On the Constituent Structure of Augmented Plurals in Russian

Abstract

This article examines augmented plurals in Russian, mostly focusing on those in -ĭj- (e.g., pero/perʲja ‘feather.sg/pl’). The accentual behavior of -ĭj-plurals is sensitive to animacy: while inanimates show stem-final stress, animates appear with inflectional stress. This is explained by different constituent structures: while for inanimates, -ĭj- combines with the stem, animate stems require complex suffix formation so as to not create neuter animates, which are not tolerated in Russian. The position of the accent is then derived from the usual assumptions about Russian stress and the hypothesis that -ĭj- is accented but unaccentable. Other plural augments are also discussed.

1. Introduction: Russian Augmented Plurals

As observed by Zaliznjak (1967b), among others, Russian plural morphology, presented in

Table 1. Nominal declension classes (after Corbett (1982)).

number	case	a	ь	o	C
main gender:		f	f	n	m
sg	nom	čert-á	cérkovʲ	božestv-ó	stól/kónʲ
pl	nom	čert-ɨ́	cérkv-i	božestv-á	stol-ɨ́/kón-i
	acc	acc=gen for animates, acc=nom for inanimates
	gen	čért-Ø	cerkv-éj	božéstv-Ø	stol-óv/kon-éj
	dat	čert-ám	cerkvʲ-ám	božestv-ám	stol-ám/konʲ-ám
	loc	čert-áx	cerkvʲ-áx	božestv-áx	stol-áx/konʲ-áx
	ins	čert-ámi	cerkvʲ-ámi	božestv-ámi	stol-ámi/konʲ-ámi
		‘line’	‘church’	‘deity’	‘table’/’horse’

, is syncretic across declension classes in non-structural cases (dative, locative, instrumental) and primarily determined by the grammatical gender in the structural cases (for a generativist treatment of exceptions as gender change in the plural, see Matushansky (2025a)).1

Despite some evidence for decomposition in non-structural cases, plural exponents are often regarded as portmanteau morphemes expressing simultaneously number and case (though see Halle (1994b), who treats the inflectional vowel as the realization of the nominal thematic suffix, whose surface form is conditioned by number and gender).

Some deviations from this picture are attested. The plurals in (1–2) contain a suffix absent from the singular: the suffix -ĭj- (surface -ʲj-/-ej-) in (1) and the suffix -es- in (2). Nouns in (3), on the other hand, have an additional suffix in the singular absent from the plural. Finally, in (4), the derivational suffix creating names of the young of the species has different forms in the singular and in the plural. Two of these classes are unproductive: (2) presents the full list of nouns with a default plural in -es-, while the class of nouns with a plural in -ĭj-, as in (1), though also closed, is much larger. Conversely, the nouns in (3) and (4) belong to open classes, determined by their derivation.

(1)	a.	brat-Ø/brát-ʲj-a ‘brother-sg.nom/aug-pl.nom’
	b.	knʲazʲ-Ø/knʲaz-ʲj-á ‘prince-sg.nom/aug-pl.nom’
	c.	kólos-Ø/kolós-ʲj-a ‘ear (of a cereal)-sg.nom/aug-pl.nom’
	d.	krɨl-ó/krɨ́l-ʲj-a ‘wing-sg.nom/aug-pl.nom’

(2)	a.	néb-o/neb-es-á ‘sky-sg.nom/aug-pl.nom’, *nebá (néba is sg.gen only)
	b.	čúd-o/čud-es-á ‘miracle-sg.nom/aug-pl.nom’, *čudá (čúda is sg.gen only)
	c.	drév-o/drev-es-á ‘tree-sg.nom/aug-pl.nom’ (obs., the normal form is dérevo)
	d.	slóv-o/slov-es-á ‘word-sg.nom/aug-pl.nom’ (obs., the normal plural is slová)
	e.	tél-o/tel-es-á ‘body-sg.nom/aug-pl.nom’ (obs., the normal plural is telá)

(3)	a.	graždan-ín-Ø ‘citizen-aug-sg.nom’	gráždan-e ‘citizen-pl.nom’
	b.	krestján-ín-Ø ‘peasant-aug-sg.nom’	krestján-e ‘peasant-pl.nom’

(4)	a.	rɨsʲ-ónok-Ø
		lynx-onok-sg.nom
		‘baby lynx’
	b.	rɨsʲ-át-a
		lynx-onok.pl-pl.nom
		‘baby lynxes’

Since my focus here is on the nouns like those in (1), I will address the patterns in (2), (3) and (4) only briefly, in Appendix A. The reason for this distinction is that the pattern exemplified in (1) is the only one limited to plural inflection, for all others the plural stem allomorph is also used in derivation.

1.1. The Plural Augment -ĭj-

Ca. 40 Russian nouns in Zaliznjak (2010) have augmented plurals in -ĭj-.2 For some nouns both augmented and regular plurals seem possible, as illustrated in (5–8). While in (5–6), the choice of the plural form distinguishes between different homonyms, the augmented plurals in (7–8) seem indistinguishable from the regular ones.

