Uses of the passé composé in Old French: evolution or revolution ?

Comp. by: SatchitananthaSivam Stage : Proof ChapterID: 0002544163 Date:15/6/15Time:19:20:12 Filepath://ppdys1122/BgPr/OUP_CAP/IN/Process/0002544163.3dDictionary : OUP_UKdictionary 178

9

The passé composé in Old Frenchand Modern French: Evolutionor revolution?

PATRICK CAUDAL

9.1 Introduction

Although the historical development of the passé composé (PC) during the OldFrench (OF) and Middle French (Mid.Fr.) periods has been the object of an impres-sive body of work, whether the OF PC was already endowed with the sort ofperfective viewpoint tense meaning it has in modern French (cf. Squartini andBertinetto’s 2000 “aoristic drift”) still remains a much debated question. Morespecifically, I intend to determine in this chapter whether we can regard examples(2)–(3) as conveying a sequence of events (SOE, henceforth) reading akin tothat of (1).

(1) Jean est parti, puis Marie est arrivée. (Mod.Fr.)Jean be-pr.3sg. leave-pp, then Marie be-pr.3sg. leave-pp.‘Jean left, then Marie arrived.’

(2) Tant li preièrent li meillur Sarrazin / Qu’el’ faldestoel s’est Marsilies asis.(Rolant, 451–52) (OF)

So him beg-ps.3pl the best Saracens / That on.the throne refl-be-pr.3sg.Marsile sit-pp.‘The best Saracens begged Marsile so much / That upon his throne he sat down.’

(3) A-dont ont lez napes leveez / Et apréz ont leurs mains laveez.(Anjou, 2625, in Martin 1971: 169) (OF)

After.that have-pr.3pl. the napkins lift-pp, and then have-pr.3pl. their handswash-pp ‘ . . . After that they lifted the napkins, and then they washed theirhands.’

Sentence and Discourse. First edition. Jacqueline GuéronThis chapter © Patrick Caudal 2015. Published 2015 by Oxford University Press

OUP UNCORRECTED PROOF – FIRST PROOF, 15/6/2015, SPi

Comp. by: SatchitananthaSivam Stage : Proof ChapterID: 0002544163 Date:15/6/15Time:19:20:13 Filepath://ppdys1122/BgPr/OUP_CAP/IN/Process/0002544163.3dDictionary : OUP_UKdictionary 179

Indeed, the rise of SOE interpretations marks an important step forward on theevolutionary path of perfects towards perfective viewpoint tenses (see e.g. Bybee et al.1994; Squartini and Bertinetto 2000). SOE uses are thus known not to be licensedwith “canonical” perfects, such as the present perfect in standard varieties of Britishor American English (cf. e.g. Portner 2003: 502), as opposed to more or less “perfec-tivized” perfects such as, for example, the Australian present perfect (Ritz 2007;Caudal and Ritz 2012) or the modern German Perfekt). It is therefore not surprisingthat SOE uses have become a crucial test for identifying bona fide perfects in thetypological literature (cf. the “perfect questionnaire” in Dahl 2000: 800–9).

Existing accounts of the OF PC fall into two broad classes, depending on the waythey analyze examples such as (2)–(3) in relation to (1). According to, for example,Wilmet (1970), Martin (1971: 257, 394), and Schøsler (1973), the Mod.Fr. PC in (1) is infact fundamentally different from the OF PC in (2)–(3); the PC in OF was not aperfective viewpoint tense (in the sense of Smith 1991) and acquired perfectivefeatures (i.e. the ability to describe events in the past, rather than merely entailthem) only at a later period in its history. I will call those accounts the “lateperfectivization accounts.” Another, smaller set of works claim that (1) and (2)–(3)are alike at least in some sense, and that the PC had already undergone some kind ofperfectivization, either semantic (Foulet 1920) or pragmatic (Caudal and Roussarie2006; Detges 2006) as soon as the OF period. I will call these the “early perfectiviza-tion accounts.”

I will propose in this chapter a novel analysis at the semantics–pragmaticsinterface, combining insights from both competing hypotheses. Following a kind ofapproach which has become prevalent in work on French tenses (cf. Fleischman 1990;de Saussure 2003; Borillo et al. 2004; Caudal and Roussarie 2005; de Swart 2007;among others), I will highlight the importance of discourse structural parameters inthe evolution of the passé composé towards novel, perfective meanings, at a timewhen it (still) exhibited considerable semantic and morphosyntactic conservatism,that is, when it was still relatively close to the Latin construction from which itoriginated, and was not yet a morphosyntactically well-established tense form. Thisrelative morphosyntactic indeterminacy will constitute a crucial precondition for theearly SOE patterns arising in discourse.

At the same time, a considerable part of the data used to substantiate my analysiswill involve clauses expressing temporal simultaneity (quand ‘when’ clauses inparticular); as we will see, it is quite likely that these recurrent syntactic patterns,combined with independent discourse-level mechanisms, played an instrumentalrole in the evolution of the passé composé—this, I will claim, illustrates a situationwhere discursive and syntactic factors become enmeshed in complex ways within aparticular grammatical evolution.

Including the present chapter within a section notably aiming at disentangling therelations between discourse structure qua discourse relations and syntax makes

OUP UNCORRECTED PROOF – FIRST PROOF, 15/6/2015, SPi

The passé composé in Old French and Modern French 179

Comp. by: SatchitananthaSivam Stage : Proof ChapterID: 0002544163 Date:15/6/15Time:19:20:13 Filepath://ppdys1122/BgPr/OUP_CAP/IN/Process/0002544163.3dDictionary : OUP_UKdictionary 180

particular sense in this respect. By showing that a broad range of inter-clausalrelations (some syntactically encoded, others relying on mere discourse relations)jointly played a significant role in the rise of “perfective” uses of the PC in OF, I willalso show how structural convergences cutting across the syntax/discourse distinc-tion can contribute to grammatical evolution—and that discourse and syntax–semantics should be regarded as intertwined in grammar. It is also crucial to notethat many inter-clausal structures were indeed at a turning point in the history of thelanguage (see e.g. Fleischman 1990: 70): the boundaries between coordination/juxta-position (where inter-clausal relations are often determined in discourse) and syn-tactic subordination were somewhat unclear (cf. e.g. apparently juxtaposed si/seclauses (Glikman 2009) or frequent cases of asyndetic subordination in OF). More-over, markers introducing subordinate causal/temporal clauses are often treated incontemporary formal discourse semantics on a par with logical connectives intro-ducing matrix clauses with similar semantic functions1 (and indeed, specific markerscan pertain to both classes: quand or lorsque ‘when/as’ in French have thus beenanalyzed as introducing specific discourse relations between the dependent andmatrix clauses, cf. Asher et al. (2007).2 These facts suggest that syntactic and dis-course structures would greatly benefit from being studied within a unifiedperspective.

Briefly, my argument will proceed as follows: Section 9.2 will present the kind ofevidence generally put forth in favor of the late perfectivization hypothesis, and itsbasic tenets; Section 9.3 will confront this hypothesis with new data originating froma corpus study, and crucially involving temporal biclausal structures. While the studyconfirms that the PC did not have full (semantic) perfective force in OF (as argued bytenants of the late perfectivization hypothesis), it also suggests that perfectivization ofthe PC was already under way at least in some pragmatic sense—in other words, thatwhile the OF PC was morphosyntactically and semantically conservative (i.e. waspartly a perfect, partly the more ancient present resultative it was derived from, inline with the late perfectivization hypothesis), it was already pragmatically innovative(i.e. its (conventional) pragmatics had already developed some perfective features, aspredicted by the early perfectivization hypothesis). Section 9.4 will attempt to giveflesh to the solution outlined in Section 9.3, combining a non-perfectivized, conser-vative semantics, with an innovative, perfectivized pragmatics. I will specifically

1 And indeed, some of these structures appear to involve some form of “mismatch” between syntax andsemantics, e.g. syntactically coordinated OF clauses involving semantic subordination, or syntacticallysubordinated OF clauses involving semantic coordination; cf. Yuasa and Sadock (2002).

2 Discourse semantic analyses for intra-sentential syntactic phenomena are increasingly common; theyhave been proposed in a number of frameworks for e.g. temporal adverbials, non-finite verbal appositions(e.g. gerund clauses), relative clauses, NP appositions (see e.g. Asher et al. 2007; Prévot et al. (2009; Vergez-Couret and Adam 2012). Relating such approaches to complementary syntactic approaches tacklingdiscourse-level phenomena thus makes perfect sense.

OUP UNCORRECTED PROOF – FIRST PROOF, 15/6/2015, SPi

180 Patrick Caudal

Comp. by: SatchitananthaSivam Stage : Proof ChapterID: 0002544163 Date:15/6/15Time:19:20:13 Filepath://ppdys1122/BgPr/OUP_CAP/IN/Process/0002544163.3dDictionary : OUP_UKdictionary 181

argue that the burden of the interpretation of the OF PC should be shared between itscompositional semantic content, and a variety of discourse structuring parameters(some syntactically marked, e.g. through temporal subordination, others purelydiscourse contextual), which contribute to determining its specific interpretation.These discourse structuring parameters will be here understood as discourse rela-tions, in the sense of the Segmented Discourse Representation Theory, see Asher andLascarides (2003), and Asher, this volume.3

9.2 Assessing the late perfectivization hypothesis:basic tenets and evidence

9.2.1 The late perfectivization hypothesis: the OF PC as an intermediateform between a resultative present and a full perfect

Although there are countless nuances in each variant of this analysis, I will overlookthose variations, and review a combination of their common core assumptions. Themain idea behind this hypothesis is that at least as late as the thirteenth or fourteenthcenturies, the PC was still partly akin to the esse ‘be’ versus habere ‘have’ + pastparticiple (PP) stative resultative constructions in which it originated, that is,respectively the Latin analytical perfect passive factum est ‘it has been done/become’for PCs with an estre ‘be’ auxiliary, and the late Latin habeo litteras scriptas ‘I have(got) written letters’ (where, crucially, the identification of the agent involved—ifany—is a matter of non-linguistic inferences), for PCs involving an aveir ‘have’auxiliary.4 Note that similar constructions exist in languages where the perfectbecame a grammaticalized tense form, as shown in (4)–(6) (Carey 1994: 104;Detges 2006).

