Language

1.1. Enclitic =to in hindi

In grammars of Hindi such as McGregor 1976 and Koul 2008, the enclitic =to is discussed in passing with a few illustrating examples. McGregor (1976:141) notes that ‘in non-initial sentence or clause position, to usually suggests that the given sentence or clause expresses an idea at variance in some way with what precedes (whether the content of a locution, an unexpressed thought or an action), or modifying it in some way’. According to Koul (2008:155), ‘to is mostly used as an emphatic [End Page 717] marker and also denotes contrast’. Montaut (2015:269) states that =to ‘constructs a particular kind of theme’, which operation involves ‘either a contrast with another term belonging to the same paradigm, or a contrast in judgements on the same term’. Another use of =to, for Montaut, is ‘limiting the relevance of an argument previously proposed by the other speaker, while pretending to confirm it’ (2015:275). To start off the discussion, I use some examples of enclitic =to from Montaut 2015, which provides a detailed exposition of =to’s distribution. Montaut labels this type of occurrence of =to, in which it is a clitic associating to its left, the ‘thematic particle’ =to.

In each example in 1, there is a sense in which the use of =to evokes contrast with contextually salient alternative propositions to the prejacent, as I have tried to convey with a relevant inference associated with each example signaled by ‘⇝’. For example, in 1a, the speaker (S) indicates that the addressee’s (A) current solitary existence contrasts with other times; in 1b the speaker indicates that the Hindi book’s being given tomorrow contrasts with when the other books are being given; and in 1c, the speaker indicates that their not being a recipient of X’s generosity is in contrast with other individuals being such.²

(1)

The contrastive function of =to is certainly salient, but when one looks systematically, it emerges that contrastive effects are only part of a wider set of effects of =to. Section 2, which focuses on =to in declarative clauses, classifies many of these effects, considering the role of salient questions in the context as well as the properties of expressions hosting =to. [End Page 718]

1.2. Background assumptions and terminology

In what follows, I assume that =to uniformly makes its meaning contribution at the propositional level regardless of its location in the clause it appears in. This means that =to comments on the way the proposition expressed by its prejacent relates to the larger discourse. The location of =to within the clause is understood to provide cues to the nature of the question/issue the prejacent is taken to address. I broadly assume the Stalnakerian model of discourse (Stalnaker 1978), according to which discourse evolves against a dynamically evolving set of background assumptions—the common ground—that models the shared aspect of interlocutors’ information states. The common ground is thus the set of propositions that interlocutors take themselves and each other to believe at any point in discourse. The utterance by an interlocutor of a declarative sentence with falling intonation brings about a proposal to add its propositional content to the common ground. The utterance of an interrogative sentence with rising intonation brings about a request to resolve the question it denotes via future discourse moves.

Throughout this article, I use the expressions question and issue interchangeably. Informally, both expressions refer to the kind of things interlocutors raise and seek to resolve in discourse by gaining more information that can be added to their evolving information states. As is usual, I take questions/issues to be semantic objects that are the denotations of interrogative sentences in a language, but have an existence that is independent of the linguistic devices used to access them. Less standardly, I distinguish between an explicitly raised question/issue and a speaker-salient question/issue. The former corresponds to a discourse move (possibly using an interrogative) that is made by an interlocutor at a context. The latter is the question the interlocutor takes their utterance to address at a context. While this distinction has not been explicitly made in the literature, work on contrastive topics and how they relate to strategies of inquiry (Büring 1997, 2003, Roberts 2012 [1996]) has shown that speakers regularly use prosodic marking to convey that they do not consider the explicit question asked by an interlocutor to be the most salient or strongest question at that context.³ Krifka (2008) implicitly assumes such a decoupling in introducing delimitation as a commonground management function. Delimitation is the indicating by a speaker that the contribution to the common ground effected by a given discourse move is not sufficient for addressing the communicative needs at that point in discourse.⁴ We will see that this divergence between the perspectives and priorities of interlocutors, with respect to what questions should be/can/are expected to be resolved at that context, is something that =to exhibits sensitivity to. [End Page 719]

2.1. Constraints on =to-hosting

The data above shows that =to may be hosted by constituents that contribute both given and new information. Moreover, there are interpretational differences associated with the prosodic realization of the =to host as well as the post-=to material. On closer observation, we can identify clear constraints on the prosodic marking on =to hosts in declaratives. These are stated with illustrations in 6. [End Page 722]

(6)

