Focalization in situ and the absence of covert focus movement in Brazilian Portuguese

3.1 The absence of “low focus movement”

In analyzing postverbal subjects in Italian, Belletti (2001, 2004) proposes that the periphery of vP includes topic- and focus-designated projections, similarly to Rizzi’s (1997) analysis of the left periphery of the sentence, as represented in (21). When a postverbal subject in Italian has focus interpretation, as in (22B), Belletti (2001, 2004) proposes that it is licensed in the specifier of a focus projection in the periphery of vP, as represented in (23).

Several works within the cartographic tradition have assumed the existence of topic and focus projections in the periphery of vP in Brazilian Portuguese (see e.g. Mioto 2003; Quarezemin 2009; Armelin 2011; Kato 2013, 2020; Quarezemin & Tescari Neto 2015; Cépeda & Cyrino 2020), especially in analyzing ditransitive and wh-in situ constructions.^[3] For instance, Cépeda & Cyrino (2020) propose that sentence (24B) has the structure in (25), while Kato (2013) proposes that sentence (26a) has the structure in (26b).

Contrary to the above works, Lacerda (2024a) presents a number of arguments in favor of analyzing postverbal focalization in Brazilian Portuguese as being truly in situ in overt syntax, without involving “low focus movement” to the periphery of vP, thus not being subject to the Focus Criterion in overt syntax. I will present here one argument regarding ditransitive constructions and one argument regarding wh-in situ and will refer the reader to the work cited for a much lengthier discussion.

Regarding ditransitive constructions in Brazilian Portuguese, Quarezemin (2009) notes that focalization of the indirect object in the non-canonical IO_F–DO order is marginal when the direct object is indefinite, as in (27B). Lacerda (2024a) notes, however, that focalization of the indirect object in the IO_F–DO order is perfectly acceptable when the direct object is definite, as in (28B).

Lacerda (2024a) takes the contrast between (27B) and (28B) to indicate that the non-canonical IO_F–DO order (with the indirect object focalized) in Brazilian Portuguese is not derived by leftward low focus movement of the indirect object past the direct object, which should in fact be immaterial to the definiteness of the direct object, but rather by right dislocation of the direct object, which is a kind of topicalization of discourse-given material. The author shows, based on a number of examples, that direct objects in the IO_F–DO order exhibit the same semantic restrictions (related to definiteness and referentiality) that subjects do in subject right dislocation. As is illustrated in the contrast between (29) and (30), right dislocation of subjects in Brazilian Portuguese is acceptable with a definite subject, but marginal with an indefinite one (and completely ruled out with quantified subjects; see Lacerda 2024a).

Regarding wh-questions, Lacerda (2024a) notes a contrast between wh-fronting and wh-in situ. In constructions with wh-fronting, both the pied-piping structure in (31a) and the sub-extraction structure in (31b) are acceptable; possible structures for (31a) and (31b) are given in (32a) and (32b), respectively.

When the relevant wh-expression appears in the postverbal area, on the other hand, the situation is different. While sentence (33a) is acceptable, sentence (33b) is ruled out. If (33a) had a structure along the lines of (34a), involving low focus movement, sentence (33b) should also be acceptable, contrary to fact, given the availability of the structure in (34b), which would be the low counterpart of the grammatical (32b). The gap in the paradigm is thus left unexplained under the assumption of low focus movement.

Based on the above, Lacerda (2024a) concludes that there is no low focus movement to the periphery of vP in Brazilian Portuguese, contra the literature following Belletti (2001, 2004), and that postverbal focalization in this language is realized truly in situ in overt syntax (see the work cited for additional arguments and further discussion).

3.2 “Low focus movement” and the presupposition problem

In this section, I will extend Lacerda’s (2024a) argumentation against low focus movement in Brazilian Portuguese. I will argue that this kind of structure fails to establish a proper focus-presupposition articulation of the clause in the postverbal area. In particular, I will argue that a focus (putatively) sitting at the edge of vP, much like a truly in situ one, does not partake in a bipartite focus-presupposition articulation of the clause, as would be mandated by the Focus Criterion (cf. (9) above).

