How much room for discourse in imperative? The lens of interface on English, Italian and Spanish*

HOW MUCH ROOM FOR DISCOURSE IN IMPERATIVE? THE LENS OF INTERFACE ON ENGLISH, ITALIAN AND SPANISH* Mara Frascarelli &  Angel L. Jim enez-Fern andez 12 Abstract. This paper discusses root phenomena in imperative clauses, assuming as diagnostics conversational dynamics and the type of discourse categories that are admitted in their C-domain, through a systematic comparative interface investigation in three languages (English, Italian and Spanish) based on an original experimental work. This novel perspective sheds new light on the syntaxsemantic mapping and the interface (syntax-prosody) properties of imperative clauses, embedding the relevant proposal in a cartographic framework of analysis. Based on a twofold distinction of root phenomena –those which are widely allowed in Common Ground-active (Type I) contexts and those which can occur in non-Common Ground-active contexts (Type II) –it is proposed that imperatives are non-Common Ground-active propositions with no update potential, thus allowing only Type II root phenomena. Syntactically, imperative clauses are dominated by a super-ordinate Speech Act Phrase, including the SPEAKER and the ADDRESSEE as co-arguments, which explains the blocking effects identified in imperatives. 1. Imperative Clauses, Root Phenomena and Information Structure: introduction and Working Hypothesis A number of recent works have examined the composition and extent of phrasal hierarchies of different clause types, distinguishing between root, root-like and (diverse types of) embedded clauses, and relevant phenomena (cf., among others Haegeman 2002, Meinunger 2004, Heycock 2006). Furthermore, recent works on the interface and structural properties of discourse categories led to a clause-related distinction for different types of Topics (cf. B€ uring 1999, Haegeman 2004, Frascarelli 2007, Krifka 2007, Bianchi & Frascarelli 2010, Frascarelli & Hinterh€ olzl 2007, 2016, Jim enez-Fern andez & Miyagawa 2014), Foci (cf., among others, Kiss 1998, ^ Ambar 1999, Krifka 2007, Leonetti & Escandell 2009, Cruschina 2011, Bianchi & Bocci 2012, Bianchi 2013, Jim enez-Fern andez 2015a,b) *Earlier versions of this paper have been presented at the “Cambridge Comparative Syntax Meeting” (UK), the “26 th Generative Linguistics in the Old World” in Gottingen (Germany), the 24 th “Colloquium on Generative Grammar” in Madrid (Spain) and the 1 st International Workshop on the “Interface of Information Structure and Argument Structure” in Seville (Spain). We are very grateful to the audiences there. We also thank two anonymous STUL reviewers for their precious comments and suggestions. The research here has been partially funded by research project PGC2018-093774-B-I00 of Spain’s Ministry of Science, Innovation, and Universities (MICINN). The order of authors is strictly alphabetical, so both of us are first authors. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 Studia Linguistica ••(•) 2020, pp. 1–61. ©2020 The Editorial Board of Studia Linguistica Published by John Wiley & Sons Ltd, 9600 Garsington Road, Oxford OX4 2DQ, UK, and 350 Main Street, Malden, MA 02148, USA S T U L 12153 Dispatch: 28.12.20 CE: Wiley Journal Code Manuscript No. No. of pages: 61 PE: Rajasekaran S. and Contrast, often associated with either Focus or Topic, but also as an independent feature (Vallduv ı & Vilkuna 1998, Moln ar 2006, Bianchi & Bocci 2012, Frascarelli & Ramaglia 2013, Bianchi 2013). To the best of our knowledge, however, the data examined in this respect mainly concern declarative or interrogative clauses, while no study of the socalled ‘left periphery of the sentence’ (cf. Rizzi 1997 and subsequent, related works) has been specifically proposed for imperative clauses. Dislocated discourse categories such as Topics are usually associated with the presence of illocutionary assertive force, as a root property. From this perspective, imperatives are expected not to allow discourse categories that are connected with the root character of the latter. Accordingly, Haegeman (2012:64) provides the examples in (1) (see also Haegeman 2010 for additional similar examples): (1) a. *Your essay, leave in my pigeon hole this afternoon. b. *The weapons leave behind. However, the empirical puzzle emerges when we are confronted with imperatives which do permit topics in their left periphery, as in (2), also from Haegeman (2012: 64): (2) a. The tie give to Bob, the aftershave give to Don. (Van der Wurff 2007) b. Anything you don’t eat put back in the fridge. (Huddleston & Pullum 2002) This raises the question as to what properties imperatives must satisfy in order to license or not certain types of discourse categories. A plausible solution to this problem can be offered if we assume a distinction between different types of Topics. In this respect, Frascarelli & Hinterh€ olzl (2007) propose a typology of topics based on the systematic correlation between their formal properties and their function in discourse, which is encoded in dedicated functional projections in the left periphery of the sentence. The authors thus produce substantial evidence for the existence of (at least) three types of Topics, namely Aboutness-Shift (A-)Topics, Contrastive (C-)Topics and Familiar/Given (G-)Topics (cf. §3 below for details; see Cardinaletti 2009 and Menza 2006 for the distinction between A- and G-Topics in different constructions). Based on this distinction, a preliminary observation can immediately lead to the conclusion that shifting (i.e. A-)Topics can hardly be 2 Mara Frascarelli &  Angel L. Jim enez-Fern andez 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 ©2020 The Editorial Board of Studia Linguistica associated with the imperative mood. Consider for instance (3a-b) below, from Italian: 1 (3) a. Sono le quattro: vai subito a fare i be.3PL the four go.IMP.2SG immediately to to.do the compiti! homework ‘It’s four o’clock: go and do your homework immediately!’ b. *Sono le quattro: i compiti, vai subito a be.3PL the four the homework go.IMP.2SG immediately to farli! to.do-them On the other hand, C-Topics (4B) and G-Topics (5) seem to be allowed (both in Italian and Spanish): (4) A. Dove posso mettere questi fiori? (Italian) ¿D onde puedo poner estas flores? (Spanish) where can.1SG put these flowers ‘Where can I put these flowers?’ B. a. Le rose, mettile nel vaso, b. Las rosas, ponlas en el jarr on, the roses put.IMP.2SG-them.CL in the vase il girasole lascialo sul tavolo. (Italian) el girasol d ejalo sobre la mesa. (Spanish) the sunflower leave. IMP.2SG.it.CL on the table ‘The roses put in the vase, the sunflower leave on the table.’ (5) [Context: A: You cannot play soccer, always keeping the ball with you...] a. ...la palla, passala anche agli altri! (Italian) b. ...la pelota, p asala tambi en a los otros! ...the ball pass-IMP.2SG-it also to the others ‘...pass the ball also to the others!’ 1 The list of abbreviations and symbols used in the article is the following: ALL =allocutive, AT =Aboutness-Shift Topic, CL =clitic (pronoun), CLLD =clitic left dislocation, CF = Corrective Focus, CG =Common Ground, CGr=Control Group, CT =Contrastive Topic, DIR =directive feature, DP =determiner phrase, F=feminine, FINP =Finiteness phrase, F 0 = Fundamental frequency, FORCEP =Force phrase, GT =Given Topic, IMP =imperative, IMPERS =impersonal, IND =indicative, IP =Inflectional phrase, IRR =interface root restriction, JUSS =jussive, LD =left dislocation, MF =Mirative Focus, NEG =negative, PL = plural, PST =past tense, REFL =reflexive, SAP =Speech act phrase, SG =singular, SUB = subordinate clause, SUBJ =subjunctive, TBU =tone bearing unit, TOP =topicalization, TP = Tense phrase, vP=Verb phrase. The present tense is assumed as deafult and, as such, it is not indicated in the glosses. Notice that throughout the paper, the same gloss will be provided for Italian and Spanish sentences when these are identical; otherwise, a different gloss will be given for each language. How much room for discourse in imperative? 3 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 ©2020 The Editorial Board of Studia Linguistica However, as translations show, English appears to provide some crosslinguistic differences. Cormany (2013) argues that non-contrastive topics are not allowed in English 2 and, in general, left-peripheral arguments are often unacceptable (from Jensen 2007), as illustrated in (1) above. Nevertheless, this is not absolute. The cases considered in (2a–b) and (the translations in) (4) show that a contrastive interpretation for the topic constituents, obtain acceptable results; hence, C-Topics can be realized in English imperative clauses. As for sentence (5), the translation shows that ‘the ball’ would be left in situ by speakers in this context. As a matter of fact, la palla/la pelota cannot be considered an A-Topic, because its mention is not used to introduce (or shift) the sentence Topic. Indeed, ‘the ball’ can be considered a background/given element in the context of a soccer game. Hence, it is a G-Topic in the framework assumed in this work and G- Topics are not dislocated in English, but rather realized through in situ destressing (as is argued in Bianchi & Frascarelli 2010). Similar to Topics, different types of Foci have been argued to exist in the literature, with different formal and discourse properties (cf., among others, Kiss 1998, Aboh, Hartmann & Zimmermann 2007, Frascarelli 2010a, Cruschina 2011, Bianchi, Bocci & Cruschina 2015). In particular, Bianchi & Bocci 2012 argue for a syntactic and semantic distinction between Mirative and Corrective Foci and, interestingly, the realization of these two Focus types in imperative clauses also obtains different results at a first observation. As for Mirative Focus (MF), the contrast between the declarative sentence (6a) and sentence (6b) immediately show that MF ‘clashes’ with the imperative mood (for reasons that are left open here, but will be fully explained later in the paper). On the other hand, Corrective Focus (CF) looks unproblematic as long as the focused element remains in situ (7-a- a’), whereas it is unacceptable if fronted (7b’-b’): 2 Cormany (2013:100-101) specifically argues that in English only contrastive topics are allowed in imperative clauses, in clear contrast with declaratives (in which both contrastive and non-contrastive topics are licensed), as is illustrated in (i)-(ii): (i) a. The book, John bought ___. b. *The book, buy ___! (ii) a. These stocks, the broker bought immediately. b. These stocks, buy immediately! (Those avoid at all costs!) This contrast is adduced to a relatively poor left periphery in imperative clauses (similar to Haegeman’s 2004 notion of ‘truncation’). As we will see in Section 5 and 6, however, a truncation-based analysis cannot account for cross-linguistic data. In addition, truncation cannot account for the grammaticality difference of examples in (1)-(2) showing that, under certain circumstances, topics can occur in the left periphery of imperatives. 4 Mara Frascarelli &  Angel L. Jim enez-Fern andez 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 ©2020 The Editorial Board of Studia Linguistica (6) a. Sei incredibile! be.2SG incredible Sei a dieta e....DUE PIZZE hai mangiato! be.2SG a diet and two pizzas have.2SG eaten ‘That’s incredible! You are on a diet and....TWO PIZZAS we have eaten!’ b. Sei incredibile! be.2SG incredible *Sei a dieta e....DUE PIZZE mangia! be.2SG a diet and two pizzas eat.IMP.2SG (7) a. Mangia LA CARNE, non solo le patate! (Italian) a’. ¡C omete LA CARNE, no solo las patatas! (Spanish) eat.IMP.2SG the meat not only the potatoes ‘Eat YOUR MEAT, not only fries!’ b. *LA CARNE mangia, non solo le patate! (Italian) b’. *¡LA CARNE c omete, no solo las patatas! (Spanish) the meat eat.IMP.2SG not only the potatoes The crucial characteristic that associates A-Topic and MF is that they have been argued to be strictly connected with illocutionary force. In particular, Bianchi & Frascarelli (2010) argue that an A-Topic constitutes a speech act on its own (an ‘initiating speech act’, following Krifka 2001), introduced by a dedicated root operator 3 and associated to the speech act expressed by the following sentence. Along similar lines, MF has been claimed to be dependent on illocutionary force (specifically, a ‘root evaluative force’), since it is a “proposal to negotiate a shared evaluation” (cf. Bianchi 2012). Since such a connection has not been argued for the other discourse categories discussed above, we have taken this preliminary observation as our leading working hypothesis for the present investigation. Accordingly, building on the assumption that a systematic connection exists between the formal (syntax-prosody) and semantic properties of discourse categories (cf. Frascarelli & Hinterh€ olzl 2007), this work takes information structure as diagnostics to provide a comprehensive explanation for the data shown above (cross-linguistic symmetries and discrepancies), define the root/nonroot quality of imperatives and provide a structural analysis (in a 3 The nature and properties of the relevant ‘root operator’ is not specified in Bianchi & Frascarelli (2010). Nevertheless, considering the formal and discourse properties of an A- Topic (cf. Section 2 below), this can be identified with Krifka’s (1995) Assert Operator insofar as it has the function of updating the common ground by asserting a proposition that is informative, non-contradictory and implies alternatives. As this Operator is encoded in a functional projection in the C-domain (cf. also Meinunger 2004), it is perfectly in line with the present syntax-prosody-semantic interface approach study. How much room for discourse in imperative? 5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 ©2020 The Editorial Board of Studia Linguistica cartographic perspective; cf. Rizzi 1997) which can account for the discourse-semantic properties of imperatives. 4 More precisely, the present working hypothesis can be formulated as follows: if root phenomena like A-Topic and MF cannot be hosted in imperative clauses, this can be taken as evidence that the latter are not endowed with illocutionary force. Consequently, according to the types of discourse categories that are allowed, an information-structural characterization will be proposed, which can also account for their semantic properties. To this purpose, a comparative interface investigation is proposed, dealing with the acceptability of A-Topics, C-Topics, G-Topics, MF and CF in English, Italian and Spanish imperative clauses. The analysis is based on judgments collected by means of an original online survey, and recordings of the relevant sentences (cf. Section 4.4. below for details). In addition, we complement the online rating test with a prosodic analysis of the different discourse categories. Based on a ToBI characterization (Pierrehumbert 1980) of the intonational contours associated to the sentences used for the online survey, the present interface study will check what type of intonation the informants associate to the discourse categories that they have judged as acceptable in the written test. This will be crucial to understand whether they assigned the intended interpretation or, rather, their acceptance is based on a (maybe collaborative) misinterpretation of the relevant discourse function –a risk that cannot be avoided in written tests, especially in languages like Italian and Spanish, in which all Topic types are realized through CLLD constructions while all Focus types are never clitic-resumed. Our work is organized as follows. In section 2 we discuss the ‘Root/ non-root distinction’ and illustrate the notion of ‘Root phenomena’ providing several examples for the relevant distinction in different languages. In section 3 we present the discourse categories addressed in the present investigation and address their role with respect to Conversational Dynamics. Section 4 provides a short survey on the major properties of imperative clauses, focusing on their mixed (root/non-root) character. In section 5 the interface analysis is presented: informants’ judgments are analyzed and confronted with the relevant prosodic realizations in the three languages under examination (English, Italian and Spanish). 4 The existence of a systematic correlation between the discourse and formal (syntaxprosody) properties of different types of Topics, put forth in Frascarelli & Hinterh€ olzl (2007) for Italian and German, has been checked and assumed in a number of works on typologically diverse languages like Somali (Puglielli & Frascarelli 2009), Tagalog (Frascarelli 2010b), English (Bianchi & Frascarelli 2010), Spanish (Jim enez-Fern andez 2016), Finnish and Russian (Frascarelli 2018), Old English (van Gelderen 2013), and Icelandic (Sigurdsson 2011). Based on this cross-linguistic validation, it is taken as a working hypothesis for the present analysis. 6 Mara Frascarelli &  Angel L. Jim enez-Fern andez 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 ©2020 The Editorial Board of Studia Linguistica Based on the evidence collected, a novel proposal on the syntactic mapping of imperatives is put forth in Section 6, capitalizing on Farkas & Bruce’s (2010) approach to conversational dynamics and examining the imperative structure in the light of the update potential of the discourse categories allowed in the relevant left periphery. Imperatives will be thus characterized as non-Common Ground active propositions, as we will see below. This means that, even though imperatives affect the context by issuing adirective act, this counts as an instruction, not as a piece of information. Notice that, in this respect, we follow Krifka (2007) and assume two dimensions for the Common Ground (CG) Management, distinguishing between (i) Conversational Moves and (ii) Instructions, where only the former can modify the CG Content. Therefore, to say that imperatives are instructions amounts to claiming that their conversational import implies an update of the Addressee’s TO-DO-LIST, while it does not count as an update of the interlocutor’s CG Content. In other words, ‘nothing is added to the CG’ (cf. also Roberts 2018), even though imperatives can have access to it (see Lauer 2013 for a framework in which information is updated with information about utterance events). In the cartographic approach that we assume to account for interpretive levels, this means that the assertive feature in ForceP is inactive and, as a consequence, all the projections connected with its activation (e.g., the projections hosting the A-Topic and the MF) are not available. To explain the imperative interpretation, it will be proposed that this is connected with a Modality projection, activated (through Agree) with a directive feature encoded in C. Finally, in order to account for the participants involved in this type of directive act, the imperative proposition will be argued to be selected as the object of a Speech Act Phrase including the ADDRESSEE and the SPEAKER as co-arguments. Section 7 is dedicated to a specific interface restriction concerning lowtoned (G-Topic) constituents in imperatives and Section 8 offers our final conclusions. 2. The Root/Non-root Distinction 2.1. What is a Root clause? To provide a clear-cut, uncontroversial definition of a ‘root clause’ is far from trivial. Since Emonds (1970, 1976), this quality has been attributed to clausal domains which could only host some types of phenomena (like Left Dislocation) and the relevant ‘root restriction’ has been connected to the availability of what can be defined as assertive force (cf. Bianchi & Frascarelli 2010) in these clauses. Hooper & Thompson (1973) make the point that this restriction only relies on semantic/pragmatic requirements How much room for discourse in imperative? 7 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 ©2020 The Editorial Board of Studia Linguistica and cannot be accounted for syntactically. 5 In this sense, assertion is connected with illocutionary force: As a positive environment we can say that [root] transformations operate only on Ss that are asserted. [...] some transformations are sensitive to more than just syntactic configurations. It does not seem possible to define the domain of an RT in terms of syntactic structures in any general way. However, [...], even if it were possible to define in syntactic terms the conditions under which RTs can apply, [...] the question of why these transformations can apply in certain syntactic environments and not others would still be unanswered (Hooper & Thompson 1973:495). This claim is supported by the observation that ‘root transformations’ are actually also allowed in syntactically embedded clauses whose content constitutes the main assertion: (8) It appears [that this book he read thoroughly]. (Hooper & Thompson 1973: 478) This challenge was then taken up by different authors in the generative framework, who tried to elaborate a syntactic account for the relevant restriction. 6 Emonds (1970, 1976) observed that for many speakers dependent clause contexts mimic the freedom of root structures in indirect discourse. However, root-like indirect discourse embedding is incompatible with most dependent clause positions (cf. Emonds 2004). A syntactic approach to root transformations was also adopted by Liliane Haegeman in a number of works, mainly focusing on adverbial clauses (e.g. 2006, 2010, 2012). Specifically, Haegeman’s (2006) first approach to the root/non-root issue strictly relies on a cartographic difference concerning the extension of the relevant C-domains (i.e., the functional projections composing the left periphery): root phenomena depend on the activation of ForceP, which is missing in the left periphery 5 In this respect Hooper & Thompson (1973) only state that the assertion of a sentence is ‘its core meaning or main proposition’, which ‘may be identified as the part that can be negated or questioned’. Sentences, however, may contain more than one assertion (e.g. coordination) and, crucially, some subordinate clauses are asserted (cf. Section 2.2). 6 Several works and proposals have been also elaborated by scholars operating in nongenerative, non-formal approaches, in which the restrictions on the use of particular constructions are sought in functional properties, whether cognitive, processing or discourse-functional, while leaving the basic structure of the construction for formalist analysis. In particular, for a semantic and discourse-pragmatic approach to root phenomena (often referred to using different terms), the interested reader is referred to McCawley (1977); Kuno & Kaburaki (1977), Kuno (1987), Prince (1981), Birner (1995), Verstraete (2005), Hengeveld & Mckenzie (2008), Ward, Birner & Kaiser (2017), among many others. 8 Mara Frascarelli &  Angel L. Jim enez-Fern andez 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 ©2020 The Editorial Board of Studia Linguistica of non-root clauses. Nevertheless, in later works the author herself pointed out the limits of such an approach, showing that the presence of illocutionary force per se is not a sufficient condition for root phenomena to apply. Hageman (2010, 2012) then argues that the incompatibility between root phenomena and some clausal environments depends on an intervention effect between the operator movement that is required by a given clausal type and the operation involved by the relevant root transformation. 7 Presupposed environments, for instance, are factive in nature, and factives have been argued to involve operator movement (cf, among many others Watanabe, 1993). In a similar vein, Haegeman & € Ur€ ogdi (2010), Jim enez-Fern andez & Miyagawa (2014) and Jim enez-Fern andez (2018) propose that non-root clauses contain an operator which prevents some discourse categories from moving to the C-domain. For Jim enez-Fern andez & Miyagawa (2014), some types of topics can move in non-root clauses because movement targets a position within the sentential (IP)-domain, after inheritance of the discourse feature by T(ense) from C(omp). Moreover, 7 Intervention effects have been originally explored in formal syntax in connection with weak-island phenomena (cf. Szabolcsi 2005 for an overview). For the purposes of the present analysis, we follow Rizzi’s (2004, 2018) account and assume a Relativized Minimality (RM) version of the intervention principle, according to which: (i) In the configuration [... X... Z... Y...] a local relation connecting X and Y is disrupted when there is a Z such that: (A) Z is of the same structural type as X, and (B) Z intervenes between X and Y A proper comprehension of this model thus requires a definition of what counts as an ‘intervener’. For the sake of space, we can briefly state that this notion implies that: (a) the intervening element must intervene hierarchically, not just linearly; (b) the class of interveners is featurally selective. Characterization (a) explains why from a baseline sentence like (iia) extraction of the temporal adjunct when is ill formed (cf. (iib), whereas from a baseline like (iiia) movement of when is possible, as in (iiib): (ii) a. Leo asked who you will arrive at five b. *When [did Bill ask [ who [ will arrive ___ ]]]] (iii) a. The doubt about who passed the exam ended at five b. When [ did [ the doubt [ about [ who passed the exam ]]] end ___ ]? On the other hand, characterization (b) explains why quantificational adverbials like beaucoup (‘a lot’) in French intervene in combien extraction (cf. iva), whereas manner adverbials like attentivement (‘carefully’) do not (cf. Laenzlinger 1998): (iv) a. *Combien a-t-il beaucoup consult e [ ___ de livres] ? ‘How many has he a lot consulted of books?’ (v) a. Combien a-t-il attentivement consult e [ ___ de livres]? ‘How many has he carefully consult of books?’ How much room for discourse in imperative? 9 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 ©2020 The Editorial Board of Studia Linguistica In this line of analysis, Bianchi & Frascarelli (2010:72) suggest that the C-Topic marking is used to ‘break down a complex proposition into a conjunction of simpler ones in which a predicate applies separately to each member of a salient set’. This is illustrated in the Italian dialogue below: (21) A: Dove stanno i tuoi fratelli adesso? where stay.3PL the your siblings now ‘Where are your siblings now?’ B: Carlo  e a casa come sempre, Carlo be.3SG at house as always Mario non lo vedo da una settimana. Mario not him.CL see.1SG for one week ‘Carlo is at home as usual, Mario, I haven’t seen for a week.’ C-Topics thus provide an instruction to the speaker and, as such, they do not constitute in themselves independent conversational moves. This means that, unlike A-Topics, C-Topics are not restricted to CG-active domains, though they cannot be hosted in a clause that is less than a proposition (i.e., properties or event modifiers). We thus conclude that C-Topics qualify as Type II root phenomena. This discourse-semantic qualification can be proved showing that C-Topics are allowed in complements to factive verbs (cf. (22)) and in the scope of a matrix QP (23), contrary to A-Topics (as is shown for Topicalization in English in (13) above): (22) Sono contenta che Mario venga alla festa, per o mi dispiace [che suo fratello, non voglia venire e che sua sorella, non l’abbiano nemmeno invitata] ‘I’m glad that Mario will come to the party but I’m sorry that [his brother doesn’t want to come, and his sister, they have not invited]’. (23) Ogni mio amico mi ha detto [che un esame, lo posso preparare in due settimane e un concorso, lo posso preparare in due mesi] (∀>∃) ‘Every friend of mine told me [that one exam, you can prepare in two weeks, and a competition, you can prepare in two months].’ Syntactically, the C-Topic is located in a position in the left periphery that is lower than the position hosting the A-Topic whereas, prosodically, it is characterized by a high tone that is aligned on the tonic vowel (H*), which is realized as a complex tone (L+H*) in a number of languages (the so-called ‘B-Accent’, cf. B€ uring 2003). 3.3. The Familiar/Given (G-)Topic G-Topics refer to given information that is somehow salient to the conversation. Since givenness is calculated on the basis of the CG 16 Mara Frascarelli &  Angel L. Jim enez-Fern andez 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 ©2020 The Editorial Board of Studia Linguistica content, it can be said that the main function of a G-Topic is the retrieval of given information that is stored in the CG repository. This retrieval can be operated either via ‘strongly given’ constituents in the discourse (i.e., definites triggering an identificational presupposition; cf. Heim 1982), or through ‘familiar’ discourse referents that are (i) perceptually accessible in the utterance, (ii) globally familiar from shared cultural knowledge, (iii) contextually entailed to exist, or (iv) implied via ‘bridging’ 13 (cf. Roberts 2003). An illustration of the distinction between a strongly given Topic (24) and a familiar G-Topic (25) is provided in the following dialogues (from Frascarelli & Hinterh€ olzl 2016): (24) A: Adoro il pollo, perch e si mangia con le adore.1SG the chicken because IMPERS eat.3SG with the mani. hands ‘I love chicken, because you can eat it with your hands’ B: Anch’ io! Anche se il pollo non lo mangio molto also I even if the chicken not it.CL eat.1SG very spesso. often ‘Me too! Even if I don’t very often eat chicken.’ (25) A: Adoro il cibo che si mangia con le mani. adore.1SG the food that IMPERS eat.3SG with the hands ‘I love that (kind of) food that can be eaten with your hands’ B: Anch’ io! Anche se il pollo non lo mangio molto also I even if the chicken not it.CL eat.1SG very spesso. often ‘Me too! Even if I don’t very often eat chicken.’ The retrieval function of G-Topics can serve either (a) to ensure topic continuity, that is to say, to maintain the current A-Topic through one or more low-toned copies in the topic chain created by the A-Topic itself; this is the function of the so-called Aboutness G-Topic; or (b) to mention a constituent that is part of the background, but it is not a copy of the current A-Topic and, importantly, it is not proposed for a topic shift (this is defined as Background G-Topic; for discussion, cf. Frascarelli 2017, 2018, and Frascarelli & Jim enez-Fern andez 2019). 13 ‘Bridging’ inferences refer to the resolution of indirect anaphoric relationships between entities in a text or a discourse, to be drawn by recipients in order to make sense of the linguistic input, which often does not fully specify the intended meaning of utterances. In this process, contextual knowledge plays a crucial role for a successful interpretation of the intended meaning of specific texts and discourses. How much room for discourse in imperative? 17 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 ©2020 The Editorial Board of Studia Linguistica Consider for instance the following passage and, in particular, the sentence underlined at the end: (26) I gladiatori k [A-Topic] entravano nell’arena sfilavano, salutavano gli spettatori e salutavano soprattutto l’imperatore, poi si recavano davanti alla tribuna [...] e c’era l’arena, che era praticamente un tavolato di legno sul quale veniva buttata della sabbia esu questa (Background G-Topic), loro k ,i gladiatori k (Aboutness G-Topics) lottavano. ‘The gladiators would enter the arena, (they) marched, (they) greeted the public, and (they) hailed the emperor first of all, then (they) used to go in front of the gallery [...] and there was the arena, which was practically a wooden stage covered with sand and on this they - the gladiators - would fight’ As is clear, the underlined pronoun loro and the DP i gladiatori are used to resume (and maintain) the current A-Topic, introduced at the beginning of this passage and heading a long chain of null subjects (cf. Frascarelli 2007, 2017; for similar cases in Spanish, see also Jim enez- Fern andez 2016). Hence, they are part of the current topic chain and qualify as Aboutness G-Topics. On the other hand, the left-dislocated PP su questa refers to ‘the arena’: an entity that is part of the background and was previously introduced in the text. However, its mention is not intended for a Topic shift: the following discourse keeps on talking about the gladiators. The arena thus qualifies as a Background G-Topic. 14 Syntactically, the G-Topic is merged in the lowest Topic position in the left periphery and its tonic vowel is realized as a low tone at the minimum of the speaker’s range, with no significant difference in comparison with the pretonic F 0 level (i.e., as a L*tone). Contrary to A- and C-Topics, G- Topics can be multiple and can appear in the right periphery of the sentence (always assuming Merge in the left periphery and deriving its final position through IP-inversion; for details, cf. Frascarelli 2000, Cardinaletti 2002). Since in our system givenness is calculated on the basis of the CG content, G-Topics clearly do not instantiate a conversational move and do not depend on illocutionary force. This implies that the G-Topic can be hosted in non CG-active propositions. Consider, for instance, the 14 Unfortunately, the limitations imposed by a written text do not allow the illustration of the prosodic contours and space restrictions impose a short exposition of the relevant contexts. In the absence of intonational cues, it is therefore important to remember that the crucial distinction between A- and G-Topics rests on the shifting function of A-Topics (whereas they share the ‘aboutness’ quality). Hence, in (26) the PP su questa is undoubtedly a G-Topic because the conversation does not shift to ‘the arena’: the speaker keeps on talking about loro (i.e., ‘the gladiators’, which is the current A-Topic). Mutatis mutandis, la vita dei romani in (17) above cannot be a G-Topic because the conversation shifts to that discourse topic after its introduction. See Frascarelli & Jim enez-Fern andez (2019) for further refinements on the role of G-Topics in topic continuity. 18 Mara Frascarelli &  Angel L. Jim enez-Fern andez 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 ©2020 The Editorial Board of Studia Linguistica Spanish example (27), in which a G-Topic is realized within a central adverbial clause (from Jim enez-Fern andez & Miyagawa 2014:298), which is argued not to be endowed with illocutionary force (cf. Haegeman 2002 and subsequent, related work): (27) Julia no ha entregado la tesis en el departamento Julia not have.3SG submitted the thesis in the department porque el cap ıtulo final no lo ha terminado because the chapter final not it.CL have.3SG finished todav ıa. yet ‘Julia has not submitted her thesis to the Department because she has not finished her final chapter yet.’ 3.4. The Mirative Focus (MF) Brunetti (2009) and Cruschina (2011) consider Mirative Focus (MF) as not purely ‘informative’. Specifically, in Bianchi (2013), Bianchi et al. (2015) and Jim enez-Fern andez (2015b), it is argued that MF provides new information and, based on the speaker’s knowledge of the hearer’s expectations, it signals that such information will be unexpected. Usually, this unexpectedness is an indication of the speaker’s surprise with respect to the fact that is being described. Consider the Italian example below (here and in the following examples small caps are used to indicate Focus): (28) Non ci posso credere! DUE BOTTIGLIE ci siamo not in.it can.1SG believe.INF two bottles REFL be.1PL bevuti! drunk ‘I can’t believe it! TWO BOTTLES we drank!’ In contrast with Information Focus, Bianchi (2013, 2015) underlines that MF does not depend on a question-answer context. Therefore, in a sentence like (28) the focused constituent does not satisfy any request in any possible question. Rather, a contrast is established with an element that is part of the CG repository of the participants and can be semantically described as a ‘proposal to negotiate a shared evaluation with the hearer(s)’ (Bianchi 2015: 64). Concentrating on this characterization, MF qualifies as a Type I root phenomenon since it depends on illocutionary force and implements a conversational move. From an intonational point of view, a fronted MF is characterized by a H*tone (cf. Bianchi et al. 2015, Frascarelli & Ramaglia 2013). 3.5. The Corrective Focus (CF) Differently from Contrastive Focus (with which it is often confused), the Corrective Focus (CF) is not just the rejection of a (question-answer) How much room for discourse in imperative? 19 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 ©2020 The Editorial Board of Studia Linguistica congruent alternative in a given set. Rather, a CF conveys an incompatible description of one and the same event (cf. van Leusen 2004, Bianchi & Bocci 2016). Specifically, using Bianchi & Bocci’s (2012:6) words: “There is one alternative proposition, already introduced in the context, which is incompatible with the proposition expressed in the corrective reply. The incompatibility import implies that accepting both the antecedent proposition and the corrective claim would lead to an inconsistency: this is what gives rise to the correction effect, whereby speaker A’s assertion is rejected by speaker B.” Syntactically, a CF can be realized both fronted (in a position that is generally defined as higher than the G-Topic, but lower than A- and C- Topics) and in situ, as is shown in the following example: (29) A: I heard you met Fred yesterday. B: No, BILL I met / No, I met BILL’ Given this characterization, CF clearly qualifies as a conversational move and, as such, it requires to be hosted in a CG-active proposition. Hence it is a Type I root phenomenon. Phonologically, a fronted CF is cross-linguistically marked by a pitch reaching high values, which is preceded by a low tone L+H*(Bianchi et al. 2015). As for the in situ realization, different prosodic characterizations can be found in the literature. Specifically, taking elicited sentences as a basis, Bianchi et al. (2015) argue for a L+H*realization (exactly as in the fronted option), while Frascarelli & Ramaglia’s (2013) study provides evidence that an in situ CF is not pitched, but integrated in the downgrading contour of the sentential Intonational Phrase. 