŭciuki dialect and its
relevance for pretonic prominence and the current analysis are discussed in Section 5.
Unfortunately, as is the case with many traditional Belarusian dialects, since mid-twentieth century
the viability of the Aŭciuki dialect has been challenged. Shortage of job opportunities in the Aŭciuki
area has forced the villagers to leave the area in search of work, and led to greater dialect mixing, as
well as more consistent exposure to Russian. The fact that dialectal pronunciation is often subject to
stigma did not help the viability of the dialect, especially in speakers working outside of the immediate
Aŭciuki area. The languages of schooling in Belarus are Russian and, to a lesser extent, standard
Belarusian, which does not provide recognition for the dialectal features. Therefore, middle-aged and
younger speakers in the Aŭciuki area today often do not retain the full dialectal phonological system.
Nevertheless, pretonic prominence is robust in the speech of older informants (over 60 y.o.). The recent
data presented in this paper comes from speakers who exhibit pronounced pretonic prominence. Still,
it should also be noted that even in older speakers pretonic prominence may not be entirely consistent:
sometimes it does not surface in the conditioning environment, and occasionally applies in unexpected
contexts (cf. also Vojtovich 1972b). This may be due to ongoing changes in the dialect, or even gradual
decline of pretonic prominence as a feature.
Given the age of the speakers, the Aŭciuki dialect is a critically endangered one. One of the purposes
of the current study, therefore, is to document a language variety that will cease to exist in a decade or
two. Most of the older recordings of the Aŭciuki dialect (and others) that have been made by Soviet
linguists have not been digitalised and many have been obliterated (Veranika Kurcova, p.c.). The hope
is that the current work will not only be of use for theoretical phonologists but can also serve as a
valuable resource for dialect documentation.
3. Previous studies
The Aŭciuki pretonic prominence has been discussed in the literature before, mainly in the work of
Soviet linguists. The works addressing it are predominantly descriptive, and the theoretical conclusions
made in the earlier analyses of pretonic prominence area are very cautious - this is true of Kurylo (1924;
1928), Kryvicki (1959) and Belaja (1974). They note that a naive listener perceives pretonic prominence
as a shift of stress to V1, and Belaja (1974) acknowledges that the results of the instrumental
investigation may be interpreted as stress retraction, but such a conclusion, nevertheless, is never made
explicitly.
To illustrate, Kurylo (1924: 14–15) notes that an unaccustomed ear perceives pretonic prominence
as a realisation of stress on V1. Kryvicki (1959: 102) writes about pretonic prominence contexts: “It
might seem at first that the pretonic syllable bears stress in such cases… and in the speech of the younger
generation, who don’t have the feature any more, it often does”. Similarly, Belaja (1974:29) mentions
that in disyllabic words with pretonic prominence “auditory analysis registers stress on the first vowel
- that is, shift of stress to the pretonic syllable”. Still, none of these investigations explicitly argue that
stress retraction is taking place in pretonic prominence contexts in the Aŭciuki dialect. The
Dialectological Atlas of Belarusian Language (1963) conveys a similar sense of uncertainty. Some
villages in the Aŭciuki dialectal area are reported to realise stress on the etymologically pretonic syllable
(V1): Navinki (Kalinkavichy rajon), Vialiki Bor (Xojniki rajon), Svedskaje (Rechyca rajon),
Spiaryzhzha (Brahin rajon). However, about the Spiaryzhzha, Navinki and Svedskaje data it also says
6
that a second investigation disconfirms earlier results and suggests that the stress in the varieties spoken
in the villages is realised on the etymologically stressed syllable (V2).
There are no available instrumental studies of the Aŭciuki phenomenon, but vowel quality in the
Upper Snov dialects, which are spoken in today’s Ukraine in an area adjacent to that of Aŭciuki and
have also exhibit pretonic prominence, have been investigated by Belaja (1974). There are some
phonological differences between the two dialects – in particular, the Upper Snov dialects are described
as possessing mid-high diphthongs /i͡ e, u͡o/, which correspond to the Aŭciuki mid-high vowels /e, o/,
and stronger and more consistent vowel neutralisation. Still, the pretonic prominence facts in the two
dialectal areas are very similar. Belaja (1974) reports on a number of experiments aimed at investigating
the acoustic properties of the vowels /a/ and /ɛ/ when they surface as V1s, both in unmarked (CaCa) and
pretonic prominence (CaCi) contexts. After reporting on a pilot study (50 words extracted from
spontaneous speech by 7 female participants), which confirmed the hypothesis that Upper Snov dialects
have pretonic prominence conditioned by vowel height, the paper presents results of a controlled
production study: measurements of duration, F0 and intensity of the pretonic and stressed vowels in
two contexts. In the first one, experimental items were embedded into declarative, interrogative and
exclamative contexts, and in the second one, they were uttered in isolation; the second context also
included non-words.3 Four female speakers (middle-aged and older) took part in the experiment; the
total number of experimental items and the position of items in the clauses is not reported. Table 1
contains a representative sample of the results for pretonic prominence items as found in different clause
types in the first experiment. Values expected from the point of view of stress realisation in East Slavic
(i.e., the stressed vowel, V2, having greater duration, intensity and F0 values than V1) are underscored;
the non-underscored values, therefore, reflect the acoustic realisation of pretonic prominence.4
Table 1. A sample of Belaja’s (1974) results for pretonic prominence in narratives
Item Clause
type
Duration,
ms
F0,
Hz
Intensity,
mm
V1 V2 V1 V2 V1 V2
vazɨ
‘carts’
declarative 240 140 167 166 13.5 0
interrogative 280 90 228 - 16.1 0
exclamative 300 340 179 196 14.3 0
vazu
‘cart.LOC’
declarative 240 80 149 159 13.1 0
interrogative 210 90 175 35 5.1 0
exclamative 300 206 220 234 23 1.9
vjazi͡ ecj
‘carry.3SG’
declarative 200 170 145 175 11.3 1.2
interrogative 220 180 191 262 18.7 2.1
exclamative 200 176 214 157 17.2 0.6
The results in Table 1 would be quite striking if Upper Snov dialects did not have pretonic
prominence. Specifically, the values for duration are consistently higher on the pretonic vowels, while
there is more variability with respect to the F0 values, and stressed vowels often have higher F0 values
than pretonic ones. The intensity results present a clear picture too, if an unexpected one: the stressed
vowels are barely registered with respect to their intensity properties, in opposition to the pretonic ones.
3 No statistical analysis is offered in Belaja (1974); accordingly, the data in this section is represented as properties
of individual experimental items, like in the source. 4 Intensity in Belaja’s (1974) study was measured as a range on an oscillogram waveform; therefore, the measuring
unit is mm.
7
Table 2 offers a sample of the results pertaining to the experimental items uttered in isolation. Here,
the picture is similar to that presented in Table 1 with respect to duration and intensity – both are greater
on V1 than V2 – and is more consistent with respect to F0: the stressed vowels have greater F0 values
than the pretonic ones one. Note that, because the majority of stimuli are disyllabic with final stress,
this may also be an artefact of list intonation, and not just the realisation of word stress.
Table 2. A sample of Belaja’s (1974) results for the word-list task
Item
Duration,
ms
F0,
Hz
Intensity,
mm
V1 V2 V1 V2 V1 V2
vazɨ ‘carts’ 300 310 157 179 14,4 0
vazu ‘cart.LOC’ 340 300 186 219 19,2 1,5
vjazi͡ ecj ‘carry.3SG’ 260 240 217 227 17,3 4,5
vazjura (non-word) 230 230 179 190 13,4 1,7
Based on these results, Belaja (1974) concludes that in the Upper Snov dialects the etymologically
stressed syllable in pretonic prominence contexts has grown weaker and lost its culminating position.
Belaja further hypothesises that this weakening of the stressed syllable leads to the compensatory
prominence that the pretonic syllable acquires. However, no explanation is offered for why the
weakening of stressed vowels took place, nor why it is limited to high and mid-high stressed vowels.