As was discussed in the previous section, the cartographic approach assumes that both left-peripheral and postverbal focalization involve the creation of a focus-presupposition bipartition of the clause, by means of the projection of a FocP above TP or above vP, as respectively schematized in (35) (see Rizzi 1997, 2004, 2006, 2010, 2015, 2018; Belletti 2001, 2004). Observe that, crucially, the distinct structures in (35a) and (35b) tacitly implicate that a left-peripheral focus and its postverbal counterpart are associated with distinct presuppositions, represented by TP and vP, respectively, where one includes tense and aspect specification (and other TP-related information and occasional adjuncts) and the other does not.

The interpretive distinction predicted to be possible by the structures in (35) is, however, not attested. The presupposition associated with the left-peripheral focus in (36B1), for example, is semantically identical to that associated with its postverbal counterpart in (36B2). In discourse, focalization in both declarative sentences (36B1) and (36B2) is a strategy to resolve the exact same gap (i.e., to identify the variable) posed by the question in (36A). In other words, both (36B1) and (36B2) are associated with the exact same presupposition stated in (37), crucially including tense and aspect information, which is not part of the vP (cf. (35b) above).

Let us assume, for the sake of the argument, that tense and aspect information could somehow be interpreted as being part of the presupposition associated with a focus in a structure like (35b) above or even that it could be argued not to actually be a necessary part of the presupposition. Still, in order for the vP to qualify as a proper presupposition at LF for a focus sitting at its edge, all the elements overtly moved out of the vP would have to reconstruct back into the vP – more precisely, all the presupposed material would have to be put back under the scope of the focus at LF in order for a covert focus-presupposition configuration to be created in the postverbal area.

Contrary to the above expectations, however, we can readily see that reconstruction of all vP-internal material cannot always be enforced. Suppose that sentence (36B2) above has the LF structure in (38), without reconstruction of the subject and the verb, which have both moved to the TP domain in overt syntax. The focus só lasanha ‘only lasagna’ in (38) is in this case associated with an “empty” vP as its presupposition, an unwelcome result. Thus, (38) cannot be the correct LF structure of (36B2). The subject and the verb must both reconstruct to their vP-internal positions at LF, as in (39), in order to be under the scope of the focus, like in its left-peripheral counterpart in (40).

Though one could presumably argue that verb movement is semantically vacuous and therefore the verb always reconstructs back into the VP, there are still many cases attested where reconstruction of the subject to a position under the postverbal focus is simply not possible.^[4] This is seen in sentences such as (41B) and (42B), where the subject must distribute over the focus or bind into the focus, respectively. The LF structures in (43) are thus ill-formed.

Upon closer scrutiny, we see that the low focus movement analysis fails to live up to its promise, namely the creation of a proper focus-presupposition articulation of the clause in the postverbal area, considering that the vP does not count as a proper presupposition at LF. Assuming that postverbal foci in Brazilian Portuguese are realized truly in situ in overt syntax and cannot involve covert focus movement to a low focus position in the periphery of vP for the reasons mentioned above, we must now raise the question of whether postverbal focalization in situ must comply with Rizzi’s (1997) Focus Criterion at LF, through covert movement to a focus position in the left periphery of the clause.

4.1 Negative concord

When proposing the Focus Criterion, Rizzi (1997) pointed out its similarity with the Neg Criterion, that is, the requirement that a negative operator be interpreted in the specifier of a negative phrase (see e.g. Haegeman & Zanuttini 1991; Haegeman 1995):

we will assume that a constituent endowed with topic or focus features must end up in a Spec/head configuration with Top or Foc, respectively; in other words, there are Topic and Focus Criteria, reminiscent of the Wh and Neg Criteria (Rizzi 1997: 287)

As is usually assumed, negative concord items must take scope from the position of the sentential negation (see e.g. Zeijlstra 2004; Giannakidou 2006 for relevant discussion); in other words, negative concord items are subject to the Neg Criterion. That being so, when a negative concord item is focalized, it should in principle simultaneously satisfy both the Neg Criterion and the Focus Criterion, which leads to a conundrum.

