Romance Linguistics and Historical Linguistics: Reflections on Synchrony and Diachrony

Rosanna Sornicola

The synchronic distribution and diachronic trajectory of Homeric-φι(ν) have been the source of long-standing debate, with the result that scholarly opinion has yet to settle on a consensus regarding the morphosyntax of forms realized by this marker. Some maintain that forms in-φι(ν) are adverbs, while others contend that they are nouns. Evidence from agreement and prepositional phrases shows that the latter analysis is correct. Homeric-φι(ν) is therefore a case exponent. More specifically, it is an oblique case marker that realizes genitive or dative case in the singular, dual, or plural across all three grammatical genders. Since other case markers exist in the language for realizing genitive and dative case, forms in-φι(ν) are an example of morphological overabundance, the realization of a paradigm cell by more than one word form. This synchronic analysis has diachronic consequences, in as much as it now becomes clearer that-φι(ν) continues the instrumental plural case marker */-bɦis/ and not the adverbial suffix */-bɦi/.

The present paper outlines a diachronic typology of changes in case systems within the Indo-European linguistic family. This study is written in the genre of identification and definition of problem: I will not attempt to offer an exhaustive treatment of the subject. Rather, I would like to draw attention to the importance of extensive research in this field in a diachronic typological perspective, which, in my opinion, may shed light on reconstruction of the linguistic prehistory of Eurasia. First, I will summarize some well-known facts about the Indo-European case and the variety of reflexes of the Proto-Indo-European case system, outlining the most important tendencies – without entering into a discussion of details of the individual case systems as well as into details of the Proto-Indo-European reconstruction. In the second part of the paper I will make an attempt to explain some of the attested developments of the original case system as resulting from contacts with non-Indo-European languages. This, in turn, will enable us to make some hypotheses about the case systems and, in general, structural types of some non-documented substrate languages which could have triggered these changes.

Suppletion is typically conceived of in diachronic terms as a discontinuous phenomenon, by which two or more originally distinct lexical items come to share slots in a single paradigm. However, it has long been known that regular sound change acting undisturbed over long periods of time can in fact give rise to what in synchronic terms must be considered suppletive stem relationships. It is proposed, on the basis of examples from ancient and modern Indo-European languages, that instances of suppletion with a morphosyntactically unmotivated distribution are more likely to have arisen by the accretion of sound change than those which neatly contrast morphosyntactic categories such as singular vs. plural, which typically result from merger of two or more lexemes. Supporting this view is the phonological origin of several well-known cases of morphomic alternations, which can serve as a gateway for the introduction of suppletive stems. However, this hypothesis is to be understood only as a general tendency, since the distribution of stems in phonological suppletion may be altered by irregular morphological developments.

