Old French, Oxford University Press, Scandinavian languages, TP, Zaring, OS, Herschensohn, Germanic languages, Modern French, TECs, subject position, the Germanic Languages, Cornell University, Scandinavian, Expletive Subjects, verbal agreement, null subjects, Medieval French, H. Thr�nsson, Main clauses, Object Shift, Topic position, Scandinavian language, Old French Eric Mathieu, Quirky Subjects, grammatical constructions, nominative case, John Benjamins, Verb Movement, logical subject, Syntactic Change
On the Germanic properties of Old French Eric Mathieu 20.1 Introduction This paper shows that a certain cluster of properties found in a subset of North Germanic languages (e.g. Icelandic, and possibly Faroese) can also be found in Old French. In addition to V2 configurations, all the following properties are available: (i) Stylistic Fronting; (ii) Quirky subjects; (iii) Object Shift; and (iv) Transitive Expletive Constructions. Building on work by Cardinaletti and Roberts (2002), Dupuis (1989), Roberts (1993), previous of work of mine (Mathieu 2006a) has already established that Stylistic Fronting (SF, henceforth) was part of the inventory of grammatical constructions of Old French, while in a recent paper I also show that Old French had Quirky Subjects (Mathieu 2006b). In the latter paper, a correlation is established between the two constructions in that, if a language has SF, then it also has Quirky Subjects (though the opposite may not be true, e.g. the case of Modern German).1 Other correlations of the sort have been made in relation to Germanic languages. A case in point is Hiraiwa (2001) who shows that Germanic languages allowing SF also allow Object Shift (OS, henceforth) while Bobaljik and Jonas (1996) following Bures (1992) show that Germanic languages allowing Transitive Expletive Constructions (TECs, henceforth) also allow OS. In fact, it can easily be shown that Germanic languages allowing SF also allow TECs. The prototypical language belonging to that group is Icelandic, an Insular Scandinavian language (there is dialectal variation for the case of Faorese).2 Mainland Scandinavian languages, on the other hand, do not tolerate SF, Quirky Subjects, OS (of fullNPs) or TECs. Since Old French has SF, the typological prediction that is made is that it should also have both TECs and OS. The aim of this paper is to show that this prediction is indeed borne out. Whereas in Mathieu (2006a) and Mathieu (2006b) I concentrated on SF and Quirky subjects respectively, the present paper therefore focuses on TECs and OS in Old French, introducing not only new data, but also a new comprehensive analysis that accounts for the distribution of all properties aforementioned, namely SF, Quirky subjects, OS, and TECs. These are argued to be possible constructions in the grammar if the pronominal features of the verb are capable of checking the [D] feature of T0 independently of the [P] feature associated
with the pre-verbal position (cf. Holmberg's 2000 split EPP) and if a special Topic position (dubbed Top+P to differentiate it from the topic phrase to which topicalized elements raise to in V2 configurations) is available/accessible. If the EPP becomes unsplit, the special topic position is no longer available/accessible and all the constructions under review become obsolete. While the idea that OS might be productive in the grammar of Old French has been put forward before by Zaring (1998), the facts about TECs in Old French are not well-known. Although I introduce new data that strengthen Zaring's original insight, I nevertheless show that many of the examples that she introduces are cases, not of OS, but of scrambling (of the kind found in West German
ic languages). The influence of Germanic on what was to become French may have been through contact, first through the invasion of Gaul by the Francs, and second, by the Normans in the North-West. The initial causes for the splitting of the EPP and the creation of a special Topic position above TP might thus be external. The reason why SF, Quirky subjects, OS, and TECs disappeared from the grammar of French is because the EPP mutated from a complex form (a bundle of two organised features, [D] and [P]) into a simple form (where [D] and [P] are one). 20.2 Object Shift The aim of this section is to show that Old French had Object Shift. First, it is important to point out that Old French is a VO language (like Icelandic), not an OV language (like German). Although Latin was an OV language, Old French lost that feature very early on. This does not mean of course that OV orders were not possible, but it must be the case that they were derived from an underlying VO order. The example in (1), from around 1180, shows that the default order is VO. The object is underlined. (1) Et cil respont que il ne quiert and this-one reply.3SG that he not ask.3SG Avoir mie desaventure have.INF FORC misfortune `And he replies that he does not seek to have his own misfortune.' (Le Chevalier а la Charrette 2650-2651) When the object has shifted to the left, it is thus a case of OS. The claim that Old French had OS is not new. Zaring (1998) is the first author to have given an analysis of such a process in the language. She bases her conclusions on the behaviour of ce `this'. In order for her 2
argument that ce `this' can undergo Objet Shift in Old French, Zaring first establishes that ce in Old French is not a clitic element. We know that it is not a clitic because it can appear in first position in V2 constructions as shown by examples such as (2).3
(2) Li rois respont: `Ce sai
the king answer.3SG this know.1SG I well
`The king answers: `This, I know well.' (Zaring 1998: 320)
Next, Zaring reports that if a lexical verb is infinitival, direct object ce never follows that verb. Rather, it precedes the infinitive and follows any matrix material as (3) shows.