Belaja also does not claim definitively that pretonic prominence is an instantiation of a stress retraction,
though she remarks that such an interpretation would be consistent with the experimental results. Belaja
(1974) further notes that pretonic /a/ is most prominent when followed by stressed high vowels /i, ɨ u,
i͡ e/, and pretonic /ɛ/ - when followed by stressed /i, u/. She confirms that neither /a/ nor /ɛ/ is subject to
pretonic prominence when followed by stressed non-high vowels /a, ɔ, ɛ, u͡o/.
The results of the current investigation, presented in Section 4, for the most part, accord with
Belaja’s. Both studies show that the acoustic prominence of a low V1 before a non-low V2 can be greater
than that of V2, and that vowel duration is the key acoustic cue for pretonic prominence. Both studies
also align with respect to the fact that pretonic prominence contexts are not consistently marked by
greater F0 values on the pretonic vowel.5,6
5 The intensity results diverge, however. In Belaja’s (1974) data, the difference between intensity values on V1
and V2 is nothing short of extreme, with the etymologically stressed vowel often being as low in intensity as post-
tonic unstressed vowels (0 mm), while in the current study, the pretonic and stressed vowels have similar intensity
values. There is no immediate explanation for this fact. 6 The instrumental results for the Upper Snov dialects also align with those available for some Russian dialects
that have pretonic prominence. In particular, Vysotskij (1973) reports on data from one speaker of a Vladimir-
Volga basin dialect, who was recorded pronouncing a set of 100-150 experimental items. The tested items were
trisyllables of CaCaCaC shape, stressed on the final syllable. Based on the collected data, Vysotskij (1973)
concludes that the duration of V1s in the dialect is equal to or greater than that of V2s, which is especially striking
when compared with data from standard Russian, where V2s have uniformly greater duration than the V1s. Similar
results were obtained by Almuxamedova and Kulsharipova’s (1980). As part of their study, at least three speakers
from each of a number of the Vladimir-Volga basin area dialects were recorded pronouncing a list of test items.
Most of the test items consisted of CV syllables, with V1 and V2 of same or different heights; the total number of
recorded items was over 20,000. According to the study results, pretonic prominence in the Vladimir-Volga basin
area dialects relies on increased duration and intensity of the pretonic vowel. Finally, Nikolaev (2009) presents
data from a single informant from Gnilovka village. The lexical items subject to pretonic prominence were
extracted from a longer narrative. In the items analysed, if V1 and V2 were of equal height or fit the CaCi pattern,
8
To recap, the stress shift account was one of the prominent ideas in earlier discussions of the
Aŭciuki phenomenon but it was not persuasively argued for or against. While Belaja’s (1974)
experimental results seem to favour it, there are several reasons for why the stress shift account cannot
be correct. First, it is evident from the hesitation with which it had been proposed that there is an
intuitive understanding that pretonic prominence and stress constitute two distinct, though interacting,
phonological phenomena in the Aŭciuki dialect. It appears to be so for the speakers too: specifically,
older speakers with robust pretonic prominence in their speech, when prompted, assign stress to its
etymological position. Another argument against a stress retraction analysis comes from vowel
neutralisation facts and is presented in Section 5.
A theoretical account of pretonic prominence is provided in Bethin (2006a; 2006b), with Bethin
(2005) also offering an Optimality Theory (OT) analysis. In these series of articles, Bethin provides an
overview of the existing literature on pretonic prominence-like phenomena in East Slavic dialects
(which is especially valuable for an English-speaking reader, given that most of the relevant sources are
published in Russian, Belarusian or Ukrainian), and suggests that pretonic prominence, at the heart of
it, is a tonal phenomenon. Specifically, Bethin (2006a; 2006b) proposes that Upper Snov dialects carry
a falling tone on the stressed syllable and a low tone on unstressed syllables, which means that the V1V2
sequence carries an LHL tonal contour. Next, citing Belaja’s (1974) data, Bethin proposes that pretonic
prominence results from a shift of the F0 peak/ H, associated with stress, from V2 to V1. She motivates
this shift by taking intrinsic phonetic length of the stressed vowel to be the crucial conditioning factor
for pretonic prominence. In particular, she argues that a non-low V2, as in pretonic prominence contexts,
does not have enough duration to accommodate the HL tonal contour, given that high vowels are
phonetically shorter than lower vowels (Lehiste 1970; Westbury & Keating 1980). Because a non-low
V2 cannot accommodate it, the leftmost of the two tones, H, in pretonic prominence contexts is shifted
to V1. As a result, the LHL contour over two syllables is still there, but H is now realised on V1. The
shift of H to V1, in turn, is what causes V1 to receive greater duration.
The analysis successfully accounts for pretonic prominence as a phonological phenomenon, but the
instrumental data, both that from Belaja (1974) and from the current investigation, presented in Section
4 below, poses some serious challenges for this account. This is because, according to the instrumental
results, V1 in pretonic prominence contexts is not characterised by an F0 peak, contrary to the
fieldworkers’ reports and the assumption that Bethin’s analysis is built on. In fact, as shown especially
by the data in the next section, it is exactly the opposite: V2 consistently retains the peak in F0 in pretonic
prominence contexts and does not differ from pretonic vowels in CaCa contexts in this respect. The
instrumental results reported in the current paper, therefore, provide empirical support for the reality of
pretonic prominence in the Aŭciuki dialect, while also attesting that the acoustic nature of pretonic
prominence is somewhat different from what was expected, based on the available impressionistic
descriptions. In this sense, the current paper also acts as a follow-up on Bethin (2006a), also published
in Phonology, in which the author called for an instrumental investigation of the Aŭciuki phenomenon.
Two other problematic aspects of Bethin’s analysis should be noted. First, there is no independent
evidence for the hypothesis that the stressed syllable is invariably associated with a falling F0 contour
HL. Tone is not contrastive in East Slavic, nor is there any evidence that stress is specified for tone in
the intensity of V1s was equal to or greater than that of V2. V1s in Nikolaev’s sample also have higher F0 values
than V2s, but V2s are reported to be of greater (or equal) duration as compared to V1s. Note that the waveforms
and spectrograms of individual lexical items that are presented in Nikolaev (2009) do not include pitch tracks,
which makes the claim about F0 properties of V1 and V2 impossible to assess.
9
the Upper Snov dialects;7 instead, the F0 properties of the stressed syllable ar determined by phrasal
prosody and, depending on the contextually-determined intonational pitch accent selected (e.g. H* or
L*), V1 may have higher or lower F0 values than V2. Second, an analysis of pretonic prominence as
driven by a shift of H from V1 to V2 wrongly predicts that pretonic prominence should apply in CiCi
contexts. Specifically, since it takes phonetic shortness of a non-low V2 to be the driving force behind
the shift of H to the V1, the shift should apply regardless of the height/phonetic length of V1. Yet this is
not the case: pretonic prominence does not apply in CiCi contexts.
Another, preliminary account of the Aŭciuki pretonic prominence is offered in Borise (2017). A
small-scale pilot study discussed in Borise (2017) shows that pretonic prominence contexts of the
Aŭciuki-type are characterised by greater intensity and duration of V1, but not consistently greater F0
values. The author tentatively suggests that pretonic prominence relies on a shift of stress-induced high
intensity values from V2 to V1. However, this result turned out to be on the wrong track as well: the
current study, based on a larger dataset and more rigorous methodology, shows that the key acoustic
cue for pretonic prominence is duration of V1 alone, as opposed to F0 or intensity.
The instrumental study reported in the next section, therefore, was motivated by the need to establish
the acoustic nature of pretonic prominence, which remained largely stipulative in the existing studies.
As the current results show, prosodic prominence is cued by increased duration of V1 (and, to some
extent, a decreased duration of V2), but not F0 or intensity.
4. The current study
4.1 Methodology The acoustic data analysed here was collected in 2014 and 2015 in the villages of Malyja Aŭciuki and
Vialikija Aŭciuki. The recordings were made using Panasonic RR-US570 and Zoom H4n voice
recorders in a quiet setting in the speakers’ homes. Due to the ongoing decrease in the number of dialect
speakers, the phenomenon of pretonic prominence is robust only in older speakers (over 60-70 y.o.); in
the speech of the next generation (ca. 40-50 y.o.), pretonic prominence is sporadic, and in speakers
younger than 40 y.o. PP is virtually non-existent. With this in mind, data from five informants, all
female (F1-F5, natives of Malyja Aŭciuki or Vialikija Aŭciuki, age range 60-92 y.o., mean age = 79.6)
was selected for analysis. Collecting controlled data from the population of this age turned out to be
difficult; therefore, the data analysed here was extracted from non-final phrases of declarative clauses
with all-new intonation, which were part of a larger, spontaneously produced narrative.