Introduction diaphasic variation as its modern Romance descendants. We therefore deliberately avoid capitalized epithets in such syntagms as 'Vulgar Latin' or 'Lale Latin', which unreasonably suggest an ill-found4d linguistic and P'.VI Imlogical demarcation between one supposcd language, Classical Latin OM die one hand, and an autonomous derivative, 'Vulgar Latin' or 'Late I .11 i n ' 011 ihe other. Rather, in thè same way that linguists regularly append di-si riprive labels like 'modern', 'spoken', 'popular', 'dialectal', 'journalistic', I 1 il ci. ii y', 'Latin-American' and such like to thè modern Romance languages i" K lei io a particular Variety' of that language (e.g., '(spoken) Barcelona < alala n', 'popular French', 'journalistic Italian', 'literary Romanian', 'LatinA I I I C I ii an Spanish'; see Wright, Volume II, for further discussion), we have l . l i ii io I|IL- iliscretion of individuai authors to indicate and identify, where nei rssaiy, (he particular register, style or variety of Latin intended by m'eans i il .ni appropriate non-capitalized epithet or periphrasis, be it 'vulgar Latin', '•pokcn Latin' or 'thè Latin of North-West Africa'. Io e OIH Inde, we should like to remember here Joseph Cremona, who died mi I 1 ) March 2003, and to whom thè present volume of The Cambridge I Ihiviy ofthe Romance Languages is dedicated - fittingly so since Joe was thè lir.i io hold thè posi of Lecturer in Romance Philology (1955—89) in thè I liiivi-rsiiy of Cambridge. During his long and eminent career, Joe firmly i-M.iblislicd, and when necessary, defended, thè study of Romance linguistics in ( .ambridge, and inspired and encouraged successive generations of stuili-nis io bccome specialists in Romance and/or generai linguistics. Jndeed, it si.mds as a testimony to his continuing legacy that a great many of those i in irntly tcaching thè history and structure of Romance languages in British miivcTsities have been his students (or, latterly, havc been taught by his simlrnts). Amongst them are two ofthe present editore and severa! ofthe contributors to thè two volumes. The subjcct ìs buoyant and rlourishing in Bri tain today, and a very large share ofthe credit goes to him. What he created was not so much a 'Cremona school' as a 'Cremona style': he argued that (mitful study of thè structure and evolution of thè Romance languages requires a thorough acquaintance with linguistic theory, and at thè same rime that thè study of linguistics, and especially historical linguistics, needs mastery ofthe kind of comparative and historical data which can be gleaned abundantly from Romance languages. It is these same issues and principles which have guided and shaped The Cambridge History of thè Romance Languages, a fitting tribute, we believe, to his memory. ROMANCE LINGUISTICS AND HISTORICAL LINGUISTICS: REFLECTIONS ON SYNCHRONY AND DIACHRONY1 Rosanna Sornicola In remembrance of Eugenio Coseriu, József Herman, Yakov Malkicl 1 Introduction I discuss here some problems of Romance diachronic morphofyntU in thè light of theoretical and methodological consideraiions on i l i r relation between diachrony and synchrony, and thè question ol lingulltil change. I first attempi to demonstrate a thesis that is perhaps noi obvions, ami rather goes against thè grain of conremporary thinking: Romanci' llngulll '• has rather more to offer generai linguistics in its thinking on thè synclirony diachrony relationship and thè problem of language change than conicin porary generai linguistics has to offer Romance linguistics. Olir diiclpllne not only possesses an extraordinary stock of data, but alsohas longhad a I H l i array of methodological and theoretical tools, which make it a pariiiulaily ideai platform for tackling thè intellectual problem of diachrony. Romani r linguistics foresaw aspects of thc modern debate, and in some rcspi-i r. oftered solutions ahead of that debate. In particular I shall be conci-nini with thè following issues: (a) (b) (e) xxii 'Laws': are there laws of transformation through time, hesidi-s laws of analogy? In other words, do diachronic structures exist, in addition to synchronic ones? The form—function relation: does this relation have thè same properties in diachrony as in synchrony? Syntactic factors in morphosyntactic change: is their role active or inert? i Soffiicela 2 Kelwccn generai and Romanie linguistici 2.1 Introduction Are thè power and potcntial of Romance diachronic linguistics obvious? 'l'hat they are not seeins to me to be sliown by thè argumeiits that bave ariseli in recent years in North American Romance linguistics, as wcll as in schools of thought rooted in diffcrent theoretical and methodological approachcs, such as diachronic typology, grammaticalization theory or generative diachronic syntax. Following a preoccupation dear to Yakov Malkiel, lounder of an authoritative American school of Romance linguisdcs, Stephen Dworkin lias repeatedly called for a 'rejuvenation' of Romance linguistics, ihrough openness to new theories and methods, and promoted Stimulating exchanges of views betwccn Romanists from difFerent countries (see also Malkiel 1988:20). But how feasible is this? And what should this lejiiveuaiion consist of? Are we to rethink old problems and domains of r i u | i i i r y , or slumld we identify new ones, from thè perspective of recent i lin 11 ies? Théories, of course, are never neutral with regard to thè data they .issume. In lact, theories impose their own specific empirica! domains. This could be one of thè cruciai points of thè question. Take tlie Chornskyan distinction between E-language ('External language') and I-language ('Internai language'), nowadays widely used in diachronic generative syntax. From thè outset research has focused on changes in f-Languagr, moreover, thè primary explanandum is taken to be changes in grammars, mental entities represcnted in die minds/brains of individuals. E-languages, and thè changes they undergo, are of little imporr, being considered mere epiphenomena (see Lightfoot 1999:74; 2003). Typical issues are: Why do Prendi children have V(erb)-to-I(nflection) raising, while English children do not, and lower their T (Lightfoot 2003:499). Or: How does thè category change from noun to preposition (e.g., Lat. CASA > Fr. chez) conform to an acquisitional principle such as 'Minimize fèature content' (see Longobardi 2001:294f.)? Such questions and thè kind of data they involve are very difFerent from what die tradidonally trained Romanist is used to. To apply thè Chornskyan distinction again, thè data may be said to involve E-language, whidi look, io boot, like rather restricted technical questions which some may icgard as imexciting, such as: If thè medieval geographical cxtension of derivai ives olii'SK in die function of'nascent article' was much greater than lod.iy, h o w d i d il subseqtientlyget reduced dìachronically, both diatopically i n . I diastratically? And how, conversely, did thè continuants of iLLE-in thè •..une- I I I I K lioii gai li ground, eventually establishing themselves over most of Rum.im e Impulsili s .nul l i t M i i i i i .il liiipn IL ihe kom.ìnu (.see Achisi liei I9'Ì8)? And aie die pi 111.11 •. m / n i l i . i l i . i n . n u l Romania n mascoline noi i MS di icu continuanti ol i he I .nm lecona < lei leu sion nominative plural infiection /, or musi we postulate .1 moie e omplexi less 'economical' and multi-staged, developmeni (see Salutini I96S)? Much of thè Strength of Romance lingiiislics resides in sudi qucsiion.s, and in answers ro theni which harness togcdier die m u l t i p l e dimensions ol spallai, social and historical enquiry, thè intcrplay of which lonns a leilmotil oi die wholc discipline (see Coseriu 1973; 1981; Malkiel 1988:20). Yet further, more wide-ranging, issues have also been raised. In 1 978, in an article on thè problem of language change (reprinted in 1 lerman 1990), thè eminent Hungarian Latinist and Romanist |ó/.scl 1 lei inaii had discussed some 'cluster changes', mainly morphosyntacric and syntactic, and datable between thè second and seventh ccnturies, which had major repercussions on thè strutture of thè noun phrase and thè senience. Illese phenomena appear to have occurred in parallel: (1) simplilìcation of Latin declension; (2) replacement of some case forins w i i h preposi) ions; (3) appearance of new prepositional elements with a more definite .spalio temperai meaning; (4) tendencies to changes in word order; ami (S) loss or wcakening of word-fina!, consonatila! and vocalic, segnu-m-.. I In m.in was convinced that what was particular to historical explaii.iiion lay in: (a) thè interrelations and mutuai causes of changes; and (li) die poiflblc connections between linguistic change and thè circumsiaiues ol l i n g u i s i u transmission (Herman 1978a=1990:362). Of course, dilleiem models ol diachrony contain thè idea of interconnected clusters of changes, rellet i in; 1 , deep strutturai adjustments (in thè language or thè granunar), yei tlie.se models ali share thè aim of radonalizing change. Think, for example, ol die typological representation of clusters of adjustments affecting languages understood as objects external to thè speakers, or thè generative models which hypothcsize 'cascades of changes' which grammars may undergo (see Lightfoot 1999; 2003; Longobardi 2001). In a more or less dircct way both hark back to thè functionalist conceptions of thè early twentieth century, according to which change is not made up of independent adjustments, but occurs as part of a System of interrelated changes. If Herman generally follows this point of view, his position is clearly distinguished by its distance from typological and generative conceptualizations of thè past few decades. I concur entirely with Malkiel, who held that Romance linguistics contains a 'reservoir of priceless data' and considered its diachronic domain one of'truly inexhaustible possibilities' (Malkiel 1988:19). There is no reason to think that he was in any way calling for thè abandonment of thè Rosanna Sornicola traditional preoccupations of thè discipline. This is clear fj-om a dream he evoked, at thè end of a resonant address in 1988 on thè co:nplex history of Romance linguistics, which appeared to him to take thè :"orm of a threeraced Janus. He called for thè new generations of Romanists who had Itnyed away (to Malkiel's regret) into synchronic studies, to return with K-ncwcd interest to thè classic themes of diachronic linguistics, but to do this without ignoring twentieth-century developmentÀ In terms that, IK-rliaps, reflected his direct exposure to thè world of ISTorth American linguistics, where diachronic studies had been reshaped by synchronic stiulii-s, Malkiel (like Herman) focused our attention on thè essentials of hisiory and diachrony, asserting their specific and autonomous nature. Yft is this weakh of data and historical problems, in space, society and i imi-, of itself a strength of Romance linguistics, or is there a risk of its DCCOming a kind of locked strongbox, to which only j tiny cliques of •,|>ci ìalists hold thè key, and whose treasures must lie largely unexploited? [iiere is also thè risk that diachronic Romance linguistics !could become a MUTI' auxiliary to diachronic speculation, a kind of 'empirical data dump', on which theories whose 'historical' nature is dubious coùld draw as they |ilc.isc. This risk may be emblematic of a new phase ih thè history of linguistics in which thè unresolved contraposition of synchrony and dia« l i i o n y in Saussurean structuralism, and thè attempts to reconcile them w i t l i i n European and then North American functionalism, appear to have H-sulicd in thè abolition of both synchrony and diachrony, in favour of a Universa! grammar lying outside time, space and society. The riches ofthe historical world: new and oldpaths in historical linguistics The classica! problems of Romance linguistics may stili be valid, but how are they to be addressed from a novel perspective? The study by Herman mentioned above seems to me to offer an exceilent vantagc point from which to assess thè distance between an authoritative point of view subscribed to by many Romanists, and some recognized approaches in diachronic linguistics which are conspicuously concerned with language as a whole. One initial difference lies in thè fundamentally sceptical view of diachronic theories manifest in Herman's work. This is not to say that he did not attempt to give an organic and coherent representation of change, but he did have a clear awareness ofthe limits of representations of historical facts. He may have hypothesized that thè five (morpho)syntactic changes Romance linguistics and historical linguistics mentioned above were manifestations of a single complex strutturai change, involving encoding of NP-internal relations and thè relation ofthe NP to thè rest ofthe sentence, but he did not think that this was a matter of causai determination in one direction or another. He concluded therefore that 'there is no reliable and generally acceprable answer to thè question of how these processes determine each other or indeed whether any of them takes priority of causai type over thè others. We could refer, at most, to a negative conjecture: an old and simplifying causai solution can in ali probabilities be excluded' (Herman 1978a; 1990:365). The point is that Herman was convìnced that historical linguistics had to be found specific and adequate models, quite different from physical-causational ones, and that thè vciy concepì of'historical explanation' in linguistics lay largely unexplored. This fundamental scepticism also involves more specific bui noi u n i i n portant issues, sudi as thè chronological delimitation of changes. I Ic-rm:m makes extremely cautious use of periodizations, knowing that thc- d.nr ol first attestation is relative and that even frequent occurrcnce canno) l>r taken as evidence ofthe passage from one stage to another. The issue ;niscs of interpretation of sources, especially written sources as reflections ol spoken language — an exquisitely historical problem with enormous con sequences for thè analysis of change. For this reason, more or less accm.iu periodizations have for Herman a less centrai role than appears from some contemporary discussioni, which retain thè legacy of Neogrammarian-stylcpositivist conceptions, apparent also in thè widespread idea that thè locus of change is language acquisition over successive generations of speakers (see Lightfoot 2003). Even further removed from Herman are models, such as diachronic typology and grammaticalization theory, which, in different ways, view change in terms of linear cycles. These are not historical cycles in thè sense of modern historiographical debate, but rather evolutionary cycles. The concepts of 'evolution' and 'history', albeit often nowadays considered intercbangeable, are profoundly different. Recali that in historical sciences this terminological fusion had already been successfully criticized, and superseded, in thè fina! decades of thè nineteenth century (see Tessitore 1991), as had thè idea ofthe predictability of change, which in many theories was allied to a biological—evolutionary view of linguistic development. Herman's approach is concerned neither with thè origin nor thè future development of a given phenomenon. From its origins Romance linguistics has been aware that a truly historical conception of language is a very different matter from a biological-evolutionary one. Consequently, however much one might agree with some recent generative critiques of typological-evolutionary models (see Lightfoot 1999:210), they look like Rosanna Somicola an i-xticmely tardy ivcognition of ideas tliat liave been argued for in thè ilicoi y .mei practicc ol Romance linguistics for two ccnturies. l'.ui ilic »ivau-si split lies in thè synchrony-diachrony relationship. I li-riiiaii (l')78.i; 1990:357) rigluly stressed an issue which stili seems lii|'Jily imporiaiH iwenty years on: che theoretical literature approached ibe |>M>hli-m of diachrony from thè perspective of synchrony, 'either by apply u i|', i lux ii ics established within synchrony toth^ history of language, or by denying die possibilità of a substantial distinction between synchrony . u n i diai lirony willi rclerencc to obviously perceptible traces of historical i luiij'.cs in synchtonic siate 1 . 1 ; Synchrony and diachrony ' l ' I n - subordinatici! of diachrony to synchrony (or their interchangeability, whicli is only apparently different) has deep roots, but emerges distinctly in various North American groups of scholars in thè 1960s. A thesis such as Hoenigswald's (1960:3), that 'much time and cffort could have been saved if historical theory had been built on more explicit synchronic foundations', although characteristic of conceptions of reconstmctionist historical linguistics, has continued to this day to influence other domains of generai linguistics concerned with diachrony (sec Lightfoot 1999:266). At thc Austin congress of 1967, Lehmann outlined thè programme for a new diachronic linguistics, built on modern descriptive linguistics and concerned primarily not with structural units, or states, but thè operations or processes which characterize thè working of languages. Taking his inspiratici! from Praguean models of dynamic functionalism, he stressed thè concepì of thè fluidity of languages with respect to synchrony and diachrony, a concepì which, he hcld, emerges conspicuously if one looks ai operations and not states. Yet thè synchronic roots of such a programme are hard to deny, for various reasons: (a) it takes as basic thè conception of .in .mivi- 'participant' in change; and (b) thè operations or processes are, alu-r ali, rcpicsciiimions of events which express more or less broad movenu-nis in lime, ilirougli descriptive schemas. In other words, thè priority of \ y i u I n o n y ovi-r di.u biony is reformulated as thè priority of descriptive ovcr i m l y h i s i i i r i i . i l lin;',uisiics (see Lehmann 1968; 1982). Such has been thè i n . . n i . I ol d i . i i l i i o n i i lypology and, in different ways, other approaches to dui I I H I M V , OVCI i l i i - Lisi loriy years. I I H in n i n n i no n.i lisi' programme had to contend with some fundameni il d i l l i , u l i h n, .in.l d i n e m.iy bave been excessive optimism about solving i l i . in i o i l i . pinblcm di ilu- metalanguage, i.e., of thè comparabili!)' Romance linguistics and historical linguìstici of different linguistic phenomena in terms of universa! .m.il\ n • ,n< ]•> » <• (b) thè problem ofhow to treat thè form—funaioli relationsliip in tli.u I I K U U , and (e) thè problem ofwhether theoretical models and sopnlltlcated |>liilo logicai practices were really compatible. Perhaps tbe gic.iicsi d i l l u nlm-. I m i . in this lasi problem, which is only apparently mctbodolojMi.il: dicmctical paradigms (whatever their nature) and historical—philological paradigmi .mfar from easy to mesh together, without banal (or d istoria I) 11 IMI meni ol onc or thè other set of paradigms. At Austin, morphology and syntax, banished to die realm ol synchrouy by early structuralism, were puf forward as new dircctions loi rcscaich on language change: diachronic syntax was brought into thè stuily ol die impact of morphological paradigms on sound cbange, and bolli in timi were brought into thè examination of thè impact of sociolinguisiic facts on linguistic structure. While Lehmann wanted wholesale iransplantation of operational—descriptive models into diachrony, Malklcl's Ausiin speech pointed in thè opposite dirèctión, with a clear attempi to bcnd synchrony towards diachronic investigation, in line with thè divani ol gralting some new branches on to thè sturdy roots and thè trunk of die olii uve ol Romance linguistics. Analogica!—synchronic factors are considerai as thè limit of regalar diachronic developmcnt, bound by so-called sound laws. These factors preseli! both a source of phonetic irregulariiy, on thè histoi iial level, and a structural explanation of diachronic irregiilarilics, on l l i c theoretical level (sec Malkiel 1968). Yet it seems ccrtain thai lor Malkiel thè historical perspective was to remain centrai and unchanged, willi ali iis attendant array of technical and methodological tools, and eoiicepiii.il problems. In a different way from typological and granuli,iik .ili/.nion basai approaches, generativism has also defined a programme wlicie s y i u b r o n y (description) controls diachrony (historical represcni.iiion). 'l'In- M i n l y ol diachrony is part of a broader programme ol biologica! lesi-.m li on m i m i , centred on thè theory of Universa! Grammar (UG) and iis rclatloni vvidi individuai grammars (see Lightfoot 1999:2661.). ' l ' I n - objci i ni eiu|iiiry is change occurring in grammars as an cffcct ol tbe 'lesetiing ol p.ii.imeieis', on thè basis of primary (external) linguistic ilata, wbieh consimile ilu'triggering experiencc', whilst changcs in thè externa] linguisti) d i v i n i l i ment, considercd accidental, are of secondary intctcsi. ( 1un|'c i-, i l m \d as a differen conclitions, which give rise to a discontimiiiy (01 'catastrophe' seiThom 1975). The tliscontinuity is an evcnl w h i i b t i a m s m • . \ i u l n n m . in individuals' minds. 7 Romance linguistics and historical linguistics Rosanna Sornicola In this mentalist framework thè dependency of historical linguists on synchronic linguists is clearly spelled out. Only a synchronically based thenry of grammar, a theory capable of accounting for thè grammar of any naturai language as emerging from normal childhood experience, would he able to explain which changes are fortuitous (i.e., attributable to environment) and which are necessary (i.e., grammatical and thus justifiable), wliile historians were bound to have but an uncertain answer (Lightfoot I < ) < ) < ):265f.)