Interpolation, verb-second, and the low left periphery in Old Spanish

69 Iberia: IJTL | Volume 5.1, 2013, 69-98 ISSN: 1989-8525 doi: tba http://revistas.ojs.es/index.php/iberia/ Interpolation, verb-second, and the low left periphery in Old Spanish1 Geoffrey Poole Newcastle University geoffrey[email protected] Abstract Interpolation is a phenomenon of a number of (chiefly Medieval) Romance varieties in which direct and indirect object pronouns may be separated from the finite verb by elements such as negation, adverbs, subjects and prepositional phrases. This paper considers both their information structure and syntax using data from the Corpus del Español (Davies 2002-). I first argue that, contra a number of analyses, interpolation is not a process of focalization. Rather, many interpolated elements are a G-Topic in the sense of Bianchi & Frascarelli (2010); in other words, a familiar or given topic. This information structure account suggests that interpolation targets a functional category in the low left periphery, and this is confirmed by a number of distributional facts: interpolation appears to target a position lower than various high topic positions, but higher than the TP domain. However, some interpolated elements (particularly sentential negation) are less plausibly analysed as topics, and I suggest that these elements are positioned in the low left periphery by ‘Formal Movement’ (Frey 2004, 2006), a semantically and pragmatically vacuous operation which attracts the element at the left edge of the TP domain in order to satisfy a verb-second requirement. Keywords: interpolation, Old Spanish, focus, G-Topic, information structure, Formal Movement, verb-second. 1. Introduction One of the most striking respects in which Old Spanish differs from Modern Spanish is in the ability of direct and indirect object pronouns to appear separated from the finite verb by an intervening constituent, as illustrated in (1)2. 1 Thanks to Maria Maza, audiences at Newcastle University and the September 2012 LAGB meeting at the University of Salford, especially Liliane Haegeman, and two anonymous reviewers for much helpful discussion. Any remaining errors are my own. 2 See, e.g., Ramsden (1963), Wanner (1992), Rivero (1992, 1997), Fontana (1993), among others for discussion. I restrict myself in this paper to consideration of interpolation in Old Spanish. ‘Interpolation’ phenomena exist in other medieval Romance varieties, but seem to have different properties from interpolation in Old Spanish, suggesting that different processes Interpolation, verb-second, and the low left periphery in Old Spanish Geoffrey Poole 70 Iberia: IJTL | Volume 5.1, 2013, 69-98 ISSN: 1989-8525 doi:tba http://revistas.ojs.es/index.php/iberia/ (1) se me non quisieres creer if me not wish.2sg believe.inf ‘if you do not wish to believe me.’ (El Libro de Alexandre (O), 14th c.) (1) contrasts sharply with the word-order possibilities of Modern Spanish, in which the object pronoun is proclitic on the finite verb and may not be separated from it by any material. ‘Interpolation’ is the label traditionally given to this phenomenon, following Chenery (1905). Sentential negation is the most commonly interpolated element, as in (1) above, but elements such as adverbs (2), short prepositional phrases (3) and subjects (4), among others, are also frequently found: (2) Et la prueua desto es que los que lo asi fizieron que se fallaron ende bien And the proof of.that is that those that it thus dis.3pl that se found.3pl there well ‘And the proof of that is that those who did it that way ended up well because of it.’ (Libro Infinido, 14th c.) (3) …mas que otro omenque se enel mundosupiese… ...morethat otherman that se in.the world knew.3sg ‘…more than any other man that was known in the world…’ (Crónica de 1344 I, 14th c.) (4) …& vsa mal del buen entendimiento que le dios dio. and uses badly of.the good understanding that him God gave.3sg ‘…and he makes poor use of the good understanding that God gave him.’ (Castigos e documentos de Sancho IV, 13th c.) Cases of interpolation in Old Spanish first appear in the 12th century, and are initially rather rare, although, as Fontana (1993:325) and Ramsden (1963:31) note, this may be due to the relative scarcity of texts (and the lack of texts in Old Spanish from any earlier). However, the phenomenon is robustly found in texts of the 13th century. Both authors note a drop in the 14th century, followed by an increase again the 15th century. However, interpolation rather abruptly disappears during the 16th century. are at work. For example, Old Portuguese (Martins 2002, 2003, 2005) is able to interpolate very large, heavy elements of a kind not seen in Spanish. Old Italian, by contrast, only allows interpolation of adverbs (Ledgeway & Lombardi 2005). In Old Catalan and Provençal the phenomenon is entirely unattested (Ramsden 1963: 134; Fischer 2003: 260). All Old Spanish examples are taken from the Corpus del Español (CdE) (Davies 2002-) unless otherwise indicated. Interpolation, verb-second, and the low left periphery in Old Spanish Geoffrey Poole 71 Iberia: IJTL | Volume 5.1, 2013, 69-98 ISSN: 1989-8525 doi:tba http://revistas.ojs.es/index.php/iberia/ This chronology is broadly confirmed by the data in the Corpus del Español. Table 1 summarizes the instances of interpolation of sentential negation immediately following a complementizer, a common interpolation pattern3. Table 1. Instances of C-ObjPn-NEG-Vfin in the Corpus del Español 13th c. 14th c. 15th c. 16th c. 17th c. Instances 1745 718 479 4 5 Per million words 259.84 268.96 58.69 0.23 0.40 Although there is an absolute dip in the number of examples from the 13th to the 14th centuries, the frequency of occurrence is virtually identical. However, by the 15th century interpolation has fallen to under 25% of its previous frequency, and in the 16th century appears to have disappeared entirely (though this particular pattern exaggerates the effect slightly). By way of illustrating the frequency of interpolation relative to non-interpolation, consider Table 2: Table 2. Instances of C-NEG-ObjPn-Vfin in CdE 13th c. 14th c. 15th c. 16th c. 17th c. Instances 1924 685 2633 12462 7886 Per million words 286.49 256.60 322.62 731.58 638.63 Table 2 summarizes the instances in which a negated finite verb and object pronoun immediately follow a complementizer, but where the negation is not interpolated. (Instead, sentential negation appears to the immediate left of the pronoun-verb cluster, just as it would in Modern Spanish). As can be seen by comparing the last row in the two tables, interpolation is about as common as non-interpolation in the 13th and 14th centuries, though non-interpolation of course eventually comes to dominate entirely4. Section 2 begins by considering the information structure value of interpolation. Contra Batllori et al (1995) and Poole (2007), among others, I suggest that interpolation is not a focalization process. Instead, I argue that, in many cases, the interpolated 3 The Corpus del Español contains no texts prior to the 13th century. 