(4) Son index est pointé . . . vers ce couple enfant-mère. (M. Thomas, 2004, Trésorsde l’art sacré dans les hautes vallées de Maurienne, p. 174) (Mod.Fr.)His forefinger be-pr.3sg point-pp towards this couple child–mother.(‘His forefinger is pointed towards this mother–child couple’.)

(5) [L’animal] a la tête dressée, les cornes développées passant derrière les oreilles;le cou est finement gravé d’un collier. (Pierre Bergé & associates, auctionbrochure, batch #67, 14/12/09) (Mod.Fr.)The-animal have-pr.3sg. the head raise-pp, the horns developed go-prp behindthe ears; the neck be-pr.3sg finely engrave-pp of-a collar.

3 Note that in contrast to the present account and Caudal and Roussarie (2006), Detges’s (2006)pragmatic approach relies on non-discourse structural parameters.

4 Unlike e.g. Detges (2006) or Squartini and Bertinetto (2000), I will neither follow nor adapt Harris’s(1982) theory of the evolution of Romance perfects, for reasons I cannot discuss here.

OUP UNCORRECTED PROOF – FIRST PROOF, 15/6/2015, SPi

The passé composé in Old French and Modern French 181

Comp. by: SatchitananthaSivam Stage : Proof ChapterID: 0002544163 Date:15/6/15Time:19:20:13 Filepath://ppdys1122/BgPr/OUP_CAP/IN/Process/0002544163.3dDictionary : OUP_UKdictionary 182

‘The animal has its head raised; its horns are well-developed, going behind itsears; its neck is finely engraved with a collar.’

(6) In this pose, Buddha has his hand raised near his shoulder. (Google) (Mod.Eng.)

According to widely accepted views on the matter (see e.g. Squartini and Bertinetto2000: 404ff.), those constructions came to receive a proto-perfect reading as follows.At first, as their PP component denoted a result state, and implicated that somecausing, prior (therefore, past) event occurred, bringing about the result state, couldhold in some contexts; subjects of the implicated past event and subject of the resultstate could refer to distinct or identical entities, depending on context (see example(5) above). In time, the past event implicature gradually strengthened, eventuallybecoming semanticized as an entailment that a past causing event occurred; it alsobecame compulsory for subject referents of the result state and entailed past events tobe identical. Additionally, agentive subjects and in fact, transitivity, became largelyrestricted to “have” Romance perfects: thus OF verbs forming their PC with the estre‘be’ auxiliary are known to be unaccusative, by and large (Dufresnes and Dupuis2010), while transitive, control OF verbs require the aveir ‘have’ auxiliary. I will arguehere that the convergence of these originally separate ‘be’ versus ‘have’ resultativeconstructions into a single analytical perfect in OF (formed with either the estre ‘be’or aveir ‘have’ auxiliary) militates for at least overlapping (if not very similar)grammaticalization paths.5

If we accept the idea that the PC in OF was an only partially grammaticalizedtense, it would seem natural to regard the semantic contribution of its morphosyn-tactic components as still transparent/compositional to some extent. Two distinctmorphological elements, and two corresponding components of meaning shouldthus be considered: a présent (PR) inflexion, with a present temporal content and anunderspecified aspectual content (the PR contributing either a imperfective orperfective viewpoint, depending on the discourse context),6 and a PP component,entailing a past, causing event and describing a result state. This idea, combined withthe well-known observation that being aspectually underspecified, the OF PR freelyadmitted SOE readings (cf. Foulet (1920: 280–1)), can lead to the hypothesis thatexamples such as (2)–(3) were instances of a special kind of SOE uses of the presentcomponent of meaning of the PC, conveying a perfective resultative meaning. Thus,

5 In sharp contrast to the present account and, e.g., Squartini and Bertinetto (2000: 405), Detges (2006:61ff.) claims that aveir ‘have’ versus estre ‘be’ PCs were still two independent constructions in OF: accordingto him, only aveir PCs entailed past causal events, whereas estre PCs were present resultatives, not perfectsat all. Accordingly, Detges takes the evolution of Romance perfects (including the OF PC) to revolvearound agentivity. I will come back to Detges’s analysis later.

6 Here I slightly modify existing accounts, most of which argue that the OF présent contributed anunambiguously imperfective aspectual viewpoint (cf. e.g. Wilmet 1970: 309; Martin andWilmet 1980). Sucha hypothesis would render the late perfectivization hypothesis impossible to defend, however, cf. Section 3.1below.

OUP UNCORRECTED PROOF – FIRST PROOF, 15/6/2015, SPi

182 Patrick Caudal

Comp. by: SatchitananthaSivam Stage : Proof ChapterID: 0002544163 Date:15/6/15Time:19:20:13 Filepath://ppdys1122/BgPr/OUP_CAP/IN/Process/0002544163.3dDictionary : OUP_UKdictionary 183

the correct interpretation of the PC clause Marsilie s’est asis in (2) would be roughlyparaphrased as ‘Marsile becomes seated’.

To make a synchronic comparison, this analysis is tantamount to saying that(2)–(3) should be interpreted like (7) in Mod.Fr. and (8) in Mod.Eng.: the presentcomponent of those constructions receives an SOE perfective interpretation, therebyimposing a perfective interpretation of the result state contributed by the PP (the PRauxiliary contributes an aspectual viewpoint function shifting that result state into achange-of-state).

(7) (Et soudain,) l’animal a la tête dressée. (Mod.Fr)(And suddenly,) the-animal have-pr.3sg. the head raise-pp.‘(And suddenly,) the animal has (got) its head raised’.

(8) A cold, blustery day in Washington, D.C . . . . People are heads-down huddles asthey walk into the wind . . . . Suddenly, the wind is gone.

(Hartford Courant, 11 March 2001) (Mod.Eng.)

If we adopt this explanation wholesale, an important corollary holds: if indeed the OFPC has retained earlier proto-perfect semantic features, then we should expect (2)–(3)not to give us any information about the controller of the past events involved. Forinstance, in (7), it is unclear by whom the animal’s head was raised (i.e. in the absenceof specific information, we cannot be certain that it was the animal itself), and in (4),we do not know who moved the subject’s finger. But it is quite clear in (2), however,that Marsile is the agent of the event described (cf. the reflexive s’)—and similarobservations hold about all the OF examples cited in this chapter. This suggests thatthe OF PC could have retained only some (not all) features of resultative construc-tions such as (7)–(8), and behaved like a “true,” grammaticalized perfect as far as,for example, semantic role information was concerned (see Detges 2006 for furtherevidence).

9.2.2 Arguments for the late perfectivization hypothesis

We will now discuss data put forth in order to substantiate the claim that the OFPC was a semantically conservative form, still exhibiting proto-perfect features(cf. Wilmet 1970: 319; Schøsler 1973: 106)—though not all, as we have seen.

9.2.2.1 Morphosyntactic evidence for the persistence of proto-perfect features in thePC As noted in e.g. Schøsler (2004: 521), two “conservative” morphosyntacticfeatures of the PC in OF appear to give credit to the late perfectivization hypothesis,namely: (i) agreement of the PP with the object remained possible in OF and (ii)linguistic material (both complements and modifiers) could be inserted between theauxiliary and the past participle. The strength of these arguments should be relativ-ized, however. The partial persistence of a set α of morphosyntactic features does not

OUP UNCORRECTED PROOF – FIRST PROOF, 15/6/2015, SPi

The passé composé in Old French and Modern French 183

Comp. by: SatchitananthaSivam Stage : Proof ChapterID: 0002544163 Date:15/6/15Time:19:20:13 Filepath://ppdys1122/BgPr/OUP_CAP/IN/Process/0002544163.3dDictionary : OUP_UKdictionary 184

necessarily warrant the persistence of the meaning formerly associated with α. And asa matter of fact, these morphosyntactic features persisted well beyond the OF period,up to the sixteenth–seventeenth centuries, at a time when the semantic perfectiviza-tion of PC was already under way (Caudal and Vetters 2007). So they cannot beregarded as reliable semantic indicators (Wilmet 1970: 239).

9.2.2.2 Analogy with narrative uses of the passé antérieur (PA) In addition tomorphological evidence, semantic evidence was also put forth in favor of the lateperfectivization hypothesis. Wilmet (1970: 322) thus claims that SOE uses of the PC inOF should be viewed as “inchoative” resultative presents, because they are semanticallysimilar to uses of the passé antérieur (PA) exemplified in (9), which he takes to express“the rapid completion of a past action.” The PA being arguably a resultative form of thepassé simple, the meaning of (9) can be viewed as an inchoative perfective resultative:7

(9) Compère le Renard . . . retint à dîner commère la Cigogne . . . . [Le] brouet futpar lui servi sur une assiette / . . . Et le drôle eut lapé le tout en un moment.

(La Fontaine, Fables, I, 18) (OF)Mister the Fox hold-ps.3sg for dinner Mistress the Stork. The soup be-ps.3sg byhim serve-pp on a plate. And the rascal have-ps.3sg lap-pp the whole in onemoment.‘Master Fox invited Mistress Stork for dinner. He served soup on a plate. And inan instant, the rascal had lapped it all.’

Although the “quickness” of the lapping event in (9) is in fact due to the temporalmodifier en un moment, not to the PA itself, I believe that the PA/PC analogy hassome substance to it. I will show in what follows that the PA and the PC in OFnarratives exhibited some amount of distributional convergence, and therefore someinterpretative similarity.