- Possible change is therefore necessary change, imposed by thè l.iws of UG. Whatever such laws are, there are gfl(ód reasons to hold that i hange is only partially and perhaps marginally connected with such generai prindples. Lightfoot's theory has some unresolved problems, such as thè relation Detwccn ontogeny and phylogeny, which contains an unjustified leap of loglc. I he properties considered specific to phylogeny are clefincd in terms "I :\l inheritance in historical linguistics which recycles late nlneteenth- and early twentieth-century ideas. What sets off thè trigger i c m . i i n s particularly obscure. Lightfoot (1999:266), well aware of these dlfficulties, sets out thè possible contribution of historical linguistics to '.viu I ironie theories: Synracticians are embarrassingly silent on what it takcs to set thè parameters which they define. What makes historical studies so interesting is that one can sometimes identify cases where grammars change at some stage in thè histoiy of thè language. If wc are lucky, we can then identify changes taking piace in thè language just prior to thè emergence of thè new grammar. In that case, if our records are good, we are in a position to identify just what it took to trigger thè new grammar. In fact, it seems to me that we can learn more about thè nature of thè triggering experience from language change than in any other way. This is no small claim, because unless syntacticians start identifying how their parameters get set by children, somebody is going to cali their bluff and show that thè emperor has no clothes. This places a heavy burden on historical linguistics, for which it does not seem to get due recognition. This theory stili implies historical movement, albcit implicitly, but only insofar as it can rationally be represented within thè grammar as a point of major change (catastrophe). A more radicai theory is that of inertia or diachronic minimalism. According to thè generativist Giuseppe Longobardi (2001:277): A priori [...] thè ideally restrittive theory of language change should probably claim that diachronic change does not exist. This is so because, if diachronic change exists, we are faced with a dilemma: either one must assume that at least some primitive change is unmotivated (i.e. largely beyond thè scope of scientific inquiry), which is incompatible with thè ideai theory: or one loses any understanding of why thè previous synchronic state was possible at ali. Since it seems to be a fact that changes exist (and previous synchronic statcs, too, of course), thè ideai (or perfectly minimalisti theory cannot be fully pursued. Consequently, thè number of primitive causes must be reduced to a minimum, some of them being cast out to thè very edge of grammatical systems, or beyond. This logicai operation is accompanied by another: thè assignment of an explanatory role to social, matetial and cultura! changes, which are extetnal to or independent of thè grammar. And it is a Romance phenomenon with which Longobardi shows this, namely thè development of French chez as an element with a prepositional function. This is thè type of change traditionally represented in terms of'grammaticalization' (passage from one grammatical category to another: Lat. CASA(M) > Fr. chez}, which within a generative framework constitutes a prime example of thè problcm of thè resetting of parameters. Longobardi ingeniously attempts to combine etymological and generai linguistic analyses. Drawing on numcTous works in Romance historical linguistics, he describes a broad range ol n o m i l i . i l constructions with continuants of Lat. CASA. His originality lies in compai ing such Romance types with thè Semitic 'constrtict state' typc (ci. I lebicw beyt ha-more, 'thè teacher's home', lit. 'thè home of thè teacher'), .u«l deriving them from principles of UG. But thè set of tinivcrsal properties of thè construct state is only thè starting point, perhaps going back to a 'predocumentary common Romance stage', of a development that in thè case of chez involves in ali rive diachronic changes: (1) (2) (3) (4) (5) thè two lexemes MANSIO 'abode, dwelling' and HOSPITALE 'abode, asylum' develop thè meaning 'house' in thè Gallo-Romance area; thè noun chiese, thè phonetically regular development of CASA, disappears; Lat. CASA(M) also follows a different phonetic development, representable as *kas> chies > chez; NOUN > PREP; thè meaning 'house' is transformed into that of 'generai and abstract position'. Longobardi (2001:298f.) concurs with various Romanists in seeing a rclationship between (1) and (2), which he expresses, however, in 'causai' terms ((1) caused (2)). His thesis is that change (2) is responsible for thè I m i . . '...i uh ni,i «•In.I. .M ni . L . H I T . i '.i i o .uul i l i . u consequendy (1) is thè originai i li.un i . i l in i l u - l'.i.nnm.ir, which gave rise to ali thè others. III. u i ih.u i l i < - ni|M',riin|. londiiion is external to thè grammar 1 . 1 1 . u . i l n w.iv i" conclusioni l.ulcn with theoretical implications: in thè M n i . i i i n l i i ' . i n i v ol i-/ir.:. i bere was no resetting of parameters, and evi u ' . M i M i i i i th.inj'.c (/'i) niighr be considercd a secondary consequence ol .1 M - i n . i n i i i change which occurred in another lexeme (Longobardi . M I D I :.'.')/ ')')). In elicci, al ihe syntactic level nothing happened. This i i | >i escili,il ion invokcs continuily, as more generally expressed in thè theory ol inriii.i: 'language is dlachronically inert unless proved otherwise'. This^ model ol ic.inalysis ol a grammatica] categoiy has its attractions, especially whcn lompaicd w i i h thè analyscs offcred by grammaticalization theory, which rcprescni ibis type of change in terms offluctuations- it matters not wheiluT diachronic or synchronic - at thè end of which there is a definitive 'leap' Irò m one category to another. Neither approach is unproblcmatic. 'l'he analysis of thè diachrony of chcz is scarcely 'historicaT, in thè sense, particular to Romance linguistics, of a systematic description of thè characteristics of lingnistic structures in their distribution in time, in space, in society and in thè culture of individuals, and in thè sense of an understanding of how sudi si ructures interact with external factors. The historical method adoptecl is really more of a typological-reconstructive one. And historical factors are exploitcd, simplistically, to demonstrate a given assuniption. Oli i he level of thè diachronic model represented, thè result is clever, rather i h a n convincing (as I intend to show elsewhere). it is in Romanci- linguistics that thè idea of thè non-existence of language change has icecivcd a major theoretical formulation, at thè hands of ( ìosciiii. K i l t bis aigumcutation is quite difterent, being of a historical— empii ii .il n.lime. As lor diachronic minimalism, some questions arise. Hl.m[',ii.i|',c is ili.ii liionically incrt, what is thè point of turning to diachrony, l u m i ihc poiiii ol view of grammatica! investigation? Does thè theory < i l u n ni.i noi ilelivei ilio coup de gràce to what was left of thè notion i.l ih.u I I U . I M .i', .1 pini ess of historical transformation dominated by gramni.u ' And iloes il noi anumnt to an unconditional surrender to external I.u lor., w h i i h un ibis account are stili haphazard and accidental? In that . i , I M . i n i i n . u n .il i.iiionalism, taken to its logicai extremes, would give rise io ,in i v i d e ni p.u.idox: die belief that everything in diachrony is purely i u n i inr.i m I he .ipp.neni siipiemacy of external factors in thè generative theory of lucrili! i di i epnve, lor ihc logic of thè theory does not favour thè identih. ni MU h fu lors. hs ultimate aim is to represent grammar unaffected II) • linguistici and h i s l i . i u .il I m p i i 1 1 . by thè perturbations ol changc, so as m l o n l n i i n i far I i 1 " <\>\- i" m ideally rcstrictive theory ol change. I leni e. i m l i k i Llghtfbot'l propoialftì thi 'triggering' factors ot change are banishcd io die pi n p h e i v ol du n u n n i u and evcn beyond. It is remarkable ih.il, uni e i b i s is dune, die {',1.1111111.11 m.iy become indiffcrently a synchronic or .1 d i . u l n o n n model. Vei .mollici paradox seems to be that this leads by anodici ionie io die ...ime generai conclusion as thè functionalists on die non disiim iness ol lynchfony .ind diachrony. This conclusion is reached in niany lespeiis l n . m die opposile direction: thè diachrony of thè incnia model is by d e l i n i l i o n si.nii , w h i l e Rinctionalisticmodels are basedon thè COIRepi ol die dyn.unii poieiuialol thè activity of speaking. The difficulties of generative approaches show bow d i l l i u i l i il is to reconcile biologica! and bistorìco—cultura! paradigmi, Tlicy cannoi simply be combined together and thè historicaJ is ultimately disiortcd by thè predominance of thè biologica!, and recluced lo a mere cpiplicnomcnon (see Lightfoot 1999:265). Many Romanisis would I n n i ibis wonying. Herman's concerns about thè lack, silice Saussure, ol hro.id cngagcinciii with thè historical dimension of language on its own terms, seems loday more relevant than ever. The 1980s bave been said to mark a turning poiìit in linguisiics, wiib die attempi to bring down thè wall betwcen generai and historiial linguistics which had stood for thè best part a century (sec Matthcws I ')') I : M.). Km this may be less straightforward than thè collapse ol die Beilin Wall. Io grasp thè potential of Romance linguistics, we need to examine why. Saussure's reflection on synchrony and diachrony appcais now.ulays, thanks to thè publication of notes from thè Cours, ed i K-I I by (lonstantin, Patois and Riedlinger (in Komatsu and Harris I9')3; Kom.usn .ind Woll 1996; 1997), much more multifacetcd and complex than wbai einei;;ed from thè edition edited by Bally and Sechehaye (Saussure l')22). I h e debate of thè 1960s on new dircctions of research in diachrony may bave been Lnfluenced by thè earlier edition. We need lo review die main poinis of this debate to assess thè imporr of thè criticai revisions which were subsequently proposed, and to grasp what is spedile and i bai.u lerisiic ol thè perspectives offered by Romance linguistics. l'unii;imeni,il .ne: (a) ibiidea of the'link between System and consciousness; (b) die pmbleni ol teleologism; and (e) reflection on thè conccpts ol phcnotnction and l.iw. The need to contrast synchronic and diachronic phenomen.i, .>•, S.mssme held (Riedlinger, Quire II [Komatsu and Wolf l ' ) l ) / : . W , l . | ) . r, looied in a theoretical conception born of thoroughgoing phllosophlcal .issesMiu-nt ol thè notion of System and that of événcmenl. Cruciai r. die i d i . i 1.1 die Rosanna Sornicola speaker's linguistic 'feeling'. A synchronic fact exists only as an element in a network of psychological relations (dependencies) which lie in thè consciousness of speakers in a collectivity (see Constantin, Quire IX [Komatsu and Harris 1993:120f.]). There, feeling and meaning are indissolubly linked (Ricdlinger, Quire III [Komatsu and Wolf 1997:49]). In other words, (or there to be a synchronic fact there must be speakers who perceive and la-I it as an entity clear and distinct from other facts, which are nonetheless e i tnnectcd to it. Only such feeling gives value to synchronic facts. The aspects of this model which post-Saussurean structuralism and l u n e lionalism in its various manifestations have most emphasized are thè logicai and formai ones, i.e., thè logicai, differential and oppositional ivlaiions between thè elements of linguistic associative networks. Yet it is (hi- rclation between synchronic fact and individuai speaker which has Eundamental theoretical consequences. A synchronic fact is not a mere linguistic 'phenomenon' or 'event' in abstract space, divorced from any •.|>r.tkc.T,s for whom it has Value',2 and speakers' awareness stands as thè oiily yardstick for determining thè dcgree of reality of a phenomenon in lynchrony and thè possibility of representing it as a structure. I IR- epistemological status of diachronic facts is quitc different. Such l.icts '.ire opposed to synchronic facts as are events to a System, are only rvcnis' and 'we do not speak via events' (Riedlinger, Quire II [Komatsu and Wolf 1997:46]; see Herman 1978a; 1990:36lf.). As an événement, every ili.n I ironie fact is determined and exists outside thè loop of linguistic, logicai .ind psychological relations which lies in thè awareness of speakers in a < olleaivity. It is an independent fact, in series with other diachronic facts (Riedlinger, Quire li [Komatsu and Wolf 1997:45]). It is not brought .ilive by thè speaker's feeling, and thereby lacks direct intcrpretation and Itmctural value. Yet thè distinction between thè two orders of facts raises a number of unresolved questions. The consequences for synchrony of thè centrality of thè speaker's perspective were not ali followed through. For this point of view can be thoroughly and coherently applied only within a time frame in which observer and speaker coexist, and of which thè linguist-observer is a direct witness. Saussure's thinking displays a rather blurred overlapping between thè concepì of 'observer' who infers thè characteristics of a given état de langue and that of'speaker' in whose feeling and consciousness they are reflected or experienced. The failure to think through thè logicai leap from speaker to linguist-observer prevents further exploration of a centrai question: thè attribution of meaning and value for états de langue in thè past which thè observer does not witness can only be a matter of conjecture. 12 Romance linguistics and historical linguistics Moreover, thè divorcc between observer and speaker's feeling, with regard to past phases of thè language, means that thè description of past states may be considered more Iike thè process whereby an observer infers (describes conjecturally) what caused thè passage from one state to another. In other words, theoretically and methodologically, it is only thè synchrony of thè observer's present that is radically different from diachrony. The centrality assigned to thè speaker throws some light on thè idea that thè study of grammatical and semantic facts belongs to synchrony, while non-grammatical facts belong to diachrony (Riedlinger, Quire IV [Komatsu and Wolf 1997:67). Grammar and meaning live only in thè consciousness of speakers\IPhey decay and melt away into mere events if we try and study them from thè perspective of thè transformations they undergo in time, for which we generally have only thè material documentation which survivt-s individuals. There are problems and contradictions in this polarization, duito lingering nineteenth-century syntactic and phonological conccpts, DUI thè status of grammatical and non-grammatical facts in relation to thè speaker's linguistic feeling and to thè role of time in language rciains iis great interest. And ali thè more interesting are Saussure's doubts about ibi 1 possibility of associative and syntagmatic facts having a history, and thè implications for thè separation of diachrony and synchrony, if indeed they did. The conclusion that synchrony and diachrony are harder to distinguisi! in thè domain of meaning and syntax than in that of phonetics, points up an awkward theoretical problem. In any case, thè distinction mentioned above has an important theoretical corollary, frequently stressed by Coseriu: thè notion of'historical grammar' is a contradiction in terms, because 'no System can straddle a succession of periods' (Riedlinger, Quire IV [Komatsu and Wolf 1997:62). In other words, structural laws of change are unthinkable. Diachrony is thè realm of événements occurring one after thè other in an unstructured way and whose logicai links only thè speakers' linguistic feeling can provide. By thè way, studying different états de langue one after thè other does not mean that one is moving in thè domain of diachrony (Riedlinger, Quire IV [Komatsu and Wolf 1997:48f). This is a conclusion of major theoretical significance, which has been ignored in many subsequent models of diachrony. But Romance linguistics has never forgotten it. Saussure's thinking also has major consequences for thè relationship between thè two dimensions, involving thè paradox of a close mutuai dependency alongsidc radicai independence and irreducibility (Riedlinger, Quire IV [Komatsu and Wolf 1997:36f.). The two domains are incommenstirable, and any attempi to reconcile them is chimerical and fraught 13 I i ' . ( U H I . III.I « u h | M M u n i . i l | n i l . i l l s (Constantin, Qui re IX [Komatsu and Harris i I ' [ i l'or examplc, a phonetic change might be considered a natura] i M ni i n i r , . I I , I v u i j ' , outside spcakcrs' consciousness, yet it becomes a M M. I l l u n i . 1.1.1 w h r i i i h c i c s i i l t i n g phonetic variants are assigned a meatimi'. K u i i l i e s\-ni I n o n i i I . n i is not cxplained by thè diachronic one. I he . uni < - p i i m i MI m mari/ed bercio gocs hand in hand with a profoundly andtcleologlcal oudook (Constantin, Qui re IX [Komatsu and Harris I'/"M: I I 11): l i is -,\n idea we havc that thè language appears to bc a mechanism i ic.iied willi ;i vicw to and in accordancc with thè concepts to be examined; wc scc how thè state was never destined to express thè ineniiiiigs il acc[iiires or to mark them according to a convention i M i v c r n i n g thè terms employed. A fortuitous state occurs and is taken ovcr. Nothing is more important from a philosophical point of view. But die state must bc carefully distinguishcd from what changes it. Tlie antitelcological position has a significant link with idcas of che concept of law, and thè need to distinguisi! bctween synchronic and diachronic laws (Constantin, Quirc IX [Komatsu and Harris 1993:117f.]): In ilie diachronic domali) a law is imperative and dynamic. It abolishes onc (liing and introduces another. It makes itself felt by its effect. It has .1 Ini. e. A di.ii (ironie law expresses an imperative which is carried out wliatcvci die rcsisiance. A synchronic law expresses an existing order. It is .1 l.ivv ni ihi- suine kiud as wlicn one asks: on what pian were trees planted in die |',.nilcn? This law captures a state of affàirs, an arrangement. • Noi ini|KT;iiivc, not dynamio 'l'wti kind.s .il generai problcm emerge, which stili deserve our attention ind.iv: dir mlr ol .sonni! change in diachrony and thc concept of die n n p r i . i i i v c n . i u i i c ol laws. In both cases thè conclusions are ofwider impor1.11 ii i i h.i 11 ni.u .11 111.M sci-in. This is apparerò", forexample, indie idea that thè io ni l.iw' ihould l>e applied widi tnuch greater care to diachronic than to •.MI. I l . u i s (sci- Kicdlinger, Quire 111 [Komatsu and Wolf 1997:48]). A n . l i l i . idi .1 ni .1 non iinpcraiivc, non-dynamic synchronic law, diat merely .1. , ulii-, .1 M.ne ni . i l l . i i i s , is not after ali far removed from thc concept of in, I i. mi. i . . u n . I in subsequent models of theoretical linguistics. But il i m i .1 I I I . I J I M difference: in these latter developments synchronic mi. IH //m» /,/. in | n i ) j i . i . i l i l c onio diachrony, while for Saussure such an "I" i (i . M i l . I In un IIIC.IHS bc i.iken for granted. I t i l i n i • I. M u l i v mie ol die major criticai points in diachronic theoryis i l i , | . i . . l . l . i n . . I i l n posuilability ol structural laws o f language changc. I i Romanie Impilisiii s .uni I U M H I H .il l u i ) ' Recent typologìcal and g ì . u n n i . l i l i . i l i / , 111011 I I . I M d niu.l. I u > • . i m i structural cycles (or laws) ol l i a n s l o m i . i i i o n , w i i h .m . \ n l i ni n u e l l e i m.i debt to thè early twenticth-cetilury thoughl ul d n l ' i . i p n '.. l u m i w h i i l stressed thè complemcntariiy ol peni i.meni e and i h.uij'.e, ih. i n i e i . n i 11 synchrony and diachrony and thè interi lianj'e.ihle n . i t i n e ni di.n l i n n n i . u n synchronic laws themselves. Rejccting ' t h è M e l i l e u n i i nctltioiil n i e l l i m i i > thè history of isolated facts'' ('l'bèses 1929:')), t h è l'i.i|',uc Sihool h.u affirmed thè nced for historical li tignisi ics (and oliici evi ih il im 1.11 \. i. m . >, ) to move from a conception of facts produce-i! a d u n a i i l y . u n i .m i d c n i . i l l y , regular as they may be, to a nomogcnctic conicpiion ol 'ioni .uen.iiioii ol evolutionary facts according to laws' (Thèstl 1929:9). Not only ,sytn l u o n y , but also evolution is taken to have structural laws w h i i h cali c x p l . i i n b o l l i phonological and grammatical changes (Thèsti 1929:8). In i b i s vicw, linguistic changes are indeed not destructive lorccs opetaling by i h a i n e and in an unstructured way, but often aim at stabili/Jng and rebullditlg ihc sysicin (Thèses 1929:8). On thè other hand, synchronic descripiion i.imiot c i u i i e l y dispense with thè concept of evolution, for 'cvcu in a sci tot cnvisagcd synchronically there exists an awareness of thè stage which is disappcaiiug, of thè present stage, and of thè stage which is coming imo being (ThèsfS 1929:8). As in neofunctionalist diachronic models, in t h i s appio.n li ilu-rc are no longer any insunnountable barriers berween (he sym bromi niodcl and thè diachronic model. Moreover, thè