Gabriel, C. 2003. Relational Elements in French: A Minimalist Approach to Grammaticalization....

31
Linguistische Berichte 193/2003 © Helmut Buske Verlag, Hamburg Beiträge aus Forschung und Anwendung Sprachwandel Relational Elements in French: A Minimalist Approach to Grammaticalization * Christoph Gabriel Abstract Data taken from the field of French prepositional elements clearly show that essential work in Grammaticalization Theory needs support from a framework which accounts in a fruitful way for instances of syntactic change. In this paper I show that prepositional forms which emerge from complex structures containing verb forms and relational nouns on the one hand and case marking elements recruited from the lexical category P on the other can plausibly be dealt with by adopting a Minimalist perspective in the spirit of Chomsky (1995). In this context I argue that some uses of certain French P elements should be interpreted as overt realizations of inherent case, comparable to the agglutinative case affixes of Turkic languages. Furthermore I demonstrate that the emergence of new prepositional forms follows the requirements of deri- vational economy. Finally it is pointed out that a cube shaped Grammaticalization model accounts in a more adaquate way for the facts observed than the scalar representation, tradi- tionally assumed by Grammaticalization theorists. 1 Introduction It is a well-known fact that different natural languages make use of different linguistic means to mark the syntactic functions of noun phrases. Among the various possibilities we can identify intonation and word order regularities on the one hand, the latter combined with agreement phenomena on the verb in most cases, and inflectional or agglutinative case affixes and prepositions on the other. The morphologically overt elements appearing on or within DPs, namely adpositions and case affixes, can plausibly be subsumed under the com- prehensive term ‘relational elements’, coined by Pottier (1962; Systématique des * Parts of this paper have been presented and discussed at the German Frankoromanistentag (Dresden, Sept 2000) as well as at the Universities of Hamburg and Konstanz. I would like to thank the participants in the discussions and especially an anonymous LB reviewer for helpful comments.

Transcript of Gabriel, C. 2003. Relational Elements in French: A Minimalist Approach to Grammaticalization....

Linguistische Berichte 193/2003 © Helmut Buske Verlag, Hamburg

Beiträge aus Forschung und Anwendung

Sprachwandel

Relational Elements in French: A MinimalistApproach to Grammaticalization*

Christoph Gabriel

Abstract

Data taken from the field of French prepositional elements clearly show that essential work inGrammaticalization Theory needs support from a framework which accounts in a fruitful wayfor instances of syntactic change. In this paper I show that prepositional forms which emergefrom complex structures containing verb forms and relational nouns on the one hand and casemarking elements recruited from the lexical category P on the other can plausibly be dealt withby adopting a Minimalist perspective in the spirit of Chomsky (1995). In this context I arguethat some uses of certain French P elements should be interpreted as overt realizations ofinherent case, comparable to the agglutinative case affixes of Turkic languages. Furthermore Idemonstrate that the emergence of new prepositional forms follows the requirements of deri-vational economy. Finally it is pointed out that a cube shaped Grammaticalization modelaccounts in a more adaquate way for the facts observed than the scalar representation, tradi-tionally assumed by Grammaticalization theorists.

1 Introduction

It is a well-known fact that different natural languages make use of differentlinguistic means to mark the syntactic functions of noun phrases. Among thevarious possibilities we can identify intonation and word order regularities onthe one hand, the latter combined with agreement phenomena on the verb inmost cases, and inflectional or agglutinative case affixes and prepositions on theother. The morphologically overt elements appearing on or within DPs, namelyadpositions and case affixes, can plausibly be subsumed under the com-prehensive term ‘relational elements’, coined by Pottier (1962; Systématique des

* Parts of this paper have been presented and discussed at the German Frankoromanistentag(Dresden, Sept 2000) as well as at the Universities of Hamburg and Konstanz. I would like to thankthe participants in the discussions and especially an anonymous LB reviewer for helpful comments.

4 Christoph Gabriel

éléments de relation).1 No language, of course, uses all of the possible means tothe same extent. A language like Modern French, which has given up case mor-phology on the noun and has only preserved some morphologically case marked(suppletive) forms in the domain of personal pronouns, uses constraints in wordorder and subject agreement on the finite verb as well as a vast array of preposi-tional elements, the latter showing considerable differences in their morphosyn-tactic properties.

My paper is divided into five parts. While section two gives an overview ofthe data concerned, section three briefly summarizes some traditional ap-proaches and points out that several problems remain unsolved in this field. Insection four I outline the framework adopted for my analysis. Section five fo-cuses on the transitions between a postulated functional category K (preposi-tional case marker) and the lexical class P and then concentrates on the transi-tions between P and the adjacent lexical categories V and N. Finally, section sixoffers some concluding remarks.

2 The Data: Relational Elements in French, Spanish, and Turkish

At first sight, three groups of relational elements can be distinguished in ModernFrench. Firstly, there is a restricted group of short prepositions like à, de or par.These items obligatorily appear in particular syntactic contexts, marking theindirect object in double-object constructions (1.a), the agentive, thematic orpossessive genitive (1.b) or the agentive DP in passive constructions (1.c).

(1) a. elle offre un livre à son frère (indirect object)‘she gives a book to her brother’

b. c’est le portrait d’Aristote de Rembrandt (thematic/agentive genitive)2

‘this is the portrait of Aristotle by Rembrandt’ce sont les baskets de Pierre (possessive genitive)‘these are Peter’s sneakers’

c. il a été invité par sa sœur (agentive DP)‘he has been invited by his sister’

Items of this type have often been qualified as ‘grammatical’ or ‘functional’prepositions or as semantically ‘empty’ or ‘colorless’ elements – in oppositionto a set of prepositions endowed with some semantic content (see section 3 fordetails). Secondly, French disposes of a (more or less closed) class of morpho-logically simple prepositions, which are selected by the speakers depending on

1 Note, however, that the concept of a relational character of both prepositions and case affixeshas been criticized in recent publications, see Tremblay (1999: 174f).

2 These are the most natural interpretations of the genitive syntagms in the context given above;likewise, the two constituents could be interpreted as agent (d’Aristote) and possessor ( Rembrandt),respectively.

Relational Elements in French 5

the requirements of a given speech act. Phrases headed by prepositions of thistype can appear in A-positions, strictly speaking as complements as in (2.a), orin A’-positions, for instance as local or temporal adjuncts, see (2.b & c).

(2) a. il mit ses baskets dans/sur/sous ... l’armoire (local complement)‘he put his sneakers into/on/under ... the wardrobe’

b. il m’attendra à/devant/derrière ... la gare (local adjunct)‘he will wait for me at / in front of / behind ... the station’

c. il me téléphona à/avant/après/vers ... dix heures (temporal adjunct)‘he called me up at/before/after/about ... ten o’clock’

Finally, a large number of periphrastic expressions, in most cases consisting ofsimple prepositions combined in a more or less rigid way with nominal or verbalelements, can be identified as a third group in this field.

(3) à côté de DP ‘beside’, à la place de DP ‘in place of / instead of’, près deDP ‘near’, au lieu de DP ‘in place of / instead of’, aux environs de DP‘around’, sous le couvert de DP ‘under the cloak of’, étant donné DP‘given’, abstraction faite de DP ‘apart from’, en ayant égard à DP ‘consi-dering’, à l’intérieur de DP ‘inside’ etc.3

Members of this group are commutable into morphologically simple prepositi-ons in most contexts (cf. il m’attendra près de la gare → ... à la gare), and theydisplay considerable differences both in their morphological complexity and intheir semantic transparency and opaqueness. It is clear that an attempt to assigna given item in a clear-cut way to groups (2) or (3) might cause serious difficul-ties. Is it convincing, for instance, to class the single termed variant of près (de)DP, as in près le Saint Siège, as a member of the second group, while the formcontaining the prepositional element de is interpreted as belonging to group (3),as proposed by Grevisse (1993: § 988)? However, a problem arises as far asitems such as excepté DP ‘except (for)’ or durant DP ‘during’ are concerned:Are they simple prepositions because of the fact that they consist of one wordonly or should they be assigned to group (3) on the basis of their close relationto the underlying verb form? Furthermore, the question of a possible limitationof the third group is difficult to answer concerning PP constructions such as à lafrontière du Kosovo ‘at the border to Kosovo’ (Le Monde) or à la fenêtre de machambre ‘at the window of my room’. Regarding surface properties, they arecompletely identical to the noun-based items quoted in (3), but they differ fromthe latter insofar as they are not stored in the lexicon as idiomatic expressions.

3 It is clear that a compilation of prepositional periphrases cannot be exhaustive. My analysis isbased on the list of prepositional periphrases established by Lang (1991: 46-52). This list contains242 items and has been completed by the relevant items taken from a corpus of press texts down-loaded from the homepage of the French daily newspaper Le Monde (http://www.lemonde.fr) duringthe months May to October, 1999.

6 Christoph Gabriel

The situation in other (Western) Romance languages is more or less thesame. In Spanish, for example, the preposition a, derived from Latin a(d) like itsFrench equivalent, marks the indirect object in double-object constructions (cf.pro regaló un libro a su hermano ‘s/he gave a book to his/her brother’), and, inaddition, the direct object of certain transitive verbs under particular circum-stances (cf. pro golpeó a su hermano ‘s/he hit his/her brother’ or pro vio (a) suhermano ‘s/he saw his/her brother’).4 Furthermore, Spanish displays equivalentsto the simple and complex French prepositions mentioned above (e.g. pro pusosus tenis en el armario / debajo del armario ‘s/he put his/her sneakers into/under the wardrobe’). It goes without saying that in Spanish, too, the classifica-tion of a given item along the lines of the procedure depicted above for Frenchraises considerable difficulties (see Kailuweit 2001: 39).