In discussions about information structure in Hindi, it has been said in passing that =to marks topics or partitions the sentences it occurs in into a ‘theme’ and a ‘rheme’ or ‘topic’ and ‘comment’ (Kidwai 2000, Butt 2014, Montaut 2015). As a particle that is said to signal the extent of backgrounded material in a clause, =to is not expected to attach to focused constituents (6a) or to be preceded by them (6b). What is surprising is that =to can be hosted by constituents that contain new information. But this is only if the relevant constituent does not carry new-information focus, as in the second example in 6c. In work on focus and focus-sensitivity, (new-information) focus marking on a constituent in a declarative utterance is standardly understood to provide cues to determining the CQ in the discourse context. I have already distinguished between explicitly raised questions and speaker-salient questions and have shown that the use of =to is appropriate in contexts where the two are understood to diverge. The observations about =to in 6 above allow us to specify a tendential condition on the relation between the =to hosts and speaker-salient issues.

(7) Salience-based constraint on =to: In a declarative utterance, =to tends to not be hosted by a constituent that completely resolves the speaker-salient issue.

The constraint in 7 can help make sense of the diverse implications arising from the use of =to-containing declaratives that we have already observed (albeit in a scattered way) in the data above. Let us consider in turn each configuration that 7 licenses.

(i) =to host contributes new information; =to host receives CT marking.

  Implication: The prejacent is a partial answer to CQ^E_c.

  Examples: the partial answers in 2d and 4a–4b [End Page 723]

(ii) =to host contributes new information; =to host is deaccented.

  Implication 1: The prejacent is a complete answer to CQ^E_c.

  Implication 2: CQ^E_c is inappropriate because it is resolvable from information at c.⁸

  Examples: the answer in 5

(iii) =to host contributes given information; =to host receives CT marking.

  Implication 1: The prejacent is a complete answer to CQ^E_c .

  Implication 2: A speaker-salient issue is implicitly raised and remains partially unresolved.

  Examples: the answer in 3d

(iv) =to host contributes given information; =to host is deaccented.

  Implication 1: The prejacent is a complete answer to CQ^E_c .

  Implication 2: CQ^E_c has an obvious resolution or is already resolved in c.

  Examples: the answers in 2c and 3c

Note that in none of the four configurations permitted by 7 is the prejacent clause understood to completely resolve a speaker-salient question distinct from the explicitly raised question. The constraint in 7 thus restrains the complex relationship that may hold between explicitly raised questions, their partial or complete answers, and the (possibly diverging) speaker-salient questions that can be raised or resolved in discourse contexts. Crucially, 7 allows for the decoupling of new information provided in a declarative utterance from the question of whether the new information resolves what the speaker takes to be the salient question. The relevant effects are tabulated in Table 1. The columns specify whether the =to host is deaccented or receives contrastive-topic marking in the declarative utterance it occurs in. The rows specify whether the =to host contributes new information or whether it denotes content that is previously mentioned/given in the context.

Click for larger view
View full resolution

Table 1.

Properties of =to-hosts and their discourse effects.

There are two subgeneralizations to be noted from Table 1. First, the presence of contrastive-topic marking on a =to host always signals an unresolved question in the discourse context, whether it is the explicitly raised question or a speaker-salient question. This suggests that =to by itself does not conventionally signal the presence of an unresolved issue in the context, but rather cooccurs quasi-categorically with such a signal (contrastive-topic marking) elsewhere in the clause. Second, when =to occurs with [End Page 724] deaccented hosts, it conveys that the prejacent, in some way, makes a weak contribution. It is understood to resolve a question that either has an obvious resolution or is already resolved in the context, or it is being used to comment on the inappropriateness of the explicitly raised question given common-ground content.

2.2. The generalization

The overall generalization so far is that =to signals that the question resolved by the prejacent is not a strong one. The question resolved by the prejacent may lack strength relative to the explicitly raised question or the speaker-salient question in virtue of being a subquestion of these. Alternately, it may lack strength in the absolute because it does not shrink the context set in a meaningful way from the speaker’s perspective. At the core of the proposal laid out here is the idea that enclitic =to uniformly signals that its prejacent resolves a weak question given the context. This is substantively different from saying that =to conveys that the prejacent offers a weak or a partial answer to the CQ and that a stronger (and complete) true answer is possible. Such a function has been claimed for a downtoner like at least by, for instance, Beaver and Clark (2008:97–98) and Zeevat (2019). The proposal defended here is that =to comments, not on the strength of the prejacent among alternative answers to a contextually fixed CQ, but on the strength of the question that the prejacent addresses, relative to the speaker’s information state, prior discourse moves, and assumptions about the common ground. In effect, =to is a generalized downtoning particle that comments on the strength of explicit and implicit questions and thereby, only indirectly, comments on the strength of the prejacent. The analysis in §3 formally implements this intuition.