notion ol l u i u t i o n . i l sysiein may be used, in different ways, in both dimensione (Thèsei 1'129:7—8). At ibibase of this conception is die idea that thè foundalion ol movcments in synchrony and diachrony is thè speaker understood .is a p a l l i i ip.nu .uni protagonist in thè functioning of thè language. Bui i b i s model lias iis theoretical problems: thè actions of spcaking Individuali in ( I n sym I n o n i e function and thè diachronic transformations w h i i h alici i l.ui|'.n.i)',e u v e i . i l i belong to mutually incompatible dimcnsions: thc lonnci .uè mi .1 sin.ili scale, thè latter on a large scale (see Herman 1978a; I9')(): UiO). In sum, even if Saussure's thinking bears thc l i . i l l n i . n l . ol .m m h . - i i i . u n c of opinions common in late nineteenth-ccnuuy hiltorlca] linguildcii his discussion of thè concept of law, and partictilaily bis n j c i i i o i i ol ibiimperative nature of laws, shows that he has ic.illy . u l v . i i n c d b e y i u i d .1 positivistic viewpoint which likened [ingulstic l.iws io (hnse < > l phyilci. Actually, it might be better to say iliai in rei nj'.iii/ini', dn I H I | H U i , m . . .il value/meaning as a guiding epistemological prindplc Im i l n M i n l y ol synchrony, Saussure already bclongs to ,m .ij'.c ih.n h.is i . | . . | . i . . l I m n i i o n as thè interpretative key in historical cni|iiiiy. I m i i h.u m limimi) 1 , i h i s principle to synchrony he shies away frorn thè conccptlUU Ictp tnai uthcr: 15 R d i n . n u e l i n g l i i s l i i s . n i . I l i r . i m i c il l i » I I I il w i i l i i» n uni.il piilalls (( lonstantin, Quire IX [Komatsu and Harris l'i'M I l,'|). l'oi examplc, a phonetic change might be considered a naturai evi-in in liseli, lying outside speakers' consciousness, yer it becomes a ;\H limnii. I.ic i wlicn thè resulting phonetic variants are assigned a meanni|', Uni ilir syiK limine lact is not explained by thè diachronic one. [*hc conccption summarized hcreto goes hand in hand with a profoundly intltclcological outlook (Constantin, Quire IX [Komatsu and Harris r>'>3:l 111): li is a 111 ist akeii idea we have that thè language appears to be a mechanism i ie.ned wiili ;\w to and in accordante with thè concepts to be cxaiiiined; wc sec how thè state was never destined to express thè iiicaniin'.s il aci[iiircs or to mark them according to a convention i;nveming thè tenns einployed. A fortuìtous state occurs and is taken over. Nothlng is more importarli from a philosophical point of view. Bui ilic siale must be carefully distinguished from what changes it. The Ultìteleologica] position has a significant link with ideas of thè concepì ot law, and thè nccd lo distinguish berween synchronic and diachronic laws (Consumiti, Quire IX [Komatsu and Harris 1993:117f.]): In ilie diachronic domani a law is imperative and dynamic. It abolishes ont- tliing and introduccs another. It inakes itself felt by its effect. It has a Ione. A diachrouie law expresses an imperative which is carried out whaU'vcr ihe resistance. A synchronic law expresses an existing order. It is .1 law ni iln' suine Uind as whcn one asks: on what pian were trees planted in ilie Barili-li? Tliis law captures a state of affaìrs, an arrangement. • Noi imperative, not dynamio I wo Idiuls <>l generai pioblctn emerge, which sdii deserve our attention lod.iy: (he mie (il sonncl change in diachrony and thè concepì of thè impei.iiivc n.nule ul laws. In both cases thè conclusions are of wider impori.un i ili.m i n . i v . u I n s i si c-iii. Tliis isapparent, for example, in thè idea thatthe I C I M I 'l.iw' simuli! be applied with inuch greater care to diachronic than to I) n, I n , u n . 1.1, is (see Kicdlinger, Quire 111 [Komatsu and Wolf 1997:48]). And ilie Kle.i ni .1 min imperative, non-dynamic synchronic law, that merely <|. . ni" i si.ne di .ill.iiis, is not after ali far removed from die concept of •.MI, I I I C I M H mi, l u n i i c l in snbsei|uent models of theoretical linguistics. But i l » i. n in u n . .1 m.ijdi dillerence: in these latter developments synchronic mi. ne i/nn /,iiin projcctablc (mio diachrony, while for Saussure such an cip. i . i c i m i c i by un nie.ins he taken for granted. l i i . i l i n . I. n wliy one o l i h e niajor criticai points in diachronic theory is i l i < l ' i ' i l ' l i i n ni die postulability of structural laws of language change. I Recent typological and gì ai m n.un illzation-baicil molici i» • structural cycles (or laws) (il iransluim.iiidii, w n l i .in • •. d. m debt to thè early twentieth-ccntury thoughl ni i l n l'i.ipn '., I • Il i il I u l n e li stressed thè complemcntariiy ol perni, i nei u e .ni, I e li.mp i l i . 11 M ,.n ni synchrony and diachrony and thè interchangeabli i un tdluchronli .uni synchronic laws themsclves. Rcjecting 'ilie siculi ami he un..» mi thod < > l thè history of isolated facts'3 ('l'hcscs 19.',9:')), thè I'I.IJ'.IK ' > c l n n i l lud affirnied thè need for historical linguistici (and Other CVOlUtlontTy IclcnCCl) to move from a conception of tacts prodnicd a d u l i . m l v .md .u e ulcm.illy, rcgular as they may be, to a noniogcnciic l o i u e p i i n n ni \n .iu I I . I I U H I di evolutionary facts according to laws' (Thitei I 1 ).','):')). Noi nnly syiu bidiiy, but also evolution is taken to have structural laws whieli i . i n ocplalr) bolli phonological and grammatical changes {Thèsts 1929:8), In iliis view, linguistic changes are indeed not destructive iorces operaling by ( li.nu e ;nid in an unstructured way, but often aim at stabili/.ing and rebuilding die sysicm (Thèses 1929:8). On thè other hand, synchronic dcscriptlon i .ninni eiuirely dispense with thè concept of evolution, for 'even in a .sei i or envisaged synchronically there exists an awareness ol thè siage which is disappearing, of thè present stage, and of thè stage which is coming imo being ( Ihèses 1929:8). As in neofunctionalist diachronic models, in ibis approach there are no longer any insurmountable barriers berween thè synchronic model and thè diachronic model. Moreover, thè notion ol lunclional systein may be used, in different ways, in both dimensions (Thèses 1929:7-8). At ihe base of this conception is die idea that thè foundaiion ol movcmcnis in synchrony and diachrony is thè speaker unclerstood as a paiticipani and protagonist in thè functioning of thè language. But tliis model has its theoretical problems: thè actions of speaking individuali in die synchronic function and thè diachronic transformations which alfect language nveiall belong to mutually incompatible dimensions: thc fornici- are on a small scale, thè latter on a large scale (see Herman 1978a; 1990:360). In sum, even if Saussure's thinking bears thè hallmark ol an inlieiiianee of opinions common in late nineteenth-century historieal linguiltici, liis discussici! of thè concept of law, and particularly bis rejeiiion ol die imperative nature of laws, shows that he has really advaiued beyond a positivistic viewpoint which likencd linguistic laws io dinse ol physies. Actually, it might be better to say that in recognizing t l i e impon.uue ni value/meaning as a guiding epistemological piimiple Ini die suuly ol synchrony, Saussure already belongs to an ;ij',e lliai has adopted l n m i i o n as thè interpretative key in historieal ciu|uiry, bui lli.u in limiting ibis principle to synchrony he shies away I co m thè conccptual le.ip ih.il ddieis 15 Romance linguistics and historical linguistics Rosanna Sornicola were to make shortly after, that of using thè concept of function as a tool for understanding how historical change happens. This is a major issue. It tuakes us face thè daunting task of imagining, on thè basis of sources which are perforce indirect, a complex of interrelated processes, with thè airn of Identifying an overarching reason for their development. It is thè task of thè 'resuscitation' of thè past by historians. But at this point thè problem is what 'rcsuscitation' means. Tliis difficulty is conspicuous in those parts ofthe C0«ndealing with thè piospective and retrospective viewpoint in diachrony - yet another issue wliicli has tetained its interest to this day. The former 'is equivalent, if we amld apply it without difficulty, to thè complete synthesis of ali facts which M l i n i - m thè history, thè evolution of languagc' (Riedlinger, Quire III | Komatsu and Wolf 1997:63a). The lattei places thè observer at a particular period of time and leads him to ask 'not what thè result of a form is, but w h a i forms gave rise to it' (Riedlinger, Quire III [Komatsu and Wolf I997:64a). This is in effect thè reconstructive method. The distinctions c( 11 ocs that made some years earlier by Meyer-Eybke in his Einfùhrung in i/iis Stiidium der romanischen Sprachwissemcha.fi (Meyer-Liibke 1 901), who dlltlngulshed in Romance linguistics a 'horizontaT method (synchronic, ni Saussurcan terms), from two 'vertical' methods, one from ancient to modi-m, thè other from modern to ancient, and who considered thè latter io IH- 'die real history of linguistics' (see Varvaro 1968:149 and n24). The prospettive approach is ideai, and difficult to apply, largely hi-i'ause 'here thè document is no longer thè observation of what is more i n k-.ss present to speakers', but something indirect (Riedlinger, Quire IV | Komatsu and Wolf 1997:63]): 'We would need an infinite mass of photographs ofthe language, of exact notations from one moment to thè next in 01 der thus to move forward following thè course of time' (Riedlinger, Quire IV [Komatsu and Wolf 1997:63]). Significantly, in observing that this iiH-ihodology may be applied to some languages and not others, Romanista are cited as scholars whose field allows thè best application of chis approach: Romance scholars are in thè best position imaginable because chey have, in thè slice of time which concerns them, thè point of departure. But even in exceptional conditions, at every moment in an infinity of compartments there will be none thè less enormous gaps which will have to be filled by abandoning narration and synthesis so as to give another direction to thè investigation, and this investigation will gencrally fall within thè retrospective point of view. (Riedlinger, Quire IV [Komarsu and Wolf 1997:64)) 16 So thè retrospective, ot compatative, point of view is indispensable (Riedlinger, Quire IV [Komatsu and Wolf 1997:65). In Romance linguistics this conclusion, even at thè time of thè Cours, would not have been subsctibed to by everybody (see Varvaro 1968:133-64). Theoretical thinking on synchrony and diachrony in Romance linguistics has played an appreciable role in clarifying thè terms of thè early twentieth-century debate, with obvious consequences for thè work of thè Romanist. As early as thè late 1950s Coseriu had rightly pointed out thè confusion caused by greater sttess being laid either on methodology, or on ontology or definitimi: What is independent of diachrony is synchronic description, not thè real state ofthe language, which is always a 'result' of another eatlier stage, and is even for Saussure a produci of historical events. The problem arises because Saussure talks about description, even if he does not clearly distinguisi! thè 'real' and thè state ofthe language as projected. Thus ilir Saussurean antinomy when mistakenly transferred to thè level of thè object is quite simply thè difference between description and liistory, and in this sense no ionger has anything Saussurean about il exci-pi ilnterminology, and cannot be suppressed or annulled, because il is a conceptual necessity. (Coseriu 1973; 19K1:1 I) Of prime importance is thè stress laid on thè complementary, raihc-r i l i . m antithetical, relation of description, history and theory ano! ihc laci i h . M 'description and history are mutually exclusive not from thè point of view ol thè object, but as operations; that is, they are distinct operations (Cosciin 1973; 1981:18). In this sense, thè idea ofthe non-separability of diachrony and synchrony seems to receive a more lucie! formulation than thè Pragmversion: thè existence of a diachrony in synchrony and a synchrony in diachrony is a matter ofthe real state ofthe langue, not thè method or i l u observer's point ofview (see Coseriu 1973; 1981: eh. 6).5 Equally important is thè centtal position given to thè concepì ol linguistic 'tradition', defined as thè 'ttansmission' of common and currc-nt modes of spcakiiig which form thè idiomatic inheritance of a language (Coseriu 1973; 1981:31, 34). This concept, profoundly imbued with a sense of history, has numerous implications. Insofar as it refers in nini io thè idea of 'traditional knowledge', quite different from abstract universa! knowledge (Coseriu 1973; 1981:38-40), it may offer a concrete historical point of contaci between thè fortuitous and irregular nature of diachrony, inaccessible to speakers' consciousness, and thè array of associative relationships which define a language at a patticular period and which 17 Uosamia Soniicola i i nnly in iln.- perception of speakers of that period. This traditional Icnowlcdgc, infieriteci by speakers initially as an impenetrable linguistic i . m i in w l i i i l i ilu-y give new life, is a key to comprehending why die / liimniityj' and E-Ltnguage dichotomy is ill-suited to an understanding of Insinui .il picnrs.si.-s. This model artificially polarizes abstract and more or li-.vs iiniviTsal nicchanisms of linguistic knowledge and thè textual objectil i i . i i i n n ul languages as external products. But thè concepì of linguistic i r . u l i i i o n also has a cruciai theoretical implication for thè modelling of ilianj',1-. ('.lungi- does not concerti phenomena taken as mere physical or medianica! lacis, bnt thè crcation of linguistic traditions, defined as 'thè lHsunk.il objectivization of what has been produced in speech' (Coseriu 1988:149). Thcrelore 'linguistic change is thè historical process by which a lanj',n.ij',c disappcars or arises, by which linguistic traditions die out or u n n i - i n n i beili;;, and by which often new traditions partially or wholly iakc (In- "pian-" of ihose dying out\in, thè system of traditions which we cali a languagc' (Coscriu 1988:150). For many Romanists, this viewpoint (.an liardly bi- avoided. Tlii-rc is a linai component characteristic of Coseriu's diachronic thinking which strikes a cord with many Romanists: thè centrality (alongside tradiiional knowledge) of speech understood as textual production, in its multiple dimensions as generai or historical activity, and as 'knowing how io speak' (Coseriii 1973; 1981:32, 38). It has long bcen a widely held eonvKtion ihai sinilying thè processes at work in spoken language at a p a i i i a i l a i s y i u h r o n i c stage may help linguistic change 'shed its contrad k i u i y naniii 1 and ils alleged mystery' (Coseriu 1973; 1981:42). ;i Sonic iasc-Miitlics i/ OU Fftnch declension A piimi 1 sliuwiasr lui ihe difference between thè diachronic and syn! l i n . i n . pi r.|H i livi-.s, and thè problematic interaction between them, Involvei a clailii mpi> ul Komance linguistics, thè genesis of thè nominai i i i l l i . i i u n . i l ,,VMI-IIIS ul uld Gallo-Romance. The early Romance case mi '.huw np as a n y i l i i n g but homogeneous and regular. They appear i , , I,, , , , i , u n . u n i r , haphazardly cobbled together from thè remains of an . , 1 , 1 , , , l l i | , . , .1 l i n i l i l i i i g . 'l'In- continuities seem sometimes to involve | , l i , , n , M, l i i n n , Miini-iinii-s morphological structure. Romance historical I m p i l i .M, •. li.r, v a i i u u s l y siicsscd thè role played by formai, or by funcl i , MI i l ' II M IH M li S, U .in, i L i . in M, i in.I L II TI ir i r i e ' v a n i I. li r, .il u n n i In I .m l'I .1 • • a ir recapitulatcd lici i1 as i l . , • i , • • n , l m i • i .Li. Ii L .|>i, i i \I ilm i .11 ( M i l ' i i s l ili'i I r l i 1 . l i . i (I1) ' Singolar l'Iui.il ',MI|'llljl l'Imal Nominativi- ROSA 'rosi-' l(l ISAI MI in is 'w.iir MURI Accusative ROSAM Rl )SAS MIIKHM M i m i i1. Genitive ROSAI' KOSAUUM MI IMI Dative ROSAE RDSIS MI ' I 1 ' i MI US Ablativo ROSA ROSIS MHI" 1 MHKIS 1,1 ,1, , 1 Il r . . U H M l l i i u l da lenitati S iugular Plural Singolar l'Iill.il Nominative CANIS CANBS HOMO 'inaii' IIOM1NCS Accusative CANEM CANÈS HOMINEM IIOMINI'S Genitive CANIS CANUM HOM1NIS IIOMINUM Dative CANI CANIBUS UOMINI IIOMINIIUIS Ablativo CANE CAN1BUS NOMINE IIOMINIIUIS 'dog' Certain facts about this System are very salient in discussions of thè evolution of thè Romance case system, notably aspccts of thè elisi ribution of thè inflectional ending -s and thè nature of thè nominative and accusative forms: • • • • -s characterizes ali accusative plurals; -s characterizes both nominative plural and accusative plural in (he third declension; -s also characterizes nominative singolar in second (and many third) declension nouns; many third declension nouns show lormal disiiiu liuns between nominative singular and thè rest of thè paradigtn, N m a l i l y , irwi-i syllables in nominative singular than cl.sewlii-ix- ('impaiisyllal)iiiiy'), and sometimes differences of stress (e.g., NOM.SG iM 'rider' vs. IMPERATÓR- cverywhere else); 19 Rosanna Sornicola • • first declension nouns are predominandy feminine; sccond declension nouns predominandy mascoline. The third declension contains bodi masculine and feminine nouns, widiout formai distinction of gender; -M, thè marker of accusative singular, was early subject to deletion, ieading, for example, to formai identity between thè nominative singular and accusative singular in thè first declension. ( )vi-r die fina! decades of thè nineteenth century, and thè beginning of thè i wrntieth, thè collapse of Latin declension and thè transition to thè nominai Inflectional systems of Romance was discussed in detail by thè major figures o( Romance historical linguistics, who were aware, sometimes acutely, of i IH- implications for linguistic theory. Aspects óf this debate remain relevant io iliis day for our understanding of thè relation between syntacticoM-mamic ractors and formai (i.e., phonetic and morphological) factors in ilic relevant diachronic developments, and also bave considerable implicalions for thè thesis that thè change brought about by thè most widespread IIKHVXS of morphosyntactic change involvedgeneralization of thè accusative i ,i\ lorm. Indeed, this issue was thè arena for two opposing theories on thè p i i n e iples of linguistic change. 1 he drst theory, asserted and defended by Diez and Meyer-Liibke, could IH- labelled 'hypostatization of forni and predominance of mental processes'. I n i l , systemic changes are examined by taking thè forms and functions of i l n - CLat. case System as thè terrns of comparison, and focusing on any functional deviations with respect to each particular form. This approach w.is adopted towards both late Latin and Romance forms. The rbllowing are i liaracteristic of this approach: • • • • Form is an absolute parameter, in terms of which cornparability under change can be assured. Mental processes are assumed to be a more characteristic aspect of change than are substantive phonetic changes associated with production. The perspective is 'teleological', in that various forces are assumed to be in play Ieading towards thè emergence of a single universal case, thè accusative. Teleologism often goes hand in hand with functionalist approaches, but in thè present instance it seems rather to be associated with thè sharp separation between form and function. Change is studied through thè comparison of successive synchronic stages in each of which thè System appears stable and fully articulated. 20 Romance linguistics and historical linguistics The second theory, maintained variously by D'Ovidio and Schuchardt, could be called thè 'theory of transitional forms and functions'. The changes between Latin and Romance are considered as processes in which form and function maintain a certain degree of stability during thè transformation, and form is taken as thè 'external' locus (or symptom) of change. Form and function are closely bound together, yet: • • • In thè processes of change, phonetics and morphological form are accordingly considered more influential than — or at any rate preconditions for — mental representation. There is wariness about thè possible teleological implications of diachronic processes. The overall mode! focuses on thè examination of individuai facts, or thè specific (accidental) characteristics of thè transitional forms which gradually dismantle thè Latin edifice. The second theory in no way implies that thè entire process is fortuitous. ' l'o follow thè diachronic dismantling of a System requires an understanding o( die incidence of certain forms as relics, each with its own raison cl'ètre, which, like thè remains of an older edifice, will form a new one, wherc dir individuai parts stili bear thè traces of their past, yet bave been, or are bdiij 1 ,, reassembled into a new construction. The 'reasons' are inextricably bound up with thè nature of thè historical processes. In history nothing is creàtcd and nothing destroyed. Cataclysmic changes excepted (but not always, rvrn then), innovation always passes through thè remodelling of pre-exlstlng structures: thè material and thè structures persist, albeit often in alterca form, but their functions are redefined. These principles are only apparently reflected in thè first theory, which considers thè persistence of forms from an absolute, rather than relative, stance and interprets functions exclusively in semantic terms. The two sidr.s of thè sign, thè static signifier and thè dynamic signified, are separateli by diachrony. The accusative form is seen as having 'usurped' thè functions o( thè other case-forms. But such an approach is profoundly anti-historical, in that entities are stili postulated which exist only relationally (inasmuch as they are dcfined in relation to one another), when they no longer survivc. Here ali thè unresolved contradictions of a structural diachrony are plain. Characteristic of these problems are thè kinds of methodological and theoretical positions assumed in a debate which involved many frontranking Romanisrs (such as Schuchardt, Ascoli, Mussarla, Tobler, MeyerLùbke), espccially with rcgard to D'Ovidio's (1873) thesis on thè origin of Latin nominai inflection. D'Ovidio had rejectcd Diez's thesis that thè 21 Rosami.i Sornicola L '..iim- w.is ilu- basic form underlying thè OFr. and OPrv. objectcase, as ut II .r. i l n - sini'Jc case form of Italian and Spanish. D'Ovidio had asked w l n - i l n - i i l n - two-casc stage had occurred not just in French and Provencal, I M n .iKo in Spanish and Italian.7 His fundamental thesis, based on reflection mi l i . i l i , i n , w.is that thè simplification of Latili declension had been thè produci of a graduai reciprocai Icvelling of ali thè cases. Avaluable article by Si lini l i . u d ì givc.s a positive analysis of D'Ovidio's argument and ali its i m p i l i iiiions, survcying each of thè various positions in thè ongoing debate .11 i l i . n limi-, and drawing a wide-ranging and sharply observed picture of i l u - pimi-ss by which thè Latin declcnsional System disintegrated, and of ihi 1 v.nienis siages in thè development of thè Romance inflectional System. 