4 A final notable fact about interpolation is that it is overwhelmingly found in subordinate clauses rather than main clauses. Chenery 1905 describes it as a non-root phenomenon, while Rivero (1997: 188) notes only ‘isolated examples’ of main clause interpolation. Both Fontana (1993: 46) and Nieuwenhuijsen (1999: Section 3.3) characterize interpolation as ‘almost exclusively’ a subordinate clause phenomenon. [Nieuwenhuijsen: ‘Sin embargo, el fenómeno de la interpolación no se producía en cualquier contexto sino casi exclusivamente en oraciones subordinadas.’] I know of no analysis that captures this generalization satisfactorily. See footnote 26. Interpolation, verb-second, and the low left periphery in Old Spanish Geoffrey Poole 72 Iberia: IJTL | Volume 5.1, 2013, 69-98 ISSN: 1989-8525 doi:tba http://revistas.ojs.es/index.php/iberia/ element is plausibly a G-Topic in the sense of Bianchi & Frascarelli (2010) -the ‘Familiar Topic’ of Frascarelli & Hinterhölzl 2007. Under their analysis, G-Topics are located in FamP, a functional projection in the low left periphery. In Section 3, it is argued on the basis of a number of word order facts that interpolation targets a position lower than various higher topic positions, but higher than TP, precisely the space suggested by the information structure analysis. Those interpolated elements which are not plausible G-Topics, however, reveal the role that verb-second plays in interpolation. Where there is no G-topic available to support the raised verb in the left periphery, a Formal Movement operation of the sort seen in other verb-second languages (Frey 2004, 2006; Light 2012) attracts the element at the left edge of the TP domain, frequently sentential negation5. 2. The information-structure value of interpolation In this section, I consider the specific information-structure value of interpolation. A number of authors (e.g., Batllori et al 1995 and Poole 2007, among others) have claimed that the interpolated element is a focus and, given the word-order facts to be discussed later, it is a logical possibility. Although detailed information-structure information can be difficult to extract from texts, there are a number of context-independent reasons to reject this analysis (in addition to the existence of seemingly unambiguous contexts). Instead, consideration of full NP and PP interpolation suggests that the interpolated element is a G-Topic (a ‘given’ or ‘familiar’ topic in the sense of Frascarelli and collaborators)6. 2.1. Interpolation is not focus movement Sitaridou (2011) identifies three different focalization processes in Old Spanish: contrastive focus, (new) information focus and verum focus. When an element undergoes contrastive focus, it is ‘singled out from a discourse set of competing alternatives’ (Sitaridou 2011:fn 14). Information focus, on the other hand, does not require a contrastive interpretation for the focalized element, but it is associated with new information (ibidem:fn. 4, 174, 177). Verum focus is ‘a kind of narrow focus on the sentence polarity, and the propositional content is presented as background information’ (Leo5 The proposal here thus differs from, for example, Uriagereka (1995a,b) and Rivero (1997), as they locate only the object pronoun in a CP-related functional projection (FP and WP respectively) and not the interpolated element. (Rivero 1992 also claims that the object pronoun is focalized.) It does bear some resemblance to Martins’ (2002; 2003; 2005) analysis of interpolation in Old Portuguese, in the sense that both the pronoun and the interpolated element are attracted to a position that hosts multiple specifiers. However, this position (AgrSP) is also not CP-related and she in fact assumes (2002: 236) that separate low left periphery projections exist within the CP field above AgrSP. 6 To reiterate, I return to a discussion of interpolation of sentential negation in Section 3.3. Interpolation, verb-second, and the low left periphery in Old Spanish Geoffrey Poole 73 Iberia: IJTL | Volume 5.1, 2013, 69-98 ISSN: 1989-8525 doi:tba http://revistas.ojs.es/index.php/iberia/ netti & Escandell-Vidal 2009:182). Although details of information structure can be difficult to extract from texts, there are a number of considerations, both context-independent and context-dependent, which clearly indicate that the interpolated element is not a focus and does not play any of these information structure roles. 2.1.1. Uniqueness One context-independent reason to suspect that interpolation is not a process of focalization is that interpolation of two short items is occasionally found: (5) si lo asi no fiziesen mostraria que no querian obedeçer if it thus not do.3pl would.show.3sg that not would.want.3pl obey.inf mandamiento de dios ni del señor temporal command of God nor of.the lord wordly ‘And if they don’t do it like this, it would show that they don’t want to obey the orders of God or their temporal master.’ (Siete Partidas, 13th c.) (6) el mal que se oy aquj faze the evil that SE today here do.3sg ‘the evil that is done here today’ (General Estoria V, 13th c.) This is unexpected if the interpolated elements are foci, since, as noted by, e.g., Zagona (2002:251), Modern Spanish allows only one pre-verbal focus. (7) *AYER los TOMATES compró Juan. yesterday the tomatoes bought-3sg Juan. ‘It was yesterday the tomatoes that Juan bought.’ (Modern Spanish, Zagona 2002:251) If the modern Spanish constraint is inherited from earlier stages of the language, rather than being a post-Golden Age innovation, then (5) and (6) are unexpected if interpolation is a focalization process. 