4. Major Properties of Imperative Clauses: a Brief Survey In this section we briefly present the main traits suggested in the literature to describe imperatives. It is needless to say that we cannot make justice to all the studies that have concentrated on the syntax and semantics of these constructions, especially when the goal of our work is to explore the information structure of imperatives and the consequences for the multifaceted analysis that we propose. 4.1. Negation and tense From a morphological viewpoint, there is no general consensus on the finite/ non-finite quality of the relevant verb form; hence, a clear-cut characterization can be hardly assessed. Insome cases the imperative form is expressed by means of some kind of finite verbal form, which is specifically identified as such across languages. This is the case of two of the three languages under examination here, namely Spanish (30) and Italian (31): 20 Mara Frascarelli &  Angel L. Jim enez-Fern andez 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 ©2020 The Editorial Board of Studia Linguistica (30) Ven pronto! come.IMP.2SG soon ‘Come soon!’ (31) Aspetta un attimo! wait.IMP.2SG a minute ‘Wait a minute!’ On the other hand, some languages seem to use the base form (English) or the participle (Dutch, classical Hebrew). However, the imperative can also be expressed through a finite form, namely the subjunctive, in languages like Greek and French (and partially in Spanish and Italian as well). This variety has opened a vivid discussion among scholars concerning the inflectional status of the imperative form, which cannot be treated in detail in the present paper. Nevertheless, since the root/non-root distinction is a major topic in the present analysis and the internal structure of imperatives will be addressed in our proposal, some further discussion is needed. In this respect, we consider as crucial the evidence provided by negative imperatives. In particular, Zanuttini (1996) convincingly shows that the realization of sentential negation in imperatives correlates with the inflectional form of the verb, Tense playing a relevant role (cf. also Rivero & Terzi 1995, Zanuttini 1996, 1997, Zeijlstra 2006). Specifically, Zanuttini (1996: 188) argues that a ban on ‘true’ negative imperatives is imposed in languages like Italian and Spanish simply because in the split- IP hierarchy the functional projection in which Negation is encoded (NegP) selects TP (i.e., the head of NegP takes TP as its complement) and imperatives lack TP, but not agreement. For this reason, the imperative form (32a) cannot be used in the presence of negation, and a suppletive form emerges (32b). On the other hand, it can be used with the forms that are identical to the indicative (33a-b). Similarly, Spanish does not allow the presence of the negative head no with second person imperatives (34a- b) and expresses the negative imperative using the subjunctive (35a-b), extracted from Zanuttini (1996:188-189): (32) a. Cammina! b. *non cammina! / non camminare! walk.IMP.2SG not walk.IMP.2SG not walk.INF ‘Walk!’ ‘Don’t walk! (33) a. Non camminate! b. Non camminiamo sopra l’ erba. not walk.IMP.2PL not walk.IMP.1PL on the grass ‘Don’t walk (2PL)!’ ‘Let’s not walk on the grass.’ (34) a. Habla! / *no habla! b. Hablad! / *no speak.IMP.2SG not speak.IMP.2SG speak.IMP.2PL not hablad! speak.IMP.2PL ‘Speak!’ ‘Speak (2PL)!’ How much room for discourse in imperative? 21 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 ©2020 The Editorial Board of Studia Linguistica (35) a. No hables! b. No habl eis! not speak.SUBJ.2SG not speak.SUBJ.2PL ‘Don’t speak!’ ‘Don’t speak (2PL)!’ We assume Zanuttini’s analysis and therefore consider the imperative as a finite defective form, lacking TP, but featuring both Mood and some form of agreement in the I-domain. This assumption will be supported by the present analysis and discussed in the structural proposal later. In other words, we assume that the IP of true imperatives (those with morphological forms unique to imperatives; cf Rivero 1994) is only endowed with agreement features. This is supported by the incompatibility of negation and true imperatives, assuming that NegP selects TP in these languages (as is argued in Zanuttini 1996; cf. above). 15 With respect to the syntactic properties, one of the main concerns in the analysis of imperatives is whether they exhibit root or non-root properties, since, though superficially independent, some evidence supports their ‘embedded’ qualification. As a matter of fact, imperatives show subordinate-like morpho-syntactic properties in languages like Latin. Considering negation once more, for instance, imperative verbs require a subordinate negative marker (cf. Evans 2007): 16 (36) a. Venias! come.SUBJ.2SG ‘Come!’/ ‘May you come!’ b. Non / *ne venias! not(SUB) not(ROOT) come.SUBJ.2SG ‘May you not come!’ The present investigation is intended to shed light on this issue, focusing on the properties of imperatives’ left periphery. 15 However, one problem that is posed by the lack of T is that temporal adverbs such as now and later are fully compatible with true imperatives, as shown in (i): (i) Hazlo ahora/despu es. (ii) *Hazlo antes. do.IMP.2SG-it now/later do.IMP.2SG-it Before ‘Do it now/later.’ ‘*Do it before.’ This compatibility points to the possibility that some sort of Tense is present in imperatives. A plausible solution can be assuming the Split TP system proposed by Sigurdsson (2016), according to which the T-system is divided into three dedicated projections, namely Utterance Time, Reference Time and Event time. Provided that the temporal adverbs allowed here are connected with the utterance time, there may be an Utterance Time Phrase in imperatives, whereas other T projections are not realized (indeed, (ii) above shows that a Reference Time projection should be missing). 16 The parameters of negation in Latin have been argued to display a complex pattern of interdependencies, the illustration of which is totally beyond the purposes of this paper. For an overview of this question and relevant references the interested reader is referred to Gianollo (2016). 22 Mara Frascarelli &  Angel L. Jim enez-Fern andez 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 ©2020 The Editorial Board of Studia Linguistica 4.2. Anaphoric relations, subjects and topics If Tense is generally assumed to be lacking in imperatives, a major question concerns agreement and the realization of subjects. Actually, true imperatives show 2 nd person agreement, independently of the overt ‘subject’ possibly preceding the verb. This is also clear in anaphoric relations, which do not ‘work’ as in standard (root) clauses, as is illustrated in (37), where a 2 nd person anaphoric possessive is the standard form used to refer to the 3 rd person ‘subject’ everybody, even though ‘their’ is also accepted and its use is becoming rather common: 17 (37) Everybody raise your ( ? their/* ? his) hand! The observation of (anti)agreement phenomena in imperatives has led authors to different explanations. Among the most influential, Downing (1969) and Ross (1970) propose that imperatives are embedded within a higher performative predicate and that the interpretation of the imperative subject is determined by an argument in the superordinate clause, which can be null because of an operation of deletion. 18 Refuting the performative analysis, authors in the GB framework have tried a characterization of the imperative subject as an empty category. In this respect, Potsdam (1988) defends the position that the English imperative has a largely unexceptional syntax and argues that no special particles nor ad hoc rules need to be assumed for the various manifestations of the syntactic imperative, which can be accounted for in terms of familiar clause structural properties (I-to-C movement, dosupport, Case as a form of spec-head agreement, and so on). As for the subject, a vocative analysis is strongly excluded: the imperative subject is located in its canonical Spec,IP and, when it is not overt, it is a pro. Surprisingly, Postdam does not linger on this final claim which –leaving aside the merits of his work –meets an immediate problem with the exceptionality of such an empty category in non pro-drop languages. Beukema & Coopmans (1989) also claim that imperative clauses can be given a ‘fairly orthodox syntactic representation’; however, they do not consider a pro analysis and suggest that the imperative subject is a variable bound by a possibly null element in a topic position. Another possibility proposed in the literature is that the imperative subject is a 17 2nd person agreement also emerges in tag questions (cf. (i)), which are generally assumed to be a root phenomenon. Nevertheless, Bianchi & Frascarelli (2010) have shown that tag questions can also occur in complements of verbs of saying and opinion, hence in ‘quasi-root’ contexts (iia-b). This means that tag questions qualify as a Type II root phenomenon, according to the finer distinction adopted in the present analysis: (i) Wash yourself, won’t you? (ii) a. I think (that) you like it, don’t you? b. Mary believes that Lucy is pregnant, isn’t she? 18 This proposal will be re-elaborated in the present structural analysis. How much room for discourse in imperative? 23 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 ©2020 The Editorial Board of Studia Linguistica PRO (Han 2000:115-160): since imperatives can pattern either with subjunctives or with non-finite forms, Han concludes that the subject is overt when the imperative patterns with a subjunctive, while the subject operator must be a PRO with infinitive forms. Though the picture might appear neat and tidy, problems arise with Case assignment and Control, since the imperative subject position seems to be case-marked and it is also not clear what element could control the imperative subject PRO. More recently in the Minimalist framework, Zanuttini (2008) posits the existence of a Jussive head in the I-domain, containing an interpretable 2 nd person feature, which semantically binds the subject and licenses it (through an Agree relation). ‘True agreement’ and the existence of a subject position in imperatives will be also addressed in this paper (cf. Section 7). 19 For the purpose of the present investigation, it is finally important to notice that imperatives clearly allow for some type of Topic in Italian and Spanish, as illustrated in the following examples: (38) a. La camicia, stirala! the shirt iron.IMP.2SG-it.CL b. ¡La camisa pl anchala ya! the shirt iron.IMP.2SG-it.CL right now ‘Iron the shirt!’ Nevertheless, Clitic Left Dislocation (CLLD) in Romance languages is compatible with both root and non-root contexts (Bianchi & Frascarelli 2010, Haegeman 2012, Jim enez-Fern andez & Miyagawa 2014), but 19 As for the realization of subjects in imperatives, notice incidentally that in V2 languages like Dutch and German the imperative verb occupies a pre-subject position, as is shown in (ia) below for Dutch and (ib) for the German polite formula (‘Hoflichkeitsform’), and that this position is also possible in Italian (ii) and Spanish (iii): (i) a. Trek jij maar je jas aan. ‘Pull you but your jacket on.’ b. Kommen Sie. ‘Come you.’ (ii) Metti tu la giacca ( =(ia)) (iii) Ponte t u la chaqueta ( =(ia)) However, the realization of a postverbal overt subject pronoun in Italian imperative clauses seems to be connected with some discourse-related intention –possibly a Contrastive Focus interpretation: (iv) Metti TU la giacca, non tuo fratello (che non ha mai freddo)! ‘YOU must put your jacket on, not your brother (who is never cold)!’ On the other hand, preverbal subjects in imperatives are rather closer to Contrastive Topics: (v) Tu, metti la giacca; tuo fratello, metter a un cappotto. ‘You –put your jacket on, you brother will wear a coat.’ We keep these observations (to be tested with informants) for future research. 24 Mara Frascarelli &  Angel L. Jim enez-Fern andez 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 ©2020 The Editorial Board of Studia Linguistica CLLD can implement different types of Topics, with different requirements with respect to illocutionary force (cf. Section 3). Hence, an investigation on the acceptability of specific discourse categories in imperative clauses is necessary to understand the root/non-root character of fronted constituents. 20 As for English Topicalization, there is no general consensus on its acceptability in imperatives, as is illustrated by apparent contrasts like the one –given above in (1) and (2), provided in Haegeman (2012: 64) – and repeated here as (39) and (40): (39) a. *Your essay, leave in my pigeon hole this afternoon. b. *The weapons leave behind. (40) a. The tie give to Bob, the aftershave give to Don. (Van der Wurff 2007) b. Anything you don’t eat put back in the fridge. (Huddleston & Pullum 2002) However, if Frascarelli & Hinterh€ olzl’s (2007) typology is adopted, the relevant grammaticality contrast is clarified: examples such as (40a-b) represent instances of C-Topics, whereas this reading is not contextually entailed in (39a-b). As a matter of fact, the idea that contrastiveness can enable topics to show up in contexts where they otherwise do not occur has been stated by different authors (cf. Prince 1998, Portner & Yabushita 1998, B€ uring 2003, Meinunger 2004). The relevance of Frascarelli & Hinterh€ olzl’s proposal rests on the systematic connection between the contrastive discourse function and a specific formal (syntaxprosody) realization in the C-domain. Hence, a finer discourse analysis proves to be crucial to understand and solve apparent inconsistencies. 4.3. Mapping the imperative meaning in syntax The syntax-semantic mapping of imperative clauses represents a complex and controversial issue in the literature and different proposals have been put forth to explain how the imperative meaning is syntactically encoded (i.e., relevant projections and dedicated features). Several authors have proposed that the relevant mapping depends on a clause-typing operation; in this type of approach, the imperative is encoded as a Force feature in the split-CP domain that composes the left periphery (also called C-domain). In particular, according to Han (2000), imperatives have an illocutionary operator with directive force, which is represented in the syntax via two specific features (i.e., [directive] and [irrealis]) encoded in C°. Schwager (2006), on the other hand, argues that 20 As a matter of fact, we can see that sentences like (38a-b) are grammatical in Italian and Spanish; however, in the absence of intonational and discourse cues, we cannot know what type of topic (i.e., A-, C- or G-Topic) is implemented by la camicia/la camisa ‘the shirt’. How much room for discourse in imperative? 25 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 ©2020 The Editorial Board of Studia Linguistica conclude that a discourse category that implies a conversational move, that is to say a request for evaluation on something apparently in contrast with shared knowledge (assumed to be stored in the CG), cannot be realized in an imperative clause. Hence, the latter is not a CG-active proposition and MF confirms its qualification as a Root I phenomenon. 5.2.3. CF in imperative clauses. Judgments on the acceptability of CF in imperative clauses show an interesting asymmetry between in situ and fronted realization: while a fronted CF (43) is rejected by informants (with values comparable to those attested for MF; cf. Table 2 above), an in situ CF (44) is fully accepted in the three languages examined. Let us then confront Tables 3 and 4, showing the two realizations of the same sample sentence: (43) a. A: Can I have a glass of wine, please? B: You must drive after the party: WATER have instead, not wine! b. A: Posso avere un bicchiere di vino per favore? can.1SG have.INF a glass of wine please ‘Can I have a glass of wine, please?’ B: Devi guidare dopo la festa. must.2SG drive.INF after the party ACQUA bevi piuttosto, non vino! water drink.IMP.2SG rather not wine ‘You must drive after the party, WATER have instead, not wine!’ (44) a. A: Can I have a glass of wine, please? B: You must drive after the party: don’t drink wine, have WATER instead! b. A: Posso avere un bicchiere di vino per favore? can.1SG have.INF a glass of wine please ‘Can I have a glass of wine, please?’ B: Devi guidare dopo la festa. must.2SG drive.INF after the party non bere VINO, bevi ACQUA piuttosto! not drink.INF wine drink.IMP.2SG water rather ‘You must drive after the party, don’t drink wine, have WATER instead!’ Table 2. MF in imperatives English Italian Spanish Informants CGr Informants CGr Informants CGr 0.12 0 0.59 1.02 0.33 0.23 32 Mara Frascarelli &  Angel L. Jim enez-Fern andez 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 ©2020 The Editorial Board of Studia Linguistica As for intonation, the elicited productions show that in situ CF is always (48/48) marked with a bi-tonal pitch accent, in which the tonic vowel can be aligned either with the high (H*+L) or with the low (H+L*)target.This alternation does not depend on syllabic structure, since the target words examined were mostly paroxytons (cf. note 24) and it seems rather connected to language-specific variations (H*+L in English and H+L*in Spanish). This variation is also left for future systematic investigation on the acoustic correlates of fronted Foci. Consider now Figures 3 and 4 below, for an immediate comparison between English and Italian fronted CF: As it can be observed, in both Figures the tonic syllable stays low in the final part of the imperative downgrading contour, marking low boundaries within the relevant Intonational phrase. Since in situ CF is fully accepted by informants, we must conclude that this discourse category is not excluded in imperative clauses. However, (i) it cannot be realized in the left periphery and (ii) it is not marked with a high pitch. Table 4. In situ CF in imperatives English Italian Spanish Informants CGr Informants CGr Informants CGr 3.18 3.33 3.11 3.31 3.16 3.25 Figure 3. In situ CF in English imperatives 5 Table 3. Fronted CF in imperatives English Italian Spanish Informants CGr Informants CGr Informants CGr 0.20 0.19 0.73 1.10 0.77 0.64 LOW RESOLUTION FIG 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 How much room for discourse in imperative? 33 ©2020 The Editorial Board of Studia Linguistica This result strongly differs from Bianchi et al.’s (2015) work, in which it is argued that CF can be realized both in the C-domain and in situ, and is marked in both cases with a L+H*pitch accent. This discrepancy leads to posit that (i) in imperative clauses a CF cannot A’-move to the left periphery because of a syntactic block (rather than for discoursesemantic reasons), which is not operative in the assertive clauses used in Bianchi et al. (2015) for their experiment, 29 and (ii) the corrective feature is interpreted in a criterial position (cf. Rizzi 2006, 2007) in the C- domain; consequently, in the absence of A’-movement, the relevant constituent cannot be interpreted at the interfaces. 30 Finally, comparing the behaviour of CF with what was attested for A- Topic and MF, we must conclude that CF is not a Type I root phenomenon (since it can be hosted by imperative clauses, at least in its in situ variant). The block on their fronted position will be dealt with later on in this paper. 5.2.4. C-Topics in imperative clauses The median values provided in Table 5 for C-Topics in imperative clauses reflect the fact that they are accepted in Italian and Spanish (with some Figure 4. In situ CF in Italian imperatives 6 29 In other words, we can conclude that the semantics of correction in itself is not incompatible with an imperative clause, but a syntactic constraint prevents the CF from being fronted. 30 On the other hand, it is possible to conjecture that the discourse properties of Corrective Foci are not exactly the same in the two positions and that focus fronting is triggered by some additional feature requirement that combines with correction, whose interpretation is basically in situ. Recent analyses have suggested, for instance, the possibility of an exhaustive character of fronted correction in some languages, on a par with Hungarian. However, since no systematic investigation has been conducted in this respect, as far as we know, we leave this question open for future research. LOW RESOLUTION FIG 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 34 Mara Frascarelli &  Angel L. Jim enez-Fern andez ©2020 The Editorial Board of Studia Linguistica variation that is, however, not significant; cf. Section 5.3 below), while they are marginal in English. Since the sentences in the test (and relevant contexts) were absolutely equivalent in the three languages, this discrepancy must be provided a comprehensive explanation. As for intonation, prosodic analysis shows that the dislocated phrases are realized with a bi-tonal L+H*pitch in 78 cases out of 98 (total occurrences are doubled because every target utterance in the test contains two sentences in this case, hence 2 C-Topics). This means that in the majority of cases C-Topics are phonetically realized as a F 0 rise with the peak aligned by the end of the TBU (as is shown in Figures 5 and 6 below). On the other hand, 20 realizations show a simple H*pitch accent; this realization generally occurs on the second C-Topic of the sequence (i.e., on the C-Topic located in the left periphery of the second sentence of the sequence), but no systematic behaviour could be attested. (45) a. A Where can I put these flowers? B: The roses put in the vase, the sunflower leave on the table. (45) b. A: ¿D onde dejo estas flores? where leave.PST.1SG these flowers ‘Where shall I leave these flowers?’ B: Las rosas, ponlas en el jarr on; the roses put.IMP.2SG-them.CL in the vase el girasol, d ejalo sobre la mesa the sunflower leave. IMP.2SG.it.CL on the table ‘The roses put in the vase; the sunflower leave on the table.’ Based on this result, we can say that the expected prosodic properties are always met for C-Topics since both H*and L+H*are pitch accents connected to contrastive elements across languages (cf. §3.2; see also Pierrehumbert 1980, Pierrehumbert & Hirschberg 1990, B€ uring 2003, Kratzer & Selkirk 2009, among many others). We can thus conclude that the relevant constituents are associated with the expected IS- interpretation and, as a consequence, that C-Topics are allowed in imperative clauses. Contrastive and Corrective discourse categories thus qualify as Type II root phenomena since both C-Topics and in situ CF are accepted in imperative clauses. However, C-Topics are not (fully) accepted in English and, in this respect, they show the same behaviour as fronted CF Table 5. C-Topics in imperatives English Italian Spanish Informants CGr Informants CGr Informants CGr 1.78 1.94 2.76 2.73 3.24 3.20 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 How much room for discourse in imperative? 35 ©2020 The Editorial Board of Studia Linguistica (excluded in all the languages examined). This suggests the existence of a syntactic block, which will be treated in detail in Section 6.3. Considering the behaviour of A-Topics/MF, on the one hand, and C- Topics/CF on the other, we can conclude that imperative clauses cannot host discourse phenomena with context update potential. Nevertheless, they cannot be considered less than propositions since they can host Type II root phenomena –a crucial claim for the purposes of the present investigation. Before dealing with the final analysis and the relevant proposal, let us finally consider the data concerning the realization of G-Topics. 