3.1. The relevant components of context

(i) Common ground: Assume that each context c is associated with a body of information info_c characterizing the joint, mutually agreed-upon public commitments of all interlocutors at c. Info_c can be construed as a set of propositions or the set of worlds yielded by their intersection (the context set). Discourse moves undertaken by interlocutors are geared toward increasing the set of their joint commitments, which amounts to a progressive reduction of the context set. A declarative sentence S uttered at c embodies a proposal to update info_c by adding the propositional content of S to info_c (or by intersecting ⟦S^c⟧ with info_c). An interrogative clause Q embodies a proposal to steer interlocutors toward choosing a proposition to update info_c from among a set of alternative propositions in the answer space of the question denoted by Q.⁹ A discourse move involving an interrogative at c thus makes public the asker’s communicative goals and preferences regarding how info_c should evolve through the immediately following discourse.

While much of the literature assumes info_c to be identical across all interlocutors, given that interlocutors only presume to know what is shared between them, unfolding discourse often reveals divergences between assumptions about info_c. Accommodation of presuppositions is required precisely in contexts where there are such divergences in the common ground across interlocutors before accommodation. Moreover, there may be situations in which some information that is publicly accessible in the discourse context [End Page 725] (e.g. a goat enters the room) is not added simultaneously to each interlocutors’ public discourse commitments. In such a situation, content may be assumed to be in the shared common ground by one interlocutor without that actually being the case. In order to capture these potential discrepancies between what is in fact common ground between interlocutors and what is assumed to be the common ground by a given interlocutor, we distinguish between info^S_c (the common ground according to the speaker’s perspective) and info^A_c (the common ground according to the addressee’s perspective).¹⁰ We also identify a distinguished set of propositions that collects the publicly accessible sensory evidence at any context c – info_pubc. Info_pubc will include propositions such as the goat entering the room or the siren of a passing ambulance. The actual common ground will be (info^S_c ∩ info^A_c) ∪ info_pubc, represented as info_c.

(ii) Current question: Each context c also provides a question/issue CQ_c that at least one of the interlocutors takes to need resolution in order to update info_c. The prosodic structure in a declarative contribution is both constrained by and constrains the CQ_c, by principles that require focus alternatives to be evoked by some part of that contribution. Adapting the model developed in Roberts 2012 [1996], Beaver and Clark (2008:48–51) introduce two core principles that regulate the felicitous flow of discourse in terms of the contextually given CQ.

(8)

a. Discourse principle: Utterances should be maximally relevant to the CQ.

b. Focus principle: Some part of a declarative utterance should evoke a set of alternatives containing all of the Rooth-Hamblin alternatives of the CQ.

It is well recognized that the question-answer sequence based picture that is presented in formal models of discourse is greatly limited in terms of connecting to the complex structuring of dialogue in real interhuman interactions. Declarative clauses may not obviously answer explicitly raised questions, and few declarative utterances occur as responses to utterances of interrogative clauses. To approximate realistic discourse, any model must allow for unuttered but inferred-and-accommodated, implicit questions. Such questions can be made salient in the context (be construed as the CQ_c) by aspects of the prior utterance or by the utterance of declaratives that convey postfacto that a particular question was salient to the speaker at the immediately preceding context. Answers regularly allow interlocutors to accommodate the CQ_c. As Beaver and Clark (2008:51) note, ‘it is only after such accommodation that the Discourse Principle and Focus Principle are satisfied’.

The fact that an implicit question can be inferred on the basis of a contributed answer and accommodated in discourse is represented here by distinguishing between CQ^E_c (the question raised by an explicit move in the discourse: the CQ according to the questioner’s perspective) and CQ^S_c (the CQ according to the speaker’s perspective). In many contexts, it will be the case that CQ^S_c = CQ^E_c, and CQ_c will suffice to represent this mutually agreed-upon question.

