Sandhi phenomena and language change

27
studia grammatica Herausgegeben von Manfred Bierwisch unter Mitwirkung von Hubert Haider, Stuttgart Paul Kiparsky, Stanford Angelika Kratzer, Amherst Jurgen Kunze, Berlin David Pesetsky, Cambridge (Massachusetts) Dieter Wunderlich, Diisseldorf

Transcript of Sandhi phenomena and language change

studia grammaticaHerausgegeben von Manfred Bierwisch

unter Mitwirkung vonHubert Haider, Stuttgart Paul Kiparsky, Stanford Angelika Kratzer, Amherst Jurgen Kunze, BerlinDavid Pesetsky, Cambridge (Massachusetts) Dieter Wunderlich, Diisseldorf

studia grammatica 41

Ursula Kleinhenz (Ed.) Interfaces in Phonology

Akademie Verlag (^ Ip

Die Deutsche Bibliothek - CIP-Einheitsaufnahme Interfaces in phonology / ed. by Ursula Kleinhenz. - Berlin : Akad. Verb, 1996

(Sludia grammatica ; 41)ISBN 3-05-002964-1

NE: Kleinhenz, Ursula [Hrsg.l; GT

ISSN 0081-6469

© Akademie Verlag GmbH, Berlin 1996Der Akademie Verlag ist ein Unternehmen der VCH-Verlagsgruppe.

Gedruckt auf chlorfrei gebleichtem Papier.Das eingesetzte Papier entspricht der amerikanischen Norm ANSI Z.39.48 - 1984 bzw. der europaischen Norm ISO TC 46.

Alle Rechte, insbesondere die der Ubersetzung in andere Sprachen, vorbehalten. Kein Teil dieses Buches darf ohne schriftliche Genehmigung des Verlages in irgendeiner Form - durch Photokopie, Mikroverfilmung oder irgendein anderes Verfahren - reproduziert oder in eine von Maschinen, insbe­sondere von Datenverarbeitungsmaschinen, verwendbare Sprache iibertragen oder iibersetzt werden. All rights reserved (including those of translation into other languages). No part of this book may be reproduced in any form - by photoprinting, microfilm, or any other means - nor transmitted or translated into a machine language without written permission from the publishers.

Druck und Bindung: GAM Media GmbH, Berlin

Printed in the Federal Republic of Germany

Contents

Introduction............................................................................................................................ ix

The Phonology-Syntax InterfaceMarina Nespor, Theresa Guasti and Anne ChristopheSelecting Word Order: The Rhythmic Activation Principle..............................................1

Zhiming BaoLocal Tree Geometry and the Phonology-Syntax Interface..............................................27

S.J. HannahsPhonological Structure and Soft Mutation in Welsh........................................................46

Loren BillingsSandhi Phenomena and Language Change.........................................................................60

Paola MonachesiOn the Representation of Italian Clitics........................................................................... 83

Sharon PeperkampOn the Prosodic Representation of Clitics...................................................................... 102

The Phonology-Morphology InterfaceSharon InkelasDominant Affixes and the Phonology-Morphology Interface....................................... 128

Rene KagerOn Affix Allomorphy and Syllable Counting................................................................ 155

Chris GolstonProsodic Constraints on Roots, Stems, and Words........................................................172

Renate RaffelsiefenGaps in Word Formation................................................................................................. 194

Sandhi Phenomena and Language Change

Loren A. Billings (Florida State University)*

Language change occurs incrementally. In this paper I show that one such diachronic phe­nomenon, so-called epenthetic n in Russian, has been conditioned by prosodic and syntactic constraints over the past 1000 years of orthographic record. Originally an instance of resyllabification, this phenomenon has progressed to a point where phonology is no longer a factor in determining this phenomenon’s distribution.

Central to this paper is the notion of minimal structure. Before presenting the data it might be useful to illustrate this notion: One instance of the notion minimal word in Eng­lish has to do with the distribution of verbs with irregular inflection. Without exception, only verbs with one-foot roots in English can undergo stem-internal changes like the ones in (1). Foot-sized roots, such as (la), can have irregular past-tense inflection; roots consisting of more than a foot, as in (lb), are not attested with irregular inflection.

(la) run/ran (lb) succumb/succumbed (lc) begin/began [McCarthy & Prince (1991:1);

Note that prefixes are not considered for weight, even when a verb can never stand without a prefix, as in beginJbegan in (lc). That is, -gin, although never attested without a prefix, is nonetheless the constituent that must be no larger than a single foot in size.

These introductory data show that language can refer somehow to a constituent’s pro­sodic size. I now turn to Russian prepositions for the remainder of the paper, where I show that similar types of reference to constituent size exist as well. This phenomenon actually takes place in all but one of the non-Balkan Slavic languages.* 1 I use Russian data, but col­leagues familiar with the other languages which have undergone this process - for example, Czech - report that, by and large, the facts below are the same for those languages as well. The historical-Russian data I use come from Hill (1977), a detailed chronology of //-initial third-person pronouns in Russian.

* Thanks for useful comments to L. Babby, S. Baehr, G. Fowler, K. Goeringer, E. Kaisse, U. Kleinhenz, K. Krivinkovd, O. Orgun, A. Prince, G. Rappaport, C. Vakareliyska, W. Wurzel, O. Yokoyama, an anonymous reviewer, and GURT and LSA audiences (cf. Billings 1994a; 1995a). Partial travel support allowing me to present this paper is gratefully acknowledged from the Slavic Languages Department, Princeton University (where 1 was enrolled), and from the Department of Chinese, Comparative Literature and Slavic Languages, Rutgers University (where I was employed). Thanks especially to Steven P. Hill for sharing his own printouts and expertise to supplement the data in his (1977) book.

1 The /i-initial form has been generalized to all (non-clitic) uses of this erstwhile ./-initial stem in all the Balkan Slavic languages (Slovene, Serbo-Croatian, Macedonian and Bulgarian), regardless whether there is a preposition. In Belarusian the opposite extreme is attested: n-initial forms have been lost altogether, even where there is a preposition. Cf. In the remaining Slavic languages the n-initial form is attested only where there is a preposition. Cf. Ferrell (1958) for the details.

Billings, Sandhi Phenomena and Language Change 61

Throughout the past millennium, during which the Russian language has had textual evidence, there has been a distinct sandhi phenomenon whereby a preposition and a third- person pronoun have an intervening nasal consonant. In (2a-d) I show the development through the historical period of Russian.2

(2a) OldRussian:

(2b) MiddleRussian:

(2c) ContemporaryRussian:

(2d) ModemInnovation:

ku nemu ku nemu ku nemu ... ku ... nemu .

PP PP

The first period - until ca. 1600, which I will call Old Russian - was subject to a strictly prosodic constraint: /z-initial forms are attested in only the maximally simple prosodic en­vironment: namely, when a preposition and pronoun together constitute a single prosodic word (PrWd). Note that nothing in (2a) refers to syntax. I will add crucial evidence to sup­port these prosodic models in Section 2 below.

The second historical period - between ca. 1600 and ca. 1800, which I will call Middle Russian - is characterized by two criteria: n-initial forms are restricted to the maximally simple syntactic environment. The preposition and pronoun in (2b) must constitute the minimal syntactic structure, which is a prepositional phrase with no other constituents aside from the preposition and pronoun. I discuss this period in Section 3.

The third period - since ca. 1800, which I will refer to as Contemporary Russian - is characterized by a more relaxed syntactic restriction: epenthesize n only if the pronoun is adjacent to the preposition and that pronoun is the head of the preposition’s complement- NP,3 as shown in (2c).

A modern innovation, attested since early this century, is shown in (2d). In this dia­chronic stage the pronoun must be head of the preposition’s complement-NP, but the preposition and pronoun no longer need to be adjacent.

I provide additional evidence below to clarify these three stages in the following sec­tions. I conclude the paper with an Optimality-theoretic analysis of these restrictions.

2 The prosodic and syntactic minimal-structure constraints in (2a-b) are originally from exx. 21a-b in Billings (1993:4). The head requirement in (2c-d) is due to L.H. Babby (lectures. Princeton, 1991).

3 Throughout this paper I use the term noun phrase (NP) to refer to the entire nominal expression that is the complement of the preposition. My proposals require personal pronouns to head the nominal ex­pressions. Crucially, a possessive pronoun does not head the complement of the preposition. (This appears to be contrary to the proposed determiner phrase in Abney (1987). The following additional abbreviations are used in this paper: ACC(usative case), DAT(ive case), EMPH(atic particle), FEM(ininc), GEN(itive case), !NST(rumental case), LOC(ative case), MASC(uline), P(reposition), PL(ural), PREP(ositional case), PP: prepositional phrase, PrWd: prosodic word, SG: singular.

