Stress, gender, and declension class in Belarusian

3.1 From Fixed paradigmatic stress to Stem-Suffix stress: Declension Ia masculines

Recall that the C-Stem-Suffix pattern opposes stress on the stem in the singular with stress on the declensional suffixes throughout the plural, i.e., SS-s in the singular but ss-S in the plural subparadigms. One ongoing innovation in Standard Belarusian is for the position of stress to change in one of the two subparadigms of a Fixed stress noun in order to achieve the target C-Stem-Suffix paradigmatic profile, either by advancing stress onto the suffixes in the plural subparadigm of A-Fixed Stem stress nouns, as in zájac ‘hare’ (6a) or by retracting stress to the stem in the singular subparadigm of B-Fixed Suffix stress nouns, as in hryb ‘mushroom’ (6b). Compare the more conservative Standard Russian (SR) stress profiles (data from Zalizniak 1977) to Standard Belarusian (SB).^[16]

This innovation is found in a good number of masculine nouns (see Loban 1957: 214–215), and even in borrowings which originally came into the language with fixed stem stress.^[18] It is clear that Standard Belarusian is now actively associating a C-Stem-Suffix paradigmatic profile with Declension Ia masculine nouns.

In fact, it isn’t only the Fixed Stress paradigms that undergo change, although this is the more interesting case. We find that original mobile patterns of other types also change to the C-Stem-Suffix type. For example, the majority of Type C₁ masculine nouns with stem stress in the singular and in the Nom/Acc plural but suffix stress in the oblique cases of the plural (Type C₁ = SS-s in the singular, SS-s also in Nom/Acc pl, but ss-S in the oblique plural case forms), as in some commonly used nouns such as hrom ‘thunder’, vécjer ‘wind’, vólas ‘hair’, front ‘front’, zvjer ‘beast’, kámjen’ ‘stone’, kóran’ ‘root’, hólub ‘dove’, žólud ‘acorn’, shifted the stress in the Nom/Acc plural from the stem to the suffixes in order to match the place of stress in the other plural forms (Nom/Acc pl SS-s became ss-S). This naturally created a clean C-Stem-Suffix pattern for the paradigm as a whole (Singular SS-s vs. Plural ss-S) which further reinforced the identification of the C-Stem-Suffix stress pattern with masculine nouns. Thus, the C-Stem-Suffix prosodic profile is now the attractor pattern for Declension Ia masculine nouns.

3.2 From Fixed paradigmatic stress to Suffix-Stem stress: Declension II feminines

In Declension II feminines, the tendency is to generalize the D-Suffix-Stem pattern (Singular ss-S vs. Plural SS-s). The A-Fixed Stem prosodic profile undergoes a stress shift in the singular from the stem to the suffix, as in vjaxá ‘perch, beacon’ (7a), while a B-Fixed Suffix noun retracts stress from the suffix to the stem in the plural subparadigm, as in dudá ‘pipe’ (7b). Both shifts converge on D-Suffix-Stress prosody in Standard Belarusian and are observed only in Declension II nouns.

Change from the A-Fixed Stem type is not as common as is change from the B-Fixed Suffix type; the latter innovation is also found in grafá ‘graph’, kumá ‘godmother’, turá ‘rook, castle’ (Dubina 2012: 102–103; Loban 1957: 226), and several others. But what is important here is the fact that both types of fixed stress patterns change to the same D-Suffix-Stem profile, and that this pattern differs from the attractor pattern in Declension Ia masculine nouns. In Declension II feminines, the attractor pattern is D-Suffix-Stem stress.

Although mobile stress patterns tend to be increasing their membership in Russian and Ukrainian as well, especially for implementing the number contrast, in those languages, unlike in Standard Belarusian, the C-Stem-Suffix pattern is increasing in all noun classes (Khazagerov 1985: 32–52) and it is not limited to Declension Ia masculines; D-Suffix-Stem stress also gains some ground in Declension II feminines (pp. 38–41). What matters for our purposes is that Standard Belarusian is much more innovative in developing a very strong correlation between stress and declension class/gender: The C-Stem-Suffix stress pattern now increasingly corresponds to Declension Ia masculines, and the D-Suffix-Stem stress characterizes Declension II feminines.

