. 2
( 17)


900 Hz, while [í ˝! ú], produced with the tongue close to the roof of the mouth,
have the narrowest aperture and the lowest values of F1 , around 250--400 Hz;
mid vowels [†] and [ó] are intermediate. The second clustering of harmonics,
F2 , can be taken as a measure of the position of articulation on the horizontal
axis, as front (high F2 ) or back (low F2 ).3 Thus [ú], the furthest back and most
strongly labialized vowel, has the lowest F2 (around 600 Hz); the value increases
as one goes around the vowel space to [ó] (700--900 Hz), [’] (1000--1400 Hz), [†]
(1600--1800 Hz), and [í], with a value of 2000 Hz or more. [“] tends to slight diph-
thongization: [uo (or [uo] after soft consonants). F2 , incidentally, is what people
´] ´
respond to when they perceive vowels with synaesthesia and characterize, for
3 A more re¬ned view is that the perception of frontness depends also on F1 and F3 , according to
the formula F2 + 0.5 (F3 ’ F2 )(F2 ’ F1 )/(F3 ’ F1 ) (Carlson et al. 1970).
Sounds 31

Table 2.1 Properties of stressed vowels

¸ ¸ CVC
F1 F2 F1 F2 F1 F2 F1 F2

[í] 10 ms. 312 2017 317 2020
30% 316 2065 346 2114
313 2121 339 2355
70% 293 2175 273 2324
’10 ms. 261 1994 210 2061

[í] 10 ms. 404 1242 380 1136
30% 393 1563 364 1787
392 1925 352 2094
70% 383 1950 346 2144
’10 ms. 337 1650 260 2050

[†] 10 ms. 599 1361 570 1386 332 2197 348 2133
30% 723 1718 567 1824 401 2216 384 2334
702 1770 548 1947 506 2000 417 2307
70% 704 1644 488 1955 569 1744 440 2258
’10 ms. 577 1547 442 1916 468 1564 304 2102

[’] 10 ms. 815 982 801 1154 432 2011 485 1979
30% 922 1285 895 1306 770 1871 833 1887
941 1346 886 1415 979 1662 912 1768
70% 896 1443 850 1560 924 1560 881 1792
’10 ms. 551 1622 494 1839 602 1579 521 1931

[j 10 ms. 560 694 402 1430 354 1985 338 1876
30% 535 738 493 1319 426 1678 437 1847
595 809 518 1213 571 1219 482 1474
70% 534 905 510 1066 566 1054 451 1418
’10 ms. 458 1297 347 1745 465 1334 323 1734

[ú] 10 ms. 425 795 435 854 360 2024 208 1871
30% 410 626 437 856 346 1696 283 1833
401 555 433 914 383 1295 309 1662
70% 386 881 446 1147 386 1100 307 1439
’10 ms. 374 1164 490 1730 392 1483 256 1789

10 ms. = measurement 10 ms. after beginning of vowel
’10 ms. = measurement 10 ms. from end of vowel
30% (50%, 70%) = measurement at a point 30% (50%, 70%) of the duration of the vowel
32 A Reference Grammar of Russian

example, [ó ú] as dark or gloomy vowels -- they have a low F2 -- and [† í] as bright
or red or cheery vowels -- they have a high F2 . Specifying the values of F1 and F2
goes a long way towards de¬ning a vowel.
It takes a little time for each vowel to reach its target position, and some of
the duration of vowels is spent in transition from the preceding to the following
consonant. Different places of articulation (labial, dental, alveo-palatal, velar)
have characteristic effects on vowels, speci¬cally on F2 . Labial consonants ([p] or
[m]) depress the value of F2 in the transition to the vowel; dentals ([t s]) raise
F2 , and velars ([k g x]) are intermediate in their effect on F2 . These effects are
similar across languages. What sets Russian apart is the way in which vowels
interact with palatalization in consonants. It is customary to de¬ne four contexts
depending on the adjacent consonants: after a hard consonant before a hard
consonant (= Co VCo ), after a hard consonant before a soft consonant (= C o VC), ¸
after a soft consonant before a hard consonant (= CVCo ), and position between
soft consonants (= CVC). One could in principle distinguish additional contexts
in which there is no consonant either before or after the vowel. A context with no
consonant is usually equivalent to a consonant with an initial hard consonant:
VCo ≈ Co VCo , VC ≈ Co VC. (The exception is [í] -- see below.) And a context with no
¸ ¸
following consonant is similar to a context in which the vowel is followed by a
hard consonant: Co V ≈ Co VCo , CV ≈ CVCo .
¸ ¸
The vowels [’ ó ú] respond to adjacent consonants in a similar way. Measure-
ments of F1 and F2 at different points in the duration of the vowel are recorded
in Table 2.1 (one token of the speech of one speaker, reading list style).4 The
behavior of [’], illustrated in Fig. 2.1, can be taken as representative of [’ ó ú].
While [’ ó ú] differ in absolute values of F1 and F2 (see the numbers in bold
italics in Table 2.1), their contours are similar.
C O N T E XT 1: C o VC o (#VC o ): The basic allophone is a central vowel, written
without diacritics as [’], which occurs between hard consonants, Co VCo (and in
initial position, VC o ). As in v’n ˜checkmate™ [m’t] (Table 2.1 and Fig. 2.1), F2 starts
low (1100 Hz) after the hard labial [mo ] and gradually rises throughout the vowel,
in anticipation of the ¬nal hard dental [to ]. Vowels [j and [ú] are similar (vj
´] ´lf
˜fashion™ [mód´], g©cnj ˜empty™ [púst´]).
C O N T E XT 2: C o VC: After a hard consonant, before a soft consonant, as in
v’nm ˜mother™ [mat]˛ , [’] begins with a similarly low F2 . In anticipation of the
¬nal palatalized consonant, F2 is higher than with v’n already at the midpoint
(1490 Hz) and then rises sharply to a much higher value in the ¬nal transition

4 Pictures from Kay™s Computerized Speech Laboratory; measurements prepared with Praat, the
phonetics program developed by Paul Boersma. Polina Barskova was kind enough to serve as native
speaker. Bondarko 1977 and Matusevich 1976 have comparable though less speci¬c data.
Sounds 33

Fig. 2.1 v’n ˜checkmate™ [mat]

(1850 Hz) (Fig. 2.2). This rise in F2 is written here with a directional subscript
indicating fronting: [’ Because [ó] and [u] have lower values of F2 , the rise of
ö]. „
F2 is quite precipitous in anticipation of the palatalized consonant of j vj ´lt
˜about fashion™ [ mjd˛ ¬], ,©hz ˜storm™ [bur˛´].
ö 5„
C O N T E XT 3: CVC o : In the third context -- after a soft consonant before hard, as
in vz ˜crumpled™ [m˛a t] (Fig. 2.3) -- F2 in the initial transition rises very quickly
from the previous labial to an early peak of more than 2000 Hz, and then
dips to a minimum after the vowel™s midpoint, rising slightly at the end in
anticipation of the ¬nal hard dental. With [o] (v=l ˜honey™) and [u] (,·cn ˜bust™),
„ „
whose F2 values are lower, the dip and the corresponding rise at the end are more
C O N T E XT 4: CVC: In the context CVC, illustrated here by vz ˜crumple™ [m5a„ t]˛
¸¸ ¸¸ ˛5
(Fig. 2.4), F2 has a similar contour to the context CVC , but F2 rises to 2000 Hz
or more at the end. With [o„] and [u„], as in j v=lt ˜about honey™ [ømo„d˛¬], j ,·cnt
˛ 55
˜about the bust™ [øb˛ us˛t¬], the dip and subsequent rise are quite signi¬cant.
55 „ ˛
The vowel [†] has a generally similar behavior, except that its natural value
for F2 is higher than with [’ o„ u]. C o VC o : After a hard consonant before a hard

consonant, as in v¦h ˜mayor™ [m†r], ;tcn ˜gesture™ [z†st], [†] is a relatively open mid

front vowel, with F1 on the order of 600--700 Hz and F2 approximately 1600--1800
34 A Reference Grammar of Russian

Fig. 2.2 v’nm ˜mother™ [mat˛]

Fig. 2.3 vz ˜crumpled™ [ mat]
˛5 „
Sounds 35

Fig. 2.4 vz ˜crumple™ [ma„t˛]
˛ 55

(Table 2.1).5 [É] in initial position, as in ¦ ˜this™ [†t´], is similarly open. C o VC: ¸
After a hard consonant but before a soft consonant, [†] is raised and fronted,
especially in its ¬nal transition, as can be seen from the lower F1 and higher F2
in j v¦ ˜about the mayor™ [m† ¬]. The effect of a palatalized consonant on [†]
can be written as [† with the same diacritic as with [’ although with [†] the
ö], ö],
effect involves raising (lowering of F1 ) as well as fronting (raising of F2 ). CVCo : ¸
After a soft consonant, [†] has a front, high transition (with an F2 in the vicinity
of 1800--2000 Hz): n†kj ˜body™ [t˛el´], v†nrf ˜mark™ [m˛etk´]. CVC: Between soft
5„ 5„
consonants, [†] remains fronted and high throughout: v†nbnm ˜aim™ [met¬t˛], with
˛ 55 „ ˛
a low F1 , around 350--450 Hz, and a high F2 , around 2100--2300 Hz.
Among high non-rounded vowels, the variant that occurs in initial position is
[í] («df ˜willow™ [ív´]) -- about the same vowel that occurs in the context CVCo , ¸
after a soft consonant before a hard consonant (CVCo : v«nhf ˜mitre™ [m˛ítr´]). In
¸ ö
this context, F2 begins and remains high throughout, but tails off a little in
the ¬nal transition to a hard consonant (Fig. 2.5). CVCo : Before a following soft