In languages displaying a rich (agglutinative) case morphology like Turkishor Azerbaijanian,5 the situation differs from the one depicted above in severalrespects, although a tripartite division of relational elements can be stated forTurkic as well. In many of the contexts where French or Spanish employ prepo-sitional elements, Turkish (and Azerbaijanian as well) use case suffixes. Thisholds for constructions with verbs assigning the RECIPIENT role to one of theircomplements, such as

(4) a. pro pro brother/sister-POSS3SG-DAT one book gift make-PAST-3SG

‘s/he gave a book to his/her brother/sister’

and for local complements and (local and temporal) adjuncts as well:

4 An accusative object [+hum] is obligatorily marked by a if the denoted entity is understood asaffected by the action described. The particular conditions are stated in detail by Torrego (1998:chap. 2). Note that there are two more idioms from the Romance family where the dative prepositiona introduces an accusative object [+hum], namely Sardinian (appo bidu a Juanne → l’appo bidu ‘Isaw Juanne/him’, Mensching 1992: 54, 61) and Corsican (aghju vistu à Pàulu → l’aghju vistu ‘Isaw Paùlu/him’, Agostini 1984: 43, 234f). Torrego (1998: 15, 25) points out that Hindi displays asignificant parallel to Spanish in using the dative postposition -ko as an overt accusative marker, ifthe object in question is characterized by the feature [+specific]. Note in this context that in Turkiclanguages as well the overt marking of accusative objects is dependent on their particular interpreta-tion: In Turkish and Azerbaijanian, for example, DPs marked with the accusative suffix - must beinterpreted as [+def]. Consider the following examples:

Turkish mektup yaz-ıyor-du-m vs. mektub-u yaz-ıyor-du-mletter write-PROG-PAST-1SG letter-ACC write-PROG-PAST-1SG

‘I was writing a letter/letters’ [–def] ‘I was writing the letter’ [+def]Azerbaijanian çör k al-ır-am vs. çör y-i al-ır-am

bread buy-PRS-1SG bread-ACC buy-PRS-1SG

‘I buy bread’ [–def] ‘I buy the bread’ [+def]5 It is commonly assumed that Turkic languages exhibit a set of six cases, marked in Turkish

and Azerbaijanian by the following suffixes: NOM - , GEN -(n) n, DAT -(y/n)E, ACC -(y/n) , LOC

-(n)dE, ABL -(n)dEn.

Relational Elements in French 7

(4) b. pro dolab-a koy-du- (local complement)prosportshoe-POSS3SG-ACC wardrobe-DAT put-PAST-3SG

‘s/he put his/her sneakers into the wardrobe’c. pro istasyon-da ben-i bekle-yecek- (local adjunct)

pro station-LOK 1SG-ACC wait-FUT-3SG

‘s/he will wait for me at the (railway) station’d. pro (saat) on-da ban-a telefon et-ti- (temporal adjunct)

pro (hour) ten-LOK 1SG-DAT telephone make-PAST-3SG

‘s/he called me up at ten (o’clock)’

The construction with the locative case affix -dE, with its allomorphe -da in(4.c), is prototypically interpreted as the less concrete localization and thus asthe equivalent to French à la gare, leaving open whether the speaker will beexpected on the platform (maybe the most natural interpretation) or, say, in frontof or behind the building. While speakers of French can replace à by otherprepositional elements in order to express a more concrete localization (devantla gare, derrière la gare etc., see (2.b) above), speakers of Turkish use a par-ticular type of complex DP, where a relational noun like ön ‘(space in) front’ orarka ‘back, (space) behind’ is connected with the entity to be situated (here:istasyon) in a special genitive-possessive-construction, usually referred to as

by Turkish grammarians:

(5) pro istasyon-un ön-ü-nde/ ben-i bekle-yecek-pro station-GEN front-POS S3SG-LOK/bac k-POS S3SG-LOK 1SG-ACC wa it-FUT -3SG

‘s/he will wait for me in front of/behind the (railway) station’

Note that the genitive suffix of the term to be situated is sometimes omitted(Johanson & Csató 1998: 222) and may even be completely abandoned once theconstruction at issue has become more fixed, entering the lexicon as an idio-matic structure and finally becoming a single element, no longer analyzable forthe purposes of (synchronic) morphosyntax. While in constructions like (5) thelocative case suffix can be replaced by other case suffixes (e.g.

station-GEN back-POSS3SG-ABL ‘from behind the station’), some structures have become fossilized, e.g. (three year interior-

POSS3SG-LOC ‘within three years’), where the relational noun ‘interior’ isamalgamated with both the possessive and the locative suffix, forming one sin-gle element for the purposes of syntax and behaving rather like a true postposi-tion P than a complex nominal structure. The form ‘within’, though per-fectly analyzable for speakers of Turkish, can thus be interpreted as a memberof the lexical category P – the third group of relational elements to be identifiedin this context –, on a par with other temporal postpositions like ‘before’ or

‘after’. Another instance of this kind of integration into the class of post-positions is , the marker of the agent DP in passive constructions(literally: side-POSS3SG-ABL ‘from his/her side’). Note that seman-

8 Christoph Gabriel

tic change, in these cases a transition from a concrete local reading to a lessconcrete temporal or abstract one, is a concomitant of this kind of developmentof new adpositional elements.

Another instance of transitions between the classes of relational elementspostulated above is to be stated in the co-occurrence of the form ‘with’ as apostposition (6.a) and as an agglutinative form -(y)lE, obedient to the require-ments of vowel harmony in the same way as other case affixes (6.b).6

(6) a. uçak ile plane with father-POSS1SG with‘with the plane, by air’ ‘with my father’

b. uçak-la baba-m-laplane-INSTR father-POSS1SG-COM

‘with the plane, by air’ ‘with my father’

Kornfilt points out that enclitic -(y)lE “remains outside the domain of the wordwith respect to the assignment of word accent” (1997: 214), unlike other agglu-tinated case affixes which receive the regular word-final stress. The most naturalconsequence of these facts should be to interpret the Turkish (and Azerbaija-nian) comitative/instrumental marker as an element of transitory status betweenadposition and case affix, occurring both as a free and a bound form.

Before entering into a short recapitulation of the approaches made withintraditional frameworks in section 2, the data mentioned so far are summarized intable (7), a representation clearly showing the problematic and ambiguous statusof several elements. The double question marks (??) signalize the difficultiesconcerning an unequivocal categorization of a given element as a member ofone of the three classes assumed so far.

6 Most scholars (e.g. Blake 1994, Schroeder 1999) interpret the enclitic use of ile as an inherentcase comitative/instrumental, belonging to a group of rather peripheral cases, commonly referred toas by Turkish grammarians ( 1974: 326). Note that the same holds forAzerbaijanian as well (Schönig 1998: 252).

Relational Elements in French 9

(7) Western Romance Southwestern Turkicfunctional/grammatical prepositions

French: à, de, par, etc.Spanish: a, de, por, etc.

case affixes

Turkish/Azerbaijanian: GEN -(n) ,DAT -(y/n)E, ACC -(y/n) , LOC -(n)dE,ABL -(n)dEn, COM/INSTR -(y)lE ??‘true’ prepositions

French: dans, sur, avant, avec, pour, devant,derrière, excepté ??, durant ??, etc.Spanish: sobre ‘on, about’, en ‘in’, con‘with’, para ‘for’, entre ‘between’, etc.

postpositions

Turkish: sonra, önce, tarafından ??,(üç yıl) içinde ?? ile ??, etc.Azerbaijanian: sonra ‘after’, etc.

prepositional periphrases

French: à côté de, à la place de, au lieu de,près de, sous le couvert de, en ayant égard à,etc.Spanish: (por) encima de ‘above’, en lugarde ‘instead of’, (con) respecto a ‘concern-ing’, etc.

constructions

Turkish: (ev)in içinde ‘inside (thehouse)’, (istasyon)un arkasından‘from behind (the station)’, etc.Azerbaijanian: ( ) ‘into(the city)’, etc.

3 Traditional Approaches

Several attempts have been made within traditional approaches to systematizethe area roughly depicted in the previous section. Scholars adopting a structura-list perspective have claimed that prepositions are grammatical morphemes, i.e.functional elements endowed with one fixed (functional-grammatical) meaningcovering all possible uses of the element in question. Given the assumption thatgrammatical elements form closed classes, it logically follows that fixed lists ofprepositions, containing as few elements as possible, ought to be established anda clear dividing line between ‘true’ prepositions and prepositional periphrasesought to be drawn. From this point of view no difference is made between a casemarking preposition like à in the double-object construction and the lexical useof the same item in contexts like (2.b & c). Rather, all prepositions are func-tional. Such a unified account for all uses of a given P element results in a se-mantic characterization of the entity concerned by means of assigning it a se-mantic feature structure (‘Systembedeutung’), that is supposed to cover everypossible use of the element concerned and which remains very vague as an in-evitable consequence.7 Following this strategy, Brøndal (1950) and, more re-cently, Lang (1991) have elaborated lists of 19 and 24 items, respectively. In

7 According to Lang (1991), the preposition à, for example, is sufficiently characterized by onesingle semantic feature, namely A à B [+contact], i.e. ‘A is in contact with B’ (1991: 478). For thepreposition de, he claims that all uses of this form are covered by assuming the semantic feature[+distance], i.e. ‘A is distant from B’ (see Lang 1991: 478).

10 Christoph Gabriel

both cases the criteria applied for exclusion and inclusion of certain items re-main highly questionable.8

As opposed to Brøndal and Lang the grammarian Grevisse simply uses anorthographic criterion and classifies all one-word items as simple prepositions.He thus establishes a list of 45 items, including for instance durant, excepté andprès, but excluding près de (1993: § 988).