(iii) Strategy of inquiry: Each context c also provides a strategy of inquiry SI_c, which at least one of the interlocutors considers optimal for satisfying joint discourse goals at c. Roberts’s (2012 [1996]) model takes discourse to be a strategic investigation [End Page 726] into the way the world is. The primary goal of interlocutors in any discourse is to determine the ways that things are, but the exchange of information that leads toward this goal must be structured and parceled into manageable portions. A strategy of inquiry is a structured sequence of questions designed to systematically refine the context set by breaking up larger conversational goals into subgoals. The idea is that once the interlocutors have mutually accepted a question, they may successively raise and resolve several smaller, more specific questions in order to resolve this larger, logically stronger question. Strategies of inquiry are thus sets of questions, partially ordered by entailment. The entailment relation between questions is defined as in Groenendijk & Stokhof 1984:16: one interrogative q₁ entails another q₂ iff every proposition that (completely) answers q₁ (completely) answers q₂ as well. In what has now become standard terminology, Roberts terms q₁ the superquestion and any q₂ it entails a subquestion. For example, a question like Who is at home? entails Are Niśa and Anu at home?, as well as Is Niśa at home?. Intuitively, an entailed question (a subquestion) is weaker (has less informative answers) than an entailing question. Roberts already suggests that a strategy of inquiry constituted by a partially ordered set of questions is not signaled in discourse solely via a sequence of explicitly raised questions. Just like implicit questions, strategies of inquiry can be inferred and accommodated as being salient at a context. We therefore allow for the possibility of diverging perspectives on the contextually salient strategy of inquiry by distinguishing between SI^S_c (the strategy of inquiry according to the speaker’s perspective) and SI^A_c (the strategy of inquiry according to the addressee’s perspective).¹¹ Again, in many contexts, it will be the case that SI^S_c = SI^A_c, and SI_c will suffice to represent this mutually agreed-upon strategy of inquiry.

3.2. The formal setup

Assume a set of worlds W, a set of propositions Prop, and a set of questions Ques.

(9)

a. Prop ⊆ ℘(W)

b. Ques ⊆ ℘(Prop), such that ∀q ∈ Ques:

i. ∀p, pʹ ∈ q : p ∩ pʹ = ∅ if p ≠ pʹ and

ii. ∪{p | p ∈ q} = W

(10) A context is a tuple 〈info_c, CQ_c, SI_c〉, such that:

a. info_c ⊆ W

b. CQ_c ∈ Ques

c. SI_c is a pair 〈Q_SIc, ≤_SIc〉 such that:

i. Q_SIc ⊆ Ques

ii. ≤_SIc is a partial strength ordering on Q_SIc

iii. CQ_c ∈ Q_SIc

iv. ∀q, qʹ ∈ Q_SIc : q ≤_SIc qʹ iff ∀p : p ∈ q → ∃pʹ ∈ qʹ : p ⊆ pʹ

v. ∃q ∈ Q_SIc : ∀qʹ ∈ Q_SIc : q ≠ qʹ → q <_SIc qʹ

According to 10c, a strategy of inquiry is a set of contextually relevant questions Q_SIc with a strength ordering ≤_SIc, such that Q_SIc includes CQ_c (10c.iii), <_SIc is an entailmentbased ranking over questions (10c.iv), and Q_SIc includes some question that is stronger than every other question in Q_SIc (10c.v).¹² Once a hierarchical structure of questions [End Page 727] such as a strategy of inquiry is included in the representation of the context, it becomes possible to characterize in a straightforward way the contribution of devices that are sensitive to strength/weakness among such questions.

3.3. Lexical entry for =to

The generalization in §2.2 was that the presence of =to signals that the question resolved by its prejacent is weak. This can be expressed straightforwardly using the lexical entry in 11.

(11) ⟦=to⟧^c = λpλw : weak_c(CQ_c). p(w)

As expected, =to combines with a proposition but adds nothing to its truth-conditional content. Its not-at-issue component simply specifies that the CQ_c (which the prejacent addresses) is weak—which means that it is informationally weak relative to the context. Of course, the usefulness of this lexical entry depends crucially on what it means to be a weak question relative to a context. We define this in 12.

(12) ∀c, q : weak_c(q) ↔

a. ∃qʹ ∈ Q_SI^A_c : qʹ <_SI^A_c q or

b. ∃qʹ ∈ Q_SI^S_c : qʹ <_SI^S_c q or

c. ∃p ∈ q : info^S_c ⊆ p

According to 12, a question q can be considered to be weak at a context c iff one of three conditions holds: q is not the strongest question in the strategy of inquiry assumed by the addressee (12a), q is not the strongest question in the strategy of inquiry assumed by the speaker (12b), or one of the answers to q is entailed by the common ground on the speaker’s perspective (12c). This last possibility amounts to saying that q is weaker than even the least inquisitive or trivial question at info^S_c.¹³ The fourth logical possibility, that of a question being weak at a context because it is entailed by the common ground of the context according to the addressee, cannot arise in discourse, since the addressee is expected to not ask superfluous questions (Groenendijk 1999).