3.3 Stress, gender, and declension class

These innovations in paradigmatic stress patterns appear to be taking place fairly rapidly. In 1957, Loban listed 140 masculine nouns with stem versus suffix stress variants either in the singular or plural subparadigms in Standard Belarusian. When Smułkowa (1978: 139–140) followed up on these particular nouns with a native speaker informant some twenty years later, most of the masculine nouns had already regularized into B-Fixed Suffix or the mobile C-Stem-Suffix patterns and only 18 still had variant forms. Furthermore, the direction of change in Smułkowa’s study is instructive. For masculine monosyllabic nouns with free variation in the singular and suffix stress in the plural subparadigm, her informant regularized Type B-Fixed Suffix stress throughout. But if free variation occurred in the plural and there was stem stress in the singular, then her informant for the most part chose suffix stress in the plural in favor of the C-Stem-Suffix pattern, rather than stem stress which would have produced the more common and expected A-Fixed Stem pattern. In disyllabic stems, too, the results favored the mobile C-Stem-Suffix pattern; only a few nouns took on A-Fixed Stem stress. Compare the Standard Russian plural subparadigm (from Zalizniak 1977) to Standard Belarusian plural forms as given in Loban (1957) and the Hramatychny sloŭnik nazoŭnika (2008) for svat ‘matchmaker’ (all show stem stress in the singular) in (8); cf. also bok ‘side’ in Loban 1957 vs. 1987 and 2008 dictionaries.

And in Declension II feminine nouns, the change is, of course, in favor of the other type of mobile stress. Loban (1957) documented 28 feminine nouns with accentual variants either in the singular or plural subparadigms in Standard Belarusian. By 1978 when Smułkowa did her study, 16 of the 28 feminine nouns had acquired the mobile D-Suffix-Stem pattern as the only pattern while 12 showed variation. The change from B-Fixed Suffix stress to the D-Suffix-Stem pattern continues to be active for Belarusian speakers. The following nouns have B-Fixed Suffix stress cognates in Russian (listed in Fedianina 1982: 94; Khazagerov 1985: 168) but D-Suffix-Stem in Belarusian: stapá ‘heel’, karčmá ‘tavern’, kačargá ‘poker’, stupnjá ‘foot’, grafá ‘graph’, turá ‘castle, rock’, and dudá ‘pipe’. Dubina (p.c. Google search of February 11, 2015) found that the Turkic loanword bašká ‘head’ borrowed from Russian where it is a B-Fixed Suffix noun has become a D-Suffix-Stem paradigm with retracted stress and a reconstructed stressed “o” in the plural stem in Belarusian: Nom pl [bóški].^[19]

To summarize, the result of the various stress shifts is that Type C₁ stress and D-Suffix-Stem patterns are not found in Declension Ia masculines, and the B-Fixed Suffix stress pattern is almost completely lost in Declension II feminines. And note that innovations in A-Fixed Stem paradigms are complementary: Declension Ia masculine nouns prefer C-Stem-Suffix stress while Declension II feminines favor D-Suffix-Stem stress, but never vice versa. Sometimes a given fixed pattern changes to the other fixed pattern, but here, too, the innovations tend to line up with declension class and gender: Declension II feminine nouns which had B-Fixed Suffix stress sometimes become A-Stem Stress types, e.g., SB kíška ‘gut, intestine’, cf. Russian kišká, and bárža ‘barge’, cf. Russian baržá; Declension I masculine nouns with A-Fixed Stem stress sometimes acquire B-Fixed Suffix stress, e.g., B-Fixed Suffix stress in SB zubr ‘bison’, Gen sg zubrá; cf. Type A in Russian zubr, Gen sg zúbra; and SB los’ ‘elk’, Gen sg lasjá, cf. SR los’ with A and C₁ stress in Zalizniak (1977). Yet it is never the case that Declension II feminine nouns take on B-Fixed Suffix stress.

The restricted distribution of these prosodic innovations led to a much tighter connection between paradigmatic stress, gender and declension class such that either gender/declension class ⇒ stress pattern or stress pattern ⇒ gender /declension class. Note that the fundamental opposition is one between the Ia masculine and II feminine noun declensions. At this point it isn’t entirely clear whether there is a necessary distinction between gender and declension class in these innovations, or whether it is the stress pattern or the gender/declension class that is implied. Developments in the neuter nouns suggest that it is gender which has come to the forefront.

3.4 Neuter nouns: gender versus declension class

In Standard Russian, neuter nouns are often given their own separate declension class. The criteria identified by Corbett (1982) for designating neuter nouns as belonging to their own Declension IV class and thus distinct from Declension I masculines are: 1) different Nom sg suffixes, with /-Ø/ for masculine nouns but /-o/ or /-e/ in neuters; 2) different exponents in the Gen pl, with /-ov/ for masculine nouns and /-Ø/ in neuters; 3) different Nom pl suffixes, with /-i/ (and some stressed /-á/) in masculine nouns and /-a/ in neuters; and 4) different paradigmatic stress patterns.