5.2.5. G-Topics in imperative clauses G-Topics provide a picture in which they are marginal in English and only “mildly” accepted in Italian and Spanish. Consider Table 6 below: Figure 6. C-Topics in Spanish imperatives 8 Figure 5. C-Topics in English imperatives 7 LOW RESOLUTION FIG LOW RESOLUTION FIG 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 36 Mara Frascarelli &  Angel L. Jim enez-Fern andez ©2020 The Editorial Board of Studia Linguistica This result could be expected in English for independent reasons since, as is argued in Bianchi & Frascarelli (2010:21), no leftward topic structure is devoted to mere givenness marking in English (although givenness may be a concurrent property of topics), and the retrieval of given information is generally implemented through simple destressing (cf. also Neeleman & Reinhart 1998, Schwarzschild 1999). However, this result was totally unexpected in Italian and Spanish, since G-Topics are not connected with CG management and should be allowed in any type of clause (cf. Section 3.3 above). Furthermore, the prosodic analysis shows that acceptance was in fact connected to a different IS-interpretation (as in the case of A-Topics). Specifically, the informants who collaborated for the intonational analysis (i.e, those who accepted these structures in the online questionnaire) always produced the dislocated phrases with a bitonal L+H*pitch accent, that is to say, with the prosodic characterization of dislocated contrastive elements. Hence, they were feasibly interpreted as C-Topics. Consider sentences (46a-c) and relevant Figures 7-8-9 below: (46) a. A: Here you have the rice. What shall we do? B: Listen to me: the rice, steam it! (46) b. A: Aqu ı est a el arroz. ¿Qu e hacemos? here stay.3SG the rice what make.1PL ‘Here is the rice. What shall we do?’ Table 6. G-Topics in imperatives English Italian Spanish Informants CGr Informants CGr Informants CGr 2.12 2.42 2.63 2.72 2.73 2.54 Figure 7. Intended G-Topics in English imperatives 9 LOW RESOLUTION FIG 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 How much room for discourse in imperative? 37 ©2020 The Editorial Board of Studia Linguistica B: Mira, ¡el arroz,cu  ecelo al vapor! Look the rice cook.IMP.2SG-it.CL at.the steam ‘Look, the rice you can steam!’ (46) c. A: Ecco il pacco del riso. Che facciamo? here the pack of.the rice what make.1PL ‘Here is the pack of rice. What shall we do?’ B: Senti, il riso, cuocilo a vapore listen the rice cook.IMP.2SG-it.CL at steam ‘Listen, the rice cook it on steam.’ In these examples it is feasible to conjecture that the steaming of the contextually given ingredient (i.e., the ‘rice’) in the relevant sentence might have been interpreted as opposed to the cooking of some other type of food (not mentioned, but inferable via bridging in a ‘cooking Figure 8. Intended G-Topics in Spanish imperatives 10 Figure 9. Intended G-Topics in Italian imperatives 11 LOW RESOLUTION FIGLOW RESOLUTION FIG 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 38 Mara Frascarelli &  Angel L. Jim enez-Fern andez ©2020 The Editorial Board of Studia Linguistica context’, cf. Roberts’ (2003); see also §3.3 and note 13). At any rate, the crucial point is that no informant provided a low-toned realization for the dislocated constituents included in the section dedicated to G-Topics. Since G-Topics are allowed in every clausal type, this outcome must be provided an independent explanation, which will be addressed in the conclusive sections. 5.3. Statistical significance and parametric technique: the ANOVA analysis of variance The acceptability judgments obtained have been analysed using the ANOVA test so as to (i) verify whether the factors involved have a significant effect on the acceptability variation between the different experimental conditions, and (ii) be able to answer the questions mentioned in Section 5.1. In particular, the factors involved imply ‘single entries’ in two cases and, specifically, in the case of ‘discourse category’ (which are 5, namely, A-Topic, C-Topic, G-Topic, MF and CF) and ‘language’ (which are 3, namely English, Italian and Spanish). Then, a third factor only concerned CF, for which we wanted to evaluate the (possible) significance of its ‘position’ (whether in situ or fronted). As ‘position’ was not tested for any other discourse category (always located in the left periphery of the sentence), the CF case represented an exception and, as such, a complication for statistical analysis (since it is not appropriate to include a variance factor that only applies for one category out of five). We thus solved this problem producing two ANOVA tests: one considering A-Topic, C-Topic, G-Topic, MF and in situ CF, the other considering A-Topic, C-Topic, G-Topic, MF and fronted CF. Importantly, the results of the two ANOVA tests carried out show that the main effect 31 of the factors involved is significant (p <0.05) in both cases. This can be seen in Tables 7 and 8 below, in which significant values are marked in red for each interaction. 32 31 The main effect of an independent variable is the effect of that variable on the dependent one, ignoring the effects of all other independent variables. 32 Tables 7 and 8 show for each factor and for their interactions the following data: a. Sum of Squares (SS): the sum of the squared deviations (i.e., differences from the mean); b. Degree of Freedom (DF): the number of values that may vary in a statistic without violating any constraint; c. Mean Squares (MS): representation of population variance, obtained by dividing the SS by the DF; d. F-Value (F): the result of the F-test calculated by dividing the MS value by the MS error; it helps to verify whether the variance between the means of two populations is significantly different; e. P-value (P): the probability of getting a result at least as extreme as the one that was actually observed, given that the null hypothesis is true. If p <0.05 the effect of the relevant factor/interaction is taken as significant. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 How much room for discourse in imperative? 39 ©2020 The Editorial Board of Studia Linguistica It should be noticed that the interaction between the factors is also significant when compared two by two and, finally, that there is a significant interaction also between the factors if they are compared together. These results provide statistical support to the claim that the variance attested in the investigation is not due to chance but, indeed, the possibility of hosting specific discourse categories in the left periphery of an imperative clause depends (i) on the discourse category and (ii) on the language. The different acceptance attested for the in situ vs. fronted position of the CF is also significant for the 3 languages examined, as is shown in Table 9 below: Table 7. Factors’ main effect and their interactions with in situ CF Effect Repeated Measures Analysis of Variance (Long Table Format) Sigma-restricted parameterization; Effective hypothesis decomposition; Std. Error of Estimate: ,3058745 SS DF MS F p Intercept 130,1623 1 130,1623 3162,116 0,000012 Error 0,1235 3 0,0412 LANG 4,9259 2 2,4630 21,413 0,001856 Error 0,6901 6 0,1150 CAT 52,7826 4 13,1957 62,179 0,000000 Error 2,5467 12 0,2122 LANG*CAT 2,7080 8 0,3385 5,107 0,000860 Error 1,5907 24 0,0663 Table 8. Factors’ main effect and their interactions with in situ CF Repeated Measures Analysis of Variance (Long Table Format) Sigma-restricted parameterization; Effective hypothesis decomposition; Std. Error of Estimate: ,3058745 Effect SS DF MS F p Intercept 237,6020 1 237,6020 2539,588 0,000017 Error 0,2807 3 0,0936 LANG 3,2842 2 1,6421 17,606 0,003086 Error 0,5596 6 0,0933 CAT 60,7466 4 15,1867 86,884 0,000000 Error 2,0975 12 0,1748 LANG*CAT 3,5614 8 0,4452 6,715 0,000125 Error 1,5910 24 0,0663 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 40 Mara Frascarelli &  Angel L. Jim enez-Fern andez ©2020 The Editorial Board of Studia Linguistica The analysis of standard deviations confirms the consistency and attainability of the present investigation since their value is always rather low. This shows that the average distance from the median values is never too large and, consequently, that the results attested are reliable. In particular, Table 10 below shows that the most dependable data concern CT, GT and in situ CF, whereas fronted CF and MF have clearly created major uncertainties to informants. We conclude this section with an illustration of the statistical data provided through a diagram (Figure 10 below), which allows for an immediate comparison between the mean values attested for the discourse categories examined across the three languages at issue: As is clear, results are very consistent and the only major difference attested concern the acceptability of CT and fronted CF in English, which will be treated in detail in Section 6.2 below. Table 9. Factors’ main effect and their interactions for in situ vs. fronted CF Repeated Measures Analysis of Variance (Long Table Format) Sigma-restricted parameterization; Effective hypothesis decomposition; Std. Error of Estimate: ,3058745 Effect SS DF MS F p Intercept 83,01026 1 83,01026 1842,991 0,000028 Error 0,13512 3 0,04504 LANG 0,33979 2 0,16989 15,335 0,004380 Error 0,06647 6 0,01108 CAT 40,10957 1 40,10957 750,948 0,000107 Error 0,16024 3 0,05341 LANG*CAT 0,47097 2 0,23549 7,712 0,021969 Error 0,18322 6 0,03054 Table 10. Mean values and standard deviations for categories and languages AT CT GT MF In situ CF Fronted CF ITA mean 1,09 2,76 2,63 0,59 3,11 0,73 St. dev. 1,1901947 1,2995660 1,3378622 0,9548604 1,0427106 1,0582323 SPA mean 1,79 3,24 2,73 0,33 3,16 0,77 St. dev. 1,4112318 0,8973735 1,2574260 0,7823406 0,9512920 1,0881123 ENG mean 1,20 1,78 2,12 0,12 3,18 0,20 St. dev. 1,1862754 1,4877976 1,4759878 0,4774156 0,9591423 0,6222048 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 How much room for discourse in imperative? 41 ©2020 The Editorial Board of Studia Linguistica by imperatives (i.e., the consistent ban imposed on Type I root phenomena) and their complement-like quality. Furthermore, the existence of a SAP can offer an explanation to the ‘antiagreement’ effects that characterize imperative clauses (cf. Section 4.2 above). Specifically, assuming that grammatical agreement starts out at C (cf. Chomsky 2005) and adapting Miyagawa’s (2010, 2012, 2017) analysis, we propose that a sort of ‘allocutive agreement’ is operative in imperative clauses as well (i.e., a default agreement with the ADDRESSEE). In particular, Miyagawa (2012) proposes that in languages like English subject agreement lowers to T, while in languages like Basque, two agreement features occur in C: one lowers to T to form the typical subject-verb agreement, while the other stays at C where it agrees with the ADDRESSEE to form the allocutive agreement (cf. Ohyarc ßabal 1993). Extending this proposal, we suggest that this parameter can also operate inter-linguistically and distinguish between different types of agreement in different clause types. Hence, even though in a given language agreement lowers to T in declarative clauses, it can stay in the C-domain in imperatives (tentatively in Fin°, following Rizzi’s 1997 association between FinP and agreement/finiteness) and agree with the allocutive probe in SA°, as we can observe in (52): We thus propose that 2 nd person agreement in imperative clauses is due to allocutive agreement. Nevertheless, considering the relevant Agree relation as a parametric option, the possibility is left open that in some languages (or sentence type, both inter-linguistically and cross-linguis- tically) allocutive agreement is not realized and an alternative form of agreement is expressed in the imperative clause. We can now finally resume and provide an explanation for the asymmetry emerged (in all the languages examined) between fronted and in situ CF and the block on C-Topics in English. 6.3. The fronting of discourse-related categories in imperative clauses: an intervention account Based on the analysis proposed in Section 6.2, we suggest that the asymmetries concerning fronted CF and C-Topics can be provided an immediate and comprehensive explanation considering their syntactic derivation. As is known, CF fronting and Topicalization (the operation deriving C-Topics in English; cf. among others Haegeman 2004, Emonds 2004, (52) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 48 Mara Frascarelli &  Angel L. Jim enez-Fern andez ©2020 The Editorial Board of Studia Linguistica Maki, Kaiser & Ochi 1999, Bianchi & Frascarelli 2010) are derived through A’-movement, whereas CLLD languages implement different strategies for Topics. Specifically, in languages like Italian (all types of) Topics are merged in the C-domain and associated with a clitic pronoun in argument position (cf. Cinque 1990, Frascarelli 2000, 2017), while in languages like Spanish G-Topics and C-Topics move to Spec,TP (cf. Jim enez-Fern andez & Miyagawa 2014). This means that fronted CF (cross-linguistically) and C-Topics in English A’-move to the C-domain, determining an intervention effect (cf. footnote 7 above; see also Haegeman 2012) with the allocutive probing operation (53a), while this is not the case in languages like Italian (53b) and Spanish (53c): Assuming for fronting the same type of Operator movement described for C-Topic in (53a), the block imposed on fronted CF –attested across languages (cf. Table 2 above) –can be also attributed to an intervention effect. This means that no inconsistencies are at stake: the block is due to independent (syntactic) reasons and CF and C-Topics can be consistently characterized as Type II root phenomena, without further ado for the present interface proposal. 