Before moving on, let me (again) point out that this analysis takes =to to comment on the weakness of the CQ_c, not the weakness of the prejacent that addresses the CQ_c. An answer is weak relative to a mutually agreed-upon question because it does not fully resolve the question at the context. Whether a question is weak depends on the speaker’s perspective on the broader context. In general, questions raised at a context are expected to be nonsuperfluous and aimed at resolving issues that are not already resolved in info_c and, if possible, resolving the strongest issues that may be reasonably resolved at that context (an instance of Grice’s quantity maxim, as Groenendijk (1999) notes). Now, if the speaker offers a complete answer to an explicit question from the addressee and signals that the addressee’s question is weak using =to, this amounts to signaling that the question has flouted quantity on the speaker’s perspective. In other words, a complete answer to an explicitly raised question can still be accompanied by a signal that the question is weak. In these uses, =to may imply that the raised question is already resolved in info^S_c, that it is obviously resolvable at c, or that it is not the strongest question in SI^S_c . This is crucially where particles commenting on the weakness of answers (like at least) differ from particles commenting on the weakness of questions (like =to). The former make reference only to stronger answers to a contextually fixed question that the speaker is not in a position to provide. The latter make reference to stronger questions that could be addressed at the same context. Of course, signals that comment on question weakness can also be used while providing partial answers to explicit questions, signaling that the prejacent answers a question weaker than the explicitly [End Page 728] raised question corresponding to SI^A_c. Only in this subset of uses does an expression like =to behave similarly to an expression like at least.

Moreover, note that by positing that =to signals its prejacent to answer a weak question, we derive the empirical generalization captured in 7—the constraint that =to cannot be hosted by a constituent that completely resolves the speaker-salient issue. A speaker that uses the prejacent to address a contextually salient question/strategy sincerely must be providing the strongest resolution they can to it, aiming to enhance the information states of all interlocutors. If they address a question that is salient for their addressee (CQ^A_c), they might comment on the weakness of this question. But to introduce a question themselves (CQ^S_c) in order to resolve it, while simultaneously commenting on its weakness, is not a strategy that coheres with Gricean quantity. The observation in 7 is therefore best understood as an interaction between =to’s conventionalized meaning and pragmatic considerations about licit discourse contributions.

3.4. Types of question weakness

There are logically three classes of contextual conditions where a speaker may seek to signal that the question resolved by the contribution they make is not strong. Given that every such class of cases will be such that the speaker and the addressee are not perfectly coordinated on some contextual component, we specify (non)coordination for each component. In each case in 13 below, CQ^S_c represents the question the speaker answers through their utterance of the prejacent.

(13)

The reader will notice that the classes of contextual conditions outlined in 13 correspond to the effects described in Table 1. However, there are two connections that remain to be made in order for the analysis to yield a (more) complete understanding of those effects. First, the interaction between contrastive-topic marking and the location of =to in explicating the salient strategy of inquiry must be spelled out. Second, 13c does not distinguish between the subtly different effects associated with deaccented new and given =to hosts. Both effects are attributed to a mismatch between the interlocutors’ understanding of common-ground content. I make these connections in §3.5 and §3.6. [End Page 729]

3.5. Interaction between contrastive-topic marking and =to hosting

One observation from §2.1 is that the presence of contrastive-topic marking on a =to host always signals an unresolved question in the context. In the analysis above, this unresolved question is taken to be a stronger question belonging to a strategy of inquiry salient in the context. There are three possible loci of contrastive-topic marking in a declarative utterance that answers a constituent question—the constituent that contributes given information, the constituent that contributes new information, and the postverbal position in the verbal complex. Each of these, we see, induces different implications about the contextually salient strategy of inquiry. For example, to an explicitly raised question like ‘Where is Niśa?’, there are three possible responses that contain contrastive topics with accompanying =to marking.

(14)

The answer in example 14a implies that according to the speaker, the salient strategy of inquiry is one in which the strongest question is ‘Where is everyone?’, of which the explicitly raised question ‘Where is Niśa?’ is a subquestion. The answer in example 14b implies that the salient strategy of inquiry is one in which the strongest question is the explicitly raised ‘Where (exactly) is Niśa?’, of which ‘Is Niśa inside the house?’ is a subquestion. The answer in example 14c implies that the salient strategy of inquiry is one in which the strongest question is something like ‘Where is Niśa and is she available for a meeting?’, of which the explicitly raised ‘Where is Niśa?’ is a subquestion.

