Scholars of structural linguistic orientation explored the possibility of decomposing grammatical case into smaller semantic features. Each grammatical case would be in this way characterized by a group of features that enter into syntagmatic and paradigmatic relations. The presence of common features among different cases would allow for replacing one case with another when they share most features, or extending the uses of a case by addition or loss of the relevant features.

One of the most renowned examples of this system is precisely decomposition of Russian case, proposed by Jakobson (1984 [1936]), and revised later by Franks (1995). These authors do not specifically address the alternation in case marking under study in this paper, but they do examine a related morphological syncretism, namely, the conflation of genitive and accusative morphological cases on animate objects (masculine singular and all plural). This conflation is illustrated in (6b):

According to these authors' system of case decomposition, genitive case would consist of the features [+oblique, –marginal, –non-ascriptive], while accusative would be defined as [–oblique, –marginal, –non-ascriptive]. In order to obtain (6b), they just erase the distinction between the two forms by the feature [oblique] in an operation that equals both forms with the characterization [–non-ascriptive, –marginal], and renders the morphological syncretism we observe in the language at the synchronic level.

Hypothetically applying the same system at the diachronic level, we could also claim that accusative and genitive morphological cases can alternate by erasing their [oblique] feature in the relevant context, leaving both forms with equal features ([–non-ascriptive, –marginal]). This operation would render the alternation between genitive and accusative in the complements of the verbs discussed in this paper.

This proposal is very appealing, at least, if we settle for a basic morphological description. Nonetheless, the mechanism of decomposition of grammatical cases—and related morphological operations—is just descriptive, and maybe not so persuasive on the basis of independent evidence. At times, the correspondence of a case to the alleged underlying semantic features is not very informative; for instance, in the specific alternation under study here, the characterization of the genitive and accusative cases does not capture the semantic values of avoidance and potentiality, which clearly differ from the usual values of these two cases in other parts of the grammar.

Another system inspired by the theory of case decomposition is the more recent idea of representing the semantic values underlying grammatical cases with the help of the so-called “maps of semantic notions.” These maps include the various semantic interpretations of the cases existing in a language or a group(s) of languages, and depict the higher or lower plausibility of syncretism or transfer between cases through the representation of the “geographic” distance between the different values.

The closest study to the phenomenon discussed in this paper is to be found in Clancy (2006), based on Haspelmath (1997). He offers a topology of Slavic case with multidimensional scaling, in which the distance between different functions or semantic values intends to capture frequencies of use, markedness of the variants, and possible changes and syncretisms. Thus, broadening or restricting the “meanings” or semantic values of a specific case should correspond to contiguous or related areas on the semantic map.

Clancy (2006, p. 24) captures the relationships between the semantic values of Slavic case in such a map of semantic notions, depicting the “distance” between those values. Such an approach is interesting from the point of view of case morphology, and in this specific study, it is very detailed. However, as it stands, it is of no use for our analysis, as the semantic notions on Clancy's map associated with the alternating cases addressed here (“dist from/afraid of,” which stands for the ablative value of the genitive case, including verbs of fear, and “understand,” which stands for regular direct objects) are too far away from each other to accommodate our variants and hypothesize a possible transfer between them. This can be due to the fact that Clancy (2006, p. 25) himself acknowledges that the dataset used for this specific map was a pilot one and thus incomplete, but in any case, such a representation of the semantic values of case is purely descriptive and does not explain the reason for the variation phenomenon under study.

Construction networks can be defined as a descriptive tool developed within the Construction Grammar, a functionalist approach to languages. Within this approach, Nesset and Kuznetsova (2015a,b) have addressed the specific phenomenon discussed in this paper. More specifically, the authors count the occurrences of the accusative and genitive variants associated with several –sja verbs in the Russian National Corpus and aim to account for the asymmetries in the use of these alternating variants.