2.1.2. Interpolation in factive complements A second context-independent reason to believe that interpolation is not a focalization process is that it is licensed in the complements of factive predicates. Zagona (2002:252) notes that focalization in Modern Spanish is impossible in the complement of factive predicates:7 (8) a. Dice que MAÑANA lo operan. “He says that it’s tomorrow that they’re operating on him.” b. *Siento que MAÑANA lo operen. “I regret that it’s tomorrow that they’re operating on him.” (Modern Spanish, Zagona 2002:252) 7 She characterizes the distinction in terms of whether or not the embedded clause is asserted or presupposed. The general impossibility of focus in the propositional complement of factive predicates is not specific to Spanish. See Haegeman (2005, in press) among many others. Interpolation, verb-second, and the low left periphery in Old Spanish Geoffrey Poole 74 Iberia: IJTL | Volume 5.1, 2013, 69-98 ISSN: 1989-8525 doi:tba http://revistas.ojs.es/index.php/iberia/ With specific reference to verum focus, Leonetti & Escandell-Vidal (2009:184) observe that since the proposition embedded under a factive predicate is presented as true, it is not appropriate for it to contain verum focus fronting since ‘the discourse function of the VFF construction is that of selecting and emphatically asserting the positive option’. Nonetheless, interpolation can be found without difficulty in the propositional complement of various factive predicates, including see (9), know (10) and forget (11). (9) Et desque vio que lo non fazia…. and since saw.3sg that it neg would.do.3sg ‘And since he saw that he wouldn’t do it….’ (El Conde Lucanor, 14th c.) (10) Empero que supiesen que lo no podian fazer sin matar a hector. But that knew.3pl that it not could.3pl do.inf without kill.inf a H. ‘However, they knew that they couldn’t do it without killing Hector.’ (Historia Troyana, 14th c.) (11) maraujlla seria si la podrias guardar njn olujdar que la non wonder would.be if it could.2sg keep and.not forget.inf that it ‘neg’ descubrieses. discovered.2sg ‘It would be a miracle if you could keep [the secret] and not forget that you had discovered it.’ (Castigos y documentos para bien vivir, 13th c.) These examples again suggest that interpolation is not an instance of focus fronting since it should not be generally possible in the complements of factive predicates. 2.1.3 Interpolation of ‘pleonastic’ negation A third context-independent reason to reject a focalization analysis of interpolation comes from examples involving ‘pleonastic’ negation. As in Modern Spanish, sentential negation is licensed in the complement clauses of verbs in Old Spanish which express, for example, doubt, fear or prohibition, without contributing a negative interpretation to that clause.8 (12) ella dudase que el angel non dixese verdad. she doubted.3sgthat the angel ‘not’ said.3sg truth ‘She doubted that the angel was telling the truth.’ (Meditations of Pseudo-Augustine, 14th c.) 8 See, e.g., Keniston (1937: 605-7) for Old Spanish and Butt & Benjamin (2000: 330) for Modern Spanish. Interpolation, verb-second, and the low left periphery in Old Spanish Geoffrey Poole 75 Iberia: IJTL | Volume 5.1, 2013, 69-98 ISSN: 1989-8525 doi:tba http://revistas.ojs.es/index.php/iberia/ (13) Yo vengo a impedirque no se case con doñaLeonor don Pedro I come.1sg to prevent that ‘not’ se marrywith d. L d. P ‘I am coming to prevent Don Pedro from marrying Doña Leonor.’ (Servir a Señor Discrete, 16th c.) From the context, it is clear that (12) and (13) are not interpreted as a ‘double’ negation. It is not that the subject in (12) doubted that the angel was not telling the truth (i.e., that she believed that the angel spoke truthfully). Nor in (13) is the speaker planning to force Don Pedro to marry Doña Leonor (*‘I am coming to prevent them from not marrying’). It seems as though sentential negation in these clauses is not contributing anything semantically. It is certainly not negating the clause that it appears in. Crucially, ‘pleonastic’ negation of this sort is able to undergo interpolation. In (14), interpolation is seen in the complement clause of the verb guardar ‘to keep/prevent’ while in (15) it is in the complement clause of negar ‘to deny’. (14) Mas conujene que nos guardemos que les non demos las viandasquando more agree.3sg that us keep.1pl thatthem‘not’give.1pl themeats when la fiebrefuere ensuestado the fever is.3sg in itsstate ‘Furthermore, we should keep from giving them meat while the fever persists.’ (Tratado de las Fiebres, 15th c.) (15) podrian quebrantar las arcas & los çilleros / & tomar lo que quisiessen could break.inf the chests the storehousesand take.infit that wanted.3pl /& despues negar que lo non tomaron. and after deny.inf that it ‘not’ took.3pl ‘They could break into the chests and storehouses and take what they wanted and then later deny that they took it.’ (Ordenamiento de Alcalá, 14th c.) Just as in (12) and (13), the interpolated morpho-syntactic negation does not negate the clause that contains it in (14) and (15). (14) is not an exhortation to restrain vegetarian impulses (*‘We should keep ourselves from not giving them meat’), nor does (15) suggest an admission of guilt (*‘They could deny that they didn’t take it’). The interpolated sentential negation appears not to contribute to the semantics of its clause at all. As such, it seems impossible to imagine that interpolation in (14) or (15) instantiates focus of any kind, much less, hypothetically speaking, some sort of polarity focus (See also example (11) in the previous section). Similar facts are also seen with another type of pleonastic negation which, again as in Modern Spanish, can appear in certain adjunct clauses, particularly those introduced by hasta ‘until’, when the main clause is negated. Interpolation, verb-second, and the low left periphery in Old Spanish Geoffrey Poole 76 Iberia: IJTL | Volume 5.1, 2013, 69-98 ISSN: 1989-8525 doi:tba http://revistas.ojs.es/index.php/iberia/ (16) ques non mouio daquel logar.fasta que ella non torno sana that.se not moved.3sg from.that place until that she ‘not’ turned.3sg healthy ‘[The people of Israel] didn’t move from there until she returned to health.’ (General Estoria I, 13th c.) In (16) the negation in the until-clause is clearly not negating the clause that contains it. (It is not even clear that such an interpretation would be coherent). However, it is possible for pleonastic negation of this type to undergo interpolation. (17) Et dixieron a Bernaldo que nunqua se partirien del; fasta que And said.3pl to B that never se leave.cond.3pl of.himuntil that el Rey le non diesse a so padre. the kingto.him ‘neg’ gave.3sg to hisfather ‘And they said to Bernaldo that they would never leave him until the king gave it (back) to his father’ (Estoria de España II, 13th c.) Once again, prima facie, it is difficult to see how a semantically vacuous element could constitute any kind of focus. 