The felicitous use of contrastive-topic marking and =to placement is determined by the rules in 15. The rules in 15a and 15b are loosely based on relevant rules for well-formed discourse in Büring 2003 but adapted to the purpose at hand.

(15)

a. An utterance u containing a contrastive topic is a felicitous move at a context c iff u indicates a salient strategy of inquiry in c.

b. u indicates a salient strategy of inquiry in c iff the constituent n, marked as contrastive topic in u, evokes a nonsingleton set of questions Q, such that (i) for some q ∈ Q, u is a complete answer to q and (ii) there is a qʹ ∈ Q such that qʹ < q.

c. In any utterance u containing a constituent n marked as contrastive topic, if =to is present, =to must be hosted by n. If present, =to signals that u does not answer the strongest question in the set of questions Q evoked by n.¹⁴ [End Page 730]

Rule 15a states that contrastive-topic marking on some constituent in an utterance is felicitous only if there is a contextually salient strategy of inquiry that is indicated by the utterance. Rule 15b states that the contrastive-topic-marked constituent must evoke a set of questions such that the utterance is an answer to one of the questions, but not an answer to the strongest question. Finally, 15c requires that if =to is present in the utterance, it should be hosted by the contrastive topic.

3.6. Deaccented =to hosts: contrasting given and new

Another observation from §2.1, summarized in Table 1, is that the presence of =to on deaccented hosts gives rise to subtly distinct effects depending on whether the =to host contributes new or previously mentioned/given information. To understand the problem, assume a context in which Anu, Bilal, and Niśa are all in different parts of the house. Anu goes to Bilal and asks: ‘Where is Niśa?’. The two minimally different answers to the question in this context with accompanying =to marking are in 16.

(16)

In 16a, the =to host is a constituent contributing given information, while in 16b, the host is a constituent that contributes new information—the answer. The contextual profile for an answer like 16a was already described in 13c. Uttered at a context c, 16a implies that, according to the speaker, the explicitly raised question either has an obvious resolution or is already resolved in c. Example 16b not only implies that the speaker considers the explicitly raised question to be already resolved in c from their perspective, but also that they think there is no reason for there to be this discrepancy between the speaker and the addressee. The speaker takes Niśa’s whereabouts to be information that is publicly accessible in the broader context and therefore included in what is commonly known. Using 16b, the speaker conveys that the addressee already has all of the resources at their disposal to resolve the question they have raised; they have simply not made use of these resources. It is as if the speaker were saying: ‘Have you tried looking around and figuring out where Niśa is from all the evidence surrounding you?’. There is an additional constraint on this particular use of =to with any prejacent p. It seems that for =to to be felicitous in an answer, the questioner must be biased toward ¬p. The English question-answer sequences below with the =to host specified and the full Hindi answer given without glosses illustrate these uses and give English speakers a sense of =to’s effects.

(17)

[End Page 731]

In both cases in 17, the constituent question is uttered in a context c in which its resolution is directly entailed by information that is publicly accessible at c. The questioner, moreover, is biased against (the proposition denoted by) the prejacent being true. The speaker’s response, where =to is hosted by the answer constituent, conveys not only that the prejacent is true but also that its truth is determinable from publicly accessible information. Remember that the notation info_pubc was introduced in §3.1 in order to represent that subset of sensory information at a context that is assumed to be commonly known by virtue of it being publicly accessible. The actual common ground, info_c, is taken to be (info^S_c ∩ info^A_c) ∪ info_pubc. The contextual conditions associated with this class of cases can be represented as follows.

(18) The questioner is biased against the prejacent. Publicly accessible information in the context indicates that the prejacent is the answer to the explicitly raised question CQ^E_c . That is:

The main difference between the profile of the context in 18 and the context in 13c is that in 18, the speaker does not note any discrepancy between their conception of the common ground and the actual common ground. Rather, the use of =to serves to comment on the addressee’s version of the common ground and its failure to include relevant, publicly accessible information at the context. The explicitly raised question CQ^E_c comes out as weak in such a context because the issue it raises is already resolved, given the publicly accessible information at c (info_pubc).

Note that this analysis does not explicitly say anything about why the questioner must be taken to be biased against the prejacent in the context in order for =to to be felicitous. However, that is the empirical observation. Consider two minimal variants of the contexts in 17a and 17b in 19a and 19b. In 19a, as soon as there is no bias on Anu’s part against the umbrella being described as yellow, the use of =to with the deaccented new-information-contributing host becomes infelicitous.¹⁵ [End Page 732]

(19)

a. [Context: Anu and Bilal had ordered an umbrella online and the color of the umbrella that arrives is hard to clearly name. She asks Bilal for help in identifying the correct label:]

  Q: What color would you call this? (I am finding it hard to name.)