As introduced in the previous section, they find differences depending on the register (whether the utterance is part of a corpus or a spontaneous production), the age of the speaker (accusative more frequent in younger speakers), and the specific lexical item. Interestingly, the verbs more often found in combination with an accusative complement are, according to their search in the National Corpus, the verbs slušat'sja “to obey,” bojat'sja “to be afraid,” dožidat'sja “to wait,” dostigat' “to reach,” and izbegat' “to avoid;” there are also differences among them, namely, the verb slušat'sja and after it, dožidat'sja, are much more frequently combined with accusative than the rest (Nesset and Kuznetsova, 2015a, p. 371–373). We will come back to this fact in the following section.

The main hypothesis in their papers is that a high level of individuation or animacy favors accusative case marking, as happens in other cases of alternation between genitive and accusative case (namely, the syncretism of animate NPs in object position, to which we will return later on). All other factors (declension class of the noun, the intensional or directional semantics of the verb, the “opacity” of the –sja suffix) are relegated by them as epiphenomenal.

After describing the conditions for case alternation, Nesset and Kuznetsova accommodate all the relevant variables in a paradigm or construction network, which can be defined as the representation of a specific construction and of some of its subtypes, together with the relevant features, such as markedness of choices, statistical significance, possible diachronic changes, variation, etc. Nesset and Kuznetsova (2015a, p. 388) provide such a construction network for three of the verbs involved in the alternation at issue. But, again, the statistics offered by these authors, as well as the intervening factors, are very illustrative of what is happening in the language, but their network is just descriptive. Another shortcoming of the account is that it overlooks the potential syntactic motivations behind the variants, which will play a fundamental role in the alternation, as we will see in the two final sections.

A final observation regards the imprecise semantic and syntactic characterization of the verbs these authors analyze. If we follow Timberlake's (2004, p. 317) classification of the Russian verbs taking lexical genitive nowadays (cf. previous section), slušat'sja, and dožidat'sja are weak intensional verbs (the only “potential” –sja verbs), bojat'sja and izbegat' are verbs of avoidance (one medial, the other active), and dostigat' is a potential verb with active form. The indistinct treatment of all these forms leads the authors to lump together verbs of different syntactic and diachronic behavior.

Bare genitive case associated with some active verbs was lost as soon as in prehistoric Slavic; other active verbs still display genitive object case in early Slavic. This shift affected several classes of active verbs, including the verbs discussed here, i.e., verbs traditionally classified in Indo-European linguistics as verbs of “separation” (avoidance) and verbs of “desire/achievement and perception” (potential).

In parallel to the loss of bare lexical genitive case, we observe in Middle Russian a significant development of the genitive form as a non-lexical case, which became either reinforced in structural positions previously existing in the language, or spread to new syntactic positions.

The structural positions undergoing the genitive/accusative case alternation are the following: (i) regular animate objects, and (ii) NPs governed by some quantificational or negative head. Let us see some examples of them.

(i) The alternation in regular objects arose in Russian with the extension of the genitive case marker to animate regular objects of the masculine singular declension (o-stems) and plural declension (all stems). The process of replacement of the old nominative-accusative form by a genitive form in the relevant animate objects started already in Old Church Slavonic (OCS) (19) and was completed in Early Middle Russian (Krys'ko, 1994). Inanimate objects belonging to these stems remained marked with nominative-accusative case (20):

In this way, learners started to be confronted with a consistent alternation between genitive and accusative cases in regular object position.

(ii) Other consistent alternations of a similar nature affected partial or partitive objects, quantified expressions, objects of negated verbs (the so-called genitive of negation), weak intensional verbs, cumulative verbs, and other similar prefixed quantificational verbs (with the prefixes do-, za-, pri-, na-; see Straková, 1961).

These types of bare genitive case are usually assumed to be licensed by some functional (rather than lexical) head and, therefore, “structurally determined” (Bailyn, 2004; Pereltsvaig, 2006; Kagan, 2013; etc.), thus giving rise to further genitive–accusative alternations in non-lexical positions.