2.1.4. The wider discourse context The previous sections have suggested that interpolation is not a focalization process without the need for any interpretation of a discourse context. However, examples of interpolation can also be readily found in which one can fairly confidently claim that the interpolated element is not focalized. Consider (18):9 (18) Et dixieron a la oliua como en uez de omne. Sey tu nuestro rey. / Respuso les ella como podria yo dexar mi grossura de que husan los Dioses & los ombres & es a grant pro & a grant seruiçio. de todos & uenir a seer rey entre los fustes. & la oliua non quiso ser so Rey: / Pues que la oliua lo non quiso dixieron lo a la figuera. & respuso les ella otrossi ‘And they spoke to the olive tree as if it were a person: “You be our king”. It replied to them: “How could I abandon my bounty which the gods and men use and which is a great benefit and service to everyone and come and be king among the trees?” And so the olive tree didn’t want to be their king. Since the olive tree didn’t want to do it, they asked the fig tree, and it replied to them similarly.’ (General Estoria II, 13th c.) In (18), the entire clause containing interpolation pues que la oliua lo non quiso ‘since the olive tree didn’t want [to do] it’ simply repeats information just given in the immediately preceding sentence la oliua non quiso ser so Rey ‘the olive tree didn’t want to be their king’. There seems to be no sense in which any of the clause containing 9 I omit for reasons of space the morpheme-by-morpheme gloss for the examples in (18) and (19) as the argument hinges on the larger discourse context, the relevant portions of which are explicated in the text. Interpolation, verb-second, and the low left periphery in Old Spanish Geoffrey Poole 77 Iberia: IJTL | Volume 5.1, 2013, 69-98 ISSN: 1989-8525 doi:tba http://revistas.ojs.es/index.php/iberia/ interpolation, much less the interpolated sentential negation itself, can be plausibly interpreted as contrastive, to constitute new information or instantiate polarity focus. Interpolation is also possible of elements which clearly represent old, rather than new, information. In (19), for example, the interpolated element esto ‘that’, resumes the recently mentioned event que se salliesen de su tierra ‘that they should leave his land’, said event having been in fact already resumed once by esto in the sentence immediately preceding the one containing the interpolated instance of esto. (19) Et estonçe les dixo el Rey que se salliesen de su tierra Et aquella gente a qujen esto dixo fueron se a la villa & tanto que les esto dixo luego se armaron muy bien & venjeron se al Rey onde yazia en su alcaçar & lidiaron conel & lo mataron ‘And then the king said to them that they should leave his land. And those people to whom he said that went to the town and as soon as he said that to them they armed themselves well and went to the king where he rested in his fortress and fought with him and killed him.’ (Crónica de 1344 I, 14th c.) The interpolated instance of esto in (19) is clearly not any kind of new information, and, like the example in (18), it seems impossible to interpret it as contrastive or emphatic in any way. 2.2. G-Topics and interpolation Having argued in the previous section that interpolation is not a focalization process, in this section I suggest that many instances of interpolation, particularly full NP and PP interpolation, are plausibly analysed as ‘given’ or ‘familiar’ topics in the sense of Bianchi & Frascarelli (2010) and Frascarelli & Hinterhölzl (2007). 2.2.1. Bianchi & Frascarelli’s (2010) typology of topics As noted by Frascarelli & Hinterhölzl (2007), the notion ‘topic’ encompasses a number of different elements which perform a number of different functions. Specifically, they identify three different types of topic: (1) the ‘aboutness topic’, which indicates in which entry within the conversational common ground the proposition expressed by the sentence should be stored, (2) the ‘contrastive topic’, which introduces alternatives with respect to other topics and (3) the ‘familiar topic’, which is used for retrieval of information already present within the conversational common ground content and is connected with topic continuity. Bianchi & Frascarelli (2010) label these three topics A-Topics, C-Topics, and G-Topics respectively. Interpolation, verb-second, and the low left periphery in Old Spanish Geoffrey Poole 84 Iberia: IJTL | Volume 5.1, 2013, 69-98 ISSN: 1989-8525 doi:tba http://revistas.ojs.es/index.php/iberia/ ‘We do not discuss here the lives of the Holy Fathers of the Old Testament before the Flood and after the Flood, 900 years and 800 years respectively, because we believe that more by faith.’ (Jardin de Floras Curiosas, 16th c.) (37) Et quando Diana andaua a caça o a correr so mont & And when D went to hunt.inf or to run.inf her mountain and la siesta la alla tomaua… the nap it there took.3sg ‘And when Diana went hunting or to climb her mountain and took a nap there…’ (General Estoria II, 13th c.) Actual clitic left-dislocation in the context of interpolation, as seen in (36) and (37), is extremely rare, as clitic-doubling of indirect and direct objects increases diachronically as interpolation is decreasing (see Fontana 1993:375-382) for discussion and Sitaridou (2011:172) for additional references). However, as noted by Sitaridou (2011:170), preposed objects are often topical in Old Spanish even without clitic left-dislocation. It therefore seems plausible to analyse esto in (36) and la siesta in (37) as occupying TopicP within an articulated left-periphery, suggesting that the interpolation cluster, which immediately follows it, occupies the low left periphery. 3.2.2. Interpolation and recomplementation Localization of the interpolation cluster in the lower area of the left periphery gains further support from its relation to so-called ‘recomplementation’ phenomena, illustrated in (38):15 (38) Et dios... quiera que los que este libro leyeren que se aprovechen And God want that the.pl that this book will read.3pl that se benefit.3pl del costunbre de.... from.the custom of ‘And God grant that those who read this book benefit from the habit of….’ (El Conde Lucanor, 14th c.) In (38), there is an instance of the complementizer que ‘that’ immediately following the verb ‘to want’, but there is also what appears to be a second complementizer following the heavy pre-verbal subject los que este libro leyeren ‘those who read this book’. Crucially, when a clause contains both recomplementation and interpolation, the interpolation cluster follows the second complementizer. 15 Fontana (1993: 234) describes recomplementation as the unmarked option with some verbs in the texts from the 13th to the 15th centuries that he examines, while Pountain (2001: 95) characterizes recomplementation as a feature of a plain, colloquial register. Interpolation, verb-second, and the low left periphery in Old Spanish Geoffrey Poole 85 Iberia: IJTL | Volume 5.1, 2013, 69-98 ISSN: 1989-8525 doi:tba http://revistas.ojs.es/index.php/iberia/ (39) ca creo que ninguna dueña que mas pare de vna criatura Because believe.1sg that no woman that more give.birth.3sg of a child que se no puede saluar de adulterio that se not can.3sg save.inf from adultery ‘because I believe that no woman who has ever had a child is immune to adultery’ (Gran Conquista de Ultramar, 13th c.) Analyses of the cartography of recomplementation differ slightly in their details, but all locate the second complementizer in an area immediately following the higher topic area within the left periphery.16 Like the data in (32) to (37) concerning subjects and topical non-subject NPs and PPs, this then suggests that interpolation occupies a low area of the left periphery, which is expected if it targets a functional projection such as FamP. 3.2.3. A note on the object pronoun Given that the object pronoun, as the leftmost element of the interpolation cluster, immediately follows the second complementizer in recomplementation structures and is in turn immediately adjacent to the interpolated G-Topic, it is presumably located in a position relatively close to FamP. Indeed, some previous analyses of interpolation (e.g., Martins) assume that the object pronoun moves to the same functional category as the interpolated element. If that were the case here, this would suggest that the object pronoun is in the outer specifier of FamP, and this seems plausible from an information structure point of view. Frascarelli & Hinterhölzl (2007:2) characterize familiar topics, for example, as ‘typically destressed and realized in a pronominal form’. This requires claiming that object pronouns in Old Spanish are (or at least can be) XPs rather than Xos, but this conclusion seems independently warranted, given that some interpolated elements, such as full NP subjects and PPs, are clearly XPs (see also Rivero 1992 for discussion). Full NPs and PP cannot be adjoined to the verb in a head position, and therefore the object pronoun which precedes them cannot be either. 3.2.4. Adverb distribution Particularly since Cinque’s (1999) influential study, the position of the finite verb relative to adverbs has become a standard test to determine how high the finite verb has raised. If, as the previous sections have suggested, the interpolation cluster does indeed occupy the low left periphery, we would expect that to be reflected in its position 16 For example, Martín-González (2002) claims that it occupies a second ForceP projection between TopicP and FocusP (a proposal tentatively adopted by Demonte & Fernández-Soriano (2009)), while Rodríguez Ramalle (2003) (cited in Demonte & Fernández-Soriano (2009)) and Villa-Garcia (2011) (cited in Haegeman (in press)) analyse it as a topic marker in the head of TopicP. See Demonte & Fernández-Soriano (2009: 44-47) for some discussion. Interpolation, verb-second, and the low left periphery in Old Spanish Geoffrey Poole 86 Iberia: IJTL | Volume 5.1, 2013, 69-98 ISSN: 1989-8525 doi:tba http://revistas.ojs.es/index.php/iberia/ relative to various high adverbs on the Cinque hierarchy. Unfortunately, Old Spanish did not possess an extensive inventory of high, speaker-oriented adverbs (compare Sitaridou’s 2012:587) remarks concerning Old Portuguese, which appears to be similar). However, an interesting picture emerges when we compare the distribution of adverbs relative to (non-)interpolation (restricting the discussion to interpolation of negation to facilitate the comparison).17 Where negation is not interpolated, the Corpus del Español contains over 700 examples of an embedded clause in which an adverb appears immediately to the left of negation (i.e., que/si – Adv – NEG – ObjPn – Vfin) during the 13th-16th century period.18 (40) is one such example: (40) si assi no lo fiziere if thus neg it do.3sg ‘if he doesn’t do it this way’ (Siete Partidas, 13th c.) By contrast, when the same set of elements displays interpolation (i.e., que/si – Adv – ObjPn – NEG – Vfin), the number of examples drops by over 99%: there are only five individual examples in which an adverb precedes an interpolation cluster, as in (41): (41) Et si assi lo non fizieredes and if thus it neg do.2pl ‘and if you don’t do it this way’ (Documentos castellanos de Alfonso X - Castilla la Vieja, 13th c.) If the relevant elements were in a lower position when negation was not interpolated, but in a higher position when it was, this radically different distribution would be a natural consequence.19 The fact that adverbs which can appear at the edge of or outside of the TP domain commonly precede an uninterpolated cluster but never 17 Just as in Tables 1 and 2 above, comparing interpolation and non-interpolation of negation means that there is only one grammatical uninterpolated word order to consider and it differs minimally from the interpolated order. 18 Tagging issues in the Corpus del Español make the precise number difficult to determine, but this is the correct order of magnitude. 