  A: #It is yellow=to.   Hindi: #ye pīlī=to hɛ.

  Contextual conditions: The actual umbrella (and the sensory input it provides) is publicly accessible. The addressee is not biased against This umbrella is yellow being true.

  b. [Context: Anu likes the biryani that Bilal makes and doesn’t eat it if it is made by anyone else. She enters the kitchen and the smell of biryani wafts up to her nose. She asks Niśa:]

  Q: Who made the biryani?

  A: #Bilal=to made it.   Hindi: #bilal=ne=to bənaji.

  Contextual conditions: The actual biryani (and the sensory input it provides) is publicly accessible. The addressee is not biased against Bilal made the biryani being true.

At this point, I leave this additional felicity condition on =to’s use with deaccented new-information-contributing hosts as an unsolved problem that may or may not require additional assumptions.

3.7. Summary

This section sketched out an analysis that attempts to capture the intuition that =to signals that its prejacent does not resolve a strong question. Conceptually, the main innovation was to assume that components of the context such as the current question, the salient strategy of inquiry, and the common ground may not be perfectly coordinated for interlocutors in unfolding discourse. Thus, while it is possible (and indeed frequently the case) that speakers straightforwardly accept a raised question or an initiated strategy of inquiry and respond fully to it in a single turn, it is also possible that they break down the raised question into an articulated strategy of inquiry or introduce a new strategy of inquiry that subsumes the explicitly raised question. The lexical entry for =to proposed in 11 accounts for the fact that =to, in conjunction with contrastive marking on particular constituents, signals the presence of a broader strategy of inquiry, of which the question resolved by the prejacent is a proper subquestion. Moreover, the same lexical entry accounts for the fact that speakers may convey that an explicitly raised question is weak relative to the common ground if the question has an obvious resolution in or is already resolved in the common ground. Further, as discussed in §3.6, the effect of =to attaching to hosts contributing new information is strictly stronger than the effect of =to attaching to hosts contributing given information. When =to attaches to new hosts, it conveys that the explicitly raised question is resolvable from publicly accessible information that the addressee should have added to their version of the common ground.

A range of empirical facts about =to’s distribution can be made sense of once we assume that it uniformly signals the weakness of the question its prejacent resolves. I turn to documenting and accounting for some of these facts in the domain of declarative clauses in §4. To the best of my knowledge, these patterns have not yet been systematically documented. Then, in §5, I provide a preliminary account of the use of =to in interrogative and imperative clauses.

4.1. Responses to assertions

All examples in §2 and §3 involved the use of =to in the answer part of an explicit question-answer sequence. However, =to may also occur [End Page 733] in responses to assertions. Assertions are standardly understood as proposals to update the common ground in a way compatible with their content (Farkas & Bruce 2010, Farkas & Roelofsen 2017). In the discourse model presented in Farkas & Bruce 2010, an assertion has the effect of adding the denoted proposition to the conversational table for consideration for addition to the common ground. An assertion thus raises an issue by ‘placing an item on the Table and it directs the conversation towards a unique resolution of that issue, namely, confirmation of the assertion’ (Farkas & Bruce 2010: 92). If the assertion receives confirmation by interlocutors, the proposition is added to the common ground, otherwise not. For our purposes, any sequence of conversational turns in which =to is used in a move asserting p that responds to a prior move asserting q, the explicitly raised question CQ^E_c will always be the same: whether it is appropriate to add q to info_c.

Consider a context where Anu asserts 20a denoting proposition g, thus raising the issue of whether g can be added to the common ground. Let this be CQ^E_c in the context. Bilal can respond to Anu’s assertion using =to, as in 20b, 20c, or 20d.

(20)

In 20b, Bilal conveys that g cannot be added to the common ground because it is inconsistent with it. CQ^E_c (whether g should be added to the common ground) is already resolved in the negative in info^S_c (Bilal’s version of the common ground). CQ^E_c thus lacks strength in the absolute because its resolution fails to meaningfully shrink the context set from Bilal’s perspective. Using 20c, Bilal conveys that he cannot take on g as a public discourse commitment but does not rule out (modulo further evidence) adding g to his public discourse commitments. The prejacent here resolves the question of whether Bilal can take g on as a public discourse commitment—a subquestion of CQ^E_c .¹⁶ The question resolved by the prejacent lacks strength relative to CQ^E_c because it resolves CQ^E_c only partially. By using 20d, Bilal conveys that CQ^E_c does not raise a strong issue in the context. Given that Niśa is always at home, it is no surprise that she is at home at the time of c. In other words, CQ^E_c has an obvious resolution in c; g can easily be added to the common ground.