(3) Mes la reine ne peut ce croire... but the queen not can.1SG this believe.INF `But the queen could not believe this,...' (Zaring 1998: 321)
If the lexical verb is a past participle
, ce immediately precedes the participle as shown by (4).
(4) Sire, por coi avйs vos ce fait ? Lord for what have.2PL you this done `Lord, why have you done this?' (Zaring 1998: 321)
If the lexical verb is finite, ce (or its dialectal variant зou) occurs either in a post-verbal position (as expected) or in a pre-verbal position, following the subject, giving SOV order, as exemplified in (5). This pattern is attested almost exclusively in embedded clauses in Zaring's corpus.
(5) ... et quant li rois зou entendi, sus est saillis,... and when the king this hear.PAST.3SG up is leapt `...and when the king heard this, he leapt up...' (Zaring 1998: 3220)
The problem with the examples introduced by Zaring is that, on the one hand, they involve compound (3) and infinitival tenses (4), and on the other, embedded clauses (5). These contexts are not possible OS environments in Scandinavian languages. These facts have been captured under Holmberg's (1999) generalization: OS is only possible when the verb has
moved to C0. In view of these facts, I suggest that most, if not all, of the examples introduced by Zaring are cases of scrambling.4 Scrambling is very similar to OS in that an object has been raised passed a negative adverb. However, contrary to OS, it is possible to scramble an object when the verb has not raised to C0. In sum, since Object movement in Old French is possible in embedded clauses and when the verb has not raised to C0, it appears that the examples introduced by Zaring (1998) are not clear cases of OS. Another problem for Zaring's analysis is that she only gives examples where an object has shifted to the left but with no adverb present in the structure. Therefore, it is not easy to check whether the object has in fact remained within the VP or whether it has actually undergone movement to the edge of the VP. Clearer examples would need to involve VP adverbs. Finally, in the examples that she gives, the object has often undergone long Object movement (she does in fact make a distinction between short OS and long OS), that is, the object has raised past the finite verb. These cases do not appear to be cases of OS either. This does not mean, however, that the operation dubbed OS was not available at all in Old French. In view of examples such as (6a and b), introduced by Arteaga (1998), I would like to argue that Old French did have OS. In (6a) the object sa poverte `his poverty' appears to the right of the adverb while in (6b) the object un hannap `a goblet' surfaces to the left of the adverb. What (6b) thus clearly shows is that the object has been moved to the edge of the VP.
(6) a. b.
Iloc deduit ledement sa
there live.PAST.3SG joyfully his.FEM.SG.OBL poverty.FEM.SG.OBL
`There he joyfully lived his poverty.'
(La vie de Saint Alexis, year 1050, line 261, in Arteaga 1998)
donna un hannap
his.MAS.SG.OBL companion.MAS.SG.OBL give.PAST.3SG a goblet.MAS.SG.OBL joyfully
`He joyfully gave his companion a goblet.'
(Dits C, line 233, in Herslund 1980:10 and Arteaga 1998)
Additional examples collected via Frantext by myself can be found in (7) and (8). In (7) the nominal home `man' is below the adverb laienz `here' whereas in (8) the nominal appears above laienz.
(7) il n'avoit
there not-have.PAST.3SG here man.MAS.SG.OBL who can.PAST.3SG speakINF `there was no man here who could speak.' (La Quкte de Saint-Graal, year 1220, p.15)
(8) Mes il n'ot
but it not-have.PAST.3SG knight.MAS.SG.OBL here
par ou il i entra
that know.PAST.3SG by where he there enter.PAST.3SG `but there was no knight here who knew from where he had come in.'
(La Quкte de Saint-Graal, year 1220, p. 7)
One might object that these examples are expletive constructions with post-verbal subjects, rather than involving objects. However, note that the nominal is in the oblique, not the nominative case. Old French showed two types of expletive constructions: one in which the verb agreed with the post-verb nominal, with that nominal surfacing in the nominative (the case of Modern English), as illustrated in (9), and another in which the verb agreed with the expletive, while the post-verbal nominal is in the oblique case (the case of Modern French).