19 See the appendix, and footnote 30 in particular, for an argument that this complementarity is not due to any kind of intervention effect (under the assumption that at least some adverbs under discussion undergo movement to a higher position, rather than being base-generated there). However, there are certainly adverbs in examples like (40) which are relatively high on the hierarchy (e.g., entonces ‘then’ and ahora ‘now’). Interpolation, verb-second, and the low left periphery in Old Spanish Geoffrey Poole 87 Iberia: IJTL | Volume 5.1, 2013, 69-98 ISSN: 1989-8525 doi:tba http://revistas.ojs.es/index.php/iberia/ precede an interpolated one indirectly suggests that the interpolated cluster does indeed occupy a higher position, one outside the TP domain.20, 21 3.3. Verb-second and interpolation of negation As has been mentioned, sentential negation is the most commonly interpolated element. However, it would seem to represent a challenge to the approach outlined thus far, in which the interpolated element is located in FamP in virtue of being a familiar or given topic. I take it that, by its very nature, sentential negation is generally incapable of serving as any kind of topic. Nonetheless, I assume that sentential negation genuinely occupies the category I have been referring to as FamP. The various category-independent generalizations concerning interpolation (the adverb facts discussed above, as well as the issue discussed in the appendix) and the obligatory adjacency between the interpolated element and the object pronoun would seem to be impossible to account for otherwise. In other words, I take an example like (1) (repeated below) to have the same structure as (27) above -that is, the structure in (42), and not for example (43), in which only the pronoun occupies FamP-: (1) se me non quisieres creer if me not wish.2sg believe.inf ‘if you do not wish to believe me’ (El Libro de Alexandre (O), 14th c.) (42) [ForceP se [FamP me non quisieres [TP creer (43) [ForceP se [FamP me [NegP non [TP quisieres creer However, examples such as (1) do raise the question of what would drive the attraction of sentential negation to a projection such as FamP. I suggest that interpolation of negation is an instance of ‘Formal Movement’ in the sense of Frey (2004, 2006). It involves attraction of the element which is at the left edge of the TP domain in order to fill a position in the left periphery, with no effect other than to satisfy an EPP feature. Formal Movement therefore is movement 20 Of course, strictly speaking, what the asymmetry suggests is that something is in a higher position in the interpolation case. However, see Section 3.3 for arguments that it is indeed all three elements of the interpolation cluster (the object pronoun, negation and the finite verb). 21 Given the absence of unambiguously very high adverbs and the well-known mobility of adverbs in general, I take the fact that various adverbs could follow the interpolation cluster not to provide the same help in fixing the location of the interpolation cluster, as it is potentially unclear what functional projection they occupy or are adjoined to. Interpolation, verb-second, and the low left periphery in Old Spanish Geoffrey Poole 88 Iberia: IJTL | Volume 5.1, 2013, 69-98 ISSN: 1989-8525 doi:tba http://revistas.ojs.es/index.php/iberia/ which has no pragmatic or semantic consequences, and it is not therefore necessary for negation to be a G-Topic. It is attracted simply as the closest available element, given that NegP immediately dominates TP in Spanish (see, e.g., Zagona 2002:195-6 for an overview and references). Attraction via Formal Movement would also account for the ability of ‘pleonastic’ negation to undergo interpolation, as in example (15) -repeated below-. (15) podrian quebrantar las arcas & los çilleros / & tomar lo que quisiessen could break.inf the chests the storehousesand take.infit that wanted.3pl /& despues negar que lo non tomaron. and after deny.inf that it ‘not’ took.3pl ‘They could break into the chests and storehouses and take what they wanted and then later deny that they took it.’ (Ordenamiento de Alcalá, 14th c.) Recall from Section 2.1.3 that these elements, though they have the morpho-phonological appearance of sentential negation, contribute nothing to the semantics of their clause. Under the assumption that these elements do indeed occupy a low left-peripheral functional projection such as FamP, they can only have been attracted purely in order to satisfy an EPP feature. Frey assumes that Formal Movement plays a key role in the verb-second constraint in German, and Light (2012) extends Frey’s analysis to account for verb-second phenomena in a number of other Germanic varieties. If Formal Movement is the operation by which non-G-Topics end up in FamP, then interpolation can be directly connected to a verb-second constraint in Old Spanish.22 By hypothesis, FamP bears a strong EPP feature, and this results, in part, in the finite verb raising to the head of FamP. When there are no other G-Topics beyond the object pronoun (which cliticizes to its left23), the element at the left edge of the TP domain is attracted, and, in cases 22 For the claim that Old Spanish possessed such a constraint, see, e.g., Fontana (1993, 1996, 1997) and Ledgeway (2011) but compare Mackenzie (2010) and Sitaridou (2011) and see Sitaridou (2011) for an overview and references. Various authors (e.g., Haegeman (1997), Roberts (1999), Poletto (2002) and Westergaard & Vangsnes (2005), among others) argue that the verb raises at least as high as Fino in verb-second clauses, and although I have used Frascarelli & Hinterhölzl’s FamP to label the functional category implicated in interpolation, I believe that the analysis is unaffected if interpolation effects were to turn out to be localized in FinP instead. See the appendix for some additional discussion concerning the cartography of the low left periphery. 23 As is well-known (e.g., Pountain (2001: 264)), it is not until the 15th century that clitics appear sentence-initially or after an intonational break. Prior to this point, clitics must be supported by some element to their left. Interpolation, verb-second, and the low left periphery in Old Spanish Geoffrey Poole 89 Iberia: IJTL | Volume 5.1, 2013, 69-98 ISSN: 1989-8525 doi:tba http://revistas.ojs.es/index.php/iberia/ such as (1), that element is sentential negation.24, 25 However, when for independent reasons there is another G-Topic occupying the inner specifier of FamP (in addition to the pronoun), Formal Movement is not triggered. The optionality of interpolation then reduces to the presence or absence of the relevant EPP feature in FamP. Viewed in this way, the proposal can be connected to Westergaard (2009)’s micro-parameter analysis of verb-second. Under her approach, verb-second is not itself a single, monolithic parameter. Rather, the traditional verb-second effect arises from the interaction of a number of different parameters regarding verb-raising relating to distinct functional heads. Certain constructions introduce