In his review of Cognola (Reference Cognola2013), Johan Brandtler (Brandtler Reference Brandtler2014) highlights the importance of its empirical contribution and raises some problems with the proposed analysis. In this reply, I would like to address these problems and respond to some of Brandtler's comments and criticisms.

Cognola (Reference Cognola2013) is a study the syntax of the finite verb in Mòcheno, a German dialect spoken in three villages of the Fersina valley (Northern Italy). Due to its peculiar sociolinguistic situation (speech island, all speakers are bilingual), most of the research on Mòcheno has highlighted its differences from modern German varieties, explaining them as the result of contact with the surrounding Italian varieties, or as conservative traits (see Rowley Reference Rowley2003 and previous work). As far as syntax is concerned, the only explanation of the facts available in the literature is in terms of contact/competing grammars (see Rowley Reference Rowley2003 and references cited there). According to this model of grammar, different word orders (such as OV and VO, or verb second (V2) vs. non-V2) co-exist and are in free variation in the same language as a consequence of the availability to speakers, typically in situations of bilingualism, of two competing grammars, each featuring one of the competing orders.

In Cognola (Reference Cognola2013:67–74), the predictions of the contact hypothesis are tested in a series of contexts which had not been considered before and lead the author to the conclusion that it does not hold in the specific case of Mòcheno. This conclusion is not based on a commitment to a particular theoretical framework, but it is the result of the empirical work carried out to test the predictions of the contact hypothesis.

There are two types of arguments which challenge the validity of the contact hypothesis for Mòcheno. The first has to do with the nature of optionality. Svenonius (Reference Svenonius and Svenonius2000:280) claims that the competing grammars hypothesis is appealing only when optionality is ‘rampant’, i.e. when it is found in all syntactic environments. The empirical facts of Mòcheno are shown not to be captured by the definition of optionality given by Svenonius, since both syntactic and information-structure factors rule the distribution of the word orders in competition. Therefore, optionality is only apparent in this language.

Let us take, for instance, the possibility of having both (i) OV and VO word orders, and (ii) V2 and V3 orders in X–V sentences with a DP subject (i.e. absence vs. presence of subject–finite verb inversion). For both phenomena, the two options are available in declarative main clauses, but not in interrogative main clauses, where only one of them is grammatical/felicitous (Cognola Reference Cognola2013:73, 159–162). The fact that optionality disappears when the syntactic context is changed, and that this happens unlike in the grammars of German and of the surrounding varieties of Italian (which are OV and VO; V2 and non-V2 coherently in all contexts) runs counter to the predictions of the contact hypothesis. Moreover, a closer investigation of the data which considers not only whether sentences are grammatical but also whether they are felicitous shows that, when two options are available, they have specialized for different functions connected with information structure. For instance, DP subject–finite verb inversion does not take place when the subject is a topic, whereas the subject must follow the verb when it is a new-information focus (Cognola Reference Cognola2013:154–162).Footnote ¹ Again, this is something not predicted by the contact hypothesis.

The second type of arguments challenging the contact hypothesis for Mòcheno concerns the presence of autonomous developments, i.e. phenomena which are absent from the contact languages: Trentino dialect and regional Italian. One clear example of this is found in the morphology and the syntax of subject pronouns. Mòcheno exhibits three morphologically distinct classes of subject pronouns, each of which has specific syntactic and discourse properties, which, crucially, differ from those of the neighbouring German varieties and of the surrounding varieties of Italian. Moreover, the syntactic and discourse properties of subject pronouns in the German and the Italian varieties are systematically excluded by all consultants, i.e. the autonomous development is the only available option in this area of Mòcheno grammar (see Cognola Reference Cognola2013:79–94, 102–109).

