Inferential Interrogatives with qu é in Spanish

: In this paper, we discuss the evidential properties of inferential interrogative sentences with qu é in Spanish. This interrogative type exhibits the shape of a wh -question but the interpretation of a polar question. These sentences have the additional particularity that they are interrogatives with evidential material, which are attested but not frequent crosslinguistically, if compared with declarative evidentials. An interesting consequence of their double interrogative and evidential nature is the fact that both discourse participants have a prominent role in the interpretation of these sentences, as the Speaker makes the inference but the Addressee is requested for conﬁrmation. To account for the construction, we assume a multiple-layered system that includes both Speech Act projection and Finiteness projection. In these two areas we simultaneously ﬁnd evidential material housing the Speaker’s inference, and a raised Addressee in its prominent interrogative position as the participant with the knowledge to provide the requested conﬁrmation of the interrogative’s truth value.


Introduction
The goal of this paper is to account for inferential interrogatives with qué in Spanish, an interrogative type with the appearance of a wh-question, since it involves the wh-pronoun qué 'what', but interpreted as an inferential yes/no question.Examples are shown in (1).This class of interrogatives has been previously identified as split interrogatives or split questions (Arregi 2007(Arregi , 2010;;Contreras and Roca 2007;López-Cortina 2003, 2009;Fernández-Soriano 2021), dislocated questions (Lorenzo 1994), and wh-doubling (Camacho 2002), since they seemingly involve a split in their structure.They have also been called compound interrogatives (Py 1971) and adjunct tags (Uriagereka 1988).
The reason why these interrogatives are identified as split is their hybrid nature.While their initial part exhibits the wh-pronoun qué 'what' as well as falling intonation, characteristic of wh-interrogatives, their final rising intonation and their interpretation as yes/no questions distinguish them from wh-interrogatives.A similar interrogative pattern has been found in Catalan (Contreras and Roca 2007) and English (López-Cortina 2009): Languages 2023, 8, 282 2 of 23 (2) a.What are you, on a diet?b.What did you have, a food fight here?(López-Cortina 2009, p. 220 [1]) More recently, these constructions have been identified as invariable qué questions (Fernández-Sánchez and García-Pardo 2020a, 2020b, 2023;Reig Alamillo 2019) and nonmatching split interrogatives (Fernández-Soriano 2021) to highlight the invariable nature of the interrogative pronoun involved in these constructions.Thus, although the tag contains a [wh] feature, it is crucially not a content wh-pronoun, hence, it is always realized as the default wh-word qué at PF, unlike other split interrogative classes.This explains the asymmetries between inferential interrogatives involving the default wh-operator qué and other split interrogative classes involving full-fledged wh-operators, such as cómo 'how', cuándo 'when', and dónde 'where' (3): (3) a.
¿Qué Interpretation-wise, split interrogatives in general appear to have a confirmational value, i.e., the speaker requests information to confirm a previous suspicion or intuition (López-Cortina 2009).In this sense, these interrogatives show a strong evidential component, as also shown in Fernández-Sánchez andGarcía-Pardo (2020b, 2023), since unlike other types of confirmation interrogatives such as tag questions, the expected reply in inferential interrogatives with qué is invariably constrained by the speaker's inferred or presupposed answer.For example, in the reading of a sentence such as (4), the speaker makes an inference about the truth value of the proposition on the basis of indirect evidence over the content of the proposition (e.g., the speaker sees the addressee while she enters the room with shopping bags).The construction then shows an interpretative behavior similar to inferential evidentials, just as the one described by Bhadra (2017Bhadra ( , 2018Bhadra ( , 2020) ) for Bangla.
The evidential element in these constructions has an unusual Speaker-oriented interpretation, since interrogatives are typically Addressee-oriented.We propose that this unexpected interpretation follows from the interaction between the discursive elements present in the Finiteness Phrase (FinP) and the Speech Act Phrase (SAP) projections.More specifically, we propose a structure that involves the interplay between the presence of the Interrogative Flip, typical of evidential interrogatives (Aikhenvald 2004;San Roque et al. 2017), the evidential component present in these sentences and realized by the presence of the Speaker participant in a Fin projection (Bhadra 2020), and the presence, above ForceP, of an SAP where the Speaker and Addressee participants are anchored to the discourse and activate the inferential and confirmational interpretations, respectively, by means of a coindexation system with the relevant clausal elements (Bianchi 2003(Bianchi , 2006)).This configuration explains why these interrogatives are interpreted as both inferential and confirmational (i.e., the speaker infers an answer in the tag, while the addressee is asked for full confirmation of the truth value of that inferred answer).This double discursive layer also accounts for the complex prosodic pattern typical of these constructions, in line with Escandell-Vidal's (2017) proposal for interrogatives with marked prosody.
The organization of this article is as follows.In Section 2, we list the grammatical properties displayed by inferential interrogatives with qué that were described in previous work such as Fernández-Sánchez and García-Pardo (2020a).In Section 3, we take a look at recent analyses of the construction.In Section 4, we discuss evidentiality in the context of interrogatives.Section 5 offers our proposal, a formal analysis based on the interaction between a Speech Act Phrase and a Finiteness Phrase.Section 6 addresses some consequences of our proposal and links our formal analysis with the grammatical properties of inferential interrogatives.Finally, in Section 7, we present the main conclusions.

Defining Characteristics of Inferential Interrogatives with qué
In this section, we review the main grammatical properties of inferential interrogatives with qué in Spanish, as they are mentioned in previous work (e.g., Fernández-Sánchez and García-Pardo 2020a), which contribute to establish crucial distinctions between these constructions and other types of interrogative clauses.

An Unexpected Intonation
Inferential interrogatives with qué exhibit a wh-pronoun in their initial part and falling intonation, characteristic of wh-interrogatives.What distinguishes these interrogatives from conventional wh-interrogatives is their final rising intonation and their interpretation as yes/no questions.This final part is frequently identified as the tag.A sentence such as (1a) above shows the intonation informally represented in (5): (5) ¿Qué vienes, de la calle?
discursive layer also accounts for the complex prosodic pattern typical of these constructions, in line with Escandell-Vidal's (2017) proposal for interrogatives with marked prosody.
The organization of this article is as follows.In Section 2, we list the grammatical properties displayed by inferential interrogatives with qué that were described in previous work such as Fernández-Sánchez and García-Pardo (2020a).In Section 3, we take a look at recent analyses of the construction.In Section 4, we discuss evidentiality in the context of interrogatives.Section 5 offers our proposal, a formal analysis based on the interaction between a Speech Act Phrase and a Finiteness Phrase.Section 6 addresses some consequences of our proposal and links our formal analysis with the grammatical properties of inferential interrogatives.Finally, in Section 7, we present the main conclusions.