1 lis siimmary olD'Ovidio's thesis contains numerous pithy observations of consideratile theoretical and methodological weight. Schuchardt observes (1874:167f.) that, in D'Ovidio's view, Diez was guilty of setting unadulleraii-d I .ai in sicle by side with thè Romance paradigm and seeking thè must diteci link between them, rather than patiently tracing thè graduai disniamling of tlie Latin System. Schuchardt acknowledges that D'Ovidio's du nonsi radon is seductively clear. Italian campo 'field' may derive from ihe nominative CAMPUS, accusative CAMPUM or dative-ablative CAMPO (bui not from genitive C:AMPI); amore 'love' only from accusative AMOREM, dative AMORI or ablative AMORI: (but not from nominative ÀMOR or genitive AMORIS); corpo 'body' only from thè nominative or accusative (hoth CORPUS),' but not from genitive CORPORIS, dative CORPORI or ablative conpoKi'.. So far as ali tlirce Italian forms have a common origin, they can only come from thè accusative. Various sound changes - loss of Latin linai M, postulated loss of -s, merger of long O with short Ù - would have Ic-d io CAMPUM and CORPUS becoming campo and corpo. But in this case, 1 )'( )vidio observes, campo can also come from CAMPUS: so is ir really righi to say l l i a i (In- all'usativi 1 forni campo usurped (replaced) both thè nominative .imi l i l i - d.itivi- ablaiive forms, which were identica! in form {campò) to thè .ic t us.iiivr? Si Ime li.neh l i m l u T synihcsizes two important principles: that in change ilii-u- .iu- un m v s i e limis forccs at work inexorably leading towards definite u - M i l i s .uni i l i . n , in generai, menta! (functional?) processes are subordinate in | i l i v - . n > l i i | ' , i i . i l (phonetic) processes. He asks (1874:168) wherc thè impulsi in i r p l . n e - l . a i i n cases with just one or two forms might come l u m i , .uni sii|',|',cMs i l i . i l il is rash to assume that such a mysterious linguistic I M I | » I | > i-, n . i l l y ai woik, lor menta] processes are subordinate to physioliij'ji .il i n M -, D'Ovidio is righi in saying that there is always amental process I m morphological transformations, but (Schuchardt proposes) Romanci- linguistica and KiltOrlcd l i i i i ' . u i i !• thè 'tracks' along which thè minii movcs .in l.nd dnwn In- |>l transformations. At this point we necd to stimma i i/o thè dlvlslon Into dcclenuloiiul i LIV.I •• common to many grammars, and reflectlng i l n application "I ' l i l i . criteria: gender, morphological s l u n i i i i e (A, \\ ('., l\. I n l n w ) .uni declensional patterns (C, H, bclow). Yei thè various accounti an nniiu .ihlv heterogencous, using these criteria in dillcivni ways, j'.ivni)', |',u .ni i \ ij'.lii in synchronic and analogica! aspects (sudi as iLissilii.nioii un iln h.r.r. ni i l i r morphological characteristics of word sirucuiic), or siii-.ssiii|>, ,i i l i . n i n n i n e approach (by considering thè originai membenhip ol .1 I H M I I I in ,i |i.nin ul.ir Latin declensional class). Some accounts inu-rwi-avi- i l » 1 v.ninns e i i u - i ì . i and perspectives. In many, especially of thè synilnoiik s i n n i m . i l lypi 1 , |',i-inln acquires special importance not only dcscriptivcly, bui alvi .rs .1 inni lor cxplaining change on thè bxsis of analogy. Actually, thè roli-ol gcndri cvolvcil gradually and in a by no means linear fashion. Tlieiv is .1 e onsidciablc diachronic continuity. The Latin first and secoml cleclensions, i In- li-xeme.s which belonged to them and their respective inflectional paiu-ms survived in an etymologically regular way in OFrench. They weti- alieady i Lirly linkcd with gender in Latin (thè first declension was largely feminine, thè second largely rnasculine), and thereforc they could bave cxerciseil analoglcal a i i i a i tion not only on Latin nouns belonging to other declensions, but also on French third declension F and M nouns. But thè liisiorie.il irality may have been much morecomplex. In any case, we sce bereanother aspa i ni i l i c i lash between diachronic (see Paris 1872) and synchronic (see Meycr-Lubke 1894:§21-24) perspectives. This clash also brings out thè important issue oftcn overlooked — of die mismatch between Latin and Romanci- ca.ses: Ini example, it is inappropriate to label thè 'subject' 'nominative', and 'suliject' is itself an unsatisfactory label since thè case-forni in questuili ciundes more than just thè subject function. Here are thè basic facts about OFrench declension classvs. Class A. Feminine. Type fitte 'daughter' (< |'||,IA/I : II,IA(M)) SGfille VLfilles This class contains thè continuants ol thè I .min I n s i decleniion largely composed of feminines. Class B. Feminine. Types^ztt 'end' (< PlNlS/FINE(M))i ffMtAv'tnother' (< MATF.R/MAJ'RE(M)), «te"city' (< I;IVII'AS/CIVIIAII ( M ) ) SGfin(s), me dre 23 PLfins, medres Rosanna Sornicola Romance linguistics and historical linguistics This class contains thè continuants (largely feminine) of thè Latin third declension, comprising both originally parisyllabic and originally imparisyllabic forms. However, thè latter had already been remodelled into a parisyllabic pattern in late Latin, through generalization of thè oblique stem (such as maìson 'house', from MANSIO/ MANSIONE(M), cité 'city' from C3vrrAS/avrTATE(M), ere.). Nouns of this class do not consistently conform to thè declensiona] pattern. Class C. Feminine. Type nonnelnonnain 'min', sner/seror 'sister' (< SOROR/SORORE(M)) is thè fact that these nouns conform to thè declensional pattern of thè type mur, characterized not only by a particular structure but also by relative regularity. It must also be stressed that thè overal) inflectional structure of this class is an entirely etymologically regular development from Latin. Class E. Masculine. Typepedre 'father' (< PATER/PATRE(M)) SG PL S nonne, suer S nonnains, serors (suers) OBL nonnain, seror (suer) OBL nonnains, serors This class also comprises a number of proper names. It lumps together different patterns of case altcrnation (thè Germanie type -el-aìn against thè Latin type -OR/-ORJS). And there are some irregularities in thè alternation between thè subject stem and thè oblique stem. Class D. Masculine. Type murs 'wall' (< MURUS/MURU(M)) SG PL S murs S mur OBL mur OBL murs SG PL S pedre(s) S pedre OKLfedre QSLpedres This class can be defined, on structural criteria, as parisyllabic masculines whose stem generally ends in a vowel. Diachronically, these are continuants of Latin third declension parisyllabics, pilli some continuants of nouns in -ER from thè second declension and some from third declension imparisyllabics in -OR, which had developed as parisyllabics (e.g., arbre 'tree', from ARBOR/ARBORK(M)). Nouns of this class conform to die declensional pattern in an inconsistent fashion, Moreover, thè plural subject case forni is clearly not an etymologically regular development from Latin. To justify i his pattern appeal has been iliade to analogica! attraction by thè subjcx \e forni of thè class D masculine Class F. Masculine. Types conslconte 'count' (< COMES/COMITE(M)), ledre/ladron 'thief (< LATRO/LATRONE(M)), enfes/enfant 'child' (< INFANS/INFANTE(M)) SG PL S cons ledre(s) enfes S comte ladron enfant OBL comte ladron enfant OBL comtes ladrons enfunz This class is generally characterized by thè fact that its members have a rather regular and specific inflectional pattern for case. By synchronic and structural criteria this class might be described as that of masculine parisyllabics whose stem ends in a consonant, but it is in fact often defined by diachronic criteria: it comprises continuants of thè Latin second declension, or of lexemes assimilated into that declension. The potential for clash between diachronic and thc other criteria is clear from some accounts which include in this class continuants of thè Latin third declension (parisyllables such as reflexes of CANIS, PANIS, or imparisyllables which have become parisyllables, such as thè reflex of LEO/LEONE(M); sec Brunot 1966, 1:181). Clearly in this case thè criterion for inclusion in class D This class comprises masculine nouns which have conserved thè Latin alternation between parisyllabic and imparisyllabic stems, thc former being specialized as thè subject case forni and die latter as thè oblique. Nouns of this class may also show stress variation: invariant (e.g., cóns/cómté) vs. variant (e.g., lédre/ladrórì). Wliere Latin stem alternante are dius preserved, structural and diachronic description coincide. But thè picture is complicated by thè fact that an originally second declension noun such as prestare (< PRESBYTER), and nouns derived from Germanie bases in -o, -óne [...] are included here. And die inflectional pattern is only pardy 24 25 Kosunna Sornicola ciymological. Besidc die lack of correspondence between die plural tubject case-Form and thè Latin nominative forms (where orice again analogica! attraction by thè mascoline class D has been invoked), it needs to be stressed diat many lexemes are notably cnadc in dieir adoption of zero or -s as thè singolar subject case inllection. Class C. Femmine indeci i nables. Typepais 'peace', voiz'voice ( llass H. Mascoline indeclinables. Typc nes 'uose', sens 'sense' Tlu- nn.sysiemaiic nature of OFrench declension is clear from many irregul.ii iiics, involving varioos kinds of theoretical issoes, which can be labelled as s i i n e -turai irregularitics (lack of structoral isomorphism between thè various nominai classcs) and empirical irregularities (thè numerous cases of failure to conlorni io die paradigms, in die manoscripts which preserve thè texts). I l.illway between thcse are 'lexical singularities', where some lexemes consiituR 1 obvious exceptions withìn a paradigm which is otherwise clearly characterizable. Forexample, thè MSG/z/y'son' (< FILIUS), invariant for case and revealing lexicali/,ation of thè originally second declension nominative -s. Illese onlionics clemonstrate diat cven continuants of thè Latin second dcdeiision, which at every stage of OFrench constitute thè most unwaveringly regolar class of mascoline noons, do not torni a onitary bloc. Sudi anomalics are of considerale theoretical interest in that they allow us to induce more generai developmentaJ principles, with regard to semanlir I.K lors sudi as |+animatc] and [+human|. Butthe roleofsoch factorscan hardly bave hcen regolar either. As Pope (1934:§805) notes, in thè paradigm »! parisyllabic nouns thè appearance of-^and -z 'carne to be regarded .is t l i r chancterlsdc flexion of thè Nom. Sing. Masc.', but this affected 11.1 n ii s o! diings somewhat earlicr than proper names. Even more irregular is tlii- developmenl ol proper names, making it difficult to invoke thè influCIH e di i l n - Animacy hicrarchy (Schosler 2001b:174, 102). I IH-IC i.s .ilso considerable case variation according to syntactic contcxt. As Woledj'r i-1 ili. ( I 967) bave shown for thè Chanson de Roland, some syntactic M I I H i m i |, i ' u n , ul.irly those in apposition or lacking an explicit predicate, i. mi in I.mini thè oblique case forni over thè subject forni (see also Moignet l ' I d i , i n i i ' i ) Vrt, in die same syntactic contexts, some noun classes (Impurliyllabics) temi lo slay closer to thè regular use of case foniis than Dilli i l u i . 11. M r. \\\. ihii s like rcw'king', and even more masculines in -e, and i. miniili . .ne le.s.s seiisiiivc lo die etfccts of syntactic context. Feminines are M n i .11 \. i li 11.il ik'. 'l'Iu-se are idiosyncratic lexical developments, reflecting i n i M | i | . i . . u n .I i m i scinaniic factors, at timcs collaborative, at times I . l i . in. . l l l r - m I I, . HI.I I I. il 1 antagonisiii (see WoledgC tf tf/ l ' i i i ' M I I . considcratìons could prevali over syntuct li "i I I il i . . i . . . D| . . I beingUSed (as al ready scen in dir < '/>,iii\t>n ili l\nl,niil I i n d i l l i n m li . mordili^ to requirementS ol assonami' and i n e l i e (\X'oKd|',r i l ,// l ' I l i l i . i , I \, 1882:6 for Anglo Nomi.in). ( rcographk al dlffcrcntlatl i | mi i un Inclecd, Stanovaia (1993) attribuita tlu- gic.n vorlablllry ni ( ' I n IH li dn Ini sion not to a 'systcm', but to t e m n a n i s ol .1 d e s i h u n l di i li ir. .il lyitcm, preservcd in some scriptae, and reanangcd in oihcis. Overall, we can hardly postulale a 'systcni'. I I H - h v l u i d n . i t i n e ol die case markers also shows that we are deallng not widi .1 e ohe-1 cui svslrm, bui with rclìcs of an older array ol. lonns w h i i h li.ive beeu rxicnded ami remodelled in different ways according io piai e and u x i n a l i i a d i i i o n s . The inflectional system seems to bave atrophled in die elidine, .1, whidi is associated with thè singulàr subject (unction (originally niainly masculine, but later spreading imo third declension feininines), or an oblique plural function, or simply plural. Bot in addition to case-emlings i h e i e is .liso diachronically notably persistent - allomorphy of thè rooi, orij'Jnaiing in Latin imparisyllabic masculine third declension animates. In ali, case marking is better preserved in thè singulàr than thè plural (see Sch0sler 2001b:170); and there is syncretism belween die subject singulàr and obliqoe plorai case forms in -s. Syncrctism, like allomorphy, is inherited from Latin, bot as fragments which are reorgani/ed according to new pattcrns of paradigmatic relations. Comparison of thè Latin and OFrench paradigms may show thè extent of what has changed, but scarcely constitutes an explanation. Rather, we may perhaps say that thè conditici) mentioned above resulted from and abetted greater unpreilictability and instability. And thè lack of alternation between subject and oblique case forms tlerived from thè first declension, both in thè singulàr and thè plural (showing thè characteristic western Romance opposition belween /ero in thè singulàr and -s in thè plural) constitutes a gap in thè sysiem. The rolc of analogy in thè relation between diachrony and synclirony is cruciai and problematic. Analogy is reaily a synclnonk lai t u r wliose dia chronicuse may clash with other mechanismsoftranslbrmaiion, and whose explanatory force remains very uncertain. Nor is il dear in what rel.iiion it stands to thè various chronological sequences of evenls or exailly wli.u iis role was in thè various diachronic stagcs under examinalion. Ali scholars agree on thè analogical nature ol die n e i i n a l i / a i i o n ol i.ise distinctions in continuants of thè Latili l'irsi dei lension, based on loss ol singulàr -M and loss of case distinctions in t h è .siii|',nlai, bui noi on duorigins of plural -s. Some trace Gallo-Roma IH e i bai k io du I . u n i .m usai i ve 27 Rosanna Sornicola Romance linguistics and historical linguistics This class contains thè continuanti (largely femmine) of thè Latin third declension, comprising both originally parisyllabic and originally imparisyllabic forms. However, thè Iatter had already been remodelled into a parisyllabic pattern in late Latin, through generalization of thè oblique stem (such as maison 'house1, from MANSIO/ MANSIONE(M), rité 'city' from CIVITAS/CIVITATE(M), etc.). Nouns of this class do not consistently conform to thè declensional pattern. Class C. Feminine. Type nonnelnonnain 'min', suer/seror 'sister' (< SOROR/SORORE(M)) is thè fact that these nouns conform to thè declensional pattern of thè type mur, characterized not only by a particular structute but also by relative rcgularity. It must also be stressed that thè overall inflectional strutture of this class is an entirely etymologically regalar development from Latin. Class E. Masculine. Typepedre 'father' (< PATER/PATRE(M)) SG PL S nonne, suer S nonnains, serors (suers) OBL nonnain, seror (suer) OBL normaiìis, serors This class also comprises a number of proper names. It lumps together different patterns of case alternation (thè Germanie type -el-ain against thè Latin type -OR/-ORIS). And there are some irregularities in thè alternation between thè subject stem and thè oblique stem. Class D. Masculine. Type murs 'wall' (< MURUS/MURU(M)) SG PL S muri S mur OBL mur OBL murs SG PL S pedre(s) S pedre OBL pedre QELpedres This class can be defined, on structural criteria, as parisyllabic masculines whose stem generally ends in a vowel. Diachronically, these are continuanti of Latin third declension parisyllabics, plus some continuants of nouns in -ER from thè second declension and some from diird declension imparisyllabics in -OR, which had developed as parisyllabics (e.g., arbre 'tree', from AKBOR/ARBORE(M)). Nouns of this class conform to thè declensional pattern in an inconsistent fashion. Moreover, thè plural subject case form is clearly not an etymologically regular development from Latin. To justify this partern appeal has been made to analogica! attraction by thè subject case form of thè class D masculine declension. Class F. Masculine. Types coni/conte 'count' (< COMES/COMITIÌ(M)), ledre/ladron 'thief (< LATRO/LATRONE(M)), enfeslenfant 'child' (< INFANS/INFANTE(M)) SG PL S coni ledre(s) enfes S comte ladron enfant OBL comte Iddron enfant OBL comtes ladrons enfanz This class is generally characterized by thè fact that its members have a rather regular and specific inflectional pattern for case. By synchronic and structural criteria this class might be described as that of masculine parisyllabics whose stem ends in a consonant, but it is in fact often defined by diachronic criteria: it comprises continuants of thè Latin second declension, or of lexemes assimilated into that declension. The potential for clash between diachronic and thè other criteria is clear from some accounts which include in this class continuants of thè Latin third declension (parisyllables sudi as reflexes of CANis, PANIS, or imparisyllables which have become parisyllables, such as thè reflex of LEO/LEONE(M); sec Brunot 1966, 1:181). Clearly in this case thè criterion for inclusion in class D This class comprises masculine nouns which have conserved thè Latin alternation between parisyllabic and imparisyllabic stems, thè fbrmer being specialized as thè subject case form and thè Iatter as thè oblique. Nouns of this class may also show stress variation: invariant (e.g., cóns/cómtè) vs. variant (e.g., lédrelladrórì). Where Latin stem alternants are thus preserved, structural and diachronic description coincide. But thè picture is complicated by thè fact that an originally second declension noun such as prestare (< PRHSBYTER), and nouns derived from Germanie bases in -o, -óne [...] are included here. And thè inflectional pattern is only partly 24 25 Rosanna Sornicola plural -AS,1 while others believe that this inflection cannot be connected with any Latin case form (see Schuchardt 1874:163). The continuants of thè Latin second declension with nominative -s are, un thè whole, a rather regalar and consistent locus of conservation of morphological and phonetic structure. Thus: S SG -s PL-0 OBL SG -0 PL-* l ' I i c major diachronic problem is thè development of thè Latin third declension (see Paris 1872:110). Diachronically, thè main division is, in i l n - lingula!, between nouns of imparisyllabic and parisyllabic origin. The l< n mei show different forms for subject and oblique singular (e.g., comi «unii-, hom/home); thè latter (e.g., thè continuants of FRATER, PATER) lack i .r,r .illnniorphy, and some have invoked analogy and paradigmaticlevelling in acvount for thè remodelling of their inflectional pattern on class D (lingula! subject murs I oblique mur), giving rise to an alternation between MibjiTt li [>eres, HJrejres and oblique le pere, lefrere.1^ Such Icvelling appears vai i.ihly in Anglo-Norman texts.16 Yet later -s was allegedly added even to Imparisyllabic nouns which already displayed allomorphy for case (e.g., in. cinpereres, sires; see Paris 1872:lllf.). For some this addition of -s took root in thè French introduced into England (thus Paris IH72:11 1 