TP-ELLIPSIS WITH A POLARITY PARTICLE IN MULTIPLE- COMPLEMENTIZER CONTEXTS IN SPANISH: ON TOPICAL REMNANTS AND FOCAL LICENSORS

This paper investigates TP-ellipsis with a polarity particle in Spanish, with particular attention to its occurrence in multiple-complementizer clauses (i.e., ... que XP que ellipsis licensor ellipsis site). The paper argues for a version of the standard ΣPaccount of TP-ellipsis, with the polarity/focal item (e.g., también/tampoco/sí/no) crucially involved in the licensing of ellipsis. It is argued that the XP-que sequence is hosted in TopicP. The XP is a contrastive topic functioning as the remnant of TP-ellipsis and que instantiates recomplementation que –a topic marker. The ellipsis licensor, for its part, sits in a CP-related projection that follows TopicP. In light of the parallelism drawn here between the polarity elements that can partake in the licensing of TP-ellipsis and run-of-the-mill foci, the paper provides a host of arguments from Spanish and related languages underscoring the role of focus in this type of ellipsis operation, which may ultimately lead to the conflation of ΣP and FocusP. Similarly, I explore the relationship established between TopicP (the remnant) and ΣP/FocusP (the ellipsis licensor) in the process of TP-ellipsis as well as offer a principled account of the inability of jussive/optative que to survive ellipsis, unlike recomplementation que. The results of this paper therefore have wide-ranging consequences for the derivation of TP-ellipsis phenomena alongside the delineation of the preverbal field in Spanish and beyond.

(1) Dice que conmigo, que no va, y que contigo, que tampoco va says that with-me that not goes and that with-you that neither Lit.'S/he says that with me, s/he is not going, and s/he also says that with you, she won't go either.'On the one hand, examples like (1) include two que complementizers framing a left-dislocated phrase (e.g., conmigo and contigo) in the left periphery of each embedded clause, and on the other, the second clause is elliptical, illustrating the phenomenon of TP-ellipsis with a polarity element in Spanish.This second clause includes a CLLDed XP remnant sandwiched between ques, followed by a polarity element (tampoco '(n)either'), which precedes the elliptical site/gap. 1The configuration of interest is schematized in (2). (2) … que XP que polarity particle ellipsis site Although the major focus of this paper is Spanish, the phenomenon of TP-ellipsis in recomplementation contexts is not only found in this language, but in other varieties displaying multiple complementizers as well.For instance, Romance languages like Asturian, Catalan, Galician, and Portuguese exhibit a pattern of behavior similar to that of Spanish, as indicated by the data in (3) ((3b), (3c), and (3d) were kindly provided to me by Francesc González i Planas, Ricardo Bermúdez-Otero, and João Costa, respectively).
(3) a. Diz que col to collaciu, que nun va, pero says that with+the your friend/brother that not goes but que col míu, que sí that with+the mine that yes 'S/he says that she's not going with your brother, but with mine.' [Asturian] b.Em van dir que si plovia, que no vindrien a la festa, cl.go say that if rains that not come to the party i que si nevava, que tampoc and that if snows that neither 'They told me that if it rains, they won't come to the party, and that they won't come to the party if it snows either.' [Catalan] c.Dille que a el, que si lle pago o billete, tell-cl.that DAT him that yes cl.
pay the ticket pero que ao seu irmán, que de ningunha maneira but that DAT +the his brother that of none manner 'Tell him that his ticket, I will pay for, but under no circumstances will I pay for his brother's ticket.' [Galician] d.Diz que com o teu amigo, que nunca vai, says that with the your friend that never goes mas que com o meu, que sim but that with the mine that yes 'S/he says that s/he's never going with your friend, but with mine.' [European Portuguese] Villa-García (2012,2015) provides an account of cases like (1) in Spanish that relies on the standard Spec-Head agreement requirement on ellipsis first proposed in Lobeck (1990Lobeck ( , 1995) ) and Saito and Murasugi (1990), as shown in (4).The dislocated phrase contigo and recomplementation que are assumed to establish feature-checking in a Spec-Head configuration, making ellipsis of the complement of the functional head que possible, on the assumption that functional heads can license ellipsis of their complement only when they undergo Spec-Head agreement.
(4) ... y que [ TopicP contigo [ Topicº que [ TP no va ]]] tampoco However, I show that this account runs into a number of empirical problems, and instead argue for a ΣP account of (1) wherein polarity particles such as tampoco '(n)either' are critically involved in the licensing of TP-ellipsis, in much the same way as has been proposed for ordinary cases of TP-ellipsis in Spanish without recomplementation.The proposal is schematized abstractly in (5): (5) .

. y que [ TopicP contigo [ Topicº que [ ΣP tampoco [ Σº Ø [ TP va]]]]]
Some welcome results of this move include the fact that sentences with and without recomplementation que are treated in the same way syntactically; the structural position and the function of polarity particles such as tampoco are no longer mysterious, unlike under the account outlined in (4); and additionally we dispense with the need to appeal to Spec-Head agreement as a crucial factor in the licensing of ellipsis, an assumption of the analysis in (4) that has been shown to be problematic on independent grounds in different languages.I also explore the overarching consequences of the construction illustrated in (1) and the analysis in (5) for the syntax of Spanish, with particular attention to the geometry of the left periphery and clausal structure more generally.It is my hope that the interaction of these two complex aspects of the syntax of Spanish, namely recomplementation and ellipsis, will shed new light upon the analysis of each phenomenon.
I also investigate the relationship between ΣP and FocusP, for there is abundant evidence that they constitute the same projection hosting focal items.Thus, I provide evidence militating in favor of a refinement of (5) to the effect that ΣP and FocusP should be conflated, or at least that FocusP is crucially involved in the process TPellipsis with a polarity particle.The configuration in (1) also raises the interesting question of how to characterize the relationship between the sandwiched remnant and the ellipsis licensor, between which an overt head (i.e., recomplementation que) can intervene.To the best of my knowledge, how to technically implement the relationship between the topic remnant and the licensor of ellipsis is far from resolved, an area to which I hope to contribute.More generally, the findings reported here have far-reaching consequences for the account of TP-ellipsis -especially regarding cases featuring a polarity particle-as well as for the geometry of leftperipheral clause structure in Spanish and related languages.
The paper is organized as follows: in Section 2, I provide an overview of TPellipsis in Spanish; in Section 3, I review the empirical arguments supporting a TopicP analysis of recomplementation in Spanish; in Section 4, I assess the Spec-Head agreement account of ellipsis in light of the recomplementation + TP-ellipsis data and submit that it cannot be the correct account of the phenomenon; in Section 5, I outline the ΣP account of the phenomenon.This section is divided into further subsections, thus: in Section 5.1, I argue for the focal status of the polarity items involved in TP-ellipsis, which in turn supports a conflation of ΣP and FocusP in the relevant environments; in Section 5.2, I investigate the connection between the topical remnant and the focal ellipsis licensor and the syntactic implementation of such a relationship; in Section 5.3, I concentrate on jussive/optative que and show that the account adopted in the paper provides an immediate explanation for why this low complementizer cannot survive TP-ellipsis; in Section 5.4 I touch briefly on the issue of phases and ellipsis.In Section 6, I offer some concluding remarks.

