Biased interrogatives in Camuno

2.1 Interrogative structures in Camuno

Subject-clitic inversion represents the default strategy to express interrogative force, in polar (1-a) and wh-questions, both in argumental (1-b) and in adjoined (1-c) position, a property common to other Northern Italian dialects (Manzini and Savoia 2005, for an overview):^[1]

Two pan-dialectal strategies are attested for wh-interrogatives: (i) fronting via (semi-)clefting (2-a); (ii) clause-internal wh-phrases (2-b):

The clause-internal wh-phrase in (2-b) is hosted in a position targeted by A-bar movement at the edge of the vP phase. In other words, (2-b) is an instance of fake in situ distribution of wh-phrases (Bonan 2017). Two additional strategies are available for at least some dialects of Camuno: The Italian-like fronting in (3-a) and wh-doubling as in (3-b):

I consider (3-a) to be an Italianism since it mirrors exactly the Italian wh-phrases distribution. The order is also more frequently attested when the topic of the conversation is related to contexts where the speaker would use Italian in real-life situations. These are topics such as education, legal matters, and health in relation to medical professionals. (3-b) is only attested in a handful of dialects of Camuno and presents a number of descriptive challenges outside the scope of this work (Fiorini and Neagu 2023, for a descriptive analysis). For these reasons, and for the sake of clarity, here I only discuss (2-a) and (2-b), which represent the most frequent strategies employed by my informants. Nevertheless, the analysis of biased interrogatives proposed here is predicted to hold for any possible question-formation strategy, as not strictly correlated with a specific syntactic behavior. The particle po is, in fact, attested in all the dialects in Valcamonica, albeit with minor interpretative differences depending on its distribution.

2.2 The discourse particle po

Cognate forms of the particle po are attested in numerous Northern Italian Dialects. For instance, see Munaro and Poletto (2005), for Piedmont varieties; Hack (2014), and Dohi (2019) for Dolomitic varieties; Hinterholzl and Munaro (2015) for Bellunese; Coniglio (2008), Cognola and Cruschina (2021), and Fiorini (2022) for Italian. Their etymological source is invariably identified in the Latin adverb post ‘afterward,’ which undergoes a process of grammaticalization from purely temporal adverb to functional particle. Different stages of this process of Grammaticalization are visible as synchronic variation across varieties spoken in the eastern Alpine region (Hack 2014).

The origin of these particles is transparent in the Italian structures in (4). In particular, in (4-a), poi has the canonical function of creating a logic (in this case, temporal) relationship between the two clauses. In contrast, poi can refer to a previous conversational event (Cardinaletti 2011) or a referent salient in the conversational context (4-b). Finally, poi can be used as a discourse marker, that is, a “sequentially dependent element which brackets units of talk” (Schiffrin 1987, 31) with an additive value or, frequently, with a contrastive, adversative, or dismissal meaning (4-c):

The literature discusses these elements as mainly attested in questions and as expressing a number of pragmatic values such as (i) astonishment (Croatto 1997), (ii) emphasis and focus (Poletto and Zanuttini 2003a), (iii) surprise and indignation (Hack 2014), (iv) ‘inability to find an answer’ or ‘concern or interest in the information being asked for’ (Coniglio 2008).

From a descriptive point of view, in the appropriate context, po in Camuno can express all these pragmatic values. Differently from cognate forms and from numerous particles attested in several Romance varieties spoken in the Italian peninsula, the distribution of po is not connected to either the type of the hosting sentence or a specific position in the clause.

2.3 Alternative semantics

Here, I adopt an alternative semantic approach, which, for space constraints, cannot be fully described here. The following is a brief overview of the fundamental tenets of this framework: the reader is referred to Rooth (2016) for a comprehensive overview.

Hamblin (1973, 1976) builds upon the seminal work of Montague (1970) on English as a formal language, expanding his line of analysis to capture the properties of questions. More recently, Rooth (1985, 1992, 1996) adopted some of the basic intuitions of the previous works to develop a formal framework to analyze focus structures. The core idea introduced by these accounts is that sets of alternatives are evoked by non-truth-conditional elements. The framework became known as Alternative Semantics and applies to the analysis of a variety of objects requiring “semantic, pragmatic, or discourse-structural operations or constraints referring to ‘alternative’ phrasal meanings” (Rooth 2016, 1).

