3.1 The construction sí Ø

The following sentences further indicate that que does not need to occur (at least overtly) below sí in sí-que cases in Spanish varieties spoken in (at least) Colombia, Mexico, and Peru (see also the data in Campos 1992: 332).⁹ Recall that the meaning of sí in (18) is conveyed via sí que in Iberian Spanish and that the absence of que in the relevant Latin American varieties is optional; thus, the complementizer could be introduced in (18) without giving rise to any syntactic or semantic change.

That sí in (18) is an instantiation of the sí-que construction (and not a polarity particle) is straightforwardly corroborated by the fact that the polarity of the proposition below sí Ø is negative in spite of the presence of the word sí, much like in Iberian Spanish cases like (8)a above. In fact, there is independent evidence corroborating the negative polarity of the sentences in question. As observed by Brucart (1999), in cases of TP ellipsis with a polarity particle, if the first conjunct is negative and the second conjunct is also negative, tampoco ‘neither’ must be used in the second, elliptical clause (and, conversely, también ‘also/too’ must be used if the polarity of both conjuncts is positive). Now, if we add to the sentences in (18) another clause including the polarity element también ‘too’ or tampoco ‘(n)either,’ the negative element tampoco must be used, as expected if the examples in (18) are indeed negative, despite the presence of sí. This is shown in (19). The judgments reported for these examples do not vary if we introduce que between sí and no.

The possibility of having the sequence sí no in Colombian and Mexican Spanish is further supported by data from the CREA (Corpus de Referencia del Español Actual) corpus, which contains examples such as the ones in (20).

In the next subsection we explore the issue of the distinction between sí Ø (sí que) and polarity sí in Colombian, Mexican, and Peruvian Spanish.¹⁰

3.2 Sí vs. sí Ø in Latin American varieties

Having mainly looked at examples involving negation, several issues arise in light of Latin American Spanish examples like (21) (pronounced with no intonational break after sí), where the polarity of the clause is positive:

As noted by Villa-García & González Rodríguez (2020), an anticipated problem when studying sí/sí que contrasts in Iberian Spanish is that, due to their seeming interchangeability, traditional grammars have usually treated sí/sí que as variants of one and the same underlying phenomenon, which is typically represented by the use of brackets around que (cf. sí (que)) in traditional descriptions. The Latin American Spanish case exemplified in (21) further exacerbates this problem, since the polarity of the proposition embedded by sí Ø can be positive and thus, sí Ø can easily be taken to instantiate polarity-encoding sí instead. In (21), the presence of the preverbal subject after sí allows us to conclude that sí is not a polarity marker, that is, that sí Ø does not instantiate sí, notwithstanding their similarity in outward appearances. As shown in Section 2, sí, unlike sí que, is a focalized constituent and therefore, requires adjacency with the verb (see below for examples involving the polarity-encoding particle sí, not sí Ø, in said Latin American varieties). As an anonymous reviewer notes, this proposal appears to make a prediction regarding the interpretation of (21): if the subject were postverbal, sí would in principle mark positive polarity, not the speaker’s firm belief that the propositional content is true. Our informants do note a semantic difference depending on whether the subject is preverbal or postverbal. However, they cannot explain clearly what this difference is, which can be due to the fact that distinguishing between positive polarity and the commitment of the speaker is subtle, especially in affirmative contexts. Similarly, the difference may well be motivated by the different information-structural statuses of preverbal and postverbal subjects.¹¹

The problem of determining whether we have a polarity marker (sí) or not (sí Ø) when the complementizer is not present in varieties of Colombian, Mexican, and Peruvian Spanish is even worse in a sentence without a preverbal subject intervening between sí and the verb, such as (22)a. Anticipating the upcoming theoretical discussion and based on the analysis pursued for Iberian Spanish in Section 2, in principle two underlying structures should exist, with two different meanings, as shown by the English translations (though evidently not easy to tell apart). The two underlying structures are given in (22)b and (22)c.¹²

A potential question to raise at this juncture is whether Latin American Spanish varieties where que can be omitted in the contexts above actually altogether lack the sí que/Ø construction and only employ the focal affirmative marker sí, in such a way that these varieties have (22)b but not (22)c. However, this possibility cannot be right. After all, several examples furnished above involve an instance of negation below sí and are interpreted as bearing negative polarity, which refutes the claim that sí in such cases encodes polarity. Likewise, que can optionally surface in these varieties, which provides direct evidence for the existence of sí-que configurations. Moreover, as will be shown below, in some syntactic environments, que is obligatory even in the relevant dialects spoken in Colombia, Mexico, and Peru.