(9) a. b. c.
Mais ill i sont venu serjanz
but EXPL there be.3PL come servants.MAS.PL.NOM and riders.MAS.PL.NOM
`But there came servants and riders.'
(Parise, 2009-2010, in Buridant 2000:324)
Il morront maint vaillant chevalier
EXPL die.FUT.3PL many brave knights. MAS.PL.NOM
`There will die many brave knights.' (Gaydon, 8327, in Buridant 2000:324)
Il i corurent .vii. roi
et .xv. duc
EXPL there run.PAST.3PL seven kings.MAS.PL.NOM and fifteen dukes.MAS.PL.NOM
`There ran there seven kings and fifteen dukes.'
(Le Couronnement de Louis, AB, 631, in Buridant 2000:324)
To summarize section 20.2: I have shown that, in addition to SF and Quirky subjects, Old French has OS, and thus patterns with Icelandic which has all the properties just mentioned. Old French also had scrambling. The idea according to which Old French had both Object Shift and scrambling is in line with the view that OS is attested not only in Scandinavian languages, but also in languages like German and Dutch, except that German
and Dutch allow scrambling as well, while Scandinavian languages allow only OS (Bobaljik and Jonas 1996). In the next section, I add one more property that Old French shares with Icelandic: Transitive Expletive Constructions.
20.3 Transitive Expletive Constructions TECs are available in Icelandic and in German, (10a) and (10b) respectively, but not in languages like English and Danish, (10c) and (10d) respectively.5 TECs are constructions where an expletive appears in the subject position while both a subject and an object appear in the postverbal position.
(10) a. b. c. d.
Юaр hafa margir jуlasveinar borрaр bърing.
there have many Christmas.trolls eaten pudding
`Many Christmas trolls have eaten pudding.' (Bobaljik and Jonas 1996:209)
Es gessen einige Mдuse Kдse in der Kьche.
there eat some mice cheese in the kitchen
`There are some mice eating cheese in the kitchen.'
(Bobaljik and Jonas 1996:209)
*There ate many Christmas trolls a pudding.
*Der har nogen spist et жble. (Danish)
there has someone eaten an apple
`Someone has eaten an apple.' (Bobaljik and Jonas 1996:208)
A popular analysis of TECs and the parametric variation behind them comes from Bobaljik and Jonas (1996). These authors argue that languages with obligatory verb raising in non-V2 environments have a split IP structure (Icelandic, German, Dutch) and that conversely languages in which the verb remains in the VP have a simple IP (Mainland Scandinavian). Split-IP languages have more specifier positions in the IP complex than languages with a simple IP. OS is made possible because by raising to AgrO0 a specifier is created for the object to move to. If the verb does not move to AgrO0, then no specifier for the object can be licensed, therefore OS is not possible (this is thus the way Holmberg's generalization is accounted for/derived). The correlation between the availability of TECs and OS is made to follow from the fact that the verb has raised to AgrO0 in the first place. Once the verb raises to AgrO0, it can raise further to T0. Therefore, two subject positions
are created: Spec-AgrSP and Spec-TP. Because Spec-TP is licensed by movement of the verb to T0, TECs are possible:
all TECs require is two subject positions. In languages where the verb has not raised to T0, only one subject position is accessible and thus TECs are not possible. Bobaljik and Thrбinsson's analysis (Bobaljik and Thrбinsson 1998) is even more radical in that they argue that clauses in Icelandic and Mainland Scandinavian languages contain a different number of functional projections. For example, Danish realizes tense and agreement in a single projection, whereas Icelandic projects separate tense and agreement phrases. This is correlated to the fact that in Danish there are no separate morphemes for tense and agreement whereas in Icelandic, there is one morpheme for tense and another for agreement (Thrбinsson 1996). Although Bobaljik and Jonas (1996) and Bobaljik and Thrбinsson (1998) address the correlation between V to I movement and the grammaticality of TECs, they do not discuss the relevance of V2. V2 appears to be a crucial factor for an account of the difference between languages that allow TECs and OS, on the one hand, and those, on the other, which do not. If all is needed for a language to have TECs and OS is V to I movement, then Modern French should exhibit both operations, since that language has V to I movement (Pollock 1989), when in fact TECs are clearly not available in its grammar. One could argue that what is really needed for the licensing of specifiers is rich agreement (Modern French has poor verbal agreement). However, the theory will face problems with languages like Italian and Spanish which clearly have V to I movement and rich agreement. Yet, these languages do not have TECs or OS. Another problem for Bobaljik and Thrбinsson is the following. If, as they argue, rich agreement correlates with two separate morphemes for tense and agreement (with thus two separate specifiers created), then not only languages like Italian, Spanish, but even Modern French qualify. The Modern French future, for example, shows that one morpheme denotes tense (-er) while another denotes agreement (-ai), thus we get je parlerai `I will speak'. Pollock's (1989) analysis, whose two subject positions thesis is adopted by Bobaljik and Jonas (1996) and Bobaljik and Thrбinsson (1998), should automatically qualify French as a TEC and OS language. In sum, French shows that it cannot be the sole presence of TP as an extra subject position that allow TECs to be licensed. It cannot be the sole presence of CP either because of the fact that Mainland Scandinavian languages do not have TECs. I want to adopt Koeneman and Neeleman's (2001) proposal according to which only if both projections are present is it possible to generate TECs. At least two projections are required and these can be taken to be Spec-TP and Spec-CP. In order to have TECs and OS, a language thus needs not only V to I, 7
but also I to C movement. As Vikner (1990, 1995) observes there is a correlation between verb movement and the grammaticality of TECs in languages that have both V to I and V2. Old French fits the bill: it has both V to I and V to C. Therefore, it is expected that Old French has TECs and the prediction is borne out as (11) and (12) show. On the other hand, Modern French only has V to I movement, therefore it is expected that the language does not have TECs, as (13) shows.