But in Belarusian, leveling has generally eliminated gender/declension class distinctions in the Nom pl (almost all nouns now take /-i/, /-y/), and the nearly complete spread of the Declension Ia Gen pl exponent /-ow/ (pronounced [-aw] when unstressed) to neuter nouns now makes them virtually identical to masculines (Bethin 2017a, 2017b). The only remaining differences in case exponents are in the Nom singular, where there is no overt suffix in Ia masculine nouns but an overt /-o/ or /-e/ gender marker in neuters. Thus, Belarusian neuter nouns share most of their declensional exponents with Ia masculines, and in Standard Belarusian they belong to the same general Declension I class.

If declension class implies a particular paradigmatic stress pattern, we might expect Ib neuter nouns to undergo changes in paradigmatic stress that are like those found in their Declension I class, and to innovate C-Stem-Suffix stress together with the Ia masculine nouns. At the same time, neuter nouns, unlike masculines, do have an overt gender and case marker in the nom/acc singular and they share this particular property with Declension II feminines, so there is also the possibility that neuter nouns could pattern with Declension II feminine nouns and innovate D-Suffix-Stem stress. A third possibility is that neuter nouns do neither.

As may be expected, Belarusian neuter nouns have also undergone significant changes in their paradigmatic stress profiles. We saw that one of the major differences between Standard Russian and Belarusian neuter nouns lies in the counts of nouns with Type C-Stem-Suffix stress: compare 43 such nouns in Russian to only three in Belarusian (see (3) and (4) above). Most neuter nouns now favor A-Fixed Stem stress. Because the singular versus plural opposition implemented by the C-Stem-Suffix pattern in Common Slavic originated with neuter nouns from where it then spread to masculine and feminine nouns (Stankiewicz 1979: 66–67), it is particularly striking that the innovation in neuter nouns is to actually eliminate C-Stem-Suffix stress and to simply default to the majority A-Fixed Stem stress pattern.^[20] Compare C-Stem-Suffix stress in Russian to A-Fixed Stem stress in Belarusian neuter nouns in (9).^[21]

This change in neuters took place fairly rapidly. The stress variation in neuters observed by Loban (1957) was uniformly resolved in favor of A-Fixed Stem stress by Smułkowa’s informant (see (10) below and Smułkowa 1978: 140; Table 15).

So, in spite of the fact that neuter nouns share many case exponents with Declension Ia masculines, they do not follow Declension I in favoring the C-Stem-Suffix pattern, unlike in Russian. In other words, there is no uniformity within Declension I with respect to what type of paradigmatic stress patterns are associated with it. And although the counts for neuter nouns with D-Suffix-Stem stress are comparable in the two languages, with 83 in Russian and 100 in Belarusian, this is not a productive or a target pattern for neuters in either language. What we find is that although frequently used neuter nouns tend to preserve original D-Suffix-Stem mobile stress in Belarusian, e.g., aknó ‘window’, pjaró ‘feather’, pis’mó ‘letter’, vjadró ‘bucket’, hnjazdó ‘nest’, bjadró ‘hip’, kryló ‘wing’, jarmó ‘yoke’ (see Loban 1957: 233), they do not retain and do not adopt a new C-Stem-Suffix or a new D-Suffix-Stem stress pattern. Neuter nouns now actually avoid having C-Stem-Suffix stress.^[22] This suggests that it might be gender rather than declension class that is becoming affiliated with paradigmatic stress, and neuter nouns simply default to the majority (unmarked) pattern of A-Fixed Stem stress.

There are other indications of the demotion of neuter gender in Belarusian. As Obnorskii (1927–1931) already pointed out for southern Russian dialects and which Loban (1957: 230) extended to Belarusian, vowel neutralization in unstressed syllables (known as strong akanne/jakanne) in which the contrast among /o/, /e/, and /a/ is neutralized to /a/ in unstressed position (Vaitovich 1968) in Standard Belarusian may have played a role in this. When a noun has stress on the stem, the /-a/ suffix for Declension II feminine nouns and the /-o/ or /-e/ suffix for Declension Ib neuters are homophonous; both are pronounced /a/ when unstressed and spelled with -a in Belarusian. So, it is not always clear from the citation form whether the noun is feminine with an etymological /-a/ or neuter with etymological /-o/, /-e/. The morphological leveling in the plural with Nom/Acc plural now being /-i/, /-y/ for all declension classes and the spread of the masculine overt Gen plural suffix /-ow/ as [-aw] to neuter and feminine nouns in Belarusian further erode any potential gender distinctions. It is not surprising that these factors have already resulted in a gender/declension class shift for some neuter nouns. For example, the Russian and Ukrainian tesló ‘adze’ is etymologically neuter, but because the Belarusian cognate has Fixed Stem stress, cjásla, and the Nom sg suffix is unstressed and pronounced as /-a/, cjásla is now grammatically feminine in Belarusian (Loban 1957: 230).