Cognola (Reference Cognola2013) takes these arguments as evidence in favour of the fact that the notion of optionality predicted by the contact hypothesis does not stand up to a more detailed investigation of the data, and proposes an alternative explanation for the observed variation – one that treats it as the result of rules internal to a single Mòcheno grammar, which is similar to that of a number of languages considered in the literature to be relaxed V2 languages (see e.g. Poletto Reference Poletto, Barbiers, Cornips and van der Kleij2002, Benincà Reference Benincà, Zanuttini, Campos and Herburger2006, Holmberg Reference Holmberg2015).Footnote ²

Brandtler (Reference Brandtler2014:108) writes: ‘the initial classification of Mòcheno as a V2 language is based more on theoretical reasoning than on undisputable empirical facts’. He notes in particular:

Brandtler's point is conceptually correct, and Cognola (Reference Cognola2013) addresses it in connection with both Old Romance and Mòcheno (see Cognola Reference Cognola2013:113–138). Drawing on the literature on relaxed V2 languages, it is shown that V2 should not be considered as a linear restriction, but as a rule forcing the finite verb and one XP to move to CP in all main clauses due to the presence of an EPP feature on a C head (see e.g. Roberts Reference Roberts2004). What is special about relaxed V2 languages is that these movements take place in an articulated left periphery (see Rizzi Reference Rizzi and Haegeman1997, Benincà & Poletto Reference Benincà and Poletto2004) and only a subtype of constituents, typically foci, ‘count’ for V2, i.e. are able to create a Spec/head configuration with the finite verb. In relaxed V2 languages constituents that do not count for V2 can be combined with XPs unable to satisfy the EPP feature, typically topics, leading to V3 word order. Crucially, the order of the constituents in the left periphery of relaxed V2 languages is fixed, with topics always preceding foci.

Given this, how can we establish whether a language is a relaxed V2 or a non-V2 language, i.e. whether or not it requires V-to-C movement in all main clauses? A relaxed V2 language has the correlated properties of V2 languages, typically subject–verb inversion and asymmetries between main and embedded clauses (see den Besten Reference den Besten and Abraham1983). Mòcheno, for instance, allows subject–verb inversion, and also has an asymmetric distribution of subject pronouns in main vs. embedded clauses, which is caused by the position of the finite verb (C° in main clauses vs. T° in embedded interrogatives, see Cognola Reference Cognola2013:194–211). Further evidence in favour of the obligatory movement of the finite verb to C° in the main clauses of relaxed V2 languages comes from the syntax of pronouns. As discussed in detail by Benincà (Reference Benincà, Zanuttini, Campos and Herburger2006) for Old Romance and Cognola (Reference Cognola2013:Chapters 3 and 4) for Mòcheno, relaxed V2 languages display a high level of syntactically-triggered (movement of the finite verb to C°) enclisis of pronouns, which does not occur in non-V2 varieties. Therefore, the fact that the enclisis of the subject pronoun is restricted to all main clauses with the order X–V in Mòcheno is a clear indication of its V2 character, and contrasts sharply with the surrounding Trentino dialect, which is non-V2 and has subject clitics, and in which enclisis occurs only in wh-main interrogative clauses (residual V2 in the sense of Rizzi Reference Rizzi, Belletti and Rizzi1996:64).

Although many relaxed V2 languages are also pro-drop or partial pro-drop languages, the connection between V2 and pro-drop has only been partially investigated in the literature on V2.Footnote ³ Cognola (Reference Cognola2013) suggests for Mòcheno that the V2 property, i.e. the movement of the finite verb and of one XP to CP, co-occurs with the possibility of having a null category in Spec,TP, where NP subjects can never appear, unlike in German and English. That is, preverbal NP subjects appear in the left periphery (Spec,TopicP or Spec,FocusP), whereas NP subjects in inversion contexts are always new-information foci that appear in Spec,vP (i.e. below sentential adverbs and negation). It is assumed that pro is an expletive null category, which is licensed by the subject clitic pronouns (see de Crousaz & Shlonsky's Reference de Crousaz and Shlonsky2003 analysis of Romance clitics and references cited there) or simply by the finite verb in C°.