62 Billings, Sandhi Phenomena and Language Chance

1. Common Slavic: Syllabic-Structure ConstraintsIn this section I recount the initial phonological incident that led to the sandhi phenomenon discussed in this paper.

1.1 Initial Syllabic Re-AnalysisThis phenomenon originally arose due to the so-called Law of the Open Syllable (prohibiting syllable-final consonants). Three prepositions, *vun ‘in’, *sun ‘with’, and *kun ‘to(ward)’ plus any front-glide-initial pronoun were reanalyzed as shown:4

(3) *ku jemu —> ku. ne . mu ‘to himoAT’

Whereas in most environments a syllable-final nasal consonant caused nasalization in the preceding vowel and was itself dropped, the environment shown in (3) - at the border be­tween a proclitic preposition and a glide-initial pronoun - allowed the final nasal consonant of the preposition to become pronounced along with the initial glide of the pronoun to form a complex n consonant.5

1.2 Generalization to Other PrepositionsAside from the three prepositions which originally ended in a nasal consonant, every other preposition ended in a vowel, and as such, resulted in an open (final) syllable. Perhaps be­cause of the sheer frequency of the three nasal-final prepositions (*vun ‘in’, *sun ‘with’, and *kun ‘to’), these other prepositions also began to take nasal-initial pronouns, as in example (4):

(4) *po jemu —> po . ne . mu ‘according-to himoAT’

This generalization to the other prepositions likewise took place prior to any Russian (or East Slavic) textual evidence, over a thousand years ago.

1.3 Suppletive /j-/ and /n-/ StemsThe data so far might be accounted for by generalizing either (i) a final nasal consonant generalized to all prepositions - i.e., */pon/ ‘according-to’ - or (ii) suppletive pronoun stems - both /j-/ and /n-/ ‘him’ - with the latter variant used after prepositions (the

4 Syllable breaks are shown with dots; h-pf. The symbol 0 is a so-called jer, historically a weak (back) vowel that alternates between 0 and a full vowel in Modem Russian (cf. Carlton 1990:171-172 for de­tails). This vowel was still consistently realized during the period when the change in (3) took place.

5 For some reason, n was not maintained consistently prior to glide-initial nouns. This might be because the third-person pronominal stem, which consists of the glide plus an adjectival ending, formed a cohesive and sufficiently frequent class of prepositional complements. The fact that only pronouns consistently maintained separate n- and ./-initial forms suggests that the two pronominal forms became re-analyzed as suppletion - both /n-/ and /j-/ stems (plus so-called pronominal-adjectival inflection)..

Billings, Sandhi Phenomena and Language Change 63

erstwhile nasal-final prepositions losing their final nasal consonant: /vu/ ‘in’, /su/ ‘with’, and /ku/ ‘to’). The proposal in (ii) is preferable for the following reasons:

First, following the period in Common Slavic when all syllables were required to be open, the vowel u (along with its front counterpart: 1) was dropped in most environments - the so-called Fall of the Jers. In East Slavic, according to Carlton (1990:171-72), this change took place during the late 1100s and early 1200s. This, for example, caused ku.ne.mu in (3) to change into kfiemu (syllabified as khe.mu ) and caused other, canonically u-final prepositions to simply lose their final consonants (e.g., bezu ‘without’ became bez, not *bezun > *bezn). That is, any third-person pronoun in Old Russian (and earlier) which was governed by a preposition was nasal-initial, regardless of whether the preposition ended in a vowel, a nasal consonant, or (later) some other consonant.

An additional argument for proposing separate underlying /j-/ and /n-/ stems is the be­havior of these pronouns in later periods (discussed below). Several centuries ago non­nasal forms began appearing in certain prepositionally governed environments. In this century nasal-initial third-person pronouns are attested even if the preposition does not immediately precede the pronoun (cf. Section 5 below). Such developments confirm that there were separate stems /j-/ and /n-/ stems (and not underlyingly nasal-final prepositions). To be precise, I am arguing that there is suppletion of these two stems.6 Unlike most suppletion, however, this phenomenon is not morphological allomorphy (i.e., differing stems depending on inflectional agreement), but rather a syntactic distinction: distinct forms to reflect the category of the governor.

2. Old Russian: Strictly Prosodic RestrictionsUp to about the year 1600, the period which I refer to as Old Russian, the distribution of the /j-/ and /n-/ stems was conditioned strictly by prosody: The /n-/ stem is attested as long as a preposition is procliticized to it. In this section I justify a prosodic-minimality constraint to account for this distribution. I discuss the introduction of two types of data: prosodically heavy prepositions and possessive pronouns.

6 Specifically, this is partial suppletion. Spencer (1991:8) distinguishes partial suppletion - “the conse­quence of a phonological change in the language that happened a long time ago [... in which] the variants still bear a fairly strong resemblance to each other" - and total suppletion, in which historically separate stems have “absolutely no phonological connection” to each other (such as English gohvent) come to function as variants of a single stem. Ironically, the third-person personal pronoun in Modem Russian is further complicated by total suppletion: The nominative-case forms ona ‘she’, ono ‘it’, on ‘he’, and oni ‘they' are formed from the erstwhile demonstrative-pronominal stem /on-/, while the other case forms are formed from the stems /h-/ (with prepositions) and /j-/ (elsewhere). Both of these partially suppletive stems take adjectival inflection with so-called nonsuffixed pronominal-adjectival stress. (Cf. Levin 1978:69-70 for discussion. Strangely, Levin does not list this particular pronoun in this otherwise thorough description; in his notation the underlying representations would be /ij-V and /j-'/.) Prepositions never govern NOM-case pronouns, thus the totally suppletive /on-/ stem is not otherwise discussed in this paper.

64 Billings, Sandhi Phenomena and Language Chance

2.1 Heavy PrepositionsIn this subsection I present the relevant data and arguments to show that prosodically heavy prepositions, which by definition fail to procliticize to the next word, were prohibited in Old Russian from triggering nasal-initial third-person pronouns.

As of the first distinctly East Slavic texts, roughly 1000 years ago, the following inven­tory of prepositions consistently triggered nasal-initial forms of third-person pronouns: vu ‘in’, su ‘with’, ku ‘to(ward)’, o/obu‘abouf, u ‘near’, do ‘(up) to’, po ‘according-to/after’, za ‘behind’, na ‘on’, izu ‘out-of/from’, otu ‘(away) from’, pri ‘close to’, bezu ‘without’, nadu ‘above’, podu ‘under’, and peredu ‘before’. Note that (if the u vowel is counted) this initial inventory consists of ten monosyllabic prepositions, six disyllabic ones, and one which is trisyllabic (cf. also Hill 1977:298). Each of these prepositions apparently procliti- cized to the first prosodic word of its complement.7

Before proceeding with the chronology, I should outline the criteria for determining a preposition’s prosodic weight. The following factors bear on such a determination: First, it is impossible to reconstruct exactly how the textual data were pronounced. Next, due to the pitch-accent history of Slavic, the exact demarcation line between light and heavy is not based solely on prosodic quantity. Finally, it is possible that there were not any non­proclitic prepositions in Old Russian (and possibly even in Middle Russian). I now deal with each of these complications in turn.

First, one cannot recover exactly how written data from an extinct language period were originally pronounced. All secondary indicators, such as voicing assimilation (across the preposition-complement boundary) spelled overtly and writing the preposition together with the following word (i.e., without a space) indicate that the prepositions above have been proclitic throughout the history of Russian.

Next, what makes a preposition large enough to be a PrWd of its own? In Modem Rus­sian most monosyllabic and all less-than-syllabic prepositions must be clitics, while most disyllabic and all larger ones are not clitics. Some mono- and disyllabic prepositions are optionally proclitic. Evidence for this has to do with whether (i) the preposition has its own word stress, (ii) it undergoes final-obstruent devoicing, and (iii) there is no regressive palatalization- (and voicing-)assimilation with the following word’s initial consonant (in connected speech). One preposition which optionally procliticizes is pered ‘in-front- of/before’, as shown in the modem examples in (5a-b).8

(5a) /pterJed/p- + /dietimiL/N* —> [ .pte.riit. ]prwd [ .diiti.mif. ]prwd ‘in-front-of (the) kids’ (5b) /pieried/p* + /dietimii/N* —>[ .pii.riidi.diitj.mil. ]prwd ‘in-front-of (the) kids’

In (5a) pered is not proclitic, and as such undergoes PrWd-final obstruent devoicing and has its own word stress. In (5b) pered procliticizes to the following word, det 'mi ‘childrenmsT PL’, as evidenced by absence of independent word stress, final devoicing of the

7 The preposition pro ‘-about/regarding’ was apparently borrowed into Russian in the 1200s (in Hill’s corpus). Even the earliest examples with third-person pronouns (in the 1300s) triggered n-initial forms.