particular functional heads, and these may or may not trigger raising of the verb (and therefore verb-second) on a head-by-head basis.26 A Formal Movement analysis may also be appropriate for at least some instances of interpolation of personal pronoun subjects. Old Spanish, like its modern counterpart, is a pro-drop language. Overt personal pronoun subjects are used for emphasis, contrast or a switch in reference. Given this usage, at least some cases of pronoun subject interpolation seem unproblematic. (44) Capitulo.cl. como vn dia que el conde entro a ver la condessa & chapter 150 how one day that the count entered to see.inf the countess and ella tenia sus tres fijos so el manto & como no se quiso she had.3sg her three children under the robe and how neg se wanted.3sg leuantar a el ni yr avn que la llamo & de como gelo dixo raise.inf to him nor go.inf even that her called.3sg and of how to.her.it said.3sg & dela respuesta que le ella dio. And of.the response that to.him she gave.3s ‘Chapter 150: how one day the count came in to see the countess, and she was breastfeeding her three children and how she didn’t want to get up for him or to go even though he called her, and of how he spoke to her and of the reply that she gave him.’ (Gran Conquista de Ultramar, 13th c.) 24 This assumes, as mentioned above, that sentential negation is an XP rather than a head. Interestingly, the ‘full’ form of sentential negation non is disproportionately favoured in interpolation contexts, particularly in the 14th and 15th centuries. For example, while the sequence non Vfin is slightly less than eight times more common than no Vfin during the 14th century, it is nearly 25 times more commonly interpolated. 25 Thus there appears to be some version in Old Spanish of a contrast discussed by Frey/Light, which in Germanic is between ‘Formal Movement’ and ‘true A´-movement’ (though I leave open the question of whether a ‘true’ G-Topic in the inner specifier is the result of movement or base-generation). 26 This is surely ultimately the locus of the explanation for why interpolation is restricted to embedded clauses, assuming that the functional head in question also attracts the object pronoun, but I must leave a detailed investigation of the issue for further research. Interpolation, verb-second, and the low left periphery in Old Spanish Geoffrey Poole 90 Iberia: IJTL | Volume 5.1, 2013, 69-98 ISSN: 1989-8525 doi:tba http://revistas.ojs.es/index.php/iberia/ In (44) the interpolated pronoun simply indicates that the subject of the verb give has switched back from ‘the count’ (the subject of the previous sentence) to ‘the countess’. It is clearly not emphatic or a contrastive focus of any kind. Since both of these third-person entities are clearly accessible from the conversational common ground, it seems reasonable to analyse ella ‘she’ in (44) as a G-Topic. However, there are other interpolated pronoun subjects which do not seem to be associated with mere topicality. (45) manifiesta cosa es que lo feziste a tuerto et sin pecado que te él fiziese manifest thing is that it did.2sg unjustly and without sin that you he did.3sg ‘It’s clear that you did it unjustly and without him having done you any wrong.’ (Calila e Dimna, 13th c., Poole 2007:13) In contrast to (44), the context in (45) appears to be contrastive and possibly emphatic. While it is true that the mere fact that a pronoun is used means that the referent is accessible from the conversational common ground, the claim that the interpolated pronoun subject in (45) is a G-Topic seems less immediately plausible. However, under the assumption that the pronoun subject in (45) is, like sentential negation, attracted by Formal Movement as the element at the edge of the TP domain, the lack of a G-Topic interpretation is expected. 4. Conclusion In this paper, I have considered the phenomenon of interpolation in Old Spanish from the point of view of both information structure and syntax. Contrary to a number of previous analyses, interpolation cannot be a species of focus movement. There are contexts which seem entirely unambiguous, but there are also a number of arguments which are context-independent. I proposed instead that many interpolated elements (including full NPs, PPs and adverbs) could be understood as given or familiar topics (G-Topics in the sense of Bianchi & Frascarelli 2010). These are the lowest of the left-peripheral topics, and various word order and distributional observations appear to confirm that the interpolation cluster occupies such a position. I suggested that specifically the interpolation cluster occupies the specifiers and head position of Fam(iliar)P, the low left-peripheral projection which is the locus of G-Topics under Bianchi & Frascarelli’s analysis. However, interpolation is possible of some elements which are less plausibly claimed to be G-Topics, and it is these that reveal the role played by the verb-second constraint. If no G-Topic is hosted by the inner specifier of FamP (the object pronoun occupies the outer specifier), sentential negation or a nonG-Topic subject at the left edge of the TP domain is attracted via Formal Movement (Frey 2004, 2006; Light 2012) to serve as the ‘first-position’ element. Interpolation, verb-second, and the low left periphery in Old Spanish Geoffrey Poole 91 Iberia: IJTL | Volume 5.1, 2013, 69-98 ISSN: 1989-8525 doi:tba http://revistas.ojs.es/index.php/iberia/ Appendix: the cartography of the low left periphery Throughout this paper I have adopted the cartographic approach to the left periphery, but this has been essentially for expository convenience. While at least some Old Spanish phenomena would seem to support the approach in broad outline (for example, recomplementation phenomena suggest the presence of at least two heads), and Fernández Ordóñez’s (2008-2009) study seems to provide the basis for a mapping, nothing in the proposed analysis of interpolation requires that there be a one-to-one (or very few-to-one) relationship between information structure roles and functional syntactic categories. However, one observation concerning interpolation does seem to have some bearing on the specifics of the low left periphery under the cartographic approach. Recall from Section 3.1 that the area between TopicP and FinP has been claimed to contain functional categories connected with both Focus and Topic, as (28) above indicated. (28) [ForceP [TopicP [FocusP [TopicP [FinP [TP Based on the