(5)	a.	koléno/koléni ‘knee.sg/pl’	-ɨ plural
	b.	koléno/koléna ‘dance move.sg/pl’	-a plural
	c.	koléno/kolénʲja ‘joint, elbow.sg/pl’	augmented plural

(6)	a.	dʲádʲa/dʲádʲi ‘uncle.sg/pl (incl. a child’s word for a male adult)’	-ɨ plural
	b.	dʲadʲjá ‘brothers of a parent’	augmented plural

(7)	a.	loskút/loskutɨ́ ‘shred.m.sg/pl’	-ɨ plural
	b.	loskútʲja ‘shreds’	augmented plural

(8)	a.	grozdʲ/grózdi ‘bunch.sg/pl’	-ɨ plural
	b.	grózdʲja ‘bunches’ (cf. archaic masculine singular grozd ‘bunch’)	augmented plural

There are in addition ca. 10 pluralia tantum nouns that are morphologically augmented plurals (9), which do not have a corresponding singular and can reasonably be argued to have object-mass semantics, as shown by their inability to combine with a cardinal (10), verified in the Russian National Corpus (henceforth RNC, https://ruscorpora.ru/, accessed on 1 December 2025).

(9)	a.	otrébʲja ‘(human) rabble.pl’ (cf. otrébʲje ‘rabble, trash.n’)
	b.	loxmótʲja ‘rags’, xlópʲja ‘flakes’

(10)	a.	*sto	xlopʲjev
		hundred	flakes.pl.gen
	b.	*pʲatʲ	loxmotʲjev
		five	rags.pl.gen

For none of these nouns is the augment -ĭj- present in any derivatives, and the fact that it can appear in mass-denoting pluralia tantum nouns argues that it is associated with morphological rather than semantic plurality. To account for these facts, I will argue that the stems that require augmentation in -ĭj- are specified as morphologically, though not semantically, singular, which means that the augment is needed to enable plural marking.

1.2. The Suffix -ĭj- and the Augment -ĭj-

The semantic drift from a collective noun to an object-mass noun to a plural noun hypothesized in Mihatsch (2016, p. 301) is well-motivated for the augment -ĭj-. Indeed, the suffix -ĭj-, creating neuter mass nouns with a “collective” meaning, is a Proto-Slavic suffix that was already present in Proto-Indo-European (see Merkulova (2006, 2020) for summary and references). Importantly, Meillet (1934, §§401, pp. 404, 521) observes the existence of two such suffixes: the non-productive one giving rise to feminine nouns, and the productive one, giving rise to neuter nouns. Both have been subject to reanalysis: the feminine object-mass bratria ‘brothers’ has become a plural in contemporary Russian (1a), while the mass zemlʲa ‘earth, ground’ has remained feminine and singular. Conversely, the neuter object-mass kamenije ‘stones’ has turned into the plural kamenʲja ‘stones’ in contemporary Russian, while the deadjectival ostrije ‘(sharp) point’ has remained neuter while becoming count. As a result, modern Russian distinguishes the productive neuter mass suffix -ĭj- (11) from the plural augment -ĭj-, though historically they are the same.3

(11)	a.	duračʲjó ‘fools’ (cf. durák ‘fool’)	animate neat mass
	b.	dubʲjó ‘cudgels’ (cf. dubína ‘cudgel’)	inanimate neat mass
	c.	starʲjó ‘old stuff’ (cf. stárɨj ‘old’)	inanimate mess mass

Besides inherent number, the two instances of -ĭj- also differ in semantics. While the singular neuter -ĭj- creates mass nouns (barring semantic drift), derived plurals in -ĭj- can combine with cardinals, as shown by the contrast between the ungrammatical pluralia tantum in (10) and the grammatical derived plurals in (12).

(12)	a.	sto	kolos-ʲj-ev
		hundred	ear (of a cereal)-aug-pl.gen
		‘a hundred cereal ears’
	b.	pʲatʲ	derev-ʲj-ev
		five	tree-aug-pl.gen
		‘five trees’
	c.	semʲ	list-jʲ-ev
		seven	leaf-aug-pl.gen
		‘seven leaves’

Compatibility with cardinals shows that augmented plurals in -ĭj- are genuine plurals. Since the neuter singular suffix -ĭj- forms mass nouns, while the augment -ĭj- is semantically vacuous, either the two must represent different lexical items or the latter has to be a null alloseme of the former.4 In addition, the two morphemes differ in accentuation: the singular -ĭj- is post-accenting, while the plural augment -ĭj- is not. In fact, at a first glance, stress in augmented plurals is unpredictable: while examples (13a–b) appear with inflectional stress in the plural, in examples (13c–f) stress precedes the augment, irrespective of the position of the stress in the singular:

(13)	a.	déverʲ/deverʲjá ‘husband’s brother.m.sg/pl’	inflectional stress
	b.	knʲazʲ/knʲazʲjá ‘prince.m.sg/pl’	inflectional stress
	c.	brat/brátʲja ‘brother.m.sg/pl’	stem-final stress
	d.	kólos/kolósʲja ‘ear (of a cereal).m.sg/pl’	stem-final stress
	e.	dérevo/derévʲja ‘tree.n.sg/pl’	stem-final stress
	f.	krɨló/krɨ́lʲja ‘wing.n.sg/pl’	stem-final stress

Actually, the accentual behavior of -ĭj-augmented plurals can be shown to be determined by animacy: stress precedes the suffix in inanimate nouns and mostly follows it for animate ones (the only exception is (13c), which will be argued to have an accented stem in Section 3). This, however, is not predicted either by their expected morphosyntactic structure or by the standard theories of Russian accentuation.

