Understanding partiality in pro-drop languages: an information-structure approach

Understanding Partiality in pro-Drop Languages: An Information-Structure Approach Mara Frascarelli and  Angel L. Jim enez-Fern andez 1 Abstract. This article explores the acceptability and interpretation of referential null subjects in several varieties of Spanish (both Peninsular and American varieties) in different syntactic contexts (matrix and embedded) and distinct clause types (under bridge verbs, under factive verbs, and in adverbial clauses). Based on the results of an original online survey, completed by almost 300 respondents all over the world, it is shown that, in contrast with a consistent-prodrop language like Italian, some Spanish varieties exhibit partial-pro-drop properties; this contributes to a recent line of research dealing with partiality in consistent-pro-drop languages. We argue that such variation can be accounted for within an information-structure approach, in which the Topic Criterion (Frascarelli 2007), the formation of topic chains, and the existence of silent topics play a crucial role. Degrees of partiality are attributed to an interface condition that combines information-structure requirements with PF visibility of overt copies in topic chains. Partial-pro-drop properties can then be explained by reference to independent syntactic conditions, such as the preference for overt minimal links and the sensitivity to islands. 1. Introduction In its original formulation (Perlmutter 1971), the pro-drop parameter was aimed at capturing the empirical observation that in some languages a definite, referential, pronominal subject in a finite clause must be expressed overtly. 1 This observation was taken up, extended, and reworked by Chomsky (1982) in the form of the ExtendedProjection Principle (EPP), a basic tenet in generative grammar. Since then, many works have been dedicated to the pro-drop parameter, defining the formal properties that determine the setting of its different options, which yield consistent-, partial-, radical-, semi-, and non-pro-drop languages (for discussion and references, see Biberauer et al. 2010, Cognola & Casalicchio 2018). Since the seminal work carried out by Jaeggli (1982) and Rizzi (1986), the nullsubject parameter has been claimed to be dependent on the /features that are specified or encoded in the relevant licensing head (i.e., the Inflectional head node). Criticism of this approach was soon elaborated (see Jaeggli & Safir 1989, Huang 1989), and an alternative analysis has recently been proposed in the Minimalist framework by Earlier versions of this article were presented at the fourth Cambridge Comparative Syntax conference in the UK, at the 26th Colloquium on Generative Grammar in Spain, and at the seventh Encuentro de Gram atica Generativa in Argentina. We thank all the audiences for their encouraging comments. We are especially grateful to the Syntax reviewers for their challenges, which definitely have contributed to the improvement of the article. The research here has been partially funded by research project FFI2013-41509P of Spain’s Ministry of Economy and Enterprise (MINECO), by research project PGC2018-093774-B-I00 of Spain’s Ministry of Science, Innovation, and Universities (MICINN), and by research project 2018/ 00000491 of the seventh Plan Propio of the Universidad de Sevilla. The order of authors is alphabetical. 1 In this work we concentrate on null subjects that are referential; other types, such as arbitrary and expletive null subjects, are left to one side. 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 ©2019 John Wiley & Sons Ltd S Y N T 12184 Dispatch: 6.7.19 CE: Wiley Journal Code Manuscript No. No. of pages: 37 PE: Rajasekaran S. Syntax 2019 DOI: 10.1111/synt.12184 Holmberg (2005), according to which the licensing of a null subject (hereafter, NS) depends on the presence of a D feature in T. (See also Camacho 2013, 2016, Saab 2016, and Ticio 2018 on the occurrence of this D feature in T in different Spanish varieties.) From a different perspective, Frascarelli (2007) shifts the focus of analysis from licensing to the interpretation of licensed NSs. Assuming with Holmberg that the licensing mechanism requires matching between interpretable and uninterpretable features in Agr 0 , Frascarelli concentrates on how a licensed pro can be correctly given a referential index and interpreted as intended in the discourse. As a matter of fact, / features only provide functional information about the person, number, and gender of apro; they do not explain how its discourse linking (D linking) is established in syntactic terms in a consistent-pro-drop language. Based on interface analysis of naturalistic spoken corpora, Frascarelli argues for an information-structure strategy. Specifically, she provides evidence that the interpretation of a referential NS depends on a matching relation (“Agree”in Minimalist terms)between it and a specifictypeof topic. Based on Italian (spontaneous) data, she identifies the latter as the aboutness-shift topic (A-Topic) of Frascarelli & Hinterh€ olzl 2007 (see section 2.1 for details, and see Jim enez-Fern andez 2016 for a similar analysis based on Spanish data). The A-Topic combines the [aboutness] feature (which Frascarelli proposes as an “extended EPP feature”) with the [shift] feature, which is the property that A-Topics have of moving the conversation from one topic to another, updating the discourse context (see Bianchi & Frascarelli 2010). A-Topics are present in every predicative sentence to establish “what the sentence is about”and, when overtly realized, to shift the topic across sentences. Frascarelli puts forth a Topic Criterion that correlates core grammar with discourse requirements: (1) Topic Criterion a. [+aboutness] is connected with an (extended) EPP feature in the high topic field that yields a specific discourse-related property, namely “aboutness.” b. The [+aboutness] topic matches with an argument in the main clause through Agree. c. When continuous, the A-Topic can be null (i.e., silent). The Topic Criterion implies that every sentence contains a position in its C domain endowed with the [aboutness] feature. However, since a shift is not realized in every sentence, it is crucial to assume that, within discourse, predication can imply a multiclausal domain in which chains of clauses are combined and refer to an established A-Topic, whose [aboutness] feature is kept continuous. Topic continuity is enabled across sentences by the presence of low-toned copies (GTopics, pronouns) and silent copies (null topics, NSs) of the established A-Topic, heading aso-calledtopic chain (see also Frascarelli 2018). In this picture, the established A-Topic is kept silent until a new A-Topic is proposed (breaking the previous topic chain). 2 2 A topic-oriented approach to NSs can also be found in previous works (among others, Giv on 1983 and more recently Adli 2011, Camacho 2013, and Pe skov a 2014). The crucial difference in the present approach is the fact that the relation between NS and topic is not “inferred from the context”but is part of discourse grammar and, as such, implies specific formal requirements. 2Mara Frascarelli and  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 ©2019 John Wiley & Sons Ltd Finally—given the (empirically based) assumptions that the interpretation of an NS depends on an Agree relation with the current A-Topic (cf. (1b); this relation is local, following Minimalist tenets) and that A-Topics are only associated with clauses endowed with illocutionary force (cf. the Interface Root Restriction in section 2.3)— the following condition must be added to the Topic Criterion. (2) Topic-Chain Condition (Frascarelli 2018:(19)) a. An A-topic chain can only be started from a root (or root-like) C domain. b. The A-Topic heading the topic chain can be silent. Indeed, the relevance of silent categories is a major issue in recent analyses, and substantial evidence has been provided to establish that, echoing Kayne’s (2016:1) words, “many more elements in the sentential projection line (and elsewhere) must be taken to be silent than is usually thought.”In this respect, see also Sigurðsson 2004. The central role of topics in discourse is widely discussed and acknowledged in the literature (see, among many others, Kuroda 1965, Giv on 1983, Lambrecht 1994). Topics have different functions, however. In this picture, it is important to emphasize that Frascarelli’s (2007) proposal (which we adopt) treats topics as an entity, one that is endowed with the discourse-semantic property of aboutness (following Reinhart 1981), and this property is captured by a formal feature as an extended EPP requirement. Furthermore, although Frascarelli 2007 focuses on consistent-NS languages, Frascarelli 2018 provides evidence that the Topic Criterion (1) can be considered a macroparameter that also accounts for partial-NS languages (see section 3.2). As a matter of fact, there is a general consensus in the relevant literature that