40 It should be noticed that we are taking intervention as a by-product of movement for evaluation; hence when two chains interfere intervention shows up. This recalls Richards’ (1999) intervention effect caused by crossing chains, instead of nesting chains. Crossing chains are only allowed in multiple movements to multiple Specifiers of a single head. In 40 We acknowledge (following a reviewer’s suggestion) that reading sentences that were created for a written test is not an optimal testing design for a prosodic evaluation and, consequently, further research is needed to validate this conclusion. Nevertheless, since the outcome proposed is very consistent across informants, we reckon that the present result can be given as conclusive in this investigation. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 How much room for discourse in imperative? 49 ©2020 The Editorial Board of Studia Linguistica our case, allocutive agreement and CLLD involve different heads. Accordingly, no crossing of the relevant chains is permitted. This can explain the intervention effect exhibited in English (cf. (53a)). On the other hand, in Italian C-Topics directly merge in ContrP, so that no crossing is at stake for the allocutive chain (cf (53b)). Finally, in Spanish C-Topics move to TP (Jim enez-Fern andez & Miyagawa 2014) and, consequently, the two chains are nesting into each other, creating no intervention effect (cf (53c)). The latter option is also attested in cases where we combine A-movement and A’-movement, as in wh-questions. Since the two resulting chains are nesting, no blocking arises. For further discussion on the role of nesting vs. crossing chains in intervention effects, see Jim enez-Fern andez (2011). The intervention effect created by the allocutive agreement raises the question as to whether initial adjuncts can occur in imperatives. Assuming Haegeman’s (2012) view that (at least some) initial adjuncts are base-generated in an initial position, the prediction is that (given there is no intervention) they are totally compatible with imperatives. This prediction is borne out by data, as illustrated in the following examples: (54) a. When you’re ready, give me a call. b. At three o’clock, come to my office and we will talk. Since there is no intervention due to the non-movement character of these adjuncts, they can perfectly occur in the left periphery of imperatives. 41 As for the case of G-Topic, it will be treated below in a separate Section, since it implies a wider reflection on the interplay between prosody and syntax. 7. Back to G-Topics: Conversational dynamics at the prosodic Interface As we have seen, G-Topics are either very marginal or rejected in the left periphery of imperative clauses. Since they are either merged in the C-domain or moved to Spec,TP in CLLD languages, relevant unacceptability cannot be attributed to the same syntactic block that was argued 41 It could be objected that fronted adjuncts are compatible with imperatives in English (and presumably in the other languages they deal with), as is shown in (i) and (ii) below: (i) When you’re ready, give me a call. (ii) At three o’clock, come to my office and we will talk. However, initial adjuncts like the ones exemplified in (i) and (ii) are ‘a different type of A’- constituents’. They are not Topics, but more properly qualified as (temporal) ‘frames’ (using Chafe’s 1976 traditional term, resumed and discussed in Krifka 2007). Hence, they are not ‘entities’, which is the semantic property that we assume for Topics (following Reinhart 1982), and their treatment is beyond the scope of this paper. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 50 Mara Frascarelli &  Angel L. Jim enez-Fern andez ©2020 The Editorial Board of Studia Linguistica for fronted CF. Similarly, a discourse-semantic block cannot be invoked, since G-Topics are assumed to be available in all types of clauses. Based on the observation that G-Topics can be realized in the right periphery of an imperative clause, as well as in other types of directive speech acts (see (55)-(56) below), we can conclude that the exclusion of G-Topics in the left periphery of imperative clauses must be due to a prosodic constraint and, specifically, to a spell-out incompatibility between an initial low-toned constituent and the H*+L intonation associated with imperative clauses. In other word, our claim is that the relevant ‘block’ can be attributed to an interface strategy emerging to avoid a prosodic clash (on the relevance of clash-avoidance strategies cf. Nespor & Vogel 1979, Prieto 2011, among others; on directive intonation cf. Pierrehumbert & Hirschberg 1990 and the discussion in §5 above). As a piece of evidence, consider sentences (55) and (56) below, which are part of an independent test, elaborated for the purposes of this proposal (with the collaboration of 15 informants from each of the examined languages): (55) A: Non voglio mangiare carote: le odio. No quiero comerme las zanahorias: las odio. not want.1SG eat.INF the carrots them hate.1SG ‘I don’t want to eat carrots: I hate them.’ B: Mangiale, le carote: sono buone per la C ometelas, las zanahorias:¡son buenas para la eat.IMP-them.CL the carrots: be.3PL good.F.PL for the salute! salud! health ‘Eat your carrots, instead: they are good for you!’ (56) A: Non voglio mangiare carote: le odio. not want. 1SG eat.INF carrots them hate.1SG No soporto comer zanahorias; no me gustan nada not bear.1SG eat.INF carrots not to.me please.3PL nothing ‘I don’t want to eat carrots: I hate them.’ B: Dovresti mangiarle, le carote: sono buone deber ıas comerte las zanahorias:¡son buenas should.2SG eat.INF-them.CL the carrots: be.3PL good.F.PL salute! per La salud! para la for the health ‘Eat your carrots, instead: they are good for you!’ As is clear, in these cases no prosodic clash can be produced with the final L% boundary of the imperative clause (see Figures in Section 5) and the relevant sentences are fully accepted by informants. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 How much room for discourse in imperative? 51 ©2020 The Editorial Board of Studia Linguistica To conclude, always sticking to a cartographic account in which the full hierarchy of projections is maintained (cf. also §6.2), no ad hoc restriction should be sought for the imperative C-domain: the cross-linguistic exclusion of G-Topics from imperatives is due to an interface incompatibility between the raising intonative contour of imperative clauses and the presence of a low-toned constituent in the left periphery. This also supports that idea that phonological information is encoded in the lexicon in the form of syntactic features (i.e., a p-feature associated with a dedicated functional projection; cf. Adger and Svenonius 2011), which are made visible to the PF interface. Hence, neither A-Topics or G-Topics are accepted in imperatives, but for different interpretive reasons at the interfaces. A-Topics are not allowed on discourse-semantic grounds: imperatives cannot host Type I root phenomena because they are non-CG-active propositions, that is to say, they are not endowed with update potential in the CG. On the other hand, the unacceptability of G-Topics is ‘milder’ because it depends on a prosodic incompatibility (not on discourse-semantic reasons). As for English, they are also excluded for independent reasons (cf. discussion after Table 6 above). A final note regarding the presence of Force (though non-assertive Force) in imperatives is connected with the possible occurrence of imperatives in relative clauses. As illustrated by the following English examples (from Haegeman 2012: 120-121), imperatives are fully compatible with relatives: (57) a. But there is, despite the book’s brevity, and the fact that it is enormously pleasurable to read (at which point let me salute the translation) much going on. (Guardian 04.12.2010, page 13, col. 5) b. He was at it again on last week’s Desert Island Discs, still complaining that the world –for which read the music press –does not appreciate his genius. (Observer 6.12.2000, page 23, col. 3) c. The top speed (which please don’t try to reach) is 220 miles an hour (Ferrari test driver BBC Radio 5, Andrew Radford, p.c.) If Rizzi (1997) is correct when he claims that relative operators move to Spec,ForceP, the conclusion is that imperatives also hold a ForceP whose Spec can be the target of a relative operator. This is precisely the case in examples (58), which support the idea that some sort of Force must be present in imperatives. This is in line with our proposal, though our claim is that imperative are not assertion but non CG-active propositions. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 52 Mara Frascarelli &  Angel L. Jim enez-Fern andez ©2020 The Editorial Board of Studia Linguistica 8. Conclusions: The C-domain of imperative clauses at the interfaces In the present work a novel syntactic analysis for imperative clauses has been proposed, crucially based on an interface investigation. Consistent evidence has been provided that in the left periphery of imperative clauses the availability of information-structural categories is partially restricted. However, rather than assuming ‘partial peripheries’ or ad hoc solutions, the present analysis has led to the conclusion that this is due to the interaction of constraints operating at different levels of grammar (conversational dynamics, discourse-semantics, syntax and prosody). Specifically: a. Imperative clauses are not endowed with assertive illocutionary force but they cannot be considered less than propositions. b. Based on the distinction between Type I and Type II root phenomena (where the former type can only occur in CG-active propositions) and assuming a compositional approach to conversational dynamics, imperatives qualify as non-CG-active propositions. As such, they have no update potential in the CG (though they can have access into the CG repository) but, rather, they affect the interlocutor’s TO-DO-LIST; c. Imperative clauses are dominated by a super-ordinate Speech Act Phrase, which includes the SPEAKER and the ADDRESSEE as coarguments; d. Given the lack of update potential, the interpretation of imperatives is not connected with assertive illocutionary Force, but with a Modality projection in the IP-split domain, which is activated by an Agree relation with the [+DIR] feature in the C-domain; e. Given the interface avoidance strategy of prosodic clashes, the unacceptability/strong marginality of G-Topics in the left periphery of imperative clauses can be comprehensively accounted for; f. Assuming that agreement is as a type of ‘allocutive agreement’ in C, the general block on fronted Corrective Foci and Contrastive Topics in languages like English can be independently attributed to an intervention effect: syntax blocks the movement of discourserelated categories that interfere with the allocutive probe in Fin° since they create crossing chains and these are banned when triggered by different heads. References 13 ABOH, E. O., K. HARTMANN &M.ZIMMERMANN (eds.) 2007. Focus Strategies in African Languages. The Interaction of Focus and Grammar in Niger-Congo and Afro-Asiatic. Berlin: Mouton de Gruyter. ADGER,D.&SVENONIUS, P. 2011. Features in Minimalist syntax. The Oxford handbook of minimalist syntax. ed. C. Boeckx, 27–51. Oxford: Oxford University Press. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 How much room for discourse in imperative? 53 ©2020 The Editorial Board of Studia Linguistica ^ AMBAR, M. 1999. Aspects of the syntax of focus in Portuguese. The grammar of focus. eds. G. Rebuschi & M. Tuller, 23–53. Amsterdam & Philadelphia: John Benjamins. BAAYEN, R. H. 2008. Analyzing Linguistic Data. A Practical Introduction to Statistics using R. Cambridge: Cambridge University Press. BEUKEMA,F.&COOPMANS, P. 1989. A Government-Binding perspective on the imperative in English. Journal of Linguistics 25:417–436. BEAVER,D.