analysis proposed here, FamP would seem to be a plausible instantiation of the lower TopicP in (28). However, interpolation appears not to co-occur with wh-operators, which are claimed (even in embedded clauses in Spanish) to occupy FocusP. Crucially, this complementarity is not seen with respect to relative clause operators, which, by hypothesis, occupy ForceP, a higher position in the articulated CP. Although arguments from non-appearance must always be taken as tentative, the observed distribution would follow from the assumption that in Old Spanish there is only one functional category below the higher TopicP, rather than two, and that this category may be have either topic properties or focus properties.27 The pre-Golden Age period, when interpolation is at its height, offers no shortage of instances in which a verb selecting for a [+wh] CP contains a wh-operator in that clause. (46) ca non sabedes quien vos quiere mal o quien vos quiere bien Because not know.2pl who you wants.3sg evil or who you wants.3sg good ‘because you don’t know who desires you ill and who desires you good.’ (Castigos y documentos para bien vivir, 13th c.) 27 Interestingly, Martín-González (2002: Ch. 2 fn 26) also notes that in Modern Spanish low left-peripheral topics are incompatible with non-D-linked wh-phrases (which he assumes occupy FocusP). Interpolation, verb-second, and the low left periphery in Old Spanish Geoffrey Poole 92 Iberia: IJTL | Volume 5.1, 2013, 69-98 ISSN: 1989-8525 doi:tba http://revistas.ojs.es/index.php/iberia/ Furthermore, when verbs which can select either a [+wh] or a [-wh] CP (such as saber ‘to know’) select a [-wh] CP, interpolation is possible in the complement clause: (47) biensabian que se no podrian defender contra la fuerça de tanta well knew.3pl that se not could.3pl to.defend against the force of such gente de moros people of Moors ‘They knew very well that they couldn’t defend themselves against the might of so many Moors.’ (Gran Conquista de Ultramar, 13th c.) However, there appear to be no examples like the constructed (48), in which a CP complement of a verb contains both a wh-operator and interpolation.28 (48) *ca non sabedes quien lo asi fiziese because not know.2pl who it thus did.3sg ‘because you don’t know who did it like that’ It is the low left periphery which is relevant, given that I assume, following Demonte & Fernández-Soriano 2009’s analysis of Modern Spanish, that wh-operators in Old Spanish target FocusP rather than ForceP in subordinate clauses. Old Spanish contains examples fully parallel to the Modern Spanish examples discussed by Demonte & Fernández-Soriano (2009:29-31), in which a wh-operator can be preceded by a complementizer: (49) E dini de monçon pregunto que quien gela pornia enla cabeça: 28 Clearly it is not the object pronoun or the verb which creates the complementarity with wh-operators. Examples such as (i) are of course plentiful: (i) & non pudo saber quien lo ferio and not was able.3SG to.know who it did.3SG ‘and he couldn’t know who had done it’ (Historia Troyana, 14th c.) The example in (ii) even contains an instance of a wh-operator in a negated clause which contains an object pronoun: (ii) y un batricajo le dio tan cascante que no sé quien and a blow him gave.3SG so shattering that NEG know.1SG who no se espante not SE frighten.3SG ‘and he gave him a blow so shattering that I don’t know who wouldn’t have been frightened’ (Teatro Completo, 16th c.) Thus the complementarity does appear to derive specifically from the interpolated element, rather than any other sub-element of the interpolation cluster. Interpolation, verb-second, and the low left periphery in Old Spanish Geoffrey Poole 93 Iberia: IJTL | Volume 5.1, 2013, 69-98 ISSN: 1989-8525 doi:tba http://revistas.ojs.es/index.php/iberia/ And D of M asked.3sg that who se.it would.put.3sg on.the head & dixo el que el obispo que es mas alto hombre de toda aquella hueste and said he that the bishop who is most high man of all those followers le tenia de coronar: him had.3sg of crown.inf ‘And Dini de Monçon asked who would put it on his head. And he said that the bishop, who was the highest ranking person among them, had to crown him.’ (Gran Conquista de Ultramar, 13th c.) (50) & començaron a grandes bozes a preguntar que que gente era aquella And began.3pl at great voices to ask.inf that what people were.3sg those que estaua encima dela sierra that were.3sg on top of.the mountain ‘And they started loudly asking who those people were who were on top of the mountain.’ (Gran Conquista de Ultramar, 13th c.) In (49), the verb preguntar ‘to ask’ is followed by a complementizer, which is then followed by the wh-operator quien ‘who’. As shown by the continuation, this construction reports a genuine request for information, since the answer to the indirect question immediately follows. In (50) it is a wh-phrase que gente ‘what people’ which follows the complementizer. In the continuing context, not included for reasons of length, the speakers entertain several possibilities and eventually are told to go up and find out, again indicating that (50) reports a genuine request for information. However, if the complementizer occupies the Force head, then the wh-operator must be in a lower position, which, following Demonte & Fernández-Soriano, I take to be the specifier of FocusP.29,30 29 The literally three counter-examples that I know of to this generalization all involve por qué ‘why’, as illustrated by (i) (see also Rivero (1992: 245)). (i) Responde / que te fiç / por que me non diste dicha en answer.IMP what you did.1SG why me not gave.2SG happiness in quantas que ame which.FPL that loved.1SG ‘Answer me. What did I do to you? Why didn’t you give me happiness from any of those that I loved?’ (Libro de Buen Amor, 14th c.) However, Rizzi (2001) argues that in Italian ‘why’ is base-generated in the specifier of INT(errogative)P, a functional projection between TopicP and FocusP, in which case these examples in fact indirectly support the proposed analysis. Also, Zubizarreta (1998) observes that por qué is the one wh-phrase which is compatible with a focused constituent, again suggesting that the two occupy a different position. Notice however that (i) is also an apparent instance of main-clause interpolation, as are the other counter-examples noted by Rivero, which does raise questions about their status. 30 One might be tempted to argue that the incompatibility between wh-operators and inter-