The Genitive-Accusative of the Personal Pronouns in Old Church Slavonic

In those Slavic languages that retain both a case system and clitic pronominal forms two case-related phenomena partially overlap: (1) Masculine animate nouns and gendered pronouns display differential object marking with sensitivity to the animacy hierarchy. Some subset of these forms with the highest score on the animacy hierarchy show the original genitive form instead of the expected accusative in contexts that otherwise call for that case, the so-called genitive-accusative. (2) Personal pronouns also show instances of the genitive for the accusative but with important differences. In languages with a clitic~stressed contrast for oblique pronominals the accusative forms generally are continued as clitics and the genitive forms as stressed. It is unlikely that the nominal and personal-pronominal gen.-acc. are unrelated. On the other hand, the case choice for nouns and gendered pronouns is sensitive to the animacy hierarchy, but for the personal pronouns the choice between genitive and accusative is phono-semantic. Whatever semantic structure evokes the stressed forms leads to the production of the gen.-acc. I suggest that gen.-acc. began with o-stem masculine personal names, the most prototypical expression of the semantic class [+human, +male, +free, +definite] and was extended to the interrogative pronoun (gen.-acc. kogo). The interrogative pronoun had just those properties that allowed the remapping of an animacy hierarchy into a tonicity distinction.


Introduction
In those Slavic languages that retain both a case system and clitic pronominal forms we find two case-related phenomena that seem partially to overlap.1 Masculine animates and gendered pronouns display differential object marking (dom; see Bossong 1984;Mardale 2008) with sensitivity to the animacy hierarchy (Silverstein 1976;Aissen 2003). Some subset of these forms with the highest score on the animacy hierarchy show the original genitive form instead of the expected accusative in contexts that otherwise call for that case.
(1) OPol. 14th cent. The Genitive-Accusative of Personal Pronouns Personal pronouns also show instances of the genitive for the accusative but with important differences. In languages with a clitic~stressed contrast for oblique forms of the personal pronoun the accusative forms generally are continued as the clitic forms and the genitive forms are the stressed ones.

Similarity, Differences, and Questions
Similarity: The use of the genitive in syntactic contexts otherwise requiring the accusative is rarely found outside of Slavic.5 Even Baltic, which shares the innovative genitive after negative verbs, does not show the Slavic phenomenon.6 It is unlikely that the nominal and personal-pronominal gen.-acc. are unrelated. Differences: The case choice for nouns and gendered pronouns is sensitive to the animacy hierarchy. The animacy hierarchy is irrelevant for the personal pronouns, which predominantly have human reference, and almost always have animate reference. In the personal pronouns the choice between genitive and accusative is phono-semantic. Whatever semantic/pragmatic structure evokes the stressed forms indirectly leads to the production of the gen.-acc. This is obviously not a factor for nominals. The personal pronouns have (or had) the contrast in the singular and plural, as is shown by the total replacement of the old acc. forms ny and vy with the exception of Bulgarian which has ni, vi as the enclitic dat./acc. contrasting with accented nas, vas.7 But nominal and gendered pronominal forms originally only made this contrast in the singular. Masculine animate nouns take the ending -a in all Slavic languages that retain case. And even in Bulgarian and Macedonian there are traces of -a in both the standard languages and in the dialects.8 Serbo-Croatian and Slovene preserve the original situation.
Questions: What is the original distribution of the gen.-acc. vs. acc. pronominal forms in ocs and Novgorodian and what can be reconstructed for ps? What is the connection between nominal and pronominal gen.-acc.? Where does the innovation originate and how did the reinterpretation happen?

The Slavic Personal and Reflexive Pronoun
Oldest Old Church Slavonic9 In the case of the dat. sg. it is clear from the two forms and the distribution of the shorter forms that there was an stressed~enclitic opposition. These clitic forms continue pie clitic forms. Cf. Ved. me, te, etc. The accusative forms11 given normally pattern as clitics, but they can occur in positions where true clitics would be excluded, e.g. clause-initial position.

The Distribution and Spread of the Pronominal Genitive-Accusative
Apparently the first change affecting this system was the introduction of gen.acc. forms (mene, tebe, sebe, nasŭ, vasŭ). In the oldest ocs texts they are very rare.19 17 The accusative forms are the norm after prepositions in ocs. For the 1st sg., for example, in the Gospel texts with one exception (j. 12.44, a, z, but even here m, o, v have mę) mę is the only form: vŭ mę (m. 18.6,18.21,Mr. 9.42,j. 7.38,11.25,11.26,12.44 11, 24v.6, 66r.26, 116v.25, 149v.24, 156r.8, 235r7, 20. 23 The form nasŭ in i věra tvoeja vŭ nasŭ da vŭzdrastetŭ et fides tua in nobis ut succrescat. "And may your faith grow in us" is almost certainly a locative. 24 van Wijk 1926:263 was suspicious of the antiquity of two of these cases. 25 For a recent discussion of focus types see Riester and Baumann 2013. In their approach corrective focus is merely an extreme case of contrastive focus. Contrastive focus differs from novelty focus in that while both generate a focus semantic value defined as a set of individuals of the same semantic type as the focused constituent, contrastive focus is (10)  The Birchbark Documents from Novgorod.41 According to Zaliznjak 2004Zaliznjak :106, 2008:137 there are basically no examples of gen.-acc. in personal pronominal forms in the oldest period and only very few from the end of the 12th century onward.42 Other Old Russian texts show an encroachment of genitive forms beginning with the plural. The fact that the genitive forms are quite uncommon in ocs or even unattested in early Novgorodian does not in itself mean that they are an innovation. The genitive forms occur only under very specific conditions, basically some kind of focus. The necessary conditions for this environment are presumably fairly infrequent and it is not surprising that they do not occur in the mainly very short early letters.
What is more significant is the fact that the old accusative forms were still viable options for the focus positions in ocs. There is only one Gospel passage transmitted unanimously in m, z, a, o and v with the genitive: All others have the accusative in at least one manuscript. This is good evidence that the gen.-acc. in pronominal forms is fairly recent.