(11) Il nel gari
EXPL not-him protect.3SG his.MAS.SG.NOM hauberk.MAS.SG.NOM white.MAS.SG.NOM
`His white hauberk didn't protect him.'
(Le Brut de Munich, 1775, in Arteaga & Herschensohn 2003:5)
(12) Li chastiaus
the castle.MAS.SG.NOM of-which EXPL spoke.3PL many.FEM.PL.NOM people.FEM.PL.NOM
`The castle that many people spoke about.'
(Montreuil, line 9312, in Arteaga & Herschensohn 2003:5)
(13) *Il mangиrent deux enfants un gвteau au chocolat. it eat.PAST.3PL two children a cake at-the chocolate `Two children ate a chocolate cake.'
In (11) and (12), the verb agrees with the nominative post-verbal logical subject, not with the expletive. This is clear in (12). The fact that the post-verbal subject is in the nominative case in (11) also shows that agreement is with the verb, since in later stages of French, the postverbal logical subject shows up in the accusative. To summarize so far: we have established that not only SF and Quirky subjects are available operations in the grammar of Old French (Mathieu 2006a and b), but so are OS (and scrambling) and TECs. A parametric account is needed to explain these facts and to differentiate Old French from Modern French, the latter not allowing any of these constructions. This parameter will also explain the differences between Insular Scandinavian languages, on the one hand, and Mainland Scandinavian languages, on the other.
20. 4 The analysis
Based on the fact that SF in Old French can target two elements, one XP and one head (in that order), I proposed in Mathieu (2006a) that SFronted elements in Old French move to a special projection dubbed Top+P. The SFronted XP raises to the specifier position of Top+P while the SFronted head raises to Top0+. I thus argued that SF is not movement to (Holmberg 2000), but through Spec-TP. In order to reach Spec-Top+, an XP must pass through Spec-TP, which must in that case be empty in order to function as an escape hatch. This is how the connection between the possibility of SF and subject gaps is accounted for. The solution avoids the inconvenience of postulating movement of phonological matrices into Spec-TP as in Holmberg (2000). Moreover, we avoid movement of heads to specifier positions, an operation that violates one of the central tenets
of generative grammar. My proposal nevertheless relies on the idea first proposed by Holmberg (2000) that the EPP can undergo feature fission between [D] (a categorial feature) and [P] (a feature requiring visibility, i.e. a specifier to be filled), with the added twist that: (i) the EPP need not undergo fission (to account for the optionality of SF); (ii) the features [D] and [P] may not necessarily come packaged as a bundle, [D] can appear on T0 while [P] can be on Top+ (this builds on work from Ritter 1992, 1993, 1995, Taraldsen 1994, Sigursson 1996, Bejar 2003 where features do not come packaged as unorganised bundles, but follow a feature hierarchy instead). (14a) gives the structure for Old French main clauses while (14b) introduces the configuration for Old French embedded clauses. Following Roberts and Roussou (2002), I assume that in V2 contexts the fronted XP is in the specifier of a Topic position, the verb in Fin0 and the subject in Spec-TP.6 In embedded clauses, FinP is not present (the verb does not raise to C0 in such environments in Old French). Instead, complementizers appear in Force0. 7
(14) a. b.