f.), but thè available evidence shows a more problematic situation. In die Cambridge Psalter -s is usually lacking in thè subject case form of Imparisyllabic nouns, although there are a few counterexamples, especially i//v\. See (Breklce 1884:8) for thè Voyage Saint Brancicai. Othcr scholars hold that -s originally appears in thè singular only in words where it is etymologically justified, its extension being a rather late phenomenon fundamentally due to analogica! adjustments, and more i li.iracteristic of Anglo-Norman and western French texts, so that careful pocis likc Wace and Chrestien only knew forms without -s17 - a view Contradictad by Woledge's findings (1979:18f.) from thè manuscript iradiiion of Chrestien de Troyes. Contrary to what one would predici from Latin, from thè very earliest Gallo-Romance documents we find no -s in thè subject plural form. This does not necessarily mean that third declension subject plural -s had already been lost in thè lower sociolinguistic registers of late Latin.18 We could, again, be dealing with attraction by thè Romance paradigm comprising nouns derived from thè Latin second declension.19 Clearly these are speculations which stress synchronic analogica! mcchanisms: Schuchardt 28 Romance linguistics and historical linguistics (1874:161, ni) saw that this could not be taken for granted. After ali, in Merovingian documents both thè nominative and accusative plural forms of thè third declension are well preserved, and indeed thè nominative has about a l O percent higher rate of conservation than thè accusative. Problematic as thè testimony of sudi documents may be, it is by no means proven that nominative plural third declension -5- was lost in late Latin. Even more problematic is thè development of Latin third declension feminine nouns (OFrench class B) conserving Latin root-final consonants (dolor(s), genz/gent,flor(s), maison(s), vertet/veitez, defension(s)). Here too we find marked inflectional variability according to region, period and text. As for second and third declension masculines (classes E and F), Chrestien de Troyes has a fairly regular two-case System: in thè singular (with a few idiosynratic exceptions) we have -s in thè subject form and no -s in thè oblique; in thè plural, like First declension feminines (class A), -s appears in subject and oblique forms alike. The long-sranding discussion as to whether this reflects an ancient state of affairs21 has been complicated by competing erymological and analogical arguments, obscuring thè faci that between Classica! Latin and Romance multiple transformations — far from regular either in time or space — must have occurred. Chrestien shows conspicuous changes with respect to Classical Latin. Firstly, levelling of Latin imparisyllabic stems on stems with a greater number of syllables, Le., remodelling of thè nominative stem on thè basis of thè oblique stem. So Latin allomorphy of thè type NOM FI.O-S GEN/DAT/ACC/ABL FI.OR- NOM MANSIO GEN/DAT/ACC/ABL MANSION- NOM UIRTU-S GEN/DAT/ACC/ABL UIRTUT- gives way to thè stems FI.OR-, MANSIONI-, UIRTUT-, whence nominatives FlDR(fi), MANSION(E), UIRTUT(E), alongside oblique forms with thè same stem. The documentation suggests that this phenomenon must already have been charactcristic of late, and especially Merovingian, Latin. The second change concerns non-etymological subject singular forms in -s and built on thè oblique root (e.g.,flor, vertezvs. Lat. FLOR, UÉRITAS). But is this an analogical development that arose during thè twelfth century and established itself in Chrestien and others, or a survivor of late Latin popular forms? The picture has been further complicated by thè discussion of thè problem in thè context of investigating thè dissolution of thè two-case system. In fact thè earlier phases need to be examined on their own terms. 29 Kosaiina Sornicola ' l ' I n - difettici] of thc diachronic process doses not emerge clearly from an i n i i n . i i i o n ol thè documentation, highly variable both geographically aiid [extually. B u i , despite appreciable differences both in and between texts, Mime Anglo-Norman texts actually do show a notable incidence of -s. The ( ) x l o i d l'suhcr, and thè Cambridge Psalter (mid twelfth century) have p l r n i y ol loims in -s, with some morphological distinctions: thus in thè ( Limbi idgc l'salter subject -s occurs in 40 percent of derivatives of thè Latin lype AS/ AI IN, and in nearly 80 percent o f derivatives of thè type -O/-ONIS. M u d i lowcr porcentagcs are found in ali other types (see Fichte 1879:81). Snbjcu A is Ircquent in later Anglo-Norman texts (see Vising 1882:96—98; mekke 1884:21). Tliis lexiual and geographical distribution makes Meyer-Liibke's (1894:§ 21) and others'22 division between an Anglo-Norman and a French ami Provcn^al dialect area look implausible. The latter area, from antiquity, allegedly distìnguished subject flon from oblique flor, held to conserve a vnlg.ii 1 .alili situatìon,23 while Anglo-Norman deviated from it by keeping -s. Tliis rcsurrects thè thesis that Galloromania should be divided into areas thai prcserve thc late Latin situation and those that rapidly broke away from it. The thesis (see Vising 1882:12f.; Schosler 1984:1/1-73), that this situation is due lo impcrfcct learning of French in England, has some sociolinguistic jusdficadon but should not be followed uncritically as an explanadon of inflecdonal vacillations in thè earliest Anglo-Norman texts. The rea I imerest of such vacilladons is that diey allow us to glimpse differences in thè reorganization of thè 'ruins' of an earlier infiectional sysicni so that perhaps we should make a criticai reappraisal both of thè Ynni i m i i t y ' thesis (FSG.OBL -s infins and maisons etc. gocs back directly to lato I,alin lypes with remodelled subject -s)2 and of thè thesis which asdihes il lo analogical innovation in OFrench Bere, differing attitudes ol i l u - w i i i n . s lowaixls existing linguistic traditions might bc decisive. That i l u - ( )xlnrd and Cambridge psaltcrs have many class B femmine forms w i i h s'iugular subject in -s might be related to their Latinizing orientation (see I u n i c i 7.007). Similar considerations might hold for thè regular I I I C M - I U i- ol i in ( 'Jncsiicn. Such orientations, characterisdc of particular ir 11/'/ni i,i or i n i n n a i cnvironments, should not immediatcly be assigned to a I I . I I I K ni.n .iiea, k-t alone to a line of diachronic development. An.ilo|',v h.is bini extensively invokcd. Both Paris and Meyer-Liibke i m u n ì i l i . n ilii- singular subject case forms without -s in thè second . 1 . , li n . . . i l .uni i l i i u l declension femmine, as found in thè earliest AngloN i n i 11.m i I-M', . u n i 11 mi i ncntal texts afterChresticndeTroyes, are due to thè m l l i i i IH r ni 11 ir ( ) I ' ' i c i i i h paradigmoffirst declension feminines (class A), . U) Roma i ni' linguistici and hlstorlcul llii| and both postulale analogical dcvelopmcnts tojuntKy forni» 111 eiyinolti^kal or non-etymological -s. Paris'smodelinvokesanalogica] mechanlimi in wliii li p/ mli i r . e \ | i l u n l \l as a powerfri plural subject case forni in Ol'Y. noims ol i I.ISM-.S I 1 , and I 1 , io m.ikr i l i c m like thè plural of class D, also cansed thc diiappeanncc "I i l i < Mii|',nl.n subject case forni of feminines of class B, bringing thefli in I m i - w i i l i i l.iss A feminines, so that ali feminine wonls dcclincil in i h c samr w,iy, or i . i i l i n that femmine declension consistcd of no more ( ha 11.1 gemici d i s i i m l i m i (see Paris 1872:114). This is strongly teleologieal, .imi il is noi i Ir.n .11 wlial diachronic (or diastratic?) point such mecli.misms might have operateti. Latili or Romance? Moreover, Paris (I 872:114) assnmes i l i . n i l i e spicail af-s into OFr. feminine singular nouns or class B musi li.ive o n i i i i e d Lue, according to an analogical influcnce exercised by musatimi- singul.ns in -s on feminine nouns, but this seems to involve a quile implansible invcision ol thè priority accorded to gender in thè mcchani.sin ol thè analogy (see Schuchardt 1874:161, n). Yet thè analogical model might have an interesting diachronic basis in i he plural. Some have held (e.g., Paris 1872), lli.ii i d i - n i i i y beiween l e m i n i n e plural subject and oblique case forms (in -s) faithfully consci vcs i he Latili morphological structures, while thè mascolina baii deviated Irom ihcii Latin antecedents. This view may have its a t n a c i i o n s , bui i h c d i l l e n - m i . i l role attributed to analogy in respect of gcnclersccins cxiessivc. On thè Other hand, thè model itself implies that if 'attractors' wcrc ai woik, ilu-se i annoi have been purely semantic, but also formai (moiphologie.il). To conclude, thelirnits to thè analogical hypoihcscs sccm io tic in i he l.u i that they postulate abstract synchronic staies which are dillu ni) lo delei mine historically, and cannot easily be reconciled w i i h i l i e ,u iu.il complexity of thè data. For example, on thè analogical accollili, ihe s|ne.ul ol i io feminine singular subjects with root-final consoli.inis w.is due io i l u - i n l l n enee of thè masculine singular subject forni in .\. M i n w c - h . i \ M ni i l u i ihe ancient cote of such forms compriseli comimi.inis ol i l i e I . u n i M I O I H | declension, while continuants of thc ihinl deilension muli iweni i o i r . u l erable vacillation in acquiring -s. So thc hypotheill "I llmultUlCOUl .imi identical analogical attraction on feminines Uld tnaicullnei lllki leemiCVCn less satisfactory than a chain of analogics l i e | > i m i i i i | - w i l l i I I I . I M n l m . . M I , I then affecting feminines. In any case, sij'.nilii .mi u.m-, ol l o i i i r . m \ Im both masculines and. thircl declension l e m i n i i h m i l i , i , I i , , I» li.un,I m Anglo-Norman texts. Analogical explanations lnmlci|U«ti uni for thè diachronic process. 31 Rosanna Sornicola plural -AS, while others believe that this inflection caiinot be connected with any Latin case Form (see Schuchardt 1874:163). l'he continuants of thè Latin second declension with nominative -s are, un die whole, a rather regolar and consistent locus of conservation of morphological and phonetic structure. Thus: SSG-; PL-0 OBL SG -0 PL-5 I IH- major diachronic problem is thè development of thè Latin third • In U'iision (see Paris 1872:110). Diachronically, thè main division is, in ilu- ilngular, between nouns of imparisyllabic and parisyllabic origin. The I O I I I H T show different forms for subject and oblique singular (e.g., consl i muli; hom/homè); thè latter (e.g., thè continuants of FRATER, PATER) lack i .INI- allomorphy, and some have invoked analogy and paradigmatic levelling io .Kcount for thè remodelling of their inflectional pattern on class D (•.nijMilar subject murs I oblique mur), giving rise to an alternation between '.ubjiri li peveé, lifreres and oblique le pere, lefrere.1^ Such levelling appears v.u i.ibly in Anglo-Norman texts.1 Yet later -s was allegedly added even to Imparisyllabic nouns which already displayed allomorphy for case (e.g., homi:, empereres, sires; see Paris 1872:1 ll£). For some this addition of -s iii-ver look root in thè French introduced imo England (thus Paris I K72:111 f), but thè available evidence shows a more problematic situation. In ihe Cambridge Psalter -s is usually lacking in die subject case forni of Imparisyllabic nouns, although there are a few counterexamples, especially MIYS 'sire'. See (Brekke 1884:8) for thè Voyage Saint Brandan. Other scholars hold that -s originally appears in thè singular only in words where it is etymologically justified, its extension being a rather late phenomenon fundamentally due to analogical adjustments, and more e haracteristic of Anglo-Norman and western French texts, so that careful pocts like Wace and Chrestien only knew forms without -s17 - a view COtltradìcted by Woledge's findings (1979:18f.) from thè manuscript uuduion of Chrestien de Troyes. Contrary to what one would predici from Latin, from thè very cari ics t Gallo-Romance documents we find no -s in thè subject plural l'orni. This does not necessarily mean that third declension subject plural -s had already been lost in thè lower sociolinguistic registers ol late Latin. We could, again, be dealing with attraction by thè Romano.- paradigm comprising nouns derived from thè Latin second declension. Clcarly illese are speculations which stress synchronic analogical mechanisms: Schuchardt 28 Romance linguistics and historical linguistics (1874:161, ni) saw that this could not be taken for granted. After ali, in Merovingian documents both thè nominative and accusative plural forms of thè third declension are well preserved, and indeed thè nominative has about a l O percent higher rate of conservation than thè accusative. Problematic as thè testimony of such documents may be, it is by no means proven that nominative plural third declension -s was lost in late Latin. Even more problematic is thè development of Latin third declension femmine nouns (OFrench class B) conserving Latin root-final consonants (dolor(s), genzJgent^flor(s), maison(s), vertet/vertez, defension(s)). Here too we find marked inflectional variability according to region, period and text. As for second and third declension masculines (classes E and F), Chrestien de Troyes has a fairly regular two-case System: in thè singular (with a few idiosynratic exceptions) we have -s in thè subject forni and no -s in thè oblique; in thè plural, like First declension feminines (class A), -s appears in subject and oblique forms alike. The long-standing discussion as to whether this reflects an ancient state of affairs21 has been complìcated by competing etymological and analogical arguments, obscuring thè fact that between Classica! Latin and Romance multiple transformations — far from regular either in tinie or space — must have occurred. Chrestien shows conspicuous changes with respect to Classica! Latin. Firstly, levelling of Latin imparisyllabic stems on stems with a greater number of syllables, Le., remodelling of thè nominative stem on thè basis of thè oblique stem. So Latin allomorphy of thè type NOM FI.O-S GEN/DAT/ACC/ABL FLOR- NOM MANSIO GEN/DAT/ACC/ABL MANSION- NOM UIRTU-S GEN/DAT/ACC/ABL UIRTUT- gives way to thè stems FI.OR-, MANSION-, UIRTUT-, whence nominatives FIJOR(K), MANSION(E), UIRTUT(E), alongside oblique forms with thè same stem. The docurnentation suggests that this phenomenon must already have been characteristic of late, and especially Merovingian, Latin. The second change concerns non-etymological subject singular forms in -s and built on thè oblique root (e.g., flor, vertezvs. Lat. FLOR, UÉRITAS). But is this an analogical development that arose during thè twelfth century and established itself in Chrestien and others, or a survivor of late Latin popular forms? The picture has been further complicated by thè discussion of thè problem in thè context of investigating thè dissolution of thè two-case System. In fact die earlier phases need to be examined on their own terms. 29 Romance linguistics and historical linguistics Rosanna Sornicola The role of syntactic factors in these complex developments is equally controversial. Por Schuchardt (1874:161), subject and oblique case forms were kept sharply distinct, in thè respective functions of subject and object/ complement, in thè earliest French texts. Schuchardt (1874:162) also bclieved that since thè oblique case forni was gaining ground from thè nominative in early medieval Latin documenta, we bave genitive, accusative .iiul aolative case forms in piace of thè nominative.But it is hard to see that syntactic factors, let alone functional ones, such as differentiation of subject .iiul object, had been at work here, for one would expect to see a much more icj'.uhir paradigmatic distribution of case allomorphs. In fact, thè two-term opposition is partial and asymmetrical. No such opposition had existed in I c i n i i i i n e singular nouns derived from thè Latin first declension since very i , n l y limes (see Schuchardt 1874:163; Sch0sler 1984; van Reenen and Siluislcr 1988). There have also been considerable oscillations in nouns 1 1 "i 1 1 i he Latin third declension, both parisyllabic and imparisyllabic. In thè l ' I n i . i l , thè two-way opposition is ali but extinct except in forms derived l u m i thè second and diird declensions (see Schosler 2001b:170). But whatcvct thè role of syntactic factors (see Moignet 1966; Woledge et al. l')(>7 69), they cannot have been thè 'engines' of thè construction of thè l ' i ( • ( .nious OFrench System: conservation of phonological structures and die aitractive torce of morphological patterns were also at work. The role of syntactic factors in thè collapse of Latiti declension I lic iole of syntactic factors in determining other types of linguistic change is one of thè most complex and controversial questions in diachronic llnguistics. Once again, early twentieth-century thinking has implications wliich have stili to be fully taken on board. Indeed, Saussure actually wonders whether syntagms and psychological associations, typically Usigncd to synchrony, do not also have their own history, and observes: As soon as we get outside of pure phonetics it is in fact much more difficult to draw thè limit or to state a radicai opposition. This is thè most diffidili pari of thè generai division, but I cannot insist on it without getting into delicate considerations. However, in an infinity of cases we will see that facts we think are grammatica! reduce to phonetic facts. (Riedlinger, Quire IV [Komatsu and Wolf 1997:67]) The fate of Latin declension is Saussure's main example. The thesis that thè complex transformation of Latin declension in Romance can be reduced to a simplifìcation due to confusion of fina! segments was as controversial tlien as 32 today. For Saussure, this is hard to prove but not wholly implausible. He holds that one has at least to acknowledge an ordered sequence of two facts: thè phonetic and diachronic confusion of final segments, and thè introduction of a grammatical - hence synchronic - System (Riedlinger, Quire V [Komatsu and Wolf 1997:68]).2S This model is consistent with thè more generai theory of thè relation between synchrony and diachrony. Yet Saussure also seems to be trying, without success, to find a way out of a problem which he sees very clearly. If, as is thè case, we can talk of a 'history of declension' (and more generally of thè history of syntagmadc groups), we also have to recognize that it has an uneven hybrid quality, including 'a multitude of isolated facts some of which will be clearly phonetic and which will join others which have a different character' (Riedlinger, Quire V [Komatsu and Wolf 1997:68]). Phonetics intrudes at every turn, inevitably, yet there is a 'residue', 'which seems to justify a grammatical history' (Riedlinger, Quire V [Komatsu and Wolf 1997:69]). This contradiction cannot be resolved: Everything that is grammatical has to be referred to a state, and there is a contradiction in saying <that> a grammatical fact has a history in time. The question ofwhat to think ofthe evolutive viewfor things which are not purely phonetic is not clear; we will not find this to be a simple matter, and phonetics will have some role to play in it (Riedlinger, Quire V [Komarsu and Wolf 1997:68f.], my italics) This pithy conclusici! might be shared by many Romanists, both ofthe old school and products of modern linguistic training. The transformation of Latin declension has long proved a rich and privileged testing ground for hypotheses about thè relation between syntactic and phonetic factors in language change. A fine example is a work by Herman on structures apparently having accusatives instead of nominatives (see also thè discussion ofthe extended accusative in Ledgeway, this volume, chapter 8, §6.2.2.1), in contexts where thè relevant NPs are not governcd by thè verb, in which he discusses their implications in respect of thè phonetic conditions governing vacillation in noun inflection. At issue are sequences of imprecations from thè Tabulae Defixtonum of Hadrumetum (short inscriptions containing curses) in which an optative mood predicate is expressed by thè present subjunctive of intransitive (or intransitively used) verbs. Such predicates are preceded by proper names in -u (second declension) and -K (third declension). The second declension forms suggest accusatives,28 but this is not uncontroversial. As Herman himself recognizes, it is unclear whether these structures are subjects or accusatives of enumeration (a type of accusative used in lists of 33 Kos;inna Sornicola oliji-i is, particularly in agricultural or medicai trcatises). His conclusici! that m die l'illudile Dejixionum thè accusative nominals are in an extra-syntactic posinoli seenis quite convincing. This position, like thè enumerative struct n i e s , .SITUI io have been thè locus of functional alternation between nominative and accusative — for thè late second century in thè intensively Romanized aix-as of Africa. Structurally, this interchangeability canno t have .ipplied lo lile rraditional functions of thè accusative or other oblique cases. I l e i inaii (1987:102) says that thè alternations -us/0 occurred with relatively lugli frequency only with nominative singular -us. Thus, with .ili due epigraphic caution, Herman states that rather than being àn orthogi.ipliical rcllection of phonetic changes in thè spoken language, this phenomcnon corresponds to a more deeply rooted and complex morphosyntaaii: conditioning. What is being suggested is that thè variability in thè occimvncc of-s really rcflected a functional perturbation in thè use of case lonns in -s, primarii/ in thè nominative. Bui whai is thè synchronic status of thè execration tablets of Hadrutnetum, within thè wider diachronic development? At that time (late second to early third century) was -S-variation an African peculiarity, or is this imprcssion a mere fluke due to thè distribution of thè texts that happen io have survived? And may we really assume that thè rarer examples of -u for -us in 'popular' inscriptions of thè imperiai epoch, in Africa, Italy and elsewhere, are also nominative—accusatives with mobile -s? Herman says ihai we can but speculate. Given that in Africa length oppositions in vowels may have been lost earlier than in thè rest of thè Empire, and that (bis happencd cspecially in unstressed (and particularly fmal) syllables, 1 lei in.in speculates, with due caution, that complete homophony between M:KVII(M) and SKRVO, I<ILIA(M) and FILLÀ may have been relatively early in Aliila. I In inaii appears to give crcdence to thè theory that thè accusative l u x . u n e extended as thè generai case form, at least in thè singular (I l i un.ni 1987:106). This allegedly began with thè interchangeability of n o i i i i i i . n i v i - and accusative, and gained momentum from thè phonetic perturbatiotu mentioned above, leading to homophony berween thè acculatlve and Othei casi- loniis. The growing frequency in typically'unmarked' m n i e x i s , M u l i .is lU-iachcd, extra-sentential structures, then further facilii.ncd ( I n - i-i|iiivali-nu- w i l l i die nominative (Herman 1987:106). I lenii.in'.s .inoliili fully displays thè complexity of marrying a phoneticphonulogical conception of change to a syntactic one, and more generally of u . mi. l i n i . - historical and descriptive-synchronic models. He seems to have io i l u - ilu-oiy ol ihe accusative as universa! case form, while giving Rolll.lllll- lilljMIISlh . I I I . I l l l M n l I. il I l l i r m Ir chronologicaJ (and phenomenologlcal?) aiul lo|iii'ul p ..... n \ p i i . . i ....... 