What happens is that the default to Type A stress in neuter nouns actually serves to foreground the two-way opposition in grammatical gender between masculine (Declension Ia) and feminine (Declension II) nouns. This [masculine] versus [feminine] grammatical gender contrast serves to raise the profile of semantic/referential gender in the minor class of common gender and masculine a-stem nouns. Furthermore, given that innovating patterns of paradigmatic stress are now so tightly bound with gender and declension class (with stem stress in the singular in Declension Ia masculine nouns and suffix stress in the singular of Declension II feminine nouns), it is not unexpected that stress affects the nature of language change elsewhere. So, it is stress together with the semantic/referential gender features [male] and [female] that now determine the inflectional morphology in common gender and masculine a-stem nouns.

4.1 Common gender nouns

The common gender nouns, a group of about 200 nouns in the 2008 dictionary, have semantic/referential gender agreement: male referent byv _masc takój _masc múrza ‘[he] was such a slovenly person’ versus female referent byla _fem takája _fem múrza ‘[she] was such a slovenly person’, in terms of attributive adjective, past tense verb form and pronoun agreement (see Biryla and Shuba 1985: 59–60; Corbett 1983; Eŭsievich 2015: 96–98; Herrity 1983 for an overview). In Standard Russian and Ukrainian today, common gender nouns still belong to Declension II where they consistently decline like Declension II feminines.

But in Belarusian, these nouns now show a systematic sensitivity to the position of stress in the singular subparadigm. If stress falls on the suffixes in the singular (i.e., B-Fixed Suffix or D-Suffix-Stem paradigmatic stress), as it does in siratá ‘orphan’, xanžá ‘bigot, hypocrite’, ljawšá ‘left-handed person’, razmaznjá ‘nincompoop’, sarvihalavá ‘daredevil’, skupjandá ‘cheapskate’, then the noun declines like a Declension II feminine noun (cf. “hard” stem siratá ‘orphan’ and “soft” stem xanžá ‘bigot, hypocrite’), regardless of its semantic gender, as exemplified in (12).

But if stress falls on the stem in the singular, then we see morphological changes: When the referent is [male], the noun takes the inflectional suffixes of Declension Ia masculine nouns in the oblique cases, Dat sg, Prep sg, and Instr sg, shown in bold in for pláksa ‘crybaby’ and múrza ‘slovenly person’ in (13a). But when the referent is [female], the noun retains canonical Declension II inflection (13b) (see Biryla and Shuba 1985: 84–86, 94–95; Eŭsievich 2015: 96–98; Lukashanets 2007: 153–158; Mayo 1976: 16).

That this innovation is entirely based on stress is shown by the common gender noun ‘cheapskate’ which has variation in the singular subparadigm. Note that when it is a Type A-Fixed Stem noun with stress on the stem, skupjénda, there are two possible paradigms and they depend on the referent (14a), but when it has stress on the singular suffixes in a D-Suffix-Stress paradigm (and vowel neutralization in the stem) as skupjandá, there is only one Declension II paradigm (14b).

So stress to a large extent controls the morphology: when stress falls on the stem in the singular, then semantic/referential gender comes to the fore and the feature [male] triggers Declension I inflections in the Dat, Prep and Instr singular cases while the feature [female] selects Declension II inflections throughout; but when stress falls on the suffixes in the singular, then the noun retains Declension II morphology.^[24]

Because most common gender nouns have A-Fixed Stem paradigmatic stress and many of them are commonly used nouns, such as kaljéha ‘colleague’, pláksa ‘crybaby’, múrza ‘slovenly person’, kapryzúlja ‘capricious person’, listanóša ‘mail carrier’, zadaváka ‘arrogant person’, pisáka ‘scribbler’, pustamélja ‘windbag’, sknára ‘skinflint’, tupíca ‘blockhead’, this morphological change involves a substantial number of common gender nouns. What is even more remarkable is that this stress-dependent morphology is also found in the a-stem masculine nouns, the overwhelming majority of which have only [male] referents.