5 [É] occurs after a hard consonant only if the consonant is unpaired or the word is a borrowing.
For this reason, Avanesov takes the position after soft consonant as basic. But the measurement
recorded in Table 2.9 shows that there is a distinct, overt transition from 2000 Hz after a soft
consonant to a target 400 or 500 Hz lower, and that transition is similar to the transition that
occurs from a palatalized consonant to other vowels such as [’].
36 A Reference Grammar of Russian

´nhf ˜mitre™ [m´tr™]
Fig. 2.5 vb

consonant, as in v«nbyu ˜meeting™ [m˛í¸¬nk], the vowel has a higher value of F2
throughout its duration (Table 2.1, Fig. 2.7).
C o VC o : After a hard consonant, the vowel that appears is [˝!] instead of the
unadulterated front vowel [í]. When the consonant following [˝!] is hard, the F2
of [˝!] starts in the vicinity of 1100--1200 Hz, a value like that of the central vowel
[’], while its F1 is similar to that of [u]. F2 then rises rapidly to a peak higher
than 2000 Hz two-thirds of the way through the vowel before falling again (see
Fig. 2.6, vßn ˜washed™ [m˝!t]). The peak value of F2 of [˝!] is nearly as high as that
of [í]. Accordingly, the increment of change in F2 over the life of the vowel is
greater than for any other stressed vowel. In this rapid and extreme change in
F2 , there is some justi¬cation for the longstanding claim that [˝!] is diphthongal.
C o VC: Before a soft consonant, as in vßnm ˜wash™ [mÈt˛], F2 , after its initial rise,
¸ 5
remains high (Fig 2.8).
The stressed vowels of Russian can be graphed as in Fig. 2.9, where the vertical
axis is the inverse of F1 and the horizontal axis is the inverse of F2 . The vowels
[í ˝!] are represented by two contextual variants each, the other vowels by four.
Fig. 2.9 re¬‚ects static, single measurements from Table 2.1 for each vowel and
context at the midpoint. Accordingly, Fig. 2.9 cannot do justice to changes over
the life of the vowel, which are especially signi¬cant for [˝!]. Despite limitations,
from Fig. 2.9 it is possible to see how the acoustic properties of values correlate
Sounds 37

Fig. 2.6 vs ˜washed™ [m˝t]
´n ´

´nbyu ˜meeting™ [m˛´ ¦¬nk]
Fig. 2.7 vb
38 A Reference Grammar of Russian

Fig. 2.8 vs ˜wash™ [m˝5t˛]
´nm ´




[í ]
[í ]

600 [o]



1000 500
2500 1500
Fig. 2.9 Midpoints of stressed vowels, contexts CVC / CVC o / C o VC / C o VC o
¸¸ ¸ ¸
Sounds 39

Table 2.2 Transcriptions of stressed vowels

Context Avanesov Panov Jones & Ward Current explicit Current simpli¬ed

C o fC o f f f ’ ’
C o tC o э э µ † †
C o ÈC o - -
È í í
s s
C o fC
¸ ™ ™ a„5
f f f ’
o tC
C¸ ˆ
э ™
э e„5
t †
C o ÈC -5 -
¸ ˆ È í í
s s
CfC o
¸ ™ ™ a„
f f f ’
µ∼ (µ6)
CtC o
¸ ™
э e„
t †
CiC o
¸ í í
b b ®
¸¸ ¦/¨/¨
ou a„
¨ f
¨ ’
¸¸ ˆ ¨
э e„
t t †
¸¸ ˆ í
5 í
b b ®

with articulatory de¬nitions; for example, [í], a high front vowel, has a low F1
and high F2 , and so on. As the lines connecting related sounds in Fig. 2.9 make
clear, the four contexts form a generally linear progression in the value of F2
from low to high: Co VCo < Co VC ¤ CVCo < CVC. The distribution of the points and
¸¸ ¸¸
lines suggests how it is possible for vowels to vary quite signi¬cantly depending
on the consonantal context and yet remain distinct from each other.
Stressed vowels, then, are affected by adjacent consonants in a consistent
fashion. Before a following palatalized consonant, all vowels are fronted and/or
raised, in the last third of the vowel and especially in the ¬nal transition.
After a soft consonant, vowels are fronted and/or raised in the ¬rst third. Be-
tween soft consonants, vowels are fronted and raised in both transitions and,
in an additive fashion, in the middle of the vowel as well. The vowels [í ˝!]
are in one sense an exception, but a motivated exception. Inasmuch as [í]
is already front (for example, in word-initial position when no consonant pre-
cedes), it is not appreciably fronted by a preceding soft consonant; instead, the
central [˝!] vowel appears after a preceding hard consonant. The generalization
that covers all vowels is that, in relative terms, transitions to adjacent soft con-
sonants are further front (higher F2 ) and/or raised (lower F1 ) than are transitions
to adjacent hard consonants.
Different systems are in use for transcribing stressed vowels in context
(Table 2.2). In the system of Avanesov 1972, for [’ o„ u], the effect of an adjacent

palatalized consonant on the vowel is indicated by a dot (a half-umlaut) above
the vowel positioned on the side of the palatalized consonant; position between
soft consonants merits a full umlaut. For [† í], the raising effect of a following
40 A Reference Grammar of Russian

soft consonant is marked by a circum¬‚ex; the effect of a preceding hard con-
sonant is indicated by using the hard-vowel letter of Cyrillic (¤э≥ and ¤s≥).
Avanesov™s system can be easily Romanized, by using ¤È≥ for Cyrillic ¤s≥ and
some distinction for Cyrillic ¤t≥ vs. ¤э≥, such as ¤t≥ vs. ¤µ≥. The Cyrillic tran-
scription of Panov 1967, which treats [†] together with [’ o„ u], transcribes the

position between two soft consonants with an umlaut and does not distinguish
CVCo and Co VC, the two environments in which a vowel is adjacent to a single
¸ ¸
palatalized consonant on one or the other side; both are marked with a centered
dot. Jones and Ward 1969 recognizes the position between soft consonants as dif-
ferent in kind from the other three positions for [’ o„ u]; this position of extreme

fronting is marked with an umlaut for [o„ u] and the digraph [¦] in the case of [’].

The basic symbol without diacritic is used for the other three positions. In the
system used here, in its most explicit form, the effect of palatalized consonants
is marked by a symbol subscripted to the vowel letter for [† ’ o„ u„], to the left side
after a palatalized consonant, to the right side before a palatalized consonant,
and with double symbols between palatalized consonants. As is conventional,
the vowel corresponding to [í] after a hard consonant is transcribed as a distinct
symbol [˝!]. There is an obvious redundancy in these transcriptions; the diacritics
re¬‚ect the contexts in which vowels can occur. Unless there is some reason for
pointing out the character of the transition to an adjacent consonant, it is of-
ten suf¬cient to omit the diacritics and transcribe with the simpli¬ed system of
[’ o„ u„ † í], with, additionally, [˝!] used after hard consonants.

2.2.2 Phonemic status of [È]
The exposition above has in effect followed the “Moscow” approach in positing
¬ve stressed vowels and in treating [˝!] as related to [í]. The incontrovertible fact
is that [˝!] is pronounced whenever [í] is put next to a hard consonant in a novel
combination, such as when a word beginning with [í] is preceded by a prepo-
sition, jn «vtyb [t˝!] ˜from the name™, or independent lexeme, xbn’k bv [lï] ˜he
read to them™. The fact that [˝!] is pronounced instead of [í] in these instances is
parallel to the fact that the vowel pronounced for ¤э≥ after prepositions is not
´njn [v†t´t], not — [v˛et´t]. Historically, whenever a consonant
fronted and raised: d ¦ 5„
has lost softness (as have c, ˇ, and ˇ), the following vowel changed from [i] to [È],
s z
as would be indicated by occasional innovative spellings in texts of ¤s≥ instead
of traditional ¤b≥. Thus, after consonants [È] and [i] are distributed complemen-
tarily, suggesting that they are related sounds: they are allophones of a single
phoneme, in a phonemic analysis.6

6 Discussion in Panov 1967:58ff.
Sounds 41

The “Leningrad” approach proceeds from a number of heterogeneous consid-
erations to argue that [È] is a phoneme distinct from [i]. One argument is the fact
that most suf¬xes begin with ¤b≥ (rather than ¤s≥) and cause “bare” softening
(palatalization) of preceding paired consonants (here termed consonant grade
Ci : §2.5.2). This distribution, however, derives from the diachronic artifact that
suf¬xes began with — i, not — y. The fact that [È] (orthographic ¤s≥) is used in
initial position in rendering exotic Asian place names (MÈqcjy in Korea) suggests
only that [È] is distinct from [i] in this one context (word initially), and then only
in a specialized lexical subsystem of not wholly assimilated lexical items.7 Over
and above these concrete observations, the basic instinct driving the Leningrad
analysis is a concern with the psychological reality of phonetics: [È] is phonemic,
ultimately, because it is psychophonetically distinct from [i].
A compromise with respect to this nagging question of the status of [i] vs. [È]
could be effected by adopting what amounts to a more radical version of the
spirit of the Leningrad approach. One might take the point of view that speakers
of Russian manipulate whole CV and VC sequences as conventionalized phonetic
units. Localizing palatalization (or its absence) in the consonant alone is an
oversimpli¬cation. For example, with respect to palatalized labials in word-¬nal
position, the palatalization in the consonant cannot be maintained or lost with-
out the preceding vowel being affected: if the labial consonant of gjpyfrj ´vmntcm
˜be acquainted!™ is pronounced without palatalization, as it often is in an infor-
mal register, the preceding vowel is also affected, hence [o„m] instead of [om]. Or 5„ ˛
when velars palatalized before [È] in the history of Russian, the change in the
consonant was correlated with a change in the vowel -- [kÈ] changed to [k˛ i].
What speakers manipulate, then, is templates of CV and VC sequences. Fine
details of phonetics have psychological reality. Among the templates are [Ci] and ¸
— [CÈ] or — [C o i], or [Ca] and [C o a] but not — [Ca] or — [C o a].8 If one
[C È] but not ¸ ¸5 „ ¸ 5„
works directly with phonetic templates, the question of whether [È] is a distinct
phoneme fades in importance.

