Hostname: page-component-74656f9884-g96k6 Total loading time: 0 Render date: 2026-11-10T10:19:00.450Z Has data issue: false hasContentIssue false
  • English
  • Français

Satisfying V2 in early Romance: Merge vs. Move1

Published online by Cambridge University Press:  19 June 2008

ADAM LEDGEWAY
ADAM LEDGEWAY*
Affiliation:
University of Cambridge & Downing College
*
Author's address: Downing College, Regent Street, Cambridge CB2 1DQ, U.K.anl21@cam.ac.uk
Rights & Permissions [Opens in a new window]

Abstract

In early Romance the finite verb raises in root clauses to the vacant C position, a movement operation frequently accompanied in turn by fronting of a pragmatically salient constituent to a preverbal position within the C-domain. Although this structural characterization of early Romance V2 word order overwhelmingly accounts for the core cases, there remains a smaller number of contexts in which the V2 effect appears to be satisfied by other means. More specifically, in addition to the move option, we argue that the V2 constraint may be met in early Romance by the Merge option: in the former case this involves movement of the finite verb to the empty C head, whereas in the latter case the V2 requirement is met by merging sì (<sic ‘thus, so’) directly in C°. This analysis radically departs from traditional analyses of Romance sì, which characterize it either as a coordinating conjunction or, in recent generative analyses, as a pleonastic phrasal category associated with a specifier position within CP. Our proposed analysis makes some interesting and novel predictions both about the rather restricted distribution of sì in Old Neapolitan, many aspects of which remain unexplained under the alternative accounts, and the C-related functional positions involved in medieval Romance V2.

Information

Type
Research Article
Information
Journal of Linguistics , Volume 44 , Issue 2 , July 2008 , pp. 437 - 470
Copyright
Copyright © 2008 Cambridge University Press

1. Introduction

Superficially, the word order of early Romance varieties appears to be less constrained than that of modern Romance, an observation which has led many linguists (cf. Vanelli Reference Vanelli and Harro1986, Reference Vanelli and Lorenzo1998; Adams Reference Adams1987; Dupuis Reference Dupuis1989; Fontana Reference Fontana1993, Reference Fontana, Ans van and Nigel1997; Roberts Reference Roberts1993; Benincà Reference Benincà, Battye and Roberts1995, Reference Benincà, Raffaella, Héctor, Elena and Paul2006; Lemieux & Dupuis Reference Lemieux, Dupuis, Battye and Roberts1995; Ribeiro Reference Ribeiro, Battye and Roberts1995; Vance Reference Vance1997; Salvi Reference Salvi2004) to argue that early Romance varieties are Verb Second (V2) languages. More recently, the same conclusion has been reached for Early Neapolitan (see Ledgeway Reference Ledgeway, Anna Laura and Arturo2007) where, on the basis of the evidence of a 50-page sample from the early 14th-century prose text Libro de la destructione de Troya (De Blasi Reference De Blasi1986: 47–97),Footnote 2 the finite verb in root clauses is shown to raise to the vacant C position, a movement frequently (though not necessarily) accompanied in turn by the fronting of one or more pragmatically salient constituents to a preverbal position within the C-domain, as in (1):

  1. (1) [CP [Spec sì fuorti cuolpi] [C′ li donava [IP … tli donava tsì fuorti cuolpi]]]  such strong blows to-him=he-gave‘he gave him such strong blows (with his sword)’ (66.12)

Although this structural characterization of Old Neapolitan word order overwhelmingly accounts for the core cases, there remains a smaller number of contexts in which the V2 effect appears to be satisfied by other means. More specifically, in addition to the Move option, the V2 constraint may be met in Old Neapolitan, and in other early Romance varieties too, by the Merge option. As we have just seen, in the former case this involves movement of the finite verb to the empty C head (cf. (1)), whereas in the latter case the V2 requirement is met by externally merging sì (<Latin sic ‘thus, so’) directly in C°, as in (2):

  1. (2) [CP [Spec spissi cuolpi mortali] [C′ sì [IP le dava tspissi cuolpi mortali]]]  frequent blows mortal sì to-him=he-gave ‘he struck him with many deadly blows’ (133.36)

In what follows, we therefore interpret V2 as a structural requirement on the C-related Fin head which, when finite and typically in root contexts, must be given PF-realization either through the Merge option (viz. lexicalization of sì) or the Move option (viz. finite verb raising). On this view, our analysis of Old Neapolitan sì departs radically from traditional analyses of Romance sì (also variously spelt sí and si), which characterize it either as a coordinating conjunction/connective (cf. Meyer-Lübke Reference Meyer-Lübke1899: sections 651, 653; Schiaffini Reference Schiaffini1926: 290–293; Rohlfs Reference Rohlfs1969: 760; Skårup Reference Skårup1975; Marchello-Nizia Reference Marchello-Nizia1985: 10ff.; Fleischman Reference Fleischman1991: 262ff.), or, in recent generative analyses, as a pleonastic phrasal category associated with the specifier position of CP or FocP (see section 3). As will emerge from the discussion below, our proposed analysis makes some interesting and novel predictions both about the rather restricted distribution of sì in Old Neapolitan and early Romance more generally, many aspects of which remain unexplained under the alternative accounts,Footnote 3 and the C-related functional positions involved in medieval Romance V2.

2. Old Neapolitan: a V2 language

2.1. Quantitative analysis

An examination of the statistical evidence from the relevant 50-page textual sample of Libro de la destructione de Troya reveals that Old Neapolitan can be broadly described as a V2 language, inasmuch as 42.9% of all root clauses were found to be superficially V2, whereas V3 and V1 orders only accounted for 24.3% and 22.3% of all root clauses, respectively.Footnote 4 Clearly, however, such statistical considerations do not provide conclusive evidence, especially in view of the other competing orders. Nevertheless, it is still possible to maintain a V2 analysis, if V2 is not understood as a superficial descriptive label, but as a syntactic constraint which requires the finite verb in root clauses to raise to the otherwise vacant C position (or, better, Fin°). In turn, this verb movement operation is generally, though not necessarily, accompanied by a further movement operation which fronts one (or more) pragmatically salient constituent(s) to a specifier position of one of the available functional projections FocP, TopP or FrameP contained within the C space, where the fronted constituent(s) receive(s) a (contrastively/informationally) focalized or thematicized interpretation. Thus, whether the verb superficially occurs in, for example, first, second, third or fourth position, technically the clause always qualifies as V2, inasmuch as the verb invariably raises to Fin°, the sole necessary structural reflex of the V2 parameter (Benincà Reference Benincà, Battye and Roberts1995: 333, 338f.).

Interpreting V2 as the result of verb movement to the vacant (lowest) C position, we expect V2 to be blocked, or at the very least more restricted, in embedded clauses since the C position is already filled by an overt complementizer. Indeed, contrary to root clauses, superficial V2 in our sample is not the most frequent word order in embedded clauses (41.4%), but comes second, albeit a close second, to V1 (44.7%). Thus, although V2 admittedly continues to prove a common superficial word order even in embedded contexts, an issue to which we return in section 4.1.2.3, what is remarkable is that the frequency of V1 orders has doubled in relation to that found in root clauses, a somewhat surprising result if Old Neapolitan were not a V2 language. Under the proposed asymmetrical V2 analysis, however, the increased frequency of V1 in embedded contexts and the concomitant decrease in V3 (10.8%), V4 (2.3%), V5 (0.6%), and V6 (0.1%) orders are quite expected.

2.2. Constituent fronting

The high frequency of root V2 orders observed above might not necessarily be a surface effect of a V2 constraint, but could perhaps simply reflect a high percentage of root clauses in which some element, most notably the subject, precedes the verb, as is the case in an SVO language. Nonetheless, this possibility is excluded: out of a total of 350 V2 root clauses only 80 of these (22.8%) were found to be subject-initial. Rather, as is to be expected of a V2 language, the preverbal position is not a privileged subject position but, rather, constitutes a pragmatically salient position licensing thematic and rhematic interpretations (Vanelli Reference Vanelli and Harro1986: 256, 270, n. 2; 1998: 259–261; Lombardi & Middleton Reference Lombardi, Middleton, Dardano and Frenguelli2004; Lombardi Vallauri Reference Lombardi Vallauri, Dardano and Frenguelli2004; Salvi Reference Salvi2004: chapter 1; Benincà Reference Benincà, Raffaella, Héctor, Elena and Paul2006: 65), a position which is potentially available to all syntactic categories, irrespective of their syntactic function and their thematic relation to the predicate. For example, consider the following representative sentences which exemplify the preverbal position variously filled by a direct object (3a), an indirect object (3b), a non-subcategorized adverbial phrase (3c), and a predicative adjectival complement (3d):

  1. (3)

    1. (a) e [CP [Spec viechy] [C′ reduceva [IP treduceva tviechy ad etate iuvenile]]]and old she-brought-back to age juvenile‘and she could make the old young again’ (56.35)

    2. (b) [CP [Spec A questo] [C′ resposse [IP tresposse Iasone tA questo]]]  to this replied Jason‘Jason replied to this’ (60.12)

    3. (c) [CP [Spec depoy queste parole] [C′ ademandao [IP tademandao lo  after these words asked themessayo licencia tde poy queste parole]]]messenger permission‘following these words the messenger asked permission (to leave)’ (55.25)

    4. (d) [CP [Spec maraviglyoso] [C′ era [IP tera lo suono tmaraviglyoso]]]  marvellous was the sound‘the sound [of the clashing swords] was tremendous’ (71.17–18)

2.3. Informational focus and subject-verb inversion

The examples in (3) above also illustrate two other significant pieces of evidence that point to the V2 nature of Old Neapolitan. First, examples like (3a) highlight how non-clitic-doubled direct objects with thematic or rhematic interpretations are frequently found to occur in clause-initial position in Old Neapolitan. In (3a), for example, the fronted object is informationally new, introducing into the narrative a referent which has not previously figured in the discourse, giving rise to an example of what is generally known as informational focus (Lambrecht Reference Lambrecht1994: 201; Vanelli Reference Vanelli and Lorenzo1998: 80f., 84–86; Benincà & Poletto Reference Benincà, Poletto and Rizzi2004: section 3; Benincà Reference Benincà, Raffaella, Héctor, Elena and Paul2006: section 3.3). Although rhematic objects can equally occur in postverbal position, especially when in wide focus together with their associated predicate, they also frequently occur in preverbal position in all early Romance varieties – a strategy not available in an SVO language like Italian, where rhematic direct objects (whether under wide or narrow focus) are restricted to occurring in postverbal position and direct objects can only be fronted under particular pragmatic conditions (e.g. contrastive focus or clitic left-dislocation). From the observed contrast between Modern Italian and Old Neapolitan we can infer that fronting of rhematic direct objects in Old Neapolitan represents movement to SpecCP licensed by prior movement of the finite verb to C°, an operation which proves impossible in Italian where generalized I-to-C movement in declarative root clauses, namely V2, fails to obtain.

Secondly, examples like (3b–c) demonstrate that, whenever a constituent different from the subject is fronted, this gives rise to verb-subject inversion whenever the subject is overtly realized. Significantly, in these and similar examples of inversion the subject does not simply follow the verb but also precedes any other sentential constituents, witness the order subject (lo messayo)+direct object (licencia) in (3c). In a non-V2 language like Italian, by contrast, postverbal subjects generally follow their associated objects and other sentential constituents (see Benincà & Salvi Reference Benincà and Renzi1988: 125; Belletti Reference Belletti and Rizzi2004: section 3), and in such cases are typically associated with rhematic interpretations, whereas the postverbal subjects in (3b–c) are clearly thematic. Now, under the proposed V2 analysis, subject inversion in these examples follows straightforwardly: by assuming an underlying S–V–XP order (at least for non-unaccusative structures), verb movement to the C position followed by fronting of some postverbal constituent XP to SpecCP will invariably result in the subject immediately following the verb and preceding any other constituents.