From the recorded narratives, a total of 496 words were selected for analysis, which included
pretonic prominence environments (CaCi) and both types of the environment in which pretonic
prominence does not apply (CaCa and CiCi). In order to control for the fact that surrounding voiceless
consonants may increase F0 values on the vowel, only words with voiced syllable onsets were
considered; in complex onsets of pretonic syllables (C1C2V1), voiceless C1s were also allowed. The
breakdown of the dataset by speaker and condition is provided in Table 3. Given the critically
endangered status of the dialect, data from speakers that contributed only a small number of items, such
as F1, was taken into consideration. Note also that there happen to be fewer examples of CiCi shape
than the other two in the lexicon of the dialect, which led to there being fewer CiCi examples in the
dataset.
Table 3. The total number of words in the dataset broken down by speaker and condition.
7 Note that Bethin (2006a; 2006b) cites Nikolaev (1995; 1996; 2000; 2002; 2003) and Dybo and Nikolaev (1998)
as providing evidence for lexical tone on monosyllables in some Ukrainian-Belarusian border dialects in
Chernihiv province. This analysis, however, did not receive wide support among slavists.
10
Environment/speaker F1 F2 F3 F4 F5 Total
CaCi 3 12 60 59 93 227
CaCa 6 15 29 88 77 215
CiCi 2 6 13 13 20 54
Total 11 33 102 160 190 496
The properties of two syllables in each of the words in the dataset, pretonic and stressed (V1 and
V2), were analysed in Praat (Boersma & Weenink 2019). The measurements made include syllable
duration, intensity, maximal F0 per syllable, and F0 values at 5 fixed points in a syllable (left edge,
25%, 50%, 75%, right edge).8
As described in Section 2, it is commonly assumed that word stress in Belarusian is acoustically
cued by greater duration and intensity of the stressed syllable, as compared to neighbouring ones, as
well as pitch accent anchoring and lack of neutralisation on the stressed vowel. However, in CaCi
contexts a direct comparison between V1 and V2 with respect to properties such as duration and intensity
cannot be carried out. This is because, as already noted, non-high vowels are intrinsically higher in
duration and intensity than non-low ones, which means that, in CaCi contexts, the effect of stress may
be offset, to a certain extent, by the fact that the pretonic vowel is lower than the stressed one. In order
to control for this, instead of making a direct comparison between the pretonic and stressed vowels in
pretonic prominence contexts, V1 in CaCi contexts is compared to V1 in CaCa contexts, and V2 in CaCi
contexts to V2 in CiCi contexts. This way, vowel height and position with respect to stress are held
constant, and any differences between the objects of comparison are attributed to pretonic prominence.
4.2 Results
First, consider the durational properties of V1 and V2. The values for vowel duration for all three
environments (CaCi, CaCa, CiCi) are provided in Table 4. The contexts to be compared to each other
within the V1 and V2 columns are shaded light grey and grey, respectively, and illustrated in Figure 1
and Figure 2.
Table 4. Vowel duration (ms) of V1 and V2 in CaCi, CaCa, and CiCi contexts
V1 V2
mean SD mean SD
CaCi 141.46 34.32 91.45 31.05
CaCa 83.15*** 21.32 131.64 30.56
CiCi 81.60 26.27 99.13 34.51
Figure 1. V1 duration in CaCi and CaCa contexts
8 The following settings were used in Praat for the acoustic analysis: pitch range = 75-400 Hz, voicing threshold = 0.6, octave
jump cost = 0.6.
11
Figure 2. V2 duration in CaCi and CiCi contexts
As Table 4 shows, in the two unmarked conditions, CaCa and CiCi, stressed vowels (V2s) are
considerably greater in duration than pretonic ones (V1s), which is expected to be the case, given that
duration is one of the key acoustic cues for stress in East Slavic. In CiCi cases, even though V2s are not
nearly as long as those in CaCa cases (which may be attributed to the fact that high vowels are
intrinsically shorter in duration that non-high ones), they are still considerably greater in duration than
V1s. Specifically, in the CaCa examples the ratio V1:V2 is 0.63:1, and in the CiCi examples it is 0.82:1.
In contrast, the relationship between the two vowels is completely reversed in CaCi cases, where the
V1:V2 ratio is 1.55:1.
For the purposes of statistical analysis, as laid out in Section 4.1, V1s in CaCi and CaCa contexts
were compared to each other, as were V2s in CaCi and CiCi contexts. The analysis was carried out using
the glmer function in the lme4 package for R (R Core Team 2017). For each of the vowels (V1 and
V2), a mixed-effects model with DURATION as the dependent variable, WORD TYPE (CaCi and CaCa, or
CaCi and CiCi) as the fixed factor, and random factors SPEAKER, ITEM, and VOWEL was run. The results
showed that V1 in CaCi contexts is significantly greater in duration that V1 in CaCa contexts (p<0.01).
In turn, V2-s in CaCi and CiCi contexts do not differ significantly (p=0.07); nevertheless, note that the
mean duration of stressed vowels in CaCi contexts is consistently shorter than that in CiCi contexts
(91.45ms and 99.13ms, respectively).
Because syllable/vowel duration is one of the main cues for stress in Belarusian, the duration results
might suggest that pretonic prominence is a retraction of stress from V2 to V1, given that V1 receives a
significant increase in duration in CaCi contexts, accompanied by a slight decrease in duration of V2.
The F0 results discussed below, however, do not lend support to this conclusion. Consider mean F0
values measured in the mid-point (50%) of V1s and V2s in all three contexts (CaCi, CaCa, CiCi), which
are provided in Table 5, with the values within the V1 and V2 columns to be compared to each other
shaded. Figure 3 further illustrates the F0 properties of V1s in CaCi and CaCa contexts, and Figure 4
does the same for V2s in CaCi and CiCi contexts.
Table 5. F0 (Hz) at the mid-point of V1 and V2 in CaCi, CaCa, and CiCi contexts
V1 V2
mean SD mean SD
CaCi 203.07 44.36 222.44 59.97
CaCa 209.15 44.97 216.27 52.64
CiCi 225.63 47.82 224.75 64.91
Figure 3. F0 at the mid-point of V1 in CaCi and CaCa contexts
12
As Table 5, Figure 3, and Figure 4 show, the F0 properties of CaCa and CaCi contexts are parallel:
in both, V2 at its mid-point has slightly higher F0 than V1. CiCi contexts exhibit a different pattern,
where the two vowels are very close in their F0 properties, which, at least in part, may be attributable
to the vowels intrinsic pitch properties. Note also that there is more variation on V2s than V1s in CiCi
contexts, which is expected, given that the F0 properties of the stressed vowel may vary, depending on
the type of an intonational pitch accent that it carries.
13
Figure 4. F0 at the mid-point of V2 in CaCi and CiCi contexts
A statistical analysis similar to that performed for the purposes of comparing vowel durations was
carried out. Specifically, for each of the vowels (V1 and V2), a mixed-effects model with F0 (50%) as
the dependent variable, WORD TYPE (CaCi and CaCa, or CaCi and CiCi) as the fixed factor, and random
factors SPEAKER, ITEM, and VOWEL was run. The results revealed no significant differences in F0
properties between V1s in CaCi and CaCa contexts, or V2s in CaCi and CiCi contexts (p=0.185 and
p=0.8, respectively).
Moreover, the F0 contours found on V1-s in in CaCi and CaCa contexts are remarkably similar: in
both cases, there is an overall fall in F0 throughout the vowel, as shown in Figure 5. Even though the
steepness of the fall varies between the two conditions (in CaCi contexts, the fall is lower than in CaCa
ones), the overall parallelism between the vowels in the two conditions is striking, especially given that
they significantly differ in their duration.
Figure 5. Average F0 contours on V1-s in in CaCi and CaCa contexts
Finally, consider the intensity properties of V1 and V2, with the average values for the three contexts
(CaCi, CaCa and CiCi) summarized in Table 6. As before, the contexts within the V1 and V2 columns
to be compared to each other are shaded light grey and grey, respectively, and also illustrated in Figure
6 and Figure 7.