Spanish TP-ellipsis with a polarity particle
As noted, the phenomenon of TP-ellipsis with a polarity element in Spanish has been investigated by a number of authors (e.g., Bosque 1984;Brucart 1987Brucart , 1999;;Brucart and MacDonald 2012;Depiante 2000Depiante , 2004;;Gallego 2016;López 1999López , 2000;;Saab 2003Saab , 2009Saab , 2010)).Unlike English, Spanish lacks VP ellipsis (Lobeck 1999 andZagona 1988), but it exhibits TP-ellipsis, as illustrated in (6).[Sampredo, La sonrisa etrusca, 1985, cited in RAE 2009: 3709] In TP-ellipsis contexts, the second conjunct does not contain a tense affix, which suggests that the elliptical site in Spanish does not correspond to VP, but to the next higher projection, that is, TP (Brucart and MacDonald 2012).That the ellipsis site in such cases includes tense is confirmed by the ungrammaticality of example (7), from Murguia (2004: 86), which shows that there cannot be a tense mismatch between the antecedent (i.e., the first conjunct) and the elliptical gap (i.e., the second conjunct) (Depiante 2000(Depiante , 2004;;Saab 2003Saab , 2009Saab , 2010)).This comes as no surprise, since tense is included in the elided part; hence it cannot differ from that of the antecedent clause.(7) *María ha leído mucho y Elena en el futuro también Mary has read much and Helen in the future too habrá leído mucho Intended meaning: 'Mary has read a lot, and Elena will have too (in the future).'Thus, irrespective of the right analysis of the licensing of ellipsis, the elliptical site in Spanish TP-ellipsis will begin with TP, as follows (though see Section 5.2 for a refinement): TP-ellipsis is readily available in turn-taking by different speakers (Brucart and MacDonald 2012), as shown in (9).In analogous fashion, a cataphoric relation can hold between the antecedent and the gap, provided that they belong to the same utterance (Brucart and MacDonald 2012), as indicated by (10).
(10) Quizás tú no pienses votar a Podemos, pero yo pienso votar a perhaps you not but I think vote to Podemos Podemos 'Maybe you won't vote for (the political party) Podemos, but I think I will.' Brucart (1987) first noted that TP-ellipsis is available in embedded contexts ((11) shows that either the embedded clause or the main one can function as the antecedent, resulting in ambiguity in this particular case): (11) Él piensa que María tiene razón, pero yo no tengo razón / pienso que he thinks that Mary has reason but I not María tiene razón 'Peter thinks that Mary is right, but I am not.' / 'Peter thinks that Mary is right, but I don't think so.' Nonetheless, the interaction of TP-ellipsis with subordination is complex (see Brucart and MacDonald 2012).As noted by López (1999), TP-ellipsis in Spanish is subject to both parallelistic and contrastive requirements with regard to the antecedent.
Similarly, this type of TP-ellipsis in Spanish relies on the overt presence of both a remnant contrastive topic and a polarity item such as tampoco '(n)either' or sí 'yes,'  in this order. 3This is straightforwardly corroborated by the following examples, where either the topical remnant (cf.(12a)) or the polarity element (cf.(12b)) are absent, resulting in strong ungrammaticality.The only licit option is for both the remnant (a Marina) and the polarity particle (sí) to both be present simultaneously, as shown by (12c).Crucially, the remnant cannot be just any preposed XP: it has to be a Clitic-Left Dislocated Phrase (CLLD), more concretely a contrastive topic (Bosque 1984, Brucart and MacDonald 2012, Depiante 2004, López 1999, RAE 2009, Saab 2010, among others).As López (1999: 266) puts it, "the remnant must have a contrastive reading vis-à-vis its correspondent in coordinate sentences."As is well known, in Spanish any phrase can be CLLDed, including subjects, direct and indirect objects, and adverbs/adverbial clauses. 4Note that parallelism does not require the phrase in the antecedent clause to which the remnant is contrasted to be left-dislocated; it can be situated the VP area: (13) Juan va a ir a Liverpool, y me dijeron que a Londres John goes to go to Liverpool and cl.told that to London también va a ir too 'John is going to Liverpool, and I've been told that he's also going to London.'An important property of contrastive topics worth mentioning at the outset is that they cannot be elided.They are similar to foci in that they enter into a contrast relationship, but they pattern with non-contrastive CLLD in that they occur with an attending clitic (see fn. 4) (RAE 2009(RAE : 2984(RAE -2985)).It is precisely their role as contrastive elements that makes them unelidable, unlike other topics.To illustrate this point, consider the non-elliptical sentence in ( 14), involving (pro)nominals that carry a contrastive function (that is, they are subjects acting as contrastive topics).While a CLLDed constituent can function as the remnant of TP-ellipsis, focalized phrases are not allowed to perform this function, as indicated by the contrast in ( 16), where capitalization indicates focus (see, e.g., Casielles-Suárez 2004 andLópez 2009 for differences between CLLD and focus in languages like Spanish).( 16 I return to the inability of focal phrases to occur higher than the polarity element below (the reader is referred to López 1999 for ample discussion of the impossibility of focused phrases to act as remnants in TP-ellipsis).
(17) María dice que a Pedro lo traen, pero que a Joaquín Mary says that DOM Peter cl. bring but that DOM Joaquín quizás no perhaps not 'Mary says that they will bring Peter, but perhaps not Joaquín.'I come back to particles other than sí/no/también/tampoco that can similarly license TP-ellipsis in Spanish in Section 5.1.I now turn to the phenomenon of recomplementation in Spanish, with particular reference to the syntactic analysis of the construction.

Recomplementation in Spanish and TopicP
Recomplementation is a phenomenon of spoken Spanish involving a CLLDed phrase sandwiched between two occurrences of the (homophonous) complementizer que /ke/ in the left periphery of an embedded clause (Demonte and Fernández-Soriano 2009 et seq.;Escribano 1991;González i Planas 2014;Rodríguez-Ramalle 2005;Uriagereka 1995;Villa-García 2012, 2015;inter alia).This construction, where the second instance of que appears to be optional, is exemplified for Spanish by the spontaneous data in (18).
(18) Dijo que como Marina ya no tenía fiebre, que la mandaron pa(ra) casa said that as Marina yet no had fever that cl.sent for home 'S/he said that because Marina no longer had fever, they sent her home.' [Principality of Asturias, Spain, August 2016] As for the analysis of such configurations, there have been a number of proposals in the literature (see Villa-García 2015 for a review).We will focus here on the TopicP analysis, first laid out in Rodríguez-Ramalle (2005), and on the arguments in its favor.Rizzi (1997 et seq.) has advanced the proposal that the CP domain should be split in (at least) four different categories, thus: Villa-García (2012, 2015) (see also references therein) provides a variety of empirical arguments to the effect that the dislocated phrase characteristic of recomplementation constructions sits in the specifier of TopicP, whose head is lexicalized by the secondary instance of que.Consequently, que functions as a reinforcement of topic or as a topic marker, which is actually the default option for embedded topicalization in Old Spanish (Fontana 1994).The TopicP analysis is provided in (20), which is simplified by only showing the constituents that are immediately relevant to the discussion at hand.First, the account in ( 20) is supported by the fact that recomplementation que relies on the occurrence of a sandwiched dislocated XP, as shown by the ungrammaticality of (21), which lacks a dislocated phrase framed by ques; therefore the second instance of que cannot occur.This is expected under (20), since the appearance of TopicP is contingent on the occurrence of a topic XP (in fact, for Rizzi 1997, TopicP and FocusP are projected on an as-needed basis, that is, when topical or focused phrases occur in the left periphery).5 (21) *Dijo que que mandó para casa a Marina porque ya no tenía fiebre said that that sent for home DOM Marina because no-longer had fever 'S/he said that they sent Marina home because she didn't have fever anymore.'Similarly, the TopicP analysis of recomplementation is strengthened by the fact that foci and wh-items cannot be hosted in sandwiched position; they must follow recomplementation que, as shown by the contrast in ( 22).Since such focal elements are assumed to occupy Spec, FocusP, it comes as no surprise that they routinely follow the material hosted in TopicP (cf.( 19 López (2009) (see also Garrett 2013) has proposed that the sequence sí que 'yes that' offers a topichood test, since the material preceding it must be topical in nature.Not surprisingly, the sequence XP + que can appear in sí-que contexts, as indicated by (24), which confirms the analysis according to which the XP and recomplementation que are hosted in TopicP -both items belong to the topic field (see Villa-García and González Rodríguez, in preparation, for differences between sí and sí que in Spanish).
(24) Gritó que a Lucía, que sí que le cobraron en el tren shouted that DAT Lucía that yes that cl.charged in the train 'S/he shouted that Lucía was in fact charged to travel by train.'A final argument that I will consider here, and which is highly relevant to the issue of TP-ellipsis, as we shall see, is the existence of yet another non-high complementizer, also lexicalized as que /ke/.In certain contexts, this que appears to be a manifestation of recomplementation que, but under closer scrutiny, it becomes clear that this que is in fact the realization of a different phenomenon.I am referring here to jussive/optative que, which is found in clauses where the verb displays subjunctive mood, and which are interpreted as desiderative or exhortative.This phenomenon is illustrated in (25b), which contrasts with (25a), which is a regular case of (optional) recomplementation que.In both cases, the matrix verb is a verb of communication which does not inherently select for either an indicative or a subjunctive complement.(Rizzi 2001), have to follow recomplementation que, perhaps due to their bearing a question feature (i.e., [+wh]), which is incompatible with the overt realization of the relevant head as que in Spanish.This is not immediately relevant to the present discussion, so I leave it for future research.
(25) a. Dice que como está amargada, (que) se va says that as is bitter that cl.leaves Ind. 'S/he says that because she is a bitter person, s/he is leaving.' b.Dice que como está amargada, *(que) se vaya says that as is bitter that cl.leave Subj.'S/he orders that she leave because she is a bitter person.' In (25b), the lower instance of que is mandatory.The embedded clause, which features a verb in the subjunctive, is interpreted as a command, as shown by the English paraphrase.Villa-García (2015) concludes on the basis of a number of differences between recomplementation que and jussive/optative que that the two homophonous complementizers constitute radically different phenomena.Thus, they should be analyzed as such: whereas recomplementation que spells out Topicº, jussive/optative que is the spell-out of Finitenessº, which Rizzi (1997) has independently argued is the locus of mood features.The analysis of the two complementizers is provided in (26) (see Villa-García 2015 for the claim that recomplementation que and jussive/optative que cannot normally be adjacent to one another and the mechanism banning their contiguous occurrence).7 The account in ( 26) immediately makes two correct predictions.For one thing, the two complementizers should be able to co-exist in the very same left periphery, since they occupy distinct positions.(27a) bears out this prediction, for in this example, the middle instance of que is optional, but the last one, which actually co-occurs with a subjunctive verb, is obligatory.Importantly, the sentence bears exhortative force.Moreover, jussive/optative que should not be dependent on a topical phrase to its left (cf.(27b,c)), unlike recomplementation que (see ( 21)).Accordingly, jussive/optative que can appear even in matrix clauses without a preceding dislocate (cf.(27c)).( 27 'S/he or I order(s) that s/he go away.'Other arguments militating in favor of a TopicP account of recomplementation include the possibility of enclisis in recomplementation contexts in a related Western Iberian Romance language, Asturian, an argument that I will not explore here for reasons of space (see Villa-García 2015: Ch. 3 for detailed discussion).In any case, the conclusion drawn from the arguments above is that the Spanish recomplementation data strongly argue for, and actually receive a unified account under, a TopicP analysis of the construction (cf.( 20)), and in turn substantiate Rizzi's split-CP approach (see, e.g., ( 26)). 8,9In what follows, I return to TP-ellipsis with a polarity particle in recomplementation contexts, the main object of study of this investigation.