In the cartographic literature, criteria are associated with the so-called “criterial freezing”, that is, the restriction that criteria cannot be satisfied en passant. As a consequence, a given element cannot move from one criterial position to another (see e.g. Rizzi 2006, 2010, 2015, 2018 for extensive discussion and possible deductions of freezing effects). Therefore, if both the Neg Criterion and the Focus Criterion are assumed to hold in a given language, a focalized negative concord item cannot comply with both at the same time. Violation of either would then lead to ungrammaticality.

Brazilian Portuguese is a negative concord language, as is evidenced by the fact that sentence (49B) involves a single interpretable negative operator, that is, it is not interpreted as a double negative. Moreover, sentence (49B) shows that the focalization of a negative concord item does not lead to ungrammaticality. In light of this, I will in this section argue that the focalization of negative concord items in Brazilian Portuguese can be accounted for if we keep the Neg Criterion and dispense with the Focus Criterion in this language.

Let us first make the case that negative concord items in Brazilian Portuguese take scope from Spec, NegP, in compliance with the Neg Criterion. Putting aside modal flavors and restricting our discussion to a deontic reading of the modal verb pode ‘can’, we observe that sentence (50) is scopally ambiguous, in that either the modal verb can take scope over the quantified direct object dois livros ‘two books’ or the direct object can take scope over the modal verb. The first option gives rise to a reading where John is allowed to read a certain number of books, the cardinality two being included; the second option gives rise to a reading where there are two (specific) books which John is allowed to read.

In sentence (51), there is a negative concord item, namely, nenhum livro ‘no book’. As opposed to sentence (50) above, sentence (51) has only one interpretation, as far as the scope interaction between the modal verb and the negative operator is concerned. The reading where pode ‘can’ takes scope over nenhum livro (i.e., John is allowed to read no books at all) is available, whereas the reading where nenhum livro takes scope over pode (i.e., no book is such that John can read it) is unavailable.^[7]

Given that the negation in (51) is part of the embedded infinitival clause and as such is structurally below the matrix modal verb, nenhum livro cannot take scope over pode. In compliance with the Neg Criterion, nenhum livro must take scope from the position of the overt negation (i.e. Spec, NegP), below pode.

Taking the above restriction regarding negative concord to our advantage, we can assess the possibility of covert focus movement in Brazilian Portuguese by analyzing the scope behavior of a focalized negative concord item and seeing whether it is interpreted at LF in the position of the negation, in compliance with the Neg Criterion, or in the left periphery of the sentence, in compliance with the Focus Criterion.

Let us look at the example in (52B), where the direct object is a negative concord item and is focalized in the context of the question in (52A). What is at stake in this dialogue is how many toys the children received and crucially the answer in (52B) only has the surface scope interpretation, whereby there are two children (x and y) which are such that they did not receive any toy.

The postulation of covert focus movement of the direct object nenhum brinquedo ‘no toy’ in (52B) creates an LF structure along the lines of (53).^[8] Though (53) complies with the Focus Criterion, it incorrectly generates an inverse scope interpretation, whereby no toy was such that two children received it, which is unavailable for (52B) in this scenario, and fails to derive the correct surface scope reading.