2.4. V1 clauses

The V2 nature of Old Neapolitan is further confirmed by consideration of a subset of apparent V1 root declaratives, so-called presentative sentences, in which, in the absence of a theme, the whole clause typically receives under wide focus a rhematic interpretation (answering the question What happened?). From a syntactic point of view, such clauses prove particularly significant in that they can be assumed to preserve the underlying order of constituents, save the verb which, in accordance with the V2 rule, moves to the vacant C position. It follows that, superficially, presentative clauses in conjunction with transitive predicates are necessarily V1 in Old Neapolitan, since only the finite verb is available to raise to the C-space, thereby stranding the other arguments of the verb in their base positions and yielding VSO words orders like that in (4a). In non-V2 languages like Italian, by contrast, presentative constructions with transitive predicates only license SVO orders, as demonstrated by (4b). In such structures the subject occurs in the preverbal position (Benincà & Salvi Reference Benincà and Renzi1988: 119), since this position (standardly associated with the SpecIP position) is informationally neutral, insofar as it is not exclusively associated with licensing either thematic (given) or rhematic (new) interpretations:

  1. (4)

    1. (a) [C′ Attesta [IP lo dicto Ovidio poeta lo nascimento de li  attests the said Ovid poet the birth of theMirmidoni]]Myrmidons   ‘Ovid the poet gives testimony to the birth of the Myrmidons’ (49.1)

    2. (b) [IP [Spec Giorgio] [I' ha buttato l'aranciata nel lavandino]]  Giorgio has thrown the orangeade in-the sink‘Giorgio has thrown the orangeade in the sink’

3. Sì: specifier of CP

Having established in the previous section that, on a par with other early Romance varieties, Old Neapolitan should be considered a V2 language, we now turn to the standard generative analysis of Romance sì (see Benincà Reference Benincà, Battye and Roberts1995: 330, 333; 2006: section 3.3; Lemieux & Dupuis Reference Lemieux, Dupuis, Battye and Roberts1995: 94–97; Vance Reference Vance, Battye and Roberts1995: 184; Salvi Reference Salvi2001; Poletto Reference Poletto, Montserrat, Maria-Lluïsa, Carme and Francesc2005: section 13.4). Generally, sì is held to belong to a limited class of adverbial elements with very little or no semantic value (van Reenen & Schøsler Reference van Reenen and Schøsler1992, Reference van Reenen, Schøsler, Susan, Pieter van and Lene2000), including, among others, or ‘now’, e(t) ‘and’, Old French ainz/Old Venetian an ‘on the contrary’, French lors/puis/après/donc/mais ‘then/then/after/thus/but’, whose main role is seen much as a last resort mechanism (cf. Fleischman Reference Fleischman1991: 260, 275f.; van Reenen & Schøsler 1993: 626, n. 5; Benincà Reference Benincà, Battye and Roberts1995: 330, 333; Ribeiro Reference Ribeiro, Battye and Roberts1995: 114; Vance Reference Vance, Battye and Roberts1995: 184; Poletto Reference Poletto, Montserrat, Maria-Lluïsa, Carme and Francesc2005: 224). In particular, especially in potential V1 contexts, it is generally claimed that ‘the verb-second requirement is often satisfied by the presence of different adverbial elements preceding the inflected verb’ (Lemieux & Dupuis 1995: 95), which function to all intents and purposes as ‘filler words’ (Benincà Reference Benincà, Battye and Roberts1995: 330, 333). As such, sì is argued to be a phrasal category in complementary distribution with those thematic/rhematic constituents (lacking a resumptive clitic copy) that undergo fronting as part of the V2 constraint to a position variously associated, albeit tacitly (Fleischman Reference Fleischman1991: 276; van Reenen & Schøsler 1995: 404; Salvi Reference Salvi2001: 3f., 6), with SpecCP (Adams Reference Adams1987; Benincà Reference Benincà, Battye and Roberts1995: 333) or, in more recent analyses within a more richly-articulated CP structure, SpecFocP (Poletto Reference Poletto, Montserrat, Maria-Lluïsa, Carme and Francesc2005; Benincà Reference Benincà, Raffaella, Héctor, Elena and Paul2006: section 3.3). In support of this analysis, it is noted that, following sì, clitics occur in proclisis (cf. (2) above) on a par with other (non-clitic-doubled) fronted rhemes and themes (cf. proclitic li in (1) above), whereas elements situated in higher topic-related positions such as SpecTop and SpecFrame invariably trigger enclisis (cf. Salvi Reference Salvi, Laura and Alberto1991; 2004: chapter 4; Benincà Reference Benincà, Battye and Roberts1995: 336f.; Poletto Reference Poletto, Montserrat, Maria-Lluïsa, Carme and Francesc2005: 226; Ledgeway Reference Ledgeway, Anna Laura and Arturo2007: 131–134), as revealed by the enclitic nde ‘thereof’ in the clitic left-dislocation example (5):

  1. (5) [SpecTopP de queste toy promissiune] voglyonde essere certa  of these your promises I-want=thereof to-be certain‘I want to be certain of these promises of yours’ (60.21–22)

Assuming sì to lexicalize SpecCP/FocP, as schematicized below in (6), the core distributional properties of sì are thus argued to follow without further stipulation (for an overview, see Fleischman Reference Fleischman1991: 268–270; Salvi Reference Salvi2001: section 1):

  1. (6) FrameP … TopP … [CP [Spec sì] [C′ Vfinite … [IP … tvFinite …]]]

More specifically, sì is predicted to occur in complementary distribution with other fronted constituents and to occur in strict adjacency to the finite verb raised to its associated head position, from which it can be separated only by intervening clitics (Fleischman Reference Fleischman1991: 270; Benincà Reference Benincà, Battye and Roberts1995: 333; Salvi Reference Salvi2001: 3f.; Poletto Reference Poletto, Montserrat, Maria-Lluïsa, Carme and Francesc2005: 224f.). Consequently, if sì occupies the preverbal position, other constituents which could potentially be fronted to this position are forced to occur either in postverbal position or in a higher, hence topic-related, left-peripheral position (viz. SpecFrameP/TopP) above sì (Benincà Reference Benincà, Battye and Roberts1995: 333; van Reenen & Schøsler Reference van Reenen, Schøsler, Susan, Pieter van and Lene2000: 59; Salvi Reference Salvi2001: 6). This explains the frequent, albeit apparently contradictory, observations that sì ‘causes obligatory subject inversion’ (cf. (7a); Benincà Reference Benincà, Battye and Roberts1995: 333) and/or serves to highlight the thematicity/topicality of the constituent immediately preceding it (cf. (7b); Fleischman Reference Fleischman1991, Reference Fleischman, Marinel and Dieter1992; van Reenen & Schøsler Reference van Reenen and Schøsler1992, Reference van Reenen, Schøsler, Susan, Pieter van and Lene2000; Salvi Reference Salvi2001: 1):

  1. (7)

    1. (a) [CP [Spec Sì ] [C′ en fu [IP la reïne ten fu moult carrociee]]]  sì of-it=was the queen very upset‘The queen was very upset about it’ (Old French, Artu, 166, cited in Benincà Reference Benincà, Raffaella, Héctor, Elena and Paul2006: 64f.)

    2. (b) Et [TopP [Spec questa scienza, cioè logica]i [CP [Spec sì] [C′ àe [IP proi tàe tre parti]]]]and this science that-is logic sì has three parts‘And this science, that is logic, has three parts to it’ (Old Florentine, Brunetto Latini, Rettorica 44.2, cited in Salvi Reference Salvi2001: 1)

Although superficially attractive, an analysis of sì which reduces it to a pleonastic filler of SpecCP/FocP only accounts for a subset of its distributional properties and fails to provide an explanation for many other aspects of the syntax of sì. Below we outline the major weaknesses and inadequacies of the ‘standard’ account before putting forward our alternative analysis.

3.1. Inadequacies and weaknesses

Phrasal vs. head categorial status

We begin by observing that under the standard analysis sì is taken to be a maximal (phrasal) category (or, to borrow Salvi's (Reference Salvi2001: 2) terminology, a parola a pieno titolo ‘fully-fledged word’), since it comes to occupy the specifier of SpecCP/FocP (henceforth simply SpecFocP). Yet, it fails to display any of the typical properties of phrasal categories. For instance, unlike other phrasal candidates for the specifier of the FocP position, sì cannot occur in other positions within the clause but, rather, is restricted to occurring within the left periphery (Fleischman Reference Fleischman1991: 268).Footnote 5 Moreover, it neither allows (adverbial) modification nor can it take complements. Rather, it bears all the hallmarks of a non-projecting head category (cf. Romance pronominal clitics). This is all the more evident when one considers that under the standard analysis sì is assumed to be nothing more than an expletive employed as a last resort strategy to satisfy the V2 requirement, which, according to Poletto (Reference Poletto, Montserrat, Maria-Lluïsa, Carme and Francesc2005: 222, 224), presents all the characteristic properties of a heavily grammaticalized element: it is a functional element that displays severe semantic impoverishment and a very limited syntactic distribution, a well-known correlate of which is the categorial reduction from phrasal (viz. XP) to head (viz. clitic, affix) status (cf. Hopper & Traugott Reference Hopper and Traugott1993: 7).Footnote 6 Within the generative paradigm the most natural interpretation of such heavily grammaticalized functional elements as sì is thus to treat them as heads (X°), rather than as phrasal categories (XP).

Sì: a last resort strategy

A further problem inherent in the standard analysis concerns the proposal that sì acts as a default specifier of FocP employed as a last resort strategy to satisfy the V2 requirement. This would imply that, in the early Romance V2 varieties, V1 clauses should invariably be excluded, since these would in some sense violate a well-principle which requires SpecFocP to be invariably lexicalized. To the contrary, however, apparent V1 structures are abundantly attested in early Romance V2 varieties, including, crucially, all those which also exhibit sì-clauses.Footnote 7 It follows that sì cannot be a last resort mechanism, since superficially V1 clauses prove to be legitimate linguistic objects in these same varieties.Footnote 8

Furthermore, FocP (and its associated specifier position), together with other functional projections contained within the C-domain such as TopP and FrameP, is standardly considered the locus of discourse-related information (Rizzi Reference Rizzi and Liliane1997), marking clausal distinctions of particular pragmatico-semantic import. Given the evidently strong pragmatico-semantic basis of such projections as FocP, it therefore seems counterintuitive to associate its specifier position with such a pragmatico-semantically impoverished element as sì, essentially an expletive (see also Ferraresi & Goldbach Reference Ferraresi and Goldbach2002: 13). Indeed, this point is persuasively made by Benincà (Reference Benincà, Raffaella, Héctor, Elena and Paul2006: 77), who, discussing V-initial main clauses in old Romance, dismisses the possibility of a default focus since ‘being relevant information, it cannot be interpreted by default’ but, rather, ‘has to be explicitly supplied’.

Sì: a topic marker

As observed in section 3, under the standard analysis sì is also frequently argued to mark the thematicity/topicality of the constituent immediately to its left (cf. (7b) above; Benincà Reference Benincà, Battye and Roberts1995: 333; Vance Reference Vance, Battye and Roberts1995: 184; Salvi Reference Salvi2001: 1; Poletto Reference Poletto, Montserrat, Maria-Lluïsa, Carme and Francesc2005: 225). Although this is not invariably the case, as we shall see below where sì will be shown to co-occur also with foci, under this analysis the function of sì is, to all intents and purposes, likened to that of a topic marker. Crucially, however, the standard analysis also argues conversely that sì occupies SpecFocP, and not Top°, which would presumably be the most appropriate (head) position for a particle with such a topicalizing function.