Within another tradition, represented by scholars like Wartburg & Zumthor(1958) and Spang-Hanssen (1963), prepositions of type (1) have been describedas semantically ‘empty’ or ‘colorless’. This point of view, however, involves acharacterization of all possible uses of an item like French à as a prépositionincolore and therefore fails in accounting for the data as soon as so-called ‘split’or ‘divergence’ phenomena (Hopper & Traugott 1993: 116f) are at play, i.e. theexistence of one overt form having various syntactic functions, here: the use of apreposition like French à both as a grammatical element as in (1.a) and as apreposition with semantic content (2.b & c).

A third possible view is adopted by Weinrich (1982) who proposes not todistinguish between complex and simple prepositions and to subsume evenfreely generated PPs like à l’entrée de DP ‘at the entrance of’ under the term‘preposition’ (Weinrich 1982: 571). While the other positions reviewed so farmake use of criteria leading to questionable subdivisions of relational elements(Brøndal, Lang, Spang-Hanssen, etc.), this approach simply ignores morpho-syntactic differences within the area of interest.

4 Grammaticalization and Generative Syntactic Theory

This section outlines the framework adopted in this paper. I assume generalfamiliarity with the GB framework and confine myself to summing up the basicassumptions of Grammaticalization Theory and Chomsky’s Minimalist Programas far as they are crucial for the analysis proposed in the following section.

4.1 Grammaticalization

As opposed to the approaches mentioned in section 3, Grammaticalization The-ory completely abstains from unnatural subdivisions and places emphasis on

8 Just a few examples should be mentioned in this context: Brøndal, who dedicates a whole sec-tion of his book to the Élimination des fausses prépositions (1950: 13f), excludes morphologicallycomplex forms like French lors de or German um ... willen ‘for the sake of’ as well as all itemsoriginating from other lexical categories like French excepté (V>P) or German kraft ‘by virtue of’(N>P). Consequently pursuing this strategy, he even excludes pendant ‘during’, although speakersof Modern French hardly see any connection with the verb pendre ‘the hang’. The criteria adoptedby Lang are no less problematic. He excludes, for instance, près (de) , arguing that this item is inreality an adverb, but he maintains avant, which can be used as adverb as well (1991: 42f). The(generative) assumption of intransitive prepositions covers these facts in a more adequate way.

Relational Elements in French 11

gradual transitions within a continuum between the two poles ‘grammar’ and‘lexicon’ (see Lehmann 1982 and Hopper & Traugott 1993, among many oth-ers). This dichotomy is problematic insofar as functional categories, which areoften derived from lexical ones, belong to the lexicon in the same way as theirlexical counterparts do. What we understand by ‘grammar’, on the under hand,is the totality of rules according to which lexical and functional categories thatare drawn from the lexicon are combined to larger syntactic units. For our pur-poses, the terms ‘lexicon’ and ‘grammar’ are subsequently replaced by the di-chotomy ‘lexical’ vs. ‘functional categories’. Grammaticalization can thus bedefined as the diachronic process in the course of which functional items arederived from lexical ones. Given the fact that such processes take place at anytime and at different speeds, we can discover within a given morphosyntacticarea items displaying different degrees of grammaticalization at any synchronicstage – on the assumption, of course, that we accept the well-known ideali-zations necessarily linked with the concept of synchrony.

The development of a functional item from a lexical one is characterized bycertain concomitants such as the declining phonetic and semantic integrity of alinguistic sign (i.e. erosion of the PF output form and loss of semantic features,often referred to as ‘desemantization’), the decreasing syntactic scope (whichcan be expressed in terms of asymmetric c-command)9, and the diminishingparadigmatic and syntagmatic variability of the item concerned. With referenceto these phenomena, the elements mentioned in (1), (2) and (3), above, can besituated on a so-called ‘scale of grammaticalization’ (Lehmann 1982) in thefollowing way:10

(8)

From this point of view the most interesting items are those situated near thetwo poles of the scale, namely case marking P elements and prepositional pe-riphrases. At this point, the grammaticalization framework must be completedby a syntactic theory capable of giving a clear insight into syntactic facts like

9 Let me briefly illustrate this with the example of the Romance future marker which is derivedfrom the Latin verb habere ‘to have, to possess’ via an intermediate stage, where habere functions asa future auxiliary comparable to English will (Roberts 2001: 114ff). While a lexical head like Latinhabet ‘s/he has’ as in [DP magnum pulcherrimumque hortum] habet ‘s/he has a big and very beautifulgarden’ asymmetrically c-commands every terminal node contained in its DP complement, the Ro-mance future affix as in French chanter-a or Spanish cantar-á ‘s/he will sing’ only modifies the ver-bal stem and hence does not c-command any other terminal node.

10 In his (1985) paper Lehmann offers a more detailed scale which distinguishes 5 different ‘de-grees of grammaticalization’: “[French] has a host of prepositional locutions of stage 1, such as àcause (de), en face (de) etc. It has also an exhaustive paradigm of prepositions which occupy posi-tions on stages 2 and 3 and between them, e.g. autour (de) or jusqu’à at stage 2, pendant and versbetween 2 and 3, and par or dans at stage 3. Finally, the prepositions de and à may be said to be atstage 4, with even some characteristics of stage 5” (Lehmann 1985: 46).

‘lexicon’lexical categories

‘grammar’functional categories

(3) (2) (1)

grammaticalization

12 Christoph Gabriel

word order variation. While syntactic theories like valency or dependencygrammar, traditionally highly appreciated within the field of Romance linguis-tics, do not account for questions of word order, work within generative gram-mar has concentrated on deriving linear ordering from underlying base configu-rations. It goes without saying that in cases where changes in word order markthe evolution of new items – and the prepositional use of a participle like ex-cepté in excepté les femmes ‘except for women’ clearly is one of these (seesection 5.2.1. for details) – a theory accounting for these facts is essential.

4.2 Reanalysis

Following Hopper & Traugott (1993) and Roberts & Roussou (1999), amongothers, I claim that syntactic reanalysis11 plays a predominant role in grammati-calization. Within the field of prepositional elements, two types of linguisticchange can be observed which can plausibly be accounted for by invoking thisconcept:

(9) a. Change in the feature structure of a given entity resulting in the tran-sition from one category to another; e.g. the prepositional use of apast participle like excepté(es) or the use of a prepositional element asa case marker.

b. Fusion of two or more syntactically distinct elements in such a waythat they form a single unit for the purposes of syntax; e.g. the fusionof three items like [P à], [D le] and [N lieu] by means of incorporationof [DP le lieu] into the preposition, yielding [P P DP] au lieu (de).

Given the fact that P is seen as a lexical category on a par with N, V and A,generative work has concentrated on prepositional elements serving as overtcase markers, i.e. on shifts from lexical to functional items, largely focusing onthe ‘grammatical pole’ of the assumed scale of grammaticalization. One objec-tive of my paper is to show that the generative framework also accounts forcategorial shifts at the ‘lexical pole’ of the scale (see section five).

4.3 Minimalist Phrase Structure Building

Under Minimalist Assumptions, the architecture of grammar is roughly designedas follows: According to the requirements of a given speech act lexical items(LI’s)12 such as α and β are taken from the lexicon as bundles of features, each

11 “Change in the structure of an expression or a class of expressions that does not involve any… modification of its surface manifestation” (Langacker 1977: 58).

12 Note that the comprehensive term ‘LI’s’ subsumes all items stored in the lexicon, i.e. bothsubstantial lexical and functional categories.

Relational Elements in French 13

comprising a matrix of semantic, phonological and formal features (SF(LI),PF(LI), and FF(LI), respectively). All feature values are fully specified forAgreement, Case, Tense, etc. This operation is commonly referred to as Select,and the array of LI’s thus obtained is called the Numeration N. The operationMerge then combines the LI’s from the Numeration to larger syntactic objects ina strictly binary way. An object thus obtained is characterized by the formalproperties FF(LI) either of α or β, i.e. either α or β projects. The structurebuilding mechanism Merge is subject to the requirements of theta-theory,among others. Let me illustrate this with the double-object construction (2.a),repeated here for convenience.

(10) il mit ses baskets dans l’armoire‘he put his sneakers into the wardrobe’

Here, the verbal head [V mit], which contains a <LOC> role in its theta-grid, ismerged with the constituent [PP dans l’armoire].13 The verbal projection [V’ mitdans l’armoire] is subsequently merged with a DP constituent like ses basketswhich receives its theta-role THEME in Spec,VP. VP is then merged with thelight verb v, to which V overtly adjoins, and the Subj theta-role AGENT is fi-nally assigned by v’ to the DP in Spec,vP.14 Once all theta-roles are discharged,vP merges with T, the functional category hosting the so-called checking fea-tures for Tense, Subject Agreement, and (nominative) Case, among others. Fi-nally, linear ordering of the terminals is derived by application of the operationAttract F: A checking feature triggers movement of the closest category en-dowed with an appropriate feature value into its checking domain (i.e. the speci-fier position of the head H hosting the relevant checking feature or any adjoinedposition within H’s maximal projection). Movement is overt if (and only if) theattracting feature is strong. In French this is the case for both the D-feature (i.e.,the EPP feature) and the V-feature in T. So V+v overtly adjoins to T, and theSubj DP moves Spec,TP. Evidence for the strong features in T comes from verb-

13 Zwarts (1992) and Rauh (1998) extend Longobardi’s (1994) view of (non-referential) NPsand (referential) DPs to the maximal projections of P elements and consequently distinguish be-tween non-referential PPs such as pour son frère or avec un ami, which cannot be assigned theta-ro-les and hence never appear in A-Positions, and referential PPs such as dans l’armoire , which arelicensed in A’-Positions as well as in A-Positions. This difference between the two types of preposi-tions is captured by assuming that the theta-grids of some prepositional elements contain a referen-tial role in the same way as nouns and verbs do. This role is called <S> (‘space’) and must be dis-charged by means of ‘theta-binding’ with an appropriate functional head, called R. This functionalhead R, which can be either empty or lexically filled with items such as ici and là, binds the referen-tial role <S> (‘space’) of a (referential) P element in the same way as D serves as theta-binder forthe referential role <R> of a noun or as Infl/T binds the event role <E> of V (Higginbotham 1985).For reasons of simplicity I abstain from distinguishing between (non-referential) PPs and (referen-tial) RPs in the following. See Zwarts (1992) and Rauh (1998) for further discussion of the theoreti-cal background and Gabriel (in press) for a detailed application to French data.