8 Note that peredti ’before’, prior to the Fall of the Jers was trisyllabic. Since then this preposition has been consistently disyllabic {pered) before any third-person pronoun. In those other environments where the final vowel is kept, as in peredo mnoj ‘before me’, the preposition continues to be optionally proclitic: [piiriidamndj].

Billings, Sandhi Phenomena and Language Change 65

preposition, or regressive palatalization-assimilation of the new obstruent cluster. Assuming that Old and Middle Russian used the same general disyllabic(-foot) yardstick as the modem language does, then none of the prepositions listed so far was heavy.9

The final complicating factor in determining just which prepositions were prosodically light and which ones were heavy has to do with an apparent expansion of the list of items that can function as prepositions in Russian; (6a-b) are two such modem examples:

(6a) otnositeTno nego ‘regarding hirnc,*’ (6b) pr62de nix ‘prior-to themGEN’

These prepositions are etymologically an adverb and a comparative adjective (respectively), which have come to function as prepositions and trigger /2-initial pronouns. Both happen to govern the GEN case.10 Based on such developments, it could be argued that the inventory of prepositions just coincidentally consisted originally of clitics. As I show below, however, before developing the ability to trigger n-initial forms, each emerging heavy preposition had a period of inability to trigger ^-initial forms. This suggests that each heavy preposition underwent a period during which a prosodic constraint restricted it from triggering /2-initial third-person pronouns.

Beginning in the 1500s other words (or combinations thereof) began to trigger nasal- initial third-person pronouns: cerez ‘through’ is first attested triggering an /2-initial third- person pronoun in 1500 and alternates between the two forms until the late 1600s; mez ‘between’ is attested with /2-initial forms in 1582 and has consistently done so since 1835, but has been replaced more recently by mezdu ‘between’. These forms, too, were most likely prosodically light during the 1500s. It was not until the 1600s that the first heavy prepositions began to trigger /2-initial third-person pronouns: pddle ‘alongside’, was first attested triggering /2-initial forms in 1623; prdtiv ‘against’ did so in 1675. Both of these are disyllabic, allowing for the possibility of being proclitics. In 1680 the trisyllabic prdtivu (an etymologically suffixed variant of prdtiv) triggered an /2-initial form, indicating the first reasonably certain attestation of a prosodically heavy preposition triggering an //-initial third-person pronoun in Russian.

Before leaving the issue of prosody, one other type of example deserves mention:

(7) mez nimi gosudari ‘between [them sovereigns 1inst.pl’[1582, Hill (1977:99), quoting PuteSestvija (1954:130)]

This is the mez datum mentioned in the preceding paragraph. This example is not a prob­lem for my prosodic account of Old Russian. Regardless of the internal structure of the words nimi gosudari, this preposition is procliticized to the first prosodic word thereof. This example’s analysis relies on an assumption that mez was a proclitic.

9 One other disyllabic preposition discussed below, Cerez ‘through’, is also optionally proclitic. I have observed, however, that the non-proclitic use is generally restricted to a specialized meaning: Cerez den' [£jiriizidi6nj]prwd ‘one day hence’, versus Cirez den' (£idriisjprwd [di£n)]prwd ‘every other day’.

10 I should point out, as Hill (1977:278, 284-85, 299-301) does, that non-canonical (i.e., prosodically heavy), DAT-assigning prepositions in modern Russian invariably require non-«-initial pronouns (e.g.,

• vopreki emu ‘despite himoAT)- There are, clearly, other necessary conditions at work in determining whether an /i-initial pronoun is used. I do not attempt to explain those other factors here.

66 Billings, Sandhi Phenomena and Language Chance

The fact that heavy prepositions initially triggered non-/t-initial third-person pronouns suggests that there was an active constraint which prohibited a preposition from triggering n-initial forms if it could not procliticize to the following pronominal stem. This restriction is supported by the initial syllabic restriction that gave rise to n-initial third-person pro­nouns, shown in (3) above; such a re-syllabification would not have happened between dis­tinct prosodic words.

2.2 Possessive and Personal PronounsIn addition to the prosodic weight of prepositions, one other distinction in the distribution of n-initial third-person pronouns is whether the pronoun is personal or possessive. The possessive pronouns in Contemporary Standard Russian are homonymous (and homo- phonous) with the genitive-case forms of the personal pronouns, as shown in (8a-c). The same homonymy is attested throughout the history of Russian. In (8a-c) I show the (transliterated) orthographic form, the historical underlying representation, and the (modem) phonetic transcription of third-person possessive pronouns.

(8a) ee (< /j-ejS/) [.jl.j6.] (8b) ego (</j-ego/) [.ji.vd.] (8c) ix (</j-ixu/) [ix.]‘herGEN^er(s)’ ‘himoENflus/itGEN/its’ ‘themGEN^dieiKs)’

For example, in (8c) ix is both the possessive pronoun ‘their’ (or ‘theirs’) and the genitive- case form of the personal pronoun ‘them’. In the modem language the possessive usage of the forms in (8) is never attested with n. Note then the minimal triplet in (9a-c), where the pronoun is possessive in each. In Old Russian, example (9a), the n-initial form is attested, while in Middle Russian, example (9b), the non-/i-initial form is attested. I add the modem form in (9c) for completeness.11

(9a) Old Russian V nego m&sto [ca, 1325, in Hill (1977:239)](9b) Middle Russian V ego mesto [ca. 1664, in Hill (1977:240)](9c) Contemporary Russian V ego/* nego mesto [elicited/LAB]

‘into his place/steadACC.SG’

The syntactic and prosodic tree diagrams for each of (9a-c) are provided in (10). The tree drawn above the segments represents the prosodic structure, while the tree drawn under­neath shows the syntactic structure:

u Although the actual meaning of the noun in (9) has changed, from Old/Middle Russian ‘stead’ to Modem Russian ‘place’ ; these changes are immaterial to either of the structures in (10) below.

Billings, Sandhi Phenomena and Language Change 67

(10) Prosodic and syntactic structure of (9a-c). PrPhrase

v (n)ego mesto

PP

Of primary significance to my proposal is the distinction between the strictly prosodic characterization in Old Russian - that the preposition and pronoun be in the same PrWd - and my characterizations of subsequent stages of Russian (discussed below). The structure in (10) allows the //-initial form in Old Russian, because the prevailing constraint is that the preposition and third-person pronoun be in the same prosodic word; there is no reference to syntactic structure in Old Russian. Hill’s (1977 and printouts) corpus confirms my charac­terizations in (9a). Only one example from before 1600 with the syntactic structure in (10) is not n-initial.12

I conclude Section 2 by summarizing the main points: Structures of the type [preposition + [possessive pronoun + head noun]] and [light preposition + pronoun] are attested with an /2-initial third-person pronoun in Old Russian, while structures of the type [heavy preposition + pronoun ... / are attested with a non-/2-initial pronoun during that period. Interestingly, the combination of the prosodic constraint in Old/Middle Russian and the syntactic one in Middle/Contemporary Russian has required that no heavy preposition has ever triggered an //-initial possessive pronoun.

In the sections below on subsequent historical periods of Russian I show that syntactic constraints prohibit possessive pronouns from being //-initial. Note that my account so far does not actually rely on whether the pronoun is “personal” or “possessive”, just whether the pronoun is in the same prosodic word with the preposition.

3. Middle Russian: The First Syntactic ConstraintIn this section I introduce data which support my syntactic characterizations of Middle Russian: //-initial third-person pronouns are attested only if the preposition and pronoun are in the maximally simple syntactic structure: [preposition [pronoun\]NP\]Pp\ that is, if the preposition and pronoun are alone in the prepositional phrase. I divide this section into two primary types of data: possessive pronouns (including those with elided head nouns) and modified pronouns. I also briefly revisit the issue of appositive objects.

12 This document is Suzdal'skaja (1927) [ca. 1377, quoted in Hill (1977:240)]. This document also lists two non-n examples with the structure in (10) and one n-initial example of this type: v ix~voli ‘in their wiIIloc’. v ix~voli ‘on [sic.] their willLoc’; vnego mesto ‘into his place/steadACc’ 1= (9a)].