Restriction to finite root clauses

Another problem with the standard analysis concerns two important restrictions on the distribution of sì. The first concerns the observation that sì is never attested in conjunction with non-finite verbal forms, and the second that sì proves most rare in, if not totally absent from, embedded contexts (Marchello-Nizia Reference Marchello-Nizia1985: 15; Fleischman Reference Fleischman1991: 267; van Reenen & Schøsler Reference van Reenen, Schøsler, Susan, Pieter van and Lene2000: 60; Ferraresi & Goldbach Reference Ferraresi and Goldbach2002: 11). For example, Lemieux & Dupuis (Reference Lemieux, Dupuis, Battye and Roberts1995: 96) note following Marchello-Nizia (Reference Marchello-Nizia1979), that ‘si is never used in subordinate clauses and never in untensed clauses’. Clearly, it might be possible under the standard analysis to interpret the rarity (e.g. Old Neapolitan) or impossibility (e.g. Old French) of sì in embedded clauses by virtue of the fact that V2 independently proves rare or impossible in these same contexts (cf. section 2 above, and Ledgeway Reference Ledgeway, Anna Laura and Arturo2007: 139–145). Yet, if sì is assumed to lexicalize SpecFocP, a priori there is no reason why it should not readily occur in, or be excluded from, embedded contexts, since embedded clauses also permit the full range of functional discourse-related projections including FocP (see Ledgeway Reference Ledgeway2003, Reference Ledgeway2005). By the same token, interpreting sì as a focus does not explain its absence from untensed contexts, since non-finite clauses do indeed license foci (8a) and other left-peripheral categories such as clitic left-dislocations (8b):

  1. (8)

    1. (a) se prometteray de reportareme a la patria toa […] e iamay per nullo tiempo me abandonareif you-will-promise of to-bring-back=me to the country your and never for no time me=to-abandon‘if you promise to take me back to your country […] and never ever leave me’ (60.4–6)

    2. (b) E de onne dommayo o crodeletate che poy mostrirrimmo de fare contra li Troyani […] zoèy […] lle donne Troyane portaremolle a gran vituperio in nostra servituteand of each damage or cruelty that after we-will-show of to-do against the Trojans that-is the women Trojan to-carry-Agr1pl=them  to great shame in our service‘All the harm and cruelty that we shall visibly commit against the Trojans […] namely […] to submit the Trojan women to great dishonour in our service’ (127.1–6)

Initiating thematic paragraphs

Another peculiarity about the distribution of sì which the standard analysis fails to capture is the fact that in our corpus sì is never found in absolute sentence-initial position at the head of what can be called a ‘thematic paragraph’.Footnote 9 Rather, it is preceded either by a topic or focus within the left periphery, or, in the absence of one of these, is preceded by the co-ordinating conjunction e(t) ‘and’ indicating some form of cohesion with the preceding sentence/clause.Footnote 10 Indeed, as will be made more precise below (cf. section 4.2), in the latter case the use of sì in Old Neapolitan strictly implies some form of topic continuity with the preceding discourse. It therefore follows that a new piece of discourse within a text, whether it takes the form of a new chapter, paragraph or section, is never introduced by sì. This same point is made by Fleischman (Reference Fleischman1991: 270), who, arguing for the subject/topic continuity function of Old French si, states that ‘si never occurs in the opening sentence of a thematic paragraph, i.e. it never follows a major break or turning point in the thematic continuity of the discourse’.

Similarly, Lemieux & Dupuis (Reference Lemieux, Dupuis, Battye and Roberts1995: 96) observe for Old French that ‘the normal position for si is at the beginning of the sentence (never at the beginning of a text)’ [emphasis added, A.L.], and van Reenen & Schøsler (Reference van Reenen and Schøsler1992: 105) report that ‘si never appears at the beginning of a given piece of thematic structure, hence never at the beginning of a major textual subdivision, nor at the start of a paragraph’ (translation A.L.; see also Rychner Reference Rychner1970: 174; Marchello-Nizia Reference Marchello-Nizia1985: 30). On the standard assumption that sì is a focus, however, its incompatibility with the start of fresh sections of discourse (‘thematic paragraphs’) proves somewhat puzzling, since foci introduce informationally new material and hence should, by definition, prove most appropriate precisely in such contexts.

Incompatibility with negation

A further weakness of the traditional analyses of sì is evidenced by the syntactic incompatibility of Old Neapolitan sì with sentential negation, witness the absence of any examples of sì with a negated finite verb in the whole of the Libro de la destructione de Troya. Similar facts have been noted for other early Romance varieties (Salvi Reference Salvi2001: 3), including Milanese (Domokos Reference Domokos1998) and French, where ‘si is never used with negative clauses’ (Lemieux & Dupuis Reference Lemieux, Dupuis, Battye and Roberts1995: 96, 105, n. 9).Footnote 11 Clearly, if sì were a focus, then there is no reason why it should be precluded from co-occurring with negation, as canonical foci do readily occur in conjunction with negated verbs.

Compatibility with fronted foci and themes

Our final and, without doubt, strongest piece of evidence for rejecting the hypothesis that sì occupies SpecFocP comes from the observation that sì in fact co-occurs with other constituents that are themselves assumed to target SpecFocP. Recall that in early Romance varieties, movement of the verb to a C position under V2 is frequently, though not always, accompanied by the fronting of a salient constituent to the specifier position of the C projection whose head is now occupied by the raised verb. The preverbal position is not therefore a privileged subject position, as is the case in SVO languages but, rather, constitutes a pragmatically-salient position licensing thematic and rhematic interpretations available to all syntactic categories. Recognizing the dual pragmatico-semantic value of this single preverbal position, Salvi (Reference Salvi2004) accordingly labels the relevant functional projection T(heme)/F(ocus)P. Along similar lines, Benincà (Reference Benincà, Raffaella, Héctor, Elena and Paul2006: 65) observes that pragmatically the preverbal position ‘can have various interpretations’ insofar as it ‘can represent a contrastive, or an unmarked focus, a “relevant” theme, or an anaphoric theme’, and hence such elements are argued to ‘occupy the Spec of FocusP, or better, a Spec in the Focus Field, as sentential operators or moved arguments’ (p. 64). Indeed, evidence that there is a single preverbal position (namely, Salvi's SpecT/FP and Benincà's SpecFocP) variously licensing thematic and rhematic interpretations of fronted constituents comes from the identical syntactic behaviour of thematic and rhematic direct objects which, when fronted to a position immediately before the verb, both occur without a clitic copy. Topicalized direct objects, in contrast, which occur in the higher topicalization space (namely, the specifier of FrameP or TopP), always require a clitic copy (Benincà Reference Benincà, Raffaella, Héctor, Elena and Paul2006: 67, 73–77).

It follows under the standard analysis, therefore, that sì is predicted not to co-occur with non-clitic-doubled direct objects, since both would be competing for the same position, namely SpecFocP. Crucially, however, our Old Neapolitan corpus offers some very important counterexamples in which sì incontrovertibly co-occurs with a non-clitic-doubled object, witness (9a–d):

  1. (9)

    1. (a) e terrimoti orribile sì faceva a sua voluntate and earthquakes horrible sì she-made to her will ‘and she could bring about horrendous earthquakes at her will’ (56.32)

    2. (b) spissi cuolpi mortali sì le davafrequent blows mortal sì to-him=he-gave‘he struck him with many deadly blows’ (133.36)

    3. (c) E multe altre parole iniuriose e sospecose sì disse Amphimaco contra Anthenoreand many other words offensive and suspicious sì said Amphimachus against Antenor‘And Amphimachus said many other offensive and accusing words against Antenor’ (243.8–9)

    4. (d) La VI.a schera Hector sì ordenao de la gente de lo riamme de Poemia sotto la guida de lo signore lorothe 6th battalion Hector sì ordered of the people of the kingdom of Poemia under the guide of the lord their‘Hector formed the 6th battalion from the people of the kingdom of Poemia under the leadership of their lord’ (150.7–8)

In (9a–b) the fronted direct objects terrimoti orribile ‘horrendous earthquakes’ and spissi cuolpi mortali ‘many deadly blows’ are clearly rhematic in interpretation, each introducing a new referent into the discourse as an example of informational focus, as further witnessed by the absence of any doubling clitic. In (9c–d), on the other hand, the fronted direct objects multe altre parole iniuriose e sospecose ‘many other offensive and accusing words’ and La VI.a schera ‘the sixth battalion’ are examples of contrastive focus, inasmuch as the quantifiers multe altre ‘many other’ and the ordinal adjective VI.a ‘sixth’ implicitly presuppose a contrast (for a discussion of inherently gradable quantifiers and adjectives in such cases, see Vanelli Reference Vanelli and Lorenzo1998: 82): in the former case the offensive words in question are contrasted with a previous series of insults, and in the latter case the ‘sixth’ battalion is singled out in relation to the other five battalions discussed in the immediately preceding discourse, as well as the other three battalions that figure in the subsequent discourse. The focalized nature of these fronted objects is once again highlighted syntactically by the absence of a resumptive clitic. Moreover, in the case of (9d) between the contrastively focused direct object and sì there intervenes the fronted thematic subject Hector, whose topicality is substantiated by the fact that Hector has already consistently figured in the immediately preceding discourse as the sole person responsible for managing the organization of the preceding five battalions (not to mention the following final three). In this respect, example (9d) proves particularly revealing in that it demonstrates that the focus position targeted by rhematic objects and the thematic position licensing fronted topics (when they do not target the higher topicalization space above the Focus field) are distinct from the position lexicalized by sì (on the projection which licenses fronted constituents under V2, see section 5 below).

Such examples are not restricted to Old Neapolitan, but are also attested in other early Romance varieties, witness the Old Perugian and Old French examples given below in (10a–b), respectively, where sì/si is preceded by the rhematic non-clitic-doubled direct objects le 20 ‘the 20’ and tote la terre ‘the entire territory’:

  1. (10)

    1. (a) e le 20 sì vole culuie ke 'l portaand the 20 sì wants he-who that them=carries‘and he who brings them wants the 20’ (Livero de l'abbecho 134v.11, in Bocchi Reference Bocchi, Michelangelo and Lorenzo2004: 142)

    2. (b) et tote la terre si tenoit Johans li rois de Blakie et de Bougrieand all the land si held Johan the king of Blakie and of Bougrie‘and Johan the king of Blakie and of Bougrie possessed the entire territory’ (Villehardouin 176: 38f., in Ferraresi and Goldbach Reference Ferraresi and Goldbach2002: 16)

Also of interest in this respect are Old Neapolitan examples such as (11a–b) where, although the fronted constituent immediately preceding sì is not the direct object of the main verb, it is clearly focalized. In (11a) the rhematic fronted subject beginning uno Iudice Guido de la Colonna de Messina is an example of informational focus (for a similar Old Lombard example, see Rohlfs Reference Rohlfs1969: 166), and example (11b) offers an unequivocal example of contrastive focus involving a fronted locative. It follows once again from examples like these that sì cannot be interpreted as the specifier of (one of) the Focus projection(s).

  1. (11)

    1. (a) uno Iudice Guido de la Colonna de Messina, homo de approbata descriptione e sottile intendemento e famoso dectatore, sì l'ave transontato in chesta presente forma latinaa judge Guido de la Colonna of Messina man of approved description and subtle understanding and famous dictator sì it=has transformed in this present form Latin‘a judge Guido de la Colonna of Messina, a man of excellent character and fine intellect and a famous dictator, translated it into its current Latin form’ (47.30–33)

    2. (b) non tanto inde li gran facti, ma inde li pizoli, sì perdeno loro saperenot so-much in the great facts but in the small sì they-lose their knowledge‘their knowledge abandons them not so much in the great, but the little, matters’ (76.15f.)

4. Sì: head of Fin°

As outlined above, the distribution of sì in Old Neapolitan and in other Romance varieties is such that it cannot be readily interpreted as the specifier of a functional projection within the C-space. Rather, all the available distributional and semantico-pragmatic evidence points to an analysis of sì as the lexicalization of the lowest C head, Fin°, in accordance with a tendency to satisfy the V2 requirement by Merge as an alternative to the more frequent Move operation.Footnote 12 However, one anonymous reviewer points out that if sì is merged in C°, we should expect to find sì immediately followed by a lexical subject in SpecTP. Contrary to the prediction, however, the sequence sì+Subject+Vfinite is not attested. Rather than constituting counterevidence to the present analysis, the impossibility of this ordering is independently predicted by the analysis of Old Neapolitan V2 presented in Ledgeway (Reference Ledgeway, Anna Laura and Arturo2007: 134–138; see also the discussion in section 2.4 above). There it is shown that presentative sentences with transitive predicates are necessarily V1 in Old Neapolitan since only the finite verb is available to raise to the C-space, thereby stranding the arguments of the verb and yielding [CP V [IP S O]] words orders like that in (4a) above. Particularly revealing in this respect are examples like (12a–b), where the position of the lower VP adverbs (Cinque Reference Cinque1999) multo ‘much’ and ancora ‘still’, to the immediate left of the subject and (prepositional) object, demonstrates that these latter arguments occupy their base positions within the v-VP:

  1. (12)

    1. (a) Meraviglyavasse multo [VP [SpecVP lo populo de la terra [Vˈ tmeraviglyavasse [PP de li muode de le costume de quisti]]]]Marvelled=self much the people of the land  of the ways of the customs of these‘The people of the land marvelled greatly at the ways and customs of these (people)’ (56.7–8)

    2. (b) Appe ancora [VP [SpecVP questa citate di Troia [Vˈ tappe [NP VI porte multo larghesseme]]]]had still this city of Troy  6 gates very very-wide‘This city of Troy had as well 6 very wide gates’ (79.1–2)

On the basis of such examples, Ledgeway (Reference Ledgeway, Anna Laura and Arturo2007) claims that Old Neapolitan lacked a dedicated preverbal subject position (namely, T lacked an EPP-feature and hence failed to project SpecTP). Rather, as we have seen, subjects, just like all other constituents, are restricted to occurring in their base position within the v-VP unless they receive particular pragmatic salience, in which case they are fronted to the C-space. Indeed, the lack of a T-related subject position in Old Neapolitan is entirely in keeping with the V2 nature of the language, a correlation independently maintained for other V2 languages such as German (see Haider Reference Haider1993; Roberts & Roussou Reference Roberts, Roussou and Peter2002: 145). Whereas in a non-V2 language like Modern Italian the dedicated preverbal subject position licenses, although not exclusively, both thematic subjects and rhematic subjects in wide focus, in a V2 language like Old Neapolitan these same pragmatic functions are typically licensed by movement of the subject to a specifier position within the C-space. There would therefore be very little motivation for a T-related subject position in a V2 language, especially if the EPP feature (whatever that turns out to be) can be satisfied by verb movement to the T head (cf. Alexiadou & Anagnostopoulou Reference Alexiadou and Anagnostopoulou1998). We conclude that the absence of the order sì+Subject+Vfinite follows without further stipulation from the failure of T to project SpecTP.Footnote 13

Let us now turn to consider the core distributional properties of sì and how these are captured by our proposed analysis.