[TopP [Top+P [FinP [TP]]]] [ForceP [Top+P[TP]]]
Main clauses Embedded clauses
Turning now to the correlation between the availability of SF and the availability of Quirky subjects, I follow previous work of mine (Mathieu 2006b). In that article, I account for the fact that oblique non-pronominal subjects in Old French are in complementary distribution with SFronted elements by assuming that they both target the Spec-Top+P position. Consider (15).8
*que [de la honte] [a sa mere] ne chaut that of the shame to his mother not matter.3SG
b. * que [a sa mere] [de la honte] ne chaut c. 9que [de la honte] ne chaut [a sa mere] d. 9qu'[a sa mere] ne chaut [de la honte] `that the shame doesn't bother his mother.' The proposal is thus that if Top+P is not available then Quirky subjects are not possible in a given language (the relevant case features of Quirky subjects are not enough). Language change can lead to a situation where both SF and Quirky subjects disappear. Fischer (2004: 208) has recently argued that in Mainland Scandinavian languages oblique subjects and SF disappear roughly at the same time as a consequence of the loss `of the extra functional material', namely the higher projection that hosts both SFronted elements and Quirky subjects. French corroborates this idea, since the older stage of the language had both SF and Quirky subjects, but the modern alternative does not have any of these constructions. Next, I would like to argue that the special position I have postulated for SFronted elements and Quirky subjects is where expletives are merged in Old French. This is a claim I have not made before and which is therefore one of the new central hypotheses of the present paper. Rather than being directly merged in Spec-CP (i.e. Spec-TopP), expletives are first merged in Spec-Top+P and only then are they raised to Spec-CP (i.e. Spec-TopP). Old French il is thus a kind of `expletive topic' (see the discussion about Icelandic in section 20.3). As in the case of SFronted elements, the expletive is an asserted topic while the element that has remained behind (here the contentful subject) is focused. This is the effect an impersonal construction with an expletive usually has. This is uncontroversial. What is new, however, is the conclusion that stems from the logic followed presently: TECs have the same format as SF constructions. Consider (16). (16) [TopP Ili Top0 [Top+P ti [FinP Fin0 nel + gari [TP ses osbers blancs T0]]]] Il nel gari ses osbers blancs (cf. 35) Expletives can never appear post-verbally when an element other than the expletive appears in Spec-TopP (the same generalization holds with Quirky Subjects). Generally, impersonal il in matrix clauses is left unexpressed if not found in CP in Old French (Arteaga and Herschensohn 2003). (17) is thus ungrammatical. (17) *Dans leur chambre mangиrent il deux enfants un gвteau au chocolat. 10
in their bedroom eat.PAST.3PL EXPL two children a cake at-the chocolate `Two children ate a cake in their bedroom.'
There are cases, as witnessed by (18), however, where impersonal il appears post-verbally with Spec-CP filled by another element. However, as Arteaga and Herschensohn (2003) correctly point out, these examples involve agreement of the verb with the pronoun. Therefore, the construction is rather different from those introduced in (9) where the postverbal subject does not appear in the nominative, but in the accusative. What we do not find is equivalents of (18) with the associate of the expletive in the nominative.
(18) Si ot
il assez en la place barons et
thus have.PAST.3SG EXPL many in the place barons.MAS.PL.OBL and
qui la voldrent retenir
knights.MAS.PL.OBL who her want.PAST.3PL retain.INF
`Thus there were many barons and knights in the place who wanted to retain her.'