9.2.2.3 The PC and aspectuo-temporal modifiers: convergences with the presentConsider now the fact that the late perfectivization hypothesis regards the OF PCas a resultative variant of the PR. A good case in favor of the late perfectivizationhypothesis could therefore be made by comparing the respective compatibility and

7 If indeed the PR is a (near) temporal equivalent of the PS in OF (see e.g. Detges 2006), and since thePA and the PC are respectively aspectual, resultative variants of the PS and PR, then the PC must be a(near) temporal equivalent of the PA. This fact is demonstrated by e.g. the possibility of freely juxtaposing,subordinating, and coordinating clauses in the PR and PS, or PC and PA, (10)—a freedom which modernFrench has largely lost:

(i) . . . il cuide soubitement enragier, et fut plus de demie heure sans soi relever de la terre ou il chey . . .(Prose Cligès, p. 12)He feel-pr.3sg suddenly be.infuriated-inf, and be-ps.3sg more of half hour without refl rise of theearth where he fall-ps.3sg‘He felt suddenly mad with rage, and for more than half an hour, could not rise from the place wherehe had fallen . . . ’

OUP UNCORRECTED PROOF – FIRST PROOF, 15/6/2015, SPi

184 Patrick Caudal

Comp. by: SatchitananthaSivam Stage : Proof ChapterID: 0002544163 Date:15/6/15Time:19:20:14 Filepath://ppdys1122/BgPr/OUP_CAP/IN/Process/0002544163.3dDictionary : OUP_UKdictionary 185

interpretation of these two tenses with a large spectrum of aspectuo-temporalmodifiers, such as the following (cf. e.g. Martin 1971; Treikelder 2006):

• Deictic temporal expressions: neither the PC nor the PR license e.g. hyer‘yesterday’; use of this item is restricted to the PS in OF and MF (cf. Caudaland Vetters 2007). Some other past deictic temporal expressions which corres-pond to some extended now interval (cf. e.g. “this week”) are compatible withboth the PR and the PC, cf. (10).

• Degree aspectuo-temporal expressions forcing an inchoative reading: tantost,tost ‘soon, quickly’, atant ‘suddenly’: both the historical PR and the PC arefound with such expressions.

• Anaphoric temporal expressions such as l’endemain ‘the following day’: boththe historical PR and the PC are felicitous with such modifiers. This suggests asimilar “perfectivizing” ability, as well as the ability to anchor an event within aspeech time interval shifted to the past.8

• Expressions of duration, including degree expressions (e.g. longuement ‘for along time’): the historical PR and the PC are both compatible with suchexpressions. One might object, however, that they could actually scope underthe PC, as is possible with the present perfect in e.g., British English when itreceives a non-continuative, experiential reading (cf. “Jean has demonstrated in1968”). Example (11) shows an equivalent non-continuative reading of the PC inOF. Here, the interval modified by par plusieurs années ‘for several years’ cannotbe continuative; that is, it cannot overlap with the speech time interval as there ispeace, not war, between the French and the Flemish at the time the chroniclewas written. Similarly, expressions referring to precise past events or datesnormally cannot combine with a perfect unless they scope under that tense;thus the British perfect accepts them as long as the intended reading is experi-ential. Example (12) shows an experiential PC in OF in which the temporalexpression might well scope under tense (see Detges 2006 for further examplesof experiential readings of the OF PC).

• Definite time expressions such as dates: neither the PC nor the PR appear tolicense such modifiers in OF, except for an experiential/non-narrative reading inthe case of the PC, as in (12); in contrast, both the historical PR and the PCaccept such expressions in modern French narrative.

(10) Trestout le bestial est peri / cest yver par la grant froidure.(Pathelin, 244, in Martin 1971: 254) (OF)

Everywhere the cattle be-pr.3sg die-pp this winter by the great cold.‘Cattle has died everywhere because of last winter’s harsh cold’.

8 See e.g. Zucchi (2006).

OUP UNCORRECTED PROOF – FIRST PROOF, 15/6/2015, SPi

The passé composé in Old French and Modern French 185

Comp. by: SatchitananthaSivam Stage : Proof ChapterID: 0002544163 Date:15/6/15Time:19:20:14 Filepath://ppdys1122/BgPr/OUP_CAP/IN/Process/0002544163.3dDictionary : OUP_UKdictionary 186

(11) . . . et encores ont les François bonne paix et ferme aux Flamens, qui moult lesont embesogniez . . . par plusieurs années.

(Froissart, Livre III, }75—XII, p. 241, 1. 16–18, ed. L. Mirot) (OF)and still have-pr.3pl the French good peace and firm to.the.pl Flemish, whomuch them have-pr.3pl keep_busy-pp.‘and the French still enjoy a good and lasting peace with the Flemish, who havekept them very busy [making war] . . . for several years’.

(12) Car nous le avons juré et promis . . . au jour de vostre coronnement.(Froissart, Book III, }38—XII, p. 156, l. 19–20, ibid.; in Martin 1971: 196) (OF)For we it have-pr.1pl swear-pp and promise-pp on.the day of your crowning.‘For we have sworn and promised [to do so] on the day you were crowned’.

These facts point to convergent aspectuo-temporal properties, especially perfectiveforce, in the historical PR and the PC, therefore substantiating the late perfectiviza-tion hypothesis. Note that both tenses then failed to combine with definite past timeadverbials; this suggests that their semantic perfective force was only relative, unliketheir modern counterparts (Caudal and Vetters 2007).9

9.2.2.4 Completing the late perfectivization picture: how did the OF PC lose its proto-perfect semantics? If we admit that the OF PC was still a resultative present, andthat the PC (i) became a full perfect at some later stage (ii) and from there, acquiredfull perfective semantic capability, then we still need to account for the “death” of theOF PC as a resultative present, and its subsequent semantic perfectivization. Andindeed, it has been claimed in, for example, Martin (1971) and Schøsler, (1973, 2004)that a significant drop in the frequency of SOE uses of the PC at the turn of thefourteenth century could have caused such a transformation. As the OF PC wasessentially a PR form, its SOE uses would have become scarce too. The PC wouldhave then lost its connection to the PR, and embarked on a solo career as a fullperfect, completing its semantic perfectivization at the end of the Classical Frenchperiod (Caudal and Vetters 2007).

Yet, contrary to the above claim, it is not the case that SOE uses of either the PC orthe PR vanished during the fourteenth century (see Section 9.3.3). In addition, it iseasy to show that calculations of the frequency of narrative PR are highly sensitive totextual genres. If we take biographies or historical works, for example, frequency ofnarrative PRs soars; but it drops if we pick up, say, newspapers articles, or argumen-tative prose (cf. Martin, 1971: 378). I must therefore set this argument aside asspeculative at best.

9 The fact that the PC (cf. Treikelder 2006: 177) was seldom used in OF to describe complex sequences ofevents also suggests a relatively limited perfective ability.

OUP UNCORRECTED PROOF – FIRST PROOF, 15/6/2015, SPi

186 Patrick Caudal

Comp. by: SatchitananthaSivam Stage : Proof ChapterID: 0002544163 Date:15/6/15Time:19:20:14 Filepath://ppdys1122/BgPr/OUP_CAP/IN/Process/0002544163.3dDictionary : OUP_UKdictionary 187

All in all, we have only proven the late perfectivization hypothesis to be partlyaccurate: the OF PC appears to behave like a resultative PR in some but not allsemantic respects, and the precise nature and causes of its evolution remain some-what mysterious. The quest for evidence settling the early versus late perfectivizationdispute must continue.

9.3 Re-assessing the perfectivization of the PC in OF:new data and analysis

Bearing this in mind, I conducted a small corpus experiment on twoMedieval Frenchtexts, and classified SOE uses of the PC, as well as of other tenses, particularly the PA,the PS, and the PR. I then analyzed relevant examples of the PC in terms of discoursestructural parameters (i.e. discourse relations)10 within the formal framework ofSegmented Discourse Representation Theory (SDRT), thus making it possible toidentify possible effects of discourse structural parameters.

The OF corpus examined comprised roughly 100,000 words, half from the Chev-alier de la Charrette (a twelfth-century verse narrative, but not a chanson de geste),and half from the Queste del Saint Graal (a thirteenth-century roman, i.e. Frenchprose narrative). I have systematically divided PC uses within narrative (as opposedto dialogue) passages into two broad aspectual groups: (i) SOE perfective-like usesand (ii) imperfective-like uses. Tables summarizing the most significant results/figures are given in the Appendix to this chapter.

Although there is a notable discrepancy between the Chevalier and the Queste interms of the frequency of SOE uses of the PC, it turns out that while comparativelyrare, since only around 1 per cent of the verbal form was used to convey sequences ofevents,11 such uses were nevertheless sufficiently well represented so as not to beconsidered as grammatically marginal.

9.3.1 The present component of the PC in narrative uses:imperfective or perfective viewpoint?

As we have seen above, the late perfectivization account is based on the assumptionthat the present component of the PC is responsible for this tense’s SOE, perfective-like readings. Somewhat paradoxically,many authors (fromWilmet 1970 to Treikelder2006, cf. note 7) claim that it should contribute an imperfective viewpoint reading.

10 I assume the reader to be familiar with the SDRT framework described in Asher and Lascarides(2003) and Asher, this volume, and in particular its inventory of discourse relations. These relations link anew discourse segment β (typically an utterance) to the current discourse context via an attachmentsite α (typically a previous utterance) in interaction with aspectuo-temporal information.

11 In comparison, SOE uses of the présent are at least ten times as numerous within both the Queste andthe Chevalier (with resp. 10.4% and 10.7% of verb occurrences). In both texts the PS remains the dominantform to convey SOE readings.

OUP UNCORRECTED PROOF – FIRST PROOF, 15/6/2015, SPi

The passé composé in Old French and Modern French 187

Comp. by: SatchitananthaSivam Stage : Proof ChapterID: 0002544163 Date:15/6/15Time:19:20:14 Filepath://ppdys1122/BgPr/OUP_CAP/IN/Process/0002544163.3dDictionary : OUP_UKdictionary 188

Such an analysis predicts that in (13), the (present) result state e5 should overlap with e4(i.e. Background is licensed)—but in the absence of any marker providing temporalanchoring/ordering information for e5 in (13), this is impossible (Background cannotbe established). Therefore, the PCmust be inherently capable of receiving a perfectiveviewpoint reading (and Narrationmust be established between the relevant discoursesegments). I will treat the PR as aspectually underspecified, that is as contributingeither a perfective or an imperfective viewpoint. In the case of SOE readings of the PC,I will claim that its PR component must receive a perfective viewpoint interpretation.

(13) Si l’ala querre (e4); . . . / A une fenestre l’a mise (e

5) (Chevalier, 3583–6) (OF)

And her-go-ps.3sg fetch-inf. To a window her have-pr.3sg put-pp.‘He [king Bademaguz] went and fetched her [Guinevere] (α), . . . sat her downat a window (β).’