2.2.3 Vowel duration
Russian does not have a phonemic distinction of quantity in vowels; there are
no words distinguished purely by (for example) a long [a ] as opposed to a short
[a]. Despite this, or perhaps because of this, vowels vary in duration in different
contexts.9 The most salient factor is position with respect to stress, but it will
be useful to mention some other factors, summarized in [1].

7 “known to very few native speakers of Russian” (Gordina [1989:21], who also notes that Sccsr-rekm
was changed to Bccsr-rekm in the 1930s).
8 -
Padgett 2001 sees the distinctive quality of [Ci ] in the velarization of the preceding consonant.
9 Shcherba 1912.
42 A Reference Grammar of Russian

[1] Duration of vowels
a [’] > [ó] > [† u]„ > [í] > [˝!]
b VR ≥ VZ ≥ VS ≥ VD ≥ VT
c V# ≥ VCV ≥ VCR(V) ≥ VCC(V)

Stressed vowels differ in their intrinsic duration, in proportion to the degree
of aperture (acoustically, F1 ) ([1](a)). The most open, [’], is the longest (about
200 milliseconds under stress). [“] is slightly longer than [†] and [u] (duration

around 155 ms.); [í] is shorter yet (140 ms.) and [˝!] the shortest of all (120 ms.).10
Unstressed vowels are appreciably shorter.
The duration of vowels varies depending on the adjacent consonants, particu-
larly the consonants that follow the vowel. L. V. Shcherba (1912:126ff.) was able
to document the effect of a number of factors. Before single consonants in the
¬rst, stressed, syllable of disyllabic words, vowels are shortest before voiceless
stops (g’gf ˜father™), a little longer before voiced stops (h’lf ˜glad™), longer still
before voiceless fricatives (h’cf ˜race™), and the longest before voiced fricatives
(gen sg h’pf ˜time™); each successive difference along this hierarchy was on the
order of 10 ms. for [’] in slow speech. The motivation for these differences may
be that absence of voicing requires an energetic gesture of opening the glottis,
and making a complete closure requires more energetic gestures than produc-
ing fricatives.11 As in [1](c), vowels were found to be shorter before clusters of
obstruents (g’cnf ˜paste™) than before single consonants (g’lfk ˜he fell™); however,
a cluster composed of obstruent plus sonorant (gen sg g’hyz ˜fellow™) allows al-
most the same duration in preceding vowels as singleton obstruents. Vowels are
longer when no consonant follows than when a consonant follows, and longer
when no consonant precedes.
These constraints on duration ([1]), familiar from other languages, suggest
the principle that consonants have negative valence: increasing complexity of
consonant articulation removes duration from vowels.12

2.2.4 Unstressed vowels
Above all, the duration of vowels depends on stress. If one compares the vowel
that appears after hard consonants for orthographic ¤f≥ and ¤j≥ to stressed
[’], the differences are striking. If stressed [’] has a duration on the order of 200
ms., the [ ] that appears in the ¬rst pretonic syllable is only half that, while the
[´] that appears in other unstressed positions is shorter yet, on the order of 80
ms. or less.13
10 Matusevich 1976, who does not indicate what kind of syllables were used in the measurements.
11 See de Jong 1991 on stops and fricatives, Kniazev 1989 on voicing.
12 On variation in duration, see Bondarko, Verbitskaia, and Zinder 1960.
13 Matusevich 1976:100--1.
Sounds 43

Because unstressed vowels are shorter than stressed vowels, there is less time
for the tongue to reach the articulatory positions of stressed vowels. Thus a
great proportion of the duration of unstressed vowels is spent in transition to
adjacent consonants. Unstressed vowels do not reach the articulatory extremes
of stressed vowels. They are neither as high nor as low, and neither as far front
nor as far back as stressed vowels. Acoustically the centralization of unstressed
vowels shows up as less extreme values for both F1 (re¬‚ecting vowel height) and
F2 (re¬‚ecting frontness vs. backness).14 The set of unstressed vowels occupies a
smaller portion of the vowel space than the set of stressed vowels. As an indirect
consequence of the reduced size of the vowel space, unstressed vowels tend to
merge. “Vowel reduction,” then, means a reduction in the duration of unstressed
vowels, and as a consequence, a reduced vowel space, and ultimately a reduced
number of distinctions made among unstressed vowels.15
Since vowels merge in unstressed position, it is something of a ¬ction to as-
sert that a given unstressed vowel derives from [’] or [o„] or [†]: once a vowel
is unstressed, and has been for at least ¬ve hundred years, in what sense is
it derived from [’] or [†]?16 We rely on various kinds of indirect evidence such
as etymology, orthography, and related word forms. The ¬ction, however, is un-
avoidable. In the following, stressed vowels and the unstressed vowels that derive
from them historically are written in curly braces as a set of vowels, termed a
s e r i e s . There are three basic positions: stressed, unstressed position after hard
consonant, and unstressed after soft consonant. (Sometimes it is necessary to
add a fourth position, position after hard immutable consonant ˇo = [s ‹ z].) In
S ‹
this way, for example, the series of vowels that includes stressed [’] would be
{’ Co C¬} or, more simply, {’ ¬}. As a shorthand for the whole, we can gen-
erally write simply {a} and refer to the set as the series {a}, meaning stressed
[’] with its variants and the unstressed vowels that are related to stressed [’] in
orthography, in other word forms, by etymology.
It is conventional to distinguish two degrees of reduction, de¬ned by posi-
tion relative to stress. F i r s t d e g r e e o f r e d u c t i o n -- a milder degree of
reduction -- occurs in the ¬rst pretonic syllable and in word-initial position

14 Bondarko 1977:111ff.
15 The relationship is not deterministic. Different dialect systems of Russian have different phonetic
implementations of vowels and different mergers, showing that reducing the phonetic space does
not lead automatically to a unique pattern of mergers.
16 Most models inevitably ascribe some primacy to the stressed vowel, and treat the unstressed vowel
as derivative. The suggestion here is that speakers learn unstressed vowels as part of a word form,
no less than they learn the identity of a stressed vowel. For example, ptvkz ˜land™ is learned as
[z˛ ¬] with its unstressed vowel in place. Support for the autonomy of unstressed vowels can be seen
in the fact that they can be manipulated analogically (§2.2.6). Certain analogies of stressed vowels
evidently rely on an identity of unstressed vowels: unstressed [¬] in ctr© ˜I cut™, analogous to [¬] in
ytc© ˜I carry™, motivates stressed c=r, analogous to y=c.
44 A Reference Grammar of Russian

(when there is no preceding consonant to cut into the duration of the vowel).
Vowels not in ¬rst pretonic position (and adjacent to consonants) -- in second
or more pretonic or in post-tonic position -- are subject to more extreme, or
s e c o n d - d e g r e e , reduction. There may be slight differences among second-
degree contexts -- post-tonic vowels are perhaps longer (though less loud) than
pretonic vowels two syllables from the stress17 -- but these are ¬ne details ignored
in transcription.

Series {i u}: Vowels of series {i u} are affected in a less obvious fashion than
other vowels. Not all transcriptions write symbols for unstressed, reduced high
vowels distinct from the stressed vowel letters (Avanesov does not).18 One might
use small caps [i i …] or, as here, (modi¬ed) Greek letters: [¬ ï √].19 No sources
distinguish between ¬rst and second degrees of reduction among high vowels.
In non-allegro style, the rounding of {u} is preserved in unstressed [√] (gen sg
gen« [p√t˛í] ˜journey™), and the backing of {i} is still audible in unstressed [ï] (fem
pst ,sk’ [bïl’] ˜she was™).

Series {e a (o)} after soft consonants: After palatalized consonants, series {e}
and {a} fall together. Until the beginning of the twentieth century, the resulting
unstressed vowel was pronounced with ekan e, that is, as a mid vowel or an
upper mid vowel with [e]-coloring, transcribed [bt ] in Cyrillic, [µ ] in Latin. In
the twentieth century, the vowel has merged with the slightly reduced vowel
of series {í}: thus the ¬rst-pretonic vowels of ,bk´n ˜ticket™ [b˛ ¬l˛et] and [b˛ ¬l˛e„t]˛
5„ 55
,tk´nm ˜become white™ are now identical. This complete merger of vowels from
the non-high series {e a} with {i} is termed ikan e.
Ikan e begins to be acknowledged as an acceptable pronunciation around the
transition from the nineteenth to the beginning of the twentieth century. In
1912 the Leningrad phonetician L. V. Shcherba (1880--1944) described a gener-
ational split: his mother distinguished fem pst vtk’ ˜she swept™ from pv fem
vbk’ ˜pleasant™, presumably [m˛¬µ l’] vs. [m˛¬l’], while he merged them, presum-
ably [m˛¬l’]. At the same time, R. Koˇuti´ (1919:39) recommended ekan e, but
conceded that “all the young people” use ikan e. Ekan e was still the pronunci-
ation that R. I. Avanesov (1972:66) recommended as recently as the last half of
the twentieth century. However, sources after Avanesov treat ekan e as conserva-
tive and outmoded, and assume that there is no longer any distinction among
vowels in the position after palatalized consonants.