The Spec-Head agreement account of ellipsis
To date, the only existing account of TP-ellipsis in multiple-complementizer environments that I am aware of is that of Villa-García (2012, 2015), based on the standard account of ellipsis dating back to Lobeck (1990Lobeck ( , 1995) ) and Saito and Murasugi (1990).These authors observed that functional heads can license ellipsis of their complement only when they undergo Spec-Head agreement (i.e., featurechecking), provided that other conditions on ellipsis, such as parallelism, are met.The following examples, adapted from Bošković (2008), illustrate this generalization.In (28a), a case of sluicing (TP-ellipsis involving a wh-item), who and Cº agree with each other (Fukui and Speas 1986; note that I use CP for ease of exposition); hence, ellipsis of the complement of Cº can be effected.In contrast, in (28b), the complementizer that in Cº does not enter into an agreement relationship with any element in its specifier (in fact, there is no constituent in Spec, CP), rendering ellipsis of its complement impossible.( 28c) is an instance of VP-ellipsis, and here again, Lucía and does establish a Spec-Head agreement relationship, making possible the elision of the complement of the functional head does.
(28) a. Chencho believes that his wife met someone but I don't know [ CP who i [ Cº  his wife met who i ]] b. *Chencho believes that his wife met a guy but I don't know [ CP [ Cº that his wife met a guy ]] c.Chencho travels a lot and [ TP Lucía [ Tº does travel a lot ]] too 8 Although I will not delve into this issue here for reasons of space and because it is not directly relevant to the discussion at hand, the reader is referred to Villa-García (2015) for extensive discussion and empirical justification of the claim that sandwiched dislocated phrases are base-generated in their surface position, framed between ques; they do not move to the left of recomplementation que, which furthermore creates an island/barrier for extraction.See López (1999) for the claim that the remnant of TP-ellipsis in Spanish is actually base-generated in the periphery (although my claim here will be that this is only the case when multiple-ques are present; otherwise, CLLD seems to be the result of movement, as argued by López 2009;Villa-García 2012, 2015;and Zubizarreta 1998;inter alia, although this is part of an on-again, off-again debate; see Saab 2010 with reference to the movement analysis of CLLD in TP-ellipsis in non-recomplementation environments; this is certainly relevant to the gap vs. internal structure debate hinted at in fn. 1). 9 The reader is referred to Villa-García (2015) for discussion of a number of open questions, such as why in matrix left-dislocation cases without a high que recomplementation is not possible.Discussing such aspects here would take us too far afield.
In Spanish, ellipsis in cases akin to English (28b) is also impossible, as shown in (29a), whose derivation is furnished in (29b In much the same way as in English, the head of CP in the Spanish sentence in (29a) is not an agreeing head; consequently, under the Spec-Head agreement account, ellipsis of the complement of Cº cannot be effected, given the agreement requirement on ellipsis (see ( 29b)).
However, as has been seen, ellipsis after recomplementation que is possible in languages like Spanish, as shown again in (30a,c).Thus, on this view, recomplementation que can presumably license ellipsis of its complement.The analysis of (30a,c) under this account is furnished in (30b,d).Note that the relevant bracketed structures employ TopicP instead of CP, as argued in the previous section.
(30) a. Me dijeron que si llueve, que no vienen a la fiesta, y cl. said that if rains that not come to the party and que si nieva, que tampoco that if snows that neither 'They told me that if it rains, they won't come to the party, and that they won't come to the party either if it snows.'(i) A: Pedro piensa que va a llover.B: Yo creo que también.
Peter thinks that goes to rain.I believe that too.'A: Peter thinks that it's going to rain.B: I also think so.'Although further investigation of cases like (i) is required, I suggest that this case is somehow idiomatic, a fixed form expressing the idea of I think so, too.In fact, the subject has to occur in the sentence obligatorily (cf.(ii)a) and it has to be preverbal (cf.(ii)b).An anonymous reviewer concurs with this claim and adds that también seems to modify the main clause, rather than the embedded clause, even though if surfaces after the complementizer.
(ii) a. A: Pedro piensa que va a llover.B: *Creo que también.Pedro thinks that goes to rain believe that too.b.A: Pedro piensa que va a llover.B: ?? Creo yo que también.

Peter thinks that goes to rain believe I that too
Raquel González Rodríguez (p.c.) adds that the idiomatic character of such constructions is reinforced by the fact that they seem to be limited to the speaker (first person).However, the issue as a whole may be more complex, since as another reviewer notes, these patterns are also possible with strongly assertive verbs such as asegurar 'assure,' which weakens the claim that they are idiomatic.In Villa-García (2012, 2015), I defended the view that the possibility of eliding the complement of recomplementation que lends support to an analysis whereby the projection headed by secondary que hosts the dislocated phrase in its specifier, with which it agrees, on the assumption that functional heads can license ellipsis of their complement only if they undergo Spec-Head agreement with their specifier (provided that other requirements on ellipsis are fulfilled).Thus, under the Spec-Head featurechecking analysis of ellipsis, elliptical recomplementation data like (30) prima facie argue that the XP sandwiched between complementizers and recomplementation que enter into a Spec-Head agreement relationship with each other, which substantiates the TopicP analysis of recomplementation configurations.
In the cited works, I also noted that TP-ellipsis is similarly licensed in ordinary cases of (root and embedded) dislocation without secondary que, as illustrated in (31a), which is the que-less counterpart of (30a).
(31) a. Me dijeron que si llueve  no vienen a la fiesta, y cl. said that if rains not come to the party and que si nieva  tampoco that if snows neither 'They told me that if it rains, they won't come to the party, and that they won't come to the party either if it snows.'b. … y que [ TopicP si nieva [ Topº Ø no vienen a la fiesta]] tampoco I claimed that examples like (31a) constitute evidence that an agreeing null head () is present in cases of (embedded) left-dislocation without secondary que, as indicated by the possibility of ellipsis.In other words, the dislocated phrases in (31a) enter into an agreement relationship with a null/covert head, which in turn can license ellipsis of its complement.
A further argument in support of the Spec-Head agreement account of elliptical structures came from the behavior of jussive/optative clauses (cf.(25b)) in ellipsis environments.As shown by the data in (32), there is a contrast between recomplementation que and jussive/optative que in the context of ellipsis.As the careful reader will note, whereas recomplementation que establishes an agreement relationship with the sandwiched dislocate (in TopicP), as in (32a), jussive/optative que, which spells out Finitenessº, does not have an element in its specifier with which to agree (i.e., … [ TopicP XP [ Topicº que [ FinitenessP [ Finitenessº que […]]]]]), rendering it unable to license ellipsis of its complement.This is confirmed by the unacceptability of (32b) and (32c).For most speakers, such sentences improve if the second instance of que in the second conjunct is absent, or if ellipsis does not occur.(I return to the asymmetry regarding the (im-)possibility of TP-ellipsis contingent on the type of low complementizer in Section 5.2, where I also offer a principled account of the contrast that follows from the analysis to be proposed.)All in all, the conclusion I reached in Villa-García (2012) was that the facts just reviewed indicate that the dislocate characteristic of recomplementation environments is located in the specifier of recomplementation que, and that the two items agree with one another, making possible the elision of the complement of recomplementation que.Nevertheless, such an account is not without problems, an issue that constitutes the focus of the next subsection.