4.2 Masculine a-stem nouns

Historically the a-stem masculines belonged to Declension II, and they still do in Russian (Matushansky 2013; Steriopolo 2018, and references therein). These are all labeled as grammatically [masculine] nouns in Biryla and Shuba (1985: 84–85), Lukashanets (2007: 153–155) and in the 2008 dictionary.^[25] Although there was some uncertainty with respect to the declension of these nouns already in 14th–16th century Belarusian, with sporadic attestations of Declension I suffixes in a few such nouns, e.g., Instr sg sy sluhóm ‘with the servant’, s staršynóm ‘with the sergeant’, Dat sg k suddzjú ‘to the judge’, these a-stem nouns were and still are basically declined as Declension II nouns throughout East Slavic (Karskii 1956 [1911]: 15–17).

But Belarusian has now undergone change here, too, and stress affects the morphology: If the noun has Type A-Fixed Stem paradigmatic stress, then it takes on Declension Ia masculine suffixes in the Dat sg, Prep sg and Instr sg cases, as in the naturally [male] mužčýna ‘man’ and nouns designating professions such as cjéslja ‘carpenter’; otherwise, these nouns decline like Declension II feminines, regardless of semantic/referential gender, e.g., staršyná ‘sergeant’, suddzjá ‘judge’, as shown in (15). This mixed declension is given for 135 nouns in the 2008 dictionary.^[26]

Compare Russian and Ukrainian, where a-stem masculines have only Declension II exponents, regardless of the position of stress (see Matviias 1969: 81; Shvedova 1980: 489–490). The situation in Belarusian a-stem masculines parallels that of common gender nouns where stress prosody determines the morphological exponent: Only nouns with A-Fixed Stem stress have the option of taking on Declension Ia suffixes if they refer to males. The generalization is: If stress falls on the stem, then semantic/referential gender determines declension class; otherwise, the default is Declension II inflection regardless of gender. In the a-stem masculines, this is not just a mechanism to distinguish the referent gender [male] from [female] of a given noun (under certain stress conditions) as in the stem-stressed common gender nouns. In those a-stem masculines where the semantic/referential gender is always [male], the morphology is primarily a consequence of the position of stress in the singular subparadigm.

Furthermore, this language change is productive in Belarusian, applying also in a set of related masculine and common gender nouns derived by the suffixation of /-ko/, as in bratkó ‘brother’, affec., bác’ka ‘father’, affec., xval’kó ‘braggart’, znawkó ‘know-it-all’ (Bulyka et al. 1979: 14–16; Karskii 1956 [1911]: 120–121). This suffix may be either stressed or unstressed and it also presents an interesting challenge in the face of vowel neutralization. These are all /-ko/ nouns in Ukrainian and in Belarusian dialects where there is no vowel neutralization of /o/ > /a/ in unstressed syllables, and they belong to Declension Ia masculines. But in those Belarusian dialects where there is strong vowel neutralization of /e/, /o/ and /a/ to /a/, as in the standard language, these /-ko/ nouns with stem stress neutralized the Nom sg /-o/ with /-a/, e.g., bác’ka /báts^jko/ ‘father’ > [báts^jka]; dzjádz’ka /dz^jádz^jko/ ‘uncle, fellow’ > [ʣ^jáʣ^jka], and the suffix is always pronounced as /a/. As a result, these types of nouns have now actually moved into the a-stem masculine noun class. And in Standard Belarusian today they have acquired the new mixed Declension I and Declension II morphology.^[27] In contrast, common gender nouns in /-ko/ with B-Fixed Suffix stress and thus a stressed Nom sg suffix /-ó/, as in hnjawkó ‘person who angers easily’, xval’kó ‘braggart’, njamkó ‘mute’, are declined completely as a Declension Ia noun if the referent is a male, but they do not decline at all if the referent is a female (Biryla and Shuba 1985: 85, Note 2; Eŭsievich 2015: 98; Lukashanets 2007: 158). The plural forms are the same for both.

This behavior reveals that there is a critical connection between stress and gender. Given that B-Fixed Suffix stress is a property of Declension Ia masculine nouns and is not active in Declension II feminines, these /-ko/ nouns avoid having Type B stress (and stressed Ia masculine suffixes) associated with the feature [female] by simply not declining in the singular subparadigm when they have female referents.