14
Table 6. Vowel intensity (dB) of V1 and V2 in CaCi, CaCa, and CiCi contexts
V1 V2
mean SD mean SD
CaCi 76.38 4.73 73.65 5.41
CaCa 75.33 4.6 75.43 4.95
CiCi 75.3 4.69 75.16 5.56
Figure 6. Intensity of V1 in CaCi and CaCa contexts
Figure 7. Intensity of V2 in CaCi and CiCi contexts
As Table 6 shows, average intensity values on the vowels in all six contexts are very close. At the
same time, Figure 6 and Figure 7 reveal that there is a tendency for V1s in pretonic prominence contexts
to have greater intensity than V1s in the unmarked (CaCa) contexts, while the opposite is true for V2s:
they have lesser intensity in pretonic prominence contexts as compared to unmarked ones (CiCi). These
trends, however, did not lead to statistical significance. Specifically, for the purposes of intensity, a
mixed-effects model with INTENSITY as the dependent variable, WORD TYPE (CaCi and CaCa, or CaCi
and CiCi) as the fixed factor, and random factors SPEAKER, ITEM, and VOWEL was run for each of the
vowels (V1 and V2). There were no significant differences in F0 properties between V1s in CaCi and
CaCa contexts, or V2s in CaCi and CiCi contexts (p=0.664 and p=0.195, respectively).
4.3 Discussion
As the instrumental results show, the distribution of prosodic prominence between V1s and V2s, in the
form of duration, F0 and intensity, is quite different in the pretonic prominence contexts, on the one
hand, and the unmarked contexts of both types on the other. Let us start with duration. As we have seen,
in CaCa and CiCi contexts, stressed vowels have considerably greater duration values than pretonic
ones. This is not surprising, given that greater duration is one of the hallmarks of a stressed vowel in
East Slavic languages. In contrast, in pretonic prominence contexts, the ration between the two vowels
is reversed: here, V1s are considerably greater in duration than V2s. The same is confirmed by the
comparison of V1s in CaCi and CaCa contexts: V1s in the former environments have significantly
greater than those in the latter. The current instrumental results, therefore, align with fieldworkers’
15
reports and the existing instrumental studies, discussed in Section 3: according to most of them, greater
duration of V1 is the most prominent prosodic characteristic of pretonic prominence contexts. What is
more surprising about the current results is that duration also turns out to be the main prosodic cue that
pretonic prominence relies on – as opposed to F0 or intensity.
Let us recap the facts. In terms of F0, we have seen that, in CaCi and CaCa contexts, V2s carry
consistently higher F0 values than V1s, and in CiCi contexts, the two vowels have comparable F0
properties. There is no evidence for V1s in CaCi contexts carrying an F0 peak, or anything that could
be described as a ‘musical’ contour that would differentiate them from V1s in the unmarked contexts.
Moreover, the F0 contours that span V1s in CaCi and CaCa contexts are nearly identical – both falling
– even though the pretonic vowels in the two contexts otherwise differ in their prosodic characteristics,
such as duration. These results, therefore, refute an analysis of pretonic prominence as relying on a
retraction of a stress-related H tone/F0 peak from V2 to V1 in pretonic prominence contexts. They are
also in conflict with the numerous fieldworkers’ reports, according to which V1s in pretonic prominence
contexts carry a distinctive tonal contour. Note, however, that the current F0 results broadly align with
those in Belaja (1974), who also found that V1s in pretonic prominence contexts are not commonly
characterised by greater F0 values than V2s.
Finally, with respect to intensity, we have seen that there are no major differences between all six
types of vowels. These results contrast with those in Belaja (1974) and Borise (2017), as well as
Almuxamedova and Kulsharipova’s (1980; for Vladimir-Volga basin dialects) who found V1s in
pretonic prominence contexts to be characterised by greater intensity than their counterparts in the
unmarked contexts.
An overall conclusion to be made based on the current data, then, is that pretonic prominence
contexts are most consistently cued by greater duration of V1s, as opposed to F0 and intensity. This, in
turn, has important implications for the possible theoretical accounts of pretonic prominence.
Specifically, the fact that V1s in pretonic prominence contexts do not carry a distinct F0 contour does
not allow one to cast pretonic prominence as a tonal phenomenon. Furthermore, the evidence laid out
in the next section confirms that pretonic prominence should not be thought of as stress retraction either.
After establishing that, the next section offers an alternative analysis of pretonic prominence, based on
its distributional parallelism with dissimilative vowel neutralization.
5. Phonological analysis
5.1 Preliminaries
Given that pretonic prominence is mainly cued by greater duration of V1, a potential analysis to rule out
is taking pretonic prominence to be a retraction of stress to V1. Indeed, recall from Section 2 that stress
retraction has been mentioned in the existing literature as possibly affecting the Aŭciuki dialect. On the
other hand, fieldworkers did not definitively commit to this view, and neither did Belaja (1974), even
though her instrumental result would be compatible with such an interpretation.
There is some evidence that a stress-retraction account of pretonic prominence would be on the
wrong track. First, vowel neutralization facts are incompatible with it, which has been noted in the
literature before (cf. Almukhamedova & Kulsharipova 1980: 49 for Vladimir-Volga basin dialects;
Bethin 2006a: 132 for Upper Snov dialects). Specifically, in those cases when pretonic prominence
affects /ɔ/ (V1) before a stressed /o/ (V2), there is no vowel neutralisation on V2, as illustrated in (5):
(5) basonožki ‘sandals’: [basɔːˈnoški], not [baˈsɔːnaški]
ɣodoŭ ‘years.GEN’: [ɣɔːˈdow], not [ˈɣɔːdaw]
16
This would be unexpected in a stress retraction context: had the stress shifted to V1, V2 would have
neutralised to /a/, as post-tonic non-high vowels in Belarusian do (Czekman & Smułkowa 1988; Mayo
1993). Since this is not the case, vowel neutralization facts serve as evidence against an analysis of
pretonic prominence as a retraction of stress to V1.
Additionally, the alignment of intonational pitch accents also suggest that stress in pretonic
prominence contexts targets its etymological position (V2). It is a known fact of intonational phonology
that intonational pitch accents anchor to lexically stressed syllables (Pierrehumbert 1980; Ladd 2008,
a.o.). At the same time, the actual F0 contours that are found on the stressed syllable can vary, depending
on the type of pitch accent (H* or L*), and on whether it has a leading or a trailing tone, which may
affect the pretonic or post-tonic vowel as well (e.g, L+H*, H*+L, etc.). In Belarusian (both standard
and dialectal varieties), emphatic contexts are marked by the presence of a pronounced H* pitch accent
on the stressed syllable. As Figure 8 and Figure 9 show, in a word such as dražnili ‘mocked.PL’ (which
is subject to pretonic prominence in the Aŭciuki dialect, [draːžˈnili]), H* is aligned with V2 both in
standard Belarusian and in the Aŭciuki dialect. Had there been stress retraction, H* in the Aŭciuki
dialect would be aligned with V1 instead.
Figure 8. H* aligned with V2 in standard Belarusian
Figure 9. H* aligned with V2 in a pretonic prominence context in Aŭciuki Belarusian
Finally, the dialect speakers, when prompted, explicitly assign stress to its etymological position
(V2). In particular, as part of fieldwork research, the informants were asked to list the names of the
17
neighbouring villages. When the fieldworker repeated those that were affected by pretonic prominence
(such as e.g. Žohli [ʒɔːˈɣli]) with stress on V1 ([ˈʒɔɣli]), the speakers corrected that.
As the evidence above shows, pretonic prominence in the Aŭciuki cannot be analysed as a retraction
of H or as stress retraction. In the remained of this section, I show that pretonic prominence, instead,
can be successfully accounted for if the connection between it and dissimilative vowel neutralisation is
recognised. Specifically, I show that both phenomena, though not necessarily dependent on each other,
rely on the same mechanism: redistribution of stress-induced prosodic prominence within the iambic
foot headed by the stressed syllable.