9.3.2 SOE uses of the OF PC: weak (inchoative) or strong perfective uses?

There are, in principle, two ways the PC could construe a perfective reading. It couldbe triggered either because the present component of the OF PC forces an inchoativeinterpretation of the result state, so that the PP result state is bounded on its initial part,but open-ended on its final part; or because the present component forces a globalperfective reading of the entire result state denoted by the PP, which is then boundedboth on its initial and final part. I will call the first reading the weak perfective orinchoative reading, and the second the strong perfective reading. Examples involvingduration adverbials such as (14) are good candidates for the strong perfective reading:

(14) Ses amis a esté cinc ans [narrative context] (Renaut de Beaujeu, Le bel inconnu(late twelfht–early thirteenth century), in Treikelder 2006: 75)Her-subj friend-subj have.pr.3sg been five years.‘He was her friend for five years . . . ’

Yet one could argue that the durative adverbial actually scopes under the aspectualcontent of the present component of the PC. In this case, (14) would not express apresent result state of “having been friends with someone” which has been going onfor five years, but rather the present result state of “having been friends with someonefor five years,” with the friendship relation being over at speech time and the durativemodifier taking scope only over the PP result state (i.e. this is not an instance of so-called “continuative” perfect; cf. Detges 2006: 55, 64 for further examples). I willtherefore put aside the question of whether such “strong” perfective resultativereadings of the PC are attested in OF.

The existence of inchoative resultative readings is, in contrast, rather easy toestablish. Recall that Wilmet (1970) compared SOE uses of the PC and of the passéantérieur (PA), as in (9). My corpus study showed that the distribution of the PR/PCsystem was partially similar to that of the PS/PA system, at least with respect to

OUP UNCORRECTED PROOF – FIRST PROOF, 15/6/2015, SPi

188 Patrick Caudal

Comp. by: SatchitananthaSivam Stage : Proof ChapterID: 0002544163 Date:15/6/15Time:19:20:14 Filepath://ppdys1122/BgPr/OUP_CAP/IN/Process/0002544163.3dDictionary : OUP_UKdictionary 189

causo-temporal subordinates (see Martin, 1971: 318, 395 and Detges 2006: 49 forrelated observations). Given the assumption that the OF PC is a resultative PR, thefact that the PR qua PS in OF implies that the PC could be used qua PA (cf. note 8).And indeed, in my corpus, Quant PA, PS structures appear to have similar aspectuo-temporal content and to be found in similar discourse contexts as Quant PC, PRstructures. They represent, in fact, the most frequent pattern of use for the SOE PC inthe Queste, making up nearly half of the observed occurrences (Appendix I). Thissuggests that syntactically encoded inter-clausal relations between clauses played aninstrumental role in the development of perfective uses for the PC. However, as thesestructures are semantically underspecified, one cannot claim that syntax alone was atplay here; additional discourse structural information was required. Within thenarratives I studied, such structures were frequently used to convey continuations(typically combined with relations such as Narration, Result, or Continuation) ofexpected events; this reflects on the fact that temporal adjuncts, whether finite or not,are presuppositional (see Bary and Haug 2011), cf. for example (15).

Quant α , β structures with discourse segment α in the PA or PC typically indicatethat event eα

12 was the intended goal of some previously initiated event, such as the‘going to the abbey in order to attend mass’ event in (15), or that the event is somehowexpected to occur by virtue of scriptal knowledge.

(15) Lors ala oïr messe en l’abeïe meïsmes. Et quant il fu entrez au mostier si vit adestre partie unes prones de fer . . . (OF)Then go-ps.3sg hear mass in the-abbey itself. And when he be-ps.3sg enter-ppat.the monastery, then see-ps.3sg on right part a gate of iron . . .‘Then he went to the abbey itself in order to attend mass. And when he hadentered the monastery, he saw an iron gate on the right . . . ’ (Queste, 176b 30–33)

Quant α PA clauses typically introduce some previously expected or planned pre-supposed event, or “close off” (i.e. set a final/result state boundary for) a previouslybegun, presupposed atelic event as in (17a)—the latter uses being often translated inMod.Fr. using finir de ‘finish’ in the Chevalier edition I used to conduct my corpusassessment, cf. (16b):

(16) a. Si manjüent molt lieemant. / Quantmangié orent longuemant, /Li vavasorsdist a son oste . . . (Chevalier, 2965–7) (OF)And eat-pr.3pl very gaily; when eat-pp have-ps.3pl long, the nobleman-subj say-ps.3sg to his guest . . .‘And they ate merrily. When they had eaten with all the appropriateslowness, / the nobleman said to his guest . . . ’.

12 Following standard SDRT notation, eα is the event underlying some speech act referent/discoursesegment α.

OUP UNCORRECTED PROOF – FIRST PROOF, 15/6/2015, SPi

The passé composé in Old French and Modern French 189

Comp. by: SatchitananthaSivam Stage : Proof ChapterID: 0002544163 Date:15/6/15Time:19:20:15 Filepath://ppdys1122/BgPr/OUP_CAP/IN/Process/0002544163.3dDictionary : OUP_UKdictionary 190

b. . . . Lorsqu’ils eurent fini de manger . . . (Mod.Fr.)when-they have-ps.3pl finish-pp cp eat-inf . . .‘when they had finished eating . . . ’.

The use of finir de highlights the event presupposition triggered by the quant PAclause; quant mangié orent longuemant presupposes that an eating event (a meal) hadalready been initiated, and indicates that the initial boundary of the associated resultstate has been reached at reference time (i.e. that the eating event has reached its finalboundary).13 Since such examples clearly demonstrate that the PA in OF could haveinchoative resultative meanings, and since the OF PC and PA apparently haveparallel uses in OF within quant structures, this strongly suggests that the OF PCalso had inchoative resultative uses.

Note that when collecting the examples discussed here, in contrast to existing studiessuch as Martin (1971: 318), I selected only the causo-temporal uses of quand/t struc-tures, rather than their solely temporal uses.14 In SOE contexts, instances of quant α, βstructures overwhelmingly receive causo-temporal interpretations such that eα tem-porally causes eβ. In a sense, α plays the role of a protasis/antecedent clause, and β thatof an apodosis/consequent clause. The acquisition of a causal meaning on top of a pre-existing temporal meaning is a well-known development path followed by (originally)temporal subordination markers (Traugott and Dasher 2002).

Consider, for example, the modern French examples (17)–(19). In (17a), we have eα< eβ or eα < � eβ (i.e. eα overlaps with (at least) the left part of eβ), versus the moreflexible eα � eβ in (18), where eα and eβ only need to overlap, without any causal link.Within the SDRT framework, this amounts to saying that quand, when endowedwith a causo-temporal meaning, monotonically triggers the Narration relation; suchgrammatical items are considered within the SDRT community to have a discourse-structural function, albeit within a complex sentence. This fact is modeled in (17b),(where ?R indicates an underspecified discourse relation). In (18), however, only theweaker relation of Occurrence is established (or possibly Parallelism, depending onthe context). And in (19), the imperfective viewpoint tense (the imparfait) in β eventriggers the Background relation.

(17) a. Quand Jean poussa l’échelle (α), Pierre tomba (β). (eα < eβ) ∨ (eα < eβ)When Jean push-ps.3sg the-ladder, Pierre fall-ps.3sg. (Mod.Fr.)‘When Jean pushed the ladder, Pierre fell [down].’

b. “Quand/t” Narration: ?R(λ,α,β)∧quand/t(α,β)∧cause(eα,eβ)!Narration(α,β)∧[(eα < eβ)∨(eα < �eβ)]

13 How the quant/d presupposition is resolved appears to vary, accommodation being a frequentpossibility in the corpus I studied.

14 I also set aside the generic/habitual/iterative uses of quant/d.

OUP UNCORRECTED PROOF – FIRST PROOF, 15/6/2015, SPi

190 Patrick Caudal

Comp. by: SatchitananthaSivam Stage : Proof ChapterID: 0002544163 Date:15/6/15Time:19:20:15 Filepath://ppdys1122/BgPr/OUP_CAP/IN/Process/0002544163.3dDictionary : OUP_UKdictionary 191

(18) Quand Jean sortit (α), le soleil apparut. (β). (Mod.Fr.)When Jean go_out-ps.3sg, the sun appear-ps.3sg.‘When Jean went out, the sun came.’ (eα�eβ)

(19) Quand Jean sortit (α), il pleuvait. (β). (Mod.Fr.)When Jean go_out-ps.3sg, it rain-ps.3sg.‘When Jean went out, it was raining.’ (eα�eβ, or eα�eβ)

My corpus study showed that Old French narratives abound with causo-temporaluses of quant similar to (17a) rather than to (18) or (19).15

Quant structures are not the only subordinating constructions allowing perfectiveuses of the PC. In the consecutive subordination structure in tant . . . que in (2) above,the antecedent clause event effectively precedes and causes the PC consequent clauseevent. Similar examples appear in different forms throughout the corpus studied(in si tost . . . com, si . . . que . . . structures, etc.).

Other facts suggest an inchoative resultative present tense analysis of the PC in OF,with a “weak” perfective reading. The corpus contains, in fact, several somewhatpuzzling, apparently formulaic, passages which decompose event structures into acore part and a result state. Some of these discourses seem to even offer resultativePRs (armé sont in (20) and désarmé sont in (21)) on top of bona fide PCs, cf. estmontez in (22).

(20) Lor armes aporter comandent (α); / L’an lor aporte tost amont (β), / Vaslet lesarment (γ): armé sont (δ); (Chevalier, 4980–1) (OF)Their arms bring-inf order-pr.3sg / Them then bring-pr.3sg quick forth,servants them arm-pr.3pl arm-pp be-pr.3pl.‘They order that their arms be brought to them. They are brought forthimmediately. Servants put arms on their backs. They are armed.’

(21) Desarmer fet les chevaliers (α), / Au desarmer les filles saillent (β); / Desarmésont (γ), puis si lor baillent / A afubler trois corz mantiax.

(Chevalier, 2548–51) (OF)Disarm-inf make-pr.3sg the knights. At.the disarming the girls rush-pr.3pl.Disarmed be-pr.3pl, after and them give-pr.3pl to wear-inf three short coats.

15 Examples of quant PA, PC structures provide clear evidence of their ability to receive causo-temporalreadings in OF, and of the perfective dimension of the PC, cf. (i).

(i) Et quant il [= a monk] li ot tot conté, / Li chevaliers l’a comandé / A Deu et a trestoz ses sainz . . .(Chevalier, 1949–51)And when he him have-ps.3sg all narrate-pp / the knight-subj him.have-pr.3sg command-pp / toGod and to all his saints.(‘And when the monk had finished telling his tale, the knight recommended [the monk’s soul] to Godand all his saints.’)