4.1. Core distributional properties of sì

The rise and fall of sì

From a diachronic perspective, one of the most significant facts regarding the distribution of sì is unquestionably the observation that it is no longer found in any of the modern Romance varieties. More precisely, it would appear that sì uniquely characterizes those early stages of the Romance languages in which the V2 constraint was still operative. Indeed, a direct correlation between V2 and the presence of sì has not gone unnoticed in the literature. For instance, Fleischman (Reference Fleischman1991: 277) argues that ‘the degrammaticalization of discourse si and its eventual disappearance from the language […] is closely connected to […] the shift from a flexible, pragmatically controlled word order of the V/2 or TVX type, to a rigid SVX type’. This view is echoed by Poletto (Reference Poletto, Montserrat, Maria-Lluïsa, Carme and Francesc2005: 227): ‘The hypothesis that sì is a Spec, Focus expletive also derives [sic] why Modern Italian has completely lost this element […] [G]iven that Modern Italian is not a V2 language, it does not have a V2 expletive.’ Under our proposed analysis, according to which merger of sì in Fin° and verb movement to the same are interpreted as complementary ways of satisfying the V2 requirement, the observed correlation falls out naturally without any further stipulation.

As we shall argue below (section 4.2), sì was employed in Old Neapolitan not only to satisfy V2, but also served, at the same time, to signal topic continuity within a given stretch of discourse. As summed up by Fleischman (Reference Fleischman1991: 270), ‘si occurs most commonly where there is no explicit expression of the subject and moreover, where the subject of the si clause is understood to be the same as that of the preceding clause’.

Given this concomitant function of sì as a topic continuity marker, we might well ask why Romance varieties did not retain sì, even once V2 was lost, at least in its topic-marking function. Once again, we claim the answer lies in the loss of the V2 constraint. In particular, it is interesting to note that in an SVO language like Modern Italian null subjects only license anaphorically coreferential readings if the antecedent is preverbal. Typically, in such cases the antecedent coincides with the subject of predication of the clause (Cordin Reference Cordin and Renzi1988: 546–549; Salvi Reference Salvi, Laura and Alberto1991: 36ff.), variously represented by the syntactic subject of the clause as in (13a) or, less frequently, by a fronted, possibly cliticized, (in)direct complement as in (13b):

  1. (13)

    1. (a) [Piero]i baciò [Maria]j e [proi/*j] scappò via Piero kissed Maria and ran away

    2. (b) [A Piero]i/ [Gli]i piace [Maria]j ma [proi/*j] si vergogna to Piero him= it-pleases Maria and  self=shames‘Piero/He likes Maria but is shy’

If, however, the syntactic subject (14a) or (in)direct complement (14b) follows the verb, then coreference with a null subject is blocked:

  1. (14)

    1. (a) Se presenta il film, [ Giovanna]i, [pro]j/*i sarà contentaif presents the film Giovanna will-be happy‘If Giovanna presents the film, she will be happy’

    2. (b) Se il film interessa [a Giovanna]i /[a lei]i, [pro]j/*i sarà contentaif the film interests to Giovanna to her will-be happy‘If the film interests Giovanna/her, she will be happy’

On the plausible assumption that even in cases like (13b) involving psych-predicates the preverbal antecedent can be argued to occupy the SpecTP position (see Cardinaletti 2004), coreferentiality in Modern Italian in such cases can be viewed as a consequence of the strong topicality uniquely associated with the preverbal SpecTP position. In Early Neapolitan, by contrast, the V2 nature of the language was such that there was more than one position frequently available to highly topical constituents. More specifically, the subject of predication, which typically coincides with the topical syntactic subject, does not necessarily occur in the preverbal SpecCP position in early Romance, but can equally occur in postverbal position if SpecCP is filled by some other, pragmatically more salient constituent (cf. the postverbal thematic subjects in (3b–c)). Thus, in contrast to Modern Italian, in a V2 language the potential antecedent of a null subject cannot simply be read off a particular clausal position,Footnote 14 witness coreferential null subjects such as (15a) licensed by a preceding postverbal subject (which would prove ungrammatical in Modern Italian), and the use of topic continuity markers like sì which overtly and unambiguously signal coreference between a given null subject and a preceding topical subject (cf. section 4.2), whether in pre- or postverbal position as in (15b).

  1. (15)

    1. (a) piglyao forza [Iasone]i e con grande ardire [pro]i/*j stese la manotook force Jason and with great courage stretched the hand‘[Jason]i summoned up strength and with great courage [he]i/*j stretched out his hand’ (65.20–21)

    2. (b) E poy che llo sappe [loduca Nestore]i incontinente [pro]i/*j sì venne a lluyand after that it=knew theduke Nestor immediately sì he-came to him‘And after [Duke Nestor]i discovered it, [he]i/*j immediately went to him’ (106.4)

Consequently, with the loss of V2 and the concomitant emergence of SpecTP as a position associated with strong topicality hosting the subject of predication, sì too fell out of usage since coreference could now be unequivocally read off a particular structural position, viz. SpecTP.

Distributional restrictions

Finiteness

Characteristic of the distribution of sì, as noted above in section 3, is the fact that it typically surfaces in affirmative, finite root clauses. Assuming sì to lexicalize Fin° as part of the V2 requirement, its incompatibility with non-finite verbs follows straightforwardly given the restriction of V2 to finite contexts. At the same time, this also explains the frequent observation by many Romance historical grammarians and philologists (see, among others, Rohlfs Reference Rohlfs1969: section 760; Ménard Reference Ménard1973: section 197; Foulet Reference Foulet1974: section 439; Fleischman Reference Fleischman1991: 262) that sì often proves impossible to translate, for, according to our analysis, its presence is motivated, at least in part, by purely structural factors, namely the V2 constraint. By the same token, the V2 analysis of Romance sì also accounts for another frequent observation (see, in particular, Marchello-Nizia Reference Marchello-Nizia1985) that sì serves as an assertive marker, essentially an evidential particle of veridiction asserting the speaker's/writer's authority with respect to the truth of the proposition (cf. also Romance sì/si(m) ‘yes’). Capitalizing on the idea proposed in many V2 analyses that movement of the finite verb to C° acts as a strategy to encode assertive force (Platzack & Holmberg Reference Platzack and Holmberg1989; Klein Reference Klein1994, Reference Klein, Norbert and Zvi1998; Brandner Reference Brandner, Horst and Susanne2004), the assertive nature often attributed to sì now falls out naturally under our proposed structural interpretation of sì as an alternative strategy of satisfying the V2 constraint. Significantly, this idea finds further support in claims which correlate assertion with finiteness (for an overview, see Nikolaeva Reference Nikolaeva and Irina2007: 10), since we too take V2 to be lexicalization of, or movement to, the head Fin°.

Negation

We cannot appeal to the supposed assertive nature of sì to explain its incompatibility with negation, since negative assertions, although possibly somewhat marked, are clearly well-formed from a semantico-pragmatic perspective, and hence legitimate linguistic objects. Moreover, as previously observed (section 3.1.6, n. 11), in some Romance varieties sì and negation do co-occur, albeit infrequently. Consequently, there must be a structural explanation for the observed incompatibility of sì and negation. Here we propose, following Cinque (Reference Cinque1999: 124), that the simple negator non in Early Neapolitan is base-generated in SpecNegP, but has to incorporate into its verbal host by Spell-Out. Consequently, under normal circumstances the finite verb, together with any pronominal clitics, targets under V2 the Fin° head, to which the negator subsequently cliticizes by left-adjunction, as in (16a).Footnote 15 However, if V2 is satisfied by merger of sì in Fin°, the negator non is not able to left-adjoin to the verbal head stranded under T°, nor can it left-adjoin to sì inasmuch as the latter lacks the necessary verbal categorial feature that non seeks in its verbal host. Consequently, the derivation will crash, as schematicized in (16b):

  1. (16)

    1. (a) [FinP [Finˈ [Neg non [T cl+V]] [NegP [Spec tnon ] [Negˈ [TP [Tˈ tcl+V [VP …]]]]]]]

    2. (b) *[FinP [Finˈ [Neg non [Fin sì ]] [NegP [Spec tnon ] [Negˈ [TP [Tˈ cl+V [VP …]]]]]]]

Under this analysis, independent differences in negation can be invoked to account for the contrast observed between (most) Romance varieties like Old Neapolitan on the one hand and those rare varieties like Early Tuscan on the other, where sì does occasionally co-occur with negation, as demonstrated in (17):

  1. (17) per questa ragione è assai manifesto, sì nollo divideròfor this reason is very clear sì not=it= I-will-divide‘for this reason it is clear, I will not divide it up’ (Dante, Vita nuova 35, 1–4, 142, 14, cited in Salvi Reference Salvi2001: 3)

In essence, we follow Zanuttini (Reference Zanuttini1997) in assuming two positions for preverbal negators, a higher position NegP1 situated above TP and a lower position NegP2 situated below TP. In the former case, as we have seen in (16a), the negator can only left-adjoin to the finite verb when the latter raises to C°, hence its incompatibility with sì when the finite verb is forced to remain under T°. In varieties like Early Tuscan, by contrast, we assume preverbal non to be merged in the lower NegP2 position. Consequently, non in Early Tuscan can always raise to left-adjoin to the finite verb under T° within the I-space, whether the finite verb ultimately raises to C° or whether sì lexicalizes C° causing the finite verb to remain in situ under T°. Under our analysis, we can thus derive the variable behaviour of sì in conjunction with negation across Romance in a principled fashion from independent differences in negation.

Embedded clauses

Finally, we turn to embedded clauses. Above (section 2) we observed that V2 orders, although frequent, are nonetheless less common in embedded clauses, where V1 orders prove more frequent. We interpreted this fact as indirect evidence of the asymmetric V2 nature of Old Neapolitan, insofar as V2 will normally be precluded in embedded clauses on account of the C position already being lexicalized by a complementizer (e.g. c(h)a, che ‘that’, se ‘if’). Given our interpretation of sì as a means of satisfying V2 through the operation Merge, we should expect a similar contrast in the distribution of sì in root and embedded contexts in our 50-page textual sample (see section 2). Indeed, this prediction is borne out: out of a total of 912 root clauses in the sample 97 of these, albeit only 10.6%, contained sì, in contrast to only 18 embedded tokens of sì, a mere 2.2%, out of a total of 808 embedded clauses. This distributional pattern is further supported by a statistical count of the number of occurrences of sì in root and embedded clauses in the rest of the text, namely 14 tokens of embedded sì in comparison to 113 tokens of root sì (yielding the overall totals: root/embedded sì: 210/32 tokens). We can thus interpret such statistical information as providing further convincing support for the conclusion that sì lexicalizes the Fin° head in satisfaction of the V2 constraint, hence its limited distribution in embedded contexts.

In Germanic and early Romance many cases of embedded V2, often constrained by various syntactic and pragmatic factors, have been noted (Vikner Reference Vikner1995: chapter 4; Salvi Reference Salvi2004: chapters 1, 3; Ledgeway Reference Ledgeway, Anna Laura and Arturo2007: 139f.). Benincà (Reference Benincà, Raffaella, Héctor, Elena and Paul2006: 72), for instance, notes in relation to early Romance that ‘[i]n dependent complement clauses governed by bridge verbs, and even in dependent relatives, the accessibility of CP appears more restricted than in main clauses. But this restriction applies only to quantity, not to quality’.