Various types of vowel neutralization in unstressed syllables are found in East Slavic. Notably, the
vowel neutralization patterns that apply to the immediately pretonic syllable are weaker than those that
apply to the posttonic and further pretonic syllables. Standard Russian, for example, exhibits two
degrees of vowel neutralization: the weaker degree applies to the immediately pretonic syllables, while
the stronger degree applies to all other unstressed syllables (further pretonic, and post-tonic). In
particular, non-high vowels in the immediately pretonic position in standard Russian neutralise to /a/,
while those in post-tonic and further pretonic positions neutralise to /ə/, as illustrated in (6):
(6) moloko ‘milk’ [məlaˈko]
karandaš ‘pencil’ [kǝranˈdaʃ]
slovo ‘word’ [ˈslovǝ]
In varieties of Eats Slavic with dissimilative vowel neutralization, the resulting quality of V1 (but
not any other unstressed vowels) is directly dependent on the quality of the stressed vowel, V2.
Specifically, non-high V1s are neutralized to [a] in the presence of a non-/a/ V2 (i.e., V2 that is
represented by any vowel other than /a/), and to [ə] if V2 is /a/ – hence the name ‘dissimilative’ vowel
neutralization. A schematic representation of the pattern is provided in Table 7, and illustrated in (7).
Table 7. The dissimilative pattern of vowel neutralisation
V1 V2
[a] ¬ [a]
[ə] [a]
(7) trava ‘grass.NOM’ [trəˈva] travy ‘grass.GEN’ [traˈvɨ]
voda ‘water.NOM’ [vəˈda], vody ‘water.GEN’ [vaˈdɨ]
Note that the mechanism behind dissimilative vowel neutralization is similar to that of pretonic
prominence: in both cases, the properties of V1 (quality and quantity, respectively) are determined by
the quality of V2. More generally, the patterns of vowel neutralization attest to the close connection that
exists between V1 and V2 in East Slavic. Specifically, V1 is either subject to a pattern of vowel
neutralization that is different from the one that affects all other unstressed syllables, or is even directly
dependent in its realization on the quality of V2. The connection between the two vowels has also been
pointed out in the literature, both descriptive and generative (Čekmonas 1987; Kasatkina 1996a; 1996b;
Crosswhite 1999; 2000, a.o.).
The connection between vowel neutralization, particularly the dissimilative kind, and vowel
quality-sensitive pretonic prominence of the Aŭciuki type that the current analysis is built on has also
18
been noted in the existing literature (Broch 1916; Vojtovich 1972b; Belaja 1974; Čekmonas 1987 and
references therein).
5.2 Theoretical approaches to vowel neutralisation in Slavic
The topic of East Slavic (dissimilative) vowel neutralization has received considerable attention in the
literature. In what follows, some of the key (theoretical) accounts are considered. Halle (1965), building
on the insight from Jakobson (1962), was one of first to cast the observation that the quality of V1 in
dissimilative vowel neutralization contexts stands in direct opposition to the quality of V2 in formal
terms. Specifically, he works out the rule that derives the opposing values of the two vowels, in different
varieties of Russian, with the help of combinations of formal phonological features, such as [± high],
[± low], etc. Davis (1970), relying on a richer inventory of ordered rules, further shows that there are
internal subgroupings within the dissimilative varieties that Halle considered, which can be subsumed
under the same set of rules.
Next, an OT account of vowel neutralization in Slavic is provided by Crosswhite (1999; 2000;
2001).9 The crux of Crosswhite’s analysis is that (dissimilative) vowel neutralization relies on different
moraic values that vowels of different qualities have, combined with requirements on footing. In
particular, building on the long-noticed connection between V1 and V2 in East Slavic, she adopts the
view that the stressed and pretonic syllable in East Slavic comprise an iambic foot (cf. also Alderete
1999: 12; Suzuki 1998; Barnes 2002). All other syllables within the word are unfooted.
Furthermore, the foot should contain at least two moras. In turn, the moraic contents of V1 and V2
reflects the intrinsic sonority of vowels (given that vowel length is not contrastive in East Slavic). In
particular, the lower the vowel, the more likely it is to contribute two moras (the cut-off point between
vowels that contribute one and two moras is dialect- specific, given that there are different subtypes of
dissimilative vowel neutralization, not discussed here). That is to say, a mora, under this approach, is a
unit of time, which reflects the intrinsic phonetic duration of a vowel, determined by its height. If the
stressed vowel is long enough and contributes two moras by itself, it also comprises a foot by itself,
which leads to strong neutralization of the pretonic vowel, as shown in the second row in Table 7. If,
on the other hand, the stressed vowel is not low enough to contribute two moras, V1 and V2 comprise a
foot together and contribute a mora each, with the pretonic vowel undergoing a weaker degree of
neutralization (first row in in Table 7). Finally, moraicity is only retained within the foot. Unfooted
vowels are non-moraic – this is reflected in stronger vowel neutralization, as well as common vowel
loss in further pretonic and post-tonic syllables.
In OT terms, Crosswhite derives this analysis via variable ranking of the constraints that require for
the stressed vowel to be bi-moraic (WSP, as in (8)), and the constraints that prohibit vowels of other
heights to contribute two moras, as in (9):
(8) WSP (Weight-to-Stress Principle): Stressed vowels should be bi-moraic (Smolensky 1993)
(9) */i,u » */e,o » */ɛ,ɔ » */a, æ
The resulting constraint ranking allows to derive the available patterns of dissimilative vowel
neutralization as a function of footing and moraic contribution of vowels of different heights.
9 An analysis similar in spirit to Crosswhite’s is also independently offered by Nesset (2002). The difference
between the two lies in the fact that, in addition to laying out an account of dissimilative vowel neutralization,
Nesset (2002) also provides a way to account for the two types of V1 assimilation following palatal(ised)
consonants, to [ǝ] and to [i]. While this contribution is very valuable for accounting for the Slavic neutralization
patterns more generally, the distinction is question is not relevant to the Aŭciuki data. Therefore, Crosswhite’s
approach is adopted here instead, as a simpler one.
19
5.3 Vowel neutralisation and pretonic prominence
The insight that Crosswhite’s approach to dissimilative vowel neutralization is built on – the
combination of variable moraic weights and requirements on footing – can also be extended to account
for the pretonic prominence facts. The advantage of such an analysis is that it allows one to make a
formal connection between two phenomena that rely on the same mechanism (and have been
hypothesized to be diachronically connected; more on this below).
I suggest that pretonic prominence, similarly to dissimilative vowel neutralizations, relies on a
sonority effect: in pretonic prominence contexts a non-low V2 is low on sonority, which leads to V1
receiving a compensatory increase in duration. Every word in the Aŭciuki dialect contains an iambic
foot, which consists of V1 and V2; all other syllables are unfooted. Within the foot, non-high vowels are
bi-moraic, and non-low vowels are mono-moraic (cf. also Crosswhite 2001 for a similar treatement of
neutralization facts in Carniolan Slovenian). As was the case for dissimilative neutralization, the key
insight is that moras are timing units, and mora count of a given vowel, in a language without vowel
length contrasts, depends on its phonetic duration, which is a function of its height. Unfooted vowels,
in turn, are non-moraic – this is reflected in stronger vowel neutralization, up to complete vowel loss,
that applies to syllables outside of the foot.
When V1 equals V2 in height, in both CaCa and CaCi contexts, they contribute a mora each. In
contrast, in pretonic prominence contexts, the sonority of the vowels is unequal, which results in V2
losing a mora (and, as a result, not contributing any), and V1 acquiring one. While it may seem that the
stressed vowel not contributing a mora is a paradoxical situation, recall that V2s in pretonic prominence
contexts undergo a degree of shortening, which provides acoustic evidence for pretonic prominence as
“moraic oozing” from V2 to V1 (Bruce Hayes, p.c; cf. also Hayes 1989 on compensatory lengthening in
moraic phonology). Acoustically, “moraic oozing” is what is responsible for V1 being longer and higher
in sonority than V2 in pretonic prominence environments.
The constraints that this analysis relies on are the following:
(10) Undominated:
MAX, DEP, IDENT
NO-FLOP-PROM ‘Corresponding prominences must have corresponding sponsors and
links’ (Alderete 1999); ensures that stress surfaces on the
etymologically stressed syllable.
RHTYPE=IAMB For every foot, assign a penalty if stress is not right-aligned in that foot
(to ensure the foot is present).
