1 Introduction

Much of the discussion in the ellipsis literature has centered around answering the identity question: that is, what is the relationship between the omitted material and its antecedent, and what type of identity is required for ellipsis to be licensed? Prior work on ellipsis (see Fiengo & May 1994, Chung et al. 1995, Lobeck 1995, Merchant 2001, Saab 2008, Aelbrecht 2010, Chung 2013, Merchant 2013, Barros 2014, Merchant 2019, Ranero 2021, Poppels 2022, and Stigliano 2022, among many others) has worked under the assumption that there should be a universal identity condition that applies to all types of ellipsis in all languages. However, recent research has delved deeply into possible and impossible mismatches between the ellipsis site and its antecedent, revealing that some mismatches can be accounted for if a more restrictive identity condition only applies to a certain portion of the structure. For instance, Kroll & Rudin 2017 and Rudin 2019 propose that only the eventive core (i.e., the vP and its constituent heads) must satisfy the requirement of syntactic identity, while the remaining elements are evaluated in terms of pragmatics. Similarly, Anand et al. 2023 enforces identity over the argument domain (i.e., an extension of the eventive core that encompasses any thematic domain). Elements outside of this domain are not required to conform to syntactic identity and are instead judged based on pragmatic and semantic calculations. Despite these recent developments, the challenge of allowing certain mismatches while maintaining identity between the ellipsis site and its antecedent has yet to be fully resolved.

The present study aims to extend this line of inquiry by proposing that certain types of ellipsis involve mixed-identity requirements. In particular, I argue that a strict syntactic identity requirement is imposed on one portion of the structure, while no identity requirements at all are imposed on a different portion of the structure. This main argument is supported by a detailed examination of an understudied empirical domain of ellipsis in Spanish: an elliptical construction that I dub topic-remnant elided wh-questions (or wh-TREQs, for short). Wh-TREQs, exemplified in B’s response to A in (1) and (2), are elliptical questions interpreted as follow-up wh-questions, as shown by A’s subsequent answer to B.1,2,3

    1. (1)
    1. A:
    1. Sonia
    2. Sonia
    1. comió
    2. ate
    1. pizza.
    2. pizza
    1. ‘Sonia ate pizza.’
    1.  
    1. B:
    1. Y
    2. and
    1. Bruno?
    2. Bruno
    1. Literal: ‘And Bruno?’
    2. Interpretation: ‘What about Bruno? What did he eat?’
    1.  
    1. A:
    1. Pasta.
    2. pasta
    1. ‘Pasta.’
    1. (2)
    1. A:
    1. Sonia
    2. Sonia
    1. comió
    2. ate
    1. pizza.
    2. pizza
    1. ‘Sonia ate pizza.’
    1.  
    1. B:
    1. Y
    2. and
    1. pasta?
    2. pasta
    1. Literal: ‘And pasta?’
    2. Interpretation: ‘What about pasta? Who ate that?’
    1.  
    1. A:
    1. Bruno.
    2. Bruno
    1. ‘Bruno.’

It’s important to note that, in general, topic-remnant elided questions could be ambiguous between a wh-question interpretation (i.e., what I call wh-TREQs) and a polar-question interpretation (i.e., polar TREQs), as exemplified in (3) and (4). Usually, context disambiguates between the two interpretations.

    1. (3)
    1. A:
    1. Sonia
    2. Sonia
    1. comió
    2. ate
    1. pizza.
    2. pizza
    1. ‘Sonia ate pizza.’
    1.  
    1. B:
    1. Y
    2. and
    1. Bruno?
    2. Bruno
    1. Literal: ‘And Bruno?’
    2. Interpretation: ‘What about Bruno? Did he also eat pizza?’
    1.  
    1. A:
    1. No.
    2. no
    1. (Está
    2.   he.is
    1. a
    2. on
    1. dieta.)
    2. diet
    1. ‘No. (He is on a diet.)’
    1. (4)
    1. A:
    1. Sonia
    2. Sonia
    1. comió
    2. ate
    1. pizza.
    2. pizza
    1. ‘Sonia ate pizza.’
    1.  
    1. B:
    1. Y
    2. and
    1. pasta?
    2. pasta
    1. Literal: ‘And pasta?’
    2. Interpretation: ‘What about pasta? Did she also eat that?’
    1.  
    1. A:
    1. No.
    2. no
    1. (Solo
    2.   only
    1. comió
    2. he.ate
    1. pizza.)
    2. pizza
    1. ‘No. (He only ate pizza.)’

An analysis of polar TREQs is outside of the scope of this article, and I refer the reader to Stigliano 2022 and 2023 for a thorough examination of this construction.

Examples such as those in (1) and (2) provide important insights into the study of ellipsis; though similar constructions have been examined before in languages like Mandarin Chinese, Korean, and Japanese, they have not been explored in such a systematic way (see section 2 for an overview of previous works). In this article, I analyze wh-TREQs as the result of the ellipsis of a wh-question, from which a topic has moved out, surviving deletion. Furthermore, I show that syntactic isomorphism between the antecedent and the ellipsis site is needed by providing evidence from connectivity effects (such as the impossibility of voice mismatches, of P-stranding under sluicing, and of ‘spray’–‘load’ alternations) and from the fact that wh-TREQs are island sensitive. However, I also show that a syntactic-identity condition is too strict and fails to account for the basic fact that the ellipsis site contains a wh-phrase, which gets deleted although it doesn’t have an identical correlate in the antecedent. In the case of (1), the wh-phrase is qué ‘what,’ and the correlate would be the NP pizza:

    1. (5)

To account for these seemingly conflicting identity requirements, I argue that wh-TREQs arise as the result of two ellipses, one triggered by an [E]-feature on C, which imposes syntactic identity over its complement (namely, the TP), and one triggered by an [E]-feature on Top, which doesn’t impose any identity requirements at all and deletes what’s remaining in its complement (namely, the wh-phrase). This is schematized in (6) for the wh-TREQ in (5); gray text represents elided material.

    1. (6)

The structure of this article is as follows. Section 2 provides an overview of previous work that has analyzed phenomena similar to what I call TREQs. Section 3 examines the syntactic distribution of wh-TREQs in Spanish and presents evidence supporting (i) an ellipsis approach to this construction, (ii) a move-and-delete analysis of the remnant, (iii) the claim that there is movement of the wh-phrase inside the ellipsis site, and (iv) the need for a syntactic-identity condition to license ellipsis. In section 4, I discuss the identity condition needed to license ellipsis in wh-TREQs: I argue that wh-TREQs are licensed under mixed-identity conditions, and I implement this by proposing that wh-TREQs are actually the result of two ellipses, triggered by two distinct [E]-features. Finally, section 5 concludes, summarizes the findings, and posits some open questions. Two appendices are included in a supplementary file: appendix A complements the data presented in section 3 regarding wh-TREQs and islands, and appendix B discusses differences and similarities between Spanish wh-TREQs and English what-about questions.

2 Previous works and background

In this section, I offer a brief overview of the previous literature that addresses constructions similar to what I refer to as TREQs. As I will show, there are significant differences between the phenomena discussed in these works and TREQs in Spanish. Crucially, none of the existing literature provides a detailed and unified analysis of TREQs, leaving it unable to account for the full range of properties exhibited by this construction. I want to underscore the importance of the work reviewed here as a foundation for the investigation I undertake in this article, which not only seeks to provide a comprehensive understanding of TREQs but also, more importantly, aims to put forth a particular theory of ellipsis licensing and identity.

The term fragment questions has been used to refer to a construction similar to what I call polar TREQs in Spanish (Stigliano 2022, 2023). According to Wei 2013, fragment questions in Mandarin Chinese, like the one in (7B), are interpreted as polar/A-not-A questions.

    1. (7)
    1. A:
    1. Zhangsan
    2. Zhangsan
    1. huilai
    2. back
    1. le.
    2. le
    1. ‘Zhangsan has already come back.’
    1.  
    1. B:
    1. Lisi
    2. Lisi
    1. ne?
    2. ne
    1. Interpretation: ‘What about Lisi? Did she come back or not?’
    1.  
    1. (Adapted from Wei 2013: (3))

Wei extends Merchant 2004’s proposal for fragment answers in English and analyzes fragment questions in Mandarin Chinese as involving focus movement of the remnant, followed by TP ellipsis. That is, fragment questions in Mandarin Chinese are considered a type of TP ellipsis, along the lines of sluicing, fragment answers, and so on. In contrast, I propose here that wh-TREQs in Spanish are a type of CP ellipsis. What is more, Wei claims that fragment questions do not require syntactic identity to be licensed but only semantic identity, despite the fact that fragment questions in this language display connectivity effects and are island sensitive.

Wei 2018 revises the proposal of Wei 2013 and argues that fragment questions are derived via topic movement (not focus movement) and TP ellipsis. According to Wei, the remnant is a topic-like constituent, the specifier of TopP, followed by a particle ne, which functions simultaneously as a topic marker and a constituent-question particle (for more details on this analysis, see Wei 2018). My analysis regarding the topicalization of the remnant is in line with Wei’s approach.

Li 2016 also analyzes fragment questions in Mandarin Chinese. However, Li’s fragment questions differ from Wei’s, as can be seen in (8B2). The crucial difference between Wei’s and Li’s fragment questions is that the latter have an explicit wh-question as their antecedent, as in (8A1).4

    1. (8)
    1. A1:
    1. Libai
    2. Libai
    1. he-le
    2. drink-asp
    1. shenme?
    2. what
    1. ‘What did Libai drink?’
    1.  
    1. B1:
    1. Hongjiu.
    2. red.wine
    1. ‘Red wine.’
    1.  
    1. A2:
    1. Dufu
    2. Dufu
    1. ne?
    2. ne
    1. Interpretation: ‘What did Dufu drink?’
    1.  
    1. B2:
    1. Baijiu.
    2. white.wine
    1. ‘White wine.’
    1.  
    1. (Adapted from Li 2016: (5))

Li proposes that fragment questions in Mandarin Chinese of the type exemplified in (8) are derived from full wh-questions, with ellipsis targeting the entire question, minus the remnant, which is a contrastive topic. On the surface, this is similar to my proposal for wh-TREQs in Spanish. However, Li argues that the interpretation of fragment questions is context dependent: they can be interpreted as wh-questions, alternative questions, A-not-A questions, and yes–no questions, depending on the question that precedes them in a dialogue. In contrast, wh-TREQs in Spanish do not require such an explicit question in the context. The following Mandarin Chinese example illustrates a fragment question interpreted as a polar question due to the presence of a polar question.5

    1. (9)
    1. A1:
    1. Libai
    2. Libai
    1. laizi
    2. come.from
    1. Shaanxi
    2. Shaanxi
    1. ma?
    2. q
    1. ‘Is Libai from Shaanxi?’
    1.  
    1. B1:
    1. Shi
    2. yes
    1. a.
    2. sfp
    1. ‘Yes, he is.’
    1.  
    1. A2:
    1. Dufu
    2. Dufu
    1. ne?
    2. ne
    1. Interpretation: ‘Is Dufu from Shaanxi?’
    1.  
    1. B2:
    1. Ye
    2. too
    1. shi.
    2. yes
    1. ‘Yes, he is too.’
    1.  
    1. (Adapted from Li 2016: (11))

The author argues that ellipsis is licensed under semantic identity, more specifically, under a focus condition on ellipsis based on Rooth 1992 (for more details, see Li 2016). This contrasts with my proposal for wh-TREQs and, more broadly, with the theory of ellipsis identity I advocate.