The change experienced by the active verbs reviewed so far can be formalized in the following way: initially, these verbs had the ability to take genitive objects, as depicted in (22), corresponding to (16a), repeated below:

Parallel changes that were taking place in the language at that time (tied to a general typological change in the language) affected mainly two structures: (i) bare genitive adjuncts being replaced by overt PPs (23), corresponding to (11a); and (ii) bare genitive objects alternating with accusative forms in non-lexical (structural) positions (24–25), representing (19b) and (21), respectively.

Finally, the result of the change, the regular accusative object case pattern is represented in (26), corresponding to example (17):

The change described here can be interpreted according to the concepts outlined in the introductory theoretical section. First, it arises because learners are confronted with innovative pieces of data as part of the Primary Linguistic Data (PLD) they receive, up to a point when their grammar stops converging with the one that generated the relevant input (“discontinuity of transmission between generations”). Language acquisition is therefore a fundamental piece in this process.

Further, as described in the introductory theoretical section, the contingency of grammar change underlies also this specific alternation of the Russian language, as it was determined by the unpredictable alteration of the conditions of the genitive—accusative alternation in other parts of Russian grammar (often related to non-syntactic factors, such as morphological syncretisms).

On the other hand, there seems to exist some bias operating in the previous changes reviewed in this section. There is little doubt that the set of changes associated with a global shift in the basic word order of the language, ultimately responsible for the replacement of bare lexical cases by PPs, is recurrent also in other Indo-European groups of languages (see references above), and seems to respond to some sort of economy or efficiency factor; in this case, to a general tendency to unify the head directionality of the language.

The second phase in the loss of bare genitive case in Russian affected the verbs including a –sja suffix, plus the prefixed active verbs izbegat' “avoid” and dostigat' “reach,” to be treated as an exception in the final section. As introduced before, almost all the verbs of fear and avoidance (except izbegat' and dostigat') are suffixed –sja forms (bojat'sja “to be afraid,” storonit'sja “to avoid,” pugat'sja “to be frightened,” strašit'sja “to dread,” lišat'sja “to be deprived,” etc.). We include in this section also the weak intensional or “potential” –sja verbs slušat'sja “to obey” and dožidat'sja “to expect” (the only non-active representatives in their group nowadays).

The history of the –sja verbal suffix in Russian is reviewed in detail in Zaliznjak (2008). This suffix has its origin in the clitic sę/sja, a free morpheme that was in fact the accusative form of the reflexive pronoun (cf. se in Romance languages). As such, it could be the object of any active verb regularly taking an accusative complement:

In Old Church Slavonic and Old Russian, unlike in other Indo-European languages (e.g. Ancient Greek), the accusative clitic sę/sja filled the internal argument position until the sixteenth century (Madariaga, 2010). In Old Russian, these elements could behave as a second-position clitic (28b), or a weak pronoun, usually following the verb (28a), but also following other elements, such as a preposition (Zaliznjak, 2008, p. 36). As we see in example (28b), when the first position in the sentence was occupied by a verb, the clitic could look much like a non-second-position -sja to learners (28a), because sja, in both patterns, followed the verb:

This kind of input could eventually lead learners to reinterpret the free sja as an element associated to a verb. Thus, the free morpheme merged in Old Russian with the verbal form, grammaticalizing later as a verbal suffix (Zaliznjak, 2008). As a final step of this morphological process, the –sja suffix underwent phonological reduction into just a palatalized -s' in certain environments (in regular conditions, after a final vowel):

Coming back to the list of the –sja verbs alternating between an accusative and genitive pattern in Russian nowadays, we can easily notice that virtually all of them are just the –sja “counterpart” of one of the active verbs of avoidance (30a) or potential verbs (30b) reviewed in the previous section. Even nowadays their morphological formation is fully transparent in most cases: the suffix –sja/-s' is just attached to the active form:

Only the verb bojat'sja “to be afraid” (Old Church Slavonic bojati sę) did not correspond to an active verb as such, although prehistoric stages of the language probably displayed an active equivalent as well. Its active counterpart can be traced back to the proto-Slavic form ^*bojati, not attested as such in historical Slavic, but related to equivalent Sanskrit or Baltic forms. All other verbs of avoidance (30a) display from Middle Russian an active form taking an accusative object, and a –sja form taking a genitive object (the one that has recently start to alternate with accusative).