(La Quкte de Saint Graal, year 1220, p. 106, 16, in Vance 1997 :234 and Arteaga and
Since Vance (1989) and Roberts (1993), the post-verbal pronouns in cases such as (18) have been treated as clitics adjoining to C0. They considered nominative pronouns (they agree with the verb), while the expletive surfacing in constructions where the post-verbal subject is nominative and the verb agrees with that post-verbal subject instead of the pronoun, is not a nominative pronoun. So far, I have accounted for the availability of three out of the four properties that Old French and Icelandic share, namely, SF, Quirky subjects, and TECs. These constructions are all made possible because a special position in a split CP layer is made available. That special position is Top+P, a special topic position in the left periphery of the clause that seems to be available if V2 is available (the reverse is not necessarily true, since Mainland Scandinavian languages, as has already been mentioned, have V2, but no SF). The other crucial factor involved in the constructions under review is the splitting of the EPP, with the scattering of the two features that make up the EPP on two different heads. The remaining property to account for is the case of OS. We cannot assume like Bobaljik and Jonas (1996) that the object raises to the specifier of AgrOP since agreement projections are no longer part of the theoretical apparatus of minimalism. On Bobaljik and
Jonas' account, movement of the verb to AgrO0 forces the creation of a specifier (hence the availability of OS), and since the verb has raised to AgrO0 in the first place, it can raise further to T0, creating another subject position by way of Spec-TP in addition to Spec-AgrSP independently available. Chomsky (1995) abandons AgrSP and AgrOP for the simple reason that they have no relevance at LF, agreement on the verb being on this view devoid of meaning. In order to account for OS, he argues that v* can have more than one specifier. v* is thus assigned an EPP feature, but only if this has a semantic effect on the outcome (optional rules are outcome-dependent). On this view of OS, it is not clear how the correlation between the availability of SF, Quirky subjects, and TECs on the one hand and OS on the other can be made to follow. I would like to propose that the correlation in question stems from the fact that after movement to the outer specifier of v*, the object further raises to the specifier position of TP (this movement in essence thus replaces the operation DISL proposed by Chomsky for Icelandic). When the EPP has been split into [D] and [P], recall that [D] probes from T0 while [P] appears on a different head, namely Top+0. Suppose the shifted object satisfies the peripheral feature associated with T0 (recall that TP is a strong phase in Old French), but that the subject or another potential goal raises to the specifier of Top+P (which is basically another subject position, in the large sense of the term) satisfying the [P] feature associated with T0. Finally, let us turn to the question as to why SF, Quirky subjects, OS and TECs are no longer possible in Modern French. I would like to tie the disappearance of SF, Quirky subjects, TECs and OS to a change in the feature make-up of the EPP. Old French had a split EPP while this was lost at some point. The loss of the split EPP goes hand in hand with the loss of strong agreement. This idea is connected to the popular view about the loss of SF in Mainland Scandinavian. It has often been claimed that in languages like Old Swedish the loss of V to T movement and the loss of SF took place simultaneously in the 16th and 17th centuries (Falk 1993:184). The generalization is that languages like Danish and Swedish do not have V to T movement, thus SF is not available whereas Icelandic has V to T movement, thus SF is a possibility in that language. According to Holmberg (2000), once the verb does not raise to T0, it cannot check the [D] feature associated with T0, thus SF is no longer possible. Instead, the subject of the sentence checks both the [D] and the [P] feature of T0. The problem with this idea is that since Modern French has not lost V to T movement (Pollock 1989), but crucially lacks SF, this account needs a slight revision. What appears to be essential for SF is that, regardless of whether the verb has raised to T0, the verbal agreement should have the relevant pronominal properties so that null subjects 12