It should be noted that while Li’s argument primarily focuses on the presence of an antecedent question for the analyzed construction, the author briefly notes that fragment questions can occur without an antecedent question:

    1. (10)
    1. A:
    1. Libai
    2. Libai
    1. he-le
    2. drink-asp
    1. kafei.
    2. coffee
    1. ‘Libai drank coffee.’
    1.  
    1. B:
    1. Na,
    2. then
    1. Dufu
    2. Dufu
    1. ne?
    2. ne
    1. ‘What about Dufu?’
    1.  
    1. (Adapted from Li 2016: (35))

These “antecedent-less” fragment questions can receive different interpretations. Although Li doesn’t provide a full account of these cases, he suggests that they are licensed by an implicit antecedent question (i.e., a question under discussion), from which they take the relevant meaning. There is no further analysis of the syntax or the distribution of this construction, nor is it clear what the predictions of this analysis would be. If it were extended to wh-TREQs in Spanish, it would predict incorrect patterns, given that it wouldn’t be able to account for the connectivity effects (such as voice mismatches and ‘spray’–‘load’ alternations) or the island-sensitivity effects found in this construction, as I will show in the rest of this article.

Kim 2021 examines fragment questions in Korean. These are similar to Li’s fragment questions in that they do have a wh-question in their antecedent. Kim proposes a direct-interpretation (i.e., non-sententialist) approach to this construction, contra, for example, Wei 2013 and 2018. Despite the fact that fragment questions in Korean display connectivity effects such as case-matching effects and are island sensitive, Kim still argues for a non-sententialist approach based on some apparent case mismatches and some contexts of island insensitivity. Although I don’t have an explanation for how these mismatches arise, I take the evidence for a sententialist account to be too overwhelming to even consider the alternative non-sententialist approach viable.

Finally, Hiraiwa & Kobayashi 2020 discusses what it calls countersluicing in Japanese, an elliptical phenomenon in which everything but a wh-phrase (and what follows it) survives deletion:

    1. (11)
    1. A:
    1. Arne
    2. Arne
    1. Jacobsen-ga
    2. Jacobsen
    1. dezainsita
    2. designed
    1. no-wa
    2. comp-top
    1. Ant
    2. Ant
    1. Chair
    2. Chair
    1. desu.
    2. cop
    1. ‘It was Ant Chair that Arne Jacobsen designed.’
    1.  
    1. B:
    1. Hans
    2. Hans
    1. J.
    2. J.
    1. Wegner-ga
    2. Wegner-nom
    1. dezainsita
    2. designed
    1. no-wa
    2. comp-top
    1. nan-toiu isu-o  (desu-ka)?
    2. what-called chair-acc cop-q
    1. Interpretation: ‘What chair was it that Hans J. Wegner designed?’
    1.  
    1. A:
    1. The
    2. The
    1. Chair
    2. Chair
    1. desu.
    2. cop
    1. ‘It was The Chair.’
    1.  
    1. (Adapted from Hiraiwa & Kobayashi 2020: (5))

Hiraiwa & Kobayashi argue that both countersluicing and sluicing in Japanese are derived from a copular source. This is a key difference from wh-TREQs in Spanish: the fact that isomorphism is needed for wh-TREQs in Spanish plays a crucial role for the type of analysis I propose for this construction. Furthermore, according to the authors, sluicing results from applying Argument Ellipsis to FinP, whereas countersluicing involves applying Argument Ellipsis to ForceP. This contrasts with the analysis proposed here for wh-TREQs and with the commonly assumed analyses for sluicing in Spanish. Additionally, it is worth noting that countersluicing in Japanese does not allow polar-question readings (unlike fragment questions in Mandarin Chinese and polar TREQs in Spanish) and is restricted to matrix clauses, which contrasts with the existence of embedded wh-TREQs (see Stigliano 2022).

To sum up, the phenomenon I analyze in this article—wh-TREQs in Spanish—shares similarities with fragment questions (of different types) in Mandarin Chinese and Korean and with countersluicing in Japanese. Some of the ingredients of the analysis I propose here, such as the claim that the remnant is a topicalized XP, have been proposed in some of the works cited above. However, there are crucial differences between wh-TREQs and the other phenomena discussed in this section. Firstly, fragment questions with a wh-question meaning in Mandarin Chinese and Korean are typically associated with an explicit antecedent wh-question. The cases lacking an antecedent wh-question are not thoroughly examined. The fact that a wh-phrase can be elided even when there is no wh-phrase or indefinite in its antecedent is a crucial property of wh-TREQs. Moreover, the analyses of fragment questions often rely on (different flavors of) semantic identity, which permits structural mismatches between the antecedent and the ellipsis site. As I will show, (some type of) syntactic identity/isomorphism is indeed needed to license ellipsis in contexts of wh-TREQs. It is crucial to reconcile this claim (and the facts that motivated it) with the empirical domain of wh-TREQs, which suggests that this syntactic-identity requirement needs to be weakened. I will achieve this by proposing that wh-TREQs are the result of two ellipses, each imposing different identity conditions—a new perspective that has not been previously discussed.

3 The distribution of wh-TREQs in Spanish

In this section I examine evidence in favor of an ellipsis approach to wh-TREQs that involves topicalization of the remnant outside the ellipsis site and wh-movement of the wh-phrase inside the ellipsis site. First, I show that there is a strict correlation between the contexts in which topicalization is allowed out of wh-questions in Spanish and the contexts that allow wh-TREQs. In particular, possible remnants for a wh-TREQ in Spanish are exactly those constituents that can otherwise be topicalized out of a wh-question (section 3.1). Likewise, those constituents that cannot be topicalized out of a wh-question cannot be remnants of wh-TREQs (section 3.2). Second, in section 3.4 I discuss the island-sensitive nature of wh-TREQs; specifically, I argue that the patterns found with respect to islands provide evidence for the claim that there is wh-movement inside the ellipsis site and that the remnant is a topicalized XP that moves out of the ellipsis site. Finally, in section 3.5, I show that wh-TREQs display connectivity effects, which provides further evidence of the need for (at least some type of) syntactic identity to license this construction.

3.1 Possible remnants

Possible remnants for wh-TREQs in Spanish are exactly those constituents that can otherwise be topicalized out of a wh-question in this language. This shows that there is indeed structure inside the ellipsis site and that the remnant moves out of it. This argument will be complemented with the analysis of those constituents that cannot be topicalized out of wh-questions and that, as expected, cannot occur as wh-TREQs (see section 3.2). In short, possible remnants include subjects and objects (both direct and indirect), prepositional phrases (both in the verbal and nominal domains), temporal and locative phrases, frequency and manner adverbs, predicative adjectives, and infinitival phrases. In what follows, I provide examples of each of them.

3.1.1 Subjects, direct objects, and indirect objects

As shown in (12), subjects can be topicalized out of wh-questions. As expected, they can also be remnants for wh-TREQs, as (13) shows.

    1. (12)
    1. (13)
    1. A:
    1. Sonia
    2. Sonia
    1. comió
    2. ate
    1. pizza.
    2. pizza
    1. ‘Sonia ate pizza.’
    1.  
    1. B:
    1. Y
    2. and
    1. Bruno?
    2. Bruno
    1. Literal: ‘And Bruno?’
    2. Interpretation: ‘What about Bruno? What did he eat?’

Likewise, the example in (14) shows that direct objects can be topicalized out of a wh-question, and the example in (15) shows that, as expected, they can be remnants for wh-TREQs.

    1. (14)
    1. (15)
    1. A:
    1. Sonia
    2. Sonia
    1. comió
    2. ate
    1. pizza.
    2. pizza
    1. ‘Sonia ate pizza.’
    1.  
    1. B:
    1. Y
    2. and
    1. pasta?
    2. pasta
    1. Literal: ‘And pasta?’
    2. Interpretation: ‘What about pasta? Who ate that?’

Finally, (16) shows that indirect objects can be topicalized out of a wh-question. They can be remnants for wh-TREQs as well, as (17) shows.

    1. (16)
    1. (17)
    1. A:
    1. Sonia
    2. Sonia
    1. le
    2. to.him
    1. dio
    2. gave
    1. una
    2. a
    1. pizza
    2. pizza
    1. a
    2. to
    1. Luciano.
    2. Luciano
    1. ‘Sonia gave a pizza to Luciano.’
    1.  
    1. B:
    1. Y
    2. and
    1. a
    2. to
    1. Bruno?
    2. Bruno
    1. Literal: ‘And to Bruno?’
    2. Interpretation: ‘What about Bruno? Who gave a pizza to him?’

3.1.2 Prepositional phrases

The following examples show that PPs can be topicalized out of a wh-question. This holds for PPs in the verbal domain, as in (18), and for PPs in the nominal domain, as in (19).

    1. (18)
    1. (19)

As expected, PPs can be remnants for wh-TREQs:

    1. (20)
    1. A:
    1. Sonia
    2. Sonia
    1. habló
    2. talked
    1. con
    2. with
    1. Bruno.
    2. Bruno
    1. ‘Sonia talked with Bruno.’
    1.  
    1. B:
    1. Y
    2. and
    1. con
    2. with
    1. Luciano?
    2. Luciano
    1. Literal: ‘And with Luciano?’
    2. Interpretation: ‘What about Luciano? Who talked with him?’
    1. (21)
    1. A:
    1. Sonia
    2. Sonia
    1. vio
    2. saw
    1. una
    2. a
    1. foto
    2. picture
    1. de
    2. of
    1. Bruno.
    2. Bruno
    1. ‘Sonia saw a picture of Bruno.’
    1.  
    1. B:
    1. Y
    2. and
    1. de
    2. of
    1. Luciano?
    2. Luciano
    1. Literal: ‘And of Luciano?’
    2. Interpretation: ‘What about Luciano? Who saw a picture of him?’

3.1.3 Temporal and locative phrases

As for temporal and locative phrases, the examples in (22) and (23) show that they can be topicalized out of wh-questions.

    1. (22)
    1. (23)

The examples in (24) and (25) show that, as expected, they can be remnants for wh-TREQs.