17 Bondarko 1977:156.
Now SRIa 1 uses [bэ ] for unstressed {i} and {e a} after soft consonants.
19 Also approximately as in Jones and Ward 1969.
Sounds 45

If one posits {o} as the series vowel where — e changed to — o under stress -- for
example, if {o} is said to be the vowel not only in y=c [n˛ os] ˜he carried™ but also
in ytck’ [n˛ ¬sl’] ˜she carried™ -- then one could say that series {o} is merged with
series {a} and {e} and ultimately series {i} after soft consonants.

Series {a o} after hard consonants: Unstressed vowels belonging to series {a} or
{o} -- that is, unstressed vowels spelled with the hard-vowel letters ¤f≥ or ¤j≥
that would be pronounced as [’] or [o] if they were stressed -- merge with each

other. Under ¬rst degree of reduction (¬rst pretonic position, position not after
consonant), the unstressed vowel is pronounced as a central, non-high, moder-
ately open vowel, written as [ ]:20 lfdyj [d vno] ˜long ago™, ljk;yj [d lznó] ˜must™,
„ ‹
´ ´
msc gen sg jlyjuj [ dn vo] ˜one™, ghbjndjh«nm [pr˛¬ tv r˛ít˛] ˜open somewhat™.

Under second degree of reduction, the unstressed vowel is [´], a vowel shorter
and less open than [ ]: second pretonic yfuhe;’nm [n´gr zat˛] ˜burden™, gjlhfcn’nm
‹5 „
[p´dr st’t˛] ˜nurture™; post-tonic v’vjxrf [m’m´c˛ ‹k´] ˜mommy™, j ´,kfxrj [obl´c˛‹k´]

˜cloud™, dtl=hjxrj [v˛¬d˛or´c˛‹k´] ˜bucket™. The merger of {o} with {a}, and the pro-
nunciation of the resulting vowel as an unrounded central vowel, is termed
akan e.

Series {e a o} after So (=[ˇ z]): For historical reasons, non-high vowels after ˇo
ˇ sˇ S
— e was
have unusual behavior. During the time when [s ‹ z] were still soft, original

˛‹ ˛‹
raised to [Se ], later [S¬ ], as it was after any soft consonant. When these conso-
nants lost palatalization, the vowel was backed to [S‹o µï ], later [S‹o ïµ ]. In the twen-
tieth century, the vowel has merged with [ï] from series {i}: ;tk†pysq [zïl˛†znïi] ‹5 ü
˜iron™, ;bk†w [zïl˛†c] ˜lodger™.21 The same vowel is pronounced for {e} in borrow-
ings after mutable consonants if they remain hard: vjltk«hjdfnm [m´dïl˛ír´v´t˛]
˜model™ (cf. vjl†km [m del˛] ˜model™).
For {a o}, there are two possibilities: an inherited pronunciation [ï] or a newer
pronunciation [ ]. How these two variants are distributed is complex (Table 2.3;
Under second degree of reduction after ˇo , vowels from the non-high series
{e a o} are pronounced as a central vowel [ï]: ;tktp’ [zïl˛¬z’] ˜gland™, bp ifkfi’

[¬s ‹ ïl s’] ˜out of the cabin™, itkrjd«wf [sïlkov˛íc´] ˜mulberry™.
‹ ‹
In absolute initial position the vowel spelled ¤э≥ in foreign borrowings is
raised though not backed (there is no preceding hard consonant), and is merged
with [¬]: эn’; [¬t’s]‹ ˜storey™, эrh’y [¬kr’n] ˜screen™, identical to buh’ [¬gr’] ˜game™.22

20 The vowel is glossed as raised and backed [a] by Jones and Ward (1969).
In Avanesov™s conservative norm, [zïµ l˛], not quite identical to [zïl˛].
‹ ‹
22 According to SRIa 1.103--4.
46 A Reference Grammar of Russian

Table 2.3 Vowel series

Co So (=[ˇ ˇ])
series sz C

{u} [] [] []
{i} [ï] [ï] [¬]
[ï]∞ [ï]∞ [¬]∞

´ ´
borrowing alternation with [V] no alternation with [V]
[ø] ([´]† ) [ï]∞ [¬]∞
{o} [ø] n.a.
[ø] ([´]† ) [ï]∞ [¬]∞
{f} [ø]Co ∼ [ï]C
¸ [ø]

x† = second-degree reduction
x∞ = Avanesov™s conservative norm = [¬µ ] ∼ [ïµ ]

The relations of stressed and unstressed members of vowel series are schema-
tized in Table 2.3 in three contexts.
As shown in Table 2.3, there are more distinctions of vowels under stress --
¬ve -- than among unstressed vowels. In the contemporary norm, three vowels
are distinguished after hard consonants, two after soft. (In the conservative style
of Avanesov, four distinctions are made after hard consonants, three after soft.)
Further reduction and merger is possible under second degree of reduction
in some varieties of speech. The troublesome question is whether unstressed
[ï ] (using breve here to mark signi¬cant shortening of an stressed vowel) is so
reduced that it merges with [´] -- whether the unstressed vowels of dat pl lj ´,hsv
and msc=nt loc sg lj ´,hjv are pronounced the same. Panov (1990) decides that
merger has long been a constant possibility in a less-than-standard, allegro style,
but has not achieved normative status. Also, in an extreme version of allegro
style, series {u} may lose its labialization and merge after soft consonants with
[¬ ] and after hard consonants with [ï ], which in this style will be identical to
[´]. At this point, only two unstressed vowel phones would be left under second
degree of reduction: [¬ ] vs. [ï ] ≈ [´]. The two vowel phones would be distributed
* *
complementarily, [¬ ] after palatalized consonants, [ï ] ≈ [´] after hard. This allegro
* *
system is not normative, in Panov™s view, but it is widespread.23

2.2.5 Unpaired consonants [ˇ ˇ c] and unstressed vocalism
As noted, [s ‹ z], which are always hard and therefore immutable and unpaired,

affect unstressed vowels in a manner different from that of ordinary mutable
hard consonants.24
As mentioned, a vowel from series {e} becomes [ï] after [s ‹ z]: ;†vxeu ˜pearl™,

;tvx©;ysq [zï]. In similar fashion, for vowels that alternate with stressed [o] „

and could be identi¬ed as series {o}, only [ï] is used after [s ‹ z]: nom pl ;=ys,

23 24
As Comrie, Stone, and Polinsky 1996 treat this merger. Kasatkin 1989.
Sounds 47

nom sg ;ty’ [zïn’] ˜wife™, nom sg i=kr, gen sg itkr’ [sïlk’] ˜silk™ (Table 2.3).25
‹ ‹
Some recent borrowings have an unstressed vowel which, because it is spelled
as ¤j≥, might be identi¬ed as belonging to series {o}. In imitation of its foreign
source, this ¤j≥ can be pronounced with only partial reduction as a shortened
mid, labialized vowel [o], for example Ijg†y ˜Chopin™ [sop†n]. As such words are
* ‹*
assimilated, this ¤j≥ is reduced to [ ] in ¬rst pretonic position: ;jyuk=h ˜juggler™
[z‹ ngl˛or], Ijg†y [s ‹ p†n], ija=h ˜chauffeur™ [s ‹ f˛jr] (Table 2.3). This pronunciation
5„ 5´
is what might be expected given the pronunciation of unstressed {o} after paired
hard consonants: djl’ [v d’] ˜water™.
Vowels of series {a} show variation between two variants, [ ] and [ï]. The older
pronunciation was [S‹o µï ], later [S‹o ïµ ], now [ï]. Throughout the nineteenth cen-
tury up until the beginning of the twentieth century, [S‹o µï ], later [Sïµ ], was used

in native words. Both variants occurred in borrowings, with a stylistic differ-
ence: [ ], which was closer to the pronunciation of the (often French) sources,
was a mark of “spoken language of good society,”26 in contradistinction to the
pronunciation that ¬t the native Russian pattern, with [S‹o µï ]/[S‹o ïµ ]. In the twenti-
eth century, sociolinguistic investigations document that there is variation and
change, but the change is not uniform; individual lexemes are regularizing us-
age, but not all lexemes are regularizing in the same direction.27 Native words
in which the unstressed vowel does not alternate with [’] have kept [ï]: h;fyj ´q
˜rye™ [rzïnoi], gen pl kjifl†q ˜horses™ [l´sïd˛ e„i]. Native words in which the pre-
‹ 5ü ‹ 55 8

tonic vowel alternates with stressed [’] are generalizing [ ]: gen sg ;fh’ (nom
sg ;’h) ˜heat™, gen sg dj;fr’ (nom sg dj;’r) ˜guide™, nom pl ifu« (nom sg
i’u) ˜step™, 3sg e;fcy=ncz ˜becomes horri¬ed™ (adj e;’cysq).28 In borrowings,
the vowel depends on the following consonant: [ï] is kept if the following conso-
nant is (or used to be) palatalized: ;fr†n ˜jacket™ [zïk˛ et], ;fcv«y ˜jasmine™. In con-
‹ 5„
trast, [ ] is being generalized in words in which the following consonant is hard:
ifn†y ˜auburn-haired person™ [s ‹ t†n], if,kj ˜clich†™, ifk’i ˜cabin™, ifvg’ycrjt

As is not surprising, since stressed [o]„ after [s ‹ z] derives from etymological — e. Here is a place where

the notion of series is revealed as something of a ¬ction. In this context, there is no evidence that
the unstressed vowel ever actually became [o]. The unstressed value here is [ï] because it remained
— e, and had the same fate as other unstressed — e after [s z].
26 See Panov 1990:260ff. Grech (1827) asserted that it was appropriate, in the “spoken language
of good society,” to say ifvgfycrjt (that is, [sø]) rather than ibvgfycrjt (a vowel of the type