an NS language of the Italian, Spanish, and Greek type may change gradually into a non-NS language like English and French (see Holmberg 2010, Saab 2016). Based on a number of properties, such as the preference for preverbal subjects over postverbal subjects, the overt use of pronominal subjects, and the impact of c-command in the interpretation of NSs, there are languages that have been claimed to be changing into partial-NS languages, including Portuguese and, to a lesser extent, Spanish (see Kato 2000, Modesto 2000, Rodrigues 2004, and Barbosa, Duarte & Kato 2005, which deal with Brazilian Portuguese, and Com ınguez 2018 and Camacho 2016, which deal with Caribbean varieties of Spanish). Nevertheless, the discussion on this topic is still lively and open (especially for Caribbean Spanish: see Camacho 2013, 2016, Com ınguez 2018, Ticio 2018; from a different perspective, see Posio 2012, taking into account usage-based sociolinguistics). In the present work we contribute to this hot topic by discussing the syntactic and interpretative properties of NSs (and, to a lesser extent, overt pronominal subjects) in two consistent-NS languages, Italian and Spanish. We present experimental results from several Spanish varieties, and we examine whether they support the claim that (i) some Spanish varieties (not only Caribbean Spanish) are becoming partial-NS languages, like Brazilian Portuguese, or that (ii) Spanish varieties are still consistentUnderstanding Partiality in pro-Drop Languages 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 ©2019 John Wiley & Sons Ltd NS languages but some of them, even so, show partial-NS properties, supporting the idea that partiality in the character of languages is a matter of degrees. 3 Frascarelli (2007) discusses this question and provides multiple arguments for the advantages of an information-structure analysis. Among other things, she provides evidence that NSs cannot be treated on a par with overt pronouns in an NS language like Italian, and she provides data involving quantified DPs that clearly indicate the necessity of a formal theory that includes some abstraction, such as null topics. In this respect, consider for instance the Italian and Spanish scope facts in (3)–(5). (3) Ogni angolo della banca era sorvegliato da una guardia. ∀,∃ Cada esquina del banco estaba vigilado por un polic ıa. each corner of.the bank be.PST.3SG guarded by a police.officer ‘Every corner of the bank was guarded by a police officer.’ As indicated, the DP una guardia/un polic ıa‘a police officer’can have a narrowor wide-scope reading. However, if we continue (3) with a sentence like (4), only an existential reading is available. (4) pro fumava come un matto!*∀,∃ ¡pro fumaba como un loco! smoke.PST.3SG like a mad ‘He smoked like a lunatic!’ We take this to be evidence that the NS in (4) does not take as its antecedent the DP una guardia in (3) but rather a silent topic in its own local C domain. This null topic is specificbydefinition and only allows for an existential reading: 4 (5) ... sorvegliato da una guardia. [〈La/questa guardia k 〉[pro k ... guarded by a police.officer the/this police.officer fumava come un matto]]. smoke.PST.3SG like a mad As we will see, silent topics will play an important role in our formalization of the interpretation of NSs. The article is organized as follows. In section 2 we present the typology of topics that is put forth in Frascarelli & Hinterh€ olzl 2007, discuss the role of the A-Topic in conversational dynamics, and illustrate its function and relevance at the syntax– 3 We believe that our information-structure approach to the encoding of D linking has advantages over a pragmatic approach to anaphora resolution. A pragmatic theory, claiming that coreferent readings are preferred unless a disjoint reading is implicated (based on the Gricean maxim of quantity; see Levinson 1991), cannot fully account for the complexity of this phenomenon; an integrated approach is required. For discussion, see Van der Sandt 1992. 4 Additional evidence for the specific interpretation of the null topic in these sentence types comes from the fact that, whenever the topic is given an explicit spellout, it must be specific. This explains why only a definite DP like la/questa guardia ‘the/this police officer’can be overtly realized in the C domain of (3)–(5) (see also Jim enez-Fern andez 2016 for Spanish). 4Mara Frascarelli and  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 ©2019 John Wiley & Sons Ltd discourse interface as the element heading topic chains. In section 3, we first provide some background on the differences between consistentand partial-NS properties and then introduce our experiment and the relevant methodology. In section 4 the experimental data are presented and discussed in detail. Based on the results obtained, section 5 gives an information-structure analysis of the data and defends the present proposal, aiming at a comprehensive explanation of the crosslinguistic variation attested. 2. Different Types of Topics, A-Topic Chains, and the Avoid-Pronoun Principle Based on an interface analysis of naturalistic spoken corpora from Italian and German, Frascarelli & Hinterh€ olzl (2007) provide substantial evidence for the existence of a systematic correlation between the formal properties of (different types of) topics and their functions in discourse. Specifically, they distinguish between the aboutness-shift topic (A-Topic), the contrastive topic (C-Topic), and the familiar/ given topic (G-Topic). The distinct function of each is encoded in a dedicated position in the C domain, in line with a cartographic and Minimalist approach to grammar (see Rizzi 1997 and subsequent works): 5 (6) [ ForceP [ ShiftP [ ContrP [ FocP [ FamP [ FinP [ IP For the purposes of the present analysis, we will specifically deal with A-Topics and G-Topics, which are briefly described in the following two subsections. 2.1. A-Topics As previewed in section 1, the A-Topic connects Reinhart’s (1981) aboutness with the property of being newly introduced or reintroduced and changed to—hence, with a shifting function in the discourse. In a cartographic approach, in which discourse properties are encoded in syntax in the form of formal (functional) features, the Shift 0 head is thus assumed to be endowed with a specification for interpretable [aboutness] and [shift] features. Since the [aboutness] feature is proposed as an extended EPP feature (see the Topic Criterion in (1)), Spec,ShiftP is a criterial position (Rizzi 2006) and must be filled in every predicational sentence. Though this notion might seem reminiscent of Lambrecht’s (1994) sentence topic, the central difference is that, adopting Reinhart’s approach, the A-Topic is an entity with a specific role in conversational dynamics: an updating effect leading to a new context. Following this line of analysis, the discourse context includes a structured repository of shared information (propositions publicly accepted by the participants), and the discourse referent denoted by the A-Topic constitutes the “address”(or file 5 The cartographic approach, in which features are assumed to be merged in dedicated projections, is fully compatible with the Minimalist Program, thought it may not seem that way at first sight. As Cinque & Rizzi (2008) note, we can take the two approaches to have a different focus. Minimalism focuses on the elementary mechanisms involved in syntactic computation, while cartography deals with the details of the syntactic structure; they are not mutually exclusive (Mao & Meng 2016, Tanaka 2016). Understanding Partiality in pro-Drop Languages 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 ©2019 John Wiley & Sons Ltd card, in the terms of Heim 1982) under which the proposition conveyed by the assertion will be stored. This conversational move updates the interlocutor’s common ground, in which propositions are stored under defining entries. Syntactically, the A-Topic is merged in the highest topic position in the C domain (Spec,ShiftP), and it cannot be iterated. As for intonation, the A-Topic has been shown to be associated with a complex L*+H tone in a number of typologically different languages (see Frascarelli 2007, Frascarelli & Hinterh€ olzl 2007, Frascarelli & Puglielli 2009). However, since the prosodic interface is not part of the present investigation, intonational properties will not be further mentioned. As an illustration of a topic shift, consider the example in (7) (from corpus data in Frascarelli 2007), in which a student is talking about computer skills with a friend. The topic constituent(s) under examination are boldfaced. (7) A: Il computer k[A-Topic 1] tu lo k avevi gi a usato qualche the computer you it have.PST.2SG already used any volta? time ‘Did you ever use a computer before?’ B: S ımi  e capitato per opro k era MacIntosh. yes to.me be.3SG happened but was Mac ‘Yes, I’ve experienced that, but it was a Mac.’ Ho sempre avuto un approccio di base ... have.1SG always had a approach of base ‘I only had basic skills ...’ Ecco, il CD-ROM z[A-Topic 2] non l z ’avevo mai well the CD-ROM not it have.PST.1SG never usato—pro z  e utilissimo! used be.3SG very.useful ‘Well, I had never used a CD-ROM—it is very useful!’ As shown, the topic proposed by speaker A is il computer (a dislocated direct object resumed by the clitic lo), but then speaker B shifts the topic to a specific tool, namely il CD-ROM. Note that the NS following this topic shift is unambiguously related to the DP il CD-ROM and cannot be linked to il computer. This is evidence that topic shifts create new topic chains. 2.2. G-Topics Quite differently from A-Topics, the discourse function of G-Topics is the retrieval of given information in the discourse. The G-Topic can therefore be considered a Dlinked constituent, either in a “strong”sense (Heim 1982) or in a weak/familiar sense (Roberts 2003). According to Frascarelli & Hinterh€ olzl’s (2007) theory, the retrieval of given information can satisfy two different discourse requirements. On the one hand, a GTopic can appear as part of a topic chain, and in this case it is a “low copy”of an 6Mara Frascarelli and  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 ©2019 John Wiley & Sons Ltd established A-Topic, serving a continuity function. On the other hand, the G-Topic can retrieve given information that is part of the background but is not linked to the current A-Topic. The former type of G-Topics are defined as aboutness G-Topics, the latter as background G-Topics. Given this discourse characterization of G-Topics, Frascarelli & Hinterh€ olzl assume that the head of FamP is endowed with the [aboutness] feature, which is however uninterpretable in this position. By Minimalist tenets, the property of (un) interpretability is the driving force behind the establishment of a syntactic dependency (Chomsky 1995:277); accordingly, we assume, as Adger & Svenonius (2011:34) do, that “uninterpretable features are those that drive the derivation, while the interpretable ones are those that are used, in the final representation, to connect with the semantic systems or the phonological ones”(i.e., with interface levels). The distinction between the two types of G-Topics thus depends on their syntactic derivation. When a G-Topic moves to Spec,FamP, as illustrated in (8), it enters an Agree relation with Shift 0 and Fam 0 , which are endowed with, respectively, the interpretable features [shift; aboutness; ref] and the uninterpretable features [aboutness; ref]. We assume that the NS’s features are unvalued but interpretable (a possible dissociation offered by Pesetsky & Torrego 2007). As a consequence of this Agree relation, the G-Topic’s interpretable [aboutness] feature is valued in the syntax and then interpreted accordingly at the interfaces. This constituent is thus part of a topic chain and is interpreted as an aboutness G-Topic. (8) [ ShiftP A-Topic z [ Shift 0 Shift 0 [shift; aboutness; ref] [ FamP G-Topic/pro z[aboutness; ref] [ Fam 0 Fam 0 [aboutness; ref] [ TP t z T[ vP v+V]]]]]] Notice that, besides [shift] and [aboutness], the Agree relation under examination also allows for the interpretation of the [ref] of the NS. The Shift head, with interpretable [aboutness], [shift], and [ref] features, thus plays a crucial role in interface interpretation by acting as a probe. In contrast, when a G-Topic is not endowed with the unvalued interpretable [aboutness] feature, movement to Spec,FamP is not triggered and an Agree relation with Shift 0 is not realized. Since FamP is not needed, it does not project. In this case the G-Topic gets a “background”realization and is not part of a topic chain. To account for the background reading of the G-Topic, we propose—as an addition to Frascarelli & Hinterh€ olzl’s (2007) theory—that a background G-Topic is interpreted in Spec,TP (drawing on Jim enez-Fern andez & Miyagawa 2014). This is illustrated in (9). The G-Topic does not refer back to the A-Topic. (9) [ ShiftP A-Topic z [ Shift 0 Shift 0 [shift; aboutness; ref] [ TP G-Topic T [ vP v+V...]]]] 2.3. Discourse Categories, Conversational Dynamics, and Topic Chains Discourse categories have different functions in conversational dynamics. In this respect Bianchi & Frascarelli (2010:51), adopting Krifka’s (2007) framework for the common ground that distinguishes between common-ground management and Understanding Partiality in pro-Drop Languages 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 ©2019 John Wiley & Sons Ltd common-ground content, provide evidence that a discourse category that triggers an update of the discourse context must occur in a clause endowed with context-update potential. They propose the following restriction. (10) Interface Root Restriction (Bianchi & Frascarelli 2010:(8)) Information-structure phenomena that affect the conversational dynamics (common-ground management) must occur in clauses endowed with illocutionary force that implement a conversational move. Evaluated in terms of this restriction, A-Topics implement a conversational move, and as such they belong to the dimension of common-ground management and are restricted to root (or root-like) clauses. By contrast, since givenness is calculated on the basis of the common-ground content, G-Topics do not depend on illocutionary force and can be found in any type of clause (adverbials included). Finally, remember that topic continuity implies the presence of silent A-Topics, whose existence is independently required by the extended EPP (see (1)). Hence, when no new topic is proposed in a sequence of discourse-related sentences, the Spec, ShiftP position is filled by a silent A-Topic, as is the case in the embedded clause in (11) (angle brackets indicate silent copies). In this case the embedded Shift 0 is endowed only with [aboutness] and [ref], and it enters the Agree relation discussed in section 2.2 for constituents moving to Spec,FamP. 6 (11) [ ShiftP Leo k [ FamP pro k [ TP t k pensa [ ForceP che [ ShiftP 〈Leo k 〉 Leo think.3SG that Leo [ FamP pro k /lui k [ TP t k comprer a una casa]]]]]]]. he buy.FUT.3SG a house ‘Leo k thinks that he k will buy a house.’ Notice that NSs and weak pronouns (of the aboutness kind) share a topic-continuity function in the discourse and can freely alternate in a topic chain (in parametric variation across languages and language varieties). In particular, the spellout of lowtoned copies can be considered a conversational device in consistent-NS languages, plausibly connected with style, processing necessities (distance, memory burden), and dialects (microvariation). Since the realization of (c)overt subject pronouns is associated with the Avoid Pronoun condition, this theory also allows for a reformulation of that condition based on information-structure considerations, as is proposed by Frascarelli (2007:(32)): (12) Avoid Pronoun Avoid using a strong pronoun, whenever it would agree with the local A-Topic. 6 The example in (11) is taken from the set of sentences used in our experiment. 8Mara Frascarelli and  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 ©2019 John Wiley & Sons Ltd 3. Comparing Consistentand Partial-NS Languages: Some Background and the Experiment 3.1. Some Background on the Differences between Consistency and Partiality It is very difficult to state what a partial-NS language is, since there is no general consensus on exactly what features such a language should show. Different partial-NS properties have been proposed in the literature (see references in section 1), with different importance across languages. In this uneven scenario, Holmberg, Nayudu & Sheehan’s (2009) proposal is often taken as a landmark; on this basis, we can assume that a partial-NS language is characterized by (i) sensitivity to locality, (ii) sensitivity to syntactic control, (iii) considerable use of explicit pronouns, and (iv) a preference for preverbal subjects. In their analysis of consistentand partial-NS languages, Holmberg, Nayudu & Sheehan concentrate on the first two of these properties, and they argue that in partial-NS languages a pro is only allowed if controlled (i.e., c-commanded) by an antecedent in a higher clause. 7 They illustrate this condition with data such as (13), from Brazilian Portuguese, a language that is generally acknowledged to instantiate partial pro drop (though, for a different view, see Kato 2011). (13) O Jo~ ao i disse [que os moleques k Brazilian Portuguese the Jo~ ao say.PST.3SG that the children acham [que *(ele i ) e esperto]]. think.3PL that he be.3SG smart ‘Jo~ ao i said that the kids think he i is smart.’(Holmberg, Nayudu & Sheehan 2009:82) They argue that partial-NS languages do not allow a control relation across another subject, even if the intervening subject has incompatible features with the NS. This is why Brazilian Portuguese prefers the explicit pronoun ele ‘he’in (13), so as to comply with the strictly-local-control property that characterizes partial pro drop (whereas an NS is used to obtain coreference with an antecedent that is internal to the sentence domain where the NS occurs). A different type of c-command-based restriction is also proposed for Spanish embedded NSs, specifically to distinguish Standard Spanish and Caribbean varieties of Spanish (see Camacho 2013, 2016, Gupton & Lowman 2013), based on Montalbetti’s (1986) effects on quantifier binding that depend on whether the pronominal subject is overt or covert. Consider the example in (14) and the possible interpretations of null and overt subject pronouns in the embedded clause. 