Defining Characteristics of Inferential Interrogatives with qué
In this section, we review the main grammatical properties of inferential interrogatives with qué in Spanish, as they are mentioned in previous work (e.g., Fernández-Sánchez and García-Pardo 2020a), which contribute to establish crucial distinctions between these constructions and other types of interrogative clauses.

An Unexpected Intonation
Inferential interrogatives with qué exhibit a wh-pronoun in their initial part and falling intonation, characteristic of wh-interrogatives.What distinguishes these interrogatives from conventional wh-interrogatives is their final rising intonation and their interpretation as yes/no questions.This final part is frequently identified as the tag.A sentence such as (1a) above shows the intonation informally represented in (5): (5) ¿Qué vienes, de la calle?
Compare with the rising intonation for the yes/no question in (6) and the falling intonation for the wh-question in (7); see Hualde ( 2005 discursive layer also accounts for the complex prosodic pattern typical of these constructions, in line with Escandell-Vidal's (2017) proposal for interrogatives with marked prosody.
The organization of this article is as follows.In Section 2, we list the grammatical properties displayed by inferential interrogatives with qué that were described in previous work such as Fernández-Sánchez and García-Pardo (2020a).In Section 3, we take a look at recent analyses of the construction.In Section 4, we discuss evidentiality in the context of interrogatives.Section 5 offers our proposal, a formal analysis based on the interaction between a Speech Act Phrase and a Finiteness Phrase.Section 6 addresses some consequences of our proposal and links our formal analysis with the grammatical properties of inferential interrogatives.Finally, in Section 7, we present the main conclusions.

Defining Characteristics of Inferential Interrogatives with qué
In this section, we review the main grammatical properties of inferential interrogatives with qué in Spanish, as they are mentioned in previous work (e.g., Fernández-Sánchez and García-Pardo 2020a), which contribute to establish crucial distinctions between these constructions and other types of interrogative clauses.

An Unexpected Intonation
Inferential interrogatives with qué exhibit a wh-pronoun in their initial part and falling intonation, characteristic of wh-interrogatives.What distinguishes these interrogatives from conventional wh-interrogatives is their final rising intonation and their interpretation as yes/no questions.This final part is frequently identified as the tag.A sentence such as (1a) above shows the intonation informally represented in (5): (5) ¿Qué vienes, de la calle?
Compare with the rising intonation for the yes/no question in (6) and the falling intonation for the wh-question in (7); see Hualde ( 2005 tions, in line with Escandell-Vidal's (2017) proposal for interrogatives with marked prosody.
The organization of this article is as follows.In Section 2, we list the grammatical properties displayed by inferential interrogatives with qué that were described in previous work such as Fernández-Sánchez and García-Pardo (2020a).In Section 3, we take a look at recent analyses of the construction.In Section 4, we discuss evidentiality in the context of interrogatives.Section 5 offers our proposal, a formal analysis based on the interaction between a Speech Act Phrase and a Finiteness Phrase.Section 6 addresses some consequences of our proposal and links our formal analysis with the grammatical properties of inferential interrogatives.Finally, in Section 7, we present the main conclusions.

Defining Characteristics of Inferential Interrogatives with qué
In this section, we review the main grammatical properties of inferential interrogatives with qué in Spanish, as they are mentioned in previous work (e.g., Fernández-Sánchez and García-Pardo 2020a), which contribute to establish crucial distinctions between these constructions and other types of interrogative clauses.

An Unexpected Intonation
Inferential interrogatives with qué exhibit a wh-pronoun in their initial part and falling intonation, characteristic of wh-interrogatives.What distinguishes these interrogatives from conventional wh-interrogatives is their final rising intonation and their interpretation as yes/no questions.This final part is frequently identified as the tag.A sentence such as (1a) above shows the intonation informally represented in (5): (5) ¿Qué vienes, de la calle?
Compare with the rising intonation for the yes/no question in (6) and the falling intonation for the wh-question in (7); see Hualde (2005): (6) ¿Vienes de la calle?(7) ¿De dónde vienes?A similar, but not identical, interrogative pattern has been found in Catalan (Contreras and Roca 2007) and English (López-Cortina 2009) What the sentences in (8-9) share with inferential interrogatives with qué is a complex prosodic pattern, as proposed in Escandell-Vidal (2017) also for other types of marked interrogatives.
2.2.The wh-Word Is Always qué 'What' Lorenzo (1994), Fernández-Sánchez and García-Pardo (2020a, 2020b), and Fernández-Soriano (2021) identify crucial distinctions between the split interrogative class and inferential interrogatives with qué.A main distinction, which we also assume, is the fact that inferential interrogatives with qué exclusively involve the wh-word qué as in (10a), whereas other split interrogatives involve any content wh-phrase, as illustrated in (10b).This is why Lorenzo (1994)  The compatibility of the interrogative element with a preposition is a good test to distinguish these two types of split interrogatives, which is particularly useful when the full-fledged interrogative is qué 'what'.It is also an indicator that the wh-word in these constructions is not a full-fledged interrogative pronoun.

The wh-Question Is Not Independent
As Fernández-Sánchez and García-Pardo (2020a, 2020b) discuss, inferential interrogatives with qué cannot involve two independent interrogatives.While the initial part of split interrogatives is independent (12), that of inferential interrogatives with qué cannot stand on its own, as seen in the ungrammaticality of (13).This contrast is an indication that inferential interrogatives with qué are monoclausal.We develop this point further below in Section 3.3.

Inferential Interrogatives with qué Accept Tags Other than DP
Also reported in Fernández-Sánchez and García-Pardo (2020a), only inferential interrogatives with qué may be equivalent to a true yes/no question, and they may accept tags beyond the DP level, as seen in ( 14), with a DP followed by a comitative PP.Other split questions are more restricted in this sense.