68 Billings, Sandhi Phenomena and Language Chance

3.1 Against N-Initial Third-Person Possessive PronounsAbove (in § 2.2) I show that in Old Russian both possessive and personal pronouns are n- initial because the only criterion is prosodic. In Middle Russian the pronoun must also be alone in the preposition’s complement-NP. Hill (1977:237-42 and unpublished printouts) reports that the last n-initial possessive pronoun was in 1551; the next datum example with such a structure (in 1649) and thereafter have all been noiwz-initial.13

The i subscripts in [preposition [pronounJsp\]pp also indicate that the pronoun must be the head of the NP complement of the preposition. In cases of ellipsis it is possible for a non-head, possessive pronoun to be the only overt word in the complement of the preposi­tion. The modem example in (11) contains two possessive pronouns:

(11) Oni zivut ne v ego dome, a v ee.they live not in hisposSESSIVE housepREP.SG hut in herposSESSIVE‘They live not in his house, but in hers.’ [Ferrell (1958:81), also quoted in Hill (1977:250]

Both of the bold-faced words in (11) are possessive, not personal, pronouns. Each is re­quired to be non-n-initial (and has been ever since Middle Russian), regardless of whether the head noun modified by it has undergone ellipsis. Hill (1977:69, 250) reports eight examples with the same structure as the latter pronoun in (11), none of which is n-initial. All of these data are since 1600.14 These data on personal pronouns confirm my proposal: n-initial pronouns are limited to [preposition [pronoun\]Npi]pp structures.

3.2 Modified PronounsThe preceding subsection, among other things, has shown that the pronoun must be the head of the preposition’s complement-NP. Here I show that when some other (non-head) word appeared in that NP in Middle Russian, then the pronoun was unable to be n-initial. This may sound obvious, since pronouns are usually alone in their NPs, even in Russian. It is actually possible, however, for a third word to occur in such a prepositional phrase. One such way is for the personal pronoun to be modified. Only a limited set of words can modify pronouns in Russian. These are usually quantifier-operators like ‘all’, ‘two’ or ‘both’. The examples in (12a-b) are relatively recent examples of such modified personal pronouns.

13 The only n-initial example after 1600 with the apparent structure in (10) is za nju 2e vinu, literally: ‘for her EMPH guiltACGSG* [ca. 1672, in Hill (1977:71)]. Hill adds that “such constructions had a very ar­chaic, Biblical flavor, especially since the enclitic ze originally functioned with such pronouns as a relative [i.e,, ‘for whosepEM.SG guiltAcc.SG’/LAB] - which does not seem to be the case here. [Ibid.]”

14 Hill offers no explanation for the absence in his corpus of structures of this type in Old Russian. Pre­sumably, this kind of possessive pronoun likewise took n-initial form during that period.

Billings, Sandhi Phenomena and Language Change 69

(12a) krome ix dvux ‘besides [them fwokjEN’ [1922, Hill (1977:230)](12b) (stydjatsja) za nix oboix ‘(they’re-ashamed) of [them MiJgen’ [1954, Hill (1977:70)]

I show modem examples in (12) because Hill’s corpus does not contain a single datum with this order from Middle Russian.15 This is a glaring gap in the data, considering that this constituent order is found in Old Russian16 and again in the contemporary period (see §4 below). This lack of data is, however, at least not inconsistent with my proposals.

3.3 Appositive Data RevisitedIn the preceding section (§2.1) I briefly discuss a 1500s example in which a preposition’s complement-NP consists of two words, a third-person personal pronoun and a noun; cf. ex. (7). Hill’s corpus reports three more tokens of this kind, all from Middle Russian:

(13a) ofnix"mjateznikov~ ‘from [them mutineers]GE^.?C[1672, Hill (1977:75), quoting Avvakum’s Zitie.]

(13b) onem" Rixmane ‘about [him Rixman (name)]pREP.sG’[1753, Hill (1977:66), quoting Na§5okin (1952:294)]

(13c) S nim Rixmanom ‘with [him Rixman (name)liNST.SG’[1753, Hill (1977:240), quoting Xrestomatija (1952:290-96)]

Unlike the Old Russian example in (7), in these examples the syntactic structure within the complement-NP of the preposition is important. I assume again that this is an appositive structure, and that the pronoun-noun combination is a morphological compound. As such, the syntactic structure, while not exactly the structure proposed so far in this section (/preposition [pronoun-JmJpp) does conform to its spirit: the pronoun is within a single syntactic word, albeit a morphologically compound one, and is coreferential with the entire complement-NP of the preposition. With this proviso, therefore, the data in (13a-c) are not problematic to the proposals in this section.

Before concluding this section, a brief mention of prosodically heavy prepositions is necessary: Above, in Section 2, I discuss the emergence of prosodically heavy prepositions, shortly after 1600. Such prepositions were likewise restricted to the same

15 In addition to the preposition + pronoun + modifier order in (12a-b), in §5.1 below I discuss the preposition + modifier + pronoun order. Neither of these orders is attested in Middle Russian. Aside from these orders, there is one other possible order for expressing a modifier: The preposition can be repeated, once before the pronoun and again immediately before the modifier. The following are examples from each period:(i) ssamemsnim ‘with vcryiNST.SG with himiNST* (=‘with this very personMAsc') [1377](ii) unixuinyx ‘near themcEN near oiherscEN.PL’ (=‘others of them [have] ...‘) [1768](iii) o nem o iivom ‘about himpREp about livingpREp.sG’ (=‘about him (who is) alive)’) [1928]

[Suzdal 'skaja (1927:1263), Russkaja (1950:240), PeSkovskij (1956:333), respectively.] Hill (1977:230-31) lists more data, with either P-pronoun-P-modifier or P-modifier-P-pronoun order from the modem period; (i) and (ii) are the only data of this kind from Old or Middle Russian. All such data have n-initial third-person personal pronouns. I discuss repeated prepositions in Billings (1996a).

16 Hill’s unpublished printouts include one token from Old Russian with the order in (12): u nej samoj ‘near [her veryJoEN* (= ‘in her very presence’) [1590, PuteSestvija (1954:125)].

70 Billings, Sandhi Phenomena and Language Chance

syntactic structures in Middle Russian. I return to a more detailed discussion of heavy prepositions in the next section.

To summarize Section 3,1 have shown that, in sharp contrast to Old Russian, in Middle Russian (ca. 1600 to ca. 1800) n-initial third-person personal pronouns are attested as long as the structure is [preposition [pronoun\]NP\]pp, this excludes all possessive pronouns and modified personal pronouns from being n-initial. With the understanding that “pronoun ” in this structure includes pronoun-noun appositives, then such appositive constructions are also accounted for.17

4. Contemporary Russian: Only Adjacent-Head PronounsIn this section I show that Contemporary Russian relaxes the syntactic requirement in Mid­dle Russian - [preposition [pronounjNPiJpp- to one in which the preposition need only be the head of the complement-NP of the preposition and adjacent to the preposition: [preposition [pronounj ... ]nPi]pp- The two data types pertinent to this structure are pro­nouns followed by a modifier and conjoined objects of a preposition. These two types of data lead me to propose an Incorporation analysis for this historical period.

4.1 Post-Modified Pronouns and Conjoined ComplementsThere is little difference between the two italicized structures in the preceding paragraph {[preposition [pronounx]Np\]ppand [preposition [pronoun\ ... JnpJpp)- Contemporary Rus­sian allows /i-initial third-person pronouns as long as they head a preposition’s complement-NP, regardless of what follows within that NP. One type of data that confirms this are the examples in (12a-b), repeated here as (14a-b):

(14a) krome ix dvux ‘besides [them /woIgen’ [1922, Hill (1977:230)](14b) (stydjatsja) za nix oboix ‘(they’re-ashamed) of [them ^o//i]gen’ [1954, Hill (1977:70)]

All but one instance of [preposition + [pronoun + modifier]] in Contemporary Russian has triggered an n-initial form; in Hill (1977 and printouts) only example (14a) is attested without n since 1800. That is, whereas examples like (14b) have n-initial pronouns in Old Russian (because of the prosodic constraint), and do not exhibit n-initial forms in Middle Russian (because the structure fails to conform to [preposition [pronoun\]NpJpp), in Contemporary Russian such additional, non-head material is allowed in the NP headed by the personal pronoun. Note once more that possessive pronouns, as in Middle Russian, are ruled out in the contemporary period by the co-indexed pronoun and NP nodes: [preposition [pronoun\ ... ]NPx ]PP.

In order to explain the lack of an n-initial pronoun in (14a), an excursus into prosodi- cally heavy prepositions is necessary once more. Above (in §2.1) I show that prosodically heavy prepositions did not develop the ability to trigger n-initial third-person pronouns until the 1600s, after the prosodic restriction of Old Russian was lifted. I add at the end of Section 3 that in Middle Russian heavy prepositions triggered n-initial third-person

17 I revise this conclusion in Section 5 to align it with preliminary analyses in Sections 3 and 4.

Billings, Sandhi Phenomena and Language Change 71

pronouns as long as the preposition and pronoun were the only two words in the prepositional phrase. That is, heavy prepositions were subject to the same restrictions in Middle Russian as light prepositions were.