41
I have chosen to focus on the Novgorod documents rather than on Old Russian manuscripts because they are less influenced by ocs norms and because they are of great antiquity and are directly transmitted. 42 An early example may be и позовало мене во погосто "and he called me into the churchyard." (531, 1200-1220).  (1) that in these Novgorod documents the gen.-acc. is observed earlier in the plural and (2) "the gen.-acc. -a in the declension of masc. sg. o-stems began to appear in the 2nd half of the 13th century and was fixed only in the 14th to 15th century" (citing Krys'ko 1993Krys'ko , 1994 The inference is that the nominal gen.-acc. is older than the pronominal one. Gendered pronouns behave like nouns rather than personal pronouns. In ocs the gen.-acc. for gendered pronominals is found, with a few exceptions, only in the singular (at j 5.21 z has the gen.-acc. sg. jegože but m and a have the acc. pl. jęže) and is sensitive to animacy.49 Different gendered pronouns have different rates of innovation. For the suppletive pronoun nom. onŭ, acc. i < *jĭ the accusative form with human reference is the most common. The gen-acc. jego is quite rare. For example in c there are 17 instances of i with human reference vs. only two cases of jego (13a 13, 13b 19). In a there are well over a hundred instance of i, the majority of them with human referent, vs. only 4 cases of jego (89a 28a, 107a 7a, 119b 10a, 119b 24a). The figures for tŭ 'this' , which has the accusative tŭ, are similar.
On the other hand, the replacement of the gen.-acc. for the acc. in the relative pronoun has gone much further and is practically the norm. In a there are 33 instances of jegože vs. 16 of iže some of which may have non-human reference. In the parallel passages in z m a and s there are 36 instances where gen.-acc. jegože is the unanimous reading of however many manuscripts attest the verse in question. There are only 3 cases where there is some disagreement, for which a number of different explanations are required. 48 A second instance of gen.-acc. клеветьника in 247 has the crucial final vowel restored. In addition we have моли Воньзда "ask Vnezd." (82, 1180-1200), на Домажировица (510, 1220-1240), на Ивана (897, 1120-1140). 49 There are  In the first sentence the gen. is regular after the negation. In the second, m has eže, but a egože. This may result from uncertainty about animacy of reference. The Greek original has the neuter hò, but the natural inference is that the Jews worship God.
In two other cases we may see the interaction of syntax and phonology. At j. 19.37 z has a gen.-acc. for the pronoun i after a preposition and the acc. for the relative pronoun: But m and a have the accusative of the pronoun and the genitive of the relative na n'ĭ egože probasę. The latter is what may be considered the general pattern since the accusative i is especially resistant to replacement after prepositions50 and, as we have said, the gen.-acc. is the norm for the relative. Thus we may suppose that if the gen.-acc. were to be introduced we would expect *na njego jegože and that this was replaced by na njego iže to avoid cacophony.51 Likewise at Mr. 14.71 z has eže, but m has egože. 50 na nĭ < *na n-jĭ is retained in the Serbian Church Slavonic Miroslav Evangelium. Pronominal accusatives survive into modern times with no animacy distinctions in Slov. vánj 'on him/it' , zánj, SCr. zȃ nj 'for him/it' (confirmed by Draga Zec). See Nahtigal 1961:213. The replacement of the acc. by the gen.-acc. seems to have come first to direct objects and only subsequently to prepositional objects. There are quite a few prepositional objects in the accusative with human referent in the Gospels, e.g. j. 6.29 vŭ tŭ, Mr. 10. 12 za n ' ŭ, and even with the relative pronoun Mt. 12.18 na n ' iže, j. In order to account for the pronominal facts we need to explain the following differences: (1) The genitive pronominal forms are found in both singular and plural, unlike the nominal forms. (2) The genitive pronominal forms are typically associated with focal stress or some other intonational marking. One and 52 van Tilburg 1988:595-596 argues that the observed extensions of the gen.-acc. to the plural pronominal forms either predominantly (Rešetar 1898:187 on 15th century SCr.) or earlier than the further extension to nominal forms supports Meillet's claim. 53 The East Baltic facts might be interpreted as pointing to a onetime accusative function of the genitive, but Old Prussian shows through its accusative forms mien, tien, sien, exactly matching ocs, that this cannot be old. 54 As one of the referees notes, this problem may be serious enough to rule out Kuryłowicz's account. In that case, the exact origin of the gen.-acc. -a is still a mystery to be solved. 55 Or maybe not. First and 2nd personal pronouns have exclusively animate reference (except in metaphorical contexts like speaking inscriptions!) but they do not necessarily refer to humans (we can address dogs and cats, for example, with 2nd person forms). only one form provides the bridge between nominal and pronominal forms, viz. kogo, the old genitive of the interrogative-indefinite kŭto.56 The nominal gen.acc. is used primarily to mark nouns that score high on the animacy hierarchy.