[µï ], subsequently [ïµ ], now [ï]). At the turn of the twentieth century, Koˇuti´ (1919) gives two
pronunciations for borrowings: [ø] (literary) and [ï] (non-literary). Interestingly, he gives only vowels
similar to [ï] in native words in which the relevant vowel alternates with stressed [’]. Thus these
two sources suggest that [ø] has long been used in borrowings.
27 Krysin 1974.
28 :fk†nm ˜pity™, with [ï], is exceptional in this regard if it is related to ;’km, ;’krj ˜feel sorry for™,
but the derivational connection is tenuous (and the following [l˛] favors [ï]).
29 Panov 1968 puts the burden on alternation, Krysin 1974:105 on the following consonant. Evidently
both are relevant.
48 A Reference Grammar of Russian

Evidently, the use of [ ] in borrowings prepared the way for using [ ] in na-
tive words in which there is alternation with stressed [á], when the unstressed
vowel is still associated with [’], and this has become normative. As noted, [ï] is
maintained in native words when the unstressed vowel does not alternate with
stressed [’]. In borrowings, both [ ] and [ï] occur, distributed according to the
following consonant. A following palatalized consonant tipped the balance in
favor of the raised variant [ï]. Thus far with series {o}, only borrowings use a
low unstressed vowel [ ]. Words in which the unstressed vowel alternates with
stressed [o„] do not use [ ]. This is an important difference between {a} and {o},
re¬‚ected in Table 2.3.
The sound [c] is, like [s ‹ z], an unpaired immutable hard consonant, but it

hardly occurs before series {o} or {a}. A visible exception is the root w’hm ˜tsar™,
in which {a} under stress alternates with [ ] under ¬rst degree of reduction:
gen sg wfhz wfh«wf ˜tsar™s wife™.

2.2.6 Post-tonic soft vocalism
In general, unstressed vowels associated with series {a o e} are pronounced as a
high front vowel [¬] after any soft consonant. For this reason, one might expect
to ¬nd [¬] in place of post-tonic vowels in grammatical endings as well. It is
regularly stated, however, that this vowel can, depending on the morpheme, be
pronounced as [´]. Grammatical morphemes differ, and there is some change --
and some disagreement among authorities. Table 2.4 lists contexts of nouns,
organized by the vowel that appears when the given morpheme is stressed.30
There is a gradation of possibilities, from regular [´] to regular [¬]. One phono-
logical condition overrides other considerations. A following soft consonant
evokes [¬], as in: Context 10 (Declension<II> ins sg lth†dytq ˜village™ [n˛ ¬8]) and
Context 12 (Declension<III> {-†j} gen pl gtx’ktq ˜sorrows™ [p¬c˛‹a„l˛¬i]). Also, [¬] has
55 8
become usual since the beginning of the twentieth century in Context 6 (ins pl
lth†dyzvb: previously [n˛ ´m˛¬], now [n˛ ¬m˛¬]).
Beyond this syntagmatic phonetic condition, the choice between [´] and [¬] de-
pends on a paradigmatic condition -- on the vowel phones that occur in the given
morpheme in other words. At one extreme, [´] is used consistently in Contexts
1--5, for example, nom sg lth†dyz ˜village™ [n˛ ´]. The vowel of these morphemes
would be [’] under stress (nom sg ujkjd’ ˜head™, ptvkz ˜land™) and [´] after hard

30 See Koˇuti´ 1919 (on Old Muscovite), Avanesov 1972:69--71, Kuz mina 1966, Panov 1968:42--56. In
summarizing Old Muscovite usage, Kuz mina claims that the adjectival endings had exclusively
[¬] (1966:7), relying on Koˇuti´™s characterization of [¬] as literary, [´] as uneducated (1919:100). But
Koˇutic „ (1950:80) transcribes gen sg cbytuj ˜blue™ as [c by f df ], exactly parallel to gen sg lj,hjuj
˜good™ [lj,hf df ]. Presumably Panov would posit [¬] in gen pl {-óv}, ins sg {-ój}, and gen pl {-†j},
contexts with closed syllables, which implies [¬] for Panov.
Sounds 49

Table 2.4 Post-tonic vowel reduction, nominal morphology

Old Kuz mina,
Context Muscovite Avanesov Panov Krysin
[CV] ∼[Co ´] ∼ [’]
1 nom=acc pl<Ib> {-’} ´ ´ ´
2 nom sg<II> {-’} ´ ´ ´
3 gen sg<I> {-’} ´ ´ ´
4 dat pl {-’m} ´ ´ ´ (∼ ¬) ´ (90%)
5 loc pl {-’x} ´ ´ ´ (∼ ¬)
6 ins pl {-’mi}
˛ ´ ´ ¬ ´ ¬
[CV] ∼ [Co ´] ∼ [ó]
7 nom=acc sg<Ib> {-ó} ´>¬ ´ (63%) ∼ ¬
´ (∼¬) ´
8 ins sg<I> {-óm} ´>¬
´ ´ ¬ ´ (81--83%)
9 gen pl<Ia> {-óv} ´ ´ ¬
10 ins sg<II> {-ój} ´ ¬
[CV] ∼ [†]
11 [dat-] loc sg<[II]I/II> {-†} ¬ ¬ ¬
12 gen pl {-†j} ¬ ¬

[V] ∼ [V] = unstressed vowel in alternation with stressed vowel
* ´
x ∼ y = x occurs in variation with y
y = x has yielded to y
x > y = x is yielding to y
ins sg<ii> = case-number form of Declension<ii>
x† = presumed, not explicitly stated

consonants (nom sg ,’,f ˜old woman™ [b’b´]). At the opposite extreme, the loca-
tive singular of Declension<Ia> and Declension<II> (also the dative singular of
Declension<II> ) is [†] under stress. The vowel of this morpheme does not occur
after hard mutable consonants, and accordingly there is no alternation with
[Co ´]. Unstressed, this vowel has the variant [¬] after C (and [ï] after immutable
S ‹ : yf gkz ´;t ˜on the beach™ [zï]).

The contexts of greatest interest are those whose vowels alternate between
[o] under stress and unstressed [´] after hard consonants: Context 7 nom=acc

sg vj ˜sea™ -- recall jryj [ knj ˜window™, gbnm= ˜drinking™ [p¬t˛jo], cn’lj [st’d´]
´ht ´ ´]
˜herd™ -- and Context 8 ins sg vj ´htv -- recall jryj [ knj ˜window™, gbnm=v
´v ´m]
[p¬t˛jö´ m], cn’ljv [st’d´m]. (Context 9, for example, gen pl ,h’nmtd ˜brothers™, be-
longs here as well.) In these contexts, the Old Muscovite style at the beginning
of the twentieth century had [´] after soft consonants.
With respect to usage after the middle of the twentieth century, there is
disagreement among commentators. Avanesov (1972:70), recalling that [´] was
the Old Muscovite norm, concedes that in the nominative-accusative singular
50 A Reference Grammar of Russian

[¬] has become possible (“widely known”) and that in the instrumental singular
[¬] has even replaced [´] (the latter “must be considered moribund”).31
Avanesov™s view contrasts with that of Kuz mina (1966) and Panov (1968), who
report on a questionnaire administered during the 1960s to 100 students of the
cohort 1940--49. In that questionnaire, 98 percent of the respondents had [¬]
in ins sg r’vytv ˜stone™ and, surprisingly, 98 percent had [´] in nom sg gj ´kt
˜¬eld™. Their results seem quite unambiguous in these two contexts; they are
dramatically less ambiguous than in other words in which the vowel is usually
thought to be pronounced as [´] without variation: loc pl lßyz[ ˜melons™ (74%
[´]) or dat pl lßyzv (only 52% [´]).32
A third view derives from the mass survey in the 1960s (Krysin 1974), according
to which [´] was reported to be basically stable, or slightly increasing, in both
contexts. In that survey, the use of [´] in nom=acc sg gj rose from just above
50 percent for the oldest generation to above 60 percent in the ¬nal cohort of
1940--49, and [´] in ins sg vtld†ltv ˜bear™ and ins sg gk’xtv ˜cry™ basically held
constant at 80 percent over the six decades of the survey.33
To summarize about the two contexts, nom=acc sg vj and ins sg vj
´ht ´htv:
Avanesov believed that both were developing towards [¬]; Kuz mina and Panov
found that they were moving in opposite directions; Krysin™s survey suggest that
both contexts were developing in the same direction, towards [´].
It is dif¬cult to resolve the discrepancy among these sources. A pilot instru-
mental investigation carried out for this study (six speakers) did not yield un-
ambiguous results. There was no consistent difference between loc sg vj ´ht, in
which only [¬] is expected, and nom=acc sg vj ´ht, in which variation between
[´] and [¬] is expected. The one reasonably clear result was that the vowel of ins
´htv had a higher F1 and lower F2 than other vowels in nouns, implying a
sg vj
more [´]-like pronunciation, evidently in anticipation of the following [m]. From
this limited investigation, it was not clear to what extent a categorial distinction
between [´] and [¬] remains in these morphemes.