In Contemporary Russian, however, heavy prepositions appear to have been subject to somewhat more restrictive distribution of ^-initial third-person pronouns: As each new word entered the prepositional inventory, starting with prosodically light mez ‘between’ and cerez ‘through’ in the 1500s, followed by other mostly prosodically heavy words be­ginning in the 1600s, it appears that each of these new prepositions underwent each of the stages listed above in Sections 1 and 2: First the preposition, if heavy, could not trigger n- initial third-person pronouns, due presumably to a prosodic restriction like the one de­scribed in Section 1. Then, perhaps for several decades, the same preposition developed the ability to trigger ^-initial forms, but only in the maximally simple syntactic structure de­scribed above in Section 2. Then, at long last, the same preposition developed the ability to trigger ^-initial pronouns even if a modifier followed in the NP (i.e., the structure described above in this subsection). The only difference, regarding heavy prepositions, is that these three stages lagged behind the periods described above (in §2-§3) by up to a couple centu­ries, depending on various factors. The preposition mezdu ‘between’ (the more common contemporary variant of synonymous mez), had the following transitional dates for each of the structures described [Hill (1977:127-30)]:

(15) Structure (P = mezdu) First /2-initial pronoun Last non-zz-initial pronouna. light P [pronoun,]/ not attested ^ ^ u ce~^ not attested sipy\*cb. heavy P [pronoun,],- 1740 1856c. heavy P [pronoun,- ...]np,<+/) 1858 1907

The row in (15a) is not applicable to this preposition because, as I assume, mezdu was never prosodically light. Row (15b) shows that - for whatever reason - mezdu did not de­velop the ability to trigger /2-initial third-person pronouns until 1740 (mezdu nimi ‘between themINST’), about a century after other heavy prepositions began to do so (cf. §2.1), with the last non-rt datum of this kind recorded in 1856 {mezdu imi). The structure described in (15c) do not conform identically to the structure discussed so far in this subsection, as indi­cated in the final i (+ j) subscript there, which is meant to indicate that the first pronoun in question is head of the first of two conjoined NPs (the second NP takes the j subscript). Specifically, (15c) allows the pronoun to be one of the preposition’s conjoined heads, as long as the preposition and pronoun are adjacent..

The earliest /i-initial data with the structure in (15c) are shown in (16); the most recent non-« data with this structure are listed in (17). (It appears to be just a coincidence that both the pronouns in (16) are masculine, while all of them in (17) are feminine.)

(16a) me2du nim i soboj ‘between himiNST and selfiNST’(16b) me2du nim i temi ‘between himiNST and thosciNST’ (= ‘between him and them’)

[both 1859, Hill (1977:129), quoting Gondarov’s Oblomov.]

72 Billings, Sandhi Phenomena and Language Chance

(17a) meidu eju i Raskornikovym‘between heriNST and [Raskol'nikov (name)]MASC.iNST.SG’[1866, Hill (1977:129), quoting Dostoevskij (1955:513)]

(17b) mezdu eju i krasnoj stenoj ‘between [[it (grey wall)] and [red wall]FEM.lNST.PL’[1903, Hill (1977:127), quoting GorTdj (1952, stage directions to Act 3)]

(17c) meidueju i soldatami ‘between [it (crowd))FEM.lNST.SG and [soldiers]inst.pl[1907, Hill (1977:127), quoting GorTdj (1907:26)]

The data in (15) through (17) demonstrate that each new preposition gradually acquires the ability to trigger /i-initial third-person pronouns, going through first the prosodic restriction (§2 - if applicable), then the strict syntactic restriction (§3), then the adjacent-head restric­tion described in this section. These phases are gradual, but distinct, lasting at least several decades.

Returning, then, to example (14a), the table in (18) chronicles the progression of krome ‘besides’ in this respect [Hill (1977:116-18)]:

(18) Structure (P = krome) First /i-initial pronoun Last non-7i-initial pronouna. light P [pronoun,], not attested a op not attestedb. heavy P [pronoun,],• 1884 1883c. heavy P [pronoun, ...]np,- not attested 1922 [ex. (14a)]

Like (15a), row (18a) is “not attested” because, most likely, krome has always been proso- dically heavy, as it is today. Row (18b) indicates that krome began triggering /i-initial forms under the simplest syntactic structure in 1884. (As it so happens, Hill’s corpus shows no transition period in (18b), the last non-w-initial token of that kind having occurred the year before.) The datum in (18c) shows that, despite the transition in 1883/1884, the additional modifier influenced the choice of a non-n-initial form in 1922; Hill’s corpus does not happen to show any n-initial data of this kind.

Data like (14a-b) are quite rare. Hill (1977:230) reports only 39 tokens. Of these, only four - counting (14a)- are not n-initial. Of the remaining three, one of the prepositions, vopreki ‘despite’, does not trigger ^-initial forms of any kind to this day, because it assigns dative case; another, pdzze ‘later (than)’, had not yet triggered any n-initial pronouns of any kind; and the last, krugdm ‘around’, had not yet fully transitioned to n-triggering in even the simple syntactic environment described in Section 2. The data in (16) and (17), although not as scarce as those in (14), are largely limited to mezdu ‘between’ and its variants mez and promez, as well as sredi ‘among’ - the prepositions that take conjoined complement-NPs. While mezdu is a very common preposition. Hill (1977:127-30) reports only four tokens like (16) and seven like (17), all with mezdu J8 18

18 Hill (1977:99) reports met nimi i toboj ‘between themjNST and theeiNsf [1868, quoting Tolstoj (1922:242)]. As I mention in §2 above, met began triggering ^-initial forms in 1582 and the last noo­n-initial token, from the simplex structure described in §3, is from 1835. Unfortunately, Hill’s corpus shows no non-H data corresponding to this one. Nor does he report any conjoined data with sredi ‘among’ or promei ‘between’ (no longer used).

Billings, Sandhi Phenomena and Language Change 73

4.2 Adjacent Abstract IncorporationThe preceding subsection has shown that a pronoun in Contemporary Russian can be n- initial if (i) it is the head of a light preposition’s complement-NP and (ii) the pronoun is adjacent to the preposition. In addition, heavy/derived prepositions gradually developed the ability to trigger /i-initial pronouns in their own analogous period of becoming prepositions. That is, such newcomers to the prepositional inventory first trigger n forms in the maximally simple prosodic structures, then the maximally simple syntactic structure, then in structures in which the preposition is the head of the preposition’s complement-NP and adjacent to the preposition. It is also possible for a pronoun that is the first of two conjoined complement-NPs of the preposition to be ^-initial, again, as long as the preposition and this pronoun are adjacent. In this subsection I propose an Incorporation analysis to explain these seemingly unrelated requirements: adjacency and head-hood.

Baker (1988) proposes a unified analysis of numerous phenomena in human languages, primarily having to do with the syntax/morphology interface, in which words are said to in­corporate onto other words. For example, English hunt heads the noun can incorporate (to the left of) the verb, to form head-hunt{er), with generally the same meaning (but different morpho-syntactic properties). I propose that the pronoun head of a preposition’s complement-NP incorporates to the preposition, as shown in (19):

(19) [ P [pronoun,-...] np« (+»]pp —> [[ P+ pronoun,]P- [/,...] np;(+»]pp

Unlike the hunt heads —> head-hunt(er) example, however, the overt order in (19) of the two constituents (i.e., the relative precedence of the preposition-incorporater and pronoun- incorporee) does not change. Baker (1988:202) defines Abstract Incorporation (also known in the linguistic literature by the term Reanalysis) as follows:

(20a) a. [Yp ... [Xj + Y]y ... [xptj... ]] b. [Yp ... Yj... [Xp X/... ]] [= his exx. 124a-b]

“In effect,” Baker adds, “the same relationship holds between the two head positions in both cases, and it does not matter where the lower head actually happens to appear phonologically.”191 interpret one part of Baker’s proposal to mean the following: Incorpo­ration can result in the two constituents having the same precedence order but with different phrase-structure bracketing; i.e., (19) instead of (20). Such a re-analyzed structure can be defined as a means of restricting the distribution of ^-initial pronouns. Thus, in Contemporary Russian a third-person pronoun is ^-initial in the following structure:

19 Elsewhere, Baker (1988:259-63) convincingly argues that so-called psuedopassives in English {Everyone talked about Fred versus Fred; was talked about t/ [- his exx 72a-b]) are likewise the result of abstract incorporation. In order to allow the preposition’s object {Fred) to govern its trace, the preposition must incorporate to the verb. Unfortunately for this English phenomenon, Baker concedes, there is no overt indication - i.e., changed constituent order or an affix perhaps - that incorporation has taken place. In Russian /i-initial third-person pronouns provide exactly such overt indication.