Inferential Interrogatives with qué Have a Confirmational Flavor
Split interrogatives in general have a confirmational value, in the sense that the speaker requests the addressee to confirm a previous suspicion or intuition in their expected answer (López-Cortina 2009).This confirmational flavor is present in other types of interrogatives.For example, an inference based on evidentiality is precisely what we find in constructions such as Bianchi and Cruschina's (2019) polar interrogatives with fronted focus in English, also found in Spanish: (15) a.
Soup are you making?(Carter and McCarthy 2006, p. 780) b. ¿Gazpacho estás haciendo?Gazpacho be:2SG making 'Gazpacho are you making?'This type of polar question has a confirmational value, which can only be understood if produced in a context that can be used to trigger the evidential reading.For example, the sentences in ( 15) are interpreted as confirmational if uttered when we enter the kitchen and see somebody preparing the necessary ingredients for the specific dish.
Interestingly, inferential interrogatives with qué show a behavior that is similar to inferential evidentials crosslinguistically.Bhadra (2018) analyzes the evidential marker naki in Bangla as a case of indirect evidence (Rooryck 2001) that can occur in different types of sentences, including interrogatives: (16) Sita baRi giy-ech-e naki?Sita home go-PERF-3SG NAKI 'Sita has gone home.Has she? (Bhadra 2018, p. 1 [1a]) Bhadra (2018) claims that one of the roles of this particle is to ask for confirmation of the positive answer expected after inferring the truth-value from indirect evidence.In (17), we find another clear example from Bangla where the evidential marker naki indicates that some indirect evidence proves that what is asserted is true.(17) Context: Ram knows that Mina has been thinking about going to America for a while now but has not made up her mind yet.Today, he suddenly sees several of her suitcases, all packed, sitting out in the hall and asks her brother: Mina amerika chol-e ja-cche naki?Mina America go-IMPERF go-3SG.PRES.PROGR NAKI '(Given what I inferred) Mina is going away to America (is it true)?' (Bhadra 2018, p. 2[3]) What is interesting about this particle is that this interpretation is only available if it appears in an interrogative sentence.If it appears in a declarative sentence, the interpretation of the particle does not have confirmation value, as seen in ( 18), which has a strictly reportative value.
(18) Context: Ram heard a rumor about his neighbor that he is now reporting to his friend Sita: Mina naki amerika chol-e ja-cche Mina NAKI America go-IMPERF go-3SG.PRES.PROG 'Mina is going away to America (I hear)' (Bhadra 2018, p. 2[2]) The behavior of this particle is evidence of the projection of evidentiality material in syntax, as well as its composition interpretation, as will be argued in this paper.Sections 5 and 6 below further develop the idea that evidentiality projects in syntax.

Both Types of Split Questions May Be Preceded by Topics
Finally, both types of split questions allow the wh-word to be preceded by topics.Both sentences in ( 19) allow the preceding topic el helado 'the ice-cream'.As just seen, the sentence in (19a) is the inferential interrogative with qué, disallowing a preceding preposition, in contrast with the split interrogative in (19b) In the next section, we show recent analyses of the construction, divided between monoclausal and biclausal approaches.

Previous Recent Analyses
In this section, we discuss two recent analyses of Spanish inferential interrogatives with qué.One is monoclausal (Fernández-Sánchez and García-Pardo 2020a), whereas the other one is biclausal (Fernández-Soriano 2021).

A Monoclausal Analysis
Fernández-Sánchez and García-Pardo (2020a) suggest an analysis based on a low FocP (à la Belletti 2001Belletti , 2005)), whereby the interrogative operator qué is base-generated in situ.More precisely, it originates in spec-CP, not involving any kind of movement.For the sentence in (20), the authors provide the analysis in ( 21): Fernández-Sánchez and García-Pardo 's analysis is based on VP-movement to spec-FocP.However, this is problematic in cases in which we have a complex verb (with auxiliaries), since it is the auxiliary part that occupies T, whereas V remains lower.In such cases, we would obtain the ordering Qué Aux Subj VP: *¿Qué has tú venido, en bicicleta?, contrary to facts.In particular, if the finite auxiliary is in T and then moves to C (as the lexical V vienes in ( 21)), the outcome involves the ordering Aux+Subject+VP, which is completely ill-formed.
Also, embedding of the inferential interrogative appears to be possible, at least in our dialect, which suggests that the operator undergoes long-distance movement to the matrix CP: come well that come:1SG, the Saturday 'Isn't it good for you that I'm coming Saturday (I infer)?' In addition, as shown by López-Cortina (2009), some Spanish varieties allow the interrogative pronoun qué to appear in situ in these constructions, and the same is found in English.Compare ( 23) and ( 24 Fernández-Sánchez and García-Pardo 's analysis is based on VP-movement to spec-FocP.However, this is problematic in cases in which we have a complex verb (with auxiliaries), since it is the auxiliary part that occupies T, whereas V remains lower.In such cases, we would obtain the ordering Qué Aux Subj VP: *¿Qué has tú venido, en bicicleta?, contrary to facts.In particular, if the finite auxiliary is in T and then moves to C (as the lexical V vienes in ( 21)), the outcome involves the ordering Aux+Subject+VP, which is completely ill-formed.
Also, embedding of the inferential interrogative appears to be possible, at least in our dialect, which suggests that the operator undergoes long-distance movement to the matrix CP: In addition, as shown by López-Cortina (2009), some Spanish varieties allow the interrogative pronoun qué to appear in situ in these constructions, and the same is found in English.Compare ( 23) and ( 24 An analysis where the operator is base-generated is inconsistent with this data, as the alternation between in situ and wh-first constructions favor a movement analysis. 5If the operator occupies [Spec, CP] from the beginning, the connection between ( 23) and ( 24) is lost.Note that their interpretation is identical, and the only difference is syntactic, i.e., no movement of qué in (23), and movement of this operator in (24).
The analysis proposed by López-Cortina ( 2009) is based on the projection of a Confirmation Phrase (ConfP), whose complement is the tag of the interrogative and whose specifier is the operator qué.In (25), we offer the analysis proposed by López-Cortina for sentences such as (24a): (25) An analysis where the operator is base-generated is inconsistent with this data, as the alternation between in situ and wh-first constructions favor a movement analysis. 5If the operator occupies [Spec, CP] from the beginning, the connection between ( 23) and ( 24) is lost.Note that their interpretation is identical, and the only difference is syntactic, i.e., no movement of qué in ( 23), and movement of this operator in ( 24).
The analysis proposed by López-Cortina ( 2009) is based on the projection of a Confirmation Phrase (ConfP), whose complement is the tag of the interrogative and whose specifier is the operator qué.In (25), we offer the analysis proposed by López-Cortina for sentences such as (24a): (25) Depending on the variety of Spanish, the operator will undergo wh-movement or remain in its original position.
This analysis captures the confirmational meaning of these interrogatives, but it fails to account for the focus interpretation of the tag.Actually, as is clear from the derivation in (25), no focus interpretation is taken into account by López-Cortina.In turn, Fernández-Sánchez and García-Pardo (2020a) propose that the tag is focused, as seen in their analysis in (21).In our analysis in Section 5, we also agree with a focus interpretation of the tag.
Below, in Section 3.3, we also argue in favor of a monoclausal analysis.In view of the data in this section, we also propose that the interrogative operator is the result of raising, as it appears to account for the behavior of these constructions regardless of the variety.