4.2

Overanswering. =to is perfectly felicitous in cases of overanswering, where a speaker answers a question with a response that is finer-grained than the alternatives in the denotation of that question.¹⁷ This is illustrated here with polar interrogatives and [End Page 734] responses to them.¹⁸ To a question like 21a, Bilal may respond with 21b, 21c, or 21d, all of which completely answer the question expressed by the polar interrogative and also provide further information. Notice that in each case, =to attaches to a deaccented constituent that contributes given information—the third-person pronoun vo in 21b–c and the definite description in 21d.

(21)

Attaching =to to the predicate constituents in 21b or 21c—whether deaccented or with contrastive-topic marking—is ruled out. So is attaching it to Mīra in 21d. In each unacceptable case, the relevant constituent contributes new information. As discussed in §3.6, =to can be hosted by a deaccented constituent contributing new information only if the speaker takes CQ^E_c to be resolvable from publicly accessible information. The infelicity of =to when attached to the predicate constituents is thus accounted for naturally given the discourse context.

How can an overanswering response contain a particle conveying that the question answered by the prejacent is weak? Here is a way to think about this. At any context, an overanswering response means that there is divergence between CQ^E_c and the question the prejacent addresses (CQ^S_c). In some contexts, both interlocutors may be aware that CQ^E_c is part of a strategy of inquiry SI^A_c initiated by the addressee. For instance, Bilal may be aware that 21a is being asked in order to determine the relationship between him and Niśa or to determine the identity of his sister and take that to be SI^A_c . In such a context, Bilal does not unilaterally take CQ^S_c to be salient but rather recognizes it as the strongest question in Q_SI^A_c, the set of questions constituting the addressee’s strategy SI^A_c. Examples 21b and 21c indicate that Bilal takes SI^A_c to be aimed at identifying the relationship between him and Niśa, while 21d indicates that Bilal takes SI^A_c to be aimed at determining the identity of his sister. In all cases, =to’s presence in the over-answer conveys that the question resolved by the prejacent has an easy, obvious resolution in the context. The contextual profile is given in 22. Max(SI_c) denotes the strongest question in the set of questions Q_SIc in any salient strategy of inquiry. [End Page 735]

(22) The strongest question in Q_SI^A_c (of which CQ^E_c is a subquestion) is already resolved in the common ground from the perspective of the speaker. That is:

A context in which the speaker signals ‘recognition’ of the addressee’s strategy of inquiry and addresses its strongest question is a cooperative overanswering context. Overanswering can also occur uncooperatively if the speaker unilaterally assumes a strategy and proceeds to resolve its strongest question without making the prior move of raising the question. To illustrate, if Bilal uses 23a to ask Anu if he looks reasonably dressed for a routine Zoom meeting with his collaborator, Anu might respond with the over-answers in 23b or 23c—both containing =to.

(23)

We might assume that Bilal does not have a high threshold for what counts as reasonably dressed and is not deeply interested in resolving the question of the degree to which he is well dressed, as long as he doesn’t look terrible at the meeting. Anu’s evaluative responses in 23b and 23c signal that she assumes a broader strategy of inquiry SI^S_c whose strongest question Max(SI^S_c) is ‘How well dressed is Bilal relative to the contextual standard for well-dressedness?’. The explicitly raised question CQ^E_c is a subquestion of Max(SI^S_c).²⁰ The associated contextual profile is in 24.

(24) The strongest question in Q_SI^S_c (of which CQ^E_c is a subquestion) is already resolved in the common ground from the perspective of the speaker. That is:

[End Page 736]

4.3

Underanswering. =to is also felicitous in cases of underanswering in polar interrogatives. In these cases, the speaker may either provide some evidence in support of one of the answers without providing definitive resolution or convey ignorance regarding the truth value of the embedded sentence radical. =to is totally at home in such responses, which fail to resolve CQ^E_c . These cases are somewhat parallel to the cases in 2, where partial resolution of an explicitly raised issue can be signaled by =to attached to a contrastively marked new-information-carrying constituent, as in 2d. But the cases diverge in that the responses in underanswering cases, although informative in the sense of refining the context set, do not offer a partial answer to CQ^E_c