Kŭto is the only pronominal form to make an animate/inanimate contrast.57 The nominal gen.-acc. occurs only in the singular in ocs and or. The pronominal gen.-acc. is found in all three numbers.58 Kŭto is number-indifferent and thus provides a model for the extension of the pattern to the plural. Like the ostems, the nom. and acc. of kŭto would have been identical. Avoidance of this homonymy is often cited as a motivation for the nominal gen.-acc. innovation. (Kuryłowicz 1962 inter alios).59 But it is noteworthy that Novgorodian has an o-stem nominative in -e and accusative in -ŭ, and also the interrogative keto60 (and thus had no nom.acc. homophony).61 If Olander 2012 is correct that *-os# gave *-ə in Proto-Slavic which merged with ŭ in all forms of Slavic except Novgorodian where it merged with e, then a Proto-Slavic nom.-acc. homonymy could not have been a crucial factor in the creation of the gen.-acc. It seems to me that the vocative hypothesis (recently Kwon 2009) has not been adequately refuted. The main objection seems to be that the wholesale morphological replacement of the nominative by the vocative is typologically unparalleled (so Zaliznjak 2004:148), but this is exactly what appears to have happened in the Latin 1st declension nom. sg. in -ă.62 Unlike the personal pronouns, the gen.-acc. kogo is the only form found in ocs Gospels. (Diels 1963:97: "The accusative is lacking. It is always represented by the genitive kogo.") My examination of the Gospels confirms this. 56 The important role of the interrogative pronoun has been noted before, e.g. by Vondrák 1908:340: "Der Impuls zum Gen.-Akk. ging also wohl vom urslav. kogo aus. Da mit kŭto-kogo nach Personen gefragt wird, so sind davon zunächst auch die Personennamen tangiert worden und zwar aus dem angegebenen Grunde zunächst die o-St., welche männlicher Personen bezeichneten. " Huntley 1980:205. But these accounts take kogo as the starting point for the spread of the gen.-acc. to nouns. 57 The loss of a distinct feminine is in itself an innovation. For traces of the expected *ka throughout Slavic see Boryś 2005:226, s.v. każdy, Majer 2012 Cf. j. 6.64 kŭto sǫtŭ nevěrujǫštei [tínes eisìn hoi mḕ pisteúontes] "who they were that believed not" (an indirect question) where the Greek has an explicit plural. 59 See also Klenin 1987and Bratishenko 1998 Staraja Rjazan' 12 and Novgorod itself in the form ketŭ (892, Lehiste and Ivić 1986:207 report that in ten sentences introduced by kada 'when' , the average f0 peak was at 269Hz; the next word, which is comparable to the first word of a simple statement, had an average f0 peak of 235Hz followed by a regular decline. The f0 on the first word of the ten corresponding statements was 230Hz. This raised pitch is comparable to "emphasis" in a declarative sentence.66 Bulgarian shows the same pattern as can be seen from the Praat spectrogram with pitch track below: Serbo-Croatian and Bulgarian are not ocs, but we will never know ocs sentence prosody. Russian and Polish are similar. Thus it is not improbable that this intonation was characteristic of Proto-Slavic.67 This would explain why the early personal pronominal gen.-acc. show the distribution they do, as stressed (often focused) variants. In particular, the extension from q > a seems natural. The gen.-acc. pronominals are employed in an intonationally marked context, predominantly focus. The gen.-acc. originates in o-stem personal names (as prototypically free, male, definite referents) in direct object function and spreads from there first to the interrogative pronoun already in Proto-Slavic. From these two bases the gen.-acc. spreads along two vectors. On the one hand, weiss Indo-European Linguistics 3 (2015) 118-144 the gen.-acc. is extended as a marker of higher animacy to the gendered pronouns (first of all to the relative pronoun) and to o-stem substantives with human referent. These extensions were not complete in Proto-Slavic and there are significant traces of the old accusative forms in these categories in the historically attested Slavic languages. On the other hand, the extension from the interrogative gen.-acc. kogo 'whom' to the personal pronouns crucially involved a re-mapping of the animacy distinction into a tonicity distinction. This extension too was only incipient in Proto-Slavic. The creation of new distinctively tonic personal-pronominal forms accelerated the downgrading of the old tonic forms to clitics.69 The cliticization of old tonic pronouns and their replacement by new tonic forms is a pattern that reoccurs in a wide variety of Indo-European languages (e.g. Latin and Romance) and which perhaps deserves to be called a close cousin of Jespersen's Cycle.