    1. (24)
    1. A:
    1. Sonia
    2. Sonia
    1. corrió
    2. ran
    1. el
    2. the
    1. martes.
    2. Tuesday
    1. ‘Sonia ran on Tuesday.’
    1.  
    1. B:
    1. Y
    2. and
    1. el
    2. the
    1. viernes?
    2. Friday
    1. Literal: ‘And on Friday?’
    2. Interpretation: ‘What about on Friday? Who ran that day?’
    1. (25)
    1. A:
    1. Sonia
    2. Sonia
    1. corrió
    2. ran
    1. en
    2. in
    1. el
    2. the
    1. parque.
    2. park
    1. ‘Sonia ran in the park.’
    1.  
    1. B:
    1. Y
    2. and
    1. en
    2. on
    1. la
    2. the
    1. playa?
    2. beach
    1. Literal: ‘And on the beach?’
    2. Interpretation: ‘What about on the beach? Who ran there?’

3.1.4 Adverbs

Similarly, frequency, manner, and temporal adverbs can be topicalized out of a wh-question:

    1. (26)
    1. (27)
    1. (28)

Likewise, examples (29), (30), and (31) show that they can also be remnants for wh-TREQs.

    1. (29)
    1. A:
    1. Sonia
    2. Sonia
    1. corre
    2. runs
    1. siempre.
    2. always
    1. ‘Sonia always runs.’
    1.  
    1. B:
    1. Y
    2. and
    1. ocasionalmente?
    2. occasionally
    1. Literal: ‘And occasionally?’
    2. Interpretation: ‘What about occasionally? Who runs occasionally?’
    1. (30)
    1. A:
    1. Sonia
    2. Sonia
    1. corre
    2. runs
    1. despacio.
    2. slow
    1. ‘Sonia runs slow.’
    1.  
    1. B:
    1. Y
    2. and
    1. rápido?
    2. fast
    1. Literal: ‘And fast?’
    2. Interpretation: ‘What about fast? Who runs fast?’
    1. (31)
    1. A:
    1. Sonia
    2. Sonia
    1. corrió
    2. ran
    1. antes.
    2. before
    1. ‘Sonia ran before.’
    1.  
    1. B:
    1. Y
    2. and
    1. después?
    2. after
    1. Literal: ‘And after?’
    2. Interpretation: ‘What about after? Who ran after?’

3.1.5 Predicative adjectives

As shown in (32), predicative adjectives can also be topicalized out of a wh-question. As expected, these adjectives can also be remnants for wh-TREQs, as (33) shows.

    1. (32)
    1. (33)
    1. A:
    1. Sonia
    2. Sonia
    1. pintó
    2. painted
    1. el
    2. the
    1. auto
    2. car
    1. rojo.
    2. red
    1. ‘Sonia painted the car red.’
    1.  
    1. B:
    1. Y
    2. and
    1. azul?
    2. blue
    1. Literal: ‘And blue?’
    2. Interpretation: ‘What about blue? Who painted the car blue?’

3.1.6 Infinitival phrases

Bare infinitives can be topicalized out of wh-questions and be remnants for wh-TREQs:

    1. (34)
    1. (35)
    1. A:
    1. Sonia
    2. Sonia
    1. quiere
    2. wants
    1. cocinar.
    2. to.cook
    1. ‘Sonia wants to cook.’
    1.  
    1. B:
    1. Y
    2. and
    1. comer?
    2. to.eat
    1. Literal: ‘And to eat?’
    2. Interpretation: ‘What about eating? Who wants to eat?’

Additionally, (36) and (37) show that infinitival phrases, too, can be topicalized out of wh-questions and be remnants for wh-TREQs.

    1. (36)
    1. (37)
    1. A:
    1. Sonia
    2. Sonia
    1. quiere
    2. wants
    1. viajar
    2. to.travel
    1. a
    2. to
    1. Chicago.
    2. Chicago
    1. ‘Sonia wants to travel to Chicago.’
    1.  
    1. B:
    1. Y
    2. and
    1. comprar
    2. to.buy
    1. un
    2. a
    1. auto?
    2. car
    1. Literal: ‘And to buy a car?’
    2. Interpretation: ‘What about buying a car? Who wants to buy a car?’

3.2 Impossible remnants

Impossible remnants for wh-TREQs in Spanish are exactly those constituents that cannot be topicalized out of wh-questions in this language. Again, I take this to show that there is indeed structure inside the ellipsis site and that the remnant must be topicalized out of it. This argument complements what I just presented in section 3.1. In short, impossible remnants include TPs, sentential adverbs, and attributive adjectives.

3.2.1 TPs

As (38) shows, TPs cannot be topicalized out of wh-questions. As predicted, they cannot be remnants for wh-TREQs either, as shown in (39).

    1. (38)
    1. (39)
    1. A:
    1.   Sonia
    2.   Sonia
    1. viajó
    2. traveled
    1. a
    2. to
    1. Chicago.
    2. Chicago
    1.   ‘Sonia traveled to Chicago.’
    1.  
    1. B:
    1. *Y
    2.   and
    1. compró
    2. bought
    1. un
    2. a
    1. auto?
    2. car
    1.   Literal: ‘And bought a car?’
    2.   Intended interpretation: ‘What about buying a car? Who did that?’

3.2.2 Sentential adverbs

The example in (40) shows that sentential adverbs cannot be topicalized out of wh-questions.6 Likewise, (41) shows that they cannot be remnants for wh-TREQs.

    1. (40)
    1. (41)
    1. A:
    1.   Va a
    2.   will
    1. renunciar
    2. quit
    1. Sonia
    2. Sonia
    1. seguramente.
    2. for.sure
    1.   ‘Sonia will quit for sure.’
    1.  
    1. B:
    1. *Y
    2.   and
    1. probablemente?
    2. probably
    1.   Literal: ‘And probably?’
    2.   Intended interpretation: ‘What about probably? Who will probably quit?’

3.2.3 Post- and pre-nominal adjectives

Post-nominal adjectives cannot be topicalized out of wh-questions, and they cannot be remnants of wh-TREQs:7

    1. (42)
    1. (43)
    1. A:
    1.   Sonia
    2.   Sonia
    1. contrató
    2. hired
    1. a
    2. dom
    1. un
    2. a
    1. amigo
    2. friend
    1. viejo.
    2. old
    1.   ‘Sonia hired an old friend (= a friend that’s old).’
    1.  
    1. B:
    1. *Y
    2.   and
    1. joven?
    2. young
    1.   Literal: ‘And young?’
    2.   Intended interpretation: ‘What about young? Who hired a young friend?’

Similarly, (44) and (45) show that, as predicted, pre-nominal adjectives cannot be topicalized out of wh-questions nor be remnants for wh-TREQs.

    1. (44)
    1. (45)
    1. A:
    1.   Sonia
    2.   Sonia
    1. contrató
    2. hired
    1. a
    2. dom
    1. un
    2. an
    1. viejo
    2. old
    1. amigo.
    2. friend
    1.   ‘Sonia hired an old friend (= long-time friend).’
    1.  
    1. B:
    1. *Y
    2.   and
    1. nuevo?
    2. new
    1.   Literal: ‘And new?’
    2.   Intended interpretation: ‘What about new? Who hired a new friend?’

3.3 Interim summary

In section 3.1 and section 3.2 I presented compelling evidence establishing a strong correlation between those constituents that can be topicalized out of wh-questions and those constituents that can be remnants in wh-TREQs in Spanish. These patterns support the argument made here that there’s structure inside the ellipsis site and that the remnant has been topicalized out of it. In the next section I will examine wh-TREQs in the context of syntactic islands. As I will show, wh-TREQs that would involve topicalization or wh-movement (or both) from inside an island are not possible. These findings provide additional evidence of the need for syntactic isomorphism between the antecedent and the ellipsis site. Furthermore, they support the claim that the remnant undergoes topicalization out of the ellipsis site and that there is also wh-movement inside the ellipsis site.

3.4 Islands

In this section I analyze the behavior of wh-TREQs in contexts of islands. To avoid making this section unnecessarily long, I focus on complex-NP islands, while other types of islands such as adjunct islands, relative-clause islands, and interrogative wh-islands are illustrated in appendix A (see supplementary material). I provide evidence for the claim that wh-TREQs involve topicalization out of the ellipsis site and wh-movement inside the ellipsis site and that the ellipsis site must be structurally isomorphic to its antecedent.

Topicalizations out of complex NPs are ungrammatical, giving rise to a complex-NP island, as illustrated in (46b). For the sake of explicitness, (46a) shows that when the topicalization doesn’t involve an island, the sentence is fully grammatical, as expected.

    1. (46)

If, as proposed here, remnants of wh-TREQs are topicalized out of the ellipsis site and the ellipsis site contains a structure that is syntactically identical to its antecedent’s, then we expect wh-TREQs that would involve a topicalization out of a complex NP not to be possible. This prediction is borne out:

    1. (47)
    1. A:
    1.   Sonia
    2.   Sonia
    1. escuchó
    2. heard
    1. el
    2. the
    1. rumor
    2. rumor
    1. de
    2. of
    1. que
    2. that
    1. comí
    2. I.ate
    1. pizza.
    2. pizza
    1.   ‘Sonia heard the rumor that I ate pizza.’
    1.  
    1. B:
    1. *Y
    2.   and
    1. pasta?
    2. pasta
    1.   Literal: ‘And pasta?’
    2.   Intended interpretation: ‘What about pasta? Who heard the rumor that you ate that?’

In other words, the ungrammaticality of the wh-TREQ in (47) suggests that its source is the following ungrammatical structure.

    1. (48)

Importantly, the example just discussed involves a topicalization from inside the island and wh-movement from outside the island, indicating that only the former movement is illicit, while the latter is allowed. However, the reverse scenario also results in an ungrammatical wh-TREQ, as shown below; this serves as supporting evidence for the claim that there is indeed wh-movement occurring inside the ellipsis site. First, to provide further clarity, the example in (49a) illustrates the possibility of moving (i.e., topicalizing) material from outside an island, while (49b) demonstrates that, as expected, a wh-phrase cannot be moved out of a complex NP.

    1. (49)

When this is tested in the context of a wh-TREQ, the result is ungrammatical:

    1. (50)
    1. A:
    1.   Le
    2.   to.her
    1. conté
    2. I.told
    1. a
    2. to
    1. Sonia
    2. Sonia
    1. el
    2. the
    1. rumor
    2. rumor
    1. de
    2. of
    1. que
    2. that
    1. comiste
    2. you.ate
    1. pizza.
    2. pizza
    1.   ‘I told Sonia the rumor that you ate pizza.’
    1.  
    1. B:
    1. *Y
    2.   and
    1. a
    2. to
    1. Bruno?
    2. Bruno
    1.   Literal: ‘And Bruno?’
    2.   Intended: ‘What about Bruno? Whati did you tell him the rumor that I ate ti?’8

Crucially, the ungrammaticality of (50) is not due to the topicalization of a Bruno ‘to Bruno,’ since it doesn’t cross an island boundary, as (49a) above shows. Instead, it is the wh-movement of qué ‘what’ that renders the sentence ungrammatical, as exemplified in (49b). This suggests that the source of the ungrammaticality in (50) can be traced to the structure in (51), in other words, that the ungrammaticality arises from the wh-movement rather than the topicalization.

    1. (51)

This, in turn, provides evidence that there is indeed wh-movement inside the ellipsis site.