Conversely, then, could it be that these Latin American varieties use only sí Ø (cf. (22)c), instead of the two constructions in (22)b,c? The answer to this question turns out to be negative as well. First of all, the relevant varieties can also have examples akin to (12), where sí que co-occurs with the emphatic positive polarity marker, as in (23), suggesting that we are dealing with the familiar sí-que construction followed by sí. (We return to these cases in the following section.)

Furthermore, Villa-García & González Rodríguez (2020) capitalize on cases involving ellipsis with a polarity particle (Brucart 1999), which, as the name indicates, feature a polarity element as the licensor of ellipsis. These cases are illustrated in (24).

Cases like these are attested throughout the Spanish-speaking world, as shown by the examples provided by RAE-ASALE (2009: 3006–3007), and clearly involve a contrasting polarity marker in the second clause. This is further substantiated by the fact that sí que, which in principle covaries with sí Ø in some Latin American Spanish varieties, cannot occur in cases like the following, which have been analyzed as a context in which the verb is elided (Brucart 1999; Depiante 2004; Villa-García & González Rodríguez 2020):

In (25)b, if sí were always the que-less variant of sí que, we would in principle expect que to be able to be pronounced, contrary to fact. What is more, if such ellipsis cases must involve polarity particles, as is standardly assumed in the literature, we would expect the bona fide focal affirmative polarity marker to be instantiated in (25)a, rather than sí que/Ø, which does not serve the function of indicating the polarity of the clause. This is the reason why (25)a is acceptable but not (25)b: only in the first case do we have a genuine polarity marker, as required by this type of ellipsis. As shown in (25)c, the combination sí que sí is rejected in these contexts. We would like to suggest that the ungrammaticality of (25)c is due to the parallelism requirement on ellipsis (i.e., we would expect coordination at the clausal level with sí que), although the data merit further attention. Note in passing that cases like (23) above (Ahora sí que sí ‘Now it is truly the case that X’) are also interpreted as elliptical clauses, which is exactly what would be expected, given that the polarity particle sí (the ellipsis licensor) occurs below sí que (and, as shown below, examples with other ellipsis-licensing polarity particles, namely no ‘not,’ también ‘also/too,’ and tampoco ‘neither,’ are also documented).

Lastly, it is important to note that sí Ø configurations do not require the inversion that one would expect in the case of a bona fide focal sí. This is indicated by (21), repeated here again in (26), which exhibits a preverbal subject intervening between sí Ø and the verb.

However, Latin American Spanish varieties such as Argentinean, Colombian, Ecuadorian, Mexican, and Peruvian Spanish do have focal sí requiring adjacency with the verb. By way of illustration, speakers report obligatory inversion in cases like (27), which again involve a contrast in polarity (no ‘not’ vs. sí ‘yes’).

This strongly indicates that sí in such cases is an emphatic affirmative polarity element, which given its focal character, requires subject-verb inversion. Overall, the evidence adduced in the preceding paragraphs confirms that both sí que/Ø and sí exist in Colombian, Mexican, and Peruvian Spanish, although the difficulty in distinguishing the two of them in certain contexts is undeniable. We now turn to the analysis of sí Ø.

4.1 Ruling out an alternative analysis

Prima facie, the by-now familiar sí-no combination in Colombian, Mexican, and Peruvian Spanish would in principle be amenable to an analysis whereby sí occupies FocusP (presumably it would have to be base-generated there) and no is located in ΣP, by virtue of the fact that it marks the negative polarity of the clause. This possibility is schematized in (29):

On this view, then, the two items (sí and no) may be able to co-occur in the same clause. This would assume that sí does not signal a commitment on the part of the speaker to the truth of the proposition —perhaps it is just a focal, emphatic affirmative polarity marker, contra the judgments of the speakers consulted. Beyond meaning, this alternative account of the sí-no sequence is also weakened on syntactic grounds by the fact that sí can be followed by an unambiguous adverb stressing the truth value of the proposition that follows it in the relevant Latin American varieties, as in (30).¹⁵

Similarly, this piece of data crucially indicates that sí in this context is indeed not a focal marker, since it does not require adjacency with the verb (see the preceding section).

Moreover, under (29), we would expect to find no-sí sequences if the facts reduced to a matter of focal polarity particle + polarity particle. However, the CREA corpus does not contain any instance of the no-sí combination in any dialect of Spanish. The corpus searches, coupled with the introspective judgments of our native-speaker consultants, suggest that no-sí sequences are fully ungrammatical, as (31) and (32) demonstrate. Note that speakers find these sentences not only syntactically ill-formed, but they actually fail to assign them a plausible interpretation.