The alternatives activated by questions are formalized as ‘answerhood conditions,’ that is, the conditions under which an interrogative can be answered. Take a wh-question like (11-a), which introduces the presupposition in (11-b):

The Hamblian set of alternatives evoked by the wh-pronoun can be represented as in (12-b), deriving from the condition in (12-a):

Following a Roothian approach, every expression has an ordinary semantic value and a focus semantic one (traditionally represented via a superscript f). The answer to a question, i.e., an f-marked element, is selected by the presuppositional operator ∼ , which adjoins to the focal-XP associating to a variable that gets its content from a target in the context. A simple NP like John can thus be represented as (13-b), where D e is contextually defined set of individuals:

In this view, the set of alternatives is quantified upon by ∼ , which selects a discourse salient argument C. This ensures that the ordinary semantic value of the relevant proposition is a subset of its contextually defined focus semantic value, i.e., man(John) & smoke(John), coherently with the presupposition introduced by the wh-phrase.

Focus associated particles like only or even play a similar role by operating on the relevant set of alternatives.^[3] In Roothian terms, only selects all and exclusively the alternative propositions true in w, and even orders the alternatives, and selects the least likely proposition in w.

Numerous proposals regarding the precise ways these operators manipulate the set of alternatives are present in the literature (e.g., Slade 2011, Liu 2018, Greenberg 2020). This article aims to explore the properties of po in relation to the sets of alternatives activated by the hosting interrogative structure. In particular, I take po to be a variable over choice functions (Cable 2007, Reinhart 1997, Slade 2011, 2019), defined as “a function which when applied to a set returns a member of that set” (Slade 2019, 21).

2.4 Semantic source for the interpretation of po

I argue that the possible interpretations discussed above are a by-product of the non-truth conditional (i.e., presuppositional) semantic import of po, which conveys the speaker’s doxastic evaluation of the po-proposition.

In particular, I argue that the meaning of po is strictly connected to the interrelated concepts of factuality and uncontroversiality, informally defined as in (14-a) and (14-b):

The definitions above are rooted in the beliefs of the speakers regarding p. Consequently, I argue that po operates on the Common Ground Management level, which “contain[s] information about the manifest communicative interests and goals of the participants” (Krifka 2008, 246). Initially proposed to handle the interaction between questions and common ground, this level of organization of the common ground operates over the speaker’s doxastic evaluation of the common ground, which can be asymmetric by definition. In other words, po refers to the way the common ground should develop according to at least one of the participants in a conversation, regardless of the actual state of affairs.

In the declarative/exclamative structure above, the use of po signals that the speaker believes that a proposition p should be uncontroversial (U). However, this does not hold at the utterance time. In other words, the speaker remains committed to the truth of p, but they have reasons to believe the addressee is not. This results in the under-specified mirative (Delancey 1997) interpretation identified in the literature regarding cognate forms of po listed above.

The typical conversational progression licensing po is articulated as follows (Bs = Speaker’s beliefs; Ba = Addressee’s beliefs), (15):

At t-1, the speaker believes that p is uncontroversial and, crucially, that p should be uncontroversial (15-a). At the utterance time t0, however, they believe that the addressee is considering ¬ p, i.e., p is not uncontroversial in the utterance world w. Consider (16):

The utterance in (16) is only felicitous if the speaker believes that the addressee is considering ¬ p (i.e., that ‘her daughter’ works somewhere other than Breno). The speaker, in fact, believes that the addressee should know that p is factual, i.e., that p should be uncontroversial in w.

Importantly, while the speaker usually has some evidence justifying their beliefs, this is not necessary for the particle to be licensed. Consider (17):

There is no reason in the context in (17) for the speaker to believe that the addressee should know p (i.e., that vowel harmony is the typological property of Turkish resulting in the pattern that the addressee is describing). Nevertheless, the sentence, albeit impolite, is grammatical and felicitous. For instance, let us assume that the speaker is an expert in language typology. In that case, the categorization of p as uncontroversial and factual would be trivial for them in a professional setting, so po is licensed. The beliefs can be challenged by the addressee, whose reaction could be along the lines of ‘I’ve never studied linguistics, po!’ since they consider the fact that they are not familiar with linguistic phenomena to be uncontroversial. In other words, the use of po is not constrained by anything but the beliefs of the speaker, no matter how implausible they may be, and it is purely a conversational implicature.