74 Billings, Sandhi Phenomena and Language Chance

(21) [ [P + pronoun,]P* [r,-... ]wi (+;-) ]PP (preliminary; cf. §5.4 below)

Crucially, there is only one set of ellipses in (21), to the right of the trace, indicating that any other material can appear in the NP as long as the trace first in the NP.

This Russian phenomenon differs from most of Baker’s data in that the head to which incorporation takes place is a preposition. Baker (1988:489, n. 4) suggests that this is an accidental gap, possibly having to do with how prepositions govern or assign case to their objects. He lists one reasonably convincing instance in Mohawk (Iroquoian) of incorpora­tion to a preposition, which bears on some of the discussion in this section:

(22) Wa’-hati-nawatst-a’rho’ ka’-nowa-ktatie’ ne Rania’te’kowa’ (Mohawk)AOR-3MPL-mud-placed P-carapace-along Great Turtle

‘They placed mud along (the edge of) the Great Turtle’s carapace.’[= ex. 43c in Baker (1988:90), citing Hewitt (1903)]

The head noun nowa ‘carapace/shell’ is incorporated into the (complex) preposition, leaving behind other material in the NP, namely the possessor [Baker (1988:453, n. 11]. In Russian only pronouns which head the preposition’s complement-NP can be n-initial; possessor-pronouns have been specifically restricted from being ^-initial (since 1600).

The structure in (21) is sufficient to account for Contemporary Russian. Abstract incor­poration takes place as long as the preposition and pronoun are adjacent and do not cause overt movement (as is not the case in the Mohawk example). I return to this discussion be­low (in §5.3), where I present newer data that require a revision of (21).

I summarize Section 3 as follows: Unlike Middle Russian, which restricted the distribu­tion of n-initial third-person pronouns to single-word complements of the preposition, in Contemporary Russian other material is allowed after the pronoun. Such “other material’’ can be a modifier or even another conjoined object. I formalize this seemingly haphazard distribution using Baker’s (1988) notion of Abstract Incorporation. In the following section I continue Baker’s theoretical extension of Abstract Incorporation to account for a more re­cent innovation in the data.

5. A Modem Innovation: P and Pronoun are SeparableIn this section I show that, since the early 1900s, a modem innovation has begun to replace the characterization described in the preceding section. Whereas Contemporary Russian re­quired third-person pronouns to be /i-initial in the structure [ P [protiounx... ]NPl (+ j)]PP, Russian now allows /i-initial pronouns in a structure [ P [... pronounx...]NPi (+J)JPP (note the ellipses both before and after the pronoun in the latter structure). That is, the third-person pronoun no longer needs to be adjacent to the preposition to be n-initial. There are three structures in which this is possible: when a modifier precedes the pronoun, when the pro­noun is the latter member of a preposition’s conjoined complement-NP, and if a second- position clitic appears between the (heavy) preposition and pronoun.

Billings, Sandhi Phenomena and Language Change 75

5.1 When a Modifier Separates the Preposition and PronounIn the preceding two sections I use modified pronouns to show that Middle Russian disal­lowed w-initial third-person pronouns if any other constituent appeared inside the PP (in §3.2), while Contemporary Russian allows other material after the pronoun within the preposition’s complement-NP (§4.1). In this subsection I discuss structures in which a modifier precedes the pronoun within the preposition’s complement-NP.

The preposition + modifier + pronoun order has been attested since Old Russian, but until 1907 triggered only non-M-initial forms. Since then, with increasing frequency, n- initial third-person pronouns are attested.20 In (23) I show two of Hill’s examples with this order: the last attested example without an /i-initial third-person pronoun, in (23a), and the first such example with an /i-initial pronoun, in (23b):

(23a) ko vsemu emu ‘to [all him]DAT’ [1962, Hill (1977:63, 233), citing Finkel' (1962:131)](23b) VO vsex nix ‘in [all them]pR£p‘ [1907, Hill (1977:233, fn. 2), quoting the newspaper Rus'.]

Hill (1977:231-33) points out that n-initial examples of this kind are quite rare. Hill’s initial corpus contained only one such example; he later also uncovered the early example in (23b).21 In my own search of a million-word corpus of modem Russian, Uppsala (1992),I was unable to find even a single token like (23b). Nonetheless, a quite ingenious informant study, Blazev (1962), shows Russian speakers to be almost evenly divided between forms like (23a-b). My own additional informal sampling of native speakers suggests that speakers in their twenties today slightly prefer the form in (23b), while older ones slightly prefer (23a). Both age groups find either type fully acceptable, however.

5.2 Conjoined Complements of a PrepositionAnother way for a preposition to be separated from its pronoun complement is if there is a conjoined complement of the preposition. Above (in §4.1) I discuss such conjoined struc­tures, dealing with initial conjoined pronouns. Such structures, in pre-1900s Russian, did not trigger n-initial third-person pronouns in the latter-conjunct position. While many such second-conjunct pronouns continue to be non-w-initial, as in (24a), Hill also reports one example from early this century22 in which an n-initial third-person pronoun is not the first conjoined complement-NP of the preposition, in (24b). I have bold-faced the third-person pronoun in each.

20 Hill lists only one example with this order from Old Russian, which predictably (in keeping with my proposals in §2 above) is not n-initial: na vsSx ix ‘on [all them]’ [ca. 1350, Hill (1977:232), quoting Cudovskij sbomik.]. Hill (1977:64, 72, 164, 232) lists several examples with this structure from Middle Russian, all of which are likewise non-/i-initial.

21 Hill (1977:232) reports another five examples, many of these from linguistic works. All of these are either vo vsex nix [= (23b)] or dlja vsex nix ‘for [all themJcEN’- Clearly, there are constraints regulating the phase-in of such separated n-initial third-person pronouns, perhaps with only prosodically light prepositions able at first to trigger such forms. Because of the scarcity of data, however, it’s impossible to verify this.

22 Neither Bla2ev nor Hill cites the year or page of ex. (24b)’s source, by Arkadij Gajdar (1904-1941).

76 Billings, Sandhi Phenomena and Language Chance

(24a) mezdu nami i imi [1924, Hill (1977:128), quoting the newspaper Izvestija.](24b) mezdu nami i nimi [Hill (1977:129), Bla2ev (1962:14), quoting Gajdar’s Skola.]

‘betweenus and them’

The vast majority of such separated pronouns, especially in written/printed Russian, are not /z-initial. Hill’s actual corpus includes only one example of each kind from the 1900s, although other studies have shown that /2-initial forms are more and more prevalent in the spoken language, especially with younger speakers.23

5.3 Particle IntrusionA final way for a preposition and pronoun to be separated is for a second-position clitic to appear between the preposition and pronoun. In Russian the clitic li ‘whether’ (yes/no marker) must follow the first prosodic word of its (embedded) clause. If the first two PrWds of that clause are a prosodically heavy preposition and third-person pronoun, as in(25), then the pronoun is optimally n-initial:

(25) Janeznaju, otnositerno li (n)ego oni govorjatI don’t know regarding Y/N himcEN they speak3.PL (with clausal stress on nego)‘I don’t know if it’s regarding him (that) they’re talking.’ [Elicited/LAB]

None of the seven examples like (25) in Hill (1977:233-35) are of any use here, because none of these seven prepositions took n-initial pronouns under any circumstances at the dates when the examples were recorded. In (25) older speakers tend to prefer the non-n form, while younger ones slightly prefer the n-initial form.

I should point out that the second-position clitic in (25) is not necessarily syntactically positioned between otnositel 1no and (n)ego, the first two PrWds of the clause. Instead, as various recent works have proposed - for example, Anderson (1995) - special (or phrasal) clitics are positioned somewhat independently of the syntactic structure. In the case of li in Russian, I have argued, Billings (1994b; 1996b), that Optimality-theoretic constraints re­quire the clitic to be as close to clause-initial as possible without actually being at the leading edge. I still consider the preposition otnositel 'no and the pronoun (n)ego to be the lone members of the PP even though these two words are separated by other phonetic material. I therefore do not consider this data in the next subsection, which deals with the syntactic-sisterhood issue.

23 As an aside, Bulaxovskij’s (1952: gl. 4, §5) grammar, the first edition of which was apparently pub­lished in 1935, suggests that when there are two conjoined third-person personal pronouns as the com­plement of a preposition, only the first of the two is n-initial (i.e., meidu nim i eju ‘between him and her’ [also in Hill (1977:128)). This characterization was generally descriptive of the state of affairs in Russian during the 1930s. My own informal work with native informants shows that this innovation has apparently progressed since 1935 to the point where n-initial third-person pronouns are even preferable as the second conjunct.