Issues for the Spec-Head-agreement account of ellipsis
Even though the account outlined in the preceding subsection accommodates the Spanish recomplementation data quite successfully, it raises a number of non-trivial questions.The major issue posed by this approach is what is the structural position of items such as también 'too.' 11 Under the Spec-Head-agreement analysis of TP-ellipsis (cf.( 33)), the ellipsis licensor is recomplementation que, by virtue of establishing Spec-Head agreement (i.e., feature-checking) with its concomitant sandwiched XP.Therefore, it is puzzling that también is not part of the ellipsis site/gap and hence absent, since it occurs after que, the putative functional head charged with the licensing of ellipsis.

(33) … y que [ TopicP XP [ Topº que/ [ ]]] también
As suggested by the bracketed structures furnished so far, items such as también are not part of the structure of the sentence and instead stay outside.This is rather problematic, since también should occupy a slot in the structure and/or be part of the elliptical site.Under (33), it is not only the position, but also the role of the polarity element también, that remains shrouded in mystery.It is important to bear in mind that at least one polarity particle should occur for TP-ellipsis to be possible, as indicated by the contrasts in (12) above.I resume the discussion of the function and status of such particles in Section 5, but what matters for our current purposes is that constituents like también are not only mandatory, but they also need to occupy a position in the structure in the construction of interest.
In addition, the agreement/feature-checking requirement on ellipsis has been contested on independent grounds.The reader is referred to Bošković (2014) for relevant discussion and counterexamples from different languages, to Eguren (2009) for analogous issues with regard nominal ellipsis, and to Ticio (forthcoming) for evidence against an agreement-based account of ellipsis in the Spanish DP from the domain of bilingual first language acquisition.For instance, no obvious Spec-Head agreement relationship holds in some cases, such as in the following English sentence illustrating VP-ellipsis, generously provided to me by Jonathan Bobaljik: (34) I don't know if Jonathan left, but I think he may have left In (34), he agrees with may, but it is not at all clear that he enters into a featurechecking relationship with have, the functional head presumably responsible for licensing ellipsis here.In fact, a similar situation can be replicated for the multiplecomplementizer cases with which this paper is concerned.Consider ( 35): (35) Gritó que a Fili, que no va, pero que a San Francisco, que sí va shouted that to Philly that not goes but that to San Francisco that yes 'S/he shouted that s/he is going to San Francisco, not to Philly.' First of all, in previous work, I did not take such examples into considerationonly examples involving también and tampoco.Such particles could be specifiers (Laka 1990, López 1999, Suñer 1995), but elements like sí and no are generally deemed to be heads.Thus, in the same way he agrees with may in the English example in (34), and not with the ellipsis licensor have, it may well be the case that a San Francisco in (35) agrees with recomplementation que (after all, both elements are topic-related constituents under TopicP), but not with sí (cf. he may have left / a San Francisco que sí va), which I argue in Section 5 is in reality the head licensing ellipsis.Put differently, it seems that in both (34) and ( 35) the (functional) head that appears to be in charge of the licensing of ellipsis does not seem to establish a featurechecking relationship with an XP hosted in its specifier, weakening the Spec-Headagreement analysis of ellipsis. 12 Overall, despite its virtues, the Spec-Head agreement account of TP-ellipsis in Spanish seems untenable, for the reasons laid out here.I what follows, I show that the alternative ΣP analysis of ellipsis, which is the account standardly assumed for regular cases of Spanish TP-ellipsis, fares well with the data, while still supporting the TopicP account of recomplementation outlined above.

The ΣP account of TP-ellipsis in regular and recomplementation cases
In the remainder of this paper, I will explore the consequences of adopting an alternative analysis of ellipsis that does not rely on feature-checking.Authors including Brucart (1999), Brucart and MacDonald (2012), Depiante (2000Depiante ( , 2004)), de Cuba and MacDonald (2013), and Saab (2009) claim that particles such as también 'too' and tampoco '(n)either' license TP-ellipsis in Spanish in Laka's (1990) polarity-12 Raquel González Rodríguez (p.c.) points out that if the Spec-Head feature-checking requirement on ellipsis were on the right track, contrastive foci should be able to license ellipsis, since they are assumed to enter into an agreement relationship with a null focus head.However, this does not seem to be the case.Clearly, feature-checking in a Spec-Head alone cannot suffice to license ellipsis.
encoding projection ƩP, as shown in (36).I submit that the by-now standard ƩP analysis of TP-ellipsis in Spanish actually accounts for the data put forth here and solves the issues posed by the Spec-Head-feature-checking analysis, while still providing support for the TopicP account of recomplementation in Spanish.The analysis of ellipsis in recomplementation constructions raises interesting questions about the characterization of the relationship between the ellipsis licensor and the contrastively CLLDed remnant, as we shall see.
(36) ΣP as the projection responsible for the licensing of TP-ellipsis ΣP Σ' Σº TP ...Under (36), the ability of the XP + recomplementation que sequence to be part of the remnant also lends credence to the proposal that a close relationship exists between the two entities, since both of them would have to be dislocated to qualify as remnants (which must be topical, as has been seen).This is fully consistent with the two elements being located in a Topic projection, as shown in the arboreal representation in (37).Consequently, even though an analysis along the lines of (37) would not argue in favor of a feature-checking relationship between the sandwiched XP and recomplementation que per se, it would still be wholly compatible with the XP and recomplementation que (or the corresponding null head Ø) being located together under the same projection.In fact, such an analysis would also capture the lack of ellipsis with jussive/optative que, since this element is not part of the topic field and therefore is not appropriate to be part of the remnant, a matter which I tackle again in Section 5.2. 14 Note that ellipsis is also possible in cases of multiple lower ques, as exemplified in (38).This is expected if TopicP, the category that qualifies as the remnant, is a recursive phrase, as has been shown.In fact, iterative CLLDs can work as contrastive remnants in recomplementationless TP-ellipsis in Spanish (López 1999: 285 andSaab 2010: 83;cf. Matos 1992: 349 for evidence that this is however banned in Portuguese).Under the analysis currently pursued, there is nothing special about cases like (38): instead of one topical sequence (i.e., XP que) surviving ellipsis and acting as a remnant, there is more than one; the ellipsis licensor is still Σº, as in the regular cases without multiple ques.Similarly, these examples confirm the recursive nature of TopicP, which hosts the sandwiched XP and recomplementation que (cf.(38b)).
(38) a. Dijo que ella, que por eso, que no venía, y que él, said that she that for that that not come, and that he que entonces, que tampoco that then that neither 'S/he said that because of that, she is not coming, and then he's not coming either.' Furthermore, ellipsis is also licensed when the recomplementation structure involves more than one dislocate sandwiched between overt ques, as (39) indicates.This example further corroborates that there can be more than one contrastive topic per clause. 15  (39) Dijo que hoy, si llueve, que no viene, y que esta tarde, said that today if rains that not come, and that this afternoon si nieva, que tampoco if snows that neither 'S/he said she won't come today if it rains, and s/he won't come this afternoon if it snows either.' In a similar vein, ellipsis can be licensed in recomplementation contexts even if the second conjunct contains just one instance of XP + que, unlike the first one, as indicated by the data in (40), which again provides evidence for the recursion of the projection headed by secondary que, as argued above.
14 Martins (1994) extends the ΣP analysis of ellipsis to Portuguese.As Michelle Sheehan (p.c.) notes, investigating the differences between the type of Spanish TP-ellipsis with which this paper is concerned and its counterparts in closely related languages is worth exploring in future research.See López (1995) and Saab (2010) for discussion of the properties of Σ° that account for its ability to license ellipsis in Spanish. 15It is not clear whether cases like (39) involve multiple TopicPs with null ques or several stacked XPs in multiple specifiers of TopicP.Since this issue is not directly relevant to the discussion at hand, I will not pursue it further here.
(40) Dijo que hoy, que si llueve, que no viene, y que said that today that if rains that not come, and that mañana, que tampoco tomorrow that neither 'S/he said she won't come today if it rains, and s/he won't come tomorrow if it rains either.'Finally, note that the instance of recomplementation que in the second conjunct can also be null in ellipsis cases in which the first conjunct involves a different number of XP-que combinations, as illustrated by ( 41 As with (38), these different possibilities do not pose a problem for the current analysis and in turn confirm the iterative character of recomplementation que.In the following subsection, I turn to the connection between ΣP and FocusP and some potential consequences for the mapping of the left periphery.Recall that a broader goal of this paper is to explore the consequences of the analysis pursued here for the much-debated architecture of the CP domain.