This characterization holds equally of Old Neapolitan, where embedded V2 is constrained by a number of factors.Footnote 16 In particular, following Ledgeway (Reference Ledgeway, Anna Laura and Arturo2007: 139f.), we note that the cases of embedded V2 identified for Old Neapolitan involve, with very few exceptions, the same environments recognized for other languages with restricted embedded V2. More specifically, embedded V2 in Old Neapolitan occurs predominantly in complement clauses characterized by a strong degree of assertion, complements of so-called ‘bridge verbs’ (de Haan & Weerman Reference de Haan, Weerman, Hubert and Martin1986; Vanelli Reference Vanelli and Harro1986: 269, 1998: 78; Hulk & van Kemenade Reference Hulk, van Kemenade, Battye and Roberts1995: 237; Kiparsky Reference Kiparsky, Battye and Roberts1995: 164, n. 15; Ribeiro Reference Ribeiro, Battye and Roberts1995: 118, 121; Santorini Reference Santorini, Battye and Roberts1995: 56; Fontana Reference Fontana, Ans van and Nigel1997: 246; Benincà Reference Benincà, Raffaella, Héctor, Elena and Paul2006: 72), as well as in specific types of adjunct clause (e.g. headed by ‘because’, ‘although’, ‘such that’). Old Neapolitan thus patterns with other asymmetric V2 languages that only license V2 in specific, lexically-restricted embedded contexts. It follows from this that, under our analysis, we should expect embedded sì to occur in the same restricted contexts, a prediction which is borne out: from a total of 32 tokens of embedded sì, 17 occurred in complement clauses to bridge verbs (18a) and 12 in the adjunct clause types identified above (18b), the remaining three examples occurring in temporal clauses headed by intanto che ‘while, until’ (2 occurrences) and che ‘when’:

  1. (18)

    1. (a) Dicate a vostro signore che nuy sì nne partirrimotell to your lord that we sì ourselves=will-leave‘Tell your lord that we shall leave’ (55.7)

    2. (b) concessa de cosa chi li Mirmidoni sì s'appellano habitaturi de Thesaliaconceded of thing that the Myrmidons sì selves=call inhabitants of Thessaly‘although the Myrmidons are called the inhabitants of Thessaly’ (48.31–32)

The question remains, however, how these infrequent occurrences of sì in embedded contexts are to be explained. The major technical issue here associated with embedded V2 is how both the finite verb, or sì in our case, and a lexical complementizer can simultaneously occupy the same C position. Following the seminal work of Benincà (Reference Benincà and Renzi1988, Reference Benincà, Battye and Roberts1995, Reference Benincà, Raffaella, Héctor, Elena and Paul2006) and Rizzi (Reference Rizzi and Liliane1997), who highlight a more richly-articulated C-space with at least two distinct complementizer positions Force° and Fin°, in addition to various Topic and Focus positions sandwiched between them, a new solution to this problem has emerged. In particular, within this enriched C-space embedded V2 structures can now be accommodated by assuming that the lexical complementizer occupies a higher position within the C-space such as Force°, and that the finite verb or sì occupies a lower C position such as Fin°, as illustrated in (19):Footnote 17

  1. (19) [ForcePche [TopP Topic [FocP Focus [FinPVfinite/sì [IP tVFinite …]]]]]

That this is indeed the correct analysis of Old Neapolitan embedded V2 (whether VFinite or sì in Fin°) is further corroborated by the form of the complementizer. As demonstrated in Ledgeway (Reference Ledgeway2003, Reference Ledgeway2005), in the early dialects of southern Italy, including Neapolitan, there occur two complementizers, che (<quod/quid) and c(h)a (<qu(i)a). Whereas all types of subjunctive clause are introduced by che, indicative complement clauses are headed by either che or c(h)a. Simplifying the facts somewhat, it will suffice to note here that either c(h)a (20a) or che (20b) can be employed when an embedded indicative clause does not contain any topic or focus, whereas che alone is found in the presence of topics (20c) or foci (20d):

  1. (20)

    1. (a) Homero […] dice a li suoy libri ca foro nave MCLXXXVIHomer says to the his books that were ships 1186‘Homer says in his books that there were 1186 ships’ (115.35)

    2. (b) purriase ben dicere che fo causa multo legierecould=self well to-say that it-was cause very light‘it could indeed be said that there was very little reason’ (53.16)

    3. (c) certa speranza  che lo re Priamo poterrànde recoperare la soro soacertain hope that the king  Priam will-be-able=therefrom to-recover the sister his‘with certain hope that King Priam will be able to rescue his sister from there’ (102.26)

    4. (d) Verace cosa èy che a li nuostri Diey fo sempremay gratatrue thing it-is that to the our gods was always welcome‘It is true that it has always pleased our gods’ (125.22)

In view of these distributional facts, it is possible to argue that Old Neapolitan had just one indicative complementizer c(h)a generated in the lowest C position Fin° which, whenever raised to a higher C position such as Force° – as is obligatory whenever topics or a focus are present – is morphologically spelt out in the form che (namely, c(h)a Fin⇉che Force). In short, the overt form of indicative che is interpreted as nothing more than the surface morphological reflex of raising c(h)a from its base position Fin° to some higher position within the C-space. Now, it is not coincidental that in all Old Neapolitan indicative complement clauses argued to be V2 structures, like those in (21a–b), the complementizer invariably surfaces in the form che (and never c(h)a). The impossibility of c(h)a in embedded V2 clauses now follows naturally: if the lowest C-position Fin° is filled by the raised finite verb, then the complementizer is forced to lexicalize the higher C-position Force°, where it will necessarily be spelt out as che.

  1. (21)

    1. (a) dice lo proverbio che a lo bove morto non fa prode de se le ponere l' erba a lo nasosays the proverb that to the ox dead not it-does advantageous of self=to-it=to-put the-grass to the nose‘the proverb says that it is not advantageous to place grass before a dead ox's nose’ (67.30–31)

    2. (b) Considerava che a quista insula de Colcosa, ove stava chisto pecoro de auro, non se nce poteva gire se non per marehe-considered that to this island of Colchis where stood this sheep of gold not self=there=could to-go if not for sea‘He believed that this island of Colchis, where this golden sheep lived, could only be reached by sea’ (51.26–27)

By the same token, it is also predicted that in indicative embedded contexts sì can only be preceded by the complementizer che, and never c(h)a, since the latter would be competing for the same position as sì. This prediction is indeed borne out, as revealed by the representative examples in (18a–b) above (see also Bocchi Reference Bocchi, Michelangelo and Lorenzo2004: 154, n. 6).

4.2. Sì: marker of topic continuity

Following Fleischman (Reference Fleischman1991), it was briefly claimed above (section 4.1.1) that besides satisfying the V2 constraint, merger of sì in Fin° also serves to mark topic continuity, inasmuch as it signals a null topic in SpecFinP (typically, though not exclusively, the subject) which is coreferential with the primary topic (generally encoded, though not exclusively, as the subject) of the preceding clause. This immediately explains a number of distributional properties of clauses containing sì, notably the failure of sì to initiate what was termed above (section 3.1.5) a ‘thematic paragraph’.Footnote 18 The following is thus a typical example of this use of sì, in which the co-ordinated clause contains a null subject (recall that referential null subjects are intrinsically topical) which is unequivocally interpreted as coreferential with the null subject of the preceding clause, Jason, a continuing topic of the immediately preceding discourse):

  1. (22) E da poy proi se voltao a [lo messayo]j e proi sì le parlaoand after self=turned to the messenger and sì to-him=spoke‘And after he turned towards the messenger and said to him’ (55.1)

Sì is not, however, a universal marker of topic continuity but, rather, surfaces where potential ambiguity might otherwise arise as to the identity of a given null topic. Under normal circumstances, the most continuous topic, which can be most readily identified in most cases, though not all, with the topical subject of the immediately preceding main clause, is simply marked through verb agreement, if there is no risk of ambiguity. However, the presence of other competing nominals/topics in the same clause or its immediate discourse environment can create the potential for topic ambiguity, especially if they can lay claim to the same (typically 3rd-person) verb morphology, thereby rendering less easily recoverable the identification of the intended subject topic (Fleischman Reference Fleischman1991: 256–258). Indeed, this is what happens in (22), where the preceding main clause (henceforth reference clause) reintroduces the animate, topical DP lo messayo ‘the messenger’, which would be a potential candidate for subject of the 3rd-person verb form parlao of the co-ordinated clause (henceforth marking clause) if it were not for the presence of sì. As a more marked device than grammatical agreement alone, sì overtly signals continuity of the most continuous/primary topic which, in accordance with general cross-linguistic principles (Givón Reference Givón and Talmy1983: 38, n. 21), standardly correlates with the grammatical subject as the argument characterized by a higher topic continuity value than all other grammatical relations. It comes as no surprise, then, that in our corpus of 242 tokens of sì, all but a handful of examples involve potentially ambiguous sequences such as (22) where two or more 3rd-person nominals/topics in the reference clause make competing claims on the 3rd-person verbal morphology in the marking clause.

By way of further illustration, consider the following representative examples:

  1. (23)

    1. (a) E [lo re Laumedonta]i per quella volta appe victoria de li suoy nemice; et ecco che lei sopervenne una novella che le portao [uno cavaliere de Troya anciano]j, [lo quale]j era feruto e venea fugendo a ccavallo e proj/*i sì le disse che la soa citate de Troya era stata invaduta e presa da li Grieciand the king Laumedonta for that time had victory of the his enemies and here that to-him=arrived a news that to-him=brought a knight of Troy old who was injured and came fleeing to horse and sì to-him=said that the his city of Troy was been invaded and taken by the Greeks‘And King Laumedonta on that occasion had victory over his enemies; and then he received some news which an elderly knight from Troy brought him who was injured and had escaped on a horse and told him that his city of Troy had been invaded and taken by the Greeks’ (73.30–33)

    2. (b) Meraviglyavasse multo [lo populo de la terra]i de li muodi e de le costume de [quisti]j, [li quali]j [a lloro]i pareano multo gran gentili huomini, e proi/*j sì ademandavano multo desiderosamente de la loro condicionemarvelled=self much the people of the land of the ways and of the customs of these who to them seemed much great gentle men and sì they-asked much with-desire of the their condition‘The people of this land marvelled greatly at the ways and the customs of these (people), who seemed to them very great gentlemen, and enquired with great desire about their condition’ (56.7–9)

In (23a) the first clause introduces as the subject King Laumedonta, which, as an established topic, is now referenced in the subsequent clause by the dative clitic le ‘to him’ alongside the introduction of the rhematic subject uno cavaliere de Troya anciano ‘an elderly knight from Troy’, which also heads the following subject relative clause. Consequently, there arises in the marking clause a degree of potential ambiguity as to the identity of the null subject, since both topical nominals can legitimately lay claim to the 3rd-person verb morphology. The presence of sì in the marking clause, however, eliminates any doubt regarding the correct interpretation of the null subject, explicitly linking the latter with uno cavaliere de Troya, which, by virtue of its subject status in the two immediately preceding clauses, counts as the topic of the thematic paragraph, outranking the topical referent of the dative clitic le to qualify as the primary/most continuous topic.

Turning now to (23b), here we have two 3rd-person topical nominals, the grammatically singular but semantically plural lo populo de la terra ‘the people of the land’ (=the inhabitants of the island of Colchis) and quisti ‘these’ (=the Greeks). The first of these functions as the postverbal subject of the reference clause and the latter as the possessor of the prepositional object nominal, in turn heading a subject relative in which the matrix subject returns as the indirect pronominal a lloro ‘to them’. In the marking clause introduced by e ‘and’, however, the verb now occurs in the 3rd-person plural and its associated null subject should therefore be most readily construed as coreferential with the plural quisti in the reference clause. Despite expectations, however, the null subject is coreferential with the grammatically singular matrix subject lo populo de la terra. This interpretation is made possible by the presence of sì, which signals that the intended referent of the null subject is the topic of the thematic paragraph and that the grammatical mismatch between the singular matrix subject and the plural null subject of the marking clause is to be overridden in favour of topic continuity.Footnote 19

The topic continuity role played by sì in this and other early Romance texts is further substantiated by the observation, frequent in the literature (Schiaffini Reference Schiaffini1926: 290–293; Marchello-Nizia Reference Marchello-Nizia1985: 161f.; Fleischman Reference Fleischman1991: 271; Salvi Reference Salvi2001: 8f.; Bocchi Reference Bocchi, Michelangelo and Lorenzo2004: 122ff.; Poletto Reference Poletto, Montserrat, Maria-Lluïsa, Carme and Francesc2005: 225), that sì is particularly favoured in main clauses preceded by a fronted subordinate clause (temporal, causal, conditional, final):

  1. (24)

    1. (a) E poy che foro suso [ Iasone et Hercules]i, per licencia de lo re, proi/*j sì se assettaro insembla co lluyand after that were up Jason and Hercules for license of the king sì selves=sat together with him‘And after Jason and Hercules went up, by permission of the king, they sat down together with him’ (56.17–19)

    2. (b) e, proi respondendo [a Iasone]j, proi/*j sì le disse cossì […]and replying to Jason sì to-him=she-said thus‘and, answering Jason, she said to him the following […]’ (60.20–21)

Significantly, it is precisely in such fronted structures that underlying binding relations are modified. If left in situ after the main clause, the subordinate clauses in (24a–b) license a disjoint interpretation, such that the matrix and subordinate subjects are necessarily interpreted as distinct (cf. English [they]isat down [after [Jason and Hercules]j/*iwent upstairs]). When fronted as in (24a–b), however, non-obligatory control obtains, such that the matrix subject is now free to refer or not to the embedded subject (cf. English [After [Jason and Hercules]jwent upstairs], [they]i/jsat down]). It follows that in such structures sì fulfils a disambiguating functioning, signalling the intended coreferential reading.