31 A point of notation: Avanesov (1972) uses three symbols: [(], a low back vowel after hard conso-
nants; [m] is unstressed {i}; and Avanesov™s [´] is the front vowel occurring after soft consonants
for series {e a o}. Other sources (Panov) collapse Avanesov™s two front vowels [m] and [´] to [m], and
Avanesov himself abandons [´] in favor of [m] in his transcribed texts (p. 356: lh†vktim has [m],
not [´]) and in the summary of phonetic variants (pp. 311--14). In Table 2.4, Avanesov™s [m] and his
[´] are both written as [¬], [(] as [´].
32 Kuz mina 1966, Panov 1968:47--48. Comrie, Stone, and Polinsky (1996:56--59), after deliberation,
side with Kuz mina. Panov™s position cannot be separated from his strong belief that the distribu-
tion of [´] vs. [¬] must be determined by phonetic factors: for underlying <o>, [´] is said to occur
only in ¬nal open syllables (therefore in nom=acc sg gjkt) while [¬] is said to occur elsewhere
(therefore in ins sg gk’xtv).
33 Krysin 1974: 114, Fig. 24.
Sounds 51

Table 2.4 above focused on endings in nouns, the richest set of morphemes
in which variation in post-tonic vocalism can be observed. In addition, soft-stem
adjectives generally have [¬] (gen sg msc=nt gh†;ytuj, dat sg msc=nt gh†;ytve,
loc sg msc=nt gh†;ytv ˜former™).34 The ¬nal vowel of singular adjectives after
[j] is [´] (nom nt sg cn’hjt ˜old™, nom fem sg cn’hfz [´i´]) but that of the plural
normally [¬] (nom pl cn’hst [ïi¬]). Present adverbial participles (lttghbxfcnbz)
have [´] (e.g., gj´vyz ˜remembering™), a pronunciation that is consistent with [’] in
lexemes with ¬nal stress (ytcz ˜carrying™). Re¬‚exive present adverbial participles
still allow [¬], under the in¬‚uence of the following soft consonant (ghbjc’yzcm
˜putting on airs™ is Panov™s example). Individual forms such as psv nom msc sg
p’yzn ˜occupied™, msc sg pst p’yzk once had only [¬] but now prefer [´].35 The
thematic ligature that marks the present tense in verbs of e-Conjugation is [o] „
(ytc=im ˜you carry™, etc.) under stress. Unstressed, the ligature in the middle
forms of the present tense is only [¬] (2sg k†ptim [l˛e„z˛ ¬s], 3sg k†ptn, 1pl k†ptv
55 ‹
˜climb™). The third plural is [´] in verbs of the i-Conjugation (gk’nzn ˜they pay™
[pl’ ´t]).
The Old Muscovite [´] in nom=acc sg vj and ins sg vj ´htv is unexpected
on phonetic grounds -- after a soft consonant in positions of reduction, original
non-high vowels have generally become [¬]. The origin of the Old Muscovite pro-
nunciation has been disputed. Most likely, it is analogy, at the level of phonetics.
The [¬] that might be expected after soft consonants was suppressed, or never
developed, in deference to the [´] that occurs after hard consonants in the same
morphemes: nom sg lth†dyz [C´] imitates nom sg ,’,f [Co ´], ins sg vtld†ltv
´krjv ˜wolf™ [Co ´m]. The fact that [´] can participate in
[C´m] imitates ins sg dj
analogy shows that unstressed vowels have some psychological independence.
The expected development to [¬] did take place in those morphemes whose vowel
would not be found unstressed after paired hard consonants (only [¬] in loc sg
´ht, 2sg prs k†ptim, etc.).
yf vj

2.2.7 Unstressed vowels in sequence
When {a} or {o} follows another vowel, it does not have to share duration with
a preceding consonant in its syllable, and it is less reduced, even two or more
syllables from the stress: e jujhj ˜near the garden™ [√ g rj ´d´], yt jnjck’k ˜did
not send off™ [n˛ ¬ t sl’l]. An {a} or {o} that is the first in a sequence of two vowels
far removed from stress will be fully reduced, as yfeu’l ˜by guesswork™ [n´√g’d],
yfbuh’nmcz ˜play much™ [n´¬gratts´], except when the ¬rst vowel is followed by

34 Avanesov 1972:71 implies there is a change from [´] to [¬]; Kuz mina 1966 mentions only [¬].
35 Panov (1968:49), who relies on Koˇuti´ 1919 and Chernyshev 1908.
52 A Reference Grammar of Russian

another [ ]. Then the ¬rst vowel assimilates to the second and does not re-
duce, as in pf jlbyj ´rbv ˜beyond isolated™ [zøød˛ ¬nok˛ ¬m], djjleitdk=y ˜inspired™
[vøød√sïvl˛on]. In allegro style the two [ ] coalesce and reduce: cjjnyji†ybt ˜inter-
‹ 5„
connection™ neutral [s tn s†en˛ ¬´], allegro [s´tnøsen˛ ¬´]. As the ¬rst of a sequence,
‹ 5„ ‹ 5„
{e} reduces normally to [¬]: ytj,[jl«vj ˜necessary™ [n˛ ¬ px d˛ ím´].

2.2.8 Unstressed vowels in borrowings
In foreign borrowings of high culture, unstressed mid vowels (the vowels written
as ¤t≥ and ¤j≥) do not necessarily reduce completely according to the rules that
apply to native words. They can instead maintain something of the pronunci-
ation of the source language and, though they are shorter, they do not merge
with series {a} or {i} according to the usual pattern: ktu’nj ˜legato™ [leg’to],
utyjw«l ˜genocide™ [genocit]. As words are assimilated, the semi-reduced foreign
pronunciation of ¤j≥ as [o] yields to [ ] ([´]), as in native words. Thus, in certain
frequently used borrowings, the usage of [o] for ¤j≥ declined quite dramatically
from the oldest cohort (1890--99) to the youngest (1940--49): rjyuh†cc ˜congress™
(63% > 27%), gjhna†km ˜notebook™ (62% > 20%), ghjw†cc ˜process™ (76% > 32%),
cjy†n ˜sonnet™ (78% > 41%), hjz ˜piano™ (51% > 25%).37

2.3 Consonants

2.3.1 Classification of consonants
The quintessential consonants are obstruents (= C), segments that involve ob-

struction or serious narrowing in the long path from the larynx to the lips.
Obstruents are listed above the internal line in Table 2.5. In addition to obstru-
ents, consonants include sonorants, a group of sounds that are heterogeneous
but share the negative property of being neither obstruents nor vowels. Sono-
rants are listed below the line in Table 2.5.
Consonants are de¬ned by a complex of articulatory activities. Consonants can
be voiced (the membranes of the vocal cords are taut and vibrate) or voiceless
(the membranes are open and relaxed, allowing air to pass without vibration).
Obstruents can be produced with different trajectories of gestures, or manners
of articulation. Obstruents can be stops, sounds that involve a sudden gesture
of complete closure (for example, the complete closure of both lips to make a
[b]), a short interval of stasis, and a sudden release. Or they can be fricatives,
which involve a more gradual restriction of the air¬‚ow without complete clo-
sure followed by an interval of incomplete closure and then a more gradual
release. Affricates are intermediate; they are produced by a stop closure and a

36 37
Avanesov 1972:174, 167--68. Glovinskaia 1976, speci¬cally Table 12.
Sounds 53

Table 2.5 Russian consonant phones

bilabial labio-dental dental (alveo-)palatal velar

k˛ †
voiceless stop pp t t˛ k
g˛ †
voiced stop b b˛ d d˛ g
voiceless affricate c c˛
Z∞ Z∞
voiced affricate *˛
s˛ ‡ x˛ †
voiceless fricative f f˛ s s˛ s * x

voiced fricative v v˛ z z˛ z z˛

glide j
nasal stop mm˛ n n˛
lateral l l˛
trill r r˛

x† = restricted distribution: / {i e}, rarely / {f j b}
x∞ = restricted distribution: / C 3/

x = normally long (/V V), shortened adjacent to C

brief interval of closure followed by a more gradual release similar to that of a
Obstruents are also de¬ned by the place of articulation, the place in the vo-
cal tract where the obstruction occurs and, correlatively, the mobile organ used
to make the restriction. One ingredient of the place of articulation is palatal-
ization. How consonants are palatalized depends on where the consonant is
articulated, but there is a basic similarity.
The matrix of obstruents in Table 2.5 is organized by place of articulation
along the top, with non-palatalized consonants listed to the left of palatalized.
Bilabial stops are produced by closing the lips together: [b], [p]. The closest frica-
tives [f f˛] and [v v˛ ] are not pure bilabials, but labio-dentals, formed by moving
the lower lip up under and close to the upper teeth, constricting the air¬‚ow.
However, with respect to voicing rules, [v v˛ ] do not quite act like well-mannered
obstruents, and can be designated as a distinct class of sounds “W” that is inter-
mediate between obstruents and sonorants (§2.3.9). When a labial or labio-dental
consonant is palatalized, at approximately the same time the primary closure
(or restriction) is made with the lips (or upper lip and lower teeth), the blade of
the tongue is arched up and raised towards the hard palate (see [p], broken line
in Fig. 2.10). In non-palatalized labials, the tongue is in a neutral position (see
[p], solid line, in Fig. 2.11).
Russian has a class of dental sounds whose obstruction is made in the region
of the upper teeth. As the dental stops [t d] or the affricate [c] are produced,
the tip and blade of the tongue touch against the upper teeth. The body of
54 A Reference Grammar of Russian