I will argue that due to the presence of a yer, which is unaccentable when not vocalized, the accentual behavior of the plural augment -ĭj- in inanimates can be explained by the hypothesis that it is accented. Its accentual behavior in animates, however, requires novel assumptions about its morphosyntax: I will argue that the plural augment -ĭj-, being neuter, is incompatible with animate stems, and hence requires the formation of a complex suffix.

2. The Accentuation of the Plural Augment in Inanimates

As is easy to see in (14–15) and all other inanimate -ĭj-augmented plurals in the paper, stress in inanimate -ĭj-augmented plurals falls on the syllable preceding the augment irrespective of the position of the stress in the singular:5

(14)	a.	kopɨ́l, kopɨlá ‘wooden hoe.nom/gen’	kopɨ́lʲja ‘wooden hoes’
	b.	kólos, kólosa ‘ear (of a cereal).nom/gen’	kolósʲja ‘ears (of a cereal)’

(15)	a.	dérevo ‘tree’	derévʲja ‘trees’
	b.	pomeló ‘broom’	pomélʲja ‘brooms’
	c.	koléno ‘elbow, joint’	kolénʲja ‘elbows, joints’

While this can be taken to mean that the augment is pre-accenting and dominant, I will now show that it is sufficient to assume that it bears an accent.

2.1. Russian Accentual System and Plural Accentuation

As discussed in Garde (1968a, 1968b, 1980/1998), Halle (1973), Zaliznjak (1985) and Melvold (1989), among others, the position of stress in Russian is primarily determined by the underlying accentuation of various morphemes. To understand Russian stress, it is necessary to distinguish stress, i.e., the surface prominence characterizing one syllable in a phonological word, from accent, which refers to an abstract property of a syllable in a morpheme. Morphemes can be accented (i.e., bearing an accent on themselves; the only open class in modern Russian, containing both roots and affixes), pre-accenting (i.e., setting the accent on the preceding syllable, possible only for suffixes), post-accenting (setting the accent on the next syllable) and unaccented (having no accentual specification of their own). The position of stress in a word is determined by the Basic Accentuation Principle:6

(16)	The Basic Accentuation Principle (Kiparsky & Halle, 1977):
	Assign stress to the leftmost accented vowel; if there is no accented vowel, assign stress to the initial vowel.

In addition, an affix can be accentually dominant, which descriptively means that its accentual specification trumps over all others. Dominance can be formally implemented as the deletion by the affix of all accents of the constituent that it combines with Poser (1984), Kiparsky (1984), Inkelas (1997), among others.

The workings of this system can be exemplified by the Russian plural. Specifically, Russian nominative plural has three productive allomorphs: the unaccented suffix -ɨ-, the accented suffix -a-, and the accented dominant suffix -a- (an underlying accent is henceforth indicated by underlining; double underlining indicates accentual dominance).7 The three types of stems (accented, unaccented and post-accenting) can be diagnosed by the singular-plural contrast in stress placement.8 As shown in (17–18), while accented and post-accenting stems preserve the stress on and after themselves, respectively, in both the singular and the plural, with unaccented stems the position of stress is not constant throughout the paradigm.

(17)	nominative singular -a-: accented, combines with a-declension stems
	nominative plural -ɨ-: unaccented, combines with non-neuter stems
	a.	accented stem wins over both accented and unaccented suffixes:
		ustric- + a → ústrica ‘oyster’
		ustric- + ɨ → ústricɨ ‘oysters’
	b.	post-accenting stem yields post-stem stress for accented and unaccented suffixes:
		konur_- + á → konurá ‘dog house’
		konur_- + ɨ → konurɨ́ ‘dog houses’
	c.	unaccented stem yields initial stress with an unaccented suffix, and stress on the suffix with an accented one:
		borod- + a → borodá ‘beard’
		borod- + ɨ → bórodɨ ‘beards’

(18)	nominative singular -o-: accented, combines with o-declension stems
	nominative plural -a-: accented, requires neuter stems (the original Indo-European neuter plural)
	a.	accented stem wins over both accented and unaccented suffixes:
		logov- + o → lógovo ‘lair, den’
		logov- + a → lógova ‘lairs, dens’
	b.	post-accenting stem yields post-stem stress for accented and unaccented suffixes:
		božestv_- + o → božestvó ‘deity’
		božestv_- + a → božestvá ‘deities’
	c.	unaccented stem yields initial stress with an unaccented suffix, and stress on the suffix with an accented one:
		zerkal- + o → zérkalo ‘mirror’
		zerkal- + a → zerkalá ‘mirrors’

In (17c) stress is inflectional with the accented nominative singular suffix and initial with the unaccented plural suffix. Conversely, in (18c) it is the singular suffix that is unaccented and yields initial stress, whereas the plural suffix is accented and yields inflectional stress. As the contrast between (18a) and (18c) shows, for a neuter noun, stem-initial stress in the nominative singular may correspond to an accented stem or an unaccented one.