On the focal behavior of the ellipsis licensor
If we take Rizzi's proposal at face value and make the addition of ΣP to our geometry of the clausal structure of Spanish, we obtain a structure along the lines of (42).Note that in earlier work (e.g., Laka 1990), the assumption was that CP occurs higher than ΣP (similar to the case of Uriagereka's 1995 FP category).
(42) ForceP ˃ TopicP ˃ FocusP ˃ FinitenessP ˃ ΣP ˃ TP … However, there is evidence that the categories of FocusP and ΣP may be the same category, at least in certain contexts, including TP-ellipsis.If this is true, then TPellipsis may be a context in which the two projections are conflated.Note, however, that the data below may be also be compatible with an alternative analysis wherein ΣP and FocusP are both present in the familiar TP-ellipsis contexts in Spanish, with ΣP being the projection where the particle originates, and FocusP its landing site, given its focal character. 16,17  16 The reader should note that I am concerned here about hierarchical positions, and my goal is to determine whether focused phrases and polarity elements in TP-ellipsis target the same projection, which appears right below the position hosting recomplementation que.Therefore, I am not that concerned about the names of the projections in question. 17There is good reason to suspect that FocusP and FinitenessP are not overtly filled in the same clause, and the question arises as to whether FinitenessP is projected in the presence of FocusP (see the discussion on the impossibility of ellipsis with jussive/optative que in the main text; see also Villa-García 2015 for further discussion).
In what follows, I provide a variety of empirical arguments to the effect that something along these lines is going on in TP-ellipsis with a polarity item.The below shows that even in non-TP-ellipsis cases, the relevant polarity particles exhibit focuslike behavior (with the potential exception of no in regular contexts, as we shall see).
As a matter of fact, Laka (1990: 129-133), a piece of work that was published prior to the advent of Rizzi's cartographic approach, proposed that ΣP is the locus of negative and affirmative particles, as well as the landing site of emphatic fronting, illustrated by examples like (43) (from Laka 1990: 132).( 43 43) correspond to what is generally referred to as focused phrases at present, which under a Rizzian approach occupy Spec, FocusP.Thus, the work of Laka (1990) proposed that polarity particles and focal phrases alike target ΣP.A different possibility is suggested by Batllori andHernanz (2008, 2013) and Hernanz (2007), for whom having ΣP alone is not enough.According to these authors, FocusP and ΣP (in their terms, PolarityP) do co-exist in the same structure.Their work focuses on the positive particle sí in non-elliptical environments.This emphatic affirmation particle is generated in PolarityP, where it satisfies its polarity feature, and then moves to FocusP to satisfy its emphatic (i.e., focal) feature.In the same spirit, Sailor (2011) argues that PolarityP is a polarity phrase in the CP layer and ΣP encodes clause-internal polarity (see also Villa-García and González Rodríguez, in preparation).If this line of analysis is correct, then both projections are needed, at least in non-elliptical contexts.
Let us entertain the hypothesis that ΣP and FocusP are one and the same category in TP-ellipsis (or at least that both categories are present, as suggested in the previous paragraph, but in any case that FocusP is crucially implicated in TP-ellipsis with a polarity particle).If ΣP and FocusP perform similar functions, then it is not surprising that when a polarity particle is present, then a focused constituent cannot co-occur with it (see also Batllori andHernanz 2013 andHernanz 2007 for Catalan and Spanish in cases of emphatic positive polarity particles in contexts that do not involve ellipsis).As is known, in languages like Spanish, only one focused phrase can occur per clause (Rizzi 1997).This would account for why an instance of focus cannot act as the remnant of TP-ellipsis in Spanish, since there cannot be two focal elements (A JOAQUÍN and no) in the same clause, as indicated by the unacceptability of (16b), repeated here as (44). 18  (44) *María dice que A PEDRO traen, pero que A JOAQUÍN no Mary says that DOM Peter bring but that DOM Joaquín not 'Mary says that PETER they will bring, but not Joaquín.' 18 There have been claims that in fact the situation may be more complex.For instance, Bazaco (2014) contends that there can be multiple instances of wh-movement -a kind of focus movement-and Ortega-Santos (2016) argues in favor of multiple focus movement.(Note however that Bazaco's data are poorly understood to date and that multiple focus movement, if available, is only possible under highly specific circumstances; see Ortega-Santos 2016 for much relevant discussion).
The same point is illustrated by the following data set from RAE (2009: 3006).According to this source, this paradigm confirms the focal nature of particles like no 'not' (though see below on the ambiguous behavior of no).
(45) a. Eso no lo digo yo that not cl.say I 'I don't say that.' b.ESO digo yo that say I 'I say so/I believe so.' c. *ESO no digo yo that not say I 'THAT, I don't say.'In (45a), eso precedes no and is concomitant with an agreeing direct object clitic (lo), a hallmark of left dislocation (rather than focalization).Note that the alternative order no  eso is fully ungrammatical (*no eso digo yo).In (45b), which is a positive sentence, ESO is stressed and is not accompanied by an attending clitic, indicating that it is an instance of focus.If we now add no to (45b), the result is ungrammatical, as shown by (45c), which suggests that the presence of a focalized phrase (ESO) before no leads to an ill-formed outcome.
Likewise, the fact that the last element -the polarity particle-in an elliptical sentence is focal is not an unnatural assumption, given that in languages like Spanish, the last element in the sentence can bear main sentence stress and thus function as focus (in general, new information focus; see Zubizarreta 1998, Ortega-Santos 2006, Villa-García 2015: Ch. 5, inter alia). 19Ortega-Santos (2016), for instance, provides a variety of arguments that postverbal subjects that appear last in the sentence and which instantiate (corrective) focus are actually located in FocusP (with remnant movement of the remaining sentence constituents to a left-peripheral position higher than FocusP).
It is important to mention that items such as sí and tampoco are emphatic and thus stressed (e.g., affirmative sí is tonic but conditional si and the interrogative complementizer si are not).The Praat spectrogram furnished in (47) indicates that in the elliptical clause (i.e., the second clause of the sentence in ( 46)), no bears sentence stress, which is symptomatic of a focus constituent (this does not happen in cases of ordinary negation necessarily, which is suggestive of the possibility that no can but need not be focal, unlike sí, which is invariably focal, a matter to which I return immediately).( 46 Dijo que a la fiesta, que sí (que) venía said that to the party that yes that would-come 'She said that she would certainly come to the party.' The apparent optionality, however, may not be so: the distribution patterns of sí and sí que are not identical (see Villa-García and González Rodríguez, in preparation).Note that in cases of ellipsis, que is banned: (ii) *Dijo que a la fiesta, que no iba, pero que al partido, que sí que said that to the party that not went but that to+the match that yes that 'S/he said that to the party, s/he was not going, but s/he also said that s/he was going to the match.' Similarly, sí que sentences do not exhibit strictly focal behavior in that contrary to sí, they can be followed by a topic or a subject (Batllori and Hernanz 2013).The position of sí que cannot be ForceP, as proposed by Battlori and Hernanz (2013) and Hernanz (2007), since it can be preceded by topics and by recomplementation-que topics, as the previous data demonstrate.It is also questionable whether sí que is hosted in FocusP, since it does not manifest full-fledged focal behavior (i.e., no S-V inversion) and as noted, it can be followed by topics.Prima facie, examples like (i) resemble exclamative clauses followed by an instance of que, which occur in many Spanish dialects, exemplified in (iii).

(iii)
Dice que la casa, que qué grande (que) es says that the house that what big that is 'She exclaims that the house is so big.'According to Rizzi (2013), similar examples in Italian should be analyzed as locating the exclamative XP in Spec, FocusP and que in the head position (i.e., Focusº).If this were the case, then in (i) sí would have to sit in Spec, FocusP/ΣP (but see above on the possibility of placing lexical material other than the verb below que in sí que sequences, which does not hold for exclamatives (cf.(iii)); exclamative sentences do show S-V inversion despite the presence of an optional low que below the exclamative phrase).I leave this issue open for future research, but the reader is referred to ongoing work by Villa-García and González-Rodríguez (in preparation) for the prospect that sí que, contrary to the polarity particle sí, actually constitutes a predicate of truth in AssertiveP.
Moreover, if the ΣP of previous accounts of TP-ellipsis is in fact FocusP, then the prediction is that other focused/emphatically fronted phrases should in principle be able to be featured in TP-ellipsis irrespective of whether or not the clause features recomplementation que.This turns out to be correct, as shown by (49a), which includes focused constituents such as de ninguna manera 'in no way' and by (49b), which displays siempre 'always,' from Depiante (2000: 130).As the careful reader will note, both de ninguna manera and siempre also mark polarity (in (49a), de ninguna manera could be replaced by tampoco and in (49b), siempre could be replaced by sí; as noted by the examples, adding tampoco and sí would result in strong ungrammaticality, which is expected, since both items instantiate foci).
(49) a. Me dijeron que si llueve, que no vienen a la fiesta, y que cl. said that if rains that not come to the party and that si nieva, que de ninguna manera / ni de coña (*tampoco) if snows that of no manner / nor of joke neither 'They told me that they will not come to the party if it rains, but that if it snows, under no circumstances/in no way will they come.'b.Ana nunca limpia su oficina, pero su cuarto siempre (*sí) Ana never cleans her office but her room always yes 'Ana never cleans her office, but she always cleans her room.' The reader is also referred to the Galician data in (3c) and to RAE (2009: 3709) for evidence that adverbs such as jamás 'never ever' also license TP-ellipsis in Spanish.However, there seems to be a certain degree of variation; see López (1999) for the view that elements like nunca 'never' cannot license ellipsis, in accordance with the judgments of his consultants. 21 Furthermore, recomplementation que can also be followed by an instance of focus/wh in non-elliptical contexts, as shown again in (50a).The parallelism is apparent: in ellipsis contexts, también occupies the position following recomplementation que, in much the same way as a focused phrase (e.g., a wh-item) occupies the position following recomplementation que in embedded questions (cf.(50b)).
(50) a. Dice que a Mánchester, que cuándo fueron says that to Manchester that when went 'S/he asks when they went to Manchester.' b.Dice que a Mánchester, que también fueron says that to Manchester that also went 'S/he says that they went to Manchester as well.' 21 A related construction that also involves a focused element followed by an elliptical gap is sluicing, TP-ellipsis featuring a wh-item, as in (i) (see also (28a) for English sluicing).Such constructions do not manifest a CLLDed remnant and are analogous to their English counterparts, as suggested by the paraphrase.