14 “The external role is a property of the v-VP configuration, and a specifier bearing this role istherefore a necessary part of the configuration” (Chomsky 1995: 316).

14 Christoph Gabriel

oriented adverbs (such as soigneusement ‘carefully’) which obligatorily followthe finite verb (Pollock 1989):

(11) il mit (soigneusement) ses baskets dans l’armoire‘he (carefully) put his sneakers into the wardrobe’

All strong features being checked, Spell-out applies and separates FF(LI) andSF(LI) from PF(LI), the former serving as input for the semantic interpretationof the structure at LF and the latter being interpreted by the phonological rulesapplying to PF representations. The linear ordering of the constituents is derivedas sketched in the following phrase marker.

(12)

4.4 Optionality and the Concept of Feature-driven Syntax

Strict Minimalism is based on the assumption that theta-roles are assigned inidentical structural configurations15 and that all instances of overt movement aretriggered by strong checking features.16 From this point of view different pat-

15 See also the following statements: “θ-relatedness is a ‘base property’, complementary to fea-ture checking, which is a property of movement. More accurately, θ-relatedness is a property of theposition of merger and its (very local) configuration ... θ-relatedness generally is a property of ‘basepositions’” (Chomsky 1995: 313). This is conform with Baker’s Uniformity of Theta AssignmentHypothesis (UTAH): “Identical thematic relationships between items are represented by identicalstructural relationships between those items at the level of D-structure” (Baker 1988: 46). In his(1996) monograph Baker offers a more detailed definition: “Thematic roles are always assigned inthe following configurations: (i) Theme to the specifier of the minimal VP; (ii) Goal (more gener-ally, path) to the complement of [the] minimal VP; (iii) Agent outside the minimal VP (typically, tothe specifier of a VP shell)” (1996: 226).

16 A similar mechanism can be seen in the ‘P(rosodically motivated)-Movement’, which hasbeen proposed by Zubizarreta (1998) in order to capture cases of focus-induced word order variationin Romance and Germanic Languages: In Zubizarreta’s framework a focus feature F (which is freelyassigned by the speaker in the course of the derivation at a level called Σ-structure) triggers eitherthe phonological prominence of the F-marked constituents or the reordering of the terminals. Thechoice between the syntactic or phonological option for signaling focus, then, depends on language-specific preferences for the former or the latter.

VP

PPdans l’armoire

V’

v’

Spec,VP[DP ses baskets]

Vmit

vP

V+vmit

Spec,vPil

Spec,TP [DP il]

T’

(V+v+)Tmit

LOC

THEME

AGENT

TP

Relational Elements in French 15

terns of word order can only be accounted for on the condition that a given at-tracting feature can alternatively be strong or weak, as this is e.g. proposed byCulicover (1999: 144) for French wh-in-situ questions. This approach clearlycauses serious conceptual and empirical problems, especially as soon as para-metric change such as the shift from Latin OV to Romance VO or the loss of V2structures in the history of French has to be accounted for. Roberts & Roussou(1999) and Roberts (2001) therefore have developed an alternative proposal toMinimalist Checking Theory, which will be discussed in connection with myanalysis in the following section.

5 The ‘Transitory Edges’ of the Scale of Grammaticalization

This section concentrates on the analysis of the two ‘transitory edges’ of theassumed scale of grammaticalization.

5.1 Transitions from P to K

Following Rauh’s (1998) proposal a subset of prepositions in certain uses canplausibly be interpreted as overt realizations of inherent case. For the purposesof syntax we are dealing with functional heads K, which occur in the extendedprojection of a nominal head N (cf. Grimshaw 1991). In light of this, a phraselike à son frère as in (1.a) elle offre un livre [KP à son frère] is to be analyzed asan overtly case marked DP, hence as a nominal projection. As shown in (13),the different functional projection levels are indicated by {F0}, {F1}, and {F2}.

(13)

I assume that we are dealing with the functional use of a preposition if (and onlyif) the occurrence of the element concerned can be required by a lexical head forthe marking of a subcategorized constituent in a certain context. This changefrom a preposition P to functional head K can be captured as an instance of the“reanalysis of lexical material as functional material” (Roberts & Roussou 1999:1011), involving

NP [+N; –V] {F0}

Nfrère

DP [+N; –V] {F1}

Dson

KP [+N; –V] {F2}

16 Christoph Gabriel

– a change in the FF(LI) matrix from [–N; –V] to [+N; –V] {F2} (but mainte-nance of the case checking feature [D:case:ACC], see below for details) and

– a complete loss of theta-theoretical properties (given the fact that a DP likeson frère in (1.a) receives its RECIPIENT role from the lexical head [V offre]and not from K).

As shown in (14), French K elements correspond to the values of the case fea-tures of lexical heads, i.e. case checking features which are specified for inher-ent case require that the relevant K item be selected for the Numeration.17 In thiscontext I distinguish between inherent cases which correspond to particular pro-forms, namely the proclitic indirect object pronoun lui (14.a) or the possessivedeterminer son (14.d), and those which allow only for replacing the DP by anon-clitic pronoun (pronom disjoint), but retaining the prepositional case markerK (14.b & c). The latter will henceforth be referred to as ‘objectoid case’ (Siller-Runggaldier 1996).

(14) a. offrir [–N; +V] <1,2,3,E> [D:case:ACC], [D:case:DAT]Marie offre un livre à son frère. Marie lui offre un livre.‘Marie gives a book to her brother / to him.’

b. penser [–N; +V] <1,2,E> [D:case:à]Marie pense à son frère. Marie pense à lui.‘Marie thinks of her brother / of him.’

c. compter [–N; +V] <1,2,E> [D:case:sur]Marie compte sur son frère. Marie compte sur lui.‘Marie relies on her brother / on him.’

d. portrait [+N ; –V] <1,2,R> [D:case:GEN], [D:case:GEN]C’est le portrait d’Aristote de Rembrandt. Rembrandti ... son i portraitd’Aristote.‘This is Rembrandt’s portrait of Aristotle. Rembrandti ... hisi portraitof Aristotle.’

The so-called ‘objectoid cases’ crucially differ from genitive and dative case inthat they exclusively mark complements, whereas the latter can also occur asadjuncts. In this case, their case feature is not checked by a subcategorizinghead. This holds for the so-called non-lexical or free dative, as in je lui tricoteun pull-over ‘I knit a sweater for him/her’, and for the non-argumental posses-sive genitive, as in les baskets de Pierre ‘Peter’s sneakers’. In order to accountfor these cases, I assume contra Chomsky (1995) that inherent case is[+interpretable] and that the relevant features [case:DAT] and [case:GEN] canremain unchecked without causing the derivation to crash.

It is important, however, to underline that case marking prepositions such asFrench [K à], [K de] or [K sur] cannot be identified on a par with case suffixes of

17 I assume that K elements are stored in the lexicon on a par with other functional categoriessuch as D and C. For a different position, see Rauh (1998: 188), who claims that K items have nolexical entry at all.

Relational Elements in French 17

inflecting languages. While the latter are combined with open nominal stemsand bear not only case features but phi-features as well, the former are mergedwith full DPs and are never endowed with features other than case.18 Conse-quently, French (and Spanish) K elements display crucial parallels with aggluti-native case affixes of Turkic languages.

The prepositional origin of French case marking à, for instance, becomesmanifest in several syntactic aspects. For example, case marking prepositions Kand lexical prepositions P are both subject to the well-known restriction calledPP island constraint, prohibiting a constituent being extracted by pronominali-zation from within an XP headed by K and P:

(15) a. j’ai parlé [KP à [DP l’auteur de ce livre]]‘I talked to the author of this book’

b. *ce livrei ... j’eni ai parlé à l’auteur(16) a. je suis allé à l’opéra [PP avec [DP l’auteur de ce livre]]

‘I went to the opera with the author of this book’b. *ce livrei ... j’eni suis allé à l’opéra avec l’auteur

Concerning case-theoretical properties, K elements again clearly reveal theirprepositional descent: They check the structural case of their DP complement inthe same way as lexical prepositions do. While this assumption relies uniquelyon theory-internal implications for French, there is an obvious manifestation ofovert case morphology in languages like English and German, where theequivalents of the French example (14.c) clearly exhibit accusative case on the(pronominal) DP:

(17) a. Mary relies [KP on [DP [case:ACC] him]]b. Marie verlässt sich [KP auf [DP [case:ACC] ihren Bruder / ihn]]

Note that this is why some scholars have argued against the assumption of casemarking prepositions. Radford, for example, states in his (1997) monographwith reference to the English genitive marker of: “[What] has always remained amystery under the genitive analysis is why of takes an objective complement”(453). Keeping in mind, however, that even agglutinative case affixes like theTurkish and Azerbaijanian comitative marker -lE require overtly case markedforms, at least as far as pronominal DPs are concerned (see (18) below), there isno reason for strictly rejecting the idea of (functional) K elements checkingstructural case, all the less as other functional elements are assumed to check

18 Note, however, that some scholars have proposed assuming inflected prepositions for Italian,see Napoli & Nevis (1987). Taking this into account, the amalgam au formed from [P à] and [D le] ina structure like elle parle au directeur ‘she talks to the director’ could be analyzed as an elementbearing case and phi-features as well. From my point of view, these forms are better accounted forby simply interpreting them as results of phonological rules applying to PF representations.