7 The authors deal with NSs in the complements of bridge and perception verbs, as well as in adjunct clauses and indirect questions. Notice that the claim that NSs in Brazilian Portuguese are subject to a c-command requirement is already made by Modesto (2000). Understanding Partiality in pro-Drop Languages 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 ©2019 John Wiley & Sons Ltd control but on an information-structure strategy. Hence, when no discourse-prosodic cues are available for a clear indication of the A-Topic, ambiguity dominates (also in PS varieties, contrary to the contrast that Ticio 2018 claims between PS and Caribbean varieties; on the ambiguities deriving from “semantic predictability”in topic continuity, see also Giv on 1983). Let us now consider the case in which an overt pronoun is realized in the embedded clause, to see whether any difference emerges between Italian and Spanish: (22) (Same context as (21)) a. Leo ha detto che lui ha comprato una casa. Leo have.3SG said that he have.3SG bought a house ‘Leo k said that he k /he z bought a house.’ b. Leo dijo que  el compr o una casa. Leo say.PST.3SG that he buy.PST.3SG a house ‘Leo k said that he k /he z bought a house.’ Table 3 shows that Spanish overt pronouns are not discourse-context oriented (either in PS or in AS) and that they prefer the matrix subject/topic as an antecedent over the embedded subject (both possible A-Topics), whereas no preference is attested in Italian data. Indeed, the values attested for ‘Leo’in table 3 are higher than those reported in table 2, and the difference between these values is significant for both PS (p=.0141) and AS (p=.0169) varieties. Particularly high values are attested for ‘Leo’in Mex and Carib, whose values with respect to table 2 are significantly different (p=.0256 and p=.0441, respectively). Since the preferred option is still ‘both’for PS and AS (in general), we can plausibly conclude that, when an exclusive reading is required, in the absence of information-structure cues the matrix subject is interpreted as an A-Topic and the default discourse interpretation associated with an overt pronominal subject is that of a familiar entity that is part of a topic chain—that is to say, an aboutness G-Topic. Despite this preference, ambiguity is still the dominant result. This supports the present information-structure approach, according to which the information-structure identification of the A-Topic is necessary for the interpretation of both NSs and overt pronouns (see the contrast between NSs and overt pronominal subjects in table 14 below). Table 3. Overt pronoun embedded under a matrix bridge verb Italian PS Anda Cast Cat Ext AS Carib Centr Mex Riv Per OK/N 128 173 111 23 35 4 91 29 14 15 18 15 Leo N 35 43 20 11 11 1 30 14 4 8 2 2 %27 25 18 48 31 25 33 48 29 53 11 13 His brother N36 21 16 1 4 0 9 2 1 0 5 1 %28 12 14 4 11 0 10 7 7 0 28 7 Both N 57 109 75 11 20 3 52 13 9 7 11 12 %45 63 68 48 58 75 57 45 64 47 61 80 16 Mara Frascarelli and  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 ©2019 John Wiley & Sons Ltd Let us now consider what happens when an NS is double embedded under two bridge verbs (condition (v) in section 3.2), thus having two possible controlling antecedents, one of which is closer than the other: 14 (23) (At the stadium) A: Look!Juan and his brother are already on the track! B: ¡S ı!Pedro dijo que Juan cree que va a ganar yes Pedro say.PST.3SG that Juan believe.3SG that go.3SG to win la carrera. the race ‘I see!Pedro k said that Juan z thinks that he k /he z is going to win the race.’ Comparing the results in table 4 with those for one level of embedding (table 2), where the second option was not proposed in the syntactic context, it is interesting to notice that values are almost identical for AS varieties (except Centr), for which the best option is again ‘both’. Ambiguity dominates in PS varieties as well. Nevertheless, PS differs from AS insofar as the embedded subject/topic (‘Juan’) is preferred to the matrix subject, and the difference between these values and those in table 2 is extremely significant (p=.0008). Hence PS varieties, especially Ext and Cat, seem to be more sensitive to syntactic requirements such as locality for the selection of the antecedent, (a point that will be taken up in section 5). Notice, however, that with the exception of Ext and Cat the best option is once more ambiguity, scoring around 60% in both Spanish macrovarieties. 4.2. NSs Embedded under Factive Verbs Let us now consider the second of the eight conditions listed in section 3.2. For the sake of space, we will not present the data involving one level of embedding under a factive verb, since the results are basically the same as with bridge verbs in section 4.1, and will instead pass immediately to NSs in complex sentences (whose matrix verb is factive) with two controlling antecedents. This condition is particularly Table 4. NS double embedded under two bridge verbs Italian PS Anda Cast Cat Ext AS Carib Centr Mex Riv Per OK/N 128 173 111 23 35 4 91 29 14 15 18 15 Pedro N 73 12 7 1 3 1 16 4 10 2 0 0 %57 7 6 4 9 25 18 14 74 9 0 0 Juan N 20 55 29 5 18 3 20 4 1 4 5 6 %17 32 26 22 50 75 22 14 5 27 28 38 Both N 30 106 75 17 14 0 55 21 3 9 13 9 %26 61 68 74 41 0 60 72 21 62 72 62 14 For the sake of space, we will not provide the Italian sentence in this and the remaining examples, except where it is important for comparative purposes. Understanding Partiality in pro-Drop Languages 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 ©2019 John Wiley & Sons Ltd interesting since the complement of a factive verb is presupposed information and as such does not have root-like properties. This means that an A-Topic is not allowed in its C domain (see Bianchi & Frascarelli 2010) and a preverbal DP like ‘Pedro’in (24) can only be considered a G-Topic. (24) A: Juan is going with Pedro to the race. B: Ya sabes, Juan lamenta que Pedro crea que va already know.2SG Juan regret.3SG that Pedro think.3SG that go.3SG a perder la carrera. to lose.INF the race ‘You know, Juan k is sorry that Pedro z thinks that he k /he z will lose the race.’ As we can see in table 5, the embedded preverbal DP is selected with high frequency as a possible antecedent for the NS, as shown by the high combined scores of the ‘Pedro’option and the ‘both’option. Even though ‘Juan’is the only possible overt ATopic in (24), it is not preferred either in Italian or in Spanish varieties. This result supports, on the one hand, the hypothesis that the A-Topic heading the chain can be silent and, on the other, that locality plays a role in the selection of the antecedent. In this respect, it must be noted that in Ext the ‘Pedro’option actually ties with the ‘both’ option (50% each), showing a degree of preference for the closest DP that is not attested in the remaining Spanish varieties—especially not in Centr, where, exceptionally, the ‘Pedro’option is the least favored answer (7%). Both these points will be taken up in section 5. 4.3. NSs in Adverbial Clauses It is generally agreed in the literature that adverbial clauses are not endowed with illocutionary force. In this respect, however, Haegeman (2012) argues for an important distinction between central and peripheral adverbial clauses. According to Haegeman, while the left periphery of central adverbials totally lacks the functional projections encoding speaker-related functions (speech time, epistemic modality, illocutionary force) and are within the scope of operators, peripheral adverbial clauses seem to admit (some) root phenomena. Since we were interested in examining the interpretation of NSs in nonroot adverbial clauses (i.e., in structural contexts that do not allow for an A-Topic in the Table 5. NS double embedded under a factive verb Italian PS Anda Cast Cat Ext AS Carib Centr Mex Riv Per OK/N 128 173 111 23 35 4 91 29 14 15 18 15 Juan N 20 16 9 3 4 0 10 2 5 1 1 1 %16 9 8 13 11 0 11 7 36 7 6 7 Pedro N 45 50 36 3 9 2 21 6 1 6 6 2 %35 29 32 13 26 50 24 21 7 40 33 13 Both N 63 107 66 17 22 2 60 21 8 8 11 12 %49 62 60 74 63 50 65 72 57 53 61 80 18 Mara Frascarelli and  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 ©2019 John Wiley & Sons Ltd embedded C domain), we limited our survey to central adverbials, considering in particular temporal and conditional clauses. Since the results obtained for these two adverbial types are very consistent, for reasons of space we will only present data concerning conditional clauses. Conditional clauses were tested in two structural conditions: with the ‘if’clause realized after and before the matrix clause. This was to check the interpretive effect of an ostensible movement operation: the postmatrix position is usually taken to be “basic”and the prematrix position is usually defined as “fronted”(see Haegeman 2012). In (25) is a sample sentence for the postmatrix position. (25) (Pedro’s friends meet for a beer. They know that Pedro is still at work with his boss. They hope he can join them later. One of them says:) Pedro puede venir si termina el trabajo. Pedro can.3SG come if finish.3SG the work ‘Pedro k can come if he k /he z finishes his work.’ The first interesting result obtained is that NSs in adverbial clauses are not always accepted by Spanish informants, unlike what we saw for NSs in the complements of bridge and factive verbs. For postmatrix conditional clauses, as seen in table 6, 14% of PS informants and 11% of AS informants gave a negative judgment. Furthermore, it should be noticed that nonacceptance is spread across all varieties. This result can be taken as a sign that Spanish varieties are not as “consistent”as Italian is (since the Italian informants provided a positive answer for all the conditions examined). An overt pronoun makes the sentence acceptable for all speakers in PS and AS; the pronoun is interpreted as referring to Pedro (Com ınguez 2011). Let us consider in detail the interpretations provided by informants who answered positively. As shown, for the first time ambiguity is not the best option, and the possibility of establishing a “silent head”for the topic chain (i.e., an A-Topic that is mentioned in the context but not overtly realized in the sentence) scores very low, except in Mex and Per, where again ambiguity wins (56% and 64%, respectively); in all the other AS and PS varieties, the matrix subject/topic ‘Pedro’qualifies as the best antecedent for the NS. Hence, the scores for the matrix preverbal DP are rather high; comparing its scores with NSs in adverbial clauses to its scores with NSs embedded Table 6. NS embedded in a postmatrix conditional clause Italian PS Anda Cast Cat Ext AS Carib Centr Mex Riv Per OK/N 128/128 149/173 94/111 19/23 33/35 3/4 81/91 28/29 12/14 9/15 18/18 14/15 Pedro N 84 87 58 9 18 2 38 12 7 4 11 4 %66 58 62 47 55 67 47 43 58 44 61 29 His boss N 16 19 10 5 4 0 12 8 2 0 1 1 %12 13 11 26 12 0 15 28.5 17 0 6 7 Both N 28 43 26 5 11 1 31 8 3 5 6 9 %22 29 27 27 33 33 38 28.5 25 56 33 64 Understanding Partiality in pro-Drop Languages 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 ©2019 John Wiley & Sons Ltd under bridge and factive verbs, the difference is extremely significant both in Italian (p<.0001) and in both Spanish varieties (PS: p<.0001; AS: p=.0150). Let us turn to the prematrix condition, in which the embedded NS is not controlled by the matrix subject in surface structure: (26) (Same context as in (25)) Si termina el trabajo, Pedro puede venir. if finish.3SG the work Pedro can.3SG come.INF ‘If he k /he z finishes his work, Pedro k can come.’ Once again, as seen in table 7, there is less than full acceptance in Spanish—less even than in the postmatrix context: 23% in PS and 18% in AS. Evidently fronting implies some additional interpretive complications. These results place the Spanish varieties under examination close to partial languages with respect to this condition (see Frascarelli 2018). As for interpretive results, the figures in table 7 clearly show that ambiguity comes back as the best choice for AS informants; they also select ‘his boss’more frequently than the matrix subject/topic. PS varieties, on the other hand, offer a mixed picture, in which Cast seems to prefer an external antecedent while Anda and Cat show no specific preference between overt and covert antecedence. Apart from the question at issue in the present article (i.e., the informationstructure approach for an understanding of partial-pro-drop properties), these data are also important because they show that fronted adverbial clauses are not reconstructed in their basic position for interpretation: otherwise, the interpretive results should be the same for both. This is in line with Haegeman’s (2007, 2012) analysis, showing that adverbial-clause fronting in clitic-left-dislocation languages produces no intervention effects since clitic-doubled left-dislocated topics “form a separate class from other A0-dependencies”(Rizzi 2004:245). That is why adverbial-clause fronting does not interfere with wh movement out of the matrix clause in languages like Italian and Spanish, as shown in (27a) and (27b) respectively. Table 7. NS embedded in a prematrix conditional clause Italian PS Anda Cast Cat Ext AS Carib Centr Mex Riv Per OK/N 128/128 133/173 86/111 16/23 28/35 3/4 75/91 26/29 13/14 6/15 17/18 13/15 Pedro N 59 34 23 3 6 2 17 10 2 0 3 2 %46 25 27 19 21 67 23 38 15 0 17 15 His boss N29 39 20 11 8 0 21 10 4 1 3 3 %23 29 23 69 29 0 28 38 31 17 17 20 Both N 40 60 43 2 14 1 37 6 7 5 11 8 %31 46 50 12 50 33 49 24 54 83 66 65 20 Mara Frascarelli and  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 ©2019 John Wiley & Sons Ltd (27) a. Se finisce il lavoro, dove andr a Pedro t? b. Si termina el trabajo, ¿ad onde ir a Pedro t? if finish.3SG the work where go.FUT.3SG Pedro ‘Where will Pedro k go if he k /he z finishes the work?’ This property shows that a fronted adverbial clause does not interfere with other types of movement and that reconstruction for the licensing and interpretation of an NS cannot be invoked. 4.4. The Interpretation of NSs with an Intervening Nonargument Pronoun Let us now turn to sentences designed to check the interpretation of an NS when a nonargument intervenes in the superordinate clause that should, according to Holmberg, Nayudu & Sheehan (2009; see discussion of (13) in section 3.1), create interference in partial-NS languages despite being semantically incompatible. This was tested by means of sentences like (28), which obtained the results given in table 8. (28) Juan dijo que fue una suerte que hubiera Juan say.PST.3SG that be.PST.3SG a luck that have.PST.3SG ganado el segundo premio en la loter ıa. won the second prize in the lottery ‘Juan k said that it was lucky that he k /he z won at the lottery.’ This type of sentence was always accepted. Ambiguity is once more the favored option, in all varieties except Centr and, to a less extent, Carib. Locality thus seems to be a relevant requirement in these two AS varieties. We will see in the remainder of this investigation whether this sensitivity to locality in Centr and Carib is confirmed. For all other varieties, we can simply conclude that an NS can be licensed and interpreted across a nonargument DP, either through a chain with an overt nonlocal preverbal subject/topic (‘Juan’) or through a covert (silent) ATopic. This provides additional support to the validity of the present informationstructure approach. Table 8. NS interpreted across a nonargument DP Italian PS Anda Cast Cat Ext AS Carib Centr Mex Riv Per OK/N 128 173 111 23 35 4 91 29 14 15 18 15 Juan N 15 20 14 2 3 1 16 5 3 3 4 2 %12 12 13 9 9 25 19 17 21 20 22 13 Someone else N31 49 31 2 12 0 20 10 6 1 2 1 %24 28 28 6 34 0 22 35 43 7 11 7 Both N 82 104 66 15 20 3 54 14 5 11 12 12 %64 60 59 65 57 75 59 48 36 73 67 80 Understanding Partiality in pro-Drop Languages 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 ©2019 John Wiley & Sons Ltd 4.5. The Licensing and Interpretation of NSs in Matrix Clauses The data examined in the previous sections supports the existence of silent A-Topics, in line with the Topic Criterion (1), according to which every predicational sentence contains a position in the C domain endowed with the [+aboutness] feature. As a consequence, when an NS is realized in a matrix clause, a silent A-Topic must be assumed in the relevant C domain (as argued in Frascarelli 2018 and Jim enezFern andez 2016). This is illustrated in (29) for Spanish; (30) is a representation. (29) Quer ıa presentarte a Julio. Es mi mejor amigo. want.PST.1SG introduce.INF.you.CL to Julio be.3SG my best friend ‘I’d like to introduce Julio k to you. He k is my best friend.’ Considering the preference for anovert local antecedent and the partial-pro-drop properties attested in many Spanish varieties, the question arises whether (and to what extent) silent A-Topics are accepted in the matrix C domains of simple clauses in these languages. For this purpose, the survey contained a number of stimuli in which a sentence containing an NS was preceded by a sentence containing two possible overt antecedents, with different syntactic functions. Let us start with the case in which the choice was between two argument DPs as antecedent for a matrix NS: (31) Juan habl o con Pedro ayer. Ahora comprende lo que Juan talk.PST.3SG with Pedro yesterday now understand.3SG the what pas o. happen.PST.3SG ‘Juan k talked to Pedro z yesterday. Now he k /he z understands what happened.’ As we can see in table 9, silent A-Topics are not fully accepted, especially in PS varieties (22% nonacceptance compared to only 5% in AS). When it comes to interpretation, informants of most Spanish varieties mostly select the ‘both’option; this ambiguity is particularly strong for PS speakers, much less so in AS. In Carib the antecedent is less ambiguous than in the Table 9. Matrix NS Italian PS Anda Cast Cat Ext AS Carib Centr Mex Riv Per OK/N 128/128 134/173 81/111 18/23 33/35 2/4 86/91 26/29 12/14 8/15 17/18 13/15 Juan N 68 40 23 6 11 0 31 12 3 2 8 6 %53 30 28 33 33 0 36 46 25 25 47 46 Pedro N 20 7 4120117 22 00 %16 5 5 6 6 0 13 27 17 25 0 0 Both N 40 87 54 11 20 2 34 7 7 4 9 7 %31 65 67 61 61 100 51 27 58 50 53 54 22 Mara Frascarelli and  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 ©2019 John Wiley & Sons Ltd remaining AS varieties. This might be interpreted as evidence against the relevance of a hierarchy of syntactic functions for the interpretation of an NS. Yet in Riv, Per, and most importantly in Carib, the subject of the previous sentence is quite prevalently selected. Let us turn now to a case in which the first sentence includes a preverbal subject DP and a nonargument DP and the second sentence contains an embedded NS and no possible antecedent for that NS: (32) Juan se fue ayer al cine con Pedro. Juan REFL go.PST.3SG yesterday to.the cinema with Pedro S e que estaba muy contento. know.1SG that be.PST.3SG very happy ‘Juan k went to the cinema with Pedro z yesterday. I know that he k /he z was very happy.’ Comparing table 10 with table 9, we can see that acceptability decreases in both PS (from 22% nonacceptance to 35%) and AS (from 5% nonacceptance to 20%). This decrease can be attributed to the fact that the NS is embedded, with no explicit DP in the matrix clause that is a possible antecedent; hence, the closest possible link is a nonlocal silent A-Topic. This clearly represents a “double”interpretive difficulty for a partial-pro-drop speaker. The fact that the acceptability decrease is relatively dramatic for PS, while AS varieties are less affected (with Mex showing no decrease at all and Per showing a small increase), can be taken as evidence that PS varieties are “more partial-pro-drop-like”than AS varieties under this particular condition. This accords with what emerged from the analysis of adverbial clauses (section 4.3) and contrasts with general claims in the literature. These points are taken up in section 5. As for the interpretive data, ambiguity is as usual the most frequent answer, even though the preverbal subject DP is definitely preferred over the nonargument DP, especially in Ext, Carib, and Centr. The final context where we tested the interpretation of NSs is one in which the overt antecedent is a ‘by’phrase, It has been argued in recent works that such an antecedent is totally excluded (see, for example, Samek-Lodovici 1996). A sample sentence is in (33), and the results for this sentence type are given in table 11 (the ‘both’option is omitted because it was never selected by informants). Table 10. Embedded NS with no compatible overt antecedent in the same sentence Italian PS Anda Cast Cat Ext AS Carib Centr Mex Riv Per OK/N 128/128 115/173 69/111 15/23 27/35 4/4 73/91 25/29 11/14 8/15 15/18 14/15 Juan N 62 42 26 4 10 2 30 15 6 1 6 2 %48 37 38 27 37 50 41 60 55 12.5 40 14 Pedro N 7 6 4200 64 11 00 %6 5 6 13 0 0 8 16 9 12.5 0 0 Both N 59 67 39 9 17 2 37 6 4 6 9 12 %46 58 56 60 63 50 51 24 36 75 60 86 Understanding Partiality in pro-Drop Languages 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 ©2019 John Wiley & Sons Ltd (33) El congreso fue inaugurado por el Decano. the conference be.PST.3SG inaugurated by the dean Despu es se fue a dar sus clases. then REFL go.PST.3SG to give.INF his classes ‘The conference was presented by the dean k . Then he k /he z /she w went to teach his/her class.’ Acceptability judgments exhibit a further worsening compared to those in tables 9 and 10. This shows that having a ‘by’phrase as a silent A-Topic represents an additional complication, and, as a consequence, it shows that a hierarchy exists between syntactic functions in terms of eligibility as a silent A-Topic. However, a syntactic block cannot be assumed (pace Samek-Lodovici 1996) but rather an interpretive restriction that is operative at the interface between discourse and syntax. 4.6. The Licensing and Interpretation of NSs across Varieties In light of the different variables tested, the data illustrated, and the results discussed in sections 4.1–4.5, let us now try and consider the different varieties as variables and see if the values obtained provide any statistical association with the dependent variable (i.e., NS acceptability and antecedent selection). Varieties have shown different degrees of partiality depending on different syntactic variables. It is important to consider this aspect in more detail, taking into account NS acceptability first of all. As shown in tables 2, 4, 5, and 8, there are some syntactic contexts in which no partial-NS properties emerge, that is, where NS acceptance is always 100%: (i) embedding under a bridge verb (including double embedding); (ii) embedding under a factive verb (including double embedding), and (iii) intervention of a third-person (non)argument subject. Hence, the creation of a topic chain with a nonlocal matrix preverbal DP is not a disturbing factor for Spanish varieties; in this respect they pattern with consistent-pro-drop languages. However, other variables have shown that Spanish varieties cannot be considered entirely consistent-NS languages. These variables and their (normalized) acceptance rates in each variety are given in table 12. The statistical significance of the deviations attested is indicated graphically: light grey for significant deviations, dark grey background for very significant deviations, and black background for extremely significant deviations. Thus, on the one hand, Centr’s and Per’s deviations with Table 11. ‘By’phrase as antecedent Italian PS Anda Cast Cat Ext AS Carib Centr Mex Riv Per OK/N 128/128 119/173 81/111 15/23 19/35 4/4 59/91 19/29 10/14 9/15 8/18 13/15 ‘By’ phrase N128 108 76 13 15 4 54 15 10 9 7 13 %100 91 94 87 79 100 92 79 100 100 88 100 Someone else N 0 11 524054 0010 %0 9 613210 821 0 012 0 24 Mara Frascarelli and  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 ©2019 John Wiley & Sons Ltd respect to the variables considered are not significant, and thus no partiality is attested in these varieties. On the other hand, Mex seems to go in the direction of qualifying as a partial-NS language, since its values are very close to those attested for Finnish in Frascarelli’s (2018) study; 15 and some degree of partiality can also be found in the other varieties. The deviation attested with adverbial clauses is generally not significant, independently of their position. In terms of the working hypothesis assumed, this result supports the conclusion that no interpretive problem is created as long as an overt antecedent is present in the matrix clause, heading the topic chain. On the other hand, partiality emerges when a silent A-Topic must be established to head the chain and serve as an antecedent. Hence, NSs in topic-less matrix clauses determine a very significant deviation in Ext and Mex, while NSs in embedded clauses with no overt preverbal DP in the matrix clause determine a significant deviation in Anda and Cast and, once more, a very significant deviation in Mex. Interestingly, the strongest deviation emerges when the silent A-Topic to be established is coreferential with a ‘by’phrase, possibly due to a discourse-semantic inconsistency (since it involves treating as an A-Topic an element that is generally considered to be focused, when overtly realized). Let us now consider how these variables determine and/or affect antecedent selection. In particular, since partial-NS languages have been argued to prefer subject antecedence (Holmberg, Nayudu & Sheehan 2009), we will only compare subject and nonsubject values for each variable, setting aside the ‘both’option. As shown by table 13, a significant preference for subject antecedence is only attested when the NS is located in a postmatrix adverbial clause. In all the other cases, a topic chain with a nonsubject element that instantiates a silent A-Topic is always possible (if not preferred, as for instance with prematrix adverbials in Cast, Cat, Centr, and Per), which is consistent with our hypothesis. We will return to these results for final analysis in section 5. Table 12. Relevant variables and significant deviations for each variety % Postmatrix conditional adverbial Prematrix conditional adverbial Matrix clause Embedded +no overt antecedent ‘By’-phrase antecedent Anda 84.5 77.5 75 62 73 Cast 82.5 69.5 82.5 65 65 Cat 94 80 97 77 54 Ext 75 75 50 100 100 Carib 96.5 89.5 93 86 65.5 Centr 85.5 93 85.5 78.5 71.5 Mex 94.5 60 53 53 60 Riv 100 94.5 94.5 83 44.5 Per 93 86.5 86.5 93 86.5 15 Specifically, 100% of Finnish informants accept NSs in postmatrix conditional clauses %, 48% in prematrix conditional clauses, 73% in matrix clauses, and 27% in embedded clauses with no overt antecedent. ‘By’phrases