A Biclausal Analysis
Fernández-Soriano (2021), in turn, suggests the following biclausal analysis of her non-matching split interrogative clauses: In her analysis, the whole IP in the second clause is subject to ellipsis à la Merchant (2004).For Fernández-Soriano, the neuter operator qué and the tag are contained in an FP or Speech Phrase, which she claims is the phrase corresponding to discourse phrases in monoclausal analyses.In our proposal, we will see that the evidentiality reading of these interrogatives is the consequence of projecting a Speech Act Phrase, but this will be located in the top of the tree (Miyagawa 2017(Miyagawa , 2022)), not in the middle field. 6In Fernández-Soriano's analysis, the evidential interpretation is the responsibility of FP in the highest CP in (26), whereas in our proposal this interpretation is the consequence of the projection of a Speech Act Phrase on top of the only CP.
A biclausal analysis is justified for split questions with content interrogative pronouns, since both parts are independently grammatical.This is, however, not what happens in the interrogatives under study here, whose first part is not grammatical on its own, as argued by Fernández-Sánchez andGarcía-Pardo (2020a, 2020b).That is, these Depending on the variety of Spanish, the operator will undergo wh-movement or remain in its original position.
This analysis captures the confirmational meaning of these interrogatives, but it fails to account for the focus interpretation of the tag.Actually, as is clear from the derivation in (25), no focus interpretation is taken into account by López-Cortina.In turn, Fernández-Sánchez and García-Pardo (2020a) propose that the tag is focused, as seen in their analysis in (21).In our analysis in Section 5, we also agree with a focus interpretation of the tag.
Below, in Section 3.3, we also argue in favor of a monoclausal analysis.In view of the data in this section, we also propose that the interrogative operator is the result of raising, as it appears to account for the behavior of these constructions regardless of the variety.

A Biclausal Analysis
Fernández-Soriano (2021), in turn, suggests the following biclausal analysis of her non-matching split interrogative clauses: In her analysis, the whole IP in the second clause is subject to ellipsis à la Merchant (2004).For Fernández-Soriano, the neuter operator qué and the tag are contained in an FP or Speech Phrase, which she claims is the phrase corresponding to discourse phrases in monoclausal analyses.In our proposal, we will see that the evidentiality reading of these interrogatives is the consequence of projecting a Speech Act Phrase, but this will be located in the top of the tree (Miyagawa 2017(Miyagawa , 2022)), not in the middle field. 6In Fernández-Soriano's analysis, the evidential interpretation is the responsibility of FP in the highest CP in (26), whereas in our proposal this interpretation is the consequence of the projection of a Speech Act Phrase on top of the only CP.
A biclausal analysis is justified for split questions with content interrogative pronouns, since both parts are independently grammatical.This is, however, not what happens in the interrogatives under study here, whose first part is not grammatical on its own, as argued by Fernández-Sánchez andGarcía-Pardo (2020a, 2020b).That is, these constructions do not consist of two independent interrogatives (i.e., a wh-interrogative followed by a yes/no question), since the wh-portion is not a well-formed independent clause in Spanish ( 27).The inferential interrogative in ( 27) cannot be used as an independent sentence, contrary to what a biclausal analysis predicts.Without a clear justification of the biclausal nature of these interrogatives, an ellipsis approach would be hard to maintain.
A monoclausal analysis where the verb moves to a focal position would be consistent with the data and the general behavior of verbs in Spanish interrogatives.We propose several tests next, which support a monoclausal analysis of inferential interrogatives with qué, while they hint at further asymmetries with other split interrogatives.

Testing the Biclausal/Monoclausal Analyses
Constituent preposing is a diagnostic associated with biclausality in Spanish ( 28 In (32), sentences exhibiting a complex tag are disallowed regardless of the position of the clitic, whereas the simplex tag renders the sentence grammatical: *¿Dónde quiere, comérselo en la cama?c.
These tests support a monoclausal analysis of inferential interrogatives with qué, and they also hint at further asymmetries with other split interrogatives.We conclude that a monoclausal analysis more accurately reflects the behavior of inferential interrogatives with qué than a biclausal analysis.In Section 6, we offer further arguments based on the prosodic contour of these constructions that a monoclausal analysis more accurately reflects the behavior of inferential interrogatives with qué than a biclausal analysis.