[p] =
[p] =
Fig. 2.10 [p], [p]. From Avanesov 1972: ¬g. 8

the tongue is ¬‚at or even depressed, which is to say that hard dental stops are
slightly velarized (see Fig. 2.11, solid line).38 In producing palatalized dental stops
(broken line in Fig. 2.11), the tongue makes contact all the way from the upper
teeth through the alveolar ridge and along the hard palate. Whereas with labials
palatalization is a somewhat independent gesture, with dentals, palatalization
is part and parcel of the articulatory gesture. For some speakers, the palatalized
dental stops [t˛ d˛ ] have begun to develop a touch of frication in their release,
especially before [í]: n«g ˜type™ [t˛s˛ íh], l«rj ˜wildly™ [d˛ z˛ ík´].39
The dental fricatives [s z] are pronounced with the tip of the tongue pointing
towards the top of the upper teeth, leaving an aperture through with which air
¬‚ows turbulently. The hard dental fricatives [s z] are noticeably velarized. The
palatalized dental fricatives [s˛ z˛ ] are made with the front of the tongue making
an arch that follows the shape of the teeth and hard palate, with the narrowest
aperture at the teeth.
Russian has a group of sounds classed together as having an alveo-palatal (or
sometimes simply palatal) place of articulation. In the hard fricatives [s ‹ z] -- the

sounds spelled by Cyrillic ¤i≥ and ¤;≥ -- the tip of the tongue approaches the
alveolar ridge, higher than is the case with [s z]. In addition, [s ‹ z] lift the sides

of the tongue and force air through a groove, while [s z] have a narrow horizon-
tal slit. These (alveo-)palatal fricatives are strongly velarized: the middle of the
tongue is depressed and the back of the tongue is arched upwards (solid line,
Fig. 2.12).40 The sounds [s ‹ z] are pronounced as hard, even when (in borrowings)

the following vowel letter is ¤/≥: ,hji·hf ˜brochure™ [su„r], gfhfi·n ˜parachute™

[sut], though sometimes ;/h« ˜jury™ [z˛‹ r˛í]. As a new (hypercorrect?) tendency,
38 39 Matusevich 1976:183.
Velarization is evident in the sketch of SRIa 1.43.
40 Avanesov 1972:40, Fig. 14; see also Matusevich 1976:182.
Sounds 55

[t] =
[t] =

Fig. 2.11 [t], [t˛]. From Avanesov 1972: ¬g. 11 Fig. 2.12 [ˇ], [ˇ˛]. From Avanesov 1972:
¬g. 14

[s ‹ z] can soften under assimilation: dxth’iybq ˜yesterday™s™ normative [sn˛], new
‹ ‹

Russian has two other alveo-palatal fricatives, [s˛ ‹ ] (the sound associated with
Cyrillic ¤o≥) and [z˛‹ ] (an older pronunciation of Cyrillic spellings ¤;;≥ or ¤p;≥
in certain words such as lhj ´;;b ˜yeast™, †p;e ˜I drive™, gj
´p;t ˜later™). These sounds
are palatalized; the tip of the tongue is pointed towards the teeth, and the blade
of the tongue curves up along the hard palate (broken line, Fig. 2.12). The alveo-
palatal affricate [ˇ˛], spelled ¤x≥, is likewise “soft” -- there is no corresponding
hard — [ˇ]. In its initial closure phase, it involves essentially the same tongue
position as [t˛]; contact is made from the alveolar ridge along the hard palate. In
its release, [c‹˛] is similar to the soft alveo-palatal fricative [s˛ ‹ ].
In the production of velars [k g x], the tongue approaches or touches the roof
of the mouth, in the region where the hard palate and soft palate meet (solid
line, Fig. 2.13). The voiced fricative [ ] is quite restricted, occurring only before
a following voiced obstruent: nh=[ly†dysq [ d˛ n˛]. Palatalized velars [k˛ g˛ x˛ ] have
basically the same tongue con¬guration as non-palatalized velars. They differ
from non-palatalized velars in that the tongue makes contact (or restriction)
further to the front of the mouth (broken line, Fig. 2.13).
Sonorants, listed below the center line of Table 2.5, are a motley group. Nasal
stops [m m˛ n n˛ ] have a complete closure in the oral cavity like that of a stop --
the place of the closure is bilabial for [m m˛], dental for [n n˛ ] -- but, in addition,
they simultaneously open the velum, allowing air to ¬‚ow into the nasal cavity
and resonate.
41 Kasatkin 2001:86.
56 A Reference Grammar of Russian

[k] =
Fig. 2.13 [k], [k˛]. From Avanesov 1972:
[k] = ¬g. 20

The approximate [j] is articulated with a tongue position like that of the vowel
[i], so that the blade of the tongue raises close to the hard palate behind the
alveolar ridge; [j] differs from [i] in that it is not the peak of syllables and involves
greater narrowing of the tongue to the front of the roof of the mouth. Given its
tongue shape, [j] is intrinsically soft.
The trills [r r˛] are made by one or more taps in the dental region. With the later-
als [l l˛], the blade of the tongue makes complete closure in the dental region but
the sides of the tongue are raised, allowing air to pass laterally (hence the term)
along its sides. Together the [r]-sounds and the [l]-sounds are liquids. Hard [r] and
especially hard [l] are velarized: the middle portion of the tongue is depressed
and the back of the tongue body is raised towards the back of the palate.
Collectively, the nasals, liquids (trills and laterals), and the glide [j] can be
grouped together as sonorants (in notation, the set “R”), a loose class of sounds
that are neither vowels nor obstruents. Sonorants can distinguish palatalization,
in this respect like obstruents. Unlike obstruents, sonorants lack a distinction of
voicing; like vowels, they are normally voiced, and do not cause preceding ob-
struents to become voiced (§2.3.9). Between an obstruent and another obstruent
or word end (the contexts C RC or C R#), sonorants can become syllabic: MXATf
// /

˜from MKhAT™ [ t ´ m x ’t´], jrnz ´,hm ˜October™, [økt˛a„b(™)r˛], h©,km ˜ruble™ [r´ b(™)l˛],
55 u
;«pym ˜life™ [z˝z˛ (™)n˛ ].42
‹5 „

2.3.2 Palatalization of consonants
Most consonants -- sonorants as well as obstruents -- can be palatalized or not.
That is, for almost every consonantal articulation -- for almost every combination
of place of articulation, manner of articulation, voicing and nasality -- there is
one sound that is not palatalized and another that is pronounced with similar
42 “I pronounce the word ;bpym as two syllables, with a ˜¬‚eeting™ *” (Trubetzkoy 1975:238).
Sounds 57

gestures but is palatalized. For example, both a palatalized voiced labial stop [b˛ ]
and a non-palatalized [b] occur, and both a palatalized voiceless dental fricative
[s˛ ] and a non-palatalized [s] occur. Palatalization is similar but not identical
for sounds of different places of articulation. Though there are these minor
differences, all palatalized consonants in¬‚uence vowels in the same way.
When a given articulation occurs in both palatalized and non-palatalized
forms, that articulation can be said to be paired, or mutable, for palataliza-
tion. Thus [b] and [b˛ ] are phonetically paired, or mutable. Most consonants are
mutable. Labials and dentals obviously are. Velars are as well, although the
palatalized forms of velars [k˛ g˛ x˛ ] are more restricted than palatalized labials or
dentals; they do not occur in all phonological contexts, and they rarely if ever
distinguish words in opposition to [k g x].
Some consonants are not mutable: the glide [j] (necessarily palatalized); the
hard affricate [c]; the soft affricate [c˛ ‹]; the hard fricatives [s]‹ (Cyrillic ¤i≥) and
[z] (Cyrillic ¤;≥). Although the alveo-palatal fricatives [s˛ ‹ z˛‹ ] are palatalized, they

are not paired with [s ‹ z] in this sense, since [s ‹ z] do not become palatalized at
‹ ‹
the end of noun stems in the locative singular (j lei† ˜about the soul™ has [s], ‹
— [s] or — [s˛ ]) nor in the conjugation of verbs (g«itim has [s], not — [s˛ ]).
not ‹ ‹ ‹
Accordingly, four groups of consonants can be distinguished:

[2] hard, immutable: [s ‹ z ‹ c]
[j c˛‹ s ˛ ‹ z˛ ‹ ]
soft, immutable:
hard, mutable: [p t k x s z], etc.
soft, mutable: [p t˛ k˛ x˛ s˛ z˛ ], etc.

Among labials and dentals, both palatalized and non-palatalized variants oc-
cur before vowels and after vowels in word-¬nal position. In both contexts,
palatalization can distinguish words. Compare: prevocalic nj ´vysq ˜languid™ [t]
vs. n=vysq ˜dark™ [t˛], g’cnm ˜fall™ [p] vs. gzcnm ˜metacarpus™ [p]; and ¬nal post-
vocalic dßgbn ˜drunk down™ [t] vs. dßgbnm ˜to drink down™ [t˛], ujnj ˜ready™ [f] vs.
´dm! ˜prepare!™ [f˛]. Because contrasts occur in ¬nal position where no vowel
follows the consonant, palatalization (or its absence) must be intrinsic to the
consonant, and in a phonemic analysis, it is the consonant, palatalized or not,
that distinguishes words in Russian. If palatalization is distinctive for some con-
sonants in that position, it can be assumed to be distinctive in position before
a vowel. Thus the contrast of [t] in nj ´vysq ˜languid™ vs. [t˛] in n=vysq ˜dark™ is
usually analyzed as a contrast of two types of dental stops, non-palatalized [t] as
opposed to palatalized [t˛].43

43 In contrast to the abstract phonology of (for example) Lightner 1972, in which there is a rich set
of vowel distinctions and consonants are intrinsically hard, becoming palatalized in the position
before (underlying) front vowels.
58 A Reference Grammar of Russian

Palatalized and non-palatalized consonants occur with different degrees of
freedom depending on the context (the position in the word) and depending on
the consonant itself.
All mutable (phonetically paired) consonants historically were palatalized be-
fore {e} within lexemes. Palatalization therefore used not to be distinctive in
the position before {e}. This historical rule, which dates from the period when
palatalization ¬rst arose in Russian (a thousand years ago, in the period around
the fall of the jers), has been eroded in various ways. Consonants at the end of
pre¬xes are not palatalized before a root-initial {e} (cэrjyj ´vbnm ˜economize<pf> ™),
nor is the ¬nal consonant of a preposition palatalized before the {e} of the
demonstrative ¦njn ˜this™ (d ¦njv ˜in that™, gjl ¦nbv ˜under that™, c ¦nbv ˜with
that™, etc., with [ve], [de], [se], not [v˛ e], — [d˛ e], — [s˛ e]).
5„ 5„ 5„
Consonants remain non-palatalized before {e} in abbreviations, when that
{e} is word-initial in the base word from which it derives, as in YЭG ([nep], not
[*n˛ ep] -- from “yjdfz эrjyjvbxtcrfz gjkbnbrf”). In borrowings, non-palatalized
consonants occur before {e}, despite the rule that consonants were historically
palatalized before {e} (§2.3.3).44 Evidently, this primordial rule is no longer pro-
ductive in all contexts.