4.3 Discussion

It is clear that these innovations are taking place in the context of the paradigm as a whole. We saw that changes in stress patterns elsewhere have resulted in (a) Declension II feminine nouns with virtually no B-Fixed Suffix stress and no C-Stem-Suffix stress but significant gains in Type D-Suffix-Stem stress, and (b) Declension Ia masculines with no D-Suffix-Stem stress, some B-Fixed Suffix stress and a systematic increase in C-Stem-Suffix stress. Thus, the distribution of these paradigmatic stress profiles in Belarusian is complementary, as shown in (17).

Recall that if stress is on the inflectional suffixes in the common gender and a-stem masculines, then Declension II holds regardless of semantic/referential gender. One possible reason for maintaining Declension II exponents under stress (aside from their perceptual salience under stress which probably also contributes to their resistance to change) is that D-Suffix-Stem is a pattern found only in Declension II nouns and not in Declension Ia masculines; it is the “attractor” pattern for Declension II feminines. In other words, the D-Suffix-Stem prosodic profile is directly related to Declension II exponence, and common gender and a-stem masculines follow this principle. But when stress falls on the stem as it does in A-Fixed Stem stress nouns, there is no particular connection between stress and declension class because A-Fixed Stem stress is found in all declension classes and with all genders. This apparently opens up the possibility for semantic/referential gender to become more instrumental in determining inflectional exponence. So, it looks like now stress together with semantic/referential gender determine the nature of inflectional exponence in common gender and a-stem masculine nouns. The reluctance of suffix-stressed -ko nouns to decline when referring to a female supports this connection between gender and stress. If paradigmatic stress is indeed a major factor determining language change in these nouns, then we would expect common gender and a-stem masculine nouns with other types of paradigmatic stress profiles to follow these trends in the language.

Thus, we might predict language change in common gender and a-stem masculine nouns that have what would be considered a “problematic” prosodic profile, that is, a paradigmatic stress pattern that does not match its declension class or gender. One such holdover pattern is B-Fixed Suffix stress in a few common gender and a-stem masculines (e.g., xanžá, tamadá). This stress pattern has lost ground in Declension II feminines so one might expect either a stress shift or a change in inflectional exponence here as well. And if gender is associated with a specific innovation in stress, then [male] nouns should pattern with [masculine] Declension Ia nouns in stress innovations. In other words, given that the Declension Ia masculine nouns with A-Fixed Stem stress often innovate C-Stem-Suffix stress, we might expect a similar change in common gender and a-stem masculine nouns with A-Fixed Stem stress and the feature [male]. Therefore, we would predict that:

1. If Declension II morphology is maintained in a common gender or a-stem masculine noun, then B-Fixed Suffix stress should not survive because this prosodic profile is inactive in Declension II; the noun would be expected to change its stress pattern and acquire the D-Suffix-Stem pattern characteristic of Declension II feminines.

2. If a common gender or a-stem masculine noun does in fact retain its B-Fixed Suffix stress, then the morphology would be expected to change and to favor Declension Ia exponents where B-Fixed Suffix stress is represented, even though the inflectional suffixes would then necessarily have to be stressed.

3. If the [male] referential gender feature is associated with some Declension Ia morphology, then male referent nouns with A-Fixed Stem stress would be expected to innovate C Stem-Suffix stress but not D-Suffix-Stem stress.

And this is exactly what we find. A count based on the 2008 Hramatychny sloŭnik nazoŭnika shows that while most common gender and a-stem nouns have A-Fixed Stem stress, approximately 7% of them have the B-Fixed Suffix or D-Suffix-Stem patterns: Type B in ljawšá ‘left-handed person’, murzá ‘petty nobleman’, mulá ‘mullah’, xanžá ‘bigot, hypocrite’, tamadá ‘toastmaster’, razmaznjá ‘nincompoop’; Type D stress in suddzjá ‘judge’, staršyná ‘sergeant’, staršynjá ‘chairman’, sarvihalavá ‘daredevil’, siratá ‘orphan’, skupjandá ‘cheapskate’. Some of these nouns do exhibit additional changes, either in colloquial speech or as new accepted variants.

According to the first prediction above, if a common gender or a-stem masculine noun keeps its Declension II morphology, then it should be more likely to lose its B-Fixed Suffix stress pattern and would be expected to shift to the D-Suffix-Stem profile which is the attractor pattern for Declension II. This is indeed a possible outcome for some speakers, and it is the solution cited by Dubina (p.c.) for mulá ‘mullah’, which retains Declension II morphology throughout its inflectional paradigm but changes from B-Fixed Suffix to D-Suffix-Stem stress.