Perhaps the most revealing evidence for the topic/subject continuity function of sì comes from rare examples like (25a–b):

  1. (25)

    1. (a) proi credo che proi/*j sì ll'aya in dubio a dicere si fo lo meglyo oy no che […]  I-believe that sì it=I-have-subj in doubt to to-say if was the better or not that ‘and I believe that I cannot say with certainty whether it was better or not that […]’ (54.16–17)

    2. (b) Et proi aveano proveduto che quisto facto proi/*j sì llo facesseroand they-had arranged that this fact sì it=they-do-subj‘And they had arranged that they would undertake this matter’ (139.30–31)

Prototypically, in Romance the selection of the subjunctive imposes a switch reference interpretation (the so-called obviation effect), such that the reference of the embedded subject is necessarily interpreted as disjoint with respect to the matrix subject. If, on the other hand, coreference is intended, this is encoded through the selection of an infinitival complement, as in the following Italian example:

  1. (26) Sandrai vuole PROi/*j diverirtsi / che proj/*i si divertaSandra wants  to-enjoy=self  that self=enjoys.subj‘Sandra wants to enjoy herself/her to enjoy herself’

Now, in both examples in (25) there are no other competing topical nominals in the reference clause other than the matrix null subject itself, but nonetheless sì occurs in the embedded clause of each example. Rather, what renders necessary the presence of sì in such examples is the unexpected coreferential reading of the embedded subject. Under normal circumstances, a null subject in an embedded subjunctive clause is interpreted in Old Neapolitan, as in Italian, as disjoint in reference vis-à-vis the matrix subject, insofar as coreferential readings are canonically licensed by infinitival clauses. Yet, in these examples a subjunctive complement has exceptionally been employed to convey the coreferential reading, an interpretation only made possible by the presence of sì, which overtly signals and licenses the otherwise exceptional topic continuity of the subject in these examples. In the absence of sì, the reference of the embedded subject in these same examples could only be interpreted as disjoint.

5. Conclusion

In this paper we have presented extensive evidence from Old Neapolitan to support the claim that the V2 requirement of medieval Romance could be satisfied not only by movement of the finite verb to the lowest C-related head Fin°, but also by merger of sì directly in Fin°, a hypothesis which explains the otherwise restricted distribution of sì in early Romance.Footnote 20 Significantly, this analysis finds parallels in a recent comparative analysis of the Germanic and Celtic C-systems by Roberts (Reference Roberts and Rizzi2004), who argues that at the relevant level of abstraction the two language families can both be analysed as V2 insofar as (root) finite Fin° must be given PF-realization, but that they differ as to the formal realization of this requirement: verb raising (Germanic) vs. merger of a ‘sentential’ particle (Celtic). Although we have seen in section 4.2 that this latter Merge option in medieval Romance often serves to mark topic continuity through the licensing of a null topic in SpecFinP coreferential with the foremost topic of the reference clause (ultimately a reflex of an EPP-feature on finite Fin°), the primary function of sì is, however, to satisfy the V2 constraint by checking the relevant V2 feature on Fin° (see also Fleischman Reference Fleischman1991: 269) – witness (27a–b) where sì apparently has no role in marking topic continuity,Footnote 21 but simply satisfies the relevant V2 feature on Fin°:Footnote 22

  1. (27)

    1. (a) Facto questo, lo tiempo de la vattaglya sì sopervennedone this the time of the battle sì arrived‘This having happened, the time of the battle arrived’ (225.3)

    2. (b) E llo re resposse a lloro e sì le disseand the king replied to them and sì to-them=said‘And the king replied to them and said to them’ (242.3)

A major claim of the present research is therefore that V2 should not be understood only as a case of structurally-induced verb raising, but ultimately as a featural requirement of the relevant C-related head which can be satisfied by either of the two core syntactic operations, (external) Merge or Move (or internal Merge). Indeed, the Old Neapolitan evidence reviewed here lends strong empirical support to the present theory, insofar as the possibility of checking the relevant V2 feature through the Merge option is entirely to be expected on standard assumptions about feature checking/valuation.

Besides throwing light on the mechanisms at work in satisfying medieval Romance V2, our analysis of the Old Neapolitan data has also highlighted that the traditional assumption of a single, syncretic preverbal position variously licensing rhematic or thematic constituents is empirically untenable. Rather, the evidence of examples such as (9d) demonstrates that in early Romance V2 varieties the dual informational values (rheme and theme) traditionally associated with the fronted preverbal position should be interpreted as indicative of two distinct positions (hierarchically ordered as rheme+theme). In light of the frequent topic-marking function of sì, which we have interpreted as licensing a null topic in SpecFinP as a reflex of an EPP feature on Fin°, we take fronted themes (not doubled by a coreferent clitic) in such cases to occupy SpecFinP, whereas fronted rhematic constituents occupy SpecFocP, hence the observed order rheme+theme.

Footnotes

[1]

The research reported here was conducted as part of a larger research project dedicated to the writing of a diachronic grammar of the Neapolitan dialect. I am grateful to the Arts and Humanities Research Council (AHRC) and the University of Cambridge for providing generous funding for this research project by way of research leave during the academic year 2005–6. I should like to thank two anonymous JL reviewers for their extremely useful comments on an earlier version of this article. I am also grateful to the audiences of the VII Incontro di dialettologia (University of Bristol, February 2006) and the XIII Giornata di dialettologia (University of Padua, June 2007), where earlier versions of this article were presented.

2 In what follows ‘Old Neapolitan’ should be understood to refer specifically to the language of this particular 14th-century text. While the text can admittedly be considered a refined piece of literary prose writing (namely, a volgarizzamento ‘vernacular translation’), the language of the text is widely recognized as an authentic reflection of the dialect, sharing most, if not all, of the typical morphosyntactic features of the dialect as represented in contemporaneous texts of different registers (see De Blasi Reference De Blasi, Günter, Michael and Christian1995: 185; Cennamo Reference Cennamo2005: 139f.), including a V2 syntax and concomitant use of sì (see Ledgeway in press: section 21.1.2).In citing from this text, the first Arabic numeral refers to the page number(s) of the text and the second Arabic numeral refers to the relevant line number(s) in the text. In glosses, the symbol ‘=’ is used to indicate cliticization.

3 For quite different reasons and with quite different results, Ferraresi & Goldbach (Reference Ferraresi and Goldbach2002) have also argued that Old French si is merged in the lowest C head, namely Fin°. However, under their analysis merger of si in Fin° is not an alternative to the move option as a means of satisfying the V2 constraint but, rather, operates in addition to verb movement to Fin°. Significantly, this incorrectly predicts, among other things, the unattested order *V+sì, since raising of the finite verb to Fin° following merger of sì can, under standard assumptions (Kayne Reference Kayne1994), only take place under left-adjunction. We do not discuss their analysis any further here.Incidentally, it should be noted that while the new empirical contribution of the present paper is principally, though not exclusively, limited to Old Neapolitan, it is claimed, and to some extent also demonstrated (cf. sections 3.1.7, 4.1.2.2, 4.2, and Footnote note 20 below), that the proposed analysis of sì can be carried over to other early Romance varieties. Thus, although the new proposal is not explored exhaustively here in relation to the syntax of all early Romance varieties, it will be the task of subsequent research to determine whether the proposal is compatible with the syntax of those early Romance varieties that are still poorly known or understood.

4 Considerably less frequent are examples of superficial V4, V5 and V6 orders, which accounted for 7.9%, 2.2% and 0.4%, respectively, of all root clauses.

5 An anonymous reviewer points out that, qua manner adverb meaning ‘thus’ or degree adverb modifying adjectives and adverbs meaning ‘so’, sì can indeed occur in positions within the clause other than the left periphery. Although this is clearly true, the semantically vacuous use of sì under discussion here cannot be equated with these semantically meaningful instances of sì, as convincingly demonstrated by Poletto (Reference Poletto, Montserrat, Maria-Lluïsa, Carme and Francesc2005: 219ff.). For example, while in the latter cases sì freely alternates with the fuller form cossì (cf. (i.a–b)), in its V2 function sì can never be replaced by cossì. Moreover, as noted by Poletto (Reference Poletto, Montserrat, Maria-Lluïsa, Carme and Francesc2005: 219, n. 19) sì and (cos)sì cannot even be readily derived from the same etymological source. In short, we conclude that, whereas qua degree/manner adverb (cos)sì can legitimately be fronted to SpecFocP under contrastive focus (cf. (ii.a)), sì in its V2 function can only be employed in preverbal position (cf. Poletto Reference Poletto, Montserrat, Maria-Lluïsa, Carme and Francesc2005: 222–224), witness the example (ii.b) in which it co-occurs with (postverbal) adverbial cossì:

  1. (i)

    1. (a) questa soa figlyola unica e cossì bellathis his daughter only and so beautiful‘this so beautiful only daughter of his’ (57.3–4)

    2. (b) e maraviglyavase de sì bella cossaand marvelled=selves of so beautiful thing ‘and they were amazed by such a beautiful thing’ (67.12)

  2. (ii)

    1. (a) cossì se facza e non altramentethus self=it-do and not otherwise‘thus it should be done and not otherwise’ (60.29)

    2. (b) E Medea sì le resposse cossìand Medea sì to-him=replied thus‘And Medea replied to him thus’ (59.37)

6 That is not to say, as pointed out by an anonymous reviewer, that expletive elements are necessarily heads, witness the canonical treatment of expletive subjects such as English it/there, French il and German es as specifiers (‘weak’ elements in Cardinaletti & Starke's (Reference Cardinaletti and Starke1994, Reference Cardinaletti, Starke and Henk1999) tripartite typology). Indeed, one possible way of reconciling the Old Neapolitan data with the traditional analyses of other Romance varieties is to view Old Neapolitan sì as having reached a more advanced stage of grammaticalization, having developed from an original XP expletive merged in SpecCP (as perhaps still witnessed in early northern Italian dialects, Old Tuscan and Old French) to the head of CP along the lines of van Gelderen (Reference van Gelderen2004; see also Bocchi Reference Bocchi, Michelangelo and Lorenzo2004: 154, n. 6).

7 As was argued above (section 2), V2 should not be understood as a simple linear surface constraint, but, rather, as a syntactic requirement that the finite verb raise to a vacant (lower) head position within the C-domain, irrespective of whether this is accompanied in turn by fronting of one or more pragmatically-salient constituents.

8 Of course, it might be argued that SpecFocP in apparent V1 structures actually hosts a phonologically null variant of sì. Indeed, in a number of V2 analyses V1 structures are assumed to contain a null operator (cf. Diesing Reference Diesing1990: 56, n. 14; Lemieux & Dupuis Reference Lemieux, Dupuis, Battye and Roberts1995: 98; Ribeiro Reference Ribeiro, Battye and Roberts1995: 122). Benincà (Reference Benincà, Raffaella, Héctor, Elena and Paul2006: 77f.), by contrast, maintains that V1 structures should be interpreted as structures containing a default Topic. The nature of such a default Topic, however, remains obscure, though, presumably, in certain contexts, it might be instantiated by a ‘discourse’ Topic. This issue will not be explored further here, inasmuch as the present analysis of V2 in terms of a featural requirement on Fin° is compatible with either analysis.