Accentual dominance is illustrated by the plural suffix -a- with masculine nouns, as in (19): no matter where stress falls in the singular, it is inflectional in the plural. The use of the nominative plural -a- for non-neuter nouns is a Russian innovation, which is restricted to masculine nouns in standard Russian (Coats, 1976; Zaliznjak, 1985; Alderete, 1999, p. 166; Timberlake, 2004, p. 136; Iordanidi, 2020; Munteanu, 2021), and dialectally extended to non-neuter nouns (Bromley & Bulatova, 1972, pp. 102–103; Iordanidi, 2020).9

(19)		[á]: accented and dominant	accented stem
	a.	proféssor ‘professor.nom’	b.	professor ‘professor.pl.nom’
		proféssora ‘professor.gen’		professormi ‘professor.pl.ins’

Whereas some phonological generalizations over the masculine nouns that are likely to switch from the default accented nominative plural suffix -ɨ- to the accentually dominant -a- in contemporary Russian have been noted (Zaliznjak, 1967a; Worth, 1983; Tabachnick, 2023), I know of no account of this phenomenon that does not treat it as a diacritic lexical property of the stem. Since this issue is not central for our purposes, no alternative will be proposed here either.

In the domain of nominal declension, this system is compounded by the phenomenon of plural stress retraction (Zaliznjak, 1963, 1967b, 1977a; Halle, 1973, 1975; Melvold, 1989; Brown et al., 1996, etc.), which will be discussed in Section 3.2.

2.2. The Accentual Behavior of Plurals in -ĭj-

As discussed above, the nominative plural suffix -a- is accentually dominant when combined with masculine nouns. Such is not the case, however, for masculine nouns that form their plural with the augment -ĭj-: though all -ĭj-augmented stems require the nominative plural suffix -a-, it is not accentually dominant. In other words, -ĭj-augmented masculine stems do not give rise only to inflectional stress, as shown by the stem-final stresses in (20) (repeated from (13c–d)):

(20)	a.	brat/brátʲja ‘brother.m.sg/pl’	stem-final stress
	b.	kólos/kolósʲja ‘ear (of a cereal).m.sg/pl’	stem-final stress

The empirical generalization characterizing augmented plurals in -ĭj- is twofold. All inanimate plurals in -ĭj- have stress on the stem-final syllable (i.e., preceding the augment), irrespective of gender (which would predict obligatory inflectional stress for masculine nouns). Conversely, in all animate plurals in -ĭj- but one (20a) stress falls on the inflection, i.e., after the augment.10

Since the nominative plural suffix -a- is accentually dominant for masculine nouns, the plural augment -ĭj- must be doing something to circumvent the accentual dominance associated with the masculine. One obvious possibility is that it renders the stem non-masculine, and this is the hypothesis that I will argue for in the next section.

2.3. The Morphophonology of Inanimate Plurals in -ĭj-

Suppose (in a way to be made precise in Section 4) that stems requiring augmentation in the plural cannot take regular plural morphology. Suppose further that to enable plural marking the semantically null augment -ĭj- is added to the stem prior to the addition of the plural suffix, as in (21). I will refer to this mode of morphological composition as iterative suffixation: each successive suffix is added to the combination of the stem and the previous suffixes.

Given that most non-evaluative suffixes of Russian are specified for phi-features and that their phi-features override those of the stem, it does not add to the complexity of the theory to assume that -ĭj- is lexically specified as [–M][–F] (i.e., as neuter) and that these values override the gender specification of the nominal stem it attaches to. As a result, -ĭj-augmented stems would not be masculine:

(21)

Whatever the gender of the nominal stem (Γ) is, the augmented stem is neuter, which means that the (nominative) plural suffix is predicted to be the accentually non-dominant -a-. Since, however, in all inanimate plurals in -ĭj- stress falls on the syllable preceding the augment, the primary hypothesis would seem to be that the augment is pre-accenting and dominant. I will show, however, that these facts can be derived from the properties of the augment vowel.

Much of the theoretical phonology of Slavic languages is dedicated to the two abstract vowels, the back and the front yers, corresponding, respectively, to the Common Slavic short lax ĭ and ŭ. As discussed by Lightner (1972), Pesetsky (1979), Gussmann (1980), and Scheer (2006), among many others, yers are vocalized cyclically left to right if followed by a yer in the next syllable, and deleted if not vocalized. The interaction of yers with stress is a very complex matter, but they are known to resist stress, and for at least some cases, discussed in Halle (1973, 1975, 1997) and Melvold (1989), if an accented yer is deleted, its accent is shifted to the preceding syllable (see Section 3.2).

Evidence that the augment -ĭj- is not underlyingly pre-accenting comes from the environment where its yer is vocalized, i.e., the genitive plural form of the few nouns that require an augment in the plural and also take the surface-zero allomorph of the genitive plural suffix (22).11 If the augment were pre-accenting, the BAP would predict stem-final stress, contrary to fact:

(22)	a.	mužjá/mužéj ‘husband.pl.nom/gen’
	b.	knʲazʲjá/knʲazéj ‘prince.pl.nom/gen’

I hypothesize that the yer in the augment -ĭj- is underlyingly accented. If this yer is realized, as in (22), the surfacing vowel (e) can be stressed. In all other cases, as will be now shown, stress is correctly predicted to surface on the stem-final syllable, just before the augment.

Out of the ca. 40 nouns with -ĭj-augmented plurals in Zaliznjak (2010), 11 have disyllabic stems and the rest have monosyllabic ones. As no trisyllabic stem gives rise to an augmented plural, only three singular stress patterns are observed: post-stem stress (23a), stem-final stress (23b) and stem-initial stress (23c), with the latter two types indistinguishable for monosyllabic stems.