Evidentiality in Inferential Interrogatives with qué
Evidentiality has been extensively studied in languages with morphological evidential particles (Aikhenvald and Dixon 2003;Aikhenvald 2004).In these languages, evidential particles explicitly mark the source of information in a number of ways.According to Aikhenvald (2004), these particles vary in different languages, and they may encode that the information was reported by someone else, that the information was experienced firsthand by the speaker, sometimes visually, sometimes non-visually (e.g., through hearing or smelling), or by means of inference.Typically, evidential markers are obligatory and morphologically contrasted, depending on the type of source they specifically encode, as seen in ( 33) for Tariana, an Arawakan language spoken in Brazilian Amazon: Some authors have specifically explored the presence of evidential particles in interrogatives (Speas and Tenny 2003;San Roque et al. 2017;Bhadra 2017Bhadra , 2018Bhadra , 2020)).According to San Roque et al. (2017), evidentials both provide information about the utterance and associate that information with the speech act participants, thanks to their perspectivizing function.For these authors, interrogative utterances also marked for evidentiality combine two facets of the expression of epistemicity in language: while evidentials express the source of information one (e.g., a discourse participant) has for a given proposition, interrogatives involve the speech act of questioning, by means of which information is requested that is unknown to the speaker.
According to these authors, this double function seems paradoxical, as one typically asks about things one knows little about.In other words, it would be a paradox for the same discourse participant to both request information about something and indicate the source of their knowledge.In fact, evidentials are incompatible or restricted with interrogatives in a number of languages (Aikhenvald 2004).
In languages in which evidentials are permitted along with interrogatives, identical evidential markers may contribute contrasted information depending on whether they appear in declaratives or interrogatives (San Roque et al. 2017;Bhadra 2017Bhadra , 2018Bhadra , 2020)), which suggests that the interpretation of evidential particles is determined compositionally.
In this section, we discuss previous work on evidentials in interrogatives, paying special attention to the contribution that evidentials make to the syntactic composition of interrogatives, particularly the left-most left periphery.We also discuss how the presence of evidential material impacts the interpretation and markedness of interrogatives as well as the participants' point of view.

Change of Perspective in Interrogatives and the Interrogative Flip in Inferential Interrogatives
A general characteristic exhibited by interrogatives with morphologically overt evidentials is the presence of the Interrogative Flip (Tenny and Speas 2004;San Roque et al. 2017), a phenomenon by means of which the same evidential particle takes the speaker perspective in a declarative sentence, while it takes the perspective of the addressee in an interrogative.In the example in (34) from Duna, a Papuan language, the evidential affix yarua is addressee-oriented in the interrogative in (34A), but the same affix is speaker-oriented in the answer (a declarative sentence) in (34B): Besides Duna, this switch in perspective is obligatory in many world languages, including English (Speas and Tenny 2003;Tenny and Speas 2004), Japanese (Tenny 2006), Cuzco Quechua (Faller 2002), and Cheyenne (Murray 2010), to name a few, and, according to Bhadra (2020), it is associated with authority, in the sense that while the speaker has the authority in a declarative sentence in the sense that it is the speaker that possesses the knowledge behind an assertion, it is the addressee's knowledge that is sought in the answer to a question.Speas and Tenny (2003) propose a system to explain syntactic structures attending to discourse participants (i.e., Speaker and Addressee) in the form of a set of syntactic projections housed in the left-most left periphery.This system is able to explain phenomena such as agreement with discourse participants instead of syntactic arguments, by means of coindexation, as seen in the case of unagreement in Spanish, shown in (35), in which the inflected verb vamos 'we go' shows first-person plural agreement with the Speaker along with the plural subject los lingüistas 'linguists'.The diagram in (36) shows Tenny and Speas' (2004) proposed structure for an interrogative, which incorporates the Interrogative Flip: 7 (36) In the system proposed by Speas and Tenny (2003) and Tenny and Speas (2004), the Seat of Knowledge, which Tenny and Speas (2004) situate within the Utterance Content, is an evidential argument that stands for 'the sentient individual who is responsible for the truth of a proposition' (Tenny and Speas 2004).In the case of interrogatives, both the Utterance Content and Seat of Knowledge are controlled by the raised Addressee.This explains why the Addressee is the discourse participant that possesses the knowledge to provide the answer to the question.For example, the Seat of Knowledge is named by evidential verbs like appear.Because of the Interrogative flip, appear will be anchored to the Speaker in a declarative, but to the Addressee in an interrogative, as seen in ( 37): (37) a. Martin appearsS to have missed his exam.(The speaker knows) b.Does Martin appearA to have missed his exam?(The addressee knows) Inferential interrogatives with qué do not seem to pattern with prototypical interrogatives with evidentials in that they do not appear to present the Interrogative Flip, at least apparently, since the inference takes the Speaker's perspective rather than the Addressee's.For example, in a sentence such as (38), it is the Speaker that makes an inference about the Addressee's prior location: 8 (38) ¿Qué vienes, de la piscina?INTER come:2SG, from the swimming.pool'Are you coming from the swimming pool (I infer)?, lit.What are you coming, from the swimming pool?' Inferential evidentials in Bangla (Bhadra 2017(Bhadra , 2018(Bhadra , 2020) also seem to lack the Interrogative Flip, as the Speaker perspective is maintained in the inference, as seen in ( 39): (39) Mina amerika cho-e ja-cche naki?Mina America go-IMPERF go-3P.PRES.PROG NAKI '(Given what I inferred) Mina is going away to America (is it true)?' (Bhadra 2018) Both in Bangla and Spanish inferential interrogatives, the Addressee still has the information that will make the requested confirmation possible.This suggests that the In the system proposed by Speas and Tenny (2003) and Tenny and Speas (2004), the Seat of Knowledge, which Tenny and Speas (2004) situate within the Utterance Content, is an evidential argument that stands for 'the sentient individual who is responsible for the truth of a proposition' (Tenny and Speas 2004).In the case of interrogatives, both the Utterance Content and Seat of Knowledge are controlled by the raised Addressee.This explains why the Addressee is the discourse participant that possesses the knowledge to provide the answer to the question.For example, the Seat of Knowledge is named by evidential verbs like appear.Because of the Interrogative flip, appear will be anchored to the Speaker in a declarative, but to the Addressee in an interrogative, as seen in ( 37): (37) a.
Martin appears S to have missed his exam.(The speaker knows) b.
Does Martin appear A to have missed his exam?(The addressee knows) Inferential interrogatives with qué do not seem to pattern with prototypical interrogatives with evidentials in that they do not appear to present the Interrogative Flip, at least apparently, since the inference takes the Speaker's perspective rather than the Addressee's.For example, in a sentence such as (38), it is the Speaker that makes an inference about the Addressee's prior location: 8 (38) ¿Qué vienes, de la piscina?INTER come:2SG, from the swimming.pool'Are you coming from the swimming pool (I infer)?, lit.What are you coming, from the swimming pool?' Inferential evidentials in Bangla (Bhadra 2017(Bhadra , 2018(Bhadra , 2020) also seem to lack the Interrogative Flip, as the Speaker perspective is maintained in the inference, as seen in ( 39): (39) Mina amerika cho-e ja-cche naki?Mina America go-IMPERF go-3P.PRES.PROG NAKI '(Given what I inferred) Mina is going away to America (is it true)?' (Bhadra 2018) Both in Bangla and Spanish inferential interrogatives, the Addressee still has the information that will make the requested confirmation possible.This suggests that the Addressee does have a prominent role in the Speech Act projection, as a consequence of the Interrogative Flip.For this reason, the activation of the Interrogative Flip in inferential interrogatives with qué in Spanish is assumed in our proposal.