9 As Fleischman (Reference Fleischman1991: 256) puts it, ‘chains of clauses combine into larger thematic units which Givón refers to as “thematic paragraphs” […], which may in turn combine into larger discourse units, such as “paragraphs,” “sections,” “chapters,” “parts,” or “stories” (Givón Reference Givón and Talmy1983: 7). The thematic paragraph is the lowest level of discourse structure at which it makes sense to talk about continuity.’

10 Poletto (Reference Poletto, Montserrat, Maria-Lluïsa, Carme and Francesc2005: 227ff.) argues on the basis of Early Tuscan that in a number of contexts e(t) functions as a topic marker situated in some higher portion of the C-domain signalling the continuation of the topic of the previous discourse. Though sì and e(t) frequently co-occur in our corpus, in numerous cases sì occurs without e(t), suggesting that the topic continuity function of sì cannot be universally related to its co-occurrence with e(t). Significant in this respect are frequent examples of fronted subordinate clauses like (24a–b), discussed below, where sì alone marks subject coreference (=topic continuity) in the main clause, as well as examples such as (25a–b), where the otherwise exceptional presence of sì (without e(t)) in a subjunctive complement clause obviates the expected switch reference interpretation of the embedded null subject.

11 This view is largely, though not entirely, supported by Marchello-Nizia (Reference Marchello-Nizia1985: 179), who notes that ‘the proportion of cases in which si introduces a negated verb is very low’ (translation A.L.; cf. also p. 52) The compatibility of si with a finite negated verb in Old Tuscan is discussed below in section 4.1.2.2.

12 This idea is also indirectly supported by Rohlfs (Reference Rohlfs1969: 166), who claims that ‘in early northern Italian dialects the function of si was so weakened that the particle can be found preposed to all verbs in main clauses preceded by one or more words’ (translation and emphasis A.L.).

13 It should also be noted, as pointed out by a reviewer, that the absence of the sequence [sì+S+V] is also naturally predicted by the pragmatic function of sì as a marker of topic continuity. As we shall see (section 4.2), sì appears in contexts in which it licenses a null topic that is co-referential with the foremost topic of the reference clause, corresponding typically to the grammatical subject. Accordingly, we would not expect a lexical subject to follow sì.

14 Also relevant here is Poletto's (Reference Poletto, Montserrat, Maria-Lluïsa, Carme and Francesc2005: 233; in press: section 2.3) observation that Early Tuscan licenses thematic null subjects in contexts which would prove pragmatically inappropriate in Modern Italian. In particular, the null subject of a main clause in Early Tuscan can be understood as coreferential with an element introduced in the preceding discourse, even when the latter is not the topic (or most topical element) of the preceding clause. Similar cases are also found in Old Neapolitan, witness (i), where the null subject of the verb andausende ‘he went off’ is to be understood as coreferential with the overt pronominal illo ‘he’ of the first clause, and not with the more immediate topic Siculo, his brother, introduced in the subsequent discourse:

  1. (i) E da poy [illo]i se partio da chella isula e lassaonce [lo frate suo]j che se clamava Siculo; e da poy se clamao la isula de Sicilia, per semeletudine de lo nome suo; et [pro]i andausende in Tuscana inde la quale fece multe terreand after he self=left from that island and left=there the brother his that self=called Siculo and after self=called the island of Sicily for similarity of the name his and went in Tuscany in the which he-made many lands‘And afterwards [he]i left that island and left there [his brother]j who was called Siculo; and subsequently the island was called Sicily on account of the similarity of his name; and [he]i went to Tuscany where he conquered many lands’ (53.32–34)

In the light of evidence like this, Poletto argues that null subjects in such examples should be interpreted as cases of topic drop.

15 The head status assumed by non when it adjoins to the verb (and any accompanying clitics) within the C-space is confirmed by the fact that it blocks further verb movement. More specifically, in V1 contexts enclisis proves the norm in conjunction with affirmative verbs, since, following Benincà (Reference Benincà, Battye and Roberts1995: 336–338; 2006: 78), we assume that the verb, once raised together with its clitic to the lowest C position, then excorporates to raise alone to a higher position within the Topic field (presumably related to the idea that V1 clauses contain a null topic; see Roberts Reference Roberts1993; Benincà Reference Benincà, Raffaella, Héctor, Elena and Paul2006: 77f.). Excorporation of V from [cl+V] under Fin° to raise to Top° is allowed because enclitics in early Romance, just like proclitics in modern Romance, were clitics and not affixes, and hence could be stranded (see Salvi Reference Salvi2004: chapter 5). However, negated verbs in V1 contexts invariably display proclisis, a fact which we interpret as a consequence of left-adjunction of non to the verbal complex [t[D cl] [VV]], which precludes excorporation of the finite verb, now too deeply embedded within this complex head (namely, [Fin [Negnon] [t [D cl] [VV]]]), as well as further movement to the Topic field.Observe that the obligatory proclisis of clitics in the presence of sì (Salvi Reference Salvi2001: 2f.; Poletto 2005: 226) also follows from the present analysis, since we have argued above, following Salvi (Reference Salvi2004: chapter 5), that clitics in old Romance left-adjoin to their verbal host directly under T°, where they remain together with their verbal host if Fin° is already lexicalized by sì.

16 The existence of embedded V2 in Old Neapolitan is supported, among other things, by the observation that in a number of embedded clauses there intervenes between the complementizer and the finite verb a constituent different from the subject. As was observed in relation to root clauses (section 2.2), the frequent preposing of a constituent distinct from the subject is a typical word order pattern in V2 languages, where such preposing signals the rhematic or thematic intepretation of the fronted constituent, and the type of constituent amenable to preposing is essentially unconstrained and ultimately not restricted to subjects. Indeed, out of a total of 327 embedded (superficially) V2 clauses, 140 of these (namely, 42.8%) were found not to be subject-initial, a fact that highlights the unrestricted nature of the preverbal position.

17 Interestingly, in a number of analyses predating the split CP analysis, V2 was interpreted as movement of the finite verb to the vacant C position to license a ‘finiteness’ feature on the verb (cf. Vikner Reference Vikner1995: section 3.4.4), an idea implicitly assumed under the present analysis of V-movement to, and sì-merger in, the lower C-head Fin(iteness).

18 Prototypically, thematic paragraphs are built around a single nominal referent which occurs as the recurring topic in each foregrounded clause of the paragraph. Coupled with the overwhelming tendency for foregrounded information to be packaged in root clauses, this accounts for the low frequency of sì in embedded clauses (Fleischman Reference Fleischman1991: 256). Moreover, this presumably explains the lower incidence of unaccusative predicates in conjunction with sì in our corpus (namely, 42 tokens (36.5%) in contrast to 73 examples with transitives/unergatives (63.5%)), since unaccusative subjects are prototypically rhematic.

19 A not-too-dissimilar case is given in (i), where the left-dislocated, topicalized 3sg. subject, tutta quella universitate de la gente Troyana ‘the entire body of Trojan people’, is coupled with a grammatically 3pl. verb. Nonetheless, the coreferential interpretation is guaranteed by sì, which identifies beyond all doubt the reference of the resumptive null subject (and hence its associated verb) with the most immediate and continuous topic, namely the preceding topicalization.

  1. (i) Allora [tutta quella universitate de la gente Troyana]i, auduto quillo sermone, [pro]i sì l'acceptarothen all that gathering of the people Trojan heard that speech  sì it=accepted-3pl‘Then [the entire body of Trojan people]i, having heard that speech, [pro]i accepted it’ (82.28–29)

20 An anonymous reviewer notes that the present analysis also seems largely compatible with the conclusions of Bocchi (Reference Bocchi, Michelangelo and Lorenzo2004) regarding the distribution of sì in the 14th-Perugian vernacular translation Livero de l'abbecho. In particular, Bocchi argues that the word order of the text is characterized by an original V2 syntax, but that, superficially, the placement of the verb in second position is synchronically often obscured by an expanding left periphery (viz. Topic field) hosting one or more left-dislocated elements (pp. 152f.), thereby pushing the verb, for example, into third or fourth position or beyond. Within this system, he argues (p. 154) that, when employed, sì comes to indicate the demarcation between the sentential core and the expanding left periphery, functioning to all intents and purposes as a ‘complementizer’ to the main verb, as evidenced by its incompatibility with the complementizer che (p. 154, n. 6). Strikingly, this account, although couched in purely descriptive terms, comes very close to that proposed in the present paper, inasmuch as a complementizer demarcating the sentential core (=I-domain) and the left periphery (=C-domain) equates, within the framework assumed here, quite squarely with the lowest C-related head, namely Fin°.

21 However, as pointed out by an anonymous reviewer, the discourse role of sì can still be maintained in examples such as (27a–b) if the topic is interpreted as the discourse topic. For example, in (27a) the preceding passage deals with the battle, such that the presence of sì in the cited example signals the continuation of ‘the battle’ as the intended topic of the thematic paragraph, though not the subject of the preceding discourse. Such a consistent interpretation of the discourse role of sì seems highly plausible and, if sustainable, preferable to that given in the text.

22 Fleischman (Reference Fleischman1991: 275) notes that in Old French si occasionally occurs in V1 unaccusative clauses with postverbal rhematic subjects. We interpret these as presentative sentences in which si not only satisfies the V2 constraint but, following Benincà's analysis (Reference Benincà, Raffaella, Héctor, Elena and Paul2006: 77f.) of V1 structures, also licenses a null theme (namely, a default topic) in SpecFinP. Indeed, in view of the class of predicates involved, namely unaccusatives, it seems plausible to associate such a null theme with a covert locative or temporal argument deictically bound to the speaker or the discourse (cf. Benincà Reference Benincà and Renzi1988: 123–125).