18 Christoph Gabriel

case as well – recall that T is usually said to check the nominative case featureof Subj.

(18) a. Turkish: ben-im-le, sen-in-le, on-un-la, etc.b. Azerbaijanian: , , on-un-la, etc.

1SG-GEN-COM, 2SG-GEN-COM, 3SG-GEN-COM, etc.‘with me, with you, with him/her’, etc.

Apart from these properties underlining the similarities of functional elements Kand lexical prepositions P, there are aspects in which case marking elementsclearly differ from the latter: With respect to their possible omission in coordi-native structures, French prepositional case markers, omissible in some re-stricted contexts only, seem to be grammaticalized to an even higher extent than,for instance, the agglutinative case affixes of Turkish, which can generally beomitted in coordinative structures (Kornfilt 1997: 122). Consider the followingexamples:

(19) a. j’ai envoyé une lettre [KP à [DP Marie et Jeanne]]‘I sent a letter to Marie and Jeanne’

b. j’ai envoyé une lettre [KP à [DP Marie et sa fille]]‘I sent a letter to Marie and her daughter’

c. *j’ai envoyé une lettre [KP à [DP Marie et le directeur]]‘I sent a letter to Marie and the director’

d. je suis allé à l’opéra [PP avec [DP Marie et le directeur]]‘I went to the opera with Marie and the director’

Note that (19.d) is perfectly acceptable, being equivalent to (19.c) except that[K à] is replaced by the lexical preposition [P avec] and the complement status ofthe coordinated DPs has been changed to adjunct status by virtue of replacingthe verb envoyer by the unaccusative aller.19

Case marking prepositions K thus clearly display some ‘prepositional’ prop-erties on the one hand, but crucially differ from lexical prepositions on the other.On the whole, they should be seen as functional elements on a par with aggluti-

19 A case marking preposition K can be omitted in a coordinating structure if condition (i) andeither of the conditions (ii) or (iii) are fulfilled.

(i) The entities denoted by the two conjuncts can plausibly be interpreted as belonging to-gether in some way (“form a single entity”; Jones 1996: 379).

(ii) The two conjuncts are connected with each other by means of a co-referential possessivedeterminer.

(iii) The two conjuncts display an identical surface representation, i.e. they both have eithertheir D position filled or empty.

Note that a prerequisite of condition (iii) is Longobardi’s (1994) assumption that proper names are Nheads, which form NPs and must be embedded in a functional projection of D whenever they appearin A-positions, as is the case for the examples given in (19) above. (19.c) is ruled out, then, becauseit fulfills neither (ii) nor (iii), although Marie and le directeur can plausibly be construed as the jointrecipients of the letter (condition i).

Relational Elements in French 19

native case affixes, but clearly differing both from lexical uses of the items con-cerned and from the case affixes of inflecting languages. In order to distinguisha K element from the lexical prepositions they derive from, they receive theirown entries in the lexicon on a par with other overt functional categories such asdeterminers or complementizers.20 These K elements are obligatorily selectedwhenever a lexical head bearing a checking feature for genitive, dative or an‘objectoid’ case is present in the Numeration (see 14, above). With reference tothe Germanic and Turkic data in (17) and (18) I have assumed that prepositionalcase markers K preserve their checking feature for (structural) accusative case.From this point of view they must be seen as case markers and case checkers atthe same time. In contrast to P elements, however, they completely lack theta-theoretical properties in not being able to assign a semantic role to the comple-ment DP, which gets its role from the verbal (or nominal) head according to thelexical entries as given in (14).

5.2 Transitions between P and the Adjacent Lexical Classes

In the remainder of this section, I will focus on categorial shifts between V andP, namely on prepositional uses of participles, and then concentrate on complexprepositions emerging from constructions with relational nouns.

5.2.1 V (past participle) > P

First I will discuss the changes observed when a past participle verb form likeexcepté is used as a preposition: [PP excepté les femmes]. In the underlying pas-sive construction [TP les femmes (sont) exceptées] the subject DP is base-gener-ated in complement position, being an internal argument of V (see phrasemarker (20.d) below for details). According to Burzio’s (1986) well-knowngeneralization passive morphology suspends both external theta-role and casechecking feature. Therefore the internal argument DP moves to Spec,TP in orderto get its (nominative) case checked. In languages which display a periphrasticpassive construction using the past participle form like French, a tensed auxil-iary must be inserted at some point in the derivation in order to provide an ap-propriate configuration for the checking of nominative case by means of the Thead which appears in its extended projection.21 Given the fact that in passiveconstructions the suspended theta-theoretical and case properties of V (past

20 Note that we are dealing with a clear case of ‘divergence’, see section three, above.21 Note that in languages displaying a synthetic passive construction such as classical Latin

(videtur ‘s/he is seen’), Turkish (gör-ül-üyor- see-PASS-PROG-3SG ‘s/he is seen’) or Azerbaijanian( - oppress-PASS-PRS-3SG ‘s/he is oppressed’) passive morphology does not trigger insertionof Aux given the fact that the passive morpheme is realized as an affix on the tensed (lexical) verbform which is consequently embedded in an articulated CP/IP structure containing T.

20 Christoph Gabriel

participle) need not be reactivated as is the case for periphrastic active forms(e.g. the passé composé; see Friedemann & Siloni 1997), a form of être is in-serted in the derivation. This can be either the finite auxiliary sont as in (20.a),the present participle form, yielding (20.b), or, finally, a phonetically null aux-iliary [e]22, yielding a so-called ‘absolute’ construction (20.c):

(20) a. [TP les femmes sont exceptées]b. [CP les femmes étant exceptées] …c. [CP les femmes [e] exceptées] …

Note that (20.b & c) must be given a CP interpretation, since these constructionsnever appear as root clauses. In all cases raising of Subj is triggered by T’sstrong EPP-feature; case and agreement of Subj are checked in Spec,TP. Fol-lowing Friedemann & Siloni (1997) I assume that participial agreement withSubj is checked within a special AgrP, which is projected immediately aboveVP. Participial agreement, then, is morphologically overt if Subj raises prior toSpellout, i.e. if Spec,AgrP is created as an intermediate landing site.23 Note thatin French (and in English as well), auxiliaries also raise prior to Spellout be-cause of their linear ordering with respect to VP-related adverbs (Pollock 1989).

22 Given the fact that Subj raises overtly in absolute constructions, the most natural way to ac-count for this fact is to assume a phonetically null auxiliary [e], selected as a bundle of formal and,perhaps, semantic features only, and licensing T in its extended projection. Another possibilitywould be to assume that the auxiliary is selected as a complete category, consisting of SF(LI),PF(LI) and FF(LI) and that phonetic features are deleted prior to Spellout.

23 See Guasti & Rizzi (2000) for a different position. While the authors’ idea that overt mor-phological agreement in some way depends on the checking of the relevant features in the overtsyntax (and thus on overt movement) is compatible with the approach sketched above, Guasti &Rizzi (2000: 15, footnote 3) attribute the (obligatory) agreement properties of the participle verbform to its overt movement to a (lower) Agr head which appears among several functional layersimmediately above VP. Strictly speaking they assume (at least) two different functional projections(AgrP and PartP; see 2000: 17) instead of my single AgrP. Among other things, this refinementprovides an account for the cross-linguistic differences concerning the placement of the past partici-ple verb form with respect to floating quantifiers (see French il a tout compris vs. Italian Gianni hacapito tutto ‘he/Gianni has understood everything’). However, this is not relevant to my discussionhere.

Relational Elements in French 21

(20) d.

In order to capture the change from the absolute passive construction (with overtraising of Subj yielding les femmes exceptées) to the prepositional construction(displaying the reverse ordering and lacking overt agreement on the verb form,i.e. excepté les femmes) it is necessary – under the assumptions exposed in sec-tion 4.3 – to assume that the strong (EPP) feature in T becomes weak and Subjtherefore remains in situ. The fact that the participle verb form displays no overtagreement can consequently be attributed to the lacking Spec-Head-Con-figuration between the Agr head and the DP. At this point of the derivation, theparticiple can be reanalyzed as a P element by the speakers; i.e. we are dealingwith a categorial shift from V to P, which is to be described as a change in thefeature structure of the item concerned, yielding [–N; –V]. Consequently, noauxiliary is inserted in the derivation and the resulting structure crucially differsfrom the former verbal construction in dispensing entirely with the FPs usuallyprojected in the articulated IP structure. Following Radford (1997: 451ff), Iadopt Chomsky’s (1995) v-VP shell structure for the analysis of prepositionalconstructions and assume that the PP is embedded in a ‘light’ projection headedby p to which P overtly adjoins.

While the structural case of the DP is checked by overt raising to Spec,TP inthe underlying verbal construction, the case feature of the DP must be checked,after reanalysis, by the prepositional head. We therefore have to assume that thereanalyzed preposition ‘reactivates’ its case checking feature which is assumedto be ‘suspended’ in the participle. Case and phi-features of the DP complementare checked, then, by raising of FF(DP) to the complex head [p P p]. Given thefact that the resulting construction displays no overt agreement between P andthe DP, we have to assume that only the D-related Agr feature in p is fully speci-fied for agreement, i.e. [D: P(erson)3, N(umber)pl G(ender)f], whereas the P-related Agr feature encodes the default value α for all the phi-features, i.e.[P:Pα Nα Gα].