Finally, for the sake of completeness, the following example shows that when both the topicalized remnant and the wh-phrase move from inside the island, the result is ungrammatical.

    1. (52)

As (53) shows, wh-TREQs are also ungrammatical in these contexts, as predicted given that the source would have been the illicit structure in (52).

    1. (53)
    1. A:
    1.   Escuché
    2.   heard
    1. el
    2. the
    1. rumor
    2. rumor
    1. de
    2. of
    1. que
    2. that
    1. Sonia
    2. Sonia
    1. comió
    2. ate
    1. pizza.
    2. pizza
    1.   ‘I heard the rumor that Sonia ate pizza.’
    1.  
    1. B:
    1. *Y
    2.   and
    1. pasta?
    2. pasta
    1.   Literal: ‘And pasta?’
    2.   Intended: ‘What about pasta? Whoi did you hear the rumor that theyi ate that?’

In summary, through an analysis of wh-TREQs in island contexts, I have presented evidence supporting the following claims. Firstly, the remnant undergoes topicali- zation and moves out of the ellipsis site. This is evident from the fact that the unavailability of topicalization also renders wh-TREQs impossible. Secondly, the ellipsis site contains a wh-phrase that has undergone movement: there is a parallel between the impossibility of moving wh-phrases out of islands in non-elliptical wh-questions and the impossibility of wh-movement out of islands in wh-TREQ contexts. Lastly, there is a requirement for syntactic identity between the ellipsis site and its antecedent—given that, when the antecedent includes a complex NP, the ellipsis site must also include this structure. In other words, evasion strategies such as copular sources or short sources are not available here.9

An outline of the analyzed patterns can be found in (54). In brief, the results are that wh-TREQs in Spanish are possible only when both topicalization and wh-movement are allowed and are ungrammatical in cases where topicalization, wh-movement, or both are not possible.

    1. (54)

In the next section I discuss connectivity effects, the last component of my argument for the need for a syntactic-identity condition to license ellipsis in wh-TREQs.

3.5 Connectivity effects

In this section, I present additional evidence for the claim that syntactic identity between the ellipsis site and the antecedent is needed to license ellipsis in wh-TREQs. This evidence stems from various connectivity effects: specifically, I examine the impossibility of case mismatches, of P-stranding under ellipsis, of voice mismatches, and of ‘spray’–‘load’ alternations.

3.5.1 Case mismatches

Although Spanish doesn’t mark all of its objects with morphological case, definite animate objects must be preceded by an a, typically referred to as differential object marking (DOM—also known as “accusative a,” “personal a,” and “a marking,” among others; see, e.g., Brugè & Brugger 1996, Torrego 1999, López 2012, Vicente 2015):

    1. (55)
    1. a.
    1. Sonia
    2. Sonia
    1. escondió
    2. hid
    1. (*a)l
    2.    dom.the
    1. tesoro.
    2. treasure
    1. ‘Sonia hid the treasure.’
    1.  
    1. b.
    1. Sonia
    2. Sonia
    1. escondió
    2. hid
    1. *(a)
    2.    dom
    1. Bruno.
    2. Bruno
    1. ‘Sonia hid Bruno.’

As the following example shows, omitting the DOM is not possible in wh-TREQs; that is, DOM cannot be absent from the remnant.

    1. (56)
    1. A:
    1. Sonia
    2. Sonia
    1. escondió
    2. hid
    1. a
    2. dom
    1. Bruno.
    2. Bruno
    1. ‘Sonia hid Bruno.’
    1.  
    1. B:
    1. Y
    2. and
    1. *(a)
    2.     dom
    1. Luciano?
    2. Luciano
    1. Literal: ‘And Luciano?’
    2. Interpretation: ‘What about Luciano? Who hid him?’

The ungrammaticality of DOM-less remnants serves as evidence against non-isomorphic, copular sources. Such sources would actually necessitate a remnant without DOM. The pivot of a copular source is ungrammatical if it bears DOM, as (57a) illustrates, whereas a topicalized object requires DOM, as (57b) illustrates.

    1. (57)
    1. a.
    1. Y
    2. and
    1. (*a)
    2.    dom
    1. Luciano,
    2. Luciano
    1. quién
    2. who
    1. es
    2. is
    1. la
    2. the
    1. persona
    2. person
    1. que
    2. that
    1. lo
    2. him
    1. escondió?
    2. hid
    1. ‘And Luciano, who is the person that hid him?’
    1.  
    1. b.
    1. Y
    2. and
    1. *(a)
    2.     dom
    1. Luciano,
    2. Luciano
    1. quién
    2. who
    1. lo
    2. him
    1. escondió?
    2. hid
    1. ‘And Luciano, who hid him?’

If non-isomorphic sources (such as copular sources) were available in ellipsis, we would predict the DOM-less version of (56B) to be grammatical, contrary to fact.

Since Ross 1969, it has been observed that case matching between the remnant and its correlate is necessary in contexts of ellipsis. As discussed in Merchant 2001, the impossibility of case mismatches follows trivially if the ellipsis site contains an elided version of the relevant case assigner (here, the verb). It should be noted that, given the nature of DOM in Spanish (i.e., that it is only required for definite animate objects), this marking needs to be present in the remnant even in contexts where there is no DOM in the antecedent, as shown in (58). The correlate object in the antecedent in (58) doesn’t bear DOM because it’s an inanimate object, but the remnant must bear DOM because it’s animate and definite.

    1. (58)
    1. A:
    1. Sonia
    2. Sonia
    1. escondió
    2. hid
    1. el
    2. the
    1. tesoro.
    2. treasure
    1. ‘Sonia hid the treasure.’
    1.  
    1. B:
    1. Y
    2. and
    1. *(a)l
    2.    dom.the
    1. tesorero?
    2. treasurer
    1. Literal: ‘And the treasurer?’
    2. Interpretation: ‘What about the treasurer? Who hid him?’

This example shows that what is relevant is not case matching, strictly speaking, but case assignment and that the ellipsis site contains a configuration in which DOM is assigned to the remnant, regardless of the (lack of) case marking in the correlate in the antecedent.

3.5.2 P-stranding under ellipsis

The preposition cannot be omitted from the remnant in wh-TREQs like the following.

    1. (59)
    1. A:
    1. Sonia
    2. Sonia
    1. habló
    2. talked
    1. con
    2. with
    1. Bruno.
    2. Bruno
    1. ‘Sonia talked with Bruno.’
    1.  
    1. B:
    1. Y
    2. and
    1. *(con)
    2.     with
    1. Luciano?
    2. Luciano
    1. Literal: ‘And with Luciano?’
    2. Interpretation: ‘What about Luciano? Who talked with him?’

Importantly, topicalization of PPs requires pied-piping the preposition:10

    1. (60)

This pattern shows the need for strict isomorphism between the linguistic antecedent and the ellipsis site, disallowing alternative structures, such as copular/cleft sources, as the source of ellipsis (this is in line with recent research on clausal ellipsis in Spanish, such as Stigliano 2022). For the sake of explicitness, the following example shows that a copular structure can indeed occur as an overt continuation to A’s utterance.

    1. (61)
    1. A:
    1. Sonia
    2. Sonia
    1. habló
    2. talked
    1. con
    2. with
    1. Bruno.
    2. Bruno
    1. ‘Sonia talked with Bruno.’
    1.  
    1. B:
    1. Y
    2. and
    1. Luciano,
    2. Luciano
    1. quién
    2. who
    1. es
    2. is
    1. la
    2. the
    1. persona
    2. person
    1. que
    2. that
    1. habló
    2. talked
    1. con
    2. with
    1. él?
    2. him
    1. ‘And Luciano, who is the person that talked with him?’

In this regard, if a cleft/copular source were a possible source for the ellipsis site, the availability of (61B) as an overt continuation would predict that the P-less version of (59B) above should be grammatical, contrary to fact.

3.5.3 Voice mismatches

Another piece of evidence for a need for syntactic identity between the ellipsis site and its antecedent is the unavailability of voice mismatches (see, e.g., Merchant 2013). An active sentence cannot be elided, as in (62B), if its antecedent is a passive sentence (62A). Importantly, a non-elliptical version of (62B) is indeed possible in this context, as shown in (62B’), ruling out an explanation based on pragmatic or congruence factors.

    1. (62)
    1. A:
    1.   La
    2.   the
    1. casa
    2. house
    1. fue
    2. was
    1. destruida
    2. destroyed
    1. por
    2. by
    1. Sonia.
    2. Sonia
    1.   ‘The house was destroyed by Sonia.’
    1.  
    1. B:
    1. *Y
    2.   and
    1. Bruno?
    2. Bruno
    1.   Literal: ‘And Bruno?’
    2.   Intended interpretation: ‘What about Bruno? What did he destroy?’
    1.  
    1. B’:
    1.   Y
    2.   and
    1. Bruno,
    2. Bruno
    1. qué
    2. what
    1. destruyó?
    2. destroyed
    1.   ‘And Bruno, what did he destroy?’

Further evidence comes from the possible answers that A can give to B’s wh-TREQ when the antecedent is passive, as in the following example. As shown, A’s answer must be a ‘by’ phrase.

    1. (63)
    1. A:
    1.   La
    2.   the
    1. casa
    2. house
    1. fue
    2. was
    1. destruida
    2. destroyed
    1. por
    2. by
    1. Sonia.
    2. Sonia
    1.   ‘The house was destroyed by Sonia.’
    1.  
    1. B:
    1.   Y
    2.   and
    1. el
    2. the
    1. auto?
    2. car
    1.   Literal: ‘And the car?’
    2.   Interpretation: ‘What about the car? Who was the car destroyed by?’
    1.  
    1. A:
    1. *(Por)
    2.     by
    1. Bruno.
    2. Bruno
    1.   ‘By Bruno.’

I argue that, based on the mandatory presence of the preposition por ‘by’ in A’s answer, the ellipsis site in (63B) must contain a passive sentence, like the following.

    1. (64)
    1. Y
    2. and
    1. el
    2. the
    1. auto,
    2. car
    1. por
    2. by
    1. quién
    2. who
    1. fue
    2. was
    1. destruído?
    2. destroyed
    1. ‘What about the car? By whom was it destroyed?’

In other words, a non-isomorphic source like the one in (65) is not possible in the context above.

    1. (65)
    1. Y
    2. and
    1. el
    2. the
    1. auto,
    2. car
    1. quién
    2. who
    1. lo
    2. it
    1. destruyó?
    2. destroyed
    1. ‘What about the car? Who destroyed it?’

Otherwise a P-less answer by A would be perfectly possible, contrary to fact.

For the sake of completeness, we see in (66) that an active–passive mismatch in the other direction (i.e., passive source given an active antecedent) is also impossible.

    1. (66)
    1. A:
    1.   Sonia
    2.   Sonia
    1. destruyó
    2. destroyed
    1. la
    2. the
    1. casa.
    2. house
    1.   ‘Sonia destroyed the house.’
    1.  
    1. B:
    1. *Y
    2.   and
    1. por
    2. by
    1. Bruno?
    2. Bruno
    1.   Literal: ‘And by Bruno?’
    2.   Intended interpretation: ‘What about Bruno? What was destroyed by him?’