References

REFERENCES

Adams, Marianne. 1987. Old French, null subjects, and verb second phenomena. Ph.D. dissertation, UCLA.Google Scholar
Alexiadou, Artemis & Anagnostopoulou, Elena. 1998. Parametrizing AGR: Word order, verb-movement and EPP checking. Natural Language & Linguistic Theory 16, 491–539.CrossRefGoogle Scholar
Battye, Adrian & Roberts, Ian (eds.). 1995. Clause structure and language change. Oxford: Oxford University Press.Google Scholar
Belletti, Adriana. 2004. Aspects of the low VP area. In Rizzi, (ed.) 16–51.Google Scholar
Benincà, Paola. 1988. L'ordine degli elementi della frase e le costruzioni marcate. Costruzioni con ordine marcato degli elementi. In Renzi, (ed.) 129–143.Google Scholar
Benincà, Paola. 1995. Complement clitics in medieval Romance: The Tobler-Mussafia law. In Battye, & Roberts, (eds.) 325–344.Google Scholar
Benincà, Paola. 2006. A detailed map of the left periphery of medieval Romance. In Raffaella, Zanuttini, Héctor, Campos, Elena, Herberger & Paul, Portner (eds.) Crosslinguistic research in syntax and semantics: Negation, tense, and clausal architecture, 53–86. Washington, DC: Georgetown University Press.Google Scholar
Benincà, Paola & Poletto, Cecilia. 2004. Topic, focus, and V2: Defining the CP sublayers. In Rizzi, (ed.) 52–75.Google Scholar
Benincà, Paola & Salvi, Giampaolo. 1988. L'ordine normali degli elementi nella frase semplice. In Renzi, (ed.) 119–129.Google Scholar
Bocchi, Andrea. 2004. «Sì» nel «Livero de l'abbecho». In Michelangelo, Zaccarello & Lorenzo, Tomasin (eds.) Storia della lingua e filologia, 121–158. Florence: SISMEL, Edizioni del Galluzzo.Google Scholar
Brandner, Ellen. 2004. Head-movement in minimalism and V2 as FORCE-marking. In Horst, Lohnstein & Susanne, Trissler (eds.) Syntax and semantics of the left periphery, 97–138. Berlin: Mouton de Gruyter.CrossRefGoogle Scholar
Cardinaletti, Anna. 2004. Toward a cartography of subject positions. In Rizzi, Trissler (ed.) 115–165.Google Scholar
Cardinaletti, Anna & Starke, Michal. 1994. The typology of structural deficiency: On the three grammatical classes. University of Venice Working Papers in Linguistics 4, 41–109.Google Scholar
Cardinaletti, Anna & Starke, Michal. 1999. The typology of structural deficiency: On the three grammatical classes. In Henk, van Riemsdijk (ed.) Clitics in the languages of Europe, 145–233. Berlin: Mouton de Gruyter.Google Scholar
Cennamo, Michela. 2005. Note sul rapporto tra il latino e il volgare nello studio della sintassi dei primi testi. La lingua italiana: Storia, strutture, testi 1, 137–143.Google Scholar
Cinque, Guglielmo. 1999. Adverbs and functional heads: A cross-linguistic perspective. Oxford: Oxford University Press.CrossRefGoogle Scholar
Cordin, Patrizia. 1988. I pronomi personali. In Renzi, (ed.) 535–549.Google Scholar
Dardano, Maurizio & Frenguelli, Gianluca (eds.). 2004. SintAnt: La sintassi dell'italiano antico. Atti del convegno internazionale di studi (Università Roma Tre, 18–21 settembre 2002). Rome: Aracne.Google Scholar
De Blasi, Nicola (ed.). 1986. Libro de la destructione de Troya: Volgarizzamento napoletano trecentesco da Guido delle Colonne. Rome: Bonacci.Google Scholar
De Blasi, Nicola. 1995. Kampanien. In Günter, Holtus, Michael, Metzeltin & Christian, Schmitt (eds.) Lexikon der romanistischen Linguistik, vol. 2(2), 175–189. Tübingen: Niemeyer.Google Scholar
de Haan, Willem & Weerman, Frank. 1986. Finiteness and verb fronting in Frisian. In Hubert, Haider & Martin, Prinzhorn (eds.) Verb second phenomena in Germanic languages, 77–110. Dordrecht: Foris.Google Scholar
Diesing, Molly. 1990. Verb movement and the subject position in Yiddish. Natural Language & Linguistic Theory 8, 41–79.CrossRefGoogle Scholar
Domokos, György. 1998. Appunti su morfologia e sintassi del milanese di Bonvesin de la Riva. Aevum 72, 619–631.Google Scholar
Dupuis, Fernande. 1989. L'expression du sujet dans les subordonnées en ancien français. Ph.D. dissertation, University of Montreal.Google Scholar
Ferraresi, Gisella & Goldbach, Maria. 2002. V2 syntax and topicalisation in Old French. Linguistische Berichte 18, 3–25.Google Scholar
Fleischman, Suzanne. 1991. Discourse pragmatics and the grammar of Old French: A functional reinterpretation of si and the personal pronouns. Romance Philology 44, 251–283.Google Scholar
Fleischman, Suzanne. 1992. Discourse and diachrony: The rise and fall of Old French SI. In Marinel, Gerritsen & Dieter, Stein (eds.) Internal and external factors in syntactic change, 433–473. Berlin: Mouton de Gruyter.CrossRefGoogle Scholar
Fontana, Josep M. 1993. Phrase structure and the syntax of clitics in the history of Spanish. Ph.D. dissertation, University of Pennsylvania.Google Scholar
Fontana, Josep M. 1997. On the integration of second position phenomena. In Ans van, Kemenade & Nigel, Vincent (eds.) Parameters of morphosyntactic change, 207–249. Cambridge: Cambridge University Press.Google Scholar
Foulet, Lucien. 1974. Petite syntaxe de l'ancien français. Paris: Champion.Google Scholar
Givón, Talmy. 1983. Topic continuity in discourse: An introduction. In Talmy, Givón (ed.) Topic continuity in discourse: Quantitative cross-language studies, 1–41. Amsterdam: John Benjamins.CrossRefGoogle Scholar
Haider, Hubert. 1993. Deutsche Syntax generativ. Tübingen: Gunter Narr.Google Scholar
Hopper, Paul & Traugott, Elizabeth. 1993. Grammaticalization. Cambridge: Cambridge University Press.Google Scholar
Hulk, Aafke & van Kemenade, Ans. 1995. Verb second, pro-drop, functional projections and language change. In Battye, & Roberts, (eds.) 227–256.Google Scholar
Kayne, Richard. 1994. The antisymmetry of syntax. Cambridge, MA: MIT Press.Google Scholar
Kiparsky, Paul. 1995. Indo-European origins of Germanic syntax. In Battye, & Roberts, (eds.) 140–169.Google Scholar
Klein, Wolfgang. 1994. Time in language. London & New York: Routledge.Google Scholar
Klein, Wolfgang. 1998. Assertion and finiteness. In Norbert, Dittmar & Zvi, Penner (eds.) Issues in the theory of language acquisition: Essays in honor of Jürgen Weissenborn, 225–245. Bern: Peter Lang.Google Scholar
Lambrecht, Knud. 1994. Information structure and sentence form: Topic, focus and the mental representations of discourse referents. Cambridge: Cambridge University Press.CrossRefGoogle Scholar
Ledgeway, Adam. 2003. Il sistema completivo dei dialetti meridionali: La doppia serie di complementatori. Rivista italiana di dialettologia 27, 89–147.Google Scholar
Ledgeway, Adam. 2005. Moving through the left periphery: The dual complementiser system in the dialects of southern Italy. Transactions of the Philological Society 103, 336–396.CrossRefGoogle Scholar
Ledgeway, Adam. 2007. Old Neapolitan word order: Some initial observations. In Anna Laura, Lepschy & Arturo, Tosi (eds.) Languages of Italy: Histories and dictionaries, 119–146. Ravenna: Longo.Google Scholar
Ledgeway, Adam. In press. Grammatica storica del dialetto napoletano. Tübingen: Niemeyer.Google Scholar
Lemieux, Monique & Dupuis, Fernande. 1995. The locus of verb movement in non-asymmetric verb-second languages: The case of Middle French. In Battye, & Roberts, (eds.) 80–109.Google Scholar
Lombardi, Alessandra & Middleton, Roberta. 2004. Alcune osservazioni sull'ordine delle parole negli antichi volgari italiani. In Dardano, & Frenguelli, (eds.) 553–582.Google Scholar
Lombardi Vallauri, Edoardo. 2004. Sintassi e informazione nell'italiano antico: L'oggetto preverbale. In Dardano, & Frenguelli, (eds.) 293–321.Google Scholar
Marchello-Nizia, Christiane. 1979. Histoire de la langue française aux XIVeet XVesiècles. Paris: Bordas.Google Scholar
Marchello-Nizia, Christiane. 1985. Dire le vrai: I'adverbe «si» en français médiéval: Essai de linguistique historique. Geneva: Droz.Google Scholar
Ménard, Philippe. 1973. Syntaxe du moyen français: Manuel du français du moyen âge. Bordeaux: SOBODI.Google Scholar
Meyer-Lübke, Wilhelm. 1899. Grammatik der romanischen Sprachen: Romanische Syntax. Leipzig: Reisland.Google Scholar
Nikolaeva, Irina. 2007. Introduction. In Irina, Nikolaeva (ed.) Finiteness: Theoretical and empirical foundations, 1–19. Oxford: Oxford University Press.CrossRefGoogle Scholar
Platzack, Christer & Holmberg, Anders. 1989. The role of AGR and finiteness in Germanic VO languages. Working Papers in Scandinavian Syntax 43, 51–76.Google Scholar
Poletto, Cecilia. 2005. Sì and e as CP expletives in Old Italian. In Montserrat, Batllori, Maria-Lluïsa, Hernanz, Carme, Picallo & Francesc, Roca (eds.) Grammaticalization and parametric variation, 206–235. Oxford: Oxford University Press.Google Scholar
Poletto, Cecilia. In press. La struttura della frase. In Lorenzo, Renzi & Giampaolo, Salvi (eds.) Grammatica dell'italiano antico. Bologna: il Mulino.Google Scholar
Renzi, Lorenzo (ed.). 1988. Grande grammatica italiana di consultazione, vol. I: La frase: I sintagmi nominale e preposizionale. Bologna: il Mulino.Google Scholar
Ribeiro, Ilza. 1995. Evidence for a verb-second phase in Old Portuguese. In Battye, & Roberts, (eds.) 110–139.Google Scholar
Rizzi, Luigi. 1997. The fine structure of the left periphery. In Liliane, Haegeman (ed.) Elements of grammar, 281–337. Dordrecht: Kluwer.Google Scholar
Rizzi, Luigi (ed.). 2004. The structure of CP and IP: The cartography of syntactic structures, vol. 2. Oxford: Oxford University Press.Google Scholar
Roberts, Ian. 1993. Verbs and diachronic syntax: A comparative history of English and French. Dordrecht: Kluwer.Google Scholar
Roberts, Ian. 2004. The C-system in Brythonic Celtic languages, V2, and the EPP. In Rizzi, (ed.) 297–328.Google Scholar
Roberts, Ian & Roussou, Anna. 2002. The extended projection principle as a condition on the Tense dependency. In Peter, Svenonius (ed.) Subjects, expletives, and the EPP, 125–155. Oxford: Oxford University Press.Google Scholar
Rohlfs, Gerhard. 1969. Grammatica storica della lingua italiana e dei suoi dialetti, vol. III: Sintassi e formazione delle parole. Turin: Einaudi.Google Scholar
Rychner, Jean. 1970. L'articulation des phrases narratives dans la ‘Mort Artu’: Formes et structures de la prose française de la prose médiévale. Neuchâtel: Faculté des Lettres & Geneva: Droz.Google Scholar
Salvi, Giampaolo. 1991. Difesa e illustrazione della legge di Wackernagel applicata alle lingue romanze antiche: La posizione delle forme pronominali clitiche. In Laura, Vanelli & Alberto, Zamboni (eds.) Per Giovan Battista Pellegrini: Scritti degli allievi padovani, 439–462. Padua: Unipress.Google Scholar
Salvi, Giampaolo. 2001. Il problema di si e l'uso riflessivo di essere. Ms., University of Budapest.Google Scholar
Salvi, Giampaolo. 2004. La formazione della struttura di frase romanza: Ordine delle parole e clitici dal latino alle lingue romanze antiche. Tübingen: Niemeyer.Google Scholar
Santorini, Beatrice. 1995. Two types of verb second in the history of Yiddish. In Battye, & Roberts, (eds.) 53–79.Google Scholar
Schiaffini, Alfredo. 1926. Testi fiorentini del Dugento e dei primi del Trecento. Florence: Sansoni.Google Scholar
Skårup, Peter. 1975. Les premières zones de la proposition en ancien français. Copenhagen: Akademisk Forlag.Google Scholar
van Gelderen, Elly. 2004. Economy, innovation, and prescriptivism: From spec to head and head to head. Journal of Comparative Germanic Linguistics 7, 59–98.CrossRefGoogle Scholar
van Reenen, Pieter & Schøsler, Lene. 1992. Ancien et moyen français: Si ‘thématique’, analyse exhaustive d'une série de textes. Vox Romanica 51, 101–127.Google Scholar
van Reenen, Pieter & Schøsler, Lene. 1993. Si ‘thématique’, étude de si en ancien et moyen français, discours direct. In Gerold, Hilty (ed.) Actes du XXecongrès international de linguistique et philologie romanes, Université de Zurich (6–11 avril 1992), vol. I, section I: La phrase, 617–628. Zurich: Francke Verlag.Google Scholar
van Reenen, Pieter & Schøsler, Lene. 1995. The thematic structure of the main clause in Old French: Or versus si. In Henning, Andersen (ed.) Current studies in linguistic theory 124: Historical linguistics. Selected papers from the 11th International Conference on Historical Linguistics, Los Angeles, 16–20 August 1993, 401–419. Amsterdam: John Benjamins.Google Scholar
van Reenen, Pieter & Schøsler, Lene. 2000. The pragmatic functions of the Old French particles ainz, apres, donc, lors, or, pluis, and si. In Susan, Herring, Pieter van, Reenen & Lene, Schøsler (eds.) Textual parameters in older languages, 59–105. Amsterdam: John Benjamins.Google Scholar
Vance, Barbara. 1995. On the decline of verb movement to Comp in Old and Middle French. In Battye, & Roberts, (eds.) 173–199.Google Scholar
Vance, Barbara. 1997. Syntactic change in medieval French: Verb-second and null subjects. Dordrecht: Kluwer.Google Scholar
Vanelli, Laura. 1986. Strutture tematiche in italiano antico. In Harro, Stammerjohann (ed.) Tema-rema in italiano, 249–273. Tübingen: Gunter Narr.Google Scholar
Vanelli, Laura. 1998. Ordine delle parole e articolazione pragmatica nell'italiano antico: La ‘prominenza’ pragmatica della prima posizione nella frase. In Lorenzo, Renzi (ed.) Italant: Per una grammatica dell'italiano antico, 73–89. Padua: Progetto Italant, Università di Padova.Google Scholar
Vikner, Sten. 1995. Verb movement and expletive subjects in the Germanic languages. Oxford: Oxford University Press.Google Scholar
Zanuttini, Raffaella. 1997. Negation and clausal structure: A comparative study of Romance languages. Oxford: Oxford University Press.CrossRefGoogle Scholar