(23)	a.	pomeló, pomelá ‘broom.n.nom/gen’ → pomélʲja ‘brooms’	post-stem
	b.	koléno, koléna ‘elbow, joint.n.nom/gen’ → kolénʲja ‘elbows, joints’	stem-final
	c.	dérevo, déreva ‘tree.n.nom/gen’ → derévʲja ‘trees’	stem-initial

Post-stem stress in the singular (23a) shows that the stem is post-accenting. Assuming that stress computation proceeds cyclically but follows segmental computation (Melvold, 1989), the stem accent is assigned to the accented augment in the first cycle, yielding one single accent (24a). Assuming the BAP, stress (formalized, following Revithiadou (1999), as the projection of an iambic foot) is assigned at the leftmost accent (24b), disregarding the underlying accent of the plural suffix -a-.

(24)

post-accenting stem

As the yer of the augment is not followed by another yer, it is not vocalized. This means that it is deleted post-cyclically (24c), which also entails the deletion of the head of the foot. As a result, stress surfaces on the former tail of the foot (24d), which correctly predicts stem-final stress:

(24)

Stem-final stress in the singular in (23b) can only arise if the stem is accented. As the BAP (16) assigns stress at the leftmost accent, this means that, irrespective of the accentuation of the augment, stress remains on the same syllable in the plural:

(25)	koléno, koléna ‘elbow, joint.n.nom/gen’		accented stem
	a.	[[kolen + ĭj]1 + a]2 → [kolenĭj_ + a]2
	b.	[kolenĭj_ + a] → kolenʲja
	c.	kolenʲja → kolénʲja ‘elbows, joints’

Given that all singular case suffixes of the C- and o-declensions are unaccented, the singular paradigm provides no way of determining whether the initial stress in (23c) is due to a stem that is accented on the first syllable or unaccented. However, an accented stem means that stress should stay on the same syllable in the singular and in the plural, incorrectly predicting stem-initial stress in the plural in (23c). I therefore conclude that the relevant disyllabic stems (three masculine ones and one neuter, see (27) below) are unaccented.

As is easy to see, the computation in (26) proceeds exactly as in (24). The first accent is that of the augment (26a), which yields the structure in (26b), fully identical to (24b). As before, the non-vocalization and subsequent deletion of the yer of the augment forces its accent to move to the preceding syllable (cf. (24c–d)).

(26)

unaccented stem

To summarize, given that the yer-containing augment introduces an accent that it generally cannot bear and that is expected to shift to the left, stress is predicted to fall on the stem-final syllable for all augmented plurals in -ĭj-.

2.4. Intermediate Summary

On the morphological side, a hypothesis was introduced that some nouns cannot combine with plural inflection directly and require an augment. To explain why augmented masculine nouns differ from other masculine a-plurals and do not take the dominant accented plural suffix -a-, the augment -ĭj- was proposed to yield neuter stems. Since most Russian suffixes are specified for some gender, this proposal does not require any motivation.

The same is true for the accentual specification of the augment: most morphemes of Russian are specified for some type of accentuation, and no special motivation is needed for postulating that the augment -ĭj- is accented. The fact that it contains a yer, which cannot bear stress when deleted, accounts for the obligatory stem-final stress for all inanimate stems. While the absence of disyllabic stems with an accented initial syllable is unexpected, it is also not surprising, given how few items we are dealing with:

(27)	a.	póvod, póvoda ‘rein.m.nom/gen’	unaccented stems
	b.	kólos, kólosa ‘ear (of a cereal).m.nom/gen’
	c.	póloz, póloza ‘runner (of a sleigh).m.nom/gen’
	d.	déverʲ/déverʲa ‘husband’s brother.m.nom/gen’
	e.	dérevo, déreva ‘tree.n.nom/gen’

(28)	a.	kopɨ́l, kopɨlá ‘wooden hoe.m.nom/gen’	post-accenting stems
	b.	budɨ́lʲ, budɨlʲá ‘dry stem or stalk.m.nom/gen’
	c.	loskút/loskutá ‘shred.m.nom/gen’
	d.	pomeló, pomelá ‘broom.n.nom/gen’

(29)	a.	koléno, koléna ‘elbow, joint.n.nom/gen’	accented stems
	b.	poléno, poléna ‘log.n.nom/gen’

What is unexpected, however, is the inflectional stress in nine out of the ten animate augmented plurals. On the one hand, since all of these nouns have stem stress in the singular, inflectional stress in the plural cannot be ascribed to post-accentuation. On the other, even if it could,12 the previous section has shown that post-accenting stems yield stem-final stress in the augmented plural.

In what follows I will argue that the animate augmented pattern can be explained by taking the micromorphological approach (cf. Stump (2017)) to plural inflection, i.e., by assuming that the augment can form a constituent with the plural suffix to the exclusion of the lexical stem.

3. The Accentuation of Animate Augmented Plurals

Ten monosyllabic masculine animate stems take an -ĭj-augment in the plural (see (45)–(47) in Appendix B). Though all have stem stress in the singular, only one of them (30b) does not have inflectional stress in the plural.

(30)	a.	zʲatʲ/zʲatʲjá ‘daughter’s husband.sg/pl’	regular animate augmented
	b.	brat/brátʲja ‘brother.sg/pl’	stem-stress animate augmented

To account for this pattern and its distinction from the inanimate one, I propose that (30b) has an accented stem and that augmentation of animate nouns requires the formation of a complex suffix, as in (31).