(11) Ranked:
FT-BIN Each foot is two syllables and two moras
*STRUC-μ Moras do not appear in output forms (Crosswhite 1999, 2000)
[+low]μ: Non-high vowel ≥ μ (‘Non-high vowels should contribute at least a mora’)
[+low]μμ: Non-high vowel = μμ (‘Non-high vowels should contribute two moras’)
*[-low]μμ: Non-low vowel ≠ μμ (‘Non-low vowels cannot contribute two moras’)
(Crosswhite 1999, 2000)
The derivations of CaCi, CaCa and CiCi contexts are provided in Table 8, Table 9, and Table 10,
respectively:
20
Table 8. Tableau deriving pretonic prominence contexts
Table 9. Tableau deriving unmarked contexts of the CaCa type
Table 10. Tableau deriving unmarked contexts of the CiCi type
Note that Bethin (2005) also considers building her OT analysis of pretonic prominence on
Crosswhite’s analysis of dissimilative vowel neutralization, but ultimately rejects it. She suggests that
a Crosswhite-style analysis of pretonic prominence as “quantity dissimilation” would be stipulative,
and instead develops an analysis of pretonic prominence as based on a shift of H from V2 to V1. As the
data is Section 4 showed, however, there is no evidence for the shift of H to the pretonic vowel, while
“quantity dissimilation” is indeed what Aŭciuki-style pretonic prominence is built on. Even more to the
point, as mentioned above, not only does V1 in pretonic prominence contexts receive greater duration,
which corresponds to acquiring the second mora, V2 in the same contexts is also somewhat shortened,
as compared to V2s in CiCi contexts, which is the manifestation of losing the mora. The “quantity
dissimilation”, therefore, is an appropriate characterization of pretonic prominence.
Now that the formal connection between dissimilative vowel neutralization and pretonic
prominence have been established, the next question to ask is the following: how does the Aŭciuki
dialect behave with respect to vowel neutralization? More specifically, if a Crosswhite-style analysis is
employed for accounting for pretonic prominence, how is vowel neutralization analyzed, if present?
The facts are provided below, and, I suggest, they provide support for the current analysis. In particular,
Aŭciuki dialect is inconsistent with respect to vowel neutralization (Zhylko 1953; Kryvicki 1959;
Vojtovich 1972b). Sitting on the boundary between dialects with and without neutralization, and
currently undergoing the process of developing vowel neutralization, the dialect exhibits a mixed
neutralization pattern (Kryvicki 1959: 99–100), shown in Table 11.
21
Table 11. The pattern of vowel neutralization in the Aŭciuki dialect
V1 V2
etymological /ɔ/ etymological /a/
[ɔ:] [a:] /i, ɨ, u, o, e/
[ɔ] or [a] [a], rarely [ɔ] /ɔ, ɛ/
[a], rarely [ɔ] [a] /a/
As Table 11 shows, there are some dissimilative tendencies in the Aŭciuki pattern of vowel
neutralization, though they do not fit the canonical pattern illustrated in Table 7. In particular, the
etymological value of V1 is most faithfully rendered in the presence of a non-low V2, while that is not
the case with a non-high V2. Furthermore, according to the author’s informal fieldwork observations,
younger dialect speakers, who do not have pretonic prominence anymore, exhibit clear dissimilative
vowel neutralization, which is absent from the speech of older informants.
The hypothesis is, then, that pretonic prominence is a temporary stage in a dialect that is developing
dissimilative vowel neutralization, during which the opposition between V1 and V2 is expressed
quantitively and not qualitatively. This view has been expressed in the existing literature (Vojtovich
1972b; 1972a; Čekmonas 1987, a.o.). Note that, while there is no agreement in the literature about the
exact diachronic connection between pretonic prominence and vowel neutralization, there is some
consensus, that dissimilative vowel neutralization (quality dissimilation) and pretonic prominence
(quantity dissimilation) are two processes that may but do not have to co-occur. That is to say,
dissimilative vowel neutralization may arise without a language variety first going through a stage of
pretonic prominence, though it may also be preceded by it. This, I suggest, is precisely what is currently
happening in Aŭciuki Belarusian: pretonic prominence is giving way to a dissimilative pattern of vowel
neutralization. Finally, it should be noted, that a type of a phenomenon such as pretonic prominence is
also expected to be short-lived for another reason: it is a sub-type of an Iambic Lengthening, a
typologically rare process that is predicted to be unstable (Barnes 2008).
6. Conclusion Let us recap the main finding of the current paper. We have seen that, in the Aŭciuki dialect of
Belarusian, immediately pretonic vowels (V1), in certain contexts, can be more acoustically prominent
than stressed ones (V2) – a fact that has been noted in numerous fieldwork accounts. In particular, this
take place in contexts in which V1 is non-high and V2 is non-low. Novel instrumental data, presented in
this paper, unambiguously shows that, acoustically, the Aŭciuki pretonic prominence mainly relies on
increased duration of the V1, and somewhat decreased duration of V2. There is no significant difference
in intensity values between the two vowels, and the two do not systematically differ in their F0 values.
More specifically, there is no evidence for there being a particular tonal contour associated with V1 in
pretonic prominence contexts; on the contrary, intonational pitch accents in Aŭciuki Belarusian align
with the etymologically stressed syllable, just as they do in standard Belarusian. The distribution of
intonational pitch accents, combined with the fact that V2s in pretonic prominence contexts do not
neutralise, also provide definitive evidence against an analysis of pretonic prominence as a retraction
of stress to V1. The fact that the Aŭciuki pretonic prominence is acoustically cued by vowel duration
also means that it may be more felicitously characterised as pretonic lengthening – a phenomenon that
occurs in some non-Slavic languages as well, such as Tiberian Hebrew, Canadian French, and Córdoba
Argentinian Spanish.
Analytically, the current paper built on the known connection between pretonic prominence and
dissimilative vowel neutralization. In particular, it has long been observed that the two phenomena rely
on the same mechanism: in both, the properties of V1 are dependent on the quality of V2, but in pretonic
22
prominence contexts this manifests as a quantity dissimilation, while in dissimilative vowel
neutralization – as quality dissimilation. The similarity in nature between the two phenomena is what
the formal account offered here is built on, too. Specifically, I adopt Crosswhite’s (1999; 2000; 2001)
analysis of dissimilative vowel neutralization to account for the Aŭciuki pretonic prominence. The main
analytical thrust of the analysis lies in the interaction of footing requirements and moraic contents of
vowel of different heights, which allows to derive the pretonic prominence facts as a function of ‘moraic
oozing’ between V1 and V2.
The two phenomena, dissimilative vowel neutralization and pretonic prominence, have also been
hypothesized to be diachronically related. The Aŭciuki facts provide some evidence for this
relationship, too: younger speakers, who do not have pretonic prominence anymore, often exhibit
dissimilative vowel neutralization. This, I suggested, shows that pretonic prominence may be a possible
temporary stage in a dialect that is undergoing the process of developing (dissimilative) vowel
neutralization.
The hope is that the analysis outlined hare may be tested on other, non-Slavic languages that have
pretonic prominence too.
Abbreviations
Glosses: 3 - third person, ACC - accusative, GEN - genitive, IMP - imperative, INS - instrumental, LOC
- locative, NOM - nominative, PL - plural, SG - singular.
References
Alderete, John D. 1999. Morphologically governed accent in Optimality Theory. University of
Massachusetts, Amherst Doctoral dissertation.
Almukhamedova, Z. M. & R.E. Kulsharipova. 1980. Redukcija glasnyx i prosodija slova v okajuščix
russkix govorax [Vowel reduction and word prosody in Russian dialects with okanje]. Kazan:
Kazan University Publishing.
Avanesov, Ruben I. 1927. O govore Pereslavl-Zalesskogo uezda Vladimirskoj gubernii [On the
dialect of the Pereslavl-Zalessky uyezd of the Vladimir guberniya]. Trudy postojannoj
komissii po dialektologii russkogo jazyka 9. 62–72.
Barnes, Jonathan. 2002. Positional neutralisation: a phonologisation approach to typological
patterns. University of California, Berkeley Doctoral dissertation.