3.5.4 ‘Spray’–‘load’ alternations

The so-called ‘spray’–‘load’ alternation is a diathesis alternation in which a verb describing caused motion of one entity to another exhibits two arguments (see Beavers 2017 and references therein for an overview of this topic):

    1. (67)
    1. a.
    1. ‘With’ variant
    1. Sonia
    2. Sonia
    1. cargó
    2. loaded
    1. el
    2. the
    1. camión
    2. truck
    1. con
    2. with
    1. libros.
    2. books
    1. ‘Sonia loaded the truck with books.’
    1.  
    1. b.
    1. ‘Onto’ variant
    1. Sonia
    2. Sonia
    1. cargó
    2. loaded
    1. libros
    2. books
    1. en
    2. in
    1. el
    2. the
    1. camión.
    2. truck
    1. ‘Sonia loaded books onto the truck.’

This alternation is disallowed under ellipsis, as the examples in (68) show for sluicing.11

    1. (68)
    1. a.
    1. *Sonia
    2.   Sonia
    1. cargó
    2. loaded
    1. algún
    2. some
    1. camión
    2. truck
    1. con
    2. with
    1. libros,
    2. books
    1. pero
    2. but
    1. no
    2. not
    1. sé
    2. I.know
    1. en
    2. in
    1. qué
    2. which
    1. camión.
    2. truck
    1.   Intended: ‘Sonia loaded some truck with books, but I don’t know onto which truck.’
    1.  
    1. b.
    1. *Sonia
    2.   Sonia
    1. cargó
    2. loaded
    1. algo
    2. something
    1. en
    2. in
    1. el
    2. the
    1. camión,
    2. truck
    1. pero
    2. but
    1. no
    2. not
    1. sé
    2. I.know
    1. con
    2. with
    1. qué.
    2. what
    1.   Intended: ‘Sonia loaded something onto the truck, but I don’t know with what.’

Only structural matching is allowed:

    1. (69)
    1. a.
    1. Sonia
    2. Sonia
    1. cargó
    2. loaded
    1. algún
    2. some
    1. camión
    2. truck
    1. con
    2. with
    1. libros,
    2. books
    1. pero
    2. but
    1. no
    2. not
    1. sé
    2. I.know
    1. qué
    2. which
    1. camión.
    2. truck
    1. ‘Sonia loaded some truck with books, but I don’t know which truck.’
    1.  
    1. b.
    1. Sonia
    2. Sonia
    1. cargó
    2. loaded
    1. algo
    2. something
    1. en
    2. in
    1. el
    2. the
    1. camión,
    2. truck
    1. pero
    2. but
    1. no
    2. not
    1. sé
    2. I.know
    1. qué.
    2. what
    1. ‘Sonia loaded something onto the truck, but I don’t know what.’

The unavailability of ‘spray’–‘load’ alternations under ellipsis is usually taken to be strong evidence of the need for syntactic identity and against semantic approaches.

Importantly, ‘spray’–‘load’ alternations are also disallowed in contexts of wh-TREQs, as (70) and (71) show. This is not for pragmatic reasons nor due to a question–answer incongruence, given that the non-elliptical counterpart of B’s wh-TREQ is possible, as shown in (70B’) and (71B’).

    1. (70)
    1. A:
    1.   Sonia
    2.   Sonia
    1. cargó
    2. loaded
    1. el
    2. the
    1. camión
    2. truck
    1. con
    2. with
    1. libros.
    2. books
    1.   ‘Sonia loaded the truck with books.’
    1.  
    1. B:
    1. *Y
    2.   and
    1. en
    2. in
    1. el
    2. the
    1. auto?
    2. car
    1.   Literal: ‘And in the car?’
    2.   Intended interpretation: ‘What about the car? What did he load in it?’
    1.  
    1. B’:
    1.   Y
    2.   and
    1. en
    2. in
    1. el
    2. the
    1. auto
    2. car
    1. qué
    2. what
    1. cargó?
    2. loaded
    1.   ‘What about the car? What did he load in it?’
    1. (71)
    1. A:
    1.   Sonia
    2.   Sonia
    1. cargó
    2. loaded
    1. libros
    2. books
    1. en
    2. in
    1. el
    2. the
    1. camión.
    2. truck
    1.   ‘Sonia loaded books in the truck.’
    1.  
    1. B:
    1. *Y
    2.   and
    1. con
    2. with
    1. revistas?
    2. magazines
    1.   Literal: ‘And with magazines?’
    2.   Intended: ‘What about magazines? What did Sonia load them with?’
    1.  
    1. B’:
    1.   Y
    2.   and
    1. con
    2. with
    1. revistas
    2. magazines
    1. qué
    2. what
    1. cargó?
    2. loaded
    1.   ‘What about magazines? What did she load them with?’

The only grammatical option is the one in which there’s structural matching between the antecedent and the ellipsis site, which I take to show that syntactic identity is indeed needed:

    1. (72)
    1. A:
    1. Sonia
    2. Sonia
    1. cargó
    2. loaded
    1. el
    2. the
    1. camión
    2. truck
    1. con
    2. with
    1. libros.
    2. books
    1. ‘Sonia loaded the truck with books.’
    1.  
    1. B:
    1. Y
    2. and
    1. con
    2. with
    1. revistas?
    2. magazines
    1. Literal: ‘And with magazines?’
    2. Interpretation: ‘What about magazines? What did she load with them?’
    1.  
    1. B’:
    1. Y
    2. and
    1. el
    2. the
    1. auto?
    2. car
    1. Literal: ‘And the car?’
    2. Interpretation: ‘What about the car? What did she load it with?’
    1. (73)
    1. A:
    1. Sonia
    2. Sonia
    1. cargó
    2. loaded
    1. libros
    2. books
    1. en
    2. in
    1. el
    2. the
    1. camión.
    2. truck
    1. ‘Sonia loaded books in the truck.’
    1.  
    1. B:
    1. Y
    2. and
    1. revistas?
    2. magazines
    1. Literal: ‘And magazines?’
    2. Interpretation: ‘What about magazines? Where did she load them?’
    1.  
    1. B’:
    1. Y
    2. and
    1. en
    2. in
    1. el
    2. the
    1. auto?
    2. car
    1. Literal: ‘And onto the car?’
    2. Interpretation: ‘What about the car? What did she load there?’

3.6 Interim summary

So far, the evidence presented supports the following conclusions. (i) In wh-TREQs, there exists underlying structure within the ellipsis site, indicating that this construction results from ellipsis; (ii) the ellipsis site contains a wh-question with a moved wh-phrase; (iii) the remnant is a topicalized XP that undergoes movement out of the ellipsis site; and (iv) some form of syntactic identity or isomorphism between the ellipsis site and its antecedent is required. A schematic summary of my proposal is the following.

    1. (74)

In the following section, I extensively discuss my proposal regarding the identity condition needed to license this elliptical construction, and I put forth a more detailed formalization of how wh-TREQs in Spanish are derived.

4 Mixed-identity requirements

In this section I present my proposal for deriving wh-TREQs in Spanish. In the previous section I established that syntactic identity is needed to license this type of ellipsis. However, I will now demonstrate that an identity condition solely based on syntactic isomorphism is too strict and actually leads to the incorrect prediction that wh-TREQs are impossible. In particular, the problem for a strict syntactic account is that the wh-phrase qué ‘what’ in (75B) gets deleted, while its correlate in the antecedent (75A) is the NP pizza ‘pizza.’

    1. (75)
    1. .

Crucially, qué ‘what’ and pizza ‘pizza’ are not syntactically identical, failing to comply with any possible version of a syntactic-identity condition.

In light of this, I propose that wh-TREQs are licensed under “mixed-identity” requirements. Specifically, syntactic identity licenses ellipsis of the lower portion of the structure, namely the TP. This aligns with recent proposals on TP ellipsis in Spanish. However, as for deletion of material above the TP, I argue that pragmatic/general conditions related to discourse congruence come into play to determine what can be included within the ellipsis site between the TP and the CP. In other words, there doesn’t seem to be a specific licensing condition for eliding material above the TP. In section 4.1 I implement my proposal by positing the existence of two [E]-features: one on C and the other on Top. These features impose different identity requirements on their complements. This is schematized in the following.

    1. (76)

Before delving into the specific implementation of my proposal, it is important to highlight the connection between this framework on the one hand and Kroll & Rudin 2017 and Rudin 2019 on the other hand. The authors of these works suggest that both syntactic grammaticality constraints and pragmatic licensing constraints play a role in explaining the range of (im)possible mismatches in sluicing in English. According to their findings, syntactic identity strictly applies to material in the verbal domain (referred to as the “eventive core”). As a result, mismatches are allowed above the vP level, as long as they comply with a pragmatic condition that requires contextual entailment between the local context and the ellipsis site.12 I claim that wh-TREQs in Spanish display a comparable pattern. That is, there is a higher domain where strict syntactic identity doesn’t appear to apply and a lower domain where it is mandatory. Although the current analysis and the empirical domain under investigation differ significantly from the aforementioned works, there is a shared characteristic: both Spanish wh-TREQs and English sluicing involve independently attested ellipsis domains, namely TP ellipsis and vP ellipsis, respectively. Furthermore, one of these domains is nested within the other, and it is the lower one that enforces the requirement of syntactic identity. This will be relevant for the analysis I propose here.

4.1 Deriving wh-TREQs

I follow Merchant 2001 in assuming that ellipsis is licensed in the complement of heads bearing an [E]-feature. However, in the context of wh-TREQs, I propose the existence of two distinct [E]-features that impose different identity conditions: I will refer to them as [Esyn] and [E]. This idea draws inspiration from the framework presented in Aelbrecht 2010, which posits that there is a specific [E]-feature in the lexicon for each elliptical phenomenon. In line with Aelbrecht’s proposal, each [E]-feature is only compatible with certain heads. This is encoded in its selectional features, as will be explained shortly.

First, I argue that [Esyn] is compatible exclusively with the C head, thereby triggering TP-ellipsis phenomena (e.g., sluicing, fragment answers, stripping) and enforcing strict syntactic identity. I adopt a head-by-head evaluation of the identity condition,13 although alternative implementations might yield similar results:

    1. (77)
    1. Licensing of [Esyn]
    1.  
    1. a.
    1. Every head h in the complement of a head z[Esyn] is deleted iff h is not dominated by an [F]-marked node or a [top]-marked node.14,15
    1.  
    1. b.
    1. A head h marked for deletion by [Esyn] is licensed iff it has an identical correlate h’ in its antecedent A.

In particular, I adopt and modify Saab 2008 and 2010’s definition of identity:

    1. (78)
    1. A morpheme α is identical to another morpheme β if and only if α and β match all their semantic and syntactic features.
    2. (Saab 2010: 102–103)

Building upon Saab’s framework, I assume that the calculation of ellipsis proceeds top down, with identity being successively evaluated for each head within the ellipsis site starting from the highest head that has been marked for deletion.