VP

DPles femmes

[D: P3 Npl Gf][Case:NOM]

Vexceptées

[V: Pα Npl Gf]

Agr’

Agr[D: P3 Npl Gf][V: Pα Npl Gf]

AgrP

Spec,AgrPles femmes

[D: P3 Npl Gf][Case:NOM]

AuxP

Auxsont/étant/e

T’

Aux + T [D: Case:NOM] sont/étant/e

TP

Spec,TPles femmes

[D: P3 Npl Gf][Case:NOM]

(CP)

(C)

22 Christoph Gabriel

(21)

Most prepositions of this kind, among others excepté, vu ‘in view of, given’ andy compris ‘including’, are likely to be derived from ‘absolute’ passive construc-tions displaying a null auxiliary [e]. But it is also possible that an overt auxiliaryform is preserved as part of the reanalyzed P, as in étant donné ‘given’.

At this point, we should return to the question raised in section 4.4: Does Mini-malist Checking Theory provide a useful account for different patterns of wordorder, given the fact that alternating feature strength must simply be accepted,and – more specifically – can the categorial change V > P be plausibly capturedas sketched in the previous paragraphs?

Let me first enter into a brief discussion of Roberts & Roussou’s (1999) pro-posal. Instead of assuming strong checking features triggering movement in theovert syntax, the authors propose that “all features have an LF-interpretation,that is, there is a universal pool of substantive features, and that languages varyin which features are required to have a PF-interpretation (be overtly) realized.A functional feature F that requires a PF-realization is notated F*” (1999: 1017).Concerning diachrony they assume that “[t]he distribution of * over {F1 ... Fn}may change as a function of time” (1999: 1020). On the assumption that Mergeis generally ‘cheaper’ than Move, diachronic change and – more specifically –grammaticalization can be characterized as “change from the F*move to theF*merge option” (1999: 1022). A well-known example is the change from theLatin lexical verb habere ‘to have, to possess’ to the Romance auxiliary avoir/haber/avere and future marker (see Roberts 1993a, 1993b): While a lexical headlike Latin habet ‘s/he has’ is base-generated internal to VP and then moved to afunctional head (option F*move and merge), an auxiliary is directly merged in a func-tional position external to VP, and a mere affix is base-generated together withV and checks only the appropriate features on T. Parametric change such as thedevelopment from Latin OV to Romance VO can also be interpreted as the lossof the F*move option: On the assumption that all languages are underlyinglyhead-initial (Kayne 1994), VO languages lack the overt movement of Obj to apreverbal position (see Roberts 2001: 120-122 for a detailed discussion of theEnglish facts and a short remark on Romance).

Roberts & Roussou illustrate their idea of a dichotomy F vs. F* with the ex-ample of the feature Q (situated in C), which the authors take responsible forgiving a clause the interpretation of a yes/no-question and which either requires

PP

DPles femmes[P3 Npl Gf][Case:ACC]

pP

Pexcepté

p[D: P3 Npl Gf][P: Pα Nα Gα]

p

Pexcepté

[D: Case:ACC]

pFF(les femmes)[P3 Npl Gf][Case:ACC]

Relational Elements in French 23

an overt realization by means of Move and/or Merge (Q*) or neither Move norMerge (Q).

(22) a. did John see Mary? English: Q*move

b. a welodd John Mary? Welsh: Q*merge

c. a-t-il vu Marie ? Literary French: Q*move and merge

d. il a vu Marie ? Colloquial French: Q is silent. (1999: 1019).

In (22.a) Q* is satisfied by Move, i.e. by raising [T did] (the closest availablehead) to C because English has no “Q-particle”, as does Welsh, where Q* issatisfied by merging the relevant item directly in C (22.b). The difference be-tween the inverted interrogative structure and the wh-in-situ question in French,finally, is accounted for by assuming that “Q is silent” for the latter case (22.c &d).24 If we apply this idea to the example discussed above, we would have tostipulate that in (20) EPP* (or rather AgrS* in Roberts & Roussous’s system) issatisfied by Move of [DP les femmes], but that EPP is silent if the DP stays insitu, i.e. we have to assume once again that one and the same formal feature canhave different properties (strong/weak and */silent, respectively). The sameholds for structures with unaccusative verbs (23.a & b) or passive constructionswith particularly heavy DPs (23.c), where the internal argument DP can option-ally remain in its base position:

(23) a. [TP la reine T [VP arrive la reine]]‘the queen arrives’(strong feature in T; EPP*move)

b. [TP T [VP arrive la reine]] ‘the queen arrives’LF : [TP FF(la reine) T [VP arrive la reine]](weak feature in T; EPP is silent)

c. [TP T [auxP sont exclues ... [DP les formes adjectives des verbes transitifsindirects ... et la très grande majorité des verbes intransitifs]]‘adjectival forms of transitive indirect verbs ... and the overwhelmingmajority of intransitive verbs are excepted’(weak feature in T; EPP is silent)

While the variants in (22.c & d) may be attributed to different grammars (i.e.formal/literary vs. colloquial French), (23.a & b) are clearly available within oneand the same register. The same holds for (23.c) where ‘heavy shift’ of thecomplement DP – which is in fact to be interpreted as a case of (covert) move-ment of FF(les formes adjectives …) – is a stylistic option of formal/literaryFrench. The conceptual problem of postulating an EPP feature which is some-times strong and sometimes weak can be obviated in assuming that featurestrength can be ‘oscillating’ with respect to some strictly limited areas: i.e. the

24 In (22.c) Q* is satisfied by Merge of the subject enclitic (which the authors analyze as an in-terrogative marker, following Friedemann 1997: 185ff) in T, adjunction of the auxiliary verb a and afurther Move of a-(t-)il to C.

24 Christoph Gabriel

selection of an unaccusative verb form or Merge of a particularly complex DPas an internal argument in some way activates the ‘oscillating property’ of thestrong EPP feature, yielding output forms like those given in (23.b & c). It isimportant to underline that these cases crucially differ from cross-linguisticvariation or from parametric change in that no general change of a grammaticalsystem is at play, but rather some kind of variation within one and the samesystem.

With this in mind, I retain the Chomskyan idea of strong features triggeringovert movement, assuming some well defined cases of ‘oscillating’ featurestrength. The categorial transition from a past participle verb form to a P ele-ment can thus be characterized by the following steps:– A prerequisite of a possible reanalysis V (past participle) > P is the fact that

V theta-marks an internal argument, e.g. excepter <1,2,E>.– Following already existent cases of ‘oscillating’ feature strength (unaccusa-

tives, heavy DPs in passive constructions, see above), the complement DPstays in its base position: [TP excepté(es) les femmes].

– The resulting surface structure is reanalyzed as a PP: [PP excepté les femmes].– Once the participle is reanalyzed as a preposition there is no overt agreement

with the relevant DP: [PP excepté(*es) les femmes].Note, however, that there are instances of graphically realized agreement:

(24) a. exceptées les années 1945-1975‘except for the years 1945-1975’(A. Glucksmann; Grevisse 1993:§311)

b. tous les favoris, exceptés Kelly et Fignon ...‘all the favorites, except for Kelly and Fignon’(France Soir 27.4.1992; Müller-Lancé 1994: 281)

Given the fact that French doesn’t display systematical agreement between anadposition and the complement DP, these cases clearly show that the speakers –or rather: writers – feel that there is quite an intimate relation between suchprepositional forms and the underlying passive construction (where overtagreement is compulsory).

5.2.2 V (present participle) > P

While the examples discussed in 5.2.1 are characterized by the fact that neitherSubj nor V/Aux undergo overt raising once the participle is reanalyzed as apreposition, the change from a an absolute participle construction involvingovert raising of Subj and V+v to T as in

Relational Elements in French 25

(25) a. [CP [TP l’étéj [[durant + v]i + T][vP [l’été]j [durant + v]i [VP duranti]]]]elle a décidé de passer ce week-end à la campagne25

‘while the summer is still lasting, she decided to spend this weekendin the country’

to the PP construction [PP durant l’été] ‘during the summer’ can eventually beaccounted for by assuming an intermediate stage which only dispenses with theraising of Subj, but maintains movement of V+v to T:

(25) b. [CP [TP [[durant + v]i + T][vP l’été [durant + v]i [VP duranti]]]

After adjoining to T, the present participle verb form could be assumed to un-dergo reanalysis as a P head:(25) c. [pP duranti + p [PP [P duranti][DP l’été]]]

The theta-grid of the item concerned undergoes considerable modification inso-far as the external role of the underlying verb form is reinterpreted as an internalrole in the (reanalyzed) prepositional theta-grid.

What still remains unclear is the question of why the overt raising of Subjshould be dispensed with in (25.b). In the previous section I interpreted the non-raising of Subj as a case of over-generalization of a general property (namelythe activation of the potential ‘oscillation property’ concerning the strength ofthe EPP feature; see the discussion of excepté les femmes) to contexts where thisproperty usually does not apply, providing a configuration which is accessiblefor the reanalysis. Concerning the development of durant, however, we canhardly find a construction which serves as a model as is the case for excepté. Ofcourse, the reanalysis of the present participle durant as a P element follows insome way the well-known example of pendant ‘during’, which was reanalyzedas a preposition in the same way during the 14th century, but we are left withthe same problem when we ask for a reason why speakers should produce a newpattern of word order yielding durant/pendant l’été, given the fact that absoluteparticiple constructions are not normally likely to activate the ‘oscillation prop-erty’ of the EPP feature. The trigger for the categorial shift from durant (partici-ple) to durant (preposition) can thus hardly be attributed to syntactic properties,as is plausible for past participle constructions, but should rather be seen as ashift which directly applies to the lexical item [V durant], i.e. as an extension ofthe lexical means expressing temporal relations: [P avant/durant/après] l’été ‘be-fore/during/after the summer’. The speakers simply use the item durant as a Pelement in analogy to pendant (which was reanalyzed before in analogy to otherprepositional constructions), and the change in word order, i.e. ‘durant DP’instead of ‘DP(Subj) durant’, should rather be seen as a consequence of the re-analysis and not as its prerequisite, as has been proposed in connection with ex-

25 The absolute participle construction is given a CP interpretation, because such constructionscannot appear as a root clause.