OUP UNCORRECTED PROOF – FIRST PROOF, 15/6/2015, SPi

The passé composé in Old French and Modern French 191

Comp. by: SatchitananthaSivam Stage : Proof ChapterID: 0002544163 Date:15/6/15Time:19:20:15 Filepath://ppdys1122/BgPr/OUP_CAP/IN/Process/0002544163.3dDictionary : OUP_UKdictionary 192

‘She has the knights disarmed / Her daughters rush to disarm them / [Finally] theyare disarmed. Then [the maidens] give them three short coats for them to wear.’

(22) Quant Perceval voit le cheval si le resgarde et l’en prent hisdor, et neporec il estbien tant hardiz qu’il monte (α) sus . . . . Et quant il est montez (β) si prent sonescu et sa lance. (OF)When Perceval see-pr.3sg the horse, then it watch-pr.3sg, and him-from.ittake-pr.3sg fright, and yet he be-pr.3sg well so brave that-he mount-pr.3sgon. And when he be-pr.3sg mount-pp, then take-pr.3sg his shield and hisspear.‘When Perceval catches sight of the horse, he is frightened; and yet he is sobrave that he mounts [it] [= climbs on its back] . . . . And when he has mountedthe horse, he takes his shield and his spear.’

The arming event eγ in (22), and disarming event eα in (21) are expressed by (non-resultative) PR segments; being semantically underspecified, they could be inter-preted either perfectively or imperfectively. But the introduction of a subsequentresultative PR or PC, marking the onset of an associated result state, forces an SOEreading of the whole sequence: the “inner stage”/“development stage” events eγ in(20) and eα in (21) receive a perfective interpretation,16 as they are followed by(inchoative) result stage events, that is eδ in (20) and eγ in (21).

To make things more truth-conditionally precise, consider (23). The inchoativeresult state analysis would claim that the inchoative resultative event e2 describedby est alez is preceded by some contextually-specified entailed causal event, forexample some preparatory event—a decision-making process, or the beginning ofthe king’s motion . . .—leading up to e2 (typically a change-of-state predicatesemantically akin to “become,” “start,” etc.). In short, inchoative statives entail acontextually determined, external causal event,17 controlled by the subject whererelevant, but no longer part of the at issue contribution of the PC than it would befor for example the English perfect (at least under an analysis such as Nishiyamaand Koenig (2010).

(23) Ez-vos le roi molt desperé (e1); / Si est a la reïne alez (e2). / «Dame, fet-il(e3) . . . » (Chevalier, 116–18) (OF)Here_is the king very desperate / And be-pr.3sg. to the queen go-pp. “Lady,say-pr.3sg he . . . ”‘And the king became desperate; he went close to the queen. “My lady, sayshe, . . . ” ’

16 Where ‘stage’ or ‘phase’ (in the sense of Smith 1991) refers to the decomposition of an event into sub-events, in the sense of Parsons (1990).

17 Thus John was suddenly sick entails an external causing event of e.g. some virus entering John’s body.

OUP UNCORRECTED PROOF – FIRST PROOF, 15/6/2015, SPi

192 Patrick Caudal

Comp. by: SatchitananthaSivam Stage : Proof ChapterID: 0002544163 Date:15/6/15Time:19:20:15 Filepath://ppdys1122/BgPr/OUP_CAP/IN/Process/0002544163.3dDictionary : OUP_UKdictionary 193

9.3.3 Why the standard late perfectivization hypothesis is not fully satisfying

The data I have presented above confirm that inchoative resultative readings ofthe PC were fully part of the grammar of OF at least as late as the thirteenthcentury, and seems to substantiate the standard version of the late perfectivizationhypothesis. I now turn to data suggesting that a more complex analysis is in factrequired.

I have so far focused on SOE PCs appearing in temporal subordinate clauses,attached to matrix clauses in some other tense. But PCmatrix clauses with a temporalsubordinate were also found, thus appearing on the opposite side of the causo-temporal relation introduced by such biclausal structures. They are realized both inmy corpus (24) and elsewhere (25) as Quand/t PR, PC, Quantd/t PS, PC,18 Quand/tPC, PC, or Quand/t PA, PC.

(24) Quant la reïne voit le roi (α), / . . . Si s’est contre le roi dreciee (β).(Chevalier, 3955–7) (OF)

When the queen see-pr.3sg the king, then refl-be-pr.3sg against the kindstand_up-pp.‘When the queen saw the king . . . , she stood up in defiance.’

(25) Quand l’oït Guenes (π1), l’espée en ad brandie (π2) / Vait s’apuier suz le pin à latige. (Roland, 499–500) (OF)When him-hear-pr.3sg Ganelon, the-sword from-that have-pr.3sg brandish-pp. Go-pr.3sg refl-lean-inf on the pine at the trunk.‘When Ganelon hears him, he has brandished [= brandished] his sword. Hegoes and leans against the trunk of the pine (tree).’

At first sight, such data might appear to be amenable to the inchoative resultativeanalysis of the OF PC. In this view, the matrix clause β in (24) contributes aninchoatively interpreted result state eβ following the event of the queen standing upin defiance, and the matrix clause π2 in (25) contributes an inchoative result state eπ2following the sword-brandishing event by Ganelon. In SDRT terms, the discourserelation Narration imposed by the quant structure guarantees a strict temporalsequence between the subordinate clause events eα / eπ1 and the perfectively inter-preted matrix result states eβ / eπ2.

But a complication arises at this point . If we note eβ0 and eπ20 the external, causingevents respectively entailed by the PC matrix clauses in (24) and (25), then it isobvious that eβ0 < eβ and eπ20 < eπ2 (“external” causes must precede their effects).However, from axiom (17b) on quand/t narrations, we can determine eα < eβ andeπ1 < eπ2—nothing more: we only dispose of a partial temporal ordering. It followsfrom this partial ordering that one cannot rule out a context in which the external

18 See also (2).

OUP UNCORRECTED PROOF – FIRST PROOF, 15/6/2015, SPi

The passé composé in Old French and Modern French 193

Comp. by: SatchitananthaSivam Stage : Proof ChapterID: 0002544163 Date:15/6/15Time:19:20:15 Filepath://ppdys1122/BgPr/OUP_CAP/IN/Process/0002544163.3dDictionary : OUP_UKdictionary 194

causing events eβ0/eπ20 (with e.g. eβ0 some motion event leading up to/preparing thebeginning of state eβ in (24)) overlap with the PR events eα / eπ1 introduced by thequant subordinates (cf. Figure 9.1).

But according to world-knowledge, only the following temporal ordering shouldbe allowed to hold: eα < eβ0 < eβ and eπ1 < eπ20 < eπ2 (or maybe at least eα < eβ0 < eβ andeπ1 < eπ20 < eπ2, with < marking (non-strict) anteriority with left-overlap19). I cannotsee a way around this without introducing the appropriate temporal orderinginformation involving eβ0/eπ20—that is, without ascribing some “strong” perfectivepower to such PC matrix clauses.20

Examples such as (24)–(25) thus suggest that as early as the OF period, the PC wasin fact capable of “strong” perfective interpretations, well beyond inchoative resulta-tive present readings. In other words, some degree of interpretative innovation mustalready have been present in the PC for such examples to be felicitous. Thisconclusion is incompatible with a simple semantic version of the late perfectivizationhypothesis, according to which SOE PCs were (inchoative) resultative PRs.

Another argument against this latter view is provided by the compared distribu-tions of the PA and the PC in narrative contexts. If the PR was used as a substitute tothe PS in narrative contexts (cf. Detges 2006), then it should follow that theinchoative resultative uses of the PC should closely parallel those of the PA. This isnot what I observed in my corpus study, however. On the contrary, the PA neverappears with a SOE reading outside quant structures or other causo-temporalcorrelative structures. This fact suggests that the OF PC had acquired independentperfective SOE uses by means of pragmatic processes which gradually strengthenedover time, as opposed to the PA, which did not receive SOE readings outside ofcorrelative causo-temporal structures.

eα ‘la reïne voit’eβ: inchoatively interpreted result state

(‘the Queen stood’-INCHOATIVE)eβ': entailed cause of eβ

FIGURE 9.1 Undesirable event ordering for (26) obtaining with a simple inchoative analysis

19 However, the presence of the pronominal causal complement en (‘because of this’) in (25) suggests astrong causal connection and strict temporal ordering between eπ1 and eπ2.

20 Again, claiming that eβ is in fact more than an inchoative result state, and corresponds e.g. to a causal,subject-controlled event of standing up, is a temptation one must resist, as this would make the OF PC asemantically perfective tense. Definite temporal adverbials should be able to quantify over perfective eventsthus denoted; yet this is impossible, cf. Section 9.2.2. Only eβ0 can have causal force—not eβ.

OUP UNCORRECTED PROOF – FIRST PROOF, 15/6/2015, SPi

194 Patrick Caudal

Comp. by: SatchitananthaSivam Stage : Proof ChapterID: 0002544163 Date:15/6/15Time:19:20:16 Filepath://ppdys1122/BgPr/OUP_CAP/IN/Process/0002544163.3dDictionary : OUP_UKdictionary 195

More should be said at this point on the alleged decline of the SOE uses of PCin late OF/ Mid.Fr. In fact, these uses were already far from abundant in the OFperiod. In Martin’s (1971: 347) figures (based on Villehardouin’s Conqueste, andthe Morte d’Artu), the PC (whether narrative or non-narrative) accounted foronly 5.06 per cent of finite verb forms in OF. In Mid.Fr., it accounted for 7.74 percent of verb forms. Martin (1971: 350) notes, however, that over the Mid.Fr.period, the PC underwent major frequency variations within the texts he studied:from a high of 12.2 per cent of finite verb forms in the early fourteenth century, itdropped to a mere 4.65 per cent at the end of that century. This suggests that thelate OF PC remained a typically secondary tense within narrative texts in general,and that its decline in the fourteenth century was only relative. Indeed, a quickcorpus foray on fourteenth-century texts demonstrated that SOE uses of the PCstill occurred, even at the worst time of its history according to Martin’s (1971)chronology, cf. (26)–(27):

(26) Quant li dieus m’ot tout cela dit / . . . Bien me souvint de la priere / Que faiteavoie darreniere, / Si qu’encor li renouvelay, / Et humblement prié li ay(Guillaume de Machaut, Le Dit dou vergier, 49; c. 1340) (Mid.Fr.)When the God-subj me-have-ps.3sg all this say-pp, well me remember-ps.3sgof the prayer which do-pp have-impf-3sg last-f, so that-again it renew-ps.1sg,and humbly pray-pp him have-pr.1sg.‘Once God had told me all these things, / I remembered well the last prayer /Which I had said / So that I said it again / And humbly I prayed.’