Evidential Projections
Four types of evidentials have been identified grammatically (Speas 2004).Different authors have proposed a compositional interpretation derived either from the different coindexations between participants and speech acts (Speas and Tenny 2003) or by means of four different projections (Cinque 1999;Speas 2004).Cinque (1999) proposes the hierarchy shown in (40) based both on the position evidential morphemes tend to exhibit within a word and adverb placement in a sentence.The Speech Act projection determines the type of Speech Act (e.g., interrogative), whereas the Evidential projection determines the source of the speaker's evidence of the truth of a proposition.Evaluative mood signals an evaluation made by the speaker and epistemological mode ascertains the degree in which a speaker is certain about a proposition (Cinque 1999;Speas 2004).In the case of morphemes, Cinque notes that speech act or speaker evaluation morphemes tend to appear further from the verb root than other morphemes, as shown in ( 41).
(41) Malagasi matetika > efa > mbola > V (O) > tsara > tanteraka > foana > intsony > ve generally already still well completely always anymore speech act (Cinque 1999, p. 43 [207]) As for adverbs, he argues that adverbs such as honestly and frankly are associated with the Speech Act projection, adverbs such as luckily are associated with the Evaluative projection, and adverbs such as obviously are associated with the Evidential projection.This has an impact on the linear order that these adverbs present in a clause, as seen in (42-44), whereby evidential adverbs follow both speech act and evaluative adverbs, but precede epistemological adverbs.
Honestly I am unfortunately unable to help you.b.
*Unfortunately I am honestly unable to help you.
Fortunately, he had evidently had his own opinion of the matter.b.
*Evidently he had fortunately had his own opinion of the matter.
Clearly John probably will quickly learn French perfectly.b.
*Probably John clearly will quickly learn French perfectly.(Cinque 1999, p. 33) Although Spanish word order restrictions generally differ from English, Spanish adverbs also appear to respond to Cinque's hierarchy in unmarked word order, at least regarding speech act adverbs and evidential adverbs as compared to other adverbs, as seen in ( 45 Different types of adverbs are oriented to different discourse participants in inferential interrogatives with qué.The Addressee, associated with the Speech Act, controls the Utterance Content and high adverbs such as francamente 'frankly,' honestamente 'honestly,' and sinceramente 'sincerely.'The Speaker, in turn, appears to be associated with the clausal level, controlling the reference of evidential adverbs such as obviamente 'obviously,' claramente 'clearly,' and evidentemente 'evidently.' In the sentence in (48), the adverb honestamente 'honestly' obligatorily appears as a high adverb, and it is associated with the Addressee in the interrogative; that is, it is the Addressee's honesty that is being requested.In the case of the evidential adverb evidentemente 'evidently' in (49), it is lower and controlled by the Speaker; that is, it is the Speaker that makes the inference rather than the Addressee, even if the sentence is interrogative.The contrast in the anchoring of the different adverbs with different clause participants in (48-49) is evidence of the activation of the interrogative flip in Spanish evidential interrogatives with qué, demonstrating the Addressee's raising to a higher position where it controls high adverbs.It also proves the existence of a clausal layer controlled by the Speaker.We explain how the Speaker's point of view becomes active in these constructions next.(2004) and Bianchi (2003Bianchi ( , 2006) ) propose the existence of logophoric pronouns in a clause, representing the discourse participants' point of view.According to Bianchi (2003), every finite clause is anchored to the time of utterance or speech event (S), whereas non-finite clauses need to be anchored to the main clause.The speech time (S) is located in Fin, in the left periphery, as shown in (50).Bianchi further proposes that only a finite construction encoding S may license person agreement, as it corresponds to a speech event, identified by Bianchi as the center of deixis, which includes the discourse participants as well as spatial and temporal coordinates determining finiteness.Sells (1987) distinguishes three distinct logophoric roles, which are relevant here, as they differentiate between different sources of information being reported: the source is the individual doing the reporting, the self is the individual whose mind is reported, and the physical point of view from which the report is made is the pivot.