2.3.3 The distribution of palatalized consonants
Not all contexts allow both palatalized and non-palatalized consonants. Palatal-
ized consonants are more restricted in their distribution, but non-palatalized
consonants occur freely in almost all positions except preceding the vowel {e}.45
The distribution of palatalization is sensitive to the type of consonant involved.
Dentals distinguish palatalization before all vowels except {e}. Dentals are
even developing a distinction before {e} in borrowings, and are doing so more
readily than other consonants. Palatalized dentals can occur when no vowel
follows. Dental stops occur palatalized in ¬nal position after a dental fricative
(i†cnm ˜six™, udj
´plm ˜nail™ [s˛ t] vs. i†cn ˜pole™, lhj ˜thrush™ [st]). At the other end
˛ ´pl
of a word, a palatalized dental stop can occur in word-initial position dissimila-
tively before a non-dental (nmv’ ˜darkness™, nma© ˜phooey™). Word-internally not
before vowels, palatalized dental obstruents occur dissimilatively before velars
and labials, but not before other dentals or palatals: n’nm,f ˜thievery™, cd’lm,f
˜wedding™, n=nmrf ˜aunt™, G†nmrf ˜Pete™. Derivational suf¬xes that now begin with
a consonant, such as {-n}, once began with etymological — m, a high front vowel
which, a thousand years ago, palatalized the preceding consonant. Now con-
sonants are not palatalized before these suf¬xes: -obr (ajy’hobr but ajy’hm

44 45
Glovinskaia 1971, Alekseeva and Verbitskaia 1989. Glovinskaia 1976.
Sounds 59

˜lantern™), -xbr (k’hxbr ˜box™ but k’hm ˜chest™), -ybr (nhj ´cnybr ˜reed™ but nhj´cnm
˜cane™), -ysq (zyn’hysq but zyn’hm ˜amber™).
Palatalized dental sonorants [r˛ l˛ n˛ ], and especially [l˛], are distributed more
freely: word-¬nally after other consonants (vßckm ˜thought™, h©,km ˜ruble™,
cgtrn’rkm ˜spectacle™, ;eh’dkm ˜crane™, dj ´gkm ˜howl™, d«[hm ˜whirlwind™, ;«pym
˜life™), in comparatives (h’ymit ˜earlier™, nj ´ymit ˜thinner™, v†ymit ˜less™), and in
adjectives from months (jrnz ´,hmcrbq ˜of October™, b·ymcrbq ˜of June™, b·kmcrbq
˜of July™). The lateral [l˛] has the widest distribution: gjhneu’kmcrbq ˜Portuguese™,
djltd«kmxbr ˜vaudeville performer™.
Labials, before vowels other than {e}, can be either non-palatalized (g’cnm ˜fall™)
or palatalized (gz ´cnm ˜metacarpus™). Labials are not palatalized internally be-
fore suf¬xes that once conditioned palatalization: — rabmsk(jm > h’,crbq ˜servile™.
Labials distinguish palatalization in word-¬nal position after vowels: rj ˜mine™
vs. jrj ˜trench™, ujnj ˜ready™ vs. ujnj ˜make ready!™. They can even occur in
´g ´d ´dm
word-¬nal position after consonants, in [jh©udm ˜standard™, d†ndm ˜branch™. Final
palatalized labials in isolated grammatical forms were lost early in the history
of Russian (athematic 1sg prs — damm > lfv ˜I give™, ins sg — -Vmm > {-om}),46 and
there is a slight tendency to lose palatalization in labials at the end of words in
´ctvm ˜eight™ [m] ∼ [m].
other instances, for example, dj ˛
Velars [k g x] can be either palatalized or non-palatalized. For the most part,
the variants are distributed in complementary fashion: the palatalized variant
occurs before {i e}, the hard variant elsewhere -- before other vowels and in a
position not before a vowel. However, exceptions to this strict complementar-
ity have begun to appear. Palatalized velars occur before the [o] functioning as
the ligature in the second singular through second plural of the present tense
of velar-stem verbs, with varying stylistic values in different words. By now, [k˛ ]
is standard in forms of nr’nm ˜weave™ (2sg nr=im, etc.), while [g˛] was used by
about half of speakers (in the survey of the 1960s) in ;†xm ˜burn™ (3sg ;u=n for
standard ;;=n), and [k˛ ] by a quarter of speakers in g†xm ˜bake™ (2sg gtr=im); in
the last two the palatalized velar is not normative. To the extent that present
adverbial participles are permitted from velar-stem verbs (they are not univer-
sally accepted), the form has a palatalized velar (,thtuz ˜protecting™) by analogy
to other obstruent-stem verbs (ytcz ˜carrying™). Palatalized velars appear before
{a o u} in borrowings in the previous century: uz©h ˜giaour™, ,hfr=h ˜inspector™,
r/h† ˜cur†™, vfybr·h ˜manicure™. Palatalized velars do not occur in ¬nal, post-
vocalic position. Non-palatalized velars do not occur before {e i} in native words,
although a non-palatalized pronunciation is normal for the [k] of the preposi-
tion r before {i} and {e}, as in r buh† ˜to the game™ or r ¦njve ˜to that™ or for

46 Shakhmatov 1925.
60 A Reference Grammar of Russian

velars in compounds, as in lde[эn’;ysq ˜two-storied™ [x]. In this way, there is a
contrast of sorts between palatalized [k˛ ] internal to morphemes (r«yenm ˜toss™ >
[k˛ í]) and non-palatalized [k] in the prepositional phrase (r «yjre ˜to the monk™ >
[k˝!]). Thus velars are moving towards developing a contrast for palatalization.
In native words, all mutable hard consonants (all hard consonants except
[c s ‹ z]) are palatalized in the position before {e}. In borrowings, a non-palatalized

pronunciation is possible to a greater or lesser extent, depending on how well
assimilated the individual word is, the familiarity of a given speaker with for-
eign languages, and systemic properties. When the question was investigated in
the 1960s, it was found that in some words -- seemingly more ordinary, domes-
tic words -- the frequency of a hard pronunciation was increasing: h†qc ˜route™,
rjyc†hds ˜conserves™, rjyrh†nysq ˜concrete™, ,th†n ˜beret™, htp†hd ˜reserve™. With
other -- more scienti¬c -- words, the percentage of the population using palatal-
ized consonants decreased from the oldest to youngest cohort: fhn†hbz ˜artery™,
by†hwbz ˜inertia™, rhbn†hbq ˜criterion™, эy†hubz ˜energy™, ,frn†hbz ˜bacteria™. And
in a third group there is no clear direction of change: ghjuh†cc ˜progress™, gfn†yn
˜patent™.47 Hard consonants are more easily maintained in stressed than in un-
stressed position. Dentals most frequently allow hard consonants, then labials,
then velars. Yet a hard pronunciation does occur with labials and with velars:
,tvj ˜b-¬‚at™ [bemj l˛], v¦h ˜mayor™ [m†r], g¦h ˜peer™ [p†r], u†vvf ˜engraved stone™

[g†m´], r†vgbyu ∼ r¦vgbyu ˜camping™ [k†mp¬ng], [¦vvjr ˜hammock™ [x†m´k],
u†nnj ˜ghetto™ ([g†] ∼ [g†]).48
Overall, the possibility of having a contrast of palatalized and non-palatalized
consonants depends on a number of parameters. The possibility of a contrast
for palatalization depends on the place (and secondarily manner) of articula-
tion of the consonant itself, dentals favoring the distinction more than labials,
which in turn favor the distinction more than velars; yet velars at least have
positional variation for palatalization, thereby ranking them ahead of the im-
mutable consonants [s ‹ z‹ s˛‹ z˛‹ ], [c], and [j]. Having a contrast in palatalization also
depends on context. A contrast for palatalization is most likely before vowels
(/ V), less likely in a position after a vowel with no vowel following; within the
latter environment, palatalization is less likely than before a consonant (/V C)
than in word-¬nal position (/V #) -- perhaps because in most instances in which
a palatalized consonant would appear word-¬nally, the given form alternates
with another form in which a vowel follows (nom sg uj ´ke,m ˜dove™ [p], gen sg
´ke,z [b˛ ´]). Palatalized consonants are infrequent in contexts not adjacent to a
vowel, though they can occur (nmv’ ˜darkness™, ;«pym ˜life™, h©,km ˜ruble™, [jh©udm
˜standard™). Among vowels, a distinction is made more readily before back vowels

47 48
Glovinskaia 1976:100--10. Glovinskaia 1971:63.
Sounds 61

Table 2.6 Palatalization assimilation and place of articulation


. 2
( 17)