Additionally, I propose that [E] is only compatible with the Top head (and not with C). As I mentioned above, I argue that [E] doesn’t impose any identity requirements:

    1. (79)
    1. Licensing of [E]
    2. Every head h in the complement of a head z[E] is deleted.

Next, I will provide an example of the derivation of a wh-TREQ, such as the one in (75), repeated here.

    1. (80)

First, as shown in (81), the wh-phrase qué ‘what’ undergoes wh-movement to the specifier of the CP; after that, ellipsis is triggered by the [Esyn] feature on C. This means that every head in the complement of C is marked for deletion and must find an identical correlate in the antecedent, which it does. It should be noted that the subject Bruno is not marked for deletion due to being [top] marked.

    1. (81)

Once ellipsis of the TP is triggered, the Top head, endowed with [E], is merged, as in (82). This head triggers topicalization of the subject, followed by deletion of the heads in the complement of Top. The heads being deleted in this step are those of the wh-phrase and C, as the other—lower—heads have already undergone deletion in the previous step.

    1. (82)

At this point, it’s important to recall that the connectivity effects discussed in section 3.5 provide strong evidence of the need for syntactic identity in wh-TREQs. In this respect, if only the [E] on Top were present in the structure, anything could be deleted without any reference to the ellipsis site’s linguistic antecedent, leading to incorrect predictions. To account for the observed patterns, I propose that the necessity for some form of syntactic isomorphism is achieved if ellipsis of material between the TP and the CP in wh-TREQs (i.e., the wh-phrase and the C head) depends on a prior application of ellipsis of the TP. This is because the ellipsis of the TP, triggered by [Esyn], enforces a stricter identity condition, imposing syntactic identity. In other words, I take this to mean that the presence of [E] depends on the presence of [Esyn] lower in the structure.

In order to implement this idea, I follow Aelbrecht 2010 in holding that ellipsis is licensed via an Agree relation. More specifically, according to Aelbrecht, each type of elliptical construction is associated with a specific ellipsis feature bundle in the lexicon. In the framework proposed here, this implies that each ellipsis-triggering feature, namely [Esyn] and [E], consists of different feature bundles. Aelbrecht argues that the compatibility between the [E]-feature and a specific head is encoded in its selectional (sel) features. Furthermore, she suggests that [E]-features also possess an inflectional (infl) feature, corresponding to the category (cat) feature of the ellipsis licensor head. Consequently, the licensing of different ellipsis-triggering features relies on establishing a checking relation with their licensor.

It is also important to note the differences between Aelbrecht’s proposal and the one put forth here. Firstly, the directionality of the Agree relation is standard (probing down) rather than upward. Secondly, I argue that an [E]-feature can require agreement with another [E]-feature (that is, the licensors do not need to be heads but can also be features).

The featural makeup of the ellipsis-triggering features in Spanish is as follows.16,17

    1. (83)

Specifically, I argue that [Esyn] in Spanish does not possess any inflectional features, thereby not requiring the establishment of Agree relations with other heads/features. The agreement relation between [E] and [Esyn] is illustrated in (84), with simplified feature matrices that highlight the relevant features.

    1. (84)

Note that the [Esyn] feature on C has remained accessible for further operations, including agreement with the [E]-feature on Top.

Before concluding this section, it is worth briefly addressing the potential for cross-linguistic variation that the existence of different [E]-features, as proposed in this analysis, may create. The specific construction I am exploring here (i.e., wh-TREQs) employs [E] in a very restricted manner, and its featural makeup requires it to enter into an Agree relation with an [Esyn] feature lower in the structure. However, someone could argue that, by the logic of cross-linguistic feature variation, one might also expect the existence of an [E]-feature that neither imposes an identity requirement nor enters into an Agree relation with other features—that is, whose presence is fairly unconstrained. If such an unconstrained [E]-feature existed, one would expect to find a language where ellipsis of any constituent could occur in any context. To the best of my knowledge, no such instances of unconstrained ellipsis have been reported. Consequently, I hypothesize that a fully unconstrained [E]-feature should not be possible at all. In light of this hypothesis, I propose that an [E]-feature should either impose some form of identity requirement with a linguistic antecedent or enter into an Agree relation with a feature that imposes such a requirement (or both). In concrete terms, this implies that [E]-features must possess, either directly or indirectly, an antecedent that satisfies the necessary conditions for ellipsis licensing and identity.

Finally, I acknowledge that alternative proposals could account for the patterns observed with respect to the mixed-identity requirements in wh-TREQs. While the implementation I put forth in this article offers one way of analyzing the data and constructing a theory of ellipsis, it’s important to recognize that alternative implementations may also be able to account for the phenomenon under investigation. For instance, an anonymous reviewer suggests a condition according to which, in Spanish, the largest domain where syntactic identity can be imposed would be the TP, regardless of whether the [Esyn] feature is borne by (and hence elides the sister of) the Top head.

Although this alternative proposal could account for the Spanish data we examined, I believe there are conceptual advantages and empirical reasons to maintain the approach I propose here. For instance, it’s standard to assume that, in Spanish, there is an [E]-feature on C that triggers TP ellipsis, leading to phenomena like sluicing, fragment answers, split questions, and pseudostripping, and that there is an [E]-feature on Num that triggers nP ellipsis (see, e.g., Saab & Lipták 2016). While it is not always explicitly stated, one can assume, in accord with the proposal of Aelbrecht 2010, that these are different [E]-features. Various facts support this hypothesis: these [E]-features are compatible with different heads (C and Num, for TP ellipsis and nP ellipsis, respectively) and are licensed under different conditions, among many other observations.

These characteristics likely vary cross-linguistically, as evidenced by differences in licensing conditions and what the available [E]-features are in each language’s lexical inventory. For instance, one could argue that English has an [E]-feature compatible with v, which gives rise to VP ellipsis, but that Spanish doesn’t have this [E]-feature in its inventory. Furthermore, there is within-language variation as well; one example in Spanish comes from Saab & Stigliano 2023. There is an elliptical phenomenon in Spanish that, on the surface, looks like VP ellipsis but that should actually be analyzed as PredP ellipsis; this type of ellipsis is triggered by an [E]-feature on the Pred head and needs to be licensed through Agree with focus/polarity, which is not required in other types of ellipsis in Spanish such as clausal ellipsis and nominal ellipsis. In other words, the range of elliptical constructions available in one language as compared to others and the different identity and licensing conditions we observe both within a language and across languages make it reasonable to assume that there isn’t just one [E]-feature but, as Aelbrecht proposes, one [E]-feature for each elliptical phenomenon. With this in mind, it’s not unreasonable to assume that TP ellipsis (sluicing, fragment answers, stripping, etc.) and CP ellipsis (wh-TREQs) involve different [E]-features, which are licensed under different conditions and which impose different identity requirements.

Besides these empirical considerations, I believe that my proposal also bears conceptual advantages, since it doesn’t need to resort to referencing specific domains (or their labels) over which identity is calculated, as the analysis suggested by the anonymous reviewer does. That is, the domain over which [E] takes scope is always the complement of the head that bears it, and the specific licensing and identity conditions depend on which [E]-feature is involved (and its featural specification). That being said, ultimately, what I proposed here is one way of implementing the empirical findings of this article, but alternative implementations could achieve similar results. I leave this issue open to debate.

4.2 On the “optionality” of deleting the wh-phrase

The claim that ellipsis of the CP depends on ellipsis of the TP but not vice versa predicts that there should be cases similar to wh-TREQs but where only TP ellipsis occurs. These cases would resemble sluicing constructions with two remnants (a wh-phrase, as in regular sluicing, and a topicalized XP, as in wh-TREQs). This prediction is borne out:18

    1. (85)
    1. A:
    1. Sonia
    2. Sonia
    1. comió
    2. ate
    1. pizza.
    2. pizza
    1. ‘Sonia ate pizza.’
    1.  
    1. B:
    1. Y
    2. and
    1. Bruno
    2. Bruno
    1. qué?
    2. what
    1. ‘And Bruno, what?’
    1.  
    1. A:
    1. Pasta.
    2. pasta
    1. ‘Pasta.’
    1. (86)
    1. A:
    1. Sonia
    2. Sonia
    1. comió
    2. ate
    1. pizza.
    2. pizza
    1. ‘Sonia ate pizza.’
    1.  
    1. B:
    1. Y
    2. and
    1. pasta
    2. pasta
    1. quién?
    2. who
    1. ‘And pasta, who?’
    1.  
    1. A:
    1. Bruno.
    2. Bruno
    1. ‘Bruno.’

Although this gives the illusion that deletion of the wh-phrase is optional, I claim that these are actually cases of TP ellipsis, the only difference being the presence of an additional remnant (i.e., the topicalized XP). In other words, in these cases there is a Top head that makes Bruno in (85) and pasta in (86) undergo topicalization, but this Top head doesn’t bear an [E]-feature—parallel to what happens, for instance, with the C[wh] head: when it doesn’t bear an [E]-feature, a regular non-elliptical wh-question arises, and when it does bear an [E]-feature, sluicing arises. In the examples above, the Top head would not bear an [E]-feature; only the C head would bear one. This means that, given that the wh-phrase has moved out of the scope of the [E]-feature on C, it doesn’t get deleted.

5 Conclusions and open questions

Altogether, this article aimed to provide answers to the identity and licensing questions in order to contribute to our understanding of ellipsis in natural language. The central claim put forward here is that certain types of ellipsis exhibit “mixed-identity” requirements, imposing a strict syntactic-identity condition on one part of the structure and no identity requirements on another part. This claim was supported by the detailed examination of an understudied elliptical construction that I dubbed topic-remnant elided wh-questions (wh-TREQs). A thorough analysis of this construction yielded significant insights that extend beyond its specific domain, contributing to the broader understanding of ellipsis.

Based on the empirical data presented in this article, I proposed an elliptical account of wh-TREQs, suggesting that they involve the ellipsis of a wh-question from which a topic has moved. A crucial finding from the analysis of the data introduced here was that while syntactic identity is needed for licensing ellipsis in wh-TREQs, it only applies to the lower portion of the elided structure, namely the TP. In contrast, no specific identity requirements were found to hold for eliding the material located between the TP and the CP.

More specifically, I initially provided evidence to support the need for syntactic identity in wh-TREQs. This evidence came from the unavailability of voice mismatches and ‘spray’–‘load’ alternations and the fact that wh-TREQs are island sensitive, among other facts. However, it became apparent that a strict syntactic-identity condition alone would incorrectly predict the impossibility of this elliptical construction. To reconcile this apparent contradiction, I proposed an implementation according to which wh-TREQs arise as the result of two ellipses: one triggered by [Esyn] on C, which imposes syntactic identity, and another one triggered by [E] on Top, which does not impose any identity requirements at all but which must enter into an Agree relation with [Esyn] lower in the structure. My proposal is schematized as follows.