(27) Adonc ot li roys grant despit du chevalier . . . si le fist mettre en prison obscureet villaine, puis a parlé a ses autres barons et dist . . .

(Bérinus, t. 1, c.1350–70, 116) (Mid.Fr.)So have-ps.3sg the king-subj great disappointment of.the knight, so himmake-ps.3sg put-inf in prison obscure and ugly, then have-pr.3sg talk-pp tohis other-pl barons and say-ps.3sg . . .‘So the king was sorely disappointed by the knight . . . ; so that the king had himsent to an obscure and ugly prison, and then talked to his other barons, andsaid . . . ’.

These data confirm that SOE uses of the PC, though perhaps rarer for a time, neverlost grammatical significance. There was more continuity in the transformation ofthe PC throughout the OF/Mid.Fr. period than is assumed in late perfectivizationapproaches such as Wilmet’s (1970) or Schøsler’s (1973, 2004). The drop in frequencyof SOE PCs in the fourteenth century was not a “mass extinction”; it should be seenas an evolution rather than as a revolution. Since these uses of the PC never died outbefore being revived in a different guise, the original “two births and a burial”scenario should be abandoned.

OUP UNCORRECTED PROOF – FIRST PROOF, 15/6/2015, SPi

The passé composé in Old French and Modern French 195

Comp. by: SatchitananthaSivam Stage : Proof ChapterID: 0002544163 Date:15/6/15Time:19:20:16 Filepath://ppdys1122/BgPr/OUP_CAP/IN/Process/0002544163.3dDictionary : OUP_UKdictionary 196

9.4 Combining an innovative pragmatics with a conservative semanticsto account for the perfective uses of the PC in OF

We are now in a somewhat paradoxical situation. On the one hand, it appears thatmany SOE uses of the PC in subordinate clauses of quant structures and othercorrelative causo-temporal structures can be accounted for in terms of a proto-perfect-like, inchoative resultative present reading of the PC, which, according tothe late perfectivization hypothesis, was still available in OF (cf. Sections 9.3.1 and9.3.2). In Section 9.2.2.3, we dismissed the possibility that the PC could have hadsemantic past perfective capability in OF, as it was not compatible with, for example,definite past temporal modifiers.21 This applies of course to SOE readings of the PC:those perfective inchoations cannot be considered as true past inchoative perfectivemeanings, as they cannot be semantically anchored in the past via dedicated linguis-tic material; their only possible past dimension was that of a so-called narrativepresent, no more, no less.

On the other hand, we saw in Section 3.3 that at least some SOE uses of the PC inOF call for a more elaborate explanation: but although they appear to militate infavor of an independent early perfectivization of the PC (cf. Section 9.3.3), it clearlycannot be of a semantic nature, given the above observations. I will thereforehypothesize that the latter uses involve a conventionalized kind of pragmatic pro-cesses (cf. e.g. Hopper and Traugott 2003; Ramm 2014). The idea is that (only) someof the SOE uses of the OF PC did not originate in its former career as a resultative PRconstruction; they were still one step short of perfective semanticization, which isknown to have occurred only sometime around the seveenteenth century, when thePC became compatible with the full range of aspectuo-temporal modifiers charac-teristic of past perfective viewpoint tenses, notably hier ‘yesterday’ (see Caudal andVetters 2007).

9.4.1 Outlining a novel, composite account

The novel account I will propose reconciles insights from both the late and earlyperfectivization analyses. Inspired by the former, I will model the OF PC as a case ofcomposite, layered semantics in the sense of Hopper (1991), combining a full perfectsemantic layer22 with a more ancient resultative PR semantic layer, inherited from itsconstructional ancestor, and allowing for inchoative resultative readings (see Sections

21 While my corpus study revealed a small number of occurrences of the PC with past temporalmodifiers, none of them referred to a past day, week, month, or year. This fact is corroborated by otherstudies (cf. Treikelder 2006: 90; Martin 1971: 251–60).

22 Including the entailment that some past causal event brought about the result state described, withthe semantic role information/argument structure being shared between the PC and this entailed event–cf.e.g. subject referents of agentive aveir PC being agents of the entailed past causal events.

OUP UNCORRECTED PROOF – FIRST PROOF, 15/6/2015, SPi

196 Patrick Caudal

Comp. by: SatchitananthaSivam Stage : Proof ChapterID: 0002544163 Date:15/6/15Time:19:20:16 Filepath://ppdys1122/BgPr/OUP_CAP/IN/Process/0002544163.3dDictionary : OUP_UKdictionary 197

9.3.1 and 9.3.2).23 And, on top of this, I assume that the OF PC had developed an“innovative” pragmatics, sensitive to discourse structural triggers, and capable ofstrong (though context-dependent) SOE perfective interpretations (i.e. not merelyinchoative resultative readings; see Section 9.3.3). The latter idea is inspired fromCaudal and Roussarie (2006), who fleshed out the already old intuition that thesemantic evolution of tenses is preceded and driven by pragmatic, discursive changes(cf. Fleischman 1983; Givón 1979).

I will argue that the dual, layered semantics of the OF PC24 makes it aspectuallyunderspecified, that is, capable of contributing both a proto-perfect, inchoative,perfective resultative present viewpoint (which gives rise to inchoative-perfectiveSOE readings), and an imperfective resultative viewpoint—in effect a fully grammat-icalized perfect. To make this clearer, let us consider Nishiyama and Koenig’s (2010)distinction between resultant states and perfect states. While the former are lexicallycontributed by the PP component of a have/be+PP perfect (e.g. the resultant state ofdie0 is dead0) and would be associated with the proto-perfect (perfective) resultativepresent reading of the OF PC, the latter corresponds to topically salient resultantevents which can be contextually inferred from the contribution of the PP, and wouldbe associated with the full perfect, non-perfective reading of the OF PC. ThusParsons’s (1990) example, My son has thrown a ball onto the roof, may be attachedby some inferential process to various presently valid, imperfectively viewed perfectstates, which are (stative) consequences of the past event entailed by the perfect, suchas, for example,my son deserves to be punished / is a naughty boy, etc. In contrast, theresultant state conveyed by the PP (thrown_on_the_roof (ball)) may no longer holdat speech time. This is typically the case with the existential/experiential readings ofthe perfect. Since such uses of the PC are attested in OF, cf. (28)–(29), we must acceptthat the OF PC could describe both (imperfective) perfect states, and the resultantstate readings which must be associated with inchoative, perfective readings, cf.examples in Sections 9.3.2 and 9.3.3; see Appedix II below for a tentative formalimplementation of this analysis.

(28) En cort al rei mult i avez ested (Roland, 351, in Detges 2006: 64)In court of.the king much there have-pr.2pl be-pp.‘You have been at the king’s court many times.’

23 What Hopper (1991) calls grammatical layering is a kind of historically created polyfunctionality,whereby a certain grammatical form acquires a new semantic content A, without having (completely)discarded its former semantic content B, so that it ends up meaning both A and B for some period of time.

24 Recall that this goes hand in hand with a yet under-determined morpho-syntax; the OF PC did notyet treat the auxiliary + past participle sequence as a discontinuous tense, and allowed a relatively freesyntactic ordering of the two elements.

OUP UNCORRECTED PROOF – FIRST PROOF, 15/6/2015, SPi

The passé composé in Old French and Modern French 197

Comp. by: SatchitananthaSivam Stage : Proof ChapterID: 0002544163 Date:15/6/15Time:19:20:16 Filepath://ppdys1122/BgPr/OUP_CAP/IN/Process/0002544163.3dDictionary : OUP_UKdictionary 198

(29) Car il ont toz jorz esté sainz et haitiez (Queste del Saint Graal, p. 4)For they have-pr.3pl all days be-pp healthy and well.‘For they have always been in good health.’

In contrast to the OF PC, canonical perfects such as the English perfect can beclaimed to be only full perfects, that is, as imperfective resultative viewpoints havingmostly25 shed their proto-perfect readings.

9.4.2 The dynamic historical perspective issue: agentivity versus SOE contexts

I would now like to try and answer the thorny question of the driving force behind theevolution of the OF PC, and recast my account within a dynamic historical perspec-tive. Although his account is similar to mine in some important respects,26 Detges(2006) (cf. note 6) treats agentivity, as expressed by the aveir PC, as the key factorexplaining the evolution of the OF PC (as well as perfects in other Romancelanguages); I will here show that an alternative hypothesis is preferable.

Detges strongly associates the agentivity exhibited by aveir PCs with the broadernotion of a subject being “morally responsible” for a certain resultative state. Yetneither “moral responsibility,” nor any agentivity-related concept, appears to patternwith the general estre versus aveir auxiliary divide, since verbs forming an aveir PCcan have a non-agentive subject, as noted in Detges (2006: 67). To accommodate thisembarrassing fact, Detges claims that non-agentive aveir PCs involving, for example,non-volitional change-of-state verbs (cf. perdre ‘to lose’), or stative verbs (cf. amer ‘tolove’), are instances of what he calls “expressives,” namely violations of grammatical,semantic rules licensed by pragmatic principles. Detges’s (2006) account is thusinherently non-compositional, and presumably non-monotonic—a highly undesir-able move in my opinion.

More importantly, it is wrong to view the OF aveir PC as a subject-orientedperfect; contra Detges (2006). It is indeed easy to show that aveir PCs do not havesuch an inherent bias; thus, in examples such as (13), the PC (mise ‘put’) conveyscrucial information about the object argument, rather than about the subject argu-ment. If aveir PCs are not subject-oriented perfects, then it is unlikely that propertiesof subjects should play such a crucial role in their grammaticalization. In a similarvein, many PC examples with estre found in my corpus exhibit the very same SOEreadings as with aveir, cf. (20)–(57): it is incorrect to view estre PCs as mereresultatives, and a unified account of both aveir and ester PCs must be proposed.