According to the second prediction above, if the B-Fixed Suffix pattern is preserved, then we could expect some speakers to have some Declension Ia morphology for male referents even though the inflectional suffix would have to be stressed. And this does happen. Dubina (p.c.) reports that Instr sg [starʃɨn^jóm] is possible when the referent is a male. A rough Google search (December 31, 2016) gave 313,000 hits for the Instr sg case of staršynjój ‘chairman’ (with the expected Declension II Instr sg /-oj/) but also 107,000 hits for the new Declension Ia Instr sg suffix /-om/ in staršynjóm.^[28] Thus, Type B-Fixed Suffix stress can survive, but when it does, it favors masculine Declension I suffixes.

And given that both A-Fixed Stem stress and C-Stem-Suffix stress characterize Declension I, it should be possible to find both of these patterns as variants in a-stem masculine nouns or in a common gender noun with a [male] feature. We have a situation like this in dzjadz’ka which now exhibits two different stress paradigms and semantic extension (18).

Both paradigms show the relevant mixed declension (I and II) because stress is on the stem in the singular subparadigm in both. What changes is the overall paradigmatic stress profile, and it changes within the limits characteristic of Declension I, permitting both A-Fixed Stem stress and C-Stem-Suffix stress.

Finally, note that Declension I inflections are found only in the Dat sg, Prep sg, and Instr sg cases while the Nom sg, Acc sg, Gen sg retain Declension II exponence. It is not clear whether we are seeing an intermediate stage of a change in progress or whether this mixed declension variant is now stable. The change in oblique case forms and the retention of Declension II morphology in the Acc sg and Gen sg may be a consequence of frequency, with the more commonly used (perhaps memorized) case wordforms in the Nom sg, Acc sg, and Gen sg being more resistant to change (Bethin 2018: 210–211). The other, less commonly used cases are more likely to be generated as needed and then if a noun is [male], the semantic/referential gender marker points to Declension Ia.

Given that the direction of language changes in Belarusian nominal declension is one which foregrounds gender, whether it be referential/semantic or grammatical, it would not be unexpected to find that for some speakers gender eventually becomes more salient than stress in the a-stem masculines. Thus, as already mentioned, the noun staršyná ‘chairman’ with B-Fixed Suffix stress does show Declension Ia exponence in the Instr sg case in spite of its stressed suffixes. And it is also interesting that the D-Suffix-Stress noun suddzjá ‘judge’, whose stress pattern is associated with Declension II, had 28,500 hits in the same Google search of the expected Instr sg suddzjój, consistent with Declension II morphology, but there were also 3,070 hits with the Ia masculine suddzjóm.

4.4 Some theoretical implications

It is not clear how morphological theory would account for the type of stress-controlled morphology exhibited by common gender and a-stem masculine nouns in Belarusian. The view of Network Morphology in Brown et al. (1996) is that “the patterns defined in the stress hierarchy can be understood as parasitic paradigms” (p. 75) dependent on declension class, and that “a noun must first change declension class before it can develop a stress pattern that would otherwise not be allowed by its own declension” (p. 105, Note 18). In Russian, for example, paradigmatic stress patterns may be seen as dependent on declension class which, as Brown et al. (1996: 73) write, “means that the stress hierarchy is parasitically indexed by the declension hierarchy. In this way the range of possible stress patterns that a noun, or adjective could have is delimited by its membership of a particular declension. This is important as it therefore characterizes the differences in stress patterns as constrained by a lexical item’s membership of a given declension class.” To some extent this also holds for Belarusian, where each declension class has a limited set of paradigmatic stress options. But the common gender and a-stem masculine nouns of Belarusian show that it is the position of stress which motivates morphological change and the adoption of some Declension I inflectional morphology. These nouns have not (yet?) fully changed declension class.

Unlike for Russian, where one can say that declension class determines gender (Corbett 1982, 1991), in the Belarusian common gender nouns it is referential gender and stress which determine declension class. There is a default to Declension II for both [male] and [female] nouns, but this default is clearly subject to the position of stress in the paradigm of [male] nouns. It is not clear where exactly this stress property would be specified in a theoretical representation when there is no obvious mechanism for stress to function in the selection of inflectional exponents.