(i)
Me parece que contrataron a alguien, pero no sé a quién contrataron cl.seems that hired DOM somebody but not know DOM who 'It seems to me that they hired somebody, but I don't know who they hired.' The reader is referred to Saab (2010) for the differences between TP-ellipsis with a polarity item and sluicing, in particular with reference to the ability of the latter -but not the former-to repair island violations.
Additionally, such focalized elements are incompatible with a polarity particle and cannot be featured in TP-ellipsis contexts (cf.( 51)).On the assumption that focus is limited to one occurrence per clause, as argued above, then the ungrammaticality of ( 51) is explained away, since cuándo ('when') and también compete for the same position and therefore cannot co-occur in the same left periphery. 22  (51) *Dice que a Leeds, que cuándo van, y que a Londres, que says that to Leeds that when go and that to London that cuándo también when too Intended meaning: 'S/he asks when they went to Manchester, and also when they went to London.' In parallel fashion, as shown in Section 3, focused phrases cannot appear inbetween ques (cf.( 22)), because the sandwiched position is reserved for topical phrases.If polarity particles that can function as specifiers (e.g., también/tampoco) are focal, then they should not be able to occur in recomplementation configurations, which can only feature topics/CLLDed phrases.( 52) bears out this prediction.
(52) *Susurró que también, que vienen 23 whispered that too that come 'S/he whispered that they are also coming.'That polarity particles including sí and también are focal is corroborated by the fact that in non-elliptical contexts, when such particles are preverbal and are accompanied by overt material in the TP, they trigger inversion, a hallmark of foci such as whelements and focused phrases in Spanish (this conclusion mirrors that of Batllori and Hernanz 2013 for the positive polarity item sí in Catalan and Spanish, which is shown to display wh-operator properties).This is indicated by the contrasts in (53), where a comparison is drawn between the behavior of the wh-item cuándo 'when' and the familiar particles sí and también.Of course, an XP like the subject can precede these elements (in a position like TopicP) (cf.(53a,b)), but when subjects follow the whitem or the polarity particle, they cannot intervene between the wh-item/polarity element and the verb, which must be adjacent to each other (cf.(53c,d)); hence subjects must be postverbal (cf.(53e,f)) -or occur in a topic-like position, as in (53a,b).
22 A note is in order, since as is well known, items such as sí and no are deemed to be heads, rather than specifiers.This means that claiming that focused phrases and wh-items occupy the same slot as elements like no would be inappropriate.Thus, it would be more accurate to contend that these focuslike elements lexicalize the same projection (i.e., FocusP), which in Spanish can only be overtly realized by either the specifier or the head (which in turn may also account for why recomplementation que cannot be preceded by foci in this variety).Exclamative phrases with a low que do not seem to adhere to this rule, however (see fn. 20). 23A superficially similar sentence that is grammatical involves a longer pause after también (the same happens with sí or nada 'nothing,' as in (i)), with the second clause, introduced by the second instance of que, being an afterthought or a restart.In these cases, nada/también would not be associated with the lower clause, which could itself involve a polarity element.If the analysis currently pursued is correct, then wh-items and polarity elements such as sí should not be able to co-occur, given that both items are focal.This prediction is correct, as shown by the ungrammaticality of (54) (Batllori and Hernanz 2013).
(54) *¿Qué sí compraste?what yes bought 'What DID you buy?' Batllori and Hernanz (2013) concentrate on sí, which is a marker of emphatic affirmation (note that simple positive polarity is expressed by having just the verb and no negative marker).This contrasts with negative sentences, where no has to be present.The three possibilities are provided in (55), which draws on Batllori and Hernanz (2013: 11). 24The sentences in (i) show that the order tampoco/también plus subject is possible (cf.(53d)): (i) a. Tampoco Juan compró los billetes (cf.[tampoco Juan] compró los billetes) neither John bought the tickets 'Neither did John (buy the tickets).'b.También María fue a Roma (cf.[también María] fue a Roma) also Mary went to Rome 'Also Mary went to Rome.'According to RAE (2009: 3686), the polarity item and the subject form a constituent in (i)a and (i)b, i.e., the polarity item has scope over the subject.Thus, in (i)a, for instance, the existence of one or more individuals who did not buy the tickets is assumed.In the alternative arrangement in which the subject precedes the polarity word, as in (ii) (see also (53b)), this interpretation is not forced.In fact, the most natural reading of such sentences is that in which the polarity word modifies the full verbal group.Hence, in (ii), where Juan precedes tampoco, it could be the case that the action mentioned, i.e., not buying the tickets, is just one of a list of things (e.g., errands) John did not do.

(ii)
Juan tampoco compró los billetes John neither bought the tickets 'John didn't buy the tickets either.'Given the analysis suggested by Batllori and Hernanz (2013), according to which polarity items are merged directly in ΣP and then move to FocusP when emphatic, it follows that unmarked polarity cases (i.e., the negative adverb no in (55a) or the phonetically null features of unmarked affirmation (cf.(55b)) do not involve FocusP.The question immediately arises as to what happens with no.In particular, the position of no is not clear, as has been mentioned in passing, since it does behave much like a focal element in triggering inversion and in being incompatible with uncontroversially focused phrases (see the RAE paradigm in ( 45)).In analogous fashion, in TP-ellipsis cases, no bears sentence stress (cf.( 47)), which points out that no can function as focus, at least in this particular environment.Moreover, no is also incompatible with wh-items (note that in (56), an aggressively non-D-linked whquestion has been employed, in order to ensure that we are dealing here with a bona fide question): (56) *¿Qué (demonios) no compraste?what demons not bought 'What (the hell) did you not buy?' Importantly, ( 56) is grammatical if interpreted as a rhetorical or echo question (preferably without demonios), in which case we are not dealing with a question in the strict sense (hence a different structural position may be implicated).The same holds for (54) above, with sí.Along the same lines, Antonio Fábregas (p.c.) notes that cases involving D-linked wh-phrases or items such as por qué ('why') and cómo ('how (come)') are in fact compatible with negation, as shown in (57).These items are likely to occupy a position other than FocusP (see, for instance, Rizzi 2001 on the possibility that the Italian counterpart of por qué is hosted in InterrogativeP).Thus, if no is focus-like, it follows that it cannot appear with bona fide questions like (56) (where the wh-phrase targets FocusP), but it can do so with D-linked and por qué-type questions.The same holds for positive sí.One possibility to explain the behavior of no would be to assume that its incompatibility with foci has to do with the old Inner Island Effect of Ross (1967) (aka the intervention of negation of Rizzi 1990), rather than with negation being an instance of foci itself, hence unable to co-occur with another focused phrase.This hypothesis would be supported by the fact that postverbal foci are not incompatible with negation, as shown in cases where tampoco occurs postverbally and thus no needs to occur in the sentence for it to be grammatical (i.e., negative concord in double negation; see Villa-García 2015: Ch. 4).Whatever the case may be, this does not prevent no from acting as focus in TP-ellipsis cases, as shown above.It may well be the case that certain instances of no are focal, while others are not (see fn. 25).Indeed, González Rodríguez (2008) argues that metalinguistic negation -which denies a previous statement-occupies a focus position.As Raquel González Rodríguez (p.c.) points out, although negation in TP-ellipsis (cf.( 46)) does not deny a prior affirmation, it is indisputable that the clause is bound to the preceding discourse, much like in cases of metalinguistic negation, which can be considered to be a further argument in favor of the proposal that in contexts such as TP-ellipsis, negation can be located in a focus position in the left periphery.
An additional argument for the analysis proposed here comes from Asturian, where both proclitics (i.e., preverbal clitics) and enclitics (i.e., postverbal clitics) are attested.Clitics in this Romance variety generally display enclisis, proclitics being reserved to particular contexts, including sentences involving focus and wh-items, but not other preverbal constituents like subjects or topics (Fernández-Rubiera 2009, Villa-García 2015).This is shown by the examples in (58), where only (58b), which is an example of a constituent question, exhibits proclisis.( 58 Lastly, assimilating ΣP to FocusP would account for the incompatibility of jussive/optative que with focused phrases.Under this account, it follows that (focal) también/tampoco/sí/no cannot occur below jussive/optative que (which heralds FinitenessP), since there is no appropriate focus position for such items below jussive/optative, hence the impossibility of licensing ellipsis in cases involving jussive/optative que, as illustrated in (32), an issue to which I return in Section 5.3. 26,27Overall, the evidence discussed above substantiates the proposal that the aforementioned polarity particles instantiate foci, especially in TP-ellipsis-with-apolarity-element cases.Alternatively, the discussion above may be compatible with the involvement of both ΣP and FocusP in the familiar TP-ellipsis contexts in Spanish (perhaps with the polarity particle being initially merged in ΣP and then moving to FocusP, or with the particle occupying Sailor's CP-related focal PolarityP; see Villa-García and González Rodríguez, in preparation, and references therein, for such an analysis of sí in non-elliptical environments). 28efore concluding this section, I would like to note that focus may be needed for independent reasons in ellipsis in general (e.g., sluicing).Put another way, other types of ellipsis have been independently argued to be intimately associated with the idea of focus.Indeed, there have been proposals underscoring the role of focus in ellipsis.The MaxElide Constraint (e.g., Merchant 2008, Takahashi and Fox 2005, Hartman 2011), for instance, an economy restriction ensuring that the biggest deletable constituent is elided within a given domain, is a case in point.The underlying idea behind MaxElide is to make ellipsis as large as possible: the system starts eliding and stops with the first element that bears focus stress.The upshot of this account is that the system first deaccents, then it elides, with focus phrases not being able to be deaccented, which renders them unelidable.Consequently, that the ellipsis licensor of TP-ellipsis in Spanish is indeed a focal polarity-marking element that overcomes ellipsis resonates with generally-held views on ellipsis phenomena.The reader should note, however, that the above does not imply that focus is all that is needed for ellipsis to occur, a matter than remains the object of intensive research in theoretical linguistics (see Section 5.4).
The data discussed in this paper also pose the non-trivial question of how to technically implement the connection between the CLLDed remnant and the ellipsis licensor.This is the main focus of the next subsection.