Moreover, the purported covert focus movement depicted in (53) violates the Neg Criterion, which requires that nenhum brinquedo remain in Spec, NegP at LF in order for its negative quantificational force to be licensed (bearing in mind the criterial freezing). Crucially, note that sentence (52B) is grammatical despite the fact that it cannot respect both the Neg Criterion and the (purported) Focus Criterion at the same time. If we dispensed with the Neg Criterion, we would lose the well-known generalization illustrated in (51) above that negative concord items must take scope from the position of the overt negation. On the other hand, if we dispense with the Focus Criterion, we can assign to sentence (52B) the LF structure represented in (54) below and correctly capture the fact that nenhum brinquedo ‘no toy’ takes scope under duas crianças ‘two children’.

In sum, when the Neg Criterion and the Focus Criterion contradictorily “compete” for the licensing of a focalized negative concord item, the data show that the former requirement is valid, whereas the latter can (in fact, must) be dispensed with in Brazilian Portuguese.

Indeed, we find additional evidence for the claim above in the interaction of focalized negative concord items and modal verbs (cf. (51) above). Let us look at the contrast in interpretation between sentences (55a) and (55b). In (55a), the matrix clause is negated (i.e., negation is above the modal verb), whereas in (55b) the embedded clause is negated (i.e., negation is below the modal verb). Sentence (55a) thus has the interpretation that John was supposed to give all the books to Mary and he did so (i.e., no book was such that John forgot to give it to Mary); sentence (55b), on the other hand, has the interpretation that John was supposed not to give any books to Mary but he forgot to do so (i.e., he gave at least one book to Mary).

Regarding (55a), it is easy to observe that the fact that nenhum livro ‘no book’ takes scope over esqueceu ‘forgot’ can be derived either by a structure meant to comply with the Neg Criterion or by a structure meant to comply with the Focus Criterion, as both options would place nenhum livro above esqueceu at LF. In (55b), however, if nenhum livro underwent covert focus movement, as represented in (56), it would necessarily take scope over esqueceu, contrary to fact, and furthermore violate the Neg Criterion. Again, if we choose to preserve the Neg Criterion and dispense with the Focus Criterion, the correct interpretation is straightforwardly derived, as represented in (57).

While (57) gives us the correct interpretation of (55b) without resorting to the Focus Criterion, one could still try to salvage it by appealing to a focus position in the periphery of the embedded clause, as in (58), which would correctly place nenhum livro under the scope of esqueceu. However, there is reason to believe that such focus position is in fact not available.

Note that the embedded sentences in (55) above can be analyzed as a type of restructuring clause that may lack a CP field altogether. As Wurmbrand (2014, 2015) argues, the possibility of negative concord across the clause boundary (as in (55a) above) indicates that the infinitival clause is of reduced size. Thus, expectedly, overt focus movement to the left periphery of the matrix clause is possible, whereas overt focus movement to the periphery of the embedded clause is ruled out, as the contrast in (59) shows. Needless to say, the structure in (58) would still violate the Neg Criterion, which I maintain we need to keep in order to account for the contrasts discussed in this section.

Before ending this discussion, I will now argue that the postulation of low focus movement (i.e., focus movement to a position in the periphery of vP, as discussed in Section 3) cannot be invoked in order to try to salvage the Focus Criterion in the event of a conflict with the requirements of the Neg Criterion. Let us then take another look at sentence (52B), repeated below in (60).

If nenhum brinquedo ‘no toy’ moved (either overtly or covertly) to Spec, FocP in the periphery of vP in order to comply with the Focus Criterion, as in (61a), it would become frozen in place (cf. the criterial freezing) and thus unable to also satisfy the Neg Criterion at LF in order to license its negative quantificational force; in a structure where nenhum brinquedo moves to Spec, NegP, as in (61b), it would then violate the Focus Criterion, which cannot be satisfied en passant.

The same problems are replicated with respect to the sentences in (55) above, whose structures with the postulation of low focus movement are given in (62) below. In the LF representation of (55a) given in (62a), nenhum livro ‘no book’ would not be able to comply with the Neg Criterion and would crucially take scope under the matrix modal verb esqueceu ‘forgot’, contrary to fact. Similarly, the LF structure in (62b), though giving rise to the correct scope relation, would violate the Neg Criterion. Needless to say, movement of nenhum livro from Spec, FocP to Spec, NegP in both structures would violate the Focus Criterion.