    1. (87)

In other words, the analysis of this understudied empirical domain led to the formulation of a theory of ellipsis licensing based on a typology of [E]-features. Concretely, each [E]-feature triggers certain operations, is licensed in particular syntactic configurations (depending on its agreement requirements), and can occur only with specific heads (C, Top). This proposal successfully accounts for the empirical patterns observed in the data under consideration here, without the need to propose construction-specific analyses or exceptional mechanisms.

A note on cross-linguistic variation is relevant here. In particular, I hypothesize that cross-linguistic variation can be found with respect to the typology of [E]-features and their featural makeup. While I proposed [Esyn] and [E] as ellipsis-triggering features in Spanish, I do not claim that these are the only ellipsis-triggering features that exist, either in this language or in others. On the contrary, my proposal acknowledges the possibility of the existence of other [E]-features with different identity requirements and different inflectional, selectional, and category features. For instance, there could be an [E]-feature that imposes semantic identity ([Esem]) or an [E]-feature that imposes a different type of syntactic identity, among other alternatives. Moreover, cross-linguistic variation may also extend to the heads that can select for these [E]-features. In Spanish, I argue that C can bear [Esyn] and Top can bear [E], but it is plausible that, in other languages, other heads are the ones that bear (these or other) [E]-features, leading to different patterns of (im)possible mismatches in ellipsis.

A note on the identity condition is also relevant at this point. I claimed that an additional identity condition is not required to license the ellipsis of material above the TP. Furthermore, I proposed that ellipsis of constituents located between TP and CP is possible, as long as other, more general pragmatic/discourse conditions are satisfied (for instance, along the lines of Büring 2003). What matters here is that the conditions governing wh-TREQs also govern their non-elliptical counterparts, indicating that these constraints are not unique to elliptical structures. A special case is related to the (un)availability of sprouting. At first sight, considering the following example, it may seem that sprouting is disallowed in the context of wh-TREQs, that is, that each deleted wh-phrase must have an overt syntactic correlate in the antecedent.

    1. (88)
    1. A:
    1.   Sonia
    2.   Sonia
    1. comió.
    2. ate
    1.   ‘Sonia ate.’
    1.  
    1. B:
    1. *Y
    2.   and
    1. pizza?
    2. pizza
    1.   Literal: ‘And pizza?’
    2.   Intended: ‘What about pizza? Who ate that?’

However, crucially, this ban on sprouting is not something special about wh-TREQs, as it is also observed in their non-elliptical counterparts:

    1. (89)
    1. A:
    1.   Sonia
    2.   Sonia
    1. comió.
    2. ate
    1.   ‘Sonia ate.’
    1.  
    1. B:
    1. #Y
    2.   and
    1. pizza,
    2. pizza
    1. quién
    2. who
    1. comió?
    2. ate
    1.   ‘And pizza, who ate?’

This suggests that the restriction on sprouting may be linked to a broader pragmatic condition (see, e.g., the Contrastive-Topic-Congruence Condition in Büring 2003). Again, why the sprouted question in (89) is not possible in the given context is outside the scope of this article, but what matters here is that the impossibility of sprouting in (88) mirrors the impossibility of sprouting in its non-elliptical counterpart. Interestingly, this contrasts with other types of ellipsis that do allow sprouting, such as sluicing; importantly, in these cases, a non-elliptical counterpart is also possible.

While it is impossible to definitively prove the absence of an identity condition, I am not aware of any data supporting the necessity of such a condition for licensing the deletion of material between the TP and CP. This, of course, remains as an open empirical question that should be revisited if new data are discovered. Worth mentioning is the case of so-called inheritance of content (see Barros 2014, Weir 2014, Jacobson 2016, and Vicente et al. 2021, among others), which indirectly demonstrates that there can be some mismatches between the antecedent and the ellipsis site. In the case of wh-TREQs, the remnants do not inherit the restriction of their correlates:

    1. (90)
    1. A:
    1. Todos
    2. all
    1. los
    2. the
    1. directores
    2. directors
    1. invitaron
    2. invited
    1. a
    2. dom
    1. un
    2. an
    1. actor.
    2. actor
    1. Por
    2. for
    1. ejemplo,
    2. example
    1. Tarantino
    2. Tarantino
    1. invitó
    2. invited
    1. a
    2. dom
    1. Harvey
    2. Harvey
    1. Keitel.
    2. Keitel
    1. ‘Every director invited an actor. For example, Tarantino invited Harvey Keitel.’
    1.  
    1. B:
    1. Y
    2. and
    1. Scorsese?
    2. Scorsese
    1. Literal: ‘And Scorsese?’
    2. Intended interpretation: ‘What about Scorsese? Who did Scorsese invite?’
    1.  
    1. A:
    1. A
    2. dom
    1. Francis
    2. Francis
    1. Ford
    2. Ford
    1. Coppola,
    2. Coppola
    1. pero
    2. but
    1. no
    2. not
    1. es
    2. is
    1. actor.
    2. actor
    1. ‘Francis Ford Coppola, but he is not an actor.’

These facts contrast, again, with the patterns observed in other types of ellipsis, such as sluicing and fragment answers, which do inherit the restriction of their correlates. This again points to fundamental differences between the type of ellipsis I’ve examined here and other types such as sluicing and fragment answers. As an anonymous reviewer points out, these facts open many questions, in particular, (i) when should we expect inheritance of content in ellipsis and when not, and (ii) how do we account for this restriction? I believe that a possible answer could be found in the identity conditions (and the lack thereof) that I proposed for this construction, although a more explicit explanation is left for future explorations.

Overall, I believe this article has addressed several questions, but it also highlights numerous avenues for future research. An immediate question that arises from the proposal I developed here concerns other types of ellipsis and the specific (featural makeup of the) [E]-features that license them. For instance, how do the [E]-features responsible for NP ellipsis, vP ellipsis, gapping, TP ellipsis, CP ellipsis, and others differ? How do these various [E]-features interact with each other and with other features or heads? Furthermore, I have proposed the existence of an [E]-feature imposing syntactic identity and another [E]-feature that doesn’t impose any identity requirements on its own. However, as I suggested above, it is conceivable that languages may possess different [E]-features, such as one imposing semantic identity or one stipulating a different, possibly weaker, form of syntactic identity. If this proves to be the case, it would be necessary to explore what these identity conditions entail. Overall, these open questions pave the way for future investigations into the nature of ellipsis and its underlying mechanisms. As a final note, to gain a deeper understanding of ellipsis phenomena, it is crucial to extend the investigation of the empirical domains analyzed in this article to other languages. This would allow us to uncover other generalizations and to explore how different types of ellipsis interact across languages.

In conclusion, this article has presented a compelling challenge to existing notions and proposals concerning the identity condition that licenses ellipsis and the underlying mechanisms responsible for ellipsis. This was achieved by performing an in-depth analysis of an understudied elliptical construction in Spanish, which allowed me to provide a complete empirical picture of the phenomena under examination. The main outcome of this investigation is a theory of ellipsis that not only accurately predicts the observed patterns in Spanish wh-TREQs but also provides a foundation for the analysis of other forms of ellipsis across languages.

Supplementary material

A file containing appendix A and appendix B can be downloaded at https://doi.org/10.16995/star.17534.s1.

Acknowledgments

I am indebted to Karlos Arregi, Vera Gribanova, Jason Merchant, and Andrés Saab for discussing these ideas with me. I am grateful to the audiences at the Center for Research in Syntax, Semantics, and Phonology (CRISSP), the Yale Syntax Reading Group, the 2021 meeting of the Linguistic Society of America, and the Morphology and Syntax Workshop at the University of Chicago for their questions, feedback, and discussion. I would like to thank the editors, Klaus Abels and Suzanne Flynn, and the three anonymous reviewers for their very useful comments and suggestions. All mistakes and shortcomings are my own.

Competing interests

The author declares that they have no competing interests.