Barnes, Jonathan. 2008. Strength and weakness at the interface: Positional neutralization in phonetics
and phonology. Berlin ; New York: Mouton de Gruyter.
Belaja, Aleksandra S. 1974. K xarakteristike kvantitativno-prosodičeskix različij v Nadsnovskix
govorax na Černigovščine [For the description of the quantitative prosodic contrasts in the
Snov dialects of the Chernigov area]. In R. I. Avanesov, S.B. Bernštejn, S.K. Požaritskaja &
F.P. Filin (eds.), Obščeslavjanskij lingvističeskij atlas: materialy i issledovanija, 23–31.
Moscow: Nauka.
Bernshtein, Samuil B. 2005. Sravnitel’naja grammatika slavianskix jazykov [A comparative grammar
of the Slavic languages]. Moscow: Nauka.
Bethin, Christina Y. 2005. On pretonic length in Belarusian and Ukrainian Nadsnovs’ki dialects. In
Steven Franks, Frank Y. Gladney & Mila Tasseva- Kurktchieva (eds.), Formal approaches to
Slavic linguistics 13: the South Carolina meeting., 52–67. AnnArbor: Michigan Slavic
Publications.
Bethin, Christina Y. 2006a. Stress and Tone in East Slavic Dialects. Phonology 23(2). 125–156.
Bethin, Christina Y. 2006b. From Pitch Accent to Stress: Peak Retraction in the Nadsnovs’ki
Dialects of Ukraine and Belarus. Harvard Ukrainian Studies 28(1/4). 69–79.
23
Bila, Oleksandra S. 1970. Dovgi golosni v zoni ukrain’sko-bilorus’kix kontaktiv [Long vowels in the
Ukrainian-Belarusian contact areas]. Praci XIII Respublikan’skoj dialektolohičnoj naradi,
251–257. Kyiv: Nauk. dumka.
Boersma, Paul & David Weenink. 2019. Praat: doing phonetics by computer [Computer program].
http://www.praat.org/ (8 November, 2019).
Bondarko, Lija V. 1966. Akustičeskie xarakteristiki bezudarnosti. [The acoustic properties of absence
of stress.]. In Ludmila A. Verbitskaya, Lev R. Zinder & Vyacheslav V. Ivanov (eds.),
Strukturnaja tipologija jazykov., 56–64. Moscow: Nauka.
Bondarko, Lija V. 1977. Zvukovoj stroj sovremennogo russkogo jazyka. Moscow: Prosveščenie.
Borise, Lena. 2017. Prominence Redistribution in the Aŭciuki Dialect of Belarusian. (Ed.) Yohei
Oseki, Masha Esipova & Stephanie Harves. Proceedings of Formal Approaches to Slavic
Linguistics (FASL) 24 94–109.
Borise, Lena. 2018. Two levels of mid-vowels in the Aŭciuki dialect of Belarusian? Yes and no.
Harvard University, ms.
Broch, Olaf. 1916. Govory k zapadu ot Mosalska. [Dialects to the west of Mosalsk]. Petrograd:
Imperatorskaja Akademija Nauk.
Burova, E.G. & Leonid L. Kasatkin. 1977. Čuxlomskoje akanje [The Chukhloma akanje]. In Ruben I.
Avanesov (ed.), Dialektologičeskie issledovanija po russlomy jazyku. Moscow: Nauka.
Čekmonas, Valerij N. 1987. Territorija zarozdenija i etapy razvitija vostocnoslavjanskogo akanja v
svete dannyx lingvogeografii [The original territory and developmental stages of East Slavic
akanje with reference to the evidence from linguistic geography]. Russian Linguistics 11.
335–349.
Crosswhite, Katherine. 1999. Vowel Reduction in Optimality Theory. UCLA. UCLA PhD Thesis.
Crosswhite, Katherine M. 2000. Vowel Reduction in Russian: A Unified Account of Standard,
Dialectal, and “Dissimilative” Patterns. (Ed.) Katherine M. Crosswhite & Joyce McDonough.
University of Rochester Working Papers in the Language Sciences, Spring(1).
http://www.rochester.edu/college/cls/assets/pdf/working/crosswhite.pdf (21 May, 2016).
Crosswhite, Katherine M. 2001. Vowel Reduction in Optimality Theory. (Ed.) Laurence Horn. New
York; London: Routledge.
Czekman, Walery & Elżbieta Smułkowa. 1988. Fonetyka i fonologia języka białoruskiego z
elementami fonetyki i fonologii ogólnej [Phonetics and phonology of the Belarusian language
with elements of general phonetics and phonology]. Warszawa: Państwowe Wydawnictwo
Naukowe.
Davis, Philip. 1970. A Classification of the Dissimilative Jakan’e Dialects of Russian. Orbis. Bulletin
international de documentation linguistique 19(2). 360–376.
Durnovo, Nikolaj. 1914. Otčet ekspedicii dlja opredelenija granicy okanja i akanja [A report of the
fieldwork expedition aimed at establishing the boundaries of okanie and akanie]. Russkij
Filologičeskij Vestnik 71. 363–382.
Dybo, Vladimir A. & Sergej L. Nikolaev. 1998. Novye dannye i materialy po balto-slavjanskoj
akcentologii: problemy slavjanskogo jazykoznanija. In Vladimir A. Dybo (ed.), Tri doklada k
XII meždunarodnomu sjezdu slavistov, 5–70. Moscow: Rossijskaja Akademija Nauk.
Fontanella de Weinberg, María Beatriz. 1971. La entonación del español de Córdoba (Argentina).
Thesaurus: Boletín del Instituto Caro y Cuervo 26(1). 11–21.
Gouskova, Maria. 2010. The phonology of boundaries and secondary stress in Russian compounds.
The Linguistic Review 27(4). 387–448.
Halle, Morris. 1965. Akan’e. The treatment of unstressed nondiffuse vowels in Southern Great
Russian dialects. Symbolae linguisticae in honorem Georgii Kurylowicz, 103–109. Warszawa:
Polska Akademia Nauk. Komisja Jezykoznawstwa.
Hayes, Bruce Philip. 1989. Compensatory lengthening in moraic phonology. Linguistic inquiry 20(2).
253–306.
Hyman, Larry M. 2012. Do all languages have word accent? UC Berkeley Phonology Lab Annual
Reports 32–54.
Jakobson, Roman. 1962. Remarques sur l’évolution de russe moderne comparée à celle des autres
langues slaves. Selected writings I: phonological studies, 92–104. The Hague: Mouton.
24
Jones, Daniel & Dennis Ward. 1969. The Phonetics of Russian. Cambridge: Cambridge University
Press.
Kasatkin, Leonid L. 1989. Fonetika [Phonetics]. In Leonid L. Kasatkin (ed.), Russkaja dialektologija.
Moscow: Prosveščenie.
Kasatkina, Rozalija F. 1996a. Mezslogovaja assimiljacija glasnyx v russkix govorax [Intersyllabic
vowel assimilation in Russian dialects]. In Tatjana M. Nikolaeva (ed.), Prosodičeskij stroj
russkoj reči, 207–221. Moscow: Russian Language Institute of the Russian Academy of
Sciences.
Kasatkina, Rozalija F. 1996b. Srednerusskie govory i ritmika slova [Central Russian dialects and the
word rhythm system]. In Tatjana M. Nikolaeva (ed.), Prosodičeskij stroj russkoj reči, 222–
235. Moscow: Russian Language Institute of the Russian Academy of Sciences.
Kasatkina, Rozalija F. 2005. Moskovskoe akan’e v svete nekotoryx dialektnyx dannyx. [The Moscow
akanje in the context of some dialectal data]. Voprosy jazykoznanija 53(2). 29–45.
Kolesov, Vladimir V. 1964. Osobennosti fonologičeskoj modeli, razvivajuščej akan’e. [Aspects of the
phonological system developing vowel neutralization]. Voprosy jazykoznanija 4. 66–79.
Kryvicki, Aliaksandr A. 1959. Fanetyčnyja asablivasci adnoj z havorak poŭdnia Belarusi [Phonetic
propertis of one of the south Belarusian dialects]. Pracy Instytuta movaznaŭstva AN BSSR,
vol. 4, 98–104. Minsk: Vydavetstva Akademii navuk BSSR.