To conclude, when a conflict arises between the Neg Criterion and the Focus Criterion from the focalization of a negative concord item, the data discussed in this section indicate that the Neg Criterion is operative in Brazilian Portuguese, whereas the Focus Criterion is not. Dispensing with the latter and preserving the former gives us the right results with respect to the scope possibilities attested in sentences involving negative concord, which provides evidence against the postulation of covert focus movement in this language.^[9] In the next section, this claim will be corroborated by data from contrastive topicalization.

4.2 Contrastive topicalization

I will now argue that in sentences where a contrastive topic and a focus co-occur in Brazilian Portuguese, the assumption of covert focus movement to the left periphery may lead to illicit information structure configurations. That problem can be circumvented if we allow for foci to be licensed in situ.

While traditional aboutness topics (in the sense of Reinhart 1981; see also Bianchi & Frascarelli 2010) are standardly assumed to be interpreted relatively to a comment in a bipartite articulation of the clause as topic-comment, as in (63), contrastive topics are instead argued to be interpreted in association with a focus (see e.g. Büring 2003, 2016; Wagner 2012).

As Büring (2003) discusses, a question like Who ate what? in (64A) can be interpreted as asking (i) for each person what that person ate or (ii) for each food who ate that food. The answer in (64B1), where Fred is pronounced as a contrastive topic and the beans as a focus, addresses the first interpretation of the question (i.e., the people function as independent variables), whereas the answer in (64B2), where Fred is pronounced as a focus and the beans as a contrastive topic, addresses the second interpretation of the question (i.e., the foods function as independent variables).

The distribution of contrastive topics relatively to their associated foci is a complex matter and subject to much ongoing debate (see e.g. Neeleman & van de Koot 2008, 2010, Lacerda 2020, and references above), which is beyond the scope of this paper. Still, two distributional restrictions observed in Brazilian Portuguese by Lacerda (2020) can be used to our advantage to assess whether foci undergo covert movement in this language. I will argue that the postulation of covert focus movement fails to capture both restrictions and incorrectly allows the creation of illicit information structure configurations.

The first restriction is that contrastive topics must precede the focus in order to newly introduce (a relevant set of) contrastive alternatives in the discourse. More precisely, if the referent denoted by the contrastive topic is not known to be part of a relevant set of contrastive alternatives in the conversational context, the contrastive topic is only licit if it precedes the focus, as we will see below.^[10]

Let us compare the dialogues in (65) and (66), where both answers are given in the canonical S–V–DO–IO order. Assume that Speaker A in (65) asks a question just about Barriers and no other books. In this case, Speaker B can reply as in (65B), with o Barriers marked as contrastive topic, raising the implicature that there are other books that John might have given to other people.

Likewise, assume that in (66) Speaker A asks a question just about Mary and no other people. Now, Speaker B cannot (pragmatically felicitously) reply as in (66B), with pra Maria marked as a contrastive topic; that is, even though pra Maria is marked as a contrastive topic, the reply fails to introduce (a relevant set of) contrastive alternatives and raise the implicature that there are other people to whom John might have given books.

In accordance with the above restriction, the sentence becomes acceptable with the relevant interpretation if the contrastive topic pra Maria ‘to Mary’ is dislocated to a position before the focus, as in (67B), or if Mary’s membership in a relevant set of contrastive alternatives is established in the discourse, as in (68B).