In a Stumpian morphological analysis (Stump 2006), the innovations in common gender and a-stem masculine nouns create a heteroclitic paradigm, where certain case forms of the paradigm have linkage to another declensional paradigm. In Belarusian the basic paradigm is Declension II exponence with the intrusion of Declension I exponents in the Dat sg, Prep sg, and Instr sg for male referent nouns. This is non-canonical heteroclisis in that it is systematic, and it creates a type of fractured paradigm because only certain cases within the singular acquire the suffixes of a competing declension class. In this respect, it parallels the Czech examples cited in Stump (2006) where a-stem nouns such as sluha ‘servant’, kolega ‘colleague’, and předseda ‘president’ inflect according to Declension II feminines except in the Dat sg and Prep [Loc] sg forms where the exponents come from the Declension I masculine paradigm. However, the Belarusian innovation is unique and more complex: It not only depends on semantic/referential gender since only male referent nouns show this type of paradigm linkage, but it also depends entirely on stress because only those paradigms with unstressed case exponents in the singular can link to Declension I. It is not clear how this interaction of gender and paradigmatic stress properties would be formalized in terms of paradigm linkages; much would depend on how stress is assigned in the theory.

These developments also present a problem for frameworks where gender is either assigned to a noun before its phonological case exponents are spelled out, as in some versions of Distributed Morphology, or where it is said to be impoverished or absent in common gender nouns (Embick and Noyer 2007; Halle 1997b; Halle and Marantz 1993, 1994; McGinnis-Archibald 2016, and many others). The changes in Belarusian support the position that common gender nouns do not have a grammatical gender feature (Matushansky 2013; Steriopolo 2018) and that the referential [male] or [female] features (or “contextually determined gender” per Steriopolo 2018: 331) are active in determining the inflectional exponents. It is not entirely clear whether gender is distributed in two syntactic positions (Steriopolo 2018: 332), as discourse (or referential) gender (D-gender) and semantic (or natural) gender (n-gender), but what is interesting is that common gender nouns with referential (or discourse) gender, such as plaksa ‘crybaby’, as well as the a-stem masculine nouns with semantic (or natural) gender, such as mužčyna ‘man’, both follow the same stress-based morphology in Belarusian. If we accept the revised Distributed Gender Hypothesis of Steriopolo (2018: 332), then whatever mechanism is devised to deal with this type of morphology, it would minimally have to encompass both versions of gender or both branches of the syntactic structure, D-gender and n-gender.

The question is how to incorporate the role of stress in the morphology of these nouns. The overwhelming majority of common gender nouns and 135 of 149 a-stem masculines in the 2008 dictionary show A-Fixed Stem stress. For this group, the place of lexical free accent (or stress) can actually be indicated in the lexical representation. Regardless of how we mark lexical stress in Belarusian, as a stress feature or as an ictus/accent or as a diacritic of some type, the morphology must somehow be able to refer to this phonological property in selecting inflectional suffixes. But a few a-stem masculines show the C-Stem-Suffix pattern, e.g., dzjadz’ka ‘fellow’, a pattern normally characteristic of an unaccented root in the lexicon, so the morphology would somehow have to be able to target these nouns as well. And common gender and a-stem masculines with B-Fixed Suffix stress and D-Suffix-Stem stress would have to be specified for or made to default to Declension II morphology. Clearly, stress plays a role in determining inflectional exponents, it is just not clear exactly where this type of stress behavior is indicated in the morphology.

Finally, there is the question of determining what a declension class is. If we don’t want to consider a paradigm split subject to stress and gender, one possibility is to say that the stem-stressed nouns have homonyms. This is proposed by Lukashanets (2007) who takes nouns with two different inflectional options to be a subset of Declension II (pp. 153–158) and considers the male versus female versions of common gender nouns to be homonyms, e.g., plaksa ‘male cry-baby’ and plaksa ‘female cry-baby’ (p. 115; also reported in Eŭsievich 2015: 98). The female nouns then simply follow the declension of feminine nouns, while the common gender nouns referring to males are treated as a separate class; the two have separate declensional paradigms. But the fact remains that even this distinction would have to be made based on stress because nouns where the stress falls on the suffixes in the singular do not have two gender-based homonyms and different paradigms; they exhibit Declension II morphology regardless of semantic/referential gender.

Given the distinct behavior of the stem-stressed [male] nouns, one could raise the question of whether Belarusian morphology now has a separate declension class for male referent nouns with stress on the stem or whether these cases are simply a type of heteroclisis. Is the Belarusian situation like that of the Czech heteroclitic nouns but with more case involvement and with the special added constraint of stress? Or do the mixed paradigms now constitute a separate declension class of their own (per Lukashanets 2007)? Are we seeing a transitional stage of change in progress toward a complete shift of the [male] nouns to Declension I or is the current state with the mixed declension a stable one? This and other questions remain for future work.