Last but not least, I believe that Detges’s assumptions about the grammatical-ization timeline of the PC are debatable. Detges argues that at some relatively early

25 See Declerck (1997: 211).26 Detges (2006) ascribes a conservative, proto-perfect semantics to the OF PC, combined with some

full perfect features, while also arguing for an innovative role of pragmatics.

OUP UNCORRECTED PROOF – FIRST PROOF, 15/6/2015, SPi

198 Patrick Caudal

Comp. by: SatchitananthaSivam Stage : Proof ChapterID: 0002544163 Date:15/6/15Time:19:20:16 Filepath://ppdys1122/BgPr/OUP_CAP/IN/Process/0002544163.3dDictionary : OUP_UKdictionary 199

stage (a) along this timeline, subjects of aveir PCs became conventionally associatedwith an agentive semantic role, long before stage (b) where past, causal eventsassociated with result states described by PC utterances became conventionallyassociated with them (as, e.g., entailments). But how could a stage (a) aveir PCsemantically convey subject agentivity without conventionally entailing that saidagent took part in the associated past event? In contrast I will simply view agentivityas one of many semantic roles possibly ascribed to subjects of OF PCs, and assumethat in early Old French, semantic role/argument structure information associatedwith the entailed event became unified with those of the PC structure itself, both foraveir and estre PCs.27

My corpus study revealed (and this was unexpected) that at some stage, SOE usesof the OF PC were largely associated with correlative, causo-temporal biclausalstructures, notably with causo-temporal quand ‘when’ clauses, where the subordinateclause event precedes (and sometimes causes) the matrix clause event. I wouldtherefore like to argue that repetitive uses of PCs within syntactic structures imbuedwith inter-clausal, discourse structural-like content, facilitated the rise of “strong”perfective uses of the PC in juxtaposed matrix clauses. As we have seen, these purelydiscursive SOE readings of the PC (i.e. not involving any syntactic “prop”) wereattested in OF, but had not yet achieved semantic status: the perfectivization of thePC was still a pragmatic phenomenon (albeit of a “conventionalized” kind, that is tosay, it was not an issue of “free” pragmatic enrichment). See Caudal and Roussarie(2006), who offer a tentative (partly) past perfectivized semantics for later stages inthe development of the PC (around the seventeenth century). It contrasts with thepurely present semantics proposed for the OF PC in Appendix II.

My analysis suggests at the very least that the interaction between the OF PC andSOE structures in general (both with and without syntactic marking, i.e. boththrough syntactically marked and discursively motivated SOE structures) is a likelydriving force behind the perfectivization of the PC.

9.5 Conclusion: an evolution rather than a revolution

I have argued in this chapter that neither the early nor the late perfectivizationhypotheses found in the literature can adequately account for the history andevolution of the French PC through the OF (and Mid.Fr.) period(s). I have formu-lated an intermediate solution which draws insights from both types of accounts.

Taking inspiration from the late perfectivization accounts, I have claimed that theOF PC inherited conservative morphosyntactic and semantic (aspectuo-temporal)features from its “parent” present resultative constructions, which made the OF PC

27 With event structure parameters, e.g. telicity, also playing a role in the distribution of estre versusaveir auxiliaries; cf. Dufresnes and Dupuis(2010).

OUP UNCORRECTED PROOF – FIRST PROOF, 15/6/2015, SPi

The passé composé in Old French and Modern French 199

Comp. by: SatchitananthaSivam Stage : Proof ChapterID: 0002544163 Date:15/6/15Time:19:20:16 Filepath://ppdys1122/BgPr/OUP_CAP/IN/Process/0002544163.3dDictionary : OUP_UKdictionary 200

capable of conveying inchoative resultative readings; this accounts for at least some ofits SOE readings. And inspired by the early pragmatic perfectivization hypothesis putforth in Caudal and Roussarie (2006), I have furthermore claimed that the OF PCindependently acquired a conventional pragmatic “strong” perfective content, not-ably within matrix clauses, before it was morphosyntactically turned into a full tense.

I have suggested that the early history of the PC was one of gradual, dove-tailingchanges, with the OF PC exhibiting a layered semantics (which I model through aspecial kind of underspecified aspectual contribution in Appedixx II). That is not tosay that there were no discontinuities at all in the history of the PC; the subsequentsemanticization of strong perfective SOE uses of the PC probably counts as the mostspectacular change of that kind—I have merely demonstrated that this change didnot take place at the OF period. Details of the subsequent evolution of the PC inMiddle French, and its numerous differences with the OF PC, must be left asquestions to be addressed by future research.

Last but not least, if we replace the above analysis within the perspective of thepresent volume, I believe it provides us with some novel indications concerning theinterplay of syntax (and semantics) with discourse. It is probably no accident thatsyntactically-semantically relational (quand subordination with causo-temporal con-tent) and discursively relational (Narrative discourse relations) structures with simi-lar “sequence of event” functions both participated in the grammaticalization ofperfective meaning for the Old French PC—and contributed to transforming amerely morphosyntactically relational (discontinuous) form into a semanticallyrelational form, that is, one capable of relating a past perfective event to a presentresultative event.28 Or to put it differently, it seems that structural convergencesbetween syntax (alongside with semantics) and discourse (alongside with pragmat-ics) can place them simultaneously at the heart of particular grammaticalizationphenomena.29

Such an idea should of course be related to the general notion, that with time andrepeated uses, more or less conventionalized pragmatic uses (e.g. entailed/implicatedcontent, possibly in special discourse contexts) can become semanticized30—or,

28 See Caudal and Roussarie (2006) for more on the view that the modern French PC is in fact a“double,” relational tense, i.e. one relating and anchoring simultaneously a past perfective event, and apresent resultative event. A consequence of this analysis is that the French PC never was a bona fide perfect,unlike e.g. the English perfect.

29 Thus it has been independently proposed that SOE discourse structures involving SOE discourserelations, e.g. Narration and Result (Caudal and Ritz 2012), or mutatis mutandis equivalent theoreticalconcepts (Fleischmann 1990), played a key role in the evolution of tenses across languages. It would bemost enlightening to qualify the syntactic environment in which these evolutions took place (e.g. seewhether these correlate with particular inter-clausal syntactic patterns).

30 Specifically, while past causal sub-events were part of the non-at-issue entailed content of “incho-ative” resultative readings of the OF PC, they became “semanticized” as bona fide, past perfective events inthe at-issue semantic content of Mod.Fr. PC.

OUP UNCORRECTED PROOF – FIRST PROOF, 15/6/2015, SPi

200 Patrick Caudal

Comp. by: SatchitananthaSivam Stage : Proof ChapterID: 0002544163 Date:15/6/15Time:19:20:17 Filepath://ppdys1122/BgPr/OUP_CAP/IN/Process/0002544163.3dDictionary : OUP_UKdictionary 201

again to put it differently, that the pragmatics of linguistic forms can indeed pave theway for future semantic (and correlatively, syntactic) developments (see Givón 1979;Caudal and Roussarie 2006). However, the present analysis goes beyond such an idea,as it suggests that discursive/pragmatic and syntactic/semantic factors are inherentlyintertwined in such grammatical developments from the onset, presumably becausethey share some important structuring functions—that is, the “driving force” behindsuch grammatical developments combines discursive/pragmatic with syntactic/semantic mechanisms. The fact that OF then exhibited somewhat fuzzier boundariesbetween “discourse” motivated versus “syntactically” encoded relations betweenclauses (and associated events) was almost certainly a facilitating factor for theevolution here studied. Moreover, the very possibility of such blurred boundariespoints to a shared functional core of syntax–semantics and discourse structure (seeFleischman 1990 for more on this).

If this view is correct, contrasting languages (or language stages) with stronglyversus weakly developed inter-clausal syntactic systems (subordination, coordin-ation) and the various types of associated discourse/syntax interplay might openinteresting avenues of research for our understanding of the development of various“relational” grammatical categories (e.g. inflection, auxiliaries, connectives, . . . ) bothin language diachrony and language typology. But obviously, these must await futuredevelopments.

Appendix I: Corpus analysis

TABLE 9.1

Change-of-state uses of the PC in narration Chevalier Queste

In paratax ‘Bare’ PC 44 1

With an aspectuo-temporal adverbial 20 2

With a discourse connective (puis, lors) 6 2

Combination beginning/endpoint adverbials 1

Total paratax per work 70 6

Total paratax 76

In hypotax Tant PR que PC 5 4

Tant PS que PC 1

Tant PC que PC 1 1

Tant PC, que PA 1

Après PC, tant que PC 1

Tant PC durative exp. que PR 1

Tant PC que PS 1 3

Si, que PC 1 1

Por ce que 1

Quant PR, PC 4

(continued)

OUP UNCORRECTED PROOF – FIRST PROOF, 15/6/2015, SPi

The passé composé in Old French and Modern French 201

Comp. by: SatchitananthaSivam Stage : Proof ChapterID: 0002544163 Date:15/6/15Time:19:20:18 Filepath://ppdys1122/BgPr/OUP_CAP/IN/Process/0002544163.3dDictionary : OUP_UKdictionary 202

TABLE 9.1 Continued

Change-of-state uses of the PC in narration Chevalier Queste

Quant PS, PC 1

Quant PA, PC 3

Quant PC, PC 1

Quant PC grant piece, PC 1

Quant PC, PR 3 25

Quant PC jusqu'à < time>, PR 1

Quant PC grant piece, PR 3

Quand PC, PS 1 4

Quant PC grant piece, PS 1

Tantost PS, que PC 1

Si tost com PC, PS 1

Si tost com PC, PR 1

Des que PS, PC 1

A poi que PC, maintenant que PR 1

PR, a poi que PC 1

Perfective relative in the PC 2 1

Total hypotax per work 30 48

Total hypotax 78

Total perfective PCs per work 100 54

Total perfective PCs 154

Uses in paratax are matrix uses, with clause coordination or juxtaposition, whereas in uses in hypotax, theinterpretation of the PC depends on a subordinating structure.

