Imperatives and their interaction with VP-ellipsis in Spanish

. Imperatives


Introduction
The interaction between verb formation and other syntactic and morphological operations has been a topic under debate since at least Chomsky's (1957) theory of English morphosyntax. In this paper, I specifically focus on the interaction between imperative formation and TP-ellipsis in Spanish. I argue against the possibility of licensing a variety of head stranding TP-ellipsis via the formation of imperative verbs. In this respect, my conclusion can be read as a reaction to Martins' (1994), who argues for a generalization according to which if a Romance language exhibits verb movement to sigma (Σ), motivated by strong features on Σ itself, then it also licenses TP-ellipsis (e.g., European Portuguese and Galician). On the contrary, Romance languages that do not have verb movement to Σ (e.g., Spanish and Catalan) do not allow for this kind of ellipsis.
According to Martins "in languages like Spanish and Catalan, where VPdeletion is in general disallowed, imperative sentences constitute an exception. In this context, VP-deletion occurs. This is relevant because with respect to clitic placement, enclisis is obligatory in imperatives" (1994: 192). Putative examples of TP-ellipsis in Spanish imperative environments are the following: (1) Spanish (Martins 1994: 194)  In her analysis, it is the strong features in Σ in affirmative imperatives that trigger verb movement, licensing head stranding TP-ellipsis. Hereafter, I will argue that examples such as (1) do not involve syntactic deletion, but exophoric definite null objects. This account is based on the syntactic and discourse conditions licensing these examples, which meet the conditions of pragmatically controlled null objects.
The squib is structured as follows. In section 2, I lay out Martins' analysis of VP-ellipsis and sum up the characterization she makes of the Σ head in Portuguese and Spanish. In subsection 2.1, I present Martins' take on Spanish declarative and imperative sentences. In section 3, I argue that the data given by Martins are not examples of ellipsis, but of exophoric null objects. In the same section, I also provide the conditions licensing definite null objects in imperative sentences. Finally, in section 4, I present some concluding remarks and certain open issues.

Martins' analysis of TP-deletion
Martins' account, framed within the Minimalist Program (Chomsky 1995), takes as a starting point a clause structure that includes the functional category ΣP in its hierarchy. ΣP is merged between CP and AgrSubjP and licenses veritative operators and imperative values (Laka 1990). Therefore, given the hierarchical structure ΣP > AgrSubjP > TP > AgrObjP > VP, enclisis can be explained by postulating verb movement from V to Σ (see also Rivero & Terzi 1995) above a clitic left-adjoined to a lower functional projection, such as AgrSubjP (Kayne 1991) as can be seen in (2): 1 (2) Therefore, according to this approach, enclisis is the result of two factors: (i) strong morphological V-features on Σ, and (ii) movement from V to Σ across the clitic to delete those strong features. This movement to a high position in the structure yields the typical verb-clitic order and, at the same time, can license Σ-stranding VP-ellipsis.
Here, I will assume that this subtype of ellipsis is actually Σ-stranding TP-ellipsis. Otherwise, the adjoined clitic would remain undeleted once the ellipsis has taken place at the VP: 2 It is worth noting that the movement of V° from a complex head such as AgrSubj yields excorporation, which is, in principle, banned due to the impossibility of leaving wordinternal traces (Baker 1988). Nevertheless, Roberts (1991) observes that verb raising could provide evidence for excorporation in Dutch, for example, and according to Martins, the same seems to apply to Spanish imperatives. 2 Martins' analysis departs from ellipsis as a deletion operation. Instead, the author considers the elided constituent a verbal pro form.
Given the hierarchy of the clause assumed by Martins, imperatives in Spanish and related languages can be derived via head movement. Concretely, she assumes that the head of ΣP has a strong feature [*V] in imperative clauses and attracts the verb from its position in V to Σ above the clitic left-adjoined to TP. Hence, on one hand, enclisis surfaces and, on the other, Σ-stranding TP-ellipsis is properly licensed: This analysis generalizes in other Romance languages that also feature enclisis in affirmative declarative clauses, like European Portuguese and Galician. In these languages, the correlation between enclisis and Σ-stranding TP-ellipsis is attested in short answers to yes/no questions. Both in (4) and (5), Portuguese and Galician, respectively, show an enclitic pattern in questions (4b)-(5b) and Σ-stranding TPellipsis in yes/no short answers (4c)-(5c).
(4) European Portuguese (Martins 1994 In short, yes/no questions can be answered by an affirmative or negative word plus the finite verb or just by repeating the finite verb, as shown by the optionality of (4c) and (5c). In Martins' analysis, if the answer consists of a single finite verb, the verb moves to Σ to check strong morphological features on Σ itself. However, if the answer includes an affirmative or negative word plus a verb, as in European Portuguese, the former (sim) adjoins ΣP and the verb moves to Σ (see Holmberg 2001 for an alternative hierarchical structure and further details).
In contrast, Romance languages that show proclitic patterns as a consequence of weak morphological V-features of Σ do not allow for this kind of ellipsis in declarative environments: Spanish (Martins 1994 gave.1SG 'Yes, I gave it to him/her.' Now, Portuguese and Spanish pattern alike in negative clauses: the clitic always precedes the verb and follows negation. Thus, while Spanish keeps its proclitic pattern (7a), Portuguese displays an asymmetric syntactic behavior between affirmative clauses with the enclitic pattern (4d) and negative ones with the proclitic pattern (8b). As such, the negative word não is placed in Σ, so the verb cannot move further than T: Spanish, (Martins 1994 In negative clauses in Portuguese, given that Σ has an overt lexical item (não), the movement of the verb to Σ does not take place due to the fact that "V-movement can be delayed until LF even though the relevant head contains strong features" (1994: 203). As a result, the Neg-CL-V order arises. However, in Spanish affirmative and negative clauses have weak features in Σ and the verb stays in T, yielding proclisis (7a).
Given Martins' generalization, the presence of strong morphological features in both affirmative and negative clauses in Portuguese properly licenses ellipsis: Portuguese (Martins 1994: 190)  As observed in (9) and (10), European Portuguese has Σ-stranding TP-ellipsis in affirmative and negative clauses because of the presence of strong morphological [*V] features in Σ. In affirmative clauses, feature checking is the result of verb movement to Σ (10ai-bi), while in negative clauses, it is the presence of the negative word that checks them, properly licensing the ellipsis (10aii-bii). 3 As I said, Spanish 3 In contrast, Galician, which according to Martins, only has strong V-features in Σ in affirmative sentences, does not license VP-deletion in negative sentences: (1) Galician (Martins 1994: 188) a. Comiche-lo bolo?
'Did you eat the cake?' does not have strong V-features in affirmative or negative clauses. In consequence, VP-deletion is not allowed in the relevant declarative environments: (11) a. ¿Compraste el regalo de Marisol? 'Did you buy Marisol's present?' i. *Compré.
Martins' theory of verb formation in declarative clauses can be represented as follows:

Portuguese Spanish
Affirmative +V strong +V weak Negative +V strong +V weak As I just mentioned, this characterization in terms of feature strength aims at explaining how movement is triggered in each language and, as a consequence, how enclisis and proclisis arise.

About TP-ellipsis in Spanish imperatives
The asymmetry in clitic placement between negative and affirmative clauses displayed in Portuguese (8) is also attested in Spanish (12). However, the asymmetry is established between indicative clauses with proclitic pattern (7a) and imperative ones with obligatory enclitic pattern: (12) Spanish (Martins 1994: 194) a. In contrast, imperative negative clauses exhibit the expected proclitic pattern: (13) Spanish (Martins 1994: 194)

Indicative Imperative
Affirmative +V weak +V strong Negative +V weak +V weak As noted with respect to the pattern in (1), repeated below as (14), Martins claims that the active presence of strong features in Σ in affirmative imperatives triggers verb movement above the clitic, as shown in Figure 1, licensing head stranding TP-ellipsis: (14) Spanish (Martins 1994: 194)  All translations are mine.

5
Although Martins argues that a ditransitive verb such as dar allows deletion of all verbal complement, this example is really odd for me. While dame is good, da seems to be ungrammatical. The same happens with other ditransitive verbs such as prestar ('to lend') or recomendar ('to recommend') in absence of their complements: (1) a. *Prestá. 'Lend (it to her/him).' b. *Recomendá.
'Recommend (it to him/her).' The impossibility of having structures such as (1) in Spanish is evidence in favor of the absence of VPE (VP-ellipsis). In contrast, languages like European Portuguese or Capeverdean, which have VPE, allow deletion of all the complements of ditransitive verbs: Conversely, the presence of weak [*V] features in imperative negative clauses has two consequences. First, it results in the emergence of the proclitic pattern previously observed in (12) and second, in the absence of ellipsis as can be seen in (15): (15) Spanish (Martins 1994: 194)  In the next subsection, I will argue that Martins' conclusion does not hold mainly because examples such as (14) do not involve Σ-stranding TP-ellipsis, but exophoric definite null objects.