Taking advantage of the contrasts in (65)–(68) with respect to the restriction described above, let us now probe into the possibility of covert focus movement by examining the contrasts in (69) below. If the focus só uma vez ‘only once’ moved covertly in the grammatical sentence (69B1), it would create an LF structure analogous to sentence (69B2), which is however ruled out under the relevant information structure (with the direct object o Barriers interpreted and intonationally accented as a contrastive topic). Given that Barriers was not previously known (by Speaker A) to be a (contrastive) member of a relevant set of books, marking it as a contrastive topic in (69B2) is illicit, as it follows the focus, and the reply fails to raise the implicature that there are other books that John read. Expectedly, sentence (69B3) is acceptable (cf. (67B) above).^[11]

Suppose for the sake of the argument that, in addition to covert focus movement, we could also allow covert topic movement in (69), in order to try to salvage the necessary topic-focus order at LF. Sentence (69B1) would then have an LF structure analogous to sentence (69B3), which is well-formed, as the contrastive topic precedes the focus, a seemingly welcome result. However, resorting to covert topic movement in addition to focus movement would incorrectly rule in the ungrammatical sentence (69B2), which would then be indistinguishable from the grammatical (69B1) and (69B3) in the relevant respects at LF. We see then that neither the postulation of covert focus movement alone nor the postulation of covert topic and focus movement combined can account for the paradigm in (69), which is in turn accounted for if covert focus movement does not take place.

That focalization in situ cannot involve covert focus movement in Brazilian Portuguese is further corroborated by a second restriction, namely that contrastive topics cannot be dislocated to a position under the focus, which now gives rise to ungrammaticality and not just pragmatic infelicity. Let us first observe the scenario in (70), where the answers in (70B1) and (70B2) associate the contrastive topic do Chomsky ‘by Chomsky’ with the postverbal focus só dois livros ‘only two books’ (the question under discussion thus being for each author how many books by that author Speaker B read). Observe that the contrastive topic can be dislocated either to the left periphery of the sentence, as in (70B1), or to the postverbal middle field, as in (70B2); in both cases, the dislocated contrastive topic precedes the focus.

Let us now look at (71) below. In contrast to (70B1) and (70B2) above, the focus associated with the contrastive topic in (71B1) and (71B2) is the preverbal subject eu ‘I’. Now, the topic do Chomsky ‘by Chomsky’ can be dislocated to the left periphery, as in (71B1), but not to the postverbal middle field, as in (71B2), which is utterly unacceptable (even if Chomsky is already known by Speaker A to be a part of the relevant set of authors; cf. the first restriction discussed above), in a sharp contrast with (70B2) above.

As the examples above show, movement to either the periphery of CP or the periphery of vP may serve to reposition a contrastive topic with respect to the focus in Brazilian Portuguese – which is successful in (70B1), (70B2), and (71B1), but not in (71B2).^[12] Taking advantage of the two possible landing sites for a dislocated contrastive topic, let us now assess the possibility of covert focus movement in these structures. Suppose that the focalized direct object só dois livros ‘only two books’ in (70B1) above moved covertly as in (72a) below. Assuming that this movement could target a position under the contrastive topic, (72a) would be a well-formed information structure configuration. Note, however, that if covert focus movement of só dois livros were obligatory in (70B1)/(72a), it would also need to take place in (70B2) above, where the contrastive topic is in the postverbal middle field; that operation would then create the structure in (72b), where the focus is in a position above the dislocated contrastive topic – which is an illicit configuration, as the ungrammaticality of (71B2) above shows.^[13]

As was the case in the discussion of (69) above, the attempted introduction of covert topic movement to try to salvage covert focus movement would create LF structures that are indistinguishable in the relevant respects for all sentences in (70) and (71), as schematized in (73), thus incorrectly ruling in the unacceptable (71B2).^[14]

In conclusion, the contrasts discussed in this section cannot be captured under the assumption that covert focus movement takes place in Brazilian Portuguese, but can be accounted for if foci in situ are allowed to be syntactically licensed and interpreted (qua informational objects) in sentence-internal positions at LF. Contrastive topicalization therefore provides further evidence that Rizzi’s (1997) Focus Criterion is not operative in Brazilian Portuguese.