1. 1 phonological aspecls, , l l j ' , l l i l i j r , 1 1 1. 1 1 si I I I e i n i l i . m-., i i p i i . . ....... i IH . u n i n ces of nominatives are in syntacticall) Indcpcndcni poiltlonii th< ' ' " I di. accusative is there compeiing w i l l i .imi s n l i s n i n i i n i ' , i l i . m i i m n . i i i M I m m (Herman 1987:106). The Hadrumeiinn i . i b l i - i s .in u n n l \v i le.n and rich examples of a morphosyntai i n v.u ulatlon w l n < l i j',c m i . i l l y , .1 I n d e later and perhaps less often, woulil also enu-ige elwwfaefCi Whether forms resulting from phoneti< emsinn <>l i l.issu .il t .ise M I I U tures can be considerai 'accusative' is donbilnl. Rei .ili I >'( )vidi<>'s aiul Schuchardt's criticisms of using classica! loim.s as .in ibiolute l e l e i e m e point for fully formed Romancc forms, wiihoin Interpreting thè Intermedi afe remnants of older forms according to dilli-u-ni e i i i e i i . i , ilio.si- ol die processes of change. The fact that forms in mobile -s, typical ol M-IOIH! declension nouns, appear in thè sanie texts alongsidc d i i n l declcnsion imparisyllabics in -E, whosc case-value is even les.s l e i i a i n , siiiy^esis ih.u none of these forms can any longer be dcscribed in tcnns ol 'noiiiin.itive', 'accusative', etc. At thè end of his study, Herman asks a fondamenta! queslion: Is ilic morphosyntactic alternation -us/-u wholly unrelated to tln- laii-r, phonetic, disappearance of ali instances of final -S in thè East and in niosi li.iloRomance dialects? Did theloss by final -S or iis morphosyntactic functionin nominai morphology contribute to thè loss of-s elsewhere in ilu- grammar? Herman admits he does not know, but could anybody ever really know? I le has at leastgiven usa wayofframing thè basic problem: Wliai is ilu- relation between phonetic and syntactic factors in thè disappearance ol Latili declension? We will never know if thè supporters of die Yxti-ndal accusative' theory were right or whether some other thcory is: IDI dlachronlc research dcals perforce in hypotheses rather than ccriaintk-s, ilu-n-by pl.u ing a limit on thè observer's capacity to undersiand tlu- liisiorii.il piot esses. Once again we come up against thè full force ol (he epistemologi< .il divide between synchrony, representable in terms of strucunv, .imi di.u l i i o n y , where structure can be represcnted only in an l i n i e r i . l i n .imi u-ni.nivi- w.iy. 3.3 The prepositional object The Romance prepositional object offers a liinhei i-x.nnpl<- ni du m-i-d n > distinguish thè synchronic perspective associatcd w n l i typology l u m i i l u event-oriented perspective of diachrony. It has typologlcal i m i n i i i p n i-, m numerous unrelatcd languages, leading some m establllh .u i n . il | u i u m ' . ni formai correspondence across languages (src Mossoiip, P > ' H , I I 35 Rosanna Sornicola 1992). The idea of a link between differendal subject and object marking and syntactic type has been reinforced through typological approaches of iliis kind, but this may not help us much in our search for a diachronic 'explanation'. The semantic properties of thè modern synchtonic situation have simply been projected onto thè past, both descriptively and in terms of explanation, often without asking whether thè modern semantic properties have any real relevance to thè diachronic explanation. The carly Romance situation must have been appreciably different from tOCUy. Fourteenth-century Sicilian and Neapolitan are illuminating in this u-spect (see Sornicola 1997). The object NP is by no means regularly | iliuman], [-referential]. Ifwe seek a 'regular' occurrence of thè semantic parameters found today, thè medieval texts offer a confused and chaotic pn i ure. But two sets of properties do stand out clearly: thè lexical properui-, D| die verb and thè properties of thè NP. And they seem to have not a si-iiiantic but a morphosyntactic basis. The verb, at some stage in thè liisiory of Latin, is one that took a dative complement (or thè rivai type AD i .u curative) as an alternative to thè accusative. The time-span ranges from .ni li.iic Latin to Christian Latin writers, and thè Christian writers seem to pl.iy a decisive role in consolidating and generalizing thè dative (or AD + accusative) complementation of numerous verbs. But there are also various i .i.si's o( genuine 'karstification',*" where thè dative construction is documented for some verb in archaic Latin, yet is not attested in Classical I..uin, only to reappear in Christian writers. And there is often an unini i i i i i p t e d continuity with thè relevant verb in old Tuscan texts, where it i.ikcs a; less frequently thè a construction appears in Tuscan documents without any detectable Latin precursor. In Sornicola (1997) 1 made a lexical examination of these verbs, their Latin antecedents and their OTsc. i «imiterparts. These are verbs such as aiutare 'help', ascoltare 'listen', audire 'bear', clamare 'cali', contraddire 'contradict' and confonare 'encourage, console'. Many such verbs take a not just in Italy but in other Romance languages. I1'.vcn French, traditionally considered to lie outside thè area of thè prepositional object, shows uninterrupted series of verbs which may take a before a direct object: prìer 'pray', supplier 'beseech', requérìr 'request' and aider 'help' (see Ludi 1978; 1981). These are diesarne lexical types as have shown thè dative constructions in southern Italy since thè earliest times. The object NPs in southern Italy reveal thè following cruciai properties: (1) whenever it is a personal pronoun it is always preceded by a preposition; 36 Romance linguistics and historical lingtiistics (2) (3) if it is any other type of pronoun (relative, indefinite, etc.) it is often, but not always preceded by a preposition; if its head is a full nanie, thè preposition is not always present. Despite differences between texts, prepositional object marking is more common with NPs whose heads are proper names (and thcrefore [+human], [+referential]), than with those whose heads are [+human], [-referential]. But there are also many cases in which NPs with [+human], [-referenrial] heads are marked prepositionally. This is exactly thè situation that emerges for thè Ibero-Romance area (Meyer-Lubke 1899:§50; Reichenkron 1951; Martin Zorraquino 1976; Villar 1983). Stimm (1986) shows that for Engadine Romansh, too, thè phenomenon occurs with lexical vetbs that took a dative construction (or AD + accusative) at some point in thè history of Latin, or where thè NP is a personal, relative or indefinite pronoun. There are therefore at least two different factors triggering thè structure which appears at later synchronic stages: thè constructional properties of certain verbs, and personal pronouns. Traditionally, a great deal of emphasis has been put on thè role of various functional factors, such as thè need to differentiate subject and object (see, for example, Bossong (1991), who adopts in typological perspective an intuition already formulated by Diez in thè Grammaire des langues romanes), and foregrounding of thè object. These syntactic 'explanadons' are multiply problematic. Both blithely project a synchronic structural model onto thè past. In particular, thè hypothesis of differential subject and object marking attributes to grammatical relations universa! values which are far from being demonstrated, especially given that 'grammatical relations' are themselves one of thè most controversial areas in modern syntactic theory. Even ar thè descriptive level for individuai languages, one cannot always uncquivocally assign a particular function, such as direct object or indirect object, to a given constituent (see Sornicola 1997). The government properties of thè verbal lexemes have a cruciai role here, for they may affect thè struciure of grammatical relations in ways incompatible with theoretical expccuuions. In recent years these problems have been adtiressed in some modcls (e.g., 'structural Case' and 'inherent Case' in generative grammar), which, as we shall see, have particular relevance for thè study of thè diachrony ol thè prepositional object. Lven il diese problems were solved, could we really maintain that differentiating subject and object is thè 'explanation'? Is it any more than a mere 37 Kosanna Sornicola pi ion' of a tendency present at a more or less recent synchronic stage? And as an explanation it is problematic even on thè synchronic level. Tvpologica] studìes have shown that morphological case marking is not i-.ssi-iitial lo cncode thè subject—object relationship, given thè availability of word order and semantic or contextual cues associated with thè head of i he NI'. So thè much-invoked role of ambiguity resulting frorn loss of tasi 1 marking from Latin to Romance requires some caution. It seems a reasonable hypothesis that word order may have acted as a major synchronic dlfferentiating factor of thè subject and object relationship even in thè past. In otlier words, appeal to ambiguity needs to be treated with thè same lamioii for thè past as for thè present. Care may be needed with appeal to object forcgrounding, which textual studies of various stages in thè history ol'tlic Romance languages show to be not especially frequent. What role has been played by thè other triggering process, involving thè pionominal properties of die NP? Hcre we have to consider thè multiple successive layers of morphosyntactic properties of personal pronouns, over a long period stretching back perhaps as far as Latin, unfolding through thè umiplex traiisition from Latin to Romance and reaching into thè formative period of Romance literary languages. To simplify here gready an extremely complcx issue, we may identify at least three phases. The first may be dcscribcd as late Latili, and shows a conservative tendency widi regard to die declcnsion of personal pronouns. In Romance languages pronouns have gcneraliy maintained declensional distinctions better dian nouns (Lofstedt I % 1:225). Thiis in late Latin texts thè functions we may label 'dative' were rxprcsscd synthetically in pronouns and analytically in nouns (using thè AD + .musativi: construction). The survival of Classica! Latin dative forms is confirmed by thè lact that in many Romance languages thè stressed oblique pionouns, whosc formai development is more easily identifiable than for UlUtreued pronouns, have preserved. a morphological structure which evidi-mly goes back to a historically underlying dative form, as is apparent in thè Sp.misli . u n i northern halian mi and ti (< DAT Mini, TIBI) - see also thè dilCUttion in Stnithi this volume, chapter 6. Romanian preserves a personal I I I O I H M I I I p.ir.uligm which distinguishes nomi native/accusative/dative. Generallzatlon ol dative forms as stressed oblique pronouns, attested in suini- moduli Romance areas, must have gone through periods of wavering In i wi-i-n I I M - .md iivunsc of dative forms, endemie throtighout thè Romania I » i « < m ihr sixih and thescventh century, thè period of'decadence' of Latin, I m i i In. ll\ |iirM-iiu-d laier by thè Iberian area. Of thè Cartulario de San , l"i . miplr, Irnnings (1940:1 50) observes that not onlywas thè use u! il. d . n i M - i im'.pu I I I H I S and wcll preserved in thè expected contexts in W R o m a n i e lingmsiii s u n i l i i M u i i i .il I m i III Latin, bui that thè pronominai p.ii.uli|',m sliows d.n n i>. n iln notiti is preceded by Al>. The situation describcd by |uiniii|',s scuir. n > churactcrl • i mori advanced phase of thc dcvelopmenl of thè personal pruni m ns MM n|',|'i ' a second phase which we will e.ili 'pseudo d.uivr'. * )M ill.iinur. In iwrrn competiiig morphological types ani! 11 u1 uvunsr of dativi Im n i . in il-.n wrll attested by some relic stressed personal pronounl fòrmi, DOth In U n n i .md in Italo-Romancc. Such forms must have cxilted buwcui i l n i r n i l i u n i i l n thirteenth centuries, only to disappear alinosi i-vuywliui-, Tlirsr n -In i .m particularly intcresting for an understandin;; ol liovv thè developmenl ni dupronouns may have come to bear on ihe lonnaiiun ol wli.n wi- rtOWCall die 'prepositional object'. They have unequivocally 'd.itivi-' morphology, uy mological in thè second person, analogica! in thè Inst, bui ihcy .ne usrd in a generai way for thè oblique case. These lorms aie wi-ll doi i i i n u u r d by Menéndez Fidai. In tenth-century cartularics from Spain we In-qm-nily (imi thè type 1SG miue, mibe, analogica! on Latin 2SCI imi, IIM-I! as siivssed oblique pronouns. Later this type is documentcd in thè lamously urnsu vative Mo7,arabic and Sephardic Spanish varicties. The first puson singnl.ir oblique form myb was stili in use in thè Mozarabic speech ol J uda I l a i .evi around 1100; a myby 'to me' is stili documentcd in thè i l m u v n i h centmy Spanish of DonTodros, arabbi at thè courts of Alfonso X and tlien Sancho IV; a myb occurs in an Arabie muwassaha by thè 'hi i nel inaii ol Ti idi-la' (dici! 1126). Note also thc type tene (cunteue 'with thee') documenied in 1034 in Leon, looking very mudi like thè southern Italian forms mcve, teve, seve (Menéndez Fidai 1956:340f.). These, too, show morphological mniimiity with Latin, thè first person singular form being obviously an.ilogually remodelled (see D'Ovidio 1905:50). For other Italo-Romancc cxampli-s, see thè references in Monaci (1955:639b). Although sudi lorms survivc to this day in some dialects of Salente and Basilicata (see Rohlls l l ) ( > 8 : l V)), they may, like thè Ibcro-Romance forms, be considerai rei ics, supplanted in Italo-Romance by riva! types which have existed silice die r.irlicsi l i i u a i y attestations. These are thè pairs mene, tene and mìa, liti, now.ulay.s widc spreàd in centra] and southern Italy. Whatever thcir etymology, ihry liavipresumably followed different developments: mene and lene u n n i - l u m i accusative forms with thè addition of -ne; mia and liti come- l u m i originai dative forms (on this, see D'Ovidio 1905). Wlial seuns sipnlii .un is (hai they share thè property of being long, disyllabic lorms. In tlu- ph.i-.r u r .m calling 'pseudo-dative', then, dative morphology had i ompletely I"'.i ir, nld case value and had been refunctionalized in thè liglit ni .1 pm-.ndn i i - n d i - i u \o prefer longer 39 Rosanna Sornicola The observation that in old Ibero- and Italo-Romance texts thè preposition a (like pe in Romanian) is especially frequent with object NPs liaving first and second person pronouns as their heads may be viewcd in .1 ncw light. The obligatory occurrence of prepositions before personal pronouns may be assigned to a third phase. While thcy are unlikely to have been exactly synchronous structural developments, thè expansion of ilii- prepositional strutture may well have been a concomitant of thè rise ni ilie stressed monosyllabic pronouns mi and ti, which in Ibero-Romance look piace at thè expense of thè pseudo-dative disyllabic forms. Tn ItaloRonuncG) thè process must have been more complicated and locally dlfferentiated — as wimess, first, thè notable polymorphism of early texts (i .)•,., copresence of Sicilian mi and mia. even in thè sanie text), and si-conci, thè modern differences between northern dialects where thè i v pc mi, ti is widespread, and central-southern dialects which display in Mime regards a more locally fragmented situation. The preposition a may iniiially have established itself before first and second persons stressed ohjivt pronouns as a mere expletive element, due to thè prosodie lengthr n i i i ^ o( monosyllabic forms. This expletive element may then have been propagateci, perhaps subsequently, to disyllabic pronominai forms, as is lUggcsted by thè fact that thè preposition also occurs with thè southern li.ili.m types mene and mia. However that may be, thè proliferation of thè preposition could be taken to be a different effect of thè sanie prosodie principle which was at work in thè first two phases. So in ali thrce phases ,i uniiary principle could be said to have been at work, structurally rc.ili/.ed in various different ways. The hypothetical 'explanation' (better 'comprehension') offered here for ihi 1 pronominai manifestation of thè prepositional object in early Romance icxis might be taken to imply an autonomous development of thè signifiant with respect to thè signifié. Yet during thè long, multi-layered process dcscrìbed, semantic and referential factors may also have played a role, sudi as thè person hierarchy. Both thè southern Italian texts I have cxumined and thè old Spanish texts examined by Reichenkron (1951), Martin Zorraquino (1976) and Villar (1983) clearly show that thè object pronouns involved in prepositional structures are preponderantly first and second persons, something which may not be accountable for in purely 'formai' terms. Actually, it may be that semantic factors contributed more to thè propagation than thè genesis of thè type. In Ibero-Romance and southern and centrai Italo-Romance, thè fact that numerotis verbs govern thè dative, and thè prosodie tendencies at work on personal pronouns, have gelled into a particular structural type due to thè propagai ion ol u 40 Romance linguistics and historical linguistics into contexts with nouns specified as [^animate], [+human], [+referential], a phenomenon where semantic factors undoubtedly played some role. We are now in a position to sketch out some further hypotheses on thè diachrony of this syntactic type. Indeed, it seems that two phases ought to be distinguished: thè genesis and thè propagation of structural conditions. Wliat in modern synchrony appears to us as a unitary type must have had a long and heterogeneous gestation, with multiple lines of genesis and development. In particular, while thè two conditions identified for thè initial phases (die verb governing thè dative and prosodie tendencies at work on thè personal pronouns) were presumably so widespread in thè Romance world as to crystallize into manifestations which appear over wide areas separated by lime and space, propagation itself may have followed different routes at different times. Thus in thè Prendi and northern Italian linguistic area, thè initial conditions have always remained endemie and in some sense distinct from each odier, without ever gelling into a unitary type, whilst in centrai and southern Italy thè two initial conditions carne together to form a structural type; finally, in some varieties of Spanish thè preposition was generalized into contexts whose NP contains a noun widi thè features [-animate] and/or [-referentialj. Of thè two 'triggering' conditions, thè prosodie factors acting on thè personal pronouns must have been thè strongest and most pervasive in thè activation of thè type, as witness thè fact that it occurs in other IndoEuropean languages and even beyond. Thus thè modern synchronic description has a somewhat accidental, epiphenomenological, relation with thè genesis of thè type and is only partially linked to its propagation. This is perhaps unsurprising, given what thè great generai linguists and Romance linguists of thè past used to say: that thè synchronic dimension and thè diachronic dimension involve different problems and methods. But thè diachrony of thè prepositional object has further implications for thè more generai problems cliscnssed in this chapter. What has emerged are conditions dependent on different textual and historical circumstances, at different periodi in timc. In addition, different 'principles', prosodie and semantic, have Invii hypothesized. Can these sets of conditions and principles rcally be considiTcd Yauscs'? The conditions are, precisely, no more than conditions, and thè linguist can do no more than speculatively attempi to link meni io liistorii al sci|in-nivs. As for thè principles, their validity can only be ilisi iTiu-d locilly', in ivl.uion to thè case in point. It does not seem logically valici io proji-i i a (univiTsally active) global principle on thè basis of these locai contexts. Of course, generai explanations of change will always be a 41 Romana- IÌIIJ;HÌMÌI.S a lui Insinui al I U U M I I tli Rosanna Sornicola genendizatìon of particular explanations advanced for individuai phenomena (sci- ( '.oscrhi 1973; 1981), but onc has to admit that such a procedure stili loinain.s a good deal of approximation and uncertainty. 