26 Christoph Gabriel

cepté. It should be pointed out, however, that pendant nowadays is entirely in-tegrated into the lexical category P, while durant also occurs as a postpositiveelement, as in the example given in (25.a). I assume that the latter is a participialconstruction, while durant in (25.c) should be seen a member of the lexical classP. Here again, we are dealing with divergence phenomena, typical of categorialshifts near both the ‘grammatical’ and the ‘lexical’ poles of a postulated scale ofgrammaticalization.

5.2.3 N > P

While the categorial shifts from V to P, treated in the previous sections, areessentially characterized by changes in the feature structure of a given lexicalitem (reanalysis 9.a), the changes to be dealt with in this section, namely thefossilization of constructions with relational nouns to complex prepositionsand/or idiomatic structures, also involve incorporation, i.e., a reanalysis of type(9.b).

It is interesting in this context that some of the complex expressions men-tioned in (3), above, e.g. à la place de, allow the substitution of their DP ‘com-plement’26 by a possessive determiner.

(26) a. b.

26 ‘Complement’ is put in quotation marks in order to signalize that the DP concerned is notbase-generated in complement position but rather originates in the typical possessor position,namely in the Spec of a higher n projection, occurring in the articulated DP structure, as proposed byCarstens (2000) and Laenzlinger (2000), among others. The structure in (26.a), however, containsonly one nP shell given the fact that only one KP, namely the possessor, has to be merged. Linearordering is derived by overt raising of [N + n] to the functional head Agr, which is referred to as F inCarstens’ articulated NP structure (I prefer, however, the label Agr, given the fact that overt ‘object’agreement, e.g. between a noun and a possessor such as in Turkish Ali’nin spor ayakkabı-sı Ali-GEN

sport shoe- POSS3SG ‘Ali’s sneakers’, is checked in this position). Agr must be assumed to bear astrong N-feature. For the moment, I leave open whether N also raises overtly in constructions suchas (26.b), where output conditions do not require the noun to raise.

Dsases

PP

DPPà

dans

N/NPplace

baskets

n

nP

p

pP

Agr

AgrP

PP

dans

n’

N/NPplace

baskets

Spec,nP[KP de Pierre][KP de Pierre] N + n

placebaskets

nP

DP

Dlales

AgrP

[N + n] + Agrplace

baskets

pP

p

Relational Elements in French 27

Note, however, that it is impossible to establish a clear-cut distinction betweenidiomatic expressions like à la place de (Pierre) ‘instead of P.’ and freely gen-erated PP constructions like dans les baskets de Pierre ‘in Peter’s sneakers’ withreference only to syntactic properties. We therefore have to recur to semanticaspects in order to establish a distinction between the two types of structures.Constructions of the type à la place (de Pierre) are crucially characterized bythe fact that their semantic content cannot be seen as a ‘sum’ of the meanings ofeach single item constituting the whole expression (Fleischer 1997: 30). There-fore idiomatic structures like these require their own lexical entries, but due tothe fact that they can undergo some syntactic processes such as the use of apossessive D, as shown in (26.b) above, they must be stored in the lexicon ascomplex structures containing two lexical heads.

Returning to the use of possessive determiners in such constructions, weshould point out that not all idiomatic structures allow this operation; see forinstance the ungrammaticality of (27.e-h). This property, however, does notcorrelate in a clear-cut way with the presence or absence of an overt D elementin the structure concerned, as might be expected at first sight:

(27) a. au sujet (de DP) → à son sujet e. au lieu (de DP) → *à son lieub. à l’insu (de DP) → à son insu f. au-dessus (de DP) → *à son dessusc. de la part (de DP) → de sa part g. en face (de DP) → *en sa faced. à la place (de DP) → à sa place h. à côté (de DP) → *à son côté

Considering these data, I claim that a structure like (27.e) au lieu (de DP), *àson lieu is nothing but a lexical head for the purposes of syntax, although itcontains an overt D element and is morphologically complex and analyzable forthe speakers without any further difficulties. Consequently, I assume complexprepositions which can be analyzed synchronically, but which display the syn-tactic behavior of simple prepositions, to be stored in the lexicon as simpleheads P. I assume furthermore that items like [P à côté ] and [P au lieu] bear aninherent case feature, specified for the objectoid case [case:de],27 while a prepo-sitional head like dans or avec is endowed with a case feature that checksstructural case (type 1).

27 The item de, occurring in à côté de Pierre or in au lieu de Pierre, is plausibly analyzed as a Kelement appearing in the extended projection of the noun Pierre, rather than being part of the fossil-ized structure itself, for the following two reasons: Firstly, de has to be repeated in coordinativestructures such as en face de l’opéra et de la bibliothèque ‘in front of the opera house and the libra-ry’. In case the prepositional element de is omitted in such structures, the same rules apply as for thedative preposition à (see footnote 19, above). Secondly, lexical material can only be inserted beforede in appropriate contexts, but never after de. Consider the two structures l’autobus s’arrête en face,je crois, de l’opéra, and *... s’arrête en face de, je crois, l’opéra. The former example is perfectlyacceptable, at least in familiar registers of spoken French. By contrast, the latter is completelyungrammatical. – In this context it should be stressed once more that French K elements serve ascase markers (i.e. realizations of inherent case features) and as case checkers (of the structural casefeature the complement DP is endowed with) at the same time (see section 5.1, above).

28 Christoph Gabriel

(28) type 1: lexical headsa. à côté P <1,2,S> [D:case:de] c. dans P <1,2,S> [D:case:ACC]b. au lieu P <1,2> [D:case:de] d. avec P <1,2> [D:case:ACC]

Apart from these simple heads P, I assume a second type of preposition stored inthe lexicon in the form of idiomatic structures containing two lexical heads. In(29) I give an example of such a lexical entry; the KP in bold characters signalsthe base position Spec,nP of the entity situated by the expression concerned.

(29) type 2: idiomatic expressionsà la place [PP P [DP D [nP [Spec,nP KP] n [NP N]]]] <1,2,S>

with [N place] bearing [D:case:GEN]

Concerning type 2 prepositions I assume that the inherent case feature[case:GEN] of the ‘complement’ is checked by the nominal head N after covertRaising of FF(KP = de Pierre) to the complex head [N + n + Agr] (see (26.a)above for the relevant phrase marker).

6 Concluding Remarks

Synchronic variation is characterized by the fact that marked constructionsemerge in addition to the unmarked structures, thereby displaying new uses oflinguistic material already present in a language. As pointed out by Roberts(1993a:153ff, 2001) and Roberts & Roussou (1999), the emergence of newconstructions follows the requirements of derivational economy , especially asfar as the reanalysis of lexical as functional material, i.e. grammaticalization, isat play. The examples discussed in section 5 show that this also holds for thecategorial transitions between the lexical classes V, N, and P: The reanalysis ofa V (past participle) as a preposition applies on the basis of a structure which ischaracterized by a more economical derivation involving covert instead of overtmovement (see 5.2.1). The categorial shift from V (present participle) to P(which I have assumed to occur without being favored by word order variation)on the other hand prevents the functional categories that usually appear in thearticulated CP/IP structure from being selected for the Numeration and thusleads to a less complex derivation, a CP like l’été durant containing more func-tional nodes and involving more instances of movement than the new pP shellconstruction durant l’été (see 5.2.2). Finally, the fusion of two or more itemssuch as à, le and lieu to a single head [P au lieu] yields less complex syntacticderivations given the fact that these expressions are taken from the lexicon as awhole and need not be merged in the syntactic component (see 5.2.3).

Given the fact that all items undergoing categorial transitions continue to oc-cur in their ‘old’ contexts, the shifts observed in the peripheral areas of the pos-tulated scale of grammaticalization can only be accounted for by invoking the

Relational Elements in French 29

concept of divergence, already mentioned in section 3. In Minimalist termssuch phenomena can be explained by assuming that one and the same outputform displays a different feature matrix when occurring in different contexts.

Considering the large number of prepositional expressions, P must be seenas a lexical category on a par with N, V and A, each of these synchronicallyexpandable by means of spontaneous word formation and/or fossilization ofmore or less complex constructions. Taking this into account, the developmentof a new preposition cannot automatically be regarded as an instance of gram-maticalization. Rather, we are dealing with a kind of lexical-lexical categorialshift when observing instances of the integration of verb forms and con-structions containing relational nouns into the ‘target’ class P.28 The develop-ment of new functional items, on the other hand, must be seen as a kind of lexi-cal-functional categorial shift, hence as instances of grammaticalization. Suchtransitions from lexical to functional elements can be observed when Frenchprepositions are selected by a subcategorizing lexical item and behave ratherlike the agglutinative case affixes of Turkic languages.

The significant differences between the two prepositional subcategories (seetypes 1 and 2 in (28) and (29), above), as sketched in the previous section, can-not be accounted for by invoking the concept of different degrees of grammati-calization. Rather, they represent a different degree of integration of a givenitem into the lexical category P. Nevertheless, it makes sense to localize prepo-sitions of type 2, i.e. the idiomatic expressions, to the left of the prepositionalheads (type 1) on a scale of grammaticalization, simply because the latter aremuch more likely to undergo a lexical-functional categorial shift, i.e. grammati-calization to a case marker:

28 See Müller-Lancé (1994) for a different position.

30 Christoph Gabriel

(30)

Furthermore, it is remarkable that new prepositional elements emerge almostexclusively from the adjacent lexical categories N and V, namely those sharingone categorial feature with the prepositional specification [–N; –V]. Elementscharacterized by [+N; +V], i.e. adjectives, on the other hand, hardly ever serveas a starting point for the emergence of new P elements.29 Finally, categorialtransitions from V or N to P and a further development of the same item to Kare not observed, at least not in the history of French. Consequently, the scalarrepresentation, traditionally assumed by grammaticalization theorists and serv-ing as a model for the table given in (30) above, turns out to be problematic aswell. It should therefore be replaced by a cube-shaped model as given in (31),accounting in a much more adequate way for the facts so far observed.