Notes

  1. In (1) and (2), A’s first utterance is the prompt, B’s utterance is the wh-TREQ, and A’s second utterance is the response to B’s wh-TREQ. In the rest of this article, I won’t include A’s responses to B’s wh-TREQs, to avoid making the examples unnecessarily long, and I will just indicate the intended interpretation. ⮭
  2. All the examples of wh-TREQs presented in this article consist of the conjunction y ‘and’ followed by the remnant. It is worth mentioning that the presence of this conjunction is not compulsory (many speakers do indeed accept the examples without the conjunction), yet it is preferable over the variant without it. Importantly, this preference is not particular to the elliptical construction under examination given that non-elliptical versions of wh-TREQs also sound better/more natural with the conjunction preceding them. I thank an anonymous reviewer for bringing up this question. ⮭
  3. The translation of all the Spanish examples in this article includes what I refer to as a what-about question, followed by a resumed wh-question. I opted for this translation/interpretation of wh-TREQs because it seems to me to be the construction that most closely resembles the meaning of wh-TREQs in Spanish, in particular considering that topicalizations out of wh-questions are ungrammatical in English:
      1. (i)
      1. A:
      1.   Sonia ate pizza.
      1.  
      1. B:
      1. *And pasta, who ate?
    Nevertheless, I caution against drawing any conclusions solely based on these translations, as I believe what-about questions and wh-TREQs in Spanish are different in nature. Although an in-depth analysis of what-about questions in English is outside of the scope of this article, I briefly discuss some of their characteristics in appendix B (see supplementary material). ⮭
  4. asp = aspectual particle. ⮭
  5. sfp = sentence-final particle. ⮭
  6. Note that the problem with (40) is the topicalization of the adverb in the context of a wh-question. Sentential adverbs are possible in wh-questions as long as they remain in their base position:
      1. (i)
      1. Quién
      2. who
      1. va a
      2. will
      1. renunciar
      2. to.quit
      1. probablemente?
      2. probably
      1. ‘Who is probably going to quit?’
    ⮭
  7. dom = differential object marking. ⮭
  8. This fragment is indeed possible under a polar-question interpretation. Although polar TREQs are outside of the scope of this article, I should note that this type of interpretation wouldn’t involve any island violations. ⮭
  9. The same line of argumentation is replicated in appendix A (see supplementary material), where I extend the analysis to adjunct islands, interrogative wh-islands, and relative-clause islands. ⮭
  10. Contrariwise, hanging topics require the preposition to be absent; they can also, optionally, be preceded by con respecto a ‘with respect to’:
      1. (i)
      1. (Con
      2.   with
      1. respecto
      2. respect
      1. a)
      2. to
      1. Luciano,
      2. Luciano
      1. quién
      2. who
      1. habló
      2. talked
      1. con
      2. with
      1. él?
      2. him
      1. ‘With respect to Luciano, who talked with him?’
    As the following example shows, a remnant derived from (i) is unavailable in this context, as evidenced by the ungrammaticality of (iiB). This is despite the fact that its non-elliptical counterpart is perfectly possible, as (iiB’) shows.
      1. (ii)
      1. A:
      1.   Sonia
      2.   Sonia
      1. habló
      2. talked
      1. con
      2. with
      1. Bruno.
      2. Bruno
      1.   ‘Sonia talked with Bruno.’
      1.  
      1. B:
      1. *Y
      2.   and
      1. (con
      2.   with
      1. respecto
      2. respect
      1. a)
      2. to
      1. Luciano?
      2. Luciano
      1.   Intended interpretation: ‘With respect to Luciano, who talked with him?’
      1.  
      1. B’:
      1.   Y
      2.   and
      1. (con
      2.   with
      1. respecto
      2. respect
      1. a)
      2. to
      1. Luciano,
      2. Luciano
      1. quién
      2. who
      1. habló
      2. talked
      1. con
      2. with
      1. él?
      2. him
      1.   ‘And with respect to Luciano, who talked with him?’
    I take these facts as evidence that remnants of wh-TREQs are topicalized XPs, not hanging topics. ⮭
  11. Crucially, a non-elliptical continuation would be perfectly possible:
      1. (i)
      1. a.
      1. Sonia
      2. Sonia
      1. cargó
      2. loaded
      1. un
      2. a
      1. camión
      2. truck
      1. con
      2. with
      1. libros,
      2. books
      1. pero
      2. but
      1. no
      2. not
      1. sé
      2. I.know
      1. en
      2. in
      1. qué
      2. which
      1. camión
      2. truck
      1. cargó
      2. loaded
      1. los
      2. the
      1. libros.
      2. books
      1. ‘Sonia loaded a truck with books, but I don’t know onto which truck she loaded the books.’
      1.  
      1. b.
      1. Sonia
      2. Sonia
      1. cargó
      2. loaded
      1. algo
      2. something
      1. en
      2. in
      1. el
      2. the
      1. camión,
      2. truck
      1. pero
      2. but
      1. no
      2. not
      1. sé
      2. I.know
      1. con
      2. with
      1. qué
      2. what
      1. cargó
      2. she.loaded
      1. el
      2. the
      1. camión.
      2. truck
      1. ‘Sonia loaded something onto the truck but I don’t know with what she loaded the truck.’
    ⮭
  12. See, however, Ranero 2021 for a response to Rudin’s proposal, based on Spanish data. ⮭
  13. For more specific proposals and detailed descriptions of particular implementations, I refer the reader to Saab 2008, 2010, Tanaka 2011, Rudin 2019, Saab 2022, and Stigliano 2022, among others. ⮭
  14. I takedeletion to be a neutral term for any implementation of non-pronunciation of syntactic material. ⮭
  15. The constraint on deletion of any [F]- or [top]-marked constituents follows naturally from the assumption that [F]- and [top]-marked constituents are usually marked with pitch accents. ⮭
  16. The featural makeup of the [E]-features in different languages is likely subject to cross-linguistic variation. This remains as an empirical open question that requires further investigation. ⮭
  17. My proposal does not suggest that the [E]-features I have introduced here are the only ones available in Spanish. There might be other [E]-features in this language, potentially responsible for triggering different types of ellipsis, such as nominal ellipsis or other phenomena. Further research is needed to explore the full range of [E]-features in this language and their specific makeup. ⮭
  18. While the English translation of (85B) and (86B) might suggest that we are dealing with a juxtaposition of two sentences, this is not the case for the Spanish examples. Unlike in English, in Spanish there is no intonation break between the topicalized constituent and the wh-phrase, which supports the claim that these are indeed single sentences, not two separate sentences. I thank an anonymous reviewer for raising this question. ⮭

References

Aelbrecht, Lobke. 2010. The syntactic licensing of ellipsis. John Benjamins Publishing Company.  http://doi.org/10.1075/la.149.

Anand, Pranav & Hardt, Daniel & McCloskey, James. 2023. The domain of formal matching in sluicing. Linguistic Inquiry.  http://doi.org/10.1162/ling_a_00495.

Barros, Matthew. 2014. Sluicing and identity in ellipsis. Doctoral thesis. Rutgers University.

Beavers, John. 2017. The spray/load alternation. In: Everaert, Martin & van Riemsdijk, Henk (editors). The Wiley Blackwell companion to syntax. Wiley Online Library.  http://doi.org/10.1002/9781118358733.wbsyncom066.

Brugè, Laura & Brugger, Gerhard. 1996. On the accusative a in Spanish. Probus 8.1–51.  http://doi.org/10.1515/prbs.1996.8.1.1.

Büring, Daniel. 2003. On D-trees, beans, and B-accents. Linguistics and Philosophy 26.5.511–545.  http://doi.org/10.1023/A:1025887707652.

Chung, Sandra. 2013. Syntactic identity in sluicing: how much and why. Linguistic Inquiry 44.1.1–44. https://www.jstor.org/stable/23358087.

Chung, Sandra & Ladusaw, William A. & McCloskey, James. 1995. Sluicing and logical form. Natural Language Semantics 3.3.239–282.  http://doi.org/10.1007/BF01248819.

Fiengo, Robert & May, Robert. 1994. Indices and identity. MIT Press.

Hiraiwa, Ken & Kobayashi, Yoshiaki. 2020. Countersluicing. Syntax 23.3.295–312.  http://doi.org/10.1111/synt.12190.

Jacobson, Pauline. 2016. The short answer: implications for direct compositionality (and vice versa). Language 92.2.331–375.  http://doi.org/10.1353/lan.2016.0038.

Kim, Jong-Bok. 2021. Fragment questions: a direct interpretation approach. Linguistic Research 38.3.445–468.  http://doi.org/10.17250/khisli.38.3.202112.002.

Kroll, Margaret & Rudin, Deniz. 2017. Identity and interpretation: syntactic and pragmatic constraints on the acceptability of sluicing. In: Lamont, Andrew & Tetzloff, Katerina (editors). NELS 47: proceedings of the forty-seventh annual meeting of the North East Linguistic Society. Graduate Linguistics Student Association, University of Massachusetts Amherst. 177–190.

Li, Haoze. 2016. Fragment questions: deleting question items. In: Hammerly, Christopher & Prickett, Brandon (editors). NELS 46: proceedings of the forty-sixth annual meeting of the North East Linguistic Society. Graduate Linguistics Student Association, University of Massachusetts Amherst. 279–292.

Lobeck, Anne C. 1995. Ellipsis: functional heads, licensing, and identification. Oxford University Press.

López, Luis. 2012. Indefinite objects: scrambling, choice functions, and differential marking. MIT Press.  http://doi.org/10.7551/mitpress/9165.001.0001.

Merchant, Jason. 2001. The syntax of silence: sluicing, islands, and the theory of ellipsis. Oxford University Press.

Merchant, Jason. 2004. Fragments and ellipsis. Linguistics and Philosophy 27.6.661–738.  http://doi.org/10.1007/s10988-005-7378-3.

Merchant, Jason. 2013. Voice and ellipsis. Linguistic Inquiry 44.1.77–108.  http://doi.org/10.1162/LING_a_00120.

Merchant, Jason. 2019. Ellipsis: a survey of analytical approaches. In: van Craenenbroeck, Jeroen & Temmerman, Tanja (editors). The Oxford handbook of ellipsis. Oxford University Press. 18–46.  http://doi.org/10.1093/oxfordhb/9780198712398.001.0001.

Poppels, Till. 2022. Explaining ellipsis without identity. Linguistic Review 39.3.341–400.

Ranero, Rodrigo. 2021. Identity conditions on ellipsis. Doctoral thesis. University of Maryland.

Rooth, Mats. 1992. Ellipsis redundancy and reduction redundancy. In: Berman, Stephen & Hestvik, Arild (editors). Proceedings of the Stuttgart ellipsis workshop. IBM Germany. 1–26.

Ross, John Robert. 1969. Guess who? In: Binnick, Robert I. & Davison, Alice & Green, Georgia M. & Morgan, Jerry L. (editors). Papers from the fifth regional meeting of the Chicago Linguistic Society. Chicago Linguistic Society. 252–286.

Rudin, Deniz. 2019. Head-based syntactic identity in sluicing. Linguistic Inquiry 50.2.253–283.  http://doi.org/10.1162/ling_a_00308.

Saab, Andrés. 2008. Hacia una teoría de la identidad parcial en las elipsis [Toward a theory of partial identity in ellipsis]. Doctoral thesis. University of Buenos Aires.

Saab, Andrés. 2010. Silent interactions: Spanish TP-ellipsis and the theory of island repair. Probus 22.1.73–116.  http://doi.org/10.1515/prbs.2010.003.

Saab, Andrés. 2022. Grammatical silences from syntax to morphology: a model for the timing of ellipsis. In: Güne, Güliz & Lipták, Anikó (editors). The derivational timing of ellipsis. Oxford University Press. 170–224.  http://doi.org/10.1093/oso/9780198849490.003.0005.

Saab, Andrés & Lipták, Anikó. 2016. Movement and deletion after syntax: licensing by inflection reconsidered. Studia Linguistica 70.1.66–108.  http://doi.org/10.1111/stul.12039.

Saab, Andrés & Stigliano, Laura. 2023. On an undocumented type of predicate ellipsis in Spanish and its consequences for the theory of ellipsis licensing. Isogloss: Open Journal of Romance Linguistics 9.4.7.  http://doi.org/10.5565/rev/isogloss.354.

Stigliano, Laura. 2022. P-omission in ellipsis in Spanish: evidence for syntactic identity. Natural Language and Linguistic Theory.  http://doi.org/10.1007/s11049-021-09535-w.

Stigliano, Laura. 2023. Unveiling Topic-Remnant Elided Polar Questions: a new elliptical construction in Spanish and its consequences for the Identity Condition. Borealis: An International Journal of Hispanic Linguistics 12.2.191–222.  http://doi.org/10.7557/1.12.2.7237.

Tanaka, Hidekazu. 2011. Syntactic identity and ellipsis. Linguistic Review 28.79–110.  http://doi.org/10.1515/tlir.2011.003.

Torrego, Esther. 1999. El complemento directo preposicional [The prepositional direct complement]. In: Demonte Barreto, Violeta & Bosque, Ignacio (editors). Gramática descriptiva de la lengua española. Espasa Calpe España. 1779–1806.

Vicente, Luis. 2015. Morphological case mismatches under sluicing. Snippets 29.16–17.  http://doi.org/10.7358/snip-2015-029-vice.

Vicente, Luis & Barros, Matthew & Messick, Troy & Saab, Andrés. 2021. On a nonargument for cleft sources in sluicing. Linguistic Inquiry 52.4.867–880.  http://doi.org/10.1162/ling_a_00390.

Wei, Ting-Chi. 2013. Fragment question and ellipsis in Chinese. Studies in Chinese Linguistics 34.3.151–198.

Wei, Ting-Chi. 2018. Fragment questions in Mandarin Chinese: topic movement and pied-piping. Language and Linguistics 19.2.266–305.  http://doi.org/10.1075/lali.00010.wei.

Weir, Andrew. 2014. Fragments and clausal ellipsis. Doctoral thesis. University of Massachusetts Amherst.  http://doi.org/10.7275/5823750.0.