could not be tested since Finnish has no passives. Understanding Partiality in pro-Drop Languages 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 ©2019 John Wiley & Sons Ltd connected with (ii)). This is expressed in the Minimal-Overt-Link Condition (40), which represents a proposal to be used in future research. In other words, these conditions can be taken as valid diagnostics of (degrees of) partiality in NS languages, supporting Chomsky’s (1995) idea that syntax feeds interpretation. Furthermore, since these conditions were designed to check the relevance of having an A-Topic available to serve as an antecedent, we can conclude that the Topic Criterion (1) and the creation of topic chains are requirements for the interpretation of pro in NS languages in general; this supporting the theory that the acceptability and interpretation of NSs crucially depends on an information-structure strategy. Acceptability judgments have also strongly supported the root quality of A-Topics, showing that in many Spanish varieties the association with a subject/topic in the matrix clause is particularly important (even though A-Topics can also be merged in the complements of bridge verbs). This requirement can also be easily derived from the Interface Root Restriction (10), which is part of the present approach and taken as an additional variable to check for partiality across NS languages. Finally, judgments have shown that acceptance decreases in the case of silent ATopics, especially when no overt low copy is available in the chain, in the form of a local aboutness G-Topic. The preference for overt copies (for which the MinimalOvert-Link Condition (40) has been elaborated) can also be taken as a criterion in evaluating partiality. To summarize, the relevance of topic chains, the antecedent role of matrix ATopics, and the importance of overt links are consistent ingredients with the framework assumed. Consequently, a comprehensive account of the NS parameter can be based on the following revision of the Topic Criterion: (45) Topic Criterion (revised) a. A third-person argument pro matches the [aboutness] [ref] features that are encoded in the A-Topic projection (ShiftP) in the C domain. b. A-Topic chains can only be started in root domains (the Interface Root Restriction). c. A-Topic chains across sentences require local links, which can be overt or silent. d. A-Topics can be silent. The first two parts, (45a) and (45b), apply to all NS languages, whereas (45c) and (45d) are subject to parametrization: languages showing partial-NS properties require overt links (according to the Minimal-Overt-Link Condition) and/or an overt A-Topic heading the topic chain. The Topic Criterion, as revised, has advantages over a pragmatic approach, since it provides a formal framework and information-structure-related diagnostics for future comparative investigations across languages. It is also superior to a narrow syntactic approach, since syntactic requirements (or conditions) define grammaticality in clearcut terms and can hardly cope with ambiguity and gradient acceptance of “fuzzy” phenomena (which may be due to ongoing change in parameter settings). 32 Mara Frascarelli and  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 ©2019 John Wiley & Sons Ltd It should be noticed that the conditions proposed to define (degrees of) partiality are independently motivated. As a matter of fact, the relevance of close, intermediate links in long-distance relations is the basis for the Minimal-Link Condition, and island sensitivity is part of the core grammar of languages. As for the importance of spelledout copies in topic chains, conditions on PF merger have been proposed in the literature to explain a number of phenomena and parameters; see, for example, Bobaljik’s (2002) treatment of the VO–OV distinction. 17 As for overt pronouns, the present experiment shows that overt pronouns are not (necessarily) discourse oriented. On the contrary, they tend to be informationstructurally related to the matrix preverbal DP (like NSs), which is taken as the most accessible overt A-Topic. More specifically, the data show that when an exclusive reading is required, in the absence of information-structure cues, the matrix subject is interpreted as an A-Topic and the default discourse interpretation associated with an overt pronominal subject is that of a familiar entity that is part of a topic chain: an aboutness G-Topic. It emerged from the analysis that different varieties can show different sensitivity to one or the other of the two partial-NS conditions. When one or more of these requirements is not met, degradation increases, in line with recent approaches (see Haegeman, Jim enez-Fern andez & Radford 2014, Jim enez-Fern andez 2017, Villata, Rizzi & Franck 2016). In this connection it remains for us to carry out a more detailed study of partial-NS properties in other languages, in contrast with non-NS languages, in order to reach a full understanding of the impact of partial-NS properties across languages and on acquisition processes. References Adger, D. & P. Svenonius. 2011. Features in minimalist syntax. In The Oxford handbook of linguistic minimalism, ed. C. Boeckx, 27–51. Oxford: Oxford University Press. Adli, A. 2011. Gradient acceptability and frequency effects in information structure: A quantitative study on Spanish, Catalan, and Persian. Doctoral thesis, Universit€ at Freiburg, Freiburg, Germany. Alonso-Ovalle, L., S. Fern andez Solera, L. Frazier & C. Clifton. 2002. Null versus overt pronouns and the topic-focus articulation in Spanish. Rivista di Linguistica 14(2):1–19. Alvar, M. 1996a. Manual de dialectolog ıa hisp anica: El espa~ nol de Am erica. Barcelona: Ariel. Alvar, M. 1996b. Manual de dialectolog ıa hisp anica: El espa~ nol de Espa~ na. Barcelona: Ariel. Barbosa, P., M. E. Duarte & M. A. Kato. 2005. Null subjects in European and Brazilian Portuguese. Journal of Portuguese Linguistics 4:11–52. Bianchi, V. & M. Frascarelli. 2010. Is topic a root phenomenon? Iberia 2:43–88. Biberauer, T., A. Holmberg, I. Roberts & M. Sheehan. 2010. Parametric variation: Null subjects in Minimalist theory. Cambridge: Cambridge University Press. Bizzarri, C. 2015. Russian as a pro-drop language: Data and analysis from a new study. Annali di Ca’Foscari 49:335–362. Bobaljik, J. D. 2002. A-chains at the PF-interface: Copies and “covert”movement. Natural Language and Linguistic Theory 20.2:197–267. 17 Specifically, Bobaljik argues that Holmberg’s Generalization is the result of a morphophonological constraint on verb inflection, requiring merger under PF adjacency (supported by VO–OV data). In short, PF can decide which copy to spell out and derive the effects. Understanding Partiality in pro-Drop Languages 33 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 ©2019 John Wiley & Sons Ltd Camacho, J. 2013. Null subjects. Cambridge: Cambridge University Press. Camacho, J. 2016. The null subject parameter revisited: The evolution from null subject Spanish and Portuguese to Dominican Spanish and Brazilian Portuguese. In The morphosyntax of Portuguese and Spanish in Latin America, ed. M. Kato & F. Ord o~ nez, 27–48. Oxford: Oxford University Press. Carminati, M. N. 2002. The processing of Italian subject pronouns. Doctoral thesis, Massachusetts Institute of Technology, Cambridge. Chamorro, G. 2018. Offline interpretation of subject pronouns by native speakers of Spanish. Glossa: A Journal of General Linguistics 3.1:1–16. Chomsky, N. 1982. Some concepts and consequences of the theory of Government and Binding. Cambridge, MA: MIT Press. Chomsky, N. 1995. The Minimalist Program. Cambridge, MA: MIT Press. Cinque, G. 1990. Types of A0-dependencies. Cambridge, MA: MIT Press. Cinque, G. & L. Rizzi. 2008. The cartography of syntactic structures. Studies in Linguistics 2:42–58. Cognola, F. & J. Casalicchio. 2018. Understanding null subjects: A synchronic and diachronic perspective. Oxford: Oxford University Press. Com ınguez, J. P. 2011. Occurrence and interpretation of subject pronouns in temporal embedded clauses in L2 English near-native speakers of L2 Spanish. Ms., Rutgers University, New Brunswick, NJ. Com ınguez, J. P. 2018. The nature and position of subjects in Puerto Rican wh-questions. In Current research in Puerto Rican linguistics, ed. M. Gonz alez-Rivera, 67–89. New York: Routledge. Emonds, J. 2004. Unspecified categories as in the key to root constructions. In Peripheries: Syntactic edges and their effects, ed. D. Adger, C. De Cat & G. Tsoulas, 75–120. Dordrecht, the Netherlands: Kluwer. Filiaci, F., A. Sorace & M. Carreiras. 2014. Anaphoric biases and null overt subjects in Italian and Spanish: A cross-linguistic comparison. Language and Cognitive Processes 29.7:825–843. 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. 