(31)

Starting on the phonological side, the plural suffix in (31) combines with a non-masculine node, so it is not dominant but still accented, just as before. Assuming that the complex PL node is a phonological cycle, the combination of two accented suffixes should yield stress on the first one (cf. (16)). Assuming, as in Section 2.3, that stress is implemented as the projection of a phonological foot, two options are possible: the projection of an exceptional trochaic foot with the tail on the nominative plural ending (32a), or the projection of a deficient foot (32b). Only the former naturally yields inflectional stress once the yer of the augment is deleted (32c):

(32)

The question now arises what happens to an accented stem in the structure (31). To obtain the correct result, i.e., stem stress, it is necessary to assume that the computation of stress proceeds from left to right, and the accent of the augment loses to the accent on the stem, as predicted by the BAP, either because stress is only computed once starting from the stem or because the leftmost accent wins. There does not seem to be any way to choose between (33a) and (33b) for stress projection:13

(33)

The distinction between (30b) and all other animate augmented plurals can therefore be derived from the assumption that only (30b) has an accented stem, all others have unaccented ones.14 However, the question arises of why a complex suffix, as in (31), is formed instead of iterative suffixation, as in the case of inanimates in (21), and I will ascribe this to a feature clash.

3.1. Gender-Animacy Clash as a Driving Force for Complex Affixation

Two issues arise with the micromorphological account I have proposed: why the usual iterative suffixation structure (21) is impossible with animate stems, and why a complex suffix, like in (31), is not formed with inanimate ones.

Starting with the former, I hypothesize that the neuter augment -ĭj- is incompatible with animate stems due to its feature specification. Assuming, as before, that the augment -ĭj- is specified as neuter, in the normal course of events the neuter specification of -ĭj- should override the gender specification of the nominal stem (as with many other nominal suffixes in Russian, e.g., the suffix -ak or the feminitive -š(a)-). This means that if an animate stem is merged in the structure (31), the resulting node is specified as simultaneously animate and neuter:

(34)

Strikingly, there are no neuter animates in Russian. Besides a few male-denoting augmentatives and diminutives, there exists only one grammatically animate noun of the o-declension (podmastérʲje ‘apprentice’), and it is masculine, while semantically animate but neuter nouns like čudóvišče ‘monster’, ditʲá ‘baby’, čádo ‘child’ or licó ‘person, lit. face’ are grammatically inanimate. Whatever prohibits the combination of the features [+animate] and [–feminine, –masculine] in Russian would also rule out the iterative suffixation structure in (21) for animate stems.

In (31), however, the intermediate PL node combines the neuter and the plural features only. Given that Russian agreement does not reflect gender distinctions in the plural (cf. Stankiewicz, 1968, p. 39; Timberlake, 2004, p. 130; Wiese, 2004, p. 352; Pertsova, 2015, p. 231, etc.), I hypothesize that gender features are lost in the context of [+ plural]. The complex PL node in (31) is then not specified for gender and can therefore combine with animate stems. In other words, the formation of the complex suffix -ĭj-a- in (31) enables augmentation of animate stems without creating an animate neuter.

Evidence for the absence of gender features in the plural comes from semantically animate neuter nouns, which behave as grammatically inanimate in the singular and as grammatically animate in the plural, as shown by their syncretism patterns, which are illustrated in (35). As is well-known but rarely formulated explicitly, the accusative case of a semantically animate neuter noun is realized as syncretic with nominative in the singular (35a), just like with inanimates, and with genitive in the plural, as appropriate for an animate (35b) (cf. Igartua & Madariaga, 2018, p. 30). If gender features are deleted or not projected in the context of [+ plural], an o-declension noun will not be specified as neuter in the plural, and its semantic animacy can become a grammatical feature. (The same reasoning applies to -ĭj-augmented animate plurals: they behave as animates even though the -ĭj- augment is grammatically neuter.)

(35)	a.	Vižu	èto	čudovišče/	lico.
		see.1sg	this.n.sg.acc=nom	monster/	person (lit. face).n.sg.acc=nom
		‘I see this monster/person.’
	b.	Vižu	ètix	čudovišč/	lic.
		see.1sg	this.pl.sg.acc=gen	monster.n.pl.acc=gen	person.n.acc=gen
		‘I see these monsters/persons.’

As to why complex suffixation does not happen with inanimate stems, I propose that it is a last-resort operation that does not occur if not needed.

3.2. Plural Retraction as an Additional Factor in Augmentation

As discussed above, the nominative plural suffix has two accented -a- allomorphs (a dominant and a non-dominant one) and one unaccented allomorph (-ɨ-). Given this, the stem-final stress found in the plural of some unaccented and post-accenting stems is altogether unexpected:15

(36)	unaccented feminine stem: inflectional stress expected, except in sg.acc and pl.nom
	a.	ruká/rúku ‘hand.nom/acc’	regular
		rúki/rukámi ‘hand.pl.nom/ins’
	b.	dušá/dúšu ‘soul.nom/acc’	retracting
		dúši/dúšami ‘soul.pl.nom/ins’

(37)	post-accenting feminine stem: inflectional stress expected throughout
	a.	čertá/čertú ‘line.nom/acc’	regular
		čertɨ́/čertámi ‘line.pl.nom/ins’
	b.	stroká/strokú ‘text line.nom/acc’	retracting
		stróki/strókami ‘line.pl.nom/ins’