'l'In 1 conditions attcnding thè 'origin' of thè historic process which has shapi-d ihe Romance prepositional object type show remarkable stability, and coniimi thè force and permanence of certain structures through time. ' l'he dal ive-govcrning properties of some Latin verbs are continued through siruciures with AD + NP or through thè refunctionali'/ation of thè debris of olii pronominai forms. The 'propagatici!' phase, both to larger classes di' vcrb lexemcs and to NPs lacking thè originai semantic and referential Iramres, is, of course, a different matter. Such analogical extensions or symactic rcanalyscs can be brought about only within given synchronic statcs. But thcn, perhaps diachrony without change actually does exist? •i Universa! explanations and historical explanations 4. i Laws, princìples and explanations 'l'In 1 problems discussed so far raise anew thè issue of thè meaningandcxtent of recourse to 'laws', 'principles'30 and 'explanations' in diachronic morphosyntax. Actually a good many contemporary diachronic (and synchronic) inodels stili bear traces, more or less explicitly, of notions of'naturalness' and 'scientilicncss' which had already been thè object of controversy in thè early twcntieth century.31 Positivistic conceptions of forces or principles alli-gcdly acting as causes of linguistic phenomena borii in synchrony and in diachrony nowadays occur in various, and more or less subtle, guises. To simplily greatly, fìvc groups of principles or causes are commonly invoked as explanatory lactors: (i) ini Ri-asons relating to thè speakers' 'mind', mcaning thè capacity for undentanding or an abstract mechanism undcrlying linguistic production, In addition to traditional principles of analogy, w h k h li'vcl paradigms, there are also cognitive factors, of thè type nidddli-d in generative grammar (e.g., structural syntactic properlies uiumed as abstract properties of thè mind), or principles of va r i i » is I u in lional kinds, which project characteristics oftheorgani/.niiin ni n i u - i a i K i - s (topic—fbcusstructure, information structure) i n i h.n.u I I - I Ì M K S ol siniitural relations (e.g., subject vs. object) as • p i , i n . u n i \. I' i n I.nni|', io typological micro-parameters, such as patterns ni l i n i a i i / a l i n i i ni l o n s i i i n r n l s and ihi'ir harmoni/.arion. The i ' theoretical basis ol Mieli m u s i m i i-, i; u n i h a i , a l i l i u i i | ' J i i l u \c becn objectified and hyjXMtatized as n u l a l i r a m i . il l i m i l i - , uvei arching thè classica! Iramework ol hinyjn' vs. />,/n//r. •,ynia|' i inaiu vs. paradigmatic, synchrony vs. diaclimny. (Hi) Reasons relating to thè degree ol complexlty 01 lUttunlnCM (m.irU edness) of a given feature or phenomenon, on ilu- basis ol iis uoss linguistic frequency. (iv) Reasons which, while they beat on ipeaken, aiv ( i l an eminenti}1 extrinsic, pragmatic nature (e.g., optimi/aiion ol i o m i n n n k ation). (v) External social reasons, whicii assume a mori1 or less din-i i rd.uionship between extra-linguistic characterisiics and i In- pii-scncc <>r development of linguistic phenomena. In any case it is worthy of note that recent years bave seen a conlhicnce of these principles in many studies of various theoretical and meihodological stamps. Whether any of these principles has real explanatory power, lailicr than bcing essentially descriptive, is doubtful. The risk of circularity is great. For example, are thè h a i m o m / a i i u n s of linearization in structural configurations, defincd on thè basis ol ilie synchronic states of individuai languages (and with a degree ol irregukrity which simply cannot be ignored) really a 'causai' factor, or jusi a mrir description imposed on diachrony? These are cpistemological issues duiacteristic of so-called 'genetic explanations' (see Amsterdamski 1981:372; also Popper 1957). And thè validity of generai laws for languages has olu-n been criticized with regard to synchronic states (see Matthews 1982). On top of these difficulties are others more specific to diachronic syntax: thè transmission of syntactic traditions is unlike thè transmission ol iradiiions of other levels of analysis, in that extra-linguistic, historical factors, wcigh more heavily on syntax (see Sornicola 1995). 4.2 Diachronic explanations in Romance linguistici Before thè North American developments discussed in §2, there had already been wide-ranging and penetrating reflection on thè problem of diachronic explanation within Romance linguistics. For Coseriu (1973; 1981:80), 'explanation certainly goes beyond mere description, to motivate or justify changes, and tind thè reasons for them'. But Coseriu stresses that that ihcse 'reasons' are not 'causes', in thè sense of necessity, but conditions, circumsrances or determining factors within which speakers have frecdom to makc linguistic choices. Such factors do not trigger change, but condition it and 43 Rosanna Sornicola may help speed up or slow down what is improperly called thè 'evolution of languages'. So thè 'generai problem of changes is [...] their condidoiiing'. The perspective of conditioning leads us to rule out thè possibility that ivasons (in thè sense given above) could lie directly in thè strutture of society. Coseriu holds that 'thè historical problem of change is not a inattcr of establishing how a particular linguistic mode began (or how it COldd be initiated), but of establishing how it took shapc or how it was able io take shape as a tradition, that is, how and under what cultural and liinctional conditions it entered or could have entered in a System of already ii.idiiional linguistic modes'.33 The functional factors within thè System (lor example, its criticai or weak points) and thè cultural factors constitute multiple conditions fot change and resistance to change. Under given hlttorical conditions one group of factors may prevaii over another. But ilu- lorce of tradition has a major role: 'a vigorous cultural norm may iiu-.iii indefinite stability for an "unbalanced" System'.' And thè distinction between thè generai problem of change and thè study of an individuai < li.mgc is an important one. In both cases we are dealing with various kind ol historical explanation, distinct from thè problem of thè mutability of luìguages - a problem which belongs purely to thè theoretical dimension .ind which can be solved only if we recognize that mutability is not a bolt-on II-.IHII-L- of linguistic systems, which needs to be explained, but a necessary, mninsic, property, ultimately grounded in thè historicity of thè linguistic i udiiions on which thè systems are built.36 l'erhaps thè most interesting point, especially in thè light of diachronic inodels of recent decades, concerns thè differentness of universal explanaiions and historical explanations. Coseriu observes that 'while we may know in generai thè causes of wars, of course we can only know thè causes of iln- l'eloponnesian war by study i ng them, because universal knowledge and gcneric knowledge are no substitute for specific historical documentation'. Such documentation is an additional problem for thè history of languages, being more difEcult and elusive than in other disciplines, and lack of sources being a frequent problem to be faced. Our awareness of thè tcchnical difficulties of doing history, which are especially acute in linguistics, means that we have to take a highly hypothetical view of explanations, especially where thè origins of a linguistic change are concerned, a viewpoint which in recent years has been reformulated as a diachronic 'scenario' (see Dressler 1997). The formation of thè Romance future is a case-study/<zr excettence of thè contrast between universal explanations and historical explanations. Earlier criticai discussions have lost none of their theoretical interest but, over and 44 Romance linguistics and historical linguistics above thè phenomenon at issue, they bear directly on thè logicai structure of two types of argumentation perennially recurrent across different theoretical models. The morphological explanation holds that thè lack of a regular paradigm of synthetic forms in thè Classical Latin future led to recourse to periphrastic forms, especially hearing in mind that sound change in so-called vulgar Latin would have made thè morphological patterns dysfunctional. The semantic/stylistic approach links thè advance of thè periphrasric future to an expressive need which conveyed modal affective values rarher than temporal ones." Both explanations have been reckoned inadequate and vulnerable.39 Coseriu (1973; 1981:115) holds that: Therc are three facts to be explained: a) thè generai instability of thè forms of thè future (not of thè category of future); b) thè periodic re-formation of thè future by forms with originally modal or prospettive value which ultimately become 'temporal'; e) thè re-formation of thè Latin future at a particular point in history. The first two are not particular to one language or a particular point in history, and thereby require a universal explanation. Nothing is explained by stating that thè forms of rhe future are re-formed because they are 'grammaticalized', because this is at best a mere attestation, which cannot account for thè direction in which thè re-formation of thè future takes piace. The dual nature of thè Riture is a universal, intrinsic structural feature of this category, perennially wavering between thè temporal and thè aspectual (modal) poles: 'temporal forms are replaced by modal ones which in turn become temporal'.41 Coseriu stresses that 'in any case a universal explanation is not of itself a historical one', for 'to explain why thè Latin future was replaced by modal forms at one particular time, it is inadequate to assert that it is something which "usually happens", pointing to thè universal reason for thè phenomenon. You have to explain why this universal (and permanent) reason carne to operate precisely in thè period of so-called vulgar Latin: in other words, thè universal expressive need must be justified as a historical need.'42 For Coseriu, thè determinant circumstance lies in attitudes and expressive necds brought to thè fore by a phenomenon of great social cultural import - Christianity; Christianity marked a profound historical rupture with linguistic as well as other implications. Note that this explanation is very different from those commonly used in sociolinguistics, which appeal to diffcrences of social level. Such differences do not express causes but rathcr thè point or direction of diffusion of a phenomenon.44 What is being emphasized bere is thè role played by Christianity in bringing to thè surface latent structural possibilities which predated any contemporary historical circumstances. 45 Rosanna Sornicola Tliis approach has becn criticized for its one-sidedness and for its downplaying of thè question of when thè periphrastk form emerged as such (scc Fleischman 1982:50). The diesis that 'thè primaiy causes of morphosyntactic changc must he sought in morphosyntactic factors' (Valesio 1969:192f; Fleischman 1982:50) may have its attractions, but appeal to 'multiple causation', postulating thè joint influence of semantic, syntactic and morphological factors, as well as cultural factors (Fleischman 1982:50), has its problems too. On thè one hand, it maintains, more or less implicitly, a naturalistic idea of thè cause of diachronic processes; on thè other, it places thè onus of an elusive justification on a summary list of causes. It may be that thè two approaches are just radically alicn to cach othcr. Attempts to establish thè diachrony of periphrastic future (or past participle) constructions, based on thè correlation between basic parameters, degree of 'boundness' and modali aspcctual/temporal values are bold, but thè resulting principles stili require supplementary historical investigation (see Adams 1991). More generally, granimaticalization models describe, in thè best cases, successive cycles of structures, but it is doubtful whether these are truly cxplanatory, unless one follows thè notoriously problematic principle ofpost hoc, propter hoc. The problem is that if one attempts to organize a series of changcs iato groups forming a genetically ordered System, thè historical data are oftcn incapable of refuting thè model. Worse, as Amsterdamski observes, thè causai nature of genetic explanations is doubtful 'because they do not generally formulate sufficient conditions for thè emergence of thè firmi state which requires explanation, or even for thè intermediate states which rnake up thè genetic sequence. So they fai! to explain why thè systern evolved in thè observed sequence, but only explain why it could have don e so' (Amsterdamski 1981:372). The historicizing explanations of Coseriu are not, by defìnition, incontrovertible, but they have been put forward with a eleni awareness that causai explanations in diachrony could lead us on an infinite wild goose chase. S Conclusions It is alinosi a ccntury since Saussure and thè Prague School formulated modell ol diachrony. Their perspectives stili reflected different stages in thè \vi.lci deli.iu- about history. The terms of thè debate within thè humanities ha ve sim e changtd greatly and linguistics may have remained rather cut off l i i i n i o l i i c i disciplines and stranded on issues which are now outdated. .Some . o i i i c n i p m a i y diachronic linguists' (especially syntacticians') 46 l i n j ' , u i . s l i i s a n d I n s i n u i .il lni|'.i critiques of historical modcls a.s mere I C | > I . M n i . i i i ni il|i| I facts harks back to thè olii poiidvildc conccptlon ol I ir.mi \, I !• m ' I J M > •. seem curiously ignotant of die widc i.iiij',iii|', d e l u i c . i l i n i u i l i ' m i i m i> , i l l \c and non-causa artcmpting to conceive a theoty ol history is .1 contnuilcl i I Central to this perspectivc is thè roteo) di I leu-i 11 epistemologlcaJ paradigmi, founded on concepts of'comprchension' and ' i m c i | > i e u i i i > i i ' . mi i l i c .ihiliiy to make sense ofsituations (sec Momigliano IT/'ì; l < W .' ' ' ' . . Peuiton 1991). Others have argued that thè leap froin die do. u m c n l s ol die pasl, which are never more than historical 'rubblc', io making scuse ni i l u - i n . . .in only be doneby being somethinglike a propilei, a medium 01 .m i n u i p i e i e r of dreams (Benjamin 1997). Everyone recogni/es i l i . i l i l i . u m c l i . i h i l i i y ol this type of knowledge and thè scope for multiple and ivl.iiivisiie inlerpretations just come with thè terrain. Yet such lliings lorm die It.isis ol new models of genetic or historical explanation, which poi m tip how 'even il t h è explanans of a genetic explanation seems to be a mete "bistorto! n a t r a i i o n " without mentioning any law linking thè successive stages ol i he evoMng System, it yet presents a theoretical structure' (Amstcrdamski 1981:372). Freed of its causai value, thè notion of'theory' is here undcrsiood as a set ol generai but clearly delimited ptinciples, justifying some ol ihe major evenis in an evolutionary process. This perspective does not claim to bc a theory ol hisiory, bui simply acknowledges thè inevitably theoretical nature (that is, thè laci i l i a t it is relative to a System of hypotheses) of any historical expl.m.ilion. It i r a u scends not only thè positivistic conception o( diachrony as ihe domali) ol thè événement and thè accidcntal, but also thè functionalistic view ol diachrony as a dimension in which there are ai work principia ol c o m a i enation of phenomena belonging to successive dlachionll M.ij',es, gOVCrned by teleological laws. The key issuc is thè representation o I i he ILIS.S.IJ^- I rom one state ot a system to another. This idea ol 'passale' CCntrtlly detUied Saussure's and early structuralism's concepì ol di.u luony. li w.is pieiiselv in this respcct that Saussure felt thè need to d i l l e i c n i i . n e i l i e ide.i ni i l i . n liiony from that of history, andtoalesserextem from that ofevolution, DOth being considered not entirely suitable for use in linguistici (Ricdllngcr, ( , ) n i i e II [Komatsu and Wolf 1997:34]:81). On (he oliici l u n d , M i m e l i i M o i i . m s have criticized thè appeal to both lomcpls i > l s v m l i \i dllchfOn) because of their limited character, w h i i h miv.i •. i l u - e ' . M mi.il m ni historical research (see Braudcl 1967- (>tv, !')(.')) In die. i , m i l , i . ,|.i\c linguistics an as different domains with dillerem presupposltions unti m. i l m . l ' . Hm I » . i l i 47 Rosanna Sornicola contain unresolved contradictions (these have been discussed for thè history of thè language by Varvaro 1972-3). Saussure realized filli well how difficult describing and justifying that 'passage' could be. He clearly perceived (without developing thè potnt) thè special diachronic status of syntax, where thè passage from one diachronic Mate lo thè next is more strongly subject to thè interaction of internai and external forces. He fought shy of developing theories on thè subject and thè IH-.M he could manage was to appeal to thè concept ofévénement— which carne i l i > w n to admittingdiat there was an insuperable limit to explanation. Equally notCWDtthy is thè fact that he was convinced of thè need to avoid imposing .1 priori categories or units indiscriminately valid both for synchrony and diac hrony (Riedlinger, Quire II [Komatsu and Wolf 1997:34f.]), a fact which inay bc considered another effect of thè difficulty of modelling diachrony. l'he l'rague School were prepared to envision explanations of thè passage, but i licy were trapped inside a realistic, immanentist, ahistorical conception of thè principia or laws regulating it, like many of thè theorists who subsequently ilicw inspiration from thè Prague School. Such mindsets, nowaciays prevalent in many areas of linguistics, are very differem from what emerges from theoretical thinking on contemporary historical research. Here, recognition o! ilic intrinsically theoretical character of any historicaJ narration or explan.iiion has long been divorced from naively realist conceptions, and thè disnission of die principles and models of analysis has reached a level of i liliali awareness which verges on ironie, disillusioned, detachment. This is a standard to which modern diachronic Romance linguistics, CUlght as it is between thè cognitive paradigms of history and diachrony, t .in come dose to achieving. In their different ways, D'Ovidio, Scluichardt, Coseriu and Herman maintained thè need to conceptualize dir principles and 'causes' of thè dynamics of change in a fashion opposed io aiiy kind of metaphysical approach. In searching for thè 'passage', they maintained a kind of sober equilibrium with regard to thè possibilities and limits of diachronic research, which was perhaps due precisely to thè fact that they were Romanists, and so researchers in a discipline distinguished by thè most imposing 'mass of photographs of thè past' of any linguistic domain, too imbued with thè historical mentality to be oblivious to thè razor's edge between history and diachrony. Even a scholar who, like Malkiel, had explored thè more strictly diachronic end of this polarization had no doubts about thè importance of thè 'hard toil of historical preparation'. At thè dose of thè nineteenth century, Schuchardt held that a Romanist should be a generai linguist before addressing problems of historical 48 Romance linguistics and historical linguistics linguistics, an idea that was very modern at that rime and long remained so. In thè rwentieth, in different ways, Coseriu and Malkiel attempted thè difHcult task of reconciling generai linguistics and historical linguistics. But their work shows thè importance of being a Romanist before being a generai linguist. A Romance diachronic morphosyntax, just as much as a Romance diachronic linguistics, may be different from orher diachronic syntaxes and other types of diachronic linguistics noi so much because Romanists have available a mass of photographs of thè past which lets them get closer to reality, but because they know that thè photographs of thè past and those of thè present may let them dream a less fragmented, and rather richer, dream. 49

RELATED PAPERS

RELATED TOPICS

Log In

Romance Linguistics and Historical Linguistics: Reflections on Synchrony and Diachrony

Romance Linguistics and Historical Linguistics: Reflections on Synchrony and Diachrony

Romance Linguistics and Historical Linguistics: Reflections on Synchrony and Diachrony

Romance Linguistics and Historical Linguistics: Reflections on Synchrony and Diachrony

Romance Linguistics and Historical Linguistics: Reflections on Synchrony and Diachrony

Related Papers

RELATED PAPERS

RELATED TOPICS