are possible. Once verbal agreement lost its pronominal properties (null subjects are not possible in Modern French), SF was no longer available. Since the [P] feature of EPP+ is dependent on [D] when both features are scattered on different heads, it becomes no longer possible for the EPP to be split. Since Quirky subjects and TECs rely on a split EPP and a Top+P position, these two constructions also disappeared form the grammar of French once the EPP could no longer be split. The absence of V to I movement has also been called for by Bobaljik and Jonas (1996) as an explanation of the lack of TECs in Mainland Scandinavian. In section 20.3 we adopted Koeneman and Neeleman's (2001) proposal according to which only if both TP and CP are present is it possible to generate TECs (cf. Vikner's Generalization). In the present theory, this can be translated as tying the split of the EPP together with the availability of TopP with the availability of V2. For SF and for the other constructions under review in this paper to be operative in a given language, it is not sufficient to have the possibility of [D] feature checking by the verb. What a language must have at its disposal is the availability of the EPP+ feature, i.e. the case where [D] and [P] are not packaged as an unorganised bundle. This option together with the creation of Top+P only seems possible if CP is not only split, but accessed via verb movement and topicalization (i.e. V2). 20. 5 Conclusion This article has shown that Old French shared with Insular Scandinavian languages, not only V2 configurations, in addition to Stylistic Fronting and Quirky subjects (as shown in previous work of mine), but also Object shift and Transitive Expletive Constructions. A proposal was given for this cluster of properties: in Old French the EPP was split between a [D] and a [P] feature and a special Topic position was available above TP. The split EPP became unsplit through time and the special Topic position responsible for all the constructions under review was no longer available/accessible in later stages of the language. References Arteaga, D. (1998). `On Optionality in the Minimalist Program and the Old French Word Order', in J. Lema, and E. Treviсo (eds.), Theoretical Research on Romance languages: Selected Papers from the Twenty-sixth Linguistic Symposium on Romance Languages. Amsterdam: John Benjamins, 2334. Arteaga, D., and Herschensohn, J. (2003). `Case, Agreement, and Epletives: A ParametriC Diff
erence in Old French and Modern French', in J. Auger, J. Clancy Clements, and B. 13
Vance (eds.), Contemporary Approaches to Romance Linguistics. Amsterdam: John Benjamins, 115. Barnes, M. (1987). `Some Remarks on Subordinate-clause Word-order in Faroese', Scripta Islandica 38: 335. Bejar, S. (2003). Phi-syntax: A Theory of Agreement. PhD thesis
, University of Toronto
. Bobaljik, J., and Jonas, D. (1996). `Subject Positions and the Roles of TP', Linguistic Inquiry 27: 195-236. Bobaljik, J., and Thrбinsson, H. (1998). `Two Heads aren't always better than One', Syntax 1: 3771. Bures, T. (1992). Re-cycling Expletive (and other) Sentences. Ms. MIT. Buridant, C. (2000). Grammaire nouvelle de l'ancien franзais. Paris: Sedes. Cardinaletti, A., and Roberts, I. (2002). `Clause Structure and X-second', in G. Cinque (ed.), Functional Structure in DP and IP: The Cartography of Syntactic Structures, Volume I. Oxford: Oxford University
Press, 123166. Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: MIT press
. Chomsky, N. (2001). `Derivation by Phase', in M. Kenstowicz (ed.), Ken Hale: A Life in Language. Cambridge, MA: MIT Press, 152. Dupuis, F. (1989). L'expression du sujet dans les subordonnйes en ancien franзais. PhD thesis, Universitй de Montrйal. Eythуrsson, T. (2000). `Dative vs. Nominative: Changes in Quirky Subject in Icelandic', Leeds Working Paper
s in Linguistics 8: 2744. Eythуrsson, T. (2002). `Changes in Subject Case-marking in Icelandic', in D. Lightfoot, Syntactic Effects of Morphological Change. Oxford: Oxford UNIVERSITY PRESS
, 196212. Eythуrsson, T., and Barрdal, J. (2005). `Oblique Subjects: A Common Germanic Inheritance', Language 81: 824881. Falk, C. (1993). Non-referential Subjects in the History of Swedish. PhD thesis, University of Lund. Fernбndez-Soriano, O. (1999). `Two Types of Impersonal Sentences in Spanish: Locative and Dative Subjects', Syntax 2: 101140. Fischer, S. (2004). `Stylistic Fronting: A Contribution to information structure
', Ms. Humboldt-University of Berlin. Foulet, L. (1928). Petite syntaxe de l'ancien franзais. Paris: Йditions Champion. Herslund, M. (1980). Problиmes de l'ancien franзais. Complйments datifs et gйnitifs. Revue Romane Series, number 21. University of Copenhagen
Hiraiwa, K. (2001). `EPP: Object Shift and Stylistic Fronting in Scandinavian', in K. Megerdoomian, and L. Bar-el (eds.), West Coast
Conference in Formal Linguistics 20 Proceedings. Somerville: Cascadilla Press, 290303. Holmberg, A. (1999). `Remarks on Holmberg's Generalization', Studia Linguistica 53: 139. Holmberg, A. (2000). `Scandinavian Stylistic Fronting: How any Category can become an Expletive', Linguistic Inquiry 31: 445483. Jonas, D. (1995). Clause Structure and Verbal Syntax on Scandinavian and English. PhD thesis, Harvard University
. Koeneman, O., and Neeleman, A. (2001). `Predication, Verb Movement and the Distribution of Expletives', Lingua 111: 189233. Labelle, M., and Hirschbьhler, P. (2005). `Changes in Clausal Organization and the Position of Clitics in Old French', in M. Batllori, and F. Roca (eds.), Grammaticalization and Parametric Change. Oxford: Oxford University Press, 6071. Mathieu, E. (2006a). `Stylistic Fronting in Old French', Probus 18: 219266. Mathieu, E. (2006b). `Quirky Subjects in Old French', Studia Linguistica 60: 282312. Moore, J., and Perlmutter, D. M. (2000). `What does it take to be a Dative Subject?', natural language