On the connection between the topical remnant and the ellipsis-licensing head
in TP-ellipsis At this point, it seems that the left-dislocated remnant XP and the ellipsis licensor Σº/Focº establish some sort of relationship, inasmuch as TP-ellipsis can only occur if both items are present, as shown by the contrasts in (12).To the best of my knowledge, the association between the remnant and Σº is well known, but how to implement their connection in a technical way is still far from settled, if at all necessary.
For simple TP-ellipsis cases such as (6), repeated here as ( 60) for convenience, it could be proposed, à la López (2009), that the putatively intimate relationship between the XP (al gato) and the polarity particle (también) is captured by assuming that they belong in the same projection (for López, the XP is adjoined to Σ').DOM +the dog cl.walked and DOM +the cat too 'They walked the dog, and the cat too.'However, the data brought to light in Villa-García (2010, 2012), which constitute the object of our study, cast doubt on this analysis, since an overtly realized head (i.e., recomplementation que) intervenes between the remnant XP (si nieva) and the polarity particle (tampoco) in examples like (61).
(61) Me dijeron que si llueve, que no vienen a la fiesta, y que cl. said that if rains that not come to the party and that si nieva, que tampoco [vienen a la fiesta] if snows that neither 'They told me that they will not come to the party if it rains or snows.' The abstract structure of these sentences under the account currently pursued is outlined in (62), which shows that a lexical item (i.e., que) intervenes between the XP and the polarity element.Furthermore, there are other pieces of evidence not involving recomplementation that also show that the remnant XP and the polarity particle need not be adjacent to each other (cf.( 63), based on Villa-García 2015).In (63), the XP con la suegra is in a higher clause than the ellipsis licensor tampoco, and yet TP-ellipsis takes place without difficulty.
(63) Dice que con la prima del marido no va a ir, says that with the cousin of+the husband not goes to go y con la suegra, me parece a mí que tampoco and with the mother-in-law cl.seems to me that neither 'S/he says that s/he's not going with his/her husband's cousin, and it seems to me that s/he is not going with his/her mother-in-law either.' The question that arises then is which relationship holds between the leftdislocated remnant XP and the positive/negative polarity element licensing ellipsis in Spanish TP-ellipsis.Recall that both items must survive ellipsis.An anonymous conference abstract reviewer notes that a technical way of capturing this relationship may be Aelbrecht's (2010) Agree-based ellipsis analysis, according to which the head selecting the ellipsis site is checked against the head licensing ellipsis in order for ellipsis to occur.Taking this approach broadly, it could be argued that Topicº (be it overt or null) establishes some sort of local head-to-head relationship with the head of FocusP or ΣP, thus: This reviewer then goes on to point out that such an approach can offer an account of the impossibility of jussive/optative cases in TP-ellipsis with overt jussive/optative que (remember that these sentences stand in stark contrast to cases in the indicative involving recomplementation que).The reviewer additionally poses the interesting question of why sin estudiar and también are not sufficient to license TP-ellipsis in cases akin to ( 65 and without study too 'S/he or I order(s) that s/he give up if s/he doesn't attend lectures, and also if s/he doesn't study.'Although this reviewer's proposal seems feasible, it does not provide a real insight, nor does it account for long-distance cases like (63), which raise the question of why jussive/optative que disrupts the relationship between the topic remnant and the ellipsis-licensing head (cf.( 65)), but not heads like the verb or the higher complementizer que as part of the material intervening between the two elements (cf.( 63)).
At this point, the technical implementation of the connection between the two categories crucially involved in TP-ellipsis in Spanish is not clear.However, it is certain that on the one hand, contrastive topics are unelidable and, on the other, focus generally survives ellipsis (cf. the MaxElide Constraint).Perhaps the two items survive ellipsis in TP-ellipsis contexts as they are needed for independent reasons (i.e., contrast).Their relationship (i.e., their co-occurrence) may therefore be somewhat accidental.Future research will determine whether this is correct.I now turn to a final note on the recurrent issue of jussive/optative que in the next section.

On the impossibility of having jussive/optative que overtly in TP-ellipsis cases
Throughout the paper, I have made passing reference to jussive/optative que and its inability to be overtly realized in TP-ellipsis cases involving jussive/optative clauses.Relevant examples are provided again in (68) (see also (67)  (iii) There is no appropriate focus position able to host the focal/polarity particle below jussive/optative que, which occupies the lowest head in the split CP, that is, Finitenessº.(iv) In keeping with Aelbrecht's Agree-based account, jussive/optative que intervenes and hence disrupts the relationship between the Topic head and the ellipsis licensing head, which need to establish some sort of head-to-head agreement relationship.
Set against this background, I advance yet another explanation which follows naturally from the analysis advocated in this paper.In light of the preceding discussion (see, especially, Section 5.1), which has resulted in the claim that the element licensing ellipsis is focal in nature and is located in a projection situated below TopicP and above FinitenessP, that is, FocusP, the impossibility of having overt jussive/optative que is accounted for straightforwardly, without further stipulation: jussive/optative que cannot surface simply because it is part of the elided material, as shown in ( 69).This immediately accounts for why (68a) is ungrammatical: que is not a focal element and in fact is structurally below the focal phrase; as a consequence, it cannot remain overtly expressed in ellipsis cases.This analysis also captures with ease why (68b) is grammatical: que is below the focal element charged with the licensing of ellipsis; hence, jussive/optative que is part of the ellipsis site, that is, que is deleted, which explains why it cannot surface. 29e therefore add a fifth hypothesis for the obligatory absence of jussive/optative que in TP-ellipsis: Jussive/optative que is below FocusP, which contains the ellipsis licensor, and therefore que is elided as part of TP-ellipsis.On this view, ellipsis cases where FinitenessP is elided would more accurately constitute instantiations of "FinitenessP-ellipsis," rather than TP-ellipsis.
Further investigations will determine whether this hypothesis (cf.(v)), which follows from the analysis pursued in the paper, holds; it would also be worth looking at its implications for the theory of ellipsis more generally (see the next section).If the answer to this question turns out to be positive, then we will have an additional argument for the proposal put forward in this paper.Be that as it may, it is my hope that the findings reported here will take us a step closer towards finding the ultimate account of the impossibility of having an overt instance of jussive/optative que in elliptical cases like ( 65) and (68a).