As for the forms in (30b), at the beginning of the twentieth century, by Peškovskij's (2001 [1938]) time, they had two interesting properties: (i) they were the only members of their lexico-semantic families preserving genitive case (not having changed into an accusative pattern or an alternating pattern determined by the semantics of the object), and (ii) they were the only members of their families with the –sja suffix.

Now, after having surveyed all the relevant data, I will propose a formalization of this shift and explain why this change is still taking place nowadays, whereas their active counterparts changed four centuries ago.

The original pattern included a free pronominal sja element in object position, which could behave as a second-position clitic, as in Old Church Slavonic (cf. example 27). The “avoided” element (i.e., the element causing fear) was associated with the semantics of separation, and marked with genitive (<ablative) case or an overt PP, as usual in most early Indo-European languages. This is illustrated in (31), representing (28b):

In Old Russian, the free morpheme sja had the possibility of staying in a lower position and being attached to the verb. This initial short movement prior to reanalysis is represented in (32), corresponding to (28a), in which the pronoun sja does not behave as a second-position clitic, but surfaces attached to the verb.

By this time, the complement slot was still occupied by an overt element, which, by virtue of Burzio's (1986) Generalization, banned accusative assignment to any other possible object. The incorporation of sja into the verbal form, which represents the initial step of the change under study here, was completed in Middle Russian (Zaliznjak, 2008, p. 217ff), but it did not automatically convey any further change in the structure at this stage.

Later on, the sja element lost its ability to behave as a clitic, and became fully incorporated into the verb in a very common diachronic process classically known as grammaticalization. These kinds of processes have been described in generative accounts as up-the-tree movements, followed by the reanalysis of the element initially moved as base-generated in the landing position (see Roberts and Roussou, 2003). Again, this grammaticalization process was a necessary previous step for later reanalysis, but still did not involve any major change in case assignment. The verbs affected were still acquired as “exceptional” in that their complement was marked with quirky genitive case.

In some speakers, however, at some point in the recent history of Russian, after the morpheme –sja started to be base-generated in V, the whole element in this position could start to be perceived as a “deponent” verb. In other words, the complement slot became free, and the historically “disturbing” bare quirky genitive could finally be reanalyzed as a regular accusative object, merged as a complement of the verb. This is depicted in (34), corresponding to (2a), repeated below:

The nature of the shift represented in these structures evidences the fact that the ultimate reason for the alternating case patterns with –sja verbs, changing recently in Russian, was in fact the reorganization of bare lexical cases four centuries before. In Middle Russian, active verbs taking a genitive complement changed into an accusative pattern (or a semantically determined alternating pattern in the case of weak intensional verbs). But –sja verbs behaved in a different way. They preserved a quirky genitive complement longer because at the crucial moment of the reorganization of the bare case system in Russian, they still fell under Burzio's Generalization; i.e., the complement slot was still filled by the element sja, and learners were not able to replace the “disturbing” genitive NP with an accusative NP, as they did in the case of active verbs.

Learners did not have any other option than acquiring the quirky pattern as “exceptional,” as it is still acquired nowadays, i.e., by means of some special morphological rule assigning genitive case to the relevant objects at the Externalization component of the language (Sigurðsson, 2012). This is also in line with the theories about competing grammars coexisting in a single speaker at different linguistic levels, as stated in the introductory theoretical section.