The type of presupposition introduced by po can be captured by the concept of ‘expressive presupposition’ proposed by Sauerland (2007) and Schlenker (2007). This type of presupposition contains three indexical elements g, which refer to the speaker (s0), the addressee (a0), and the situation in the utterance context (w0). The speaker is the only authority regarding their beliefs on both the addressee and the context (Potts 2015), consequently, the presupposition is self-verifying (hence the felicity of (17). Modeling it after Grosz (2020)’s proposal for German ja, and the adaptation in Fiorini (2022), the (expressive) presupposition licensing po can be expressed as in (18)^[5]:

3.1 Manipulating sets of alternatives

In Section 2.2, I introduced the concept of common ground management as the level on which po operates. In the licensing utterance context, in fact, it signals that the host proposition (po-p) is no longer believed to be factual. However, the speaker believes that po-p should be true, albeit not factual and, more specifically, that it is the only appropriate alternative belonging to 〚 p 〛 f . This situation is clearly paradoxical and cannot be captured within the common ground content since, at the purely epistemic level, p can only be either true or false. Conversely, at the doxastic level handled by the common ground management, the indication of p to be interpreted as the only possible true alternative(s) is identified as the source of the bias. The latter can be considered to be due minimally to the number of alternatives that the po-interrogative activates. Consider the unmarked version of (19) in (23-a), and an appropriate answer like (23-b):

Following the Hamblian line of the analysis presented above, (23-a) activates the set of alternatives in (24), labeled D food , i.e., a contextually defined set of dishes:

As standardly assumed, the wh-question introduces the existential presupposition in (25-a), where D C is the set of contextually defined alternatives (i.e., possible answers). That is, in this case, D C = D food . The variable x is then solved by one of the propositions part of 〚 p 〛 f , (25-b) (‘rabbit,’ in this case). In the answer, one of the alternatives is selected and the variable value defined, resulting in the answer in (25-c):

An interrogative hosting po has the same properties of unmarked cases like (23-a). However, the particle must be interpreted by the semantic component as evoking a different set 〚 p 〛 f , consisting of the same members of D food , except for the alternative (or subset of alternatives) deemed non-factual by the speaker. One possible way to capture these facts is to assume that po-p simply refers to a different set of alternatives, in which case it could be thought of as an operator like ∼ , which ensures the appropriateness of the set with respect to the context which, in this case, excludes the most likely proposition(s) from the set. A second possible line of analysis can be related to the quantifying properties of po, which in this case could be thought as operating on 〚 p 〛 f independently, similarly to, in this case, only, which as mentioned above selects all and exclusively the true proposition in w. Developing the exact semantic mechanism responsible for these operations must be left to future research. Focusing on the effects of the semantic import of po on the pragmatic interpretation, I believe that considering po as an operator over sets of alternatives is a more straightforward way to account for the data.

3.2 Po in polar questions

One last meaning associated with po is attested exclusively in polar questions, and it is reminiscent of the one described for English echo questions. These types of interrogatives are characterized by a non-genuine nature in that they express surprise over a proposition newly introduced to the common ground (Sobin 2010 : for an overview of the properties of echo questions).^[6] Consider (33):

The use of po in (33) conveys the usual mirative reading by seeking confirmation of the veridicality of the unlikely proposition p, which the speaker, however, has already accepted and added to the common ground.

I argue that this reading directly results from the presupposition introduced by po. Recall that in wh-questions, po signals that at least one of the propositions part of 〚 p 〛 f is non-factual. In Yes/No questions, 〚 p 〛 f consists of two alternatives (34):

The speaker believes that the unlikely proposition p must be, indeed, surprisingly factual. In the majority of cases, this is inferred by the context. In particular, p is considered to be unlikely, but also the only possible true alternative in w. The negation of p (i.e., ¬ p) is, in fact, believed to be false, albeit more likely. In other words, as for the other types of interrogatives, po eliminates one of the focus alternatives to p. In the case of polar questions, by eliminating one of the possible answers, only p can be true in w, and the speaker expresses their surprise over this information.

The characteristics of the focus semantic of a polar question have immediate consequences on the distribution of po-interrogatives. In fact, some speakers do not accept po in yes/no questions because consider it to be redundant. While polar po-interrogatives are accepted by the vast majority of my informants, all the elicited occurrences are characterized by marked prosody. When prompted, (35a) is interpreted exactly as (35-b):

The interpretation of both questions in (35) is marked similarly to the po-interrogatives in the previous sections. In this case, the speaker accepts the proposition and uses (35) to convey their surprise over the veridicality of the unlikely proposition Carla made rabbit. Prosody alone is sufficient to obtain an echo interpretation so that (35-b) can be perceived as redundant and not accepted by numerous speakers.^[7]

Conversely, wh-questions marked by po and wh-questions with echo interpretation are incompatible.

This is the direct consequence of the property of po-p to be considered to be non-factual. The licensing contexts for (36-a) and (36-b) are, in fact, incompatible. The former requires p to be known by the speaker at the utterance time, while the latter requires the speaker to believe ¬ p.