This is scarcely surprising if sí in the relevant Latin American Spanish examples does not really mark polarity, in much the same way as its sí-que homolog in Iberian Spanish, which is our claim here, contra (29). No in Spanish, for its part, can only mark negative polarity, and thus the presence of sí creates a contradiction. More generally, the ungrammaticality of (31) and (32) is owing to syntactic and semantic reasons: on the one hand, sí and no compete for the same base-generated position, since both are polarity particles initially merged directly in ΣP (Hernanz 2007, inter alia); on the other hand, the two polarity particles cannot co-occur because a clause cannot bear contradictory polarity values simultaneously. Thus, the data above rule out an analysis along the lines of (29) wherein both sí and no in (31) and (32) are polarity markers. The alternative order, sí no, is possible in the relevant varieties because the high element sí is not really a polarity marker, but the sí of sí que in XP, whilst no is a polarity marker located in ΣP. Thus, we propose that the sentences in which sí is followed by no (cf. (18), (20), and (30)) are null-que variants of the sí-que-no sequence. Consequently, the analysis proposed for sí-que cases can be maintained for their que-less counterparts, as in (33) (see also (28)).¹⁶

As a reviewer kindly points out, it is important to consider the possibility of having null-que variants in other environments. In this regard, Martínez Vera (2019) argues that Latin American Spanish varieties lack overt que in the double-que (i.e., recomplementation) construction found in Iberian Spanish, as investigated by Villa-García (2015). As Michelle Sheehan (pers. comm.) observes, perhaps Latin American varieties are more permissive in terms of the omission of que as a C head more generally. Thus, the phenomenon discussed here seems to be an instantiation of a more general case (see also fn. 5 for much relevant discussion). In essence, the omission of que in sí-que environments is not an isolated case in the left periphery of Spanish; other Cºs can be overt and null as well (Villa-García in preparation).

4.2 Back to sí que in the relevant Latin American varieties: Arguments

Given that the sí que – sí Ø alternation seems to be in free variation in varieties such as Colombian and Mexican Spanish (cf. (15)–(17)), a natural way of analyzing the relevant cases is to assume, as we are doing here, that que can remain silent or, put differently, that the head of XP —under the analysis currently pursued— does not have to be spelled out under normal circumstances. Since the semantic contribution and the syntactic behavior of sí Ø in such varieties mirror those of sí que in Iberian Spanish, it follows that the two elements should be analyzed similarly, precisely as assumed by our proposed account in (33). In the following paragraphs, we provide support for the claim that the absence of que reduces to a case of null PF spellout.

Indeed, that the overtness or non-overtness of que in the sí-que sequence has to do with PF realization of the complementizer is supported by the following facts. First, the optional presence of que has no semantic effect in the Latin American varieties at issue, which is wholly compatible with its realization being reducible to PF considerations.¹⁷ Additionally, remember that in Iberian Spanish, sí que can in principle be followed by a proposition that itself contains an emphatic positive polarity marker, thus making the sequence sí que sí legitimate. If some varieties of Latin American Spanish permit the omission of que, as we have seen, then the prediction is that such varieties should in theory allow sí-sí sequences. The reason is that the first instance of sí would be in Spec,XP and the second instance would be the focal positive polarity marker located in FocusP, making the two items compatible in principle. However, the CREA corpus does not contain any examples of sí sí in Colombian Spanish, and the same result holds for Mexican Spanish. This fact leads us to suggest that the prediction that the combination sí+sí should in theory be possible is not borne out, as shown by (34). This result resonates with the judgments of our native-speaker consultants from different varieties, who categorically reject such sentences, regardless of whether or not ellipsis is effected.

This may seem like a counterexample to the analysis proposed here, but in fact it constitutes strong evidence in its favor: in such contexts, que has to surface obligatorily for speakers of the aforementioned varieties, who find que non-compulsory (and in fact tend to omit it) in other contexts. Thus, the grammatical counterpart of (34) is identical to the Iberian Spanish version, provided in (35).

The fact that que needs to be overt in sí-sí contexts but not in sí-no contexts (cf. (18)) confirms that we are dealing here with an instance of the sí-que construction.¹⁸ One natural explanation for the contrast between sí no and *sí sí is to assume that whereas que can remain phonologically unrealized in the first case, it cannot in the second, possibly due to a constraint against adjacent homophonous forms (i.e., haplology). This is in keeping with the judgments of our consultants, who intuitively deem sí-sí sentences overly cacophonic. As a result, que must be PF-realized in order for the illicit sí-sí sequence to be disrupted (i.e., sí que sí).¹⁹²⁰

Lastly, that sí-sí sequences are ruled out on phonological grounds and not because sí que is incompatible with a focal polarity particle immediately below it is corroborated by the example in (36), which features the focal polarity particle tampoco ‘neither’ after sí que. Put another way, while the presence of the emphatic affirmative polarity sí requires que to surface to avoid contiguous sí sí forms, other focal polarity particles causing no haplology issue do not trigger the overt realization of the complementizer in between items:

The existence of examples like (36) further strengthens the parallelism with Iberian Spanish sí-que patterns, which can also be followed by emphatic polarity-encoding elements like tampoco and también (see Villa-García & González Rodríguez 2020, for empirical evidence; see also (13)a).²¹

The facts just reviewed lend plausibility to an account wherein que is in principle optional in the sí-que sequence in most contexts in Latin American varieties such as Colombian and Mexican Spanish, thus reducing its presence or absence to a PF matter.²² Further, a welcome advantage of this analysis is that it allows for a unified account of sí que and sí Ø across dialects.

4.3 Excluding a lexical account (sí que vs. sí Ø)

In this subsection, we would like to briefly consider an alternative account to derive the sí que case. This competing analysis is outlined in (37):

Under the proposal in (37), que and Ø are distinct lexical items, rather than a single lexical item with two different spellouts, namely an overt spellout (i.e., que) and a null spellout (i.e, que), as assumed under our account (cf. (33)).

Nonetheless, arguments such as the following seem to favor the null-que analysis in (33) over the two-different-lexical-items analysis in (37): in Colombian, Mexican, and Peruvian Spanish, que in sí que always seems to be an option, in that we have found no cases involving sí Ø as a way of signaling the speaker’s commitment to the truth of the proposition where que would have to remain absent. In other words, it seems that que is always pronounceable. Moreover, the analysis in (37) would be at variance with the facts reviewed above where on occasion, que must be spelled out owing to PF reasons (cf. (34)/(35)). In this connection, in a model wherein syntactic structure is sent to PF and to L(ogical)F(orm), and PF checks whether the structure adheres to PF requirements such as the avoidance of contiguous homophonous forms (e.g., sí sí), it is not at all clear how the head Ø could end up being spelled out as que in such cases (since under this account, que and Ø are different lexical items, not just the result of two different spellouts of the same item). On this view, we would have to stipulate that que, as a maximally underspecified left-peripheral form (Villa-García 2015), is an expletive inserted in PF, much like contentless/expletive no ‘not’ in (i) in fn. 19.²³ Although we will defer further discussion of this alternative account for future work due to space limitations, it appears that the que alternative provides a more straightforward account of the sí-que-sí facts (as well as a unitary treatment of the facts across dialects), and in actuality the account in (33) fares rather well with the account of micro-variation to be pursued in the following subsection. More generally, our findings provide additional cross-dialectal support for the analysis of sí/sí que proposed in Villa-García & González Rodríguez (2020), which in turn is based in part on previous accounts in the literature (e.g., Hernanz 2007; Batllori & Hernanz 2013).

4.4 Toward a (micro-)parameterization of the dialectal contrast

The approach adopted here (cf. (33)) coheres well with the claim made by Rizzi (2013: 201–202) and Rizzi & Bocci (2017: 13) that in some cases, the difference between overt and null C/left-peripheral heads (i.e., que/que) can be attributed to a superficial spell-out (i.e., pronunciation) parameter, which we would like to recast also as a micro-parameter accounting for intra-linguistic differences (see, e.g., Kandybowicz 2009; Chomsky 2010; Richards 2010; Berwick & Chomsky 2011; and Boeckx 2012, for an approach that treats linguistic variation as a matter of externalization). In the words of Rizzi (2013: 201–202), this is “a familiar and widely attested kind of low-level parameterization.” This (micro-)parameter would thus lie at the heart of the observed dialectal split between Iberian Spanish sí que on the one hand, where que is overt, and Latin American sí que/que, on the other, where que can receive a silent spellout, hence accounting for its optional character. On this view, the presence or absence of que in sí que would be a matter of PF: whereas in Iberian Spanish, que requires an overt spellout, in some Latin American varieties at least, it can receive a silent/null spellout in most contexts, a view for which we have provided empirical support in the preceding subsections. This approach to micro-variation is indeed in keeping with Barbiers (2009) contention that variation more generally is restricted to lexical properties and PF spellout (see also the abovementioned references). Further research will determine whether this approach is extendable to other left-peripheral heads/Cºs.

As an alternative account of the observed microvariation, Adger & Smith (2005 et seq.) advocate the view that the feature system assumed in much work in Minimalism in the last three decades allows for variable phonetic (i.e., audible) outputs with the same interpretation (in our case, que and its null counterpart, que), by virtue of a particular head containing one additional formal feature, with visible phonetic consequences but not semantic ones. For reasons of space, we will not explore this alternative possibility here. Further research will draw a comparison between competing accounts of microvariation in sí que/que cases.