A brief note on TP-ellipsis in a phase-based approach
The analysis of ellipsis in Chomsky's generative paradigm remains to a large extent shrouded in mystery.However, there have been serious attempts to account for this phenomenon, pervasive in natural language.For instance, in the context of phasebased theory, some authors have recently advocated that ellipsis is phase-constrained: only phases and complements of phasal heads can in principle undergo ellipsis (Bošković 2014) (see also Gallego 2009 for the hypothesis that ellipsis can only target the complement domain of a phasal head).Under this approach, we have two potential scenarios for the data at issue.First, FinitenessP, which some authors have independently argued is a phase (e.g., Fernández-Rubiera 2009), would be elided (e.g., (68b)).In (indicative) recomplementation cases where FinitenessP is seemingly not lexicalized (e.g., (1)), if we assume that this projection is absent, then it would be the category TP that would be elided (see Gallego and Uriagereka 2007a,b for the claim that V-to-T movement in Spanish actually makes TP a phase).If it is complements of phases that undergo ellipsis instead, then we might assume, with Wurmbrand (2017), that FocusP is a phase in the case at hand and that its complement constitutes a spell-out domain that can remain unpronounced.Intuitively, constraining ellipsis by means of phases makes sense under current assumptions, since the elided chunk is regarded as a spell-out domain that is not realized in PF (Wurmbrand 2017).Based on a number of linguistic varieties, Sailor (2011) raises the possibility that categories such as CP, TP, and DP (all of which are in fact standardly regarded as phases in Spanish) each contain both a projection of polarity and a projection of focus, a possibility fully consistent with the findings reported here.The actual mechanisms underlying ellipsis are certainly amongst the major questions that future research should care to address, for tackling this issue is far beyond the goal of this paper.

Conclusion
This paper has investigated multiple-que sentences featuring TP-ellipsis with a polarity particle in Spanish.I first explored the properties of ordinary TP-ellipsis in Spanish and then went on to provide arguments in favor of a TopicP account of recomplementation configurations in Spanish, based on Villa-García (2012, 2015).
I reviewed the Spec-Head agreement analysis of ellipsis in light of the recomplementation data at issue and showed that this proposal falls short of capturing a crucial property of TP-ellipsis in Spanish: the obligatoriness of a focal/polarity particle such as también ('too').I then proposed to adopt a version of the standard ΣP account of TP-ellipsis in Spanish, which can easily be extended to account for the cases involving TP-ellipsis and recomplementation, while circumventing the problems raised by the Spec-Head agreement analysis.Based on a number of empirical arguments, I explored the nature of the polarity elements critically involved in the licensing of TP-ellipsis in Spanish and concluded that such elements are focal in nature, for they exhibit the host of properties traditionally linked to wh-operators and focal elements.Thus, I put forward the claim that TP-ellipsis in Spanish necessitates a focal element that licenses ellipsis of its complement.The evidence adduced in this paper therefore militates in favor of the conflation of ΣP and FocusP in the context of TP-ellipsis, or at least in favor of the involvement of (a polarityoriented) FocusP in this process (PolarityP in Sailor's terms; note that postulating a left-peripheral focal polarity phrase is not incompatible in principle with ΣP cooccurring in the same clause).
TP-ellipsis of the type discussed here involves both a contrastive topic remnant along with the focal element responsible for ellipsis.The former can be followed by recomplementation que.Since the remnant must be topical, it is then not surprising that the sequence XP-que can precede the ellipsis licensor, which confirms the TopicP analysis of recomplementation patterns argued for in Villa-García (2015).I subsequently moved on to the potential relationship between the topical remnant and the focal element, a question left somewhat open but which cannot ignore two major facts: first, the remnant is contrastive and topical, and contrastive topics are known to be unelidable on independent grounds; second, focus may be needed in ellipsis as a general phenomenon (e.g., the MaxElide Constraint), but this should not be taken to mean that focus by itself triggers ellipsis, which would incorrectly predict foci to obligatorily license.As a result, both contrastive topics and foci may be required independently of each other, which dispenses with the need to establish a formal relationship between the two.In any event, such a relationship could potentially be captured by adopting (a version of) Aelbrecht's Agree-based theory of ellipsis, wherein there is a head-to-head agreement relationship between Topicº and Focusº, although such an account is also faced with certain shortcomings.
In addition to cases involving recomplementation que (Topicº), I discussed recalcitrant cases featuring jussive/optative que (Finitenessº).A crucial difference between the two low complementizers for purposes of this paper is that whereas the former can survive in TP-ellipsis environments, the latter cannot.In order to address this contrast, I offered a non-stipulative account of the inability of jussive/optative que to surface in TP-ellipsis which follows naturally from the focal character of the ellipsis-licensing element: jussive/optative que must be elided simply because FinitenessP must be part of the ellipsis site, due to the fact that a higher head, namely Focusº, is in charge of licensing ellipsis (i.e., ForceP  TopicP  FocusP  FinitenessP  …).
Some ancillary conclusions of this investigation include the status of negative and positive polarity particles, which seem to behave like foci/wh-items, at least in the context of interest, the connection between ellipsis and phases, and the geometry of the left edge of the Spanish clause.Thus, the results reported here have varied consequences for the analysis of TP-ellipsis with a polarity particle, the role of focus, polarity, and, more generally, the architecture of the CP field in Spanish and beyond.
Time will tell whether the conclusions reached in this paper are substantiated by further studies.In the meantime, I hope to have contributed to the analysis of multiple-complementizer sentences alongside the debate over the mysterious yet fascinating phenomenon of ellipsis in natural language.
yes 'They didn't invite Paula, but they did invite Marina.'

( 14 )
Yo doy clases en la uni y ella/María curra en el I give classes in the uni and she Mary works in the Tesco Tesco 'I teach at the university and she/Mary works for Tesco.' Example (14) stands in marked contrast with (15a,b, and c), where one of the contrastive elements is not overtly expressed in the clause, resulting in inappropriate outcomes.(15) a. #Doy clases en la uni y ella/María curra en el Tesco give classes in the uni and she/Mary works in the Tesco b. #Yo doy clases en la uni y curra en el Tesco I give classes in the uni and works in the Tesco c. #Doy clases en la uni

( 47 )
Praat spectogram of the second clause of (46) …pero al resto noIf particles such as sí are focal, then it follows that they should be able to be preceded by topical material, since topics routinely precede foci.Therefore, a Lucía, que in (48) precedes the focal element sí (see also Battlori and Hernanz 2013) (see below for the same distribution with the remaining polarity particles).20   (48) Gritó que a Lucía, que sí le cobraron en el tren shouted that DAT Lucía that yes cl.charged in the train 'S/he shouted that Lucía was in fact charged to travel on the train.'

Table 1 .
Distribution of polarity particles contingent on the polarity of the antecedent clause también and tampoco form a natural class different from that of sí and no in that they only connect clauses with the same polarity values.As noted by /he ordered by shouting that with that attitude, if they continue reducing staff's pay, their children should work.' b.Le dije que cómo que que me pirara cl. said that how that that cl.go-away Subj.
'I asked him/her how come s/he dared tell me to go away.'[Spontaneousspeech, Madrid, Spain, September 2014] They told me that Albert went to the party but I don't think so.' b. *… pero no pienso [ CP [ C' que Alberto fue a la fiesta]] . … y que [ TopicP si nieva [ Topº que no vienen a la fiesta]] tampoco c.Dijeron que Isa, que ayuda a todos, y They said that Liz helps everyone and Peter told me Mary does too.' d. … que [ TopicP María [ Topº que ayuda a todos]] también Overall, these examples appear to this reviewer to constitute a counterargument to the claim that the complement of Cº cannot undergo ellipsis.These patterns certainly merit further investigation, a matter that future research should care to address.b is licit in such sentences (cf.(i)a), but not the emphatic polarity element sí (cf.(i)b), could be that sí targets FocusP (not available since FocusP precedes FinitenessP).However, negation can stay in PolarityP/ΣP, below FinitenessP.Still, it is of note that negation does exhibit focal behavior in certain respects, e.g., it needs to be adjacent to the verb.If wh-items such as qué and particles like nun/sí/tamién/tampocu share focal properties and thus display similar distribution patterns, we should then expect Asturian examples involving any of the familiar polarity elements to manifest proclitics -the word order routinely triggered by foci.This prediction is fulfilled, as demonstrated by the data in (59), since the syntactic distribution of polarity particles parallels that of unambiguously focal elements. negation ), with jussive/optative que.Here, according to the reviewer, the problem is that jussive/optative que intervenes between the null Topic head and the ellipsis licensor, making ellipsis impossible.This is shown schematically in (66).(66).. que[ TopicP XP [ Topicº Ø [ ΣP [ FintenessP [ Finº que [ P [ º sí/no…[ TP ]]]]]]]]This problem does not arise in the jussive/optative que-less counterpart of (65), furnished in (67), since jussive/optative que does not (overtly) intervene between Topicº and the ellipsis licensor.
in the previous section).As shown in (68b), jussive/optative sentences are not semantically incompatible with ellipsis, since the version without que is grammatical.+the dog that cl.wash Subj.and DOM +the cat too 'S/he or I demand(s) that they wash the dog, and the cat too.'So far, we have different potential accounts of why jussive/optative que cannot survive TP-ellipsis in such cases, namely: DOM