Exophoric null objects
Hankamer & Sag (1976) identify two types of anaphoric phenomena: surface anaphora, in which there is deletion of syntactic structure (16), and deep anaphora, in which there is a null pronoun without internal structure (17).
(16) a. VP-ellipsis, Portuguese, (Cyrino & Matos 2002: 178) A Ana já tinha lido o livro à irmã the Ana already had read the book to-the sister mas a Paula não tinha lido o livro à irmã. but the Paula not [-] 'Ana had already read that book to her sister, but Paula had not.' b. TP-ellipsis, Spanish (Saab 2008: 43) Juan fue al cine y María fue al cine también. 'John went to the cinema and Mary did too.' (The translation is mine) The facts in (17c) are of importance. Here, I will also take indefinite null objects in Spanish as instances of deep anaphora (cf. Campos 1986 andVerdecchia 2022). For the present section, it is worth noting that the data presented below not only covers phenomena of Rioplatense Spanish, but other varieties of Spanish as well.
All the deep anaphora examples in (17) are anaphoric null objects (NO [an] ), which are licensed by a linguistic antecedent. Furthermore, with respect to Martins' sentences in (14), following Masullo (2017), I claim these cases do not involve head stranding ellipsis nor NO [an] but exophoric null objects (NO [ex] ), which can only be pragmatically licensed through ostension. These pragmatically controlled null objects are not restricted to imperative contexts, such as those in (18) below; indeed, they are massively attested in other non-imperative sentences, whenever there are ostensive or strongly typified environments: (18) Spanish (Masullo 2017: 55-57) 7 a. ¡Tené (me)! (e.g., el paquete) 'Hold (for me).' (e.g., the package) b. Bajá, que no oigo. (e.g., el volumen) 'Turn (it) down. I can't hear.' (e.g., the volume) c. Médico: A ver, abrí. (e.g., la boca) 'Doctor: Let's see. Open (it).' (e.g., the mouth) In interrogative sentences, they are also licensed: (20) a. ¿Cerraste? (e.g. la puerta) 'Did you close (the door)?' b. ¿Escuchaste? (e.g., la noticia) 'Did you hear (the news)?' (20) were instances of V-stranding TP-ellipsis, it would be possible to have interrogative sentences with enclitic patterns as in (21). The fact that the sentences in (21) are ungrammatical shows that clitic placement seems to be irrelevant when it comes to definite null objects licensing. 8 Finally, as Masullo also shows, in discourse structures that have a definite linguistic antecedent, exophoric null objects are not licit.  ) and (24) sharply contrast with those in (22) and (23) in that contextual salience is a requirement for definite null objects. Indefinite null objects, instead, can also be linguistically licensed by an antecedent in the linguistic discourse. In this regard, Spanish does not license sentences like (23) with definite anaphoric null objects even in imperative contexts. This follows again from Masullo's  (14) with those in (24), which are perfectly grammatical and felicitous once (i) we add the required contextual information, and (ii) we add an act of demonstration / ostension:  (15)), I would like to note that the same specific discourse conditions that make (24)  This is not the expected scenario under Martins' analysis. On the exophoric null object analysis, instead, the pattern in (25) above is fully expected. Moreover, we also predict that clitic placement is irrelevant: to the extent that the discourse conditions are met, definite null objects can easily occur in non-imperative sentences, such as (19) and (20), or in negative imperative sentences, which trigger proclisis, not enclisis.

Note that if the examples in
I would like to conclude by mentioning an additional piece of evidence for the analysis defended here. As shown in (26) below, there are no animate NOs [ex] , a fact that reinforces my hypothesis, since, as is well known, null objects in Spanish are licit only when they are specified as inanimate (see also Goldberg 2005 As shown in figure 4, it is only possible to have null objects in imperative ostensive contexts. Moreover, anaphoric null objects of Spanish follow the animacy restriction and cannot be definite or indefinite in imperative clauses. Here, I sum up the conditions discussed so far:

Conclusions
I have reassessed Martins' generalization that verb head movement to Σ licenses verbal predicate ellipsis, which I assume as instantiating Σ-stranding TP-ellipsis. Contra her analysis I have shown that Spanish does not have this kind of ellipsis in affirmative imperative contexts; the grammatical silence we observe surfaces in particular instances of ostensive but definite null objects which are subject to pragmatic conditions.

10
Although Martins argues that a ditransitive verb such as dar allows deletion of all verbal complement, this example is really odd for me. While dame is good, da seems to be ungrammatical. The same happens with other ditransitive verbs such as prestar ('to lend') or recomendar ('to recommend') in absence of their complements: (3) a. *Prestá. 'Lend (it to her/him).' b. *Recomendá.
'Recommend (it to him/her).' The impossibility of having structures such as (1) in Spanish is evidence in favor of the absence of VPE (VP-ellipsis). In contrast, languages like European Portuguese or Capeverdean, which have VPE, allow deletion of all the complements of ditransitive verbs: As a final remark, I would like to add that, while the reaction of this squib was directed to Martins' own proposal (although similar criticism extends to other recent accounts like Thoms 2018), it remains to be answered which are the necessary and sufficient conditions that allow verbal stranding ellipsis in Romance and beyond.