29 The item sauf can perhaps be seen as an instance of an A > P transition. – Further evidencecomes from Italian: As Vincent (1997:212) points out, the most important lexical sources for Italianprepositions are nouns and verbs: ABSENTIA > senza ‘without’, FINE(M) > fino ‘until’, MALO GRATO

> malgrado ‘despite’; DURANTE > durante ‘during’, PRESSU(M) > presso ‘near’, UERSU(M) > verso‘towards’, and many others.

functional category K

(parler) à DP

(le portrait) de DP

dative/genitivecase

à la fenêtrede machambre

lexical category Pfree syntaxe.g.concernant exclu-sivement les com-merçants ...

à la frontièredu Kosovo

à la place de DP

au sujet de DP

type 2: idioma-tic expressions

à l’insu de DP

e.g.

reanalysis

(9.a) or (9.b)

dans les basketsde Pierre

à l’entréede l’église

fossili-zation reanalysis

(9.b)(compter) sur DP

(parler) avec DP

(penser) à DP

objectoidcases

reanalysis

(9.a)

calization

les femmes ex-ceptées ...

sur le bureaude Pierre

pendant DP

à côté de DP

avant DP

après DP

durant DP

excepté DP

e.g.

type 1:lexical heads

au lieu de DP

avec DP

concernant DP

en face de DP

alentour de DP

sur DP

grammati-

Relational Elements in French 31

(31)

As a whole, this paper should hopefully have demonstrated that both the gen-erative categorial system, which makes use of distinctive categorial features,and the Minimalist idea of a strictly feature-driven syntax can be integrated intothe concept of grammaticalization theory in a fruitful way, thereby offering anadequate account for considerable aspects of both synchronic variation anddiachronic change.

References

Baker, Mark C. (1988): Incorporation. A Theory of Grammatical Function Changing. Chi-cago: The University of Chicago Press.

Baker, Mark C. (1996): The Polysynthesis Parameter. Oxford: Oxford University Press., (1974):

Blake, Barry J. (1994): Case. Cambridge: CUP.Brøndal, Viggo (1950) : Théorie des prépositions. Kopenhagen: Munksgaard.Burzio, Luigi (1986): Italian Syntax: A Government-Binding Approach. Dordrecht: Reidel.Carstens, Vicki (2000): “Concord in Minimalist Theory.” Linguistic Inquiry 31, 319-355.Chomsky, Noam (1995): “Categories and Transformations.” The Minimalist Program. Cam-

bridge, MA: MIT Press, 219-394.Chomsky, Noam (2000): “Minimalist Inquiries: the Framework.” Step by Step. Martin, Roger

et al. (eds.). Cambridge, MA: MIT Press, 89-155.Culicover, Peter W. (1999): Syntactic Nuts. Hard Cases, Syntactic Theory, and Language

Acquisition. Oxford: Oxford University Press.Fleischer, Wolfgang (1997): Phraseologie der deutschen Gegenwartssprache. 2., durchgesehe-

ne und ergänzte Auflage. Tübingen: Niemeyer.Friedemann, Marc-Ariel (1997): Sujets syntaxiques. Positions, inversions et pro. Bern: Lang.Friedemann, Marc-Ariel & Tal Siloni (1997). “Agrobject is not Agrparticiple.” The Linguistic

Review 14, 69-96.Gabriel, Christoph (in press): “Präpositionen, referenzielle Rollen und Grammatikalisierung.”

Semantische Rollen in der Romania. Hummel, Martin & Rolf Kailuweit (eds.). Tübingen:Narr.

Grevisse, Maurice (1993): Le bon usage. Grammaire française. Paris-Gembloux: Duculot.Grimshaw, Jane (1991): “Extended Projection.” Ms., Waltham, MA: Brandeis University.Guasti, Maria Terea & Luigi Rizzi (2000): “Agreement and Tense as Distinct Syntactic Posi-

tions: Evidence from Acquisition.” Ms., Siena: University of Siena [38 p].Hopper, Paul J. & Elizabeth C. Traugott (1993): Grammaticalization. Cambridge: CUP.

A[+N; +V]

V[–N; +V]

N[+N; –V]

P[–N; –V]

lexical categories

functional categories

lexical categories

functional categories

32 Christoph Gabriel

Johanson, Lars & Éva Ágnes Csató (1998): “Turkish.” The Turkic Languages. Johanson, Lars& Éva Ágnes Csató (eds.). London: Routledge, 248-260.

Jones, Michael A. (1996): Foundations of French Syntax. Cambridge: CUP.Kailuweit, Rolf (2001): “Lexeme, Kasusmarker, Relatoren? Überlegungen zu den spanischen

Präpositionen unter dem Gesichtspunkt der Grammatikalisierung.” Grammatikalisierungin den iberoromanischen Sprachen. Barbara Schäfer-Prieß et al. (eds.). Wilhelmsfeld:Egert, 33-62.

Kayne, Richard S. (1994): The Antisymmetry of Syntax. Cambridge, MA: MIT Press.Kornfilt, Jaklin (1997): Turkish. London: Routledge.Laenzlinger, Christopher (2000): “French Adjective Ordering: Perspectives on DP-internal

Movement Types.” Generative Grammar in Geneva 1, 55-104.Lang, Jürgen (1991): Die französischen Präpositionen. Funktion und Bedeutung. Heidelberg:

Winter.Langacker, Ronald W. (1977): “Syntactic Reanalysis.” Mechanisms of Syntactic Change. Li,

Charles N. (ed.). Austin: University of Texas Press, 57-139.Lehmann, Christian (1982): Thoughts on Grammaticalization. München: Lincom (21995).Lehmann, Christian (1985): “The Role of Grammaticalization in Linguistic Typology.” Lan-

guage Invariants and Mental Operations: International Interdisciplinary Conference heldat Gummersbach/Cologne, Germany, sept 18-23, 1983. Seiler, Hansjakob & Gunter Brett-schneider (eds.). Tübingen: Niemeyer, 41-52.

Longobardi, Giuseppe (1994): “Reference and Proper Names.” Linguistic Inquiry 25, 609-665.

Mensching, Guido (1992): Einführung in die sardische Sprache. Bonn: Romanistischer Ver-lag.

Müller-Lancé, Johannes (1994): Absolute Konstruktionen vom Altlatein bis zum Neufranzö-sischen. Tübingen: Narr.

Napoli, Donna Jo & Joel Nevis (1987): “Inflected Prepositions in Italian.” Phonology Year-book 4, 195-209.

Pollock, Jean-Yves (1989): “Verb Movement, UG, and the Structure of IP.” Linguistic Inquiry20, 365-424.

Pottier, Bernard (1962) : Systématique des éléments de relation. Étude de morphosyntaxestructurale romane. Paris: Klincksieck.

Radford, Andrew (1997): Syntactic Theory and the Structure of English. A Minimalist Ap-proach. Cambridge: Cambridge University Press.

Rauh, Gisa (1998): “Zur Struktur von Präpositionalphrasen im Englischen.” Zeitschrift fürSprachwissenschaft 15, 178-230.

Roberts, Ian (1993a): Verbs and Diachronic Syntax. A Comparative History of English andFrench. Dordrecht: Kluwer.

Roberts, Ian (1993b): “A Formal Account of Grammaticalisation in the History of RomanceFutures.” Folia Linguistica Historica 13, 219-258.

Roberts, Ian (2001): “Language Change and Learnability.” Language Acquisition and Learn-ability. Bertolo, Stefano (ed.). Cambridge: Cambridge University Press, 81-125.

Roberts, Ian & Anna Roussou (1999). “A Formal Approach to ‘Grammaticalization’.” Lingui-stics 37, 1011-1041.

Schönig, Claus (1998): “Azerbaijanian.” The Turkic Languages. Johanson, Lars & Éva ÁgnesCsató (eds.). London: Routledge, 248-260.

Schroeder, Christoph (1999): The Turkish Nominal Phrase in Spoken Dialogue. Wiesbaden:Harrassowitz.

Siller-Runggaldier, Heidi (1996): Das Objektoid. Eine neue syntaktisch-funktionale Kategorie,aufgezeigt anhand des Italienischen. Wilhelmsfeld: Egert.

Relational Elements in French 33

Spang-Hanssen, Ebbe (1963): Les prépositions incolores du français moderne. Kopenhagen:Gads.

Torrego, Esther (1998): The Dependencies of Objects. Cambridge, MA: MIT Press.Tremblay, Mireille (1999). “Du statut des prépositions dans la grammaire.” Revue québécoise

de linguistique 27, 167-183.Vincent, Nigel (1997): “Prepositions.” The Dialects of Italy. Maiden, Martin (ed.). London:

Routledge, 208-213.Wartburg, Walther von & Paul Zumthor (1958) : Précis de syntaxe du français contemporain.

Bern: Francke.Weinrich, Harald (1982): Textgrammatik der französischen Sprache. Stuttgart: Klett.Zwarts, Joost (1992) : X’-Syntax – X’-Semantics. On the Interpretation of Functional and

Lexical Heads. Utrecht: LEd.Zubizarreta, Maria Luisa (1998): Prosody, Focus, and Word Order. Cambridge, MA: MIT

Press.

Osnabrück Christoph Gabriel

Universität Osnabrück, FB 7, Sprach- und Literaturwissenschaft, Neuer Graben 40, D-49069Osnabrück. E-mail: [email protected].