However, after the whole verbal form was reanalyzed as a unique element merged in V, freeing up the complement slot, reanalysis of the “avoided” element as a regular accusative object became available, which is in fact what eventually happened in colloquial language. Again, as explained in the introductory theoretical section, contingent unpredictable conditions determine here the possibility for a syntactic phenomenon to undergo change in a regular way, or the need to “wait” until something else happens in the language, making the conditions for change favorable. The case addressed in this paper is of special interest, because it also confirms the idea that variation can correspond to successive discrete changes spreading further to new syntactic environments.

As said before, the seemingly gradualness of a change in progress can often correspond to a diversity of linguistic environments successively affected by one unique change. Here too, what seems as a gradual change can be reduced to a series of discrete changes affecting different items or structures at different moments, according to a “third factor” effect, namely, Input Generalization (“maximize available features”); see the introductory theoretical section. If change spreading proceeds in this particular way, we expect the presence of different “splits” between competing variants according to different features, structures, or lexical items. These considerations also apply in our case study.

The major split between the alternating patterns at issue was between active and –sja verbs, which have been shown to feature a very clear structural contrast (the availability or not of a free complement slot in the structure).

But other minor splits in this process must also be taken into account, namely, those determined by (in)animacy and declension classes⁸. Some speakers favor accusative case only when the object of the verb bojat'sja conveys an animate feature (cf. footnote 2); others reject accusative assignment when the object belongs to III declension class, even if it is animate. These patterns are illustrated in (35):

Splitting alternating patterns according to some specific semantic feature or morphological class is a recurrent way of pinpointing a change process. In the specific case of Russian, animacy and declension class have played this role before: animacy determine the case patterns for masculine singular I class, and all plural objects (see example 6).

Other speakers, however, have gone further in this process and are able to use accusative case almost regardless of the animacy feature of the object, whether of class I or II (36a-b). Some are tolerant with III declension class objects, too (37a-b)⁹.

A final observation on splits in the alternating patterns concerns the spread of the new accusative form according to different lexical items, thus rendering again a succession of discrete changes, as explained in the introductory theoretical section of this paper. As expected, more frequent verbs are more prone to be used with accusative case than others. As noted by Nesset and Kuznetsova (2015a), there are differences between them even in the case of frequently used verbs; namely, slušat'sja and dožidat'sja are changing faster than bojat'sja. This correlation is perhaps not random: as shown in (30b) above, both slušat'sja and dožidat'sja have active counterparts, which had changed into an accusative pattern earlier, while bojat'sja never had one. This fact suggests that bojat'sja was maybe less prone to Input Generalization, and therefore more resistant to change, regardless of its frequent use in the language.

To conclude this section, let us now recall the only active verbs that exceptionally preserved bare genitive objects in the Russian language: dostigat' and izbegat'. Although they are active, these verbs behave as –sja verbs in the sense that they started to change later, and nowadays display, in principle, the alternation addressed in this paper, as shown in (40):

These two verbs are prefixed forms in their basic form, unlike the other active verbs that changed early in the language from genitive into accusative case. On the other hand, they are the only ones lacking a -sja counterpart; in other words, like bojat'sja (this one lacking an active counterpart), they did not enter the alternation active–medial illustrated in (30) above.

Their special morphology could probably help them to preserve bare genitive case marking as not so “disturbing” to be acquired by learners. In fact, we have independent evidence that bare genitive case was preserved until late Middle Russian, when it was associated with prefixed verbs (Černyx, 1952). The examples in (41) include a verbal prefix, called preverb in the indoeuropeanist tradition, which is identical to the “missing” preposition, making it easily recoverable¹¹.

Likewise, the prefixes in dostigat' “reach” and izbegat' “avoid,” as well as the lack of voice alternation, could contribute to the longer preservation of the bare genitive complement of the verbs. Again, this development is in line with the introductory notions about successive realizations of one unique change in different morpho-syntactic conditions (see the introductory theoretical section).