This analysis implies that Mòcheno should be analysed as a partial pro-drop language, a claim that may appear to be controversial, given that referential null subjects cannot be licensed in Mòcheno (the only exception being the second person singular in inversion contexts in the varieties of Fierozzo and Roveda). In fact, Brandtler (Reference Brandtler2014:110) writes:

Mòcheno is claimed to be a partial pro-drop language on the basis of the fact that, despite its lacking of the possibility of licensing referential subjects in all contexts, it has three out of the four properties ascribed in the literature to pro-drop languages, namely (i) free inversion, (ii) absence of that-trace effects, and (iii) expletive null subjects. If we take Roberts & Holmberg's (Reference Roberts and Holmberg2010:10–12) implicational scale of the properties of pro-drop languages, we see that Mòcheno is positioned between a consistent null-subject language such as Italian, and German, a language that only allows expletive null subjects (Cognola Reference Cognola2013:Chapter 5).Footnote ⁴

1. (1)

   

The claim that Mòcheno has expletive null subjects is considered problematic in Brandtler's review. Mòcheno patterns fully with standard German when impersonal passive constructions are considered. The expletive pronoun appearing in both languages (es) is a CP expletive, i.e. it disappears when another XP is fronted, leading to a null-subject sentence. Therefore, both Mòcheno and German allow for expletive null-subjects in passive constructions. When active sentences involving a generic reading are considered, German and Mòcheno differ (Cognola Reference Cognola2013:179–180). In both languages the generic pronoun is es, which in German behaves like a CP expletive, i.e. it disappears in all inversion contexts, whereas it is a TP expletive element in Mòcheno, i.e. it cannot be null in inversion. The behaviour of generics in Mòcheno is problematic for its classification as a partial pro-drop language, since, as discussed by Holmberg (Reference Holmberg2005:540), generics should be null in partial pro-drop languages, as Mòcheno is claimed to be. However, in the light of the other independent evidence in favour of the pro-drop nature of Mòcheno, the presence of null expletives in impersonal passives of Mòcheno in considered to be a sufficient condition for its classification as a language allowing for null expletives, and the absence of null generics is seen as an issue of Mòcheno grammar that needs to be better understood. Biberauer & Cognola (Reference Biberauer and Cognola2014) have shown – through an examination of a wide range of contexts involving a generic reading (see e.g. Gast & van der Auwera Reference Gast, van derAuwera, Bakker and Haspelmath2013) – that Mòcheno does actually display null generics, though not in all generic contexts. This finding indicates that the connection between generics and pro-drop may be subtler than has been proposed by Holmberg (Reference Holmberg2005), and that the possibility of dropping the expletive in the sentence-internal position leading to pure pro-drop does not affect all generic contexts in all partial pro-drop languages.

In the light of these facts, and of the theoretical account, Brandtler's (2014:110) comment that

can be seen as oversimplifying the whole picture. In order to arrive at a comprehensive account of the observed variation and to capture the complex Mòcheno data set beyond a simple descriptive account, the theoretical tools need to be pushed as far as possible. Cognola (Reference Cognola2013) tries to build a theory starting from the empirical data, rather than fitting the empirical data into an existing theory. In doing this, Cognola (Reference Cognola2013) mainly applies the inductive method, i.e. making observations, formulating generalization and tentative hypotheses, testing them, and trying to arrive at an account. However, as this is a study of syntax, which has a long tradition in generative grammar, the deductive method is also applied, i.e. the account is grounded in the theory (as in the case of the classification of Mòcheno as a type of V2 language), to the extent that the theory is corroborated by the data.

To sum up, the main achievements of Cognola (Reference Cognola2013) are its challenging of the tenets of the received analysis of Mòcheno syntactic variation in terms of contact/presence of competing grammars, and its contribution to an understanding of the nature of the V2 phenomenon through a novel analysis of a relaxed V2 language. Whether this alternative account is worth pursuing will be determined by further research.