TABLE 9.2

Chevalier Queste

Percentage of perfective PCs per work 1.250% 0.754%Global percentage of perfective PCs 1.016%

TABLE 9.3

Imperfective uses of the PC in narration (except in reported speech) Chevalier Queste

Matrix narrative imperfective PCs 4 7

Matrix non-narratives Imperfective PCs 1 2

Narrative imperfective PCs in subordinates 4 17

Non-narrative imperfective PCs in subordinates 3 3

Total narrative imperfective PCs 8 16

Total imperfective PCs per work 18 29

Percentage of imperfective PCs per work 0.225% 0.405%Percentage of narrative imperfective PCs per work 0.175% 0.321%Global percentage of imperfective PCs 0.310%

OUP UNCORRECTED PROOF – FIRST PROOF, 15/6/2015, SPi

202 Patrick Caudal

Comp. by: SatchitananthaSivam Stage : Proof ChapterID: 0002544163 Date:15/6/15Time:19:20:18 Filepath://ppdys1122/BgPr/OUP_CAP/IN/Process/0002544163.3dDictionary : OUP_UKdictionary 203

Appendix II: Formal implementation

As I cannot offer here a fully detailed implementation of the above analysis for wantof space, I will only outline one, couched within Type Composition Logic (TCL)proposed in Asher (2011), and Asher and Lascarides’ (2003) SDRT framework (seeAsher, this volume).

Following Asher and Hunter (2012), I assume that the event denotation of acompositional verbal complex is best accounted for by resorting to event realizers,which are functions from propositional contents (roughly the verb complex qua aproposition) to events. This allows for an elegant treatment of aspectual coercion astriggered by, for example, tenses or aspectual modifiers (see Asher and Hunter 2012for details), where aspectuo-temporal meanings are higher-order meanings.

As the PC contributes an aspectually underspecified and at least partly compos-itional verb complex with two discontinuous elements, I take the PP component ofthe OF PC to introduce a realization function mapping the propositional content ofthe verb complex (either with estre or avoir) onto the super-type of resultative statesres-state, encompassing both perfect states and resultant states (i.e. perf-statevres-state and resultantvres-state) and being in this sense semantically layered.I note this function res-realizes, cf. (30); it is essentially a kind of aspectual, stativizingresultative function—as such, it does not bind any event variable; the finite (= tense)component of the PC given in (31) will take care of this.31

Example (30) offers a much simplified notation, where arguments of the verb arenot treated as DP-types for clarity’s sake. Note that π is an additional argumentrepresenting the stack of (presupposed) semantic restrictions attached to any predi-cate; π*arg1

res-realizes: res-state thus means that the resultative component of the PCmust apply to resultant states. Last but not least, (31) captures the contribution of thefinite component of the PC, here treated as an underspecified tense function çPR-RES(i.e. a TCL polymorphic function), combining with the above (layered) resultativeaspect component.

(30) λPλeλx1 . . . xnλπ [res-realizes(e,^P(x1 . . . ,xn), π*arg1res-realizes :res-state)]32

(31) λQλπ0 ∃z:event [çPR-RES(HD(Q))(z)∧Q(z)(π0)]

The SDRT framework of Asher and Lascarides (2003) and Asher (this volume) willnow help us illustrate how the PC can be made sensitive to discourse structuraltriggers. Example (32a) states that a discourse segment β in the PC attached to thediscourse context via the Narration or Result discourse relations will trigger a

31 I leave aside the issue of the temporal anchoring function of the PC proper, as it hinges on a temporaltreatment of the présent in OF; but see e.g. Zucchi (2006).

32 res-state can be likened to Nishyama and Koenig’s (2010) context-dependent “perfect state” function X.

OUP UNCORRECTED PROOF – FIRST PROOF, 15/6/2015, SPi

The passé composé in Old French and Modern French 203

Comp. by: SatchitananthaSivam Stage : Proof ChapterID: 0002544163 Date:15/6/15Time:19:20:18 Filepath://ppdys1122/BgPr/OUP_CAP/IN/Process/0002544163.3dDictionary : OUP_UKdictionary 204

perfective viewpoint interpretation of çPR-RES. Conversely, (32b) states that if β isattached by, for example, Background or Explanation,33 çPR-RES will receive animperfective resultative viewpoint (= perfect) interpretation.

(32) Typing effects of discourse relations on çPR-RES:a. (Narration(α,β)∨Result(α,β))∧β:{���çPR-RES(e) . . . } ! (∊(PR-RES)v

PERFECTIVE-RES)b (Background(α,β)∨Explanation(α,β))∧β:{���çPR-RES(e) . . . }!(∊(PR-RES)v

IMPERFECTIVE-RES)

I also take the following to hold by definition (with n the utterance interval):

(33) Axioms on perfective versus imperfective readings of the PC OF:a. 8e,x1, . . . ,xn,π çPERFECTIVE-RES(e, x1, . . . ,xn,π))$evnb. 8e,x1, . . . ,xn.,π çIMPERFECTIVE-RES(e,x1, . . . ,xn,π))$nve

But most importantly, I take çPERFECTIVE-RES to be a perfective viewpoint functionranging over change-of-state events (type cos), and çIMPERFECTIVE-RES a imperfectiveviewpoint function reanging over perfect states (type perf-state).34 This means thatwhile the application of (31) to (30) is always well-formed with Background orExplanation in OF (çIMPERFECTIVE-RES has its input type requirement met), it isnever with Narration or Result: the cos type restriction borne by çPERFECTIVE-RESclashes with the res-state type offered by res-realizes.

Following solutions proposed in Asher and Hunter (2012) and Caudal et al. (2012),for example, to similar problems, I will argue that the type bridging function (34)then intervenes to resolve this type clash, by applying to (30) before combining itagain with (31) (see derivation (35))—this is an instance of so-called dependent eventtype coercion. çinchoation is a polymorphic function over change-of-state events (typecos—it is therefore compatible with çPERFECTIVE-RES) and resultant state predicates.

35

Subscripts (HD(P0), HD(x1) . . . HD(xn) in (35) indicate that the interpretation of çinchoationmay be sensitive to the semantic type (given by the functionHD) of the resultant statepredicate P0 it applies to, and to that of its arguments. Finally, result(e0,e) indicatesthat e0 results in e at most as a weak, internal cause, with e0��e (e0 abuts with andimmediately precedes e).

(34) a. λP0λe0λy1 . . . ynλπ1,π2∃e:resultant[çinchoation(HD(P0),HHD(y1), . . . HD(yn))(e0,

y1, . . . , yn,π1*arg1çinchoation:cos)∧result(e0,e)∧e0��e∧P0 (π2)(e)(x1) . . . (xn)]

33 Other discourse relations should be considered as well, but I cannot discuss them here.34 As perfect states are a subtype of type res-state (perf-state � res-state), no type clash arises

between res-realizes and çPR-RES if the latter receives an imperfective, perfect interpretation.35 cf. e:resultant in (34)—this forces a resultant state reading of the PP component of the PC

contributed by res-realizes; the selectional restriction borne by π is locally satisfied in (35b).

OUP UNCORRECTED PROOF – FIRST PROOF, 15/6/2015, SPi

204 Patrick Caudal

Comp. by: SatchitananthaSivam Stage : Proof ChapterID: 0002544163 Date:15/6/15Time:19:20:19 Filepath://ppdys1122/BgPr/OUP_CAP/IN/Process/0002544163.3dDictionary : OUP_UKdictionary 205

(35) a. λP0λe0λy1 . . . ynλπ1∃e:resultant[çinchoation(HD(P0),HD(y1), . . . HD(yn))(e0,y1, . . . ,

yn,π1*arg1çinchoation:cos)∧result(e0,e)∧e0��e∧P0 (π1)(e)(x1) . . . (xn)]

(λPλe000λx1 . . . xnλπ[res-realizes(e00,^P(x1, . . . ,xn),π*arg1res-realizes:res-

state)])b. ⇝λPλe0λx1 . . . xnλπ∃e:resultant[çinchoation(HD(P),HD(x1), . . . HD(xn))(e0,

x1, . . . ,xn,π*arg1çinchoation:cos) ∧result(e0,e)∧e0��e ∧res-realizes(e,^P

(x1, . . . ,xn),π)]

Applying (31) to (35b) in a Narration context yields the following (with n being theutterance time; note that the selectional restriction π*arg1

çinchoation:cos in (35b) islocally satisfied by çperfective-res, as it ranges over cos events).

(36) λPλx1 . . . xnλπ∃z:cos[çperfective-res(z)∧z�n∧∃e:res-state[çinchoation(HD(P),HD(x1), . . .HD(xn))(z,x1, . . . ,xn,π)∧result(z,e)∧z��e∧res-realizes(e,^P(x1, . . . ,xn),π)]]

In short, çPERFECTIVE-RES roughly denotes a context-dependent inchoation eventbringing about the resultant state denoted by res-realizes. Such an inchoation eventis not equivalent to a past perfective event description as denoted by, for example, thebase verb in the PS. Indeed, example (2) really means something akin to “Marsilebecame seated” (with a weak causal meaning)—and not “Marsile sat down” (with astrong causal meaning).

I have furthermore claimed that stronger SOE uses of the OF PC illustrated byexamples (24)/(57), involve a more clearly perfective breed of interpretation, albeit ofa pragmatic kind. I suggest that we adopt Caudal and Roussarie’s (2006) proposalthat these uses are triggered by conventionalized pragmatic enrichment processesassociated with the PC, to be distinguished from free, speaker-meaning basedpragmatic processes such as conversational implicatures (cf. Caudal 2012). Thegeneral intuition is that a rule such as (36) must introduce within the Logic ofInformation Packaging component of SDRT (roughly, the pragmatics) a propos-itional content KP, corresponding to a “causing event” propositional content P(x1, . . . ,xn)—with KP, as a proposition, being true prior to the run-trace ofçINCHOATION (as such inchoations of result states must correspond at least to theculmination part of an e:KP event).36 This means that subsequent causo-temporalreasoning in discourse can give us the desired temporal ordering for (24)/(25)normally associated with stronger SOE readings (this prevents the incorrect eventordering illustrated in Figure 9.1).

(37) ?(α,β)∧β:{çPR-RES(eβ,x1, . . . ,xn)∧res-realizes(s,^P(x1, . . . ,xn),π)}>KP

36 This also means that KP and the denotation of the OF PC have the same argument structuralproperties—this reflects a defining property of grammaticalized perfects.

OUP UNCORRECTED PROOF – FIRST PROOF, 15/6/2015, SPi

The passé composé in Old French and Modern French 205