&CLARK, B. 2008. Sense and sensitivity: How focus determines meaning. Oxford: Blackwell. BIANCHI, V. 2013. On Focus movement in Italian. Information Structure and Agreement. eds. V. Camacho-Taboada,  A. L. Jim enez-Fern andez, J. Mart ın- Gonz alez & M. Reyes-Tejedor, 194–215. Amsterdam & Philadelphia: John Benjamins. BIANCHI, V. 2015. Focus Fronting and the syntax-semantics interface. Beyond Functional sequence. The Cartography of Syntactic Structures 10. ed. U. Shlonsky, 60–712. Oxford: Oxford University Press. BIANCHI,V.&BOCCI, G. 2012. Should I stay or should I go? Optional Focus movement in Italian, ed. C. Pinon, Empirical Issues in Syntax and Semantics 9, 1–16. Paris: EISS. BIANCHI, V., BOCCI,G.&CRUSCHINA, S. 2015. Focus Fronting and its implicatures. Romance Languages and Linguistic Theory 2013. eds. E. Aboh, J. Schaeffer & P. Sleeman, 3–20. Amsterdam & Philadelphia: John Benjamins. BIANCHI,V.&FRASCARELLI, M. 2010. Is topic a root phenomenon? Iberia 2:43–88. BIANCHI,V.&FRASCARELLI, M. 2012. How to be rooted in a context. Paper presented at the 35 th GLOW Colloquium, University of Postdam. BIRNER, B. 1995. Pragmatic constraints on the verb in English inversion. Lingua 97(4):233–256. BRUNETTI, L. 2009. On the pragmatics of post-focal material in Italian (left peripheral focus looked from the other side. Les linguistiques du d etachement. Actes du Colloque de Nancy, Juin 7–9, 2006. eds. D. Apoth eloz, B. Combettes & F. Neveu, 151–162. Berne: Peter Lang. B€ URING, D. 1999. Topic. Focus, linguistic cognitive and computational perspectives. eds. P. Bosch & R. van der Sandt, 142–165. Cambridge: Cambridge University Press. B€ URING, D. 2003. On D-trees, beans, and B-Accents. Linguistics and Philosophy 26:511–545. CAMACHO-TABOADA,V.&JIM ENEZ-FERN ANDEZ, A. L. 2014. Focus Fronting and Root Phenomena in Spanish and English. Language Use and Linguistic Structure. eds. J. Emonds & M. Janebov a, 41–60. Olomouc: Palacky University. CARDINALETTI, A. 2002. Against optional and zero clitics. Right dislocation vs. marginalization. Studia Linguistica 56:29–57. CARDINALETTI, A. 2009. On a (wh-)moved Topic in Italian, compared to Germanic. Advances in Comparative Germanic Syntax. eds. A. Alexiadou, J. Hankmaer, T. McFadden, J. Nuger & F. Schaeffer, 3–40. Amsterdam & Philadelphia: John Benjamins. CHAFE, W. 1976. Givenness, contrastiveness, definiteness, subjects, topics, and point of view. Subject and Topic. ed. Ch. N. Li, 25–55. New York: Academic Press. CHIERCHIA, G. 1985. Formal semantics and the grammar of predication. Linguistic Inquiry 16:417–443. CHIERCHIA,G.&MCCONNELL-GINET, S. 1990. Meaning and Grammar: An Introduction to Semantics. Cambridge, MA: MIT Press. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 54 Mara Frascarelli &  Angel L. Jim enez-Fern andez ©2020 The Editorial Board of Studia Linguistica CHOMSKY, N. 2008. On phases. Issues in Linguistic Theory. Essays in Honor of Jean-Roger Vergnaud. eds. R. Freidin, C. Peregrin Otero & M. L. Zubizarreta, 133–166. Cambridge, MA: MIT Press. CINQUE, G. 1990. Types of A’-Dependencies. Cambridge, MA: MIT Press. CORMANY, E. 2013. A preference analysis of imperatives connecting syntax, semantics and pragmatics. University of Cornell: Ph.D dissertation. CRUSCHINA, S. 2011. Fronting, dislocation and the syntactic role of discourserelated features. Linguistic Variation 11:1–34. DAYAL,V.&GRIMSHAW, J. 2009. Subordination at the interface. Unpublished manuscript, Rutgers University. DOWNING, B. 1969. Vocatives and third-person imperatives in English. Papers in Linguistics 1:570–592. EMONDS, J. 1970. Root and Structure-Preserving Transformations. MIT: Ph.D. dissertation. EMONDS, J. 1976. A transformational approach to English syntax: Root, structurepreserving, and local transformations. New York: Academic Press. EMONDS, J. 2004. Unspecified categories as the key to root constructions. Peripheries: Syntactic Edges and their Effects. eds. D. Adger, C. de Cat & G. Tsoulas, 75–120. Dordrecht: Kluwer. ESTEBAN-VILAPLANA,E.&PRIETO, P. 2010. Castilian Spanish intonation. Transcription of Intonation of the Spanish Language. eds. Pilar Prieto & P. Roseano, 17–48. M€ unchen: Lincom Europa. EVANS, N. 2007. Insubordination and its uses. Finiteness: Theoretical and Empirical Foundations. ed. I. Nikolaeva, 366–431. Oxford: Oxford University Press. FARKAS,D.F.&BRUCE, K. B. 2010. On reacting to assertions and polar questions. Journal of Semantics 27:81–118. FRASCARELLI, M. 2000. The Syntax-Phonology Interface in Focus and Topic Constructions in Italian. Studies in Natural Language and Linguistic Theory 50. Dordrecht: Kluwer. FRASCARELLI, M. 2007. Subjects, Topics and the Interpretation of Referential pro. An interface approach to the linking of (null) pronouns. Natural Language and Linguistic Theory 25:691–734. FRASCARELLI, M. 2010a. Narrow Focus, Clefting and Predicate Inversion. Lingua 120(9):2121–2147. FRASCARELLI, M. 2010b. Intonation, Information Structure and the Derivation of Inverse VO Languages: An Interface Analysis of Tagalog and Malagasy. Austronesian and Theoretical Linguistics. eds. R. Mercado, E. Potsdam & L. deMena Travis, 81–101. Amsterdam & Philadelphia: John Benjamins. FRASCARELLI, M. 2017. Dislocations and Framings. Manual of Romance Morphosyntax and Syntax. eds. E. Stark & A. Dufter, 472–510. Berlin: Mouton de Gruyter. FRASCARELLI, M. 2018. The interpretation of pro in consistent and partial NS languages: A comparative interface analysis. Null-Subjects in Generative Grammar. A synchronic and diachronic perspective. eds. F. Cognola & J. Casalicchio, 211–239, Part IIB, Ch. 9. Oxford University Press. Oxford/New York. FRASCARELLI, M. 2019. Topics, Conversational dynamics and the Root/Non-root distinction: Adverbial clauses at the Discourse-syntax Interface. Architecture of Topics. eds. V. Moln ar, V. Egerland & S. Winkler, 139–170. Berlin: Mouton De Gruyter. FRASCARELLI,M.&CARELLA, G. 2019. Topic chains and the interpretation of null subjects: The acquisition of discourse-related strategies in Italian. The Linguistic Review 36(4):637–674. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 How much room for discourse in imperative? 55 ©2020 The Editorial Board of Studia Linguistica FRASCARELLI,M.&HINTERH € OLZL, R. 2007. Types of topics in German and Italian. On Information Structure, Meaning and Form. eds. K. Schwabe & S. Winkler, 87–116. Amsterdam & Philadelphia, John Benjamins. FRASCARELLI,M.&HINTERH€ OLZL, R. 2016. German scrambling meets Italian Right-Dislocation. Paper presented at the 42th Incontro di Grammatica Generativa, University of Salento FRASCARELLI,M.&JIM ENEZ-FERN ANDEZ, A. L. 2019. Understanding partiality in pro-drop languages: An information-structure approach. Syntax (Special Issue on Information Structure) 22(2-3):162–198. FRASCARELLI,M.&RAMAGLIA, F. 2013. Phasing contrast at the interfaces. Information Structure and Agreement. eds. V. Camacho-Taboada,  A. L. Jim enez-Fern andez, J. Mart ın-Gonz alez & M. Reyes-Tejedor, 55–82. Amsterdam & Philadelphia: John Benjamins. G€ ARTNER, H.-M. 2002. On the force of V-2 declaratives. Theoretical Linguistics 28:33–42. GIANOLLO, C. 2016. The Latin System of Negation at the Syntax-Semantics Interface. Rivista di Grammatica Generativa 38:115–135. GILI-FIVELA, B., AVESANI, C., BARONE, M., BOCCI, G., CROCCO, C., D’IMPERIO, M., GIORDANO, R., MAROTTA, G., SAVIO,M.&SORIANELLO, P. 2015. Varieties of Italian and their intonational phonology. Intonational Variation in Romance. eds. S. Frota & P. Prieto, 140–197. Oxford: Oxford University Press. GREGORY,M.&MICHAELIS, L. 2001. Topicalization and Left Dislocation: A functional opposition revisited. Journal of Pragmatics 33:1665–1706. HAEGEMAN, L. 2002. Anchoring to Speaker, adverbial clauses and the structure of CP. Georgetown University Working Papers in Theoretical Linguistics 2:117– 180. HAEGEMAN, L. 2004. Topicalization, CLLD and the left periphery. ZAS Papers in Linguistics 35: Proceedings of the Dislocated Elements Workshop. eds. C. Maienborn, W. Frey & B. Shaer, 157–192. Berlin. HAEGEMAN, L. 2006. Conditionals, factives and the left periphery. Lingua 116:1651–1669. HAEGEMAN, L. 2010. The internal syntax of adverbial clauses. Lingua 120:628– 648. HAEGEMAN, L. 2012. Adverbial clauses, main clause phenomena, and the composition of the left periphery. Oxford: Oxford University Press. HAEGEMAN, L. 2014. West Flemish verb-based discourse markers and the articulation of the Speech Act Layer. Studia Linguistica 68(1):116–139. HAEGEMAN,L.&HILL, V. 2013. The syntactization of discourse. Syntax and its Limits. eds. R. Folli, C. Sevdali & R. Truswell, 370–390. Oxford: Oxford University Press. HAEGEMAN,L.&MIYAGAWA, S. 2016. The syntax of participants. Keynote address GLOW workshop. https://www.uni-goettingen.de/en/510390.html. HAEGEMAN,L.& € UR€ AGDI, B. 2010. Referential CPs and DPs: An operator movement account. Theoretical Linguistics 36(2-3):111–152. HAN, C.-H. 2000. The Structure and Interpretation of Imperatives: Mood and Force in Universal Grammar. Garland, Outstanding Dissertations in Linguistics. HAN, C.-H. 2001. Force, Negation and Imperatives. The Linguistic Review 18:289–325. HARTMANN, K. 2001. Right node raising and gapping. Amsterdam & Philadelphia: John Benjamins. HEIM, I. 1982. The Semantics of Definite and Indefinite Noun Phrases. University of Massachusetts at Amherst, Ph.D. dissertation. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 56 Mara Frascarelli &  Angel L. Jim enez-Fern andez ©2020 The Editorial Board of Studia Linguistica HENGEVELD,K.&MACKENZIE, L. 2008. Functional Discourse Grammar. Oxford: Oxford University Press. HEYCOCK, C. 2006. Embedded Root Phenomena. The Blackwell Companion to Syntax. eds. M. Everaert & H. van Riemsdijk, 174–209. Oxford: Basic Blackwell. HILL, V. 2007a. Romanian adverbs and the pragmatic field. The Linguistic Review 24:61–86. HILL, V. 2007b. Vocatives and the pragmatics-syntax interface. Lingua 117:2077– 2105. HILL, V. 2008. Pragmatics markers as syntactic heads: a case study from Romanian. Syntaxfest Cyprus Conference. eds. K. Grohmann & P. Panagiotidis, 237–265. Newcastle-upon-Tyne: Cambridge University Scholars Press. HOOPER,J.&THOMPSON, S. 1973. On the applicability of root transformations. Linguistic Inquiry 4:465–497. HUDDLESTON,R.&PULLUM, G. K. 2002. The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press. JENSEN, B. 2007. In favour of a truncated imperative clause structure: evidence from adverbs. Working Papers in Scandinavian Syntax 80:163–185. JIM ENEZ-FERN  ANDEZ, A. L. 2011. On the Order of Multiple Topics and Discourse–feature Inheritance. Dilbilim Aras ßtırmaları (Journal of Linguistic Research) 1:5–32. JIM ENEZ-FERN  ANDEZ, A. L. 2015a. When focus goes wild: An empirical study of two syntactic positions for information Focus. Linguistics Beyond and Within 1:119–133. JIM ENEZ-FERN  ANDEZ, A. L. 2015b. Towards a typology of focus: Subject position and microvariation at the discourse-syntax interface. Ampersand: An International Journal of General and Applied Linguistics 2:49–60. JIM ENEZ-FERN  ANDEZ, A. L. 2016. When discourse met null subjects. Borealis: An International Journal of Hispanic Linguistics 5(2):173–189. JIM ENEZ-FERN  ANDEZ, A. L. 2018. Negative Preposing: Intervention and parametric variation in complement clauses. Atlantis: Journal of the Spanish Association for Anglo-American Studies 40(1):11–37. JIM ENEZ-FERN  ANDEZ, A. L. 2020. Syntax-information structure interactions in the sentential, verbal and nominal peripheries. Newcastle: Cambridge Scholars Publishing. JIM ENEZ-FERN  ANDEZ, A. L. & MIYAGAWA, S. 2014. A feature-inheritance approach to root phenomena and parametric variation. Lingua 145:276–302. KAUFMANN, M. 2012. Interpreting imperatives. Dordrecht: Springer (Studies in Linguistics and Philosophy 88). KISS,K. E. 1998. Identificational focus versus information focus. Language 74:245–273. KOOPMAN, H. 2007. Topics in imperatives. Imperative Clauses in Generative Grammar. ed. W. van der Wurff, 153–180. Amsterdam & Philadelphia: John Benjamins. KRATZER,A.&SELKIRK, E. 2009. Distinguishing contrastive, new and given information. Paper presented at the Workshop on Prosody and Meaning, Barcelona. KRIFKA, M. 1992. The semantics and pragmatics of polarity items. Linguistic Analysis 25:209–257. KRIFKA, M. 2001. Quantifying into question acts. Natural Language Semantics 9:1–40. KRIFKA, M. 2007. Basic Notions of Information Structure. Interdisciplinary Studies on Information Structure. eds. C. F ery & M. Krifka, 13–55. ISIS, Universit€ atverlag Potsdam. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 How much room for discourse in imperative? 57 ©2020 The Editorial Board of Studia Linguistica