Billings, Sandhi Phenomena and Language Change 77

5.4 Non-Adjacent Abstract IncorporationAbove (in §4.2) I argue that ^-initial third-person pronouns are restricted to the following structure: [ [P + pronoun\]P• ft ... ]NPx (+ j; ]PP[= (21)]. Only those head pronouns that can incorporate to the preposition without disrupting constituent order are allowed to be n- initial. To account for the modern innovation described in this section, it is necessary to extend this abstract-incorporation proposal as follows, with ellipses both before and after the trace, as shown in (26):

(26) [ [P + pronoun;] P- ]np/ (+j) ]PP (preliminary)

Such a proposal becomes theoretically much more complicated, because the structure in (26) suggests that the pronoun must change places with any material which precedes it in the preposition’s complement-NP. In fact, no such overt re-ordering is attested.

Baker (1988:202-03) extends his analysis of abstract incorporation (summarized above in §4.2), by proposing that “reanalysis is actually true incorporation happening at the map­ping between S-structure and LF [Logical Form], rather than in the mapping between D- structure and S-structure.” None of Baker’s data appear to be analogous to the separated- pronouns data in this section, in which the evidence of incorporation is overt on the incor­porated word without the overt requisite word-order evidence. That is, the pronouns in (23b), (24b) and (25) are ^-initial without the pronoun’s having incorporated into (i.e., moved to) the preposition overtly.

Applied to the Russian data, this would suggest that the pronoun undergoes the move­ment shown in (26) at LF (an abstract level of representation which is not realized in the overt utterance). The selection of the ^-initial pronoun variant, however, is realized in overt syntax (in Baker’s terms, “between D-structure and S-structure”).

Another possibility is that the material between the preposition and pronoun moves away at LF, leaving these two words adjacent at that level. This possibility is worth considering, since all the modifiers that separate the preposition from an ^-initial third- person pronoun are operator-quantifiers (cf. § 5.1 above). In my view, the movement of operators to a c-commanding position at LF is the most convincing argument for this ab­stract level. Such movement does not, however, account for conjoined object-pronouns separated from the preposition (§5.2). It might be possible to salvage this operator- movement-at-LF suggestion by assuming a copy-and-erase theory with regard to the prepo­sition. Hill (1977:130) reports one very old example of mezju (another historical form meaning ‘between’) repeated overtly, once before each conjoined object (neither of which is a third-person pronoun), which supports the possibility that the modem language deletes the latter instance of the preposition. I deal with repeated prepositions in Billings (1996a).

Whichever strategy is employed, it remains necessary for the incorporation (i.e., joining of the preposition and pronoun) to take place only after some movement at LF. In order not to resort to a multi-level (overt and LF) model of this phenomenon, I revert to Baker’s (1988:202) original proposal regarding abstract incorporation in (20a) above: the structures [yp ... [X\ + Y]y ... [xp h... J] and [Yp ... Yt ... [XpXj... ]] are essentially equivalent. By the same reasoning, [ [P + pronoun;//>• [ ... tj... ]NPx (+ j; ]PP [= (26)] is essentially equivalent to [ P° [ ... pronounx •••/ NPi(+j) Jpp in the relevant respects (cf. (19)], as long as the relationship between the preposition and pronoun is defined. It is important to recognize,

78 Billings, Sandhi Phenomena and Language Chance

however, that the subscripts in and X* in Baker’s structure do not represent preferentiality (as I have been using in my structures so far), but rather defining a particular syntactic relationship between two heads, specifically, between a governing head y and the head X of the complement of Y. With that in mind, the abstract incorporation this century is defined as follows. (The h subscripts stand for “head” as explained immediately above.)

(27) [ P* [... pronoun;,... ]Np ]pp (final)

To summarize this section, I have shown that in this century it has become possible for an n-initial third-person pronoun to be separated from the preposition that licenses it. The material separating the preposition and pronoun can be a modifier (§5.1), another conjoined complement-NP (§5.2), or a second-position clitic (§5.3). Baker’s extension of Abstract Incorporation account for these most recent Russian data as well.

6. An Optimality-Theoretic AnalysisThe preceding sections have presented the facts and shown how the notions of minimal prosodic structure, minimal syntactic structure, and Incorporation can account for the Old, Middle, and Contemporary periods of Russian, including the modern innovation in Section 5.1 now present a series of constraints that explain the historical progression. My proposals are in the framework of Optimality Theory (Prince & Smolensky 1993), as applied to syn­tax in Grimshaw (1993), Legendre et al. (1995), and elsewhere.

Old Russian is characterized by a prosodic restriction on the distribution of n-initial third-person pronouns. A similar prosodic restriction exists in so-called approximative in­version in Modem Russian. The usual way to express a numeral and noun is in that order: pjat'stolov ‘five tables’. This order can be reversed to express approximation: stolovpjat' ‘about five tables’. I argue in Billings (1995b: 162-89) that such inversion is impossible if the quantified element consists of more than one prosodic word. For example, whereas pjat' malen'kix stolov ‘five small tables’ is licit, it is impossible to apply approximative inversion to this structure, because the quantified portion, malen'kix stolov ‘small tables’, consists of more than one PrWd. I therefore propose the following constraint to restrict application of approximative inversion:

(28) LONE-WD [N", approximative inversion, PrWd]:There is no [approximative inversion] if the [N"] (i.e., the constituent quantified by the numeral) consists of more than one [prosodic word]. [=ex. 175 in Billings (1995b:213]

The same type of constraint, (29), can be used to capture the distributional restrictions on n-initial third-person pronouns during Old Russian: the n-initial form was attested so long as the preposition and pronoun are within a single PrWd. Another constraint, in (30), is needed to restrict n-initial forms generally:

Billings, Sandhi Phenomena and Language Change 79

(29) LONE-WD [[P + pronoun], /n-/, PrWd]:Use the [n-initial third-person pronoun] if the [preposition + pronoun] consist of one [prosodic word].

(30) SUPPLETE [/n-/ stem]:Do not use the /n-/ variant of the /j-/ third-person pronominal stem.

The constraint in (30) follows from a common-sense restriction against suppletion in lan­guage. In Old Russian LONE-WD was ranked above SUPPLETE, thus generating the n- initial variant as long as the preposition and third-person pronoun were in the same PrWd.

In Middle Russian (and later), the relative ranking of these two constraints was reversed. A new constraint emerged to license n-initial third-person pronouns:

(31) INC [ P/, [ pronoun* ]np]pp

Use ^-initial third-person pronoun if the pronoun in the preposition’s NP complement is the head of that NP and alone in it.

Note the lack of ellipses in this constraint. Note also that in (31) I employ a version of In­corporation in which no other material is allowed in the preposition’s complement-NP. In Middle Russian this constraint was ranked higher than SUPPLETE.

In Contemporary (pre- 1900s) Russian a similar constraint arose, which still required that the pronoun be the head, but which allowed other material after the pronoun:

(32) INC [ P/, [ pronoun/, ... ]Np]ppUse n-initial third-person pronoun if the pronoun in preposition’s NP complement is the head of that NP and at the beginning of this NP.

Note the single ellipses after the pronoun in this constraint. In Contemporary Russian INC [ P/, [ pronoun/, ... ]np]pp dominates SUPPLETE.

Finally, in this century, a third INC constraint dominates SUPPLETE:

(33) INC [ P/, [... pronoun/,... ]Np]ppUse n-initial third-person pronoun if the pronoun in the preposition’s NP complement is the head of that NP (positioned anywhere in the NP).

This last constraint lacks any positional restrictions within NP, as indicated by ellipses on either side of the pronoun.

Before continuing, it appears that the three INC constraints are universally ranked, with the least restrictive constraint (INC [ P/, [ ... pronoun/, ... ]Np]pp [= (33)]) dominating the next most restrictive one (INC [ P/, [ pronoun/, ... ]np]pp [= (32)]), which in turn dominates the most restrictive constraint (INC [ P/, [ pronoun/, ]np]pp [= (32)]). Other studies - e.g., Legendre et al (1995) - propose such universally fixed micro-hierarchies.

80 Billings, Sandhi Phenomena and Language Chance

That said, with the understanding that all of the constraints in (29) through (32) have existed in all four historical periods, I propose the following rankings for each period (with the constraint names abbreviated somewhat in order to fit them into a single row):24

(34a) Old Russian:LONE-WD » SUPPLETE» lNC[...pron...] » lNC[pron...] » lNC[pron]

(34b) Middle Russian:lNC[...pron...] » lNC[pron...] » lNC[pron] » SUPPLETE » LONE-Wd

(34c) Contemporary Russian:lNC[...pron...]» lNC[pron...] » SUPPLETE » {LONE-WD »lNC[pron]}

(34d) Contemporary Russian with modem innovation:lNC[...pron...]» SUPPLETE» {LONE-WD » { lNC[pron...] » lNC[pron]}}

Viewed in terms of these ranking hierarchies, specifically how SUPPLETE is ranked relative to the other three constraints, the general tendencies toward expansion/contraction of the distribution of /i-initial third-person pronouns are also captured:

During the transition from Old to Middle Russian there was a contraction of some n- initial forms and an expansion of others: Whereas both personal and possessive pronouns could be /i-initial in Old Russian, in Middle Russian only personal pronouns are found; nonetheless, whereas only prosodically light prepositions triggered /i-initial forms in Old Russian, in Middle Russian heavy propositions developed this ability.