2017. Romance pro-drop languages at the interfaces: A comparative analysis. Paper presented at the 31st Going Romance symposium, University of Bucharest, Bucharest, Romania, December 8. Frascarelli, M. 2018. The interpretation of pro in consistent and partial null-subject languages: A comparative interface analysis. In Null subjects in generative grammar: A synchronic and diachronic perspective, ed. F. Cognola & J. Casalicchio, 211–239. Oxford: Oxford University Press. Frascarelli, M. & R. Hinterh€ olzl. 2007. Types of topics in German and Italian. In On Information structure, meaning, and form, ed. K. Schwabe & S. Winkler, 87–116. Amsterdam: John Benjamins. Frascarelli, M. & A. Puglielli. 2009. Position, function, and interpretation of topics in Somali. In Information structure and its interfaces, ed. L. Mereu, 325–348. Berlin: Gruyter. Giv on, T. 1983. Topic continuity in discourse: An introduction. In Topic continuity in discourse: A quantitative cross-language study, ed. T. Giv on, 5–41. Amsterdam: John Benjamins. Gupton, T. & S. Lowman. 2013. An F projection in Cibe~ no Dominican Spanish. In Selected proceedings of the 16th Hispanic Linguistics symposium, ed. J. Cabrelli Amaro, J. Lord, A. de Prada P erez & J. Alana Aaron, 338–348. Somerville, MA: Cascadilla Press. Guti errez-Bravo, R. 2008. Topicalization and preverbal subjects in Spanish wh-interrogatives. In Selected Proceedings of the 10th Hispanic Linguistics Symposium, ed. J. Bruhn de Garavito & E. Valenzuela, 225–236. Somerville, MA: Cascadilla Press. Haegeman, L. 2007. Operator movement and topicalization in adverbial clauses. Folia Linguistica 18:485–502. 34 Mara Frascarelli and  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 ©2019 John Wiley & Sons Ltd Haegeman, L. 2012. Adverbial clauses, main clause phenomena, and the composition of the left periphery. Oxford: Oxford University Press. Haegeman, L.,  A. L. Jim enez-Fern andez & A. Radford. 2014. Deconstructing the Subject Condition in terms of cumulative constraint violation. Linguistic Review 31:73–150. Heim, I. 1982. The semantics of definite and indefinite noun phrases. Doctoral thesis, University of Massachusetts, Amherst. Heycock, C. 2006. Embedded root phenomena. In The Blackwell companion to syntax, ed. M. Everaert & H. van Riemsdijk, vol. 2, 174–209. Oxford: Blackwell. Holmberg, A. 2005. Is there a little pro? Evidence from Finnish. Linguistic Inquiry 36:535–564. Holmberg, A. 2010. Null subject parameters. In Parametric variation: Null subjects in minimalist theory, ed. T. Biberauer, A. Holmberg, I. Roberts & M. Sheehan, 88–124. Cambridge: Cambridge University Press. Holmberg, A., A. Nayudu & M. Sheehan. 2009. Three partial null-subject languages: A comparison of Brazilian Portuguese, Finnish, and Marathi. Studia Linguistica 63:59–97. Huang, C.-T. J. 1989. Pro-drop in Chinese: A generalized control theory. In The null subject parameter, ed. O. Jaeggli & K. Safir, 185–214. Dordrecht, the Netherlands: Kluwer. Jaeggli, O. 1982. Topics in Romance syntax. Dordrecht, the Netherlands: Foris. Jaeggli, O. & K. Safir. 1989. The null subject parameter and parametric theory. In The null subject parameter, ed. O. Jaeggli & K. Safir, 1–44. Dordrecht, the Netherlands: Kluwer. 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. 2017. Prepositions and islands: Extraction from dative and accusative DPs in psych verbs. In Language use and linguistic structures: Proceedings of the Olomouc Linguistics Colloquium 2016, ed. J. Emonds & M. Janevob a, 155–172. Olomouc: Palack y University Olomouc Press. Jim enez-Fern andez,  A. L. & S. Miyagawa. 2014. A feature-inheritance approach to root phenomena and parametric variation. Lingua 145:275–302. Kato, M. 2000. The partial pro-drop nature and the restricted VS order in Brazilian Portuguese. In Brazilian Portuguese and the null subject parameter, ed. M. A. Kato & E. Negr~ ao, 223– 258. Frankfurt, Germany: Vervuert. Kato, M. 2011. Acquisition in the context of language change: The case of Brazilian Portuguese. In The development of grammar: Language acquisition and diachronic change, ed. E. Rinke & T. Kupisch, 309–330. Amsterdam: John Benjamins. Kayne, R. S. 2016. The silence of heads. Studies in Chinese Linguistics 37.1:1–37. Krifka, M. 2007. Basic notions of information structure. Interdisciplinary Studies on Information Structure 6:13–55. Kuroda, S.-Y. 1965. Generative grammatical studies in the Japanese language. Cambridge, MA: MIT Press. Lambrecht, K. 1994. Information structure and sentence form: Topic, focus, and the mental representation of discourse referents. Cambridge: Cambridge University Press. Levinson, S. C. 1991. Pragmatic reduction of the Binding Conditions revisited. Journal of Linguistics 27:107–161. Manzini, M. R. 1992. Locality: A theory and some of its empirical consequences. Cambridge, MA: MIT Press. Mao, T. & F. Meng. 2016. The cartographic project of the generative enterprise: An interview with Guglielmo Cinque. Language and Linguistics 17.6:917–936. Meinunger, A. 2004. Verb position, verbal mood, and the anchoring (potential) of sentences. In The syntax and semantics of the left periphery, ed. H. Lohnstein & S. Trissler, 313–341. Berlin: Mouton de Gruyter. Modesto, M. 2000. Null subject without “rich”agreement. In Brazilian Portuguese and the null subject parameter, ed. M. A. Kato & E. Negr~ ao, 147–174. Frankfurt, Germany: Verveurt. Montalbetti, M. M. 1986. How pro is it? In Studies in Romance linguistics, ed. O. Jaeggli & C. Silva-Corval an, 137–152. Dordrecht, the Netherlands: Foris. Understanding Partiality in pro-Drop Languages 35 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 ©2019 John Wiley & Sons Ltd Ojea, A. 2017. Core intentional features in the syntactic computation: Deriving the position of the subject in Spanish. Lingua 195:72–91. Ord o~ nez, F. & A. Olarrea. 2006. Microvariation in Caribbean/non Caribbean Spanish interrogatives. Probus 18.1:59–96. Perlmutter, D. 1971. Deep and surface constraints in syntax. New York: Holt, Rinehart and Winston. Pesetsky, D. & E. Torrego. 2007. The syntax of valuation and the interpretability of features. Amsterdam: John Benjamins. Pe skov a, A. 2014. Information structure and the use of pronominal subjects in Spanish. In Methodological issues in the study of information structure, ed. D. El Zarka & S. Heidinger, 43–67. Graz, Austria: Grazer Linguistische Studien. Posio, P. 2012. Pronominal subjects in Peninsular Spanish and European Portuguese: Semantics, pragmatics, and formulaic sequences. Helsinki: University of Helsinki Press. Reinhart, T. 1981. Pragmatics and linguistics: An analysis of sentence topics. Philosophica 27:53–94. Rizzi, L. 1986. Null objects in Italian and the theory of pro.Linguistic Inquiry 17:501–558. Rizzi, L. 1990. Relativized Minimality. Cambridge, MA: MIT Press. Rizzi, L. 1997. The fine structure of the left periphery. In Elements of grammar: Handbook in generative syntax, ed. L. Haegeman, 281–337. Dordrecht, the Netherlands: Kluwer. Rizzi, L. 2004. Locality and left periphery. In The cartography of syntactic structures, vol. 3, Structures and beyond, ed. A. Belletti, 223–251. Oxford: Oxford University Press. Rizzi, L. 2006. On the form of chains: Criterial positions and ECP effects. In Wh-movement: Moving on, ed. L. L.-S. Cheng & N. Corver, 97–134. Cambridge, MA: MIT Press. Roberts, C. 2003. Uniqueness in definite noun phrases. Linguistics and Philosophy 26:287–350. Rodrigues, C. 2004. Impoverished morphology and A-movement out of case domains. Doctoral thesis, University of Maryland, College Park. Saab, A. 2016. On the notion of partial (non-)pro-drop in Romance. In The morphosyntax of Portuguese and Spanish in Latin America, ed. M. A. Kato & F. Ord o~ nez, 49–57. Oxford: Oxford University Press. Samek-Lodovici, V. 1996. Constraints on subjects: An optimality theoretic analysis. Doctoral thesis, Rutgers University, New Brunswick, NJ. Sigurðsson, H. 2004. Meaningful silence, meaningless sounds. Linguistic Variation Yearbook 4:235–259. Tanaka, H. 2016. A minimalist analysis of English topicalization: A phase-based cartographic complementizer phrase (CP) perspective. Journal of UOEH 38.4:289–389. Ticio, M. E. 2018. On Puerto Rican subjects. In Current research in Puerto Rican linguistics, ed. M. Gonz alez-Rivera, 90–108. New York: Routledge. Toribio, A. J. 2000. Setting parametric limits on dialectal variation in Spanish. Lingua 110.5:315–341. van der Sandt, R. 1992. Presupposition projection as anaphora resolution. Journal of Semantics 9:333–377. Villata, S., L. Rizzi & J. Franck. 2016. Intervention effects and Relativized Minimality: New experimental evidence from graded judgments. Lingua 179:76–96. Mara Frascarelli Universit a degli Studi Roma Tre Dipartimento di Lingue Letterature e Culture Straniere Via del Valco di San Paolo, 19 Roma 00146 Italy mara.frascarel[email protected] 36 Mara Frascarelli and  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 ©2019 John Wiley & Sons Ltd  Angel L. Jim enez-Fern andez Universidad de Sevilla Departamento de Filolog ıa Inglesa (Lengua Inglesa) Calle Palos de la Frontera, s/n Sevilla 41004 Spain [email protected] Understanding Partiality in pro-Drop Languages 37 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 ©2019 John Wiley & Sons Ltd