The phenomenon of plural stress retraction offers an independent piece of evidence in favor of ascribing the systematic stem-final stress in inanimate augmented plurals and the inflectional stress in inanimate augmented plurals to the inability of a deleted yer to bear stress. While the interaction of yers with stress is a very complex matter, a very similar case arises with some yer-containing nouns like kolʲcó ‘ring’ (stem -kolĭc-), as discussed in Halle (1973, p. 320). Given that all singular forms of this noun bear inflectional stress (38a), inflectional stress is expected in the plural, too, but this noun turns out to be subject to plural stress retraction. If the stem-final yer is vocalized, as in the genitive plural, stress falls on it (38b), but if the stem-final yer is deleted, as in all other plural forms, including the nominative and the dative, stress falls on the preceding syllable (38c).16

(38)	a.	kolʲcó/kolʲcá ‘ring.nom/gen’
	b.	koléc ‘ring.pl.gen’
	c.	kólʲca/kólʲcam ‘ring.pl.nom/dat’

Whatever the proper account of plural retraction is (for various proposals see Melvold, 1989; Alderete, 1999; Butska, 2002; Feldstein, 2006, 2017; Dubina, 2012; Yanovich & Steriade, 2010; Osadcha, 2019, among others), a natural question to ask is whether the systematic stem-final stress in inanimate augmented plurals could arise from this phenomenon.

Indeed, a possible alternative to my micromorphological account could be that the inflectional stress observed in nine out of ten animates is the default (expected if the augment -ĭj- is post-accenting, like the contentful neuter singular suffix -ĭj-), and the stem-final stress in inanimate augmented nouns is caused by plural stress retraction.

Several problems arise with this alternative. The first one is that it does not explain the contrast between animate and inanimate stems. Moreover, when considering the accentual behavior of the two nouns derived with the neuter singular mass suffix -ĭj- that exhibit inflectional stress in the singular and have a plural, no uniform picture emerges, even though both are inanimate:17

(39)	a.	pitʲjó/pitʲjá ‘drink.sg/pl’	stress after Zaliznjak (2010)
	b.	žnivʲjó/žnívʲja ‘stubble-field.sg/pl’

Secondly, since no explanation of plural retraction is available (beyond the hypothesis that it is a strategy to maximize the distinction between the singular and the plural, cf. Alderete (1999) and other OT proposals), it is usually formalized as a diacritic feature on the root. To account for stress retraction in all inanimate augmented plurals, this feature will have to act non-locally, across the plural augment.

I conclude that plural stress retraction by itself cannot account for the position of the stress in augmented plurals.

Conversely, one obvious advantage of appealing to plural stress retraction is the restoration of the historical link between the contentful post-accenting neuter mass suffix -ĭj- (11) and the pre-accenting plural augment -ĭj-. This becomes possible on the assumption that (39b) rather than (39a) represents the default case and the suffix itself triggers plural stress retraction. While this proposal would easily account for inanimate augmented plurals, in animate augmented plurals stress retraction to the stem would have to be ruled out in (31) by an appeal to cyclicity. As this would require a much deeper understanding of how yer deletion interacts with accents, plural retraction and stress assignment, I will not attempt to tackle this issue here, retaining the hypothesis that the augment -ĭj- is synchronically accented even though the historically related contentful post-accenting neuter mass suffix -ĭj- is post-accenting.

3.3. Intermediate Summary

I have proposed that plural augmentation with -ĭj- in Russian applies to a class of nominal stems that are semantically count but incompatible with plural morphology (with the technical details to still be spelled out, see Section 4). The augment -ĭj- is used to create a pluralizable stem, and since it is grammatically neuter, the resulting stem combines with the accented non-dominant allomorph of the plural suffix whatever the gender of the non-augmented root. As a result, the obligatory stem-final stress for inanimate augmentable nouns is predicted.

The incompatibility of the neuter augment with a grammatically and semantically animate stem forces the formation of a complex plural suffix, as in (31), repeated below. This option is a last-resort one and hence unavailable for inanimate nouns.

(31)

Since the augment yer cannot be stressed when not vocalized, the stress in the complex plural suffix migrates to the inflection, predicting inflectional stress in all animate augmented plurals unless they have an accented stem (i.e., in the sole case of brat/brátʲja ‘brother.sg/pl’ in (30b)).

The difference in the stress position with animate and inanimate augmented plurals is therefore made to follow from the difference in structure, and that difference was proposed to be due to the incompatibility between animacy and the neuter gender in Russian.

The link between the semantically null accented augment -ĭj- and the contentful mass neuter post-accenting suffix -ĭj- is historically well-motivated. To account for it on the phonological side, the fact that the augment -ĭj- is an accented morpheme (as opposed to the post-accenting neuter suffix -ĭj-) can be due to plural stress retraction (cf. Zaliznjak, 1963, 1967b, 1977a; Halle, 1973, 1975; Melvold, 1989; Brown et al., 1996, among others). As for the semantic connection, the augment -ĭj- can be derived from the suffix -ĭj- under the assumption that it is semantically null in the context of plural, which would be a natural case of semantic haplology given that mass and plural semantics can both be reduced to cumulativity (Link, 1983).