Speas
Bianchi formally identifies the speech event as a Logophoric Center.Each Logophoric Center obligatorily projects an animate participant (i.e., the Speaker or Source), a typically optional Addressee in speech events, although the Addressee may be obligatory depending on the nature of the speech event (e.g., commands, questions).The Logophoric Center also contains temporal and spatial coordinates.This system is interesting because these logophoric pronouns can be coindexed, either with clausal arguments or with discursive participants, which affects phenomena such as agreement with discursive participants.For example, it would explain the agreement of the verbal features with the speaker together with a clausal argument in interrogative sentences such as (51), typical between a medical doctor and a patient (Jiménez-Fernández and Tubino-Blanco 2022).
(51) ¿Cómo estamos hoy? how be:1PL today 'How are we today?' The presence of the Logophoric Center controlling a Speech Act projection would also account for restrictions associated with the inclusive or exclusive interpretation of first-person plural pronouns and their interplay with information structure in Spanish.This fact was shown, for example, in Jiménez-Fernández and Tubino-Blanco (2022).Their system explains the restriction to use an overt first-person plural pronoun in an out-of-the-blue context, as seen in ( 52). (52) A: What's the plan?B: #Nosotros vamos a la playa (if intended as inclusive) 1PL go:1PL to the beach 'We're going to the beach' (Jiménez-Fernández and Tubino-Blanco 2022, p. 157 [38]) Blain and Déchaine (2007) argue that operators may enter a clause at different levels (ie., vP, AspP, AgreeP, CP), leading to different semantic consequences.At the CP level, evidentials affect the Speech Act; at the vP level, they introduce the Speaker's perspective in the predicate.In Speas' ( 2004) system, a logophoric pronoun may not be the same as the prominent participant in the matrix Speech Act.For example, in (53) the Speaker would be Evaluator but someone different would be Witness and Perceptor: In Bangla (Mukherjee 2008;Bhadra 2017Bhadra , 2018Bhadra , 2020)), the evidential morpheme naki changes its evidential contribution depending on the Speech Act in which it appears.If the morpheme appears in a declarative sentence and middle position, the morpheme has a reportative interpretation, but if it appears in an interrogative sentence and final position, it has an inferential interpretation.In this sense, it is similar to inferential interrogatives with qué.In fact, Mukherjee (2008) proposes that in the case of interrogatives, naki is a confirmation particle, an interpretation that has been previously proposed also for the tag portion in Spanish inferential interrogatives (López-Cortina 2003, 2009).Bhadra nonetheless proposes that, in both cases, naki is an indirect evidential and its interpretation is compositional.We see examples in (54):  Bhadra (2018) proposes two additional coordinates with respect to the ones proposed by Bianchi, which correspond with the discourse participants.These are different from the participants in Speas and Tenny's Speech Act projection.Bhadra proposes that either within or below the SaP projection we can find a FinP projection that accounts for the finite clause's point of view, which may correspond or not to the reference of discourse participants.This projection can be integrated either within SaP or below.(55) an interpretation that has been previously proposed also for the tag portion in Spanish inferential interrogatives (López-Cortina 2003, 2009).Bhadra nonetheless proposes that, in both cases, naki is an indirect evidential and its interpretation is compositional.We see examples in ( 54): (54) Mukherjee (2008) a. Shila naki gaan Sikh-ch-e Shila INFER song learn-PROG-3 'Shila is learning music, as I have heard' b.Sita baRi giy-ech-e naki?Sita home go-PERF-3 CONFIR '(From what I infer) Sita has gone home.Has she?' Bhadra (2018) proposes two additional coordinates with respect to the ones proposed by Bianchi, which correspond with the discourse participants.These are different from the participants in Speas and Tenny's Speech Act projection.Bhadra proposes that either within or below the SaP projection we can find a FinP projection that accounts for the finite clause's point of view, which may correspond or not to the reference of discourse participants.This projection can be integrated either within SaP or below.
(55) Keeping all these pieces in mind, we proceed to our analysis, which is spelled out in the next section.

A Formal Proposal for Inferential Interrogatives with qué
Our proposal is based on the following premises: (i) The Addressee is located in a higher position within the Speech Act Phrase, which also explains the interpretation of the construction as an interrogative Speech Act (Miyagawa 2022).In this sense, it is the Addressee who controls the Utterance Content and Seat of Knowledge positions (Speas and Tenny 2003).(ii) The Speaker maintains its coindexation with the evidential material by means of coindexation with FinP logophoric projection, where the confirmation proposition is located (Bhadra 2018).
This explains why these sentences appear to be hybrid, in the sense that they appear to exhibit both partial and total interrogative behavior.It also explains why both Speaker and Addressee perspectives are present in the clause.
Evidence in favor of this hybrid structure is that the speech act adverb honestamente 'honestly' in ( 48) is anchored to the addressee, whereas the evidential adverb evidentemente 'evidently' is anchored to the speaker in (49).This leads us to propose two contrasted levels in the internal composition of these sentences, The Utterance level above is configured as an interrogative Speech Act, where the Addressee is the prominent role as a consequence of the Interrogative Flip, and where Speech Act adverbs are licensed.Below this level the clausal level can be found, where evidential adverbs are licensed and controlled by the Speaker, which represents this level's point of view (see Kim 2012 for a similar view Keeping all these pieces in mind, we proceed to our analysis, which is spelled out in the next section.

A Formal Proposal for Inferential Interrogatives with qué
Our proposal is based on the following premises: (i) The Addressee is located in a higher position within the Speech Act Phrase, which also explains the interpretation of the construction as an interrogative Speech Act (Miyagawa 2022).In this sense, it is the Addressee who controls the Utterance Content and Seat of Knowledge positions (Speas and Tenny 2003).(ii) The Speaker maintains its coindexation with the evidential material by means of coindexation with FinP logophoric projection, where the confirmation proposition is located (Bhadra 2018).
This explains why these sentences appear to be hybrid, in the sense that they appear to exhibit both partial and total interrogative behavior.It also explains why both Speaker and Addressee perspectives are present in the clause.
Evidence in favor of this hybrid structure is that the speech act adverb honestamente 'honestly' in ( 48) is anchored to the addressee, whereas the evidential adverb evidentemente 'evidently' is anchored to the speaker in (49).This leads us to propose two contrasted levels in the internal composition of these sentences, The Utterance level above is configured as an interrogative Speech Act, where the Addressee is the prominent role as a consequence of the Interrogative Flip, and where Speech Act adverbs are licensed.Below this level the clausal level can be found, where evidential adverbs are licensed and controlled by the Speaker, which represents this level's point of view (see Kim 2012 for a similar view and further argumentation on the anchoring of different types of adverbs by distinct discourse roles).
This configuration results in evidential interrogatives with qué, where we simultaneously find evidential material that houses the Speaker's point of view and classic interrogative material with the Addressee as the Seat of Knowledge, manifesting as the request for confirmation of the interrogative's truth value.To account for this configuration, we propose the construction in (56b) based on the sentence in (56a):