and Linguistic Theory 18: 373416. Pollock, J-Y. (1989). `Verb Movement, Universal Grammar and the Structure of IP', Linguistic Inquiry 20: 365424. Ritter, E. (1992). `Cross-linguistic Evidence for Number Phrase', Canadian Journal
of Linguistics 37: 197218. Ritter, E. (1993). `Where's Gender?', Linguistic Inquiry 24: 795803. Ritter, E. (1995). `On the Syntactic Category of Pronouns and Agreement', Natural Language and Linguistic Theory 13: 405443. Roberts, I. (1993). Verbs and Diachronic Syntax: A Comparative History of English and French. Dordrecht: Kluwer. Roberts, I., and Roussou, A. (2002). `The Extended Projection Principle as a Condition on the Tense Dependency', in P. Svenonius (ed.), Subjects, Expletives, and the EPP. Oxford: Oxford University Press, 125155. Sigurрsson, H. (1992). Verbal Syntax and Case in Icelandic in a Comparative GB Approach. PhD thesis, The Institute of Linguistics, University of Iceland. Sigursson, H. (1996). `Icelandic Finite Verb Agreement', Working Papers in Scandinavian Syntax 57: 146. Taraldsen, K. T. (1994). `Reflexives, Pronouns, and Subject/verb Agreement in Icelandic and 15
Faroese', Working Papers in Scandinavian Syntax 54: 4358. Thrбinsson, H. (1979). On Complementation in Icelandic. New-York: Garland. Thrбinsson, H. (1996). `On the (non)-Universality of Functional Categories', in A. Werner, S. D. Epstein, H. Thrбinsson, and J-W. Zwart (eds.), Minimal Ideas: Syntactic Studies in the Minimalist Framework. Philadelphia/Amsterdam: John Benjamins, 253281. Vance, B. (1989). Null Subjects and Syntactic Change in Medieval French. PhD thesis, Cornell University
. Vance, B. (1997). Syntactic Change in Medieval French. Verb-second and Null Subjects. Dordrecht: Kluwer. Vikner, S. (1990). Verb Movement and the Licensing of NP Positions in the Germanic Languages. PhD thesis, University of Geneva. Vikner, S. (1995). Verb Movement and Expletive Subjects in the Germanic languages. Oxford: Oxford University Press. Zaring, L. (1998). `Object Shift in Old French', in A. Schwegler, B. Tranel, and M. UribeEtxebarria (eds.), Romance Linguistics: Theoretical Perspectives. Amsterdam: John Benjamins, 319332. Zubizarreta, M.-L. (1998). Prosody, Focus, and Word Order. Cambridge, MA: MIT Press. 1 Although, Moore and Perlmutter (2000) argue that German does not have Quirky subjects, but inverted objects, Eythуrsson and Barрdal (2005) make a good case for the idea that German has Quirky subjects after all. Whether German does or does not have Quirky subjects is, however, orthogonal to the point that I am presently making. 2 For example, not all speakers accept TECs (cf. Jonas 1995). There are also differences between Icelandic and Faroese with regard to SF. Whereas DPs can be SFronted in Faroese (Barnes 1987) , only abstract denoting DPs can undergo the stylistic operation in Icelandic (Holmberg 2000). Finally, whereas Quirky subjects are still very productive in Icelandic, they are in the process of being lost in Faroese (Eythуrsson 2000, Eythуrsson 2002). 3 Further evidence for the idea that ce `this' is not a clitic is provided by the facts that: i) it can also occur in isolation; ii) it can be modified by tout `all'; iii) it can be conjoined. All these properties are known to be impossible for clitic pronouns. 16
4 There is an alternative proposal according to which left dislocation of objects in Old French compound and infinitival clauses is a case of SF. This is what I argue in fact in Mathieu (2006a). However, in that paper I do not deal with scrambling or OS, and the solution that is provided in the present paper in terms of scrambling stems from the logic developed here. 5 Although, if the subject is shifted to the right edge of the clause, the sentence is improved as shown in (i) this observation is attributed to Richard Kayne by Chomsky (2001), but see also footnote 17, p. 208, in Bobaljik and Jonas (1996) : (i) ??There ate a pudding many Christmas trolls. 6 Although Old French is not explicitly discussed in Roberts and Roussou (2002), it is clear that the ideas they develop for V2 languages can be extended to Old French (see Labelle and Hirschbьhler 2005 for such an extension). 7 It must be noted that the Top+P position is not associated with presupposed, but asserted topics. The process behind SF is one that allows an element to simply get out of the way, as it were, so that the most embedded element becomes focalized. The process is thus akin to what Zubizarreta (1998) calls P-movement, except that according to the analysis developed here movement of Stylistic elements happens in the narrow syntax, not at PF. There is independent evidence that SF is relevant for narrow syntax from the behaviour of auxiliaries. Although these are potential candidates in terms of ccommand for raising to Top+0, they nevertheless remain in situ leaving other elements to be SFronted. I follow Holmberg (2000) in viewing this fact as evidence that narrow syntax is where SF is derived. 8 The situation with pronominal Quirky Subjects is different: since they are clitics they adjoin to T0 (and possibly further up to Top+0) directly and thus do not move to any specifier position. 17