During the transition from Middle to Contemporary Russian and during the transition into the modem innovation this century, there were expansions in the distribution of n- initial forms: As the most restrictive INC constraints were demoted below SUPPLETE, more and more syntactic environments allowed n-initial pronouns.

In this section I have shown how Optimality-theoretic constraints can be used to capture the distinctions between the various periods of Russian with regard to n-initial third-person pronouns. I show that four licensing constraints, ranked relative to an inhibiting constraint, adequately generate the data in this phenomenon.

7. ConclusionIn this paper I have outlined the history of n-initial third-person pronouns in Russian. I have shown that differences in the distribution of n-initial forms in Old Russian can be explained phonologically, while Middle and Contemporary Russian require syntactic mechanisms. In these later periods I employ an analysis based on Baker’s (1988) notion of Abstract Incorporation. The strictly phonological criterion in Old Russian does not distinguish between personal and possessive third-person pronouns, while the more recent periods exclude possessive pronouns because they do not head the preposition’s complement-NP. On the other hand, whereas the early periods had a restriction which

24 For those unfamiliar with Optimality notation, the symbol » is shorthand for “dominates”; braces around two or more constraint names with intervening > signify that these constraints are not crucially ranked with respect to each other. For example, in (34c) SUPPLETE dominates each of the two constraints to its right, but there is no relative ranking between the last two.

Billings, Sandhi Phenomena and Language Change 81

required a preposition to be adjacent to the n-initial pronoun, the most recent innovation even allows the preposition and pronoun to be separated by some other constituent(s).

This one phenomenon shows how a phonetic incident has progressed through prosodic and syntactic stages to the point where there is no longer any phonology involved. My Optimality-theoretic explanation of the phenomenon shows how constraint re-rankings ac­count for slight changes in a particular phenomenon’s distribution.

8. ReferencesAbney, S.E. (1987) The English Noun Phrase in Its Sentential Aspect. Ph.D. thesis. Massachusetts

Institute of Technology.Anderson, Stephen R. (1995) “How to put your clitics in their place, or why the best account of second-

position phenomena may be a nearly optimal one." Manuscript, Yale University (ftp://sapir.ling.yale.edu/pub/ Clitics in place.ps); to appear in The linguistic review.

Billings, Loren A. (1993) “On interfaces between phonology, morphology, and syntax: The s + accusative construction in modem Russian.” Talk presented at the Slavic-phonology panel, AATSEEL [American Association of Teachers of Slavic and East European Languages] conference, Toronto.

Billings, Loren A. (1994a) “Prosodic limitations in marked syntactic constructions.” Talk presented at the Slavic-linguistics presession, GURT [Georgetown Univ. Roundtable on Linguistics], Washington.

Billings, Loren A. (1994b) “An Optimality approach to the syntax and prosody of discourse particles.” Talk presented at the Slavic-phonology panel, AATSEEL conference, San Diego.

Billings, Loren A. (1995a) “Syntactic and prosodic constraints on pronoun suppletion in Russian.” Talk presented at, Linguistic Society of America conference. New Orleans.

Billings, Loren A. (1995b) Approximation in Russian and the single-word constraint. Ph.D. dissertation, Princeton University.

Billings, Loren A. (1996a) “Negated prepositional phrases in Slavic." Talk presented at FASL-4 [Formal Approaches to Slavic Linguistics], Cornell Univ. To appear in the proceedings. W. Browne (ed.)

Billings, Loren A. (1996b) “Conflicting prosodic and syntactic constraints on special clitics.” Talk pre­sented at Workshop on the syntax, morphology and phonology of clitics, Zentrum fur Allgemeine Sprachwissenschaft, Berlin. To appear in the proceedings. J. Komfilt (ed.) (= ZAS working papers in linguistics.)

Blafcev, B.I. (1962) “K voprosu ob upotreblenii liCnogo mestoimenija 3-go lica s nafial'nym n.” Russkij jazyk v Skole 23:2, 13-16.

Bulaxovskij, Leonid (1952) Kurs russkogo literatumogo jazyka 1. 5th ed. Kiev: RadjansTca §kola.Carlton, Terence R. (1990) Introduction to the phonological history of the Slavic languages. Columbus,

Ohio: Slavica Publishers.Dostoevskij, Fedor (1955) Prestuplenie i nakazanie. Moskva: GIXL.Ferrell, James (1958) “The systems of the third person pronoun with particular reference to the category of

the preposition in Slavic languages.” In American contributions to the Fourth International Congress of Slavicists. (= Slavistic printings and reprintings 21.) The Hague: Mouton, 77-100.

Finkel', Aleksandr M. (1962) Proizvodnye pridinnye predlogi v sovremennom russkom literaturnom jazyke. Xarlcov: Izdatel'stvo Xar'kovskogo universiteta.

Gor'kij, Maksim" (1907) Devjatoe janvaija. Berlin: J. Ladischnikow-Verlag G.m.b.H.Gorlcij, Maksim (1952) “Na dne.” In his P'esy. Moskva/Leningrad: Gosudarstvennoe izdatel'stvo detskoj

literatury, 119-88.

82 Billings, Sandhi Phenomena and Language Chance

Grimshaw, Jane (1993) Minimal projection, heads and Optimality. Manuscript. [Available as a technical report from the Rutgers Univ. Center for Cognitive Science ([email protected]).] To appear as “Projection, heads and Optimality” in Linguistic inquiry.

Hewitt, John N.B. (1903) “Iroquoian cosmology (first part).” In Annual report of the Bureau of American Ethnology to the Secretary of the Smithsonian Institution 21. Washington: U.S.G.P.O.

Hill, Steven P. (1977) The n-factor and Russian prepositions: Their development in llth-20th century texts. (= Slavic printings and reprintings 118.) The Hague-Paris-New York: Mouton, 127-360.

Legendre, GCraldine, Colin Wilson, Paul Smolensky, Kristin Homer & William Raymond (1995) “Optimality and wh-extraction.” In Papers in Optimality Theory. (= University of Massachusetts occa­sional papers 18.). Amherst: Graduate Linguistic Student Association, 606-36.

Levin, Maurice I. (1978) Russian declension and conjugation: A structural description with exercises. Co­lumbus, Ohio: Slavica Publishers.

McCarthy, John & Alan Prince (1991) “Prosodic circumscription.” Handout accompanying meeting 5 of Linguistics 240: Prosodic morphology, LS A Linguistic Institute, Univ. of California, Santa Cruz.

NaSCokin, Vasilij (1952) “Zapiski Vasil'ja Aleksandrova syna NaSCokina.” In Xrestomatija po istorii russkogo jazyka (Cast') 2. S.P. Obnorskij & S.G. Barxudarov (eds.). Moskva: Gosudarstvennoe uCebno-pedagogiCeskoe izdatel'stvo Ministcrstva prosveSCenija RSFSR, 290-96.

Orr, Robert A. (1987) ‘The origin of N-mobile in Slavic.” Folia Slavica 8:2-3, 302-313.Pelkovskij, Aleksandr M. (1956) Russkij sintaksis v nauCnom osveSCenii. 7th ed. Moskva:

Gosudarstvennoe utebno-pedagogiCeskoe izdatel'stvo Ministerstva prosveScenija RSFSR.Prince, Alan & Paul Smolensky (1993) Optimality Theory: Constraint interaction in generative grammar.

Manuscript. [Available as a technical report from the Rutgers University Center for Cognitive Science. To appear from MIT Press.]

PuteSestvija russkix poslov XVI-XVII vv. Statejnye spiski. (1954) D.S. LixaCev (ed.). Moskva/Leningrad: Izdatel'stvo Akademii nauk SSSR.

Russkaja proza XVIII veka 1. (1950) A.V. Zapadov & G.P. Makogonenko (eds.). Moskva: GIXL.Spencer, Andrew (1991) Morphological theory: an introduction to word structure in generative grammar.

Oxford, England/Cambridge, Massachusetts: Basil Blackwell.Suzdal'skaja IStopis' po Lavrent'evskomu spisku. (1927) 2nd ed. (= Polnoe sobranie russkix letopisej

(tom) 1: (vyp.) 2.) Leningrad: Izdatel'stvo Akad. nauk SSSR.Uppsala University Russian text corpus. (1992) Lennert Lonngren et al. (programmers). Uppsala: Uppsala

University.