4. The Morphosemantics of the Augment

The natural intuition behind the use of the augment is that stems requiring augmentation in the plural cannot otherwise combine with regular plural morphology. Since this constraint does not seem to be semantic, I hypothesize that such stems are underlyingly specified as grammatically singular.

Evidence for underlying grammatical number that does not correspond to the denotation comes from pluralia tantum nouns. As shown by McCawley (1975), Gillon (1992), Corbett (2000), Ojeda (2005), Acquaviva (2008) and Lasersohn (2011), among others, pluralia tantum nouns can be both semantically atomic (countable; 40a), and non-atomic (40b). In the former case, their plural morphology is purely grammatical (even if historically motivated).

(40)	a.	odni	sani/	štany
		one.pl	sleigh.pl/	trousers.pl
		‘one sleigh’, ‘one pair of trousers’
	b.	*odni		drova/	denʲgi/	kanikulɨ
		one.pl		firewood.pl/	money.pl/	summer holidays.pl

To account for the observed picture, I will assume two relevant number features: [±cumulative] (corresponding to semantic cumulativity, including plurality) and [±singular] (corresponding to semantic atomicity). Crucially, I assume that number, like gender, can be purely formal and these features can be lexically specified on a noun without being motivated semantically: thus, a plurale tantum, irrespective of its atomicity, is lexically marked as [–singular], and nouns requiring a plural augment are lexically marked as [–cumulative]. I will further assume that all nouns not lexically specified for these features can have their values set semantically (mass and plural NPs are [+cumulative]) or syntactically (the [–singular] value on an NP can result from agreement with a plurality-introducing higher head, which can be either Link’s (1983) ∗-operator or a cardinal).

Assuming that non-plural environments are unmarked for either [±singular] or [±cumulative], a plurale tantum noun that is semantically count but formally [–singular] is compatible with both semantically singular environments (40a) and semantically plural ones. On the other hand, lexically specified [–cumulative] nouns like brat ‘brother’ would be incompatible with plural environments, where the feature value [–singular] assigned to the noun would conflict with its underlying [–cumulative] feature (unlike in Harbour’s (2011, 2014) system, where the feature combination [–additive][–singular] defines the paucal).

To explain why such a combination should be excluded, it is reasonable to recall at this point that the two features postulated, [±cumulative] and [±singular], in the count domain can easily be replaced with the two values of the same feature, e.g., [±singular]. It then becomes necessary that nouns can be specified for the same feature twice, i.e., for a plurale tantum to bear both an inherent formal feature value “non-singular” (i.e., [–singular] _inherent) and an unvalued feature ([αsingular] _syntactic) for agreement purposes. While so far using different labels for the inherent and the assigned features has made it easier to distinguish the two in discussion, morphological and syntactic plurality should at some level correspond to the same feature, which means that conflicting values can be reasonably assumed to be impossible.

The role of the augment is therefore to override the underlying [–cumulative] specification of the nominal stem. To make this intuition compatible with the Y-model architecture of the grammar (Chomsky, 1995, et seq.), it is enough to assume that, as in syntax, if the -ĭj- augment is present in the numeration along with the [–cumulative] noun, the derivation can converge, otherwise it will fail.18

5. Conclusions

I have shown that stress in augmented plurals in -ĭj- is determined by the animacy of the stem: stress is stem-final for inanimate nouns and predominantly inflectional for animate ones. I then proposed that the two types of stems require different constituent structure. Assuming that inanimate stems combine first with the accented augment and then with the plural suffix, stem-final stress can be accounted for by the inability of a deleted yer of the augment to bear stress. Conversely, animate augmented plurals involve a different constituent structure, where the augment and the plural suffix are combined prior to merging with the lexical stem. I showed that the two possible structures, the iterative suffixation in (21) and complex suffixation in (31), correctly derive the accentual behavior of the augment in the two contexts: it is the constituent structure that determines where the stress of the yer surfaces when the yer is deleted.

To explain why animate and inanimate nouns project different structures, I argued that the augment -ĭj-, being neuter, can combine directly with an inanimate stem but cannot do so with an animate one, which requires complex suffix formation. To answer the general question of why some nouns need augmentation in the plural, I proposed that they are lexically specified as formally singular ([–cumulative]). Since such a specification excludes them from plural environments, an augment is required to create a stem that can be pluralized.

The -ĭj- augment is historically derived from the contentful neuter mass suffix -ĭj-. I argued that the loss of meaning results from semantic deletion or impoverishment applied to -ĭj- in the context of [–singular], and I speculated that the augment was reanalyzed as inherently accented (in contrast to the post-accenting neuter mass suffix) as a result of plural stress retraction.

The entire project, beyond accounting as it does for the link between the [α animate] feature of augmented plurals in -ĭj- and the position of stress (before or after the augment), also provides evidence for micromorphology: the hypothesis (Stump, 2017, et seq.) that affixes may be combined with each other in a constituent not containing the lexical root.

Several issues remain. Given the existence of other types of augmented plurals (2–4), the question of their constituent structure and what it entails forms an important issue in theoretical morphology, but also morphosyntax and morphosemantics.

Appendix A concerns other instances of plural augmentation in Russian, since the four classes in (1–4) turn out not to have the same morphosyntactic status. Appendix B deals with several unrelated issues arising with animate augmented plurals: the two exceptional nouns that belong to the feminine a- and ĭ-declension classes in the singular, the two doubly augmented nouns, and the exponence of genitive plural.