Kurylo, Olena B. 1924. Fonetyčni ta dejaki morfolohični osoblyvosti hovirky sela Xorobyčiv
Horodnians’koho rajonu Černihivs’koj oblasti. [Phonetic and some morphological aspects of
the dialect of the Khorobychiv village of the Horodnya rajon in the Chernihiv oblast]. Kyiv:
Vseukrajins’ka Akademija Nauk.
Kurylo, Olena B. 1928. Do pytannia pro umovy rozvytku dysymiliatyvnoho akannia. [On the question
of the circumstances of the development of dissimilative vowel neutralization]. Zapysky
istoryčno-filolohičnoho viddilu VUAN 16. 48–72.
Kuznetsov, V. B., Arvo Ott & A. V. Ventsov. 1987. Inherent vowel duration in Russian: production
and perception data. Proceedings of the XIth International Congress on Phonetic Sciences,
366–369.
Ladd, D. Robert. 2008. Intonational phonology. Cambridge: Cambridge University Press.
Lang-Rigal, Jennfer Ruth. 2014. A Perceptual and Experimental Phonetic Approach to Dialect
Stereotypes: The Tonada Cordobesa of Argentina. University of Texas at Austin PhD
Dissertation.
Lehiste, Ilse. 1970. Suprasegmentals. Cambridge MA: MIT Press.
Malmberg, Bertil. 1950. Études sur la Phonétique de l’Espagnol Parlé en Argentine. Lund: Glerrup.
Matusevich, Margarita I. 1976. Sovremennyj russkij jazyk: fonetika. [Contemporary Russian
language: phonetics]. Moscow: Prosveščenie.
Mayo, Peter. 1993. East Slavonic Languages: Belorussian. In Bernard Comrie & Greville G. Corbett
(eds.), The Slavonic languages, 887–946. London: Routledge.
McCarthy, John J. 1981. Stress, pretonic strengthening, and syllabification in Tiberian Hebrew.
Linguistics Department Faculty Publication Series 76.
Nesset, Tore. 2002. Dissimilation, Assimilation and Vowel Reduction: Constraint Interaction in East
Slavic Dialects with so-called Dissimilative Akan’e and Jakan’e. Poljarnyj vestnik Feb 1. 77–
101.
Nikolaev, Sergej L. 1995. Novye dannye o vostocnoslavjanskom vokalizme i prosodii. Russian
Linguistics 19. 349–370.
Nikolaev, Sergej L. 1996. Histoire d’O. In A. Gippijs, Tatjana M. Nikolaeva & V.N. Toporov (eds.),
Rusistika, Slavistika, Indoevropeistika: sbornik k 60-letiju Andreja Anatol’evila Zaliznjaka.,
2013–242. Moscow: Indrik.
Nikolaev, Sergej L. 2000. Tindinskaja i inxokvarinskaja prosodija i ee značenie dlja
severnokavkazskoj rekonstrukcii [The Tindi and Inkhokwari prosody and its significance for
the reconstruction of North Caucasian]. Problemy izučenia dal’nego rodstva jazykov na
rubeže tretjego tysiačeletija [The problems of researching distant linguistic relationships at
the turn of the third millenium], Moscow, 29 May - 2 June 2000.
Nikolaev, Sergej L. 2002. Nekotorye rezul’taty instrumental’nogo izu6enija fonetiki i prosodii
vostocno-slavjanskix govorov. In L.E. Kalnyn’, M.I. Ermakova & G.P. Klepikova (eds.),
25
Issledovanija po slavjanskoj dialektologii., vol. 8, 142–167. Moscow: Rossijskaja Akademija
Nauk.
Nikolaev, Sergej L. 2003. Novye dannye o fonetike i prosodii vostotnoslavjanskix govorov.
Slavjanskoe jazykoznanie: XIII Mezdunarodnyj s’jezd slavistov, Ljubljana 2003. Doklady
rossijskoj delegacii., 432–448. Moscow: Rossijskaja Akademija Nauk.
Nikolaev, Sergej L. 2009. Količestvennaja oppozicija glasnyx v 1-om predudarnom sloge v a-osnovax
v tverskom govore d. Gnilovka [Quantitative distinctions between vowels in the first pretonic
syllable in a-stems in the Tver dialect of the Gnilovka village]. Issledovanija po slavjanskoj
dialektologii 14. 144–157.
Pierrehumbert, Janet. 1980. The phonetics and phonology of English intonation. MIT Doctoral
dissertation.
Pozharitskaja, Sofja K. 2005. Russkaja dialektologija [Russian dialectology]. Moscow: Paradigma.
R Core Team. 2017. R: A language and environment for statistical computing. Vienna: R Foundation
for Statistical Computing. https://www.R-project.org/.
Sinjavskyj, O. 1934. Zamitka pro movu s. Ljubeča na Černihivščyni. Movoznavstvo 1. 91–97.
Smolensky, Paul. 1993. Harmony, markedness, and phonological activity. Rutgers Optimality
Workshop, vol. 1, 87–0000.
Sussex, Roland & Paul Cubberly. 2006. The Slavic Languages. Cambridge: Cambridge University
Press.
Suzuki, Keiichiro. 1998. A typological investigation of dissimilation. University of Arizona PhD
Thesis.
Vajtovich, Nina T. 1968. Nenaciskny vakalizm narodnyx havorak Belarusi [Unstressed vocalism in
the Belarusian dialects]. Minsk: Navuka i texnika.
Vidal de Battini, Berta Elena. 1964. El español de la Argentina. Buenos Aires: Consejo Nacional de
Educación.
Vojtovich, Nina T. 1971. O strukturnom parallelizme tipov bezudarnogo vokalizma i narušenijax ego
v belorusskix govorax (k probleme akanja) [On structural paralelism between types of
unstressed vowels and departures from it in Belarusian dialects (on the issue of vowel
neutralization)]. Voprosy jazykoznanija 2. 83–92.
Vojtovich, Nina T. 1972a. O svjazi vokalizma s ritmiko-intonacionnoj sistemoj v russkix i
belorusskix govorax. Russkoe i slavjanskoe jazykoznanie. K 70-letiju člena-korrespondenta
AN SSSR R.I.Avenesova, 57–63. Moscow: Nauka.
Vojtovich, Nina T. 1972b. K voprosu o putjax razvitija akan’ja v vostočnoslavjanskix jazykax.
[On the question of the ways of vowel neutralization development in East Slavic languages].
In Ruben I. Avanesov, Samuil B. Bernshtein, Sofja K. Pozharitskaja & Fedot P. Filin (eds.),
Obščeslavjanskij lingvističeskij atlas: materialy i issledovanija, 17–39.
Vysotskij, S. S. 1973. O zvukovoj strukture slova v russkix govorax [On phonological word structure
in Russian dialects]. In Ju. S. Azarkh, S. V. Bromlej & L. N. Bulatova (eds.), Issledovanija po
russkoj dialektologii, 17–41. Moscow: Nauka.
Walker, Douglas C. 1984. The Pronunciation of Canadian French. Ottawa: University of Ottawa
Press.
Westbury, John R. & Patricia A. Keating. 1980. Central representation of vowel duration. The Journal
of the Acoustical Society of America 67(S1). S37–S37.
Yorio, Carlos Alfredo. 1973. Phonological style in the dialect of Spanish of Córdoba, Argentina.
University of Michigan PhD Thesis.
Zakharova, K.F. 1970. Vostočnye srednerusskie govory [The Eastern Central Russian dialects]. In
V.G. Orlova (ed.), Obrazovanie severnorusskogo narečija i srednerusskix govorov, 314–391.
Moscow: Nauka.
Zhylko, Fedot T. 1953. Perexidni hovirky vid ukrajins’koji do bilorus’koji movy v pivnično-
zaxidnyx rajonax Černihivščyny. [Transitional Belarusian-Ukrainian dialects in north-western
Chernihivshchina]. Dialektolohičnyj biuleten’ 4. 11–12.
Zlatoustova, Liubov V. 1954. Dlitel’nost’ glasnyx i soglasnyx zvukov russkogo jazyka [Duration of
vowels and consonants in Russian]. Učenye Zapiski Kazanskogo Gosudarstvennogo
Universiteta 114. 99–123.
Top Related