course roles).
This configuration results in evidential interrogatives with qué, where we simultaneously find evidential material that houses the Speaker's point of view and classic interrogative material with the Addressee as the Seat of Knowledge, manifesting as the request for confirmation of the interrogative's truth value.To account for this configuration, we propose the construction in (56b) based on the sentence in (56a): In the configuration in (56), the interrogative operator and the inferred material are base-generated in a Small Clause (SC) in vP.This accounts for the connection between the operator and the tag, given that the operator is clearly requesting information about the content of the tag.The inferential material remains in a lower position, which is typical of foci in Spanish.Both the evidential and discursive component of the structure take place at the CP layer, where the point of view and speech act relations take place.These evidential interrogatives are characterized by the fact that the evidential material is oriented to the Speaker rather than the Addressee.This is possible since it is the Speaker that occupies the [Spec, FinP] position, hence controlling the point of view of the material within the In the configuration in (56), the interrogative operator and the inferred material are base-generated in a Small Clause (SC) in vP.This accounts for the connection between the operator and the tag, given that the operator is clearly requesting information about the content of the tag.The inferential material remains in a lower position, which is typical of foci in Spanish.Both the evidential and discursive component of the structure take place at the CP layer, where the point of view and speech act relations take place.These evidential interrogatives are characterized by the fact that the evidential material is oriented to the Speaker rather than the Addressee.This is possible since it is the Speaker that occupies the [Spec, FinP] position, hence controlling the point of view of the material within the clause, including the tag (i.e., the SC).At the Speech act level, the interrogative character of the construction has additional consequences.The SC has a focus feature, which results in the movement of the operator qué to the [Spec, ForceP] position, passing through [Spec, FocP], since this ForceP projection has an Edge Feature (i.e., EPP in interrogatives).The reason why the SC has a focus feature is because one of its members (the tag) will end up being interpreted as the focus.The verb in turn moves to T and then to the ForceP head.At the discursive level, the Addressee takes a prominent place to reflect the interrogative character of the construction, moving from its neutral position in the complement of SaP to The sentence in (66) shows a canonical yes/no interrogative whereby the Speaker asks some specific information that the Addressee is expected to provide.In (67), the Speaker asks a question the answer to which she already knows and is ready to provide.In (68), the interrogative is an echo-question.According to Escandell-Vidal, the canonical low-rise contour is the consequence of unspecified sentence polarity corresponding to the wh-operator, but the marked contours indicate the presence of evidential material in the sentence, indicating the information source.In her proposal, in the case of high-rise contours, the source of information would be the Self, that is, the evidential would be controlled by the Speaker.In the case of rise-fall contours, the information source would be the Other, that is, the evidential is addressee-oriented, according to Escandell-Vidal.
Although Escandell-Vidal does not discuss inferential interrogatives with qué, the marked fall-rise contour associated with these interrogatives is consistent with her proposal that marked intonation contours suggest the presence of evidential material.In this case, the source of information is hybrid, whereby the 'Self' (the Speaker) does the inferring, and the 'Other' (the Addressee) possesses the Seat of Knowledge.
For Escandell-Vidal (2017), the interrogative feature introduces a set of propositions {p, ~p}, where only one proposition is true, over which the evidential operates.The fallrise contour associated with inferential interrogatives with qué indicates that the Seat of Knowledge is the 'Other', with scope over the true option and marked with the fall part of the contour.The inferential evidentiality part, associated with the 'Self' and rising intonation, operates over a hypothesis, which is explicit in the question.
Moreover, the intonational contour associated with these constructions is further evidence that they are monoclausal rather than biclausal constructions and that the tag is fully integrated within the interrogative, forming one prosodic unit, as also argued by Fernández-Sánchez and García-Pardo (2020b).Proof of this is that the tag cannot be an independent sentence, as shown also by Wiltschko and Heim (2016)

Conclusions
Inferential interrogatives with qué are monoclausal interrogatives in which the interrogative operator moves to the CP area to validate [wh], [foc], and [EF] features.The interrogative operator is base-generated in a Small Clause within vP.The inference remains at the clausal level and takes the Speaker's perspective, thanks to the presence of a Speaker-oriented logophoric pronoun within FinP.This is not incompatible with the availability of the Interrogative Flip in these questions, which is addressee-oriented and occurs in the left-most left periphery.The consequence at the interpretive level is an unconventional interrogative clause with speaker-oriented evidential interpretation, the An ellipsis-based analysis is also proposed for Split Interrogatives by Arregi (2010).Recall that these interrogatives as similar to the inferential interrogatives studied here, but crucially the operator is a full wh-phrase in the former and the neuter wh-word qué in the latter: The ellipsis analysis is provided in (ii): (ii) [CP what tree i planted Juan t i ] [CP an oak j planted Juan t j ] In the second CP the verb and the subject are deleted, leaving only the preposed element which stands for the tag in the outcome.Speas and Tenny's original idea of having discourse categories such as SaP in the syntactic tree has been developed in Haegeman and Hill (2013); Miyagawa (2012Miyagawa ( , 2017Miyagawa ( , 2022)); Portner et al. (2019); or Wiltschko (2014Wiltschko ( , 2017Wiltschko ( , 2021Wiltschko ( , 2022)), among others.

8
As a reviewer rightly points out, at least in some instances the addressee themselves provide certain types of evidence that facilitate the inference (e.g., the way they are dressed, wet hair, the addressee's routine, etc.).It is important to clarify that in these inferential interrogatives, the inference relies on contextual information that is inferred by the speaker.While contextual, this information is crucially not offered by the addressee, at least linguistically, and even if inferred by the speaker on the basis of the addressee's physical attributes, behavior, or habits known by the speaker, the inference still shows the speaker's perspective.
It is important to distinguish between the inference made by the speaker (which is strictly the inference based on contextual information based on what the speaker observes about the addressee or any other kind of contextual information) and the actual information the addressee knows, which is what is requested by means of the confirmation component of these sentences.
coming from the supermarket (I infer)?, Lit.what have you gone, to the supermarket?' ¿Vienes de la calle?(7) ¿De dónde vienes?A similar, but not identical, interrogative pattern has been found in Catalan (Contreras and Roca 2007) and English (López-Cortina 2009), as we indicated earlier: you do it, in the oven?' (Contreras and Roca 2007, p. 145 [1]) (9) a.What are you, crazy?b.What is he, your lawyer?Compare with the rising intonation for the yes/no question in (6) and the falling intonation for the wh-question in (7); see Hualde (2005):(6) ¿Vienes de la calle?
'Where does he want to eat it in the bed?' b.
all come by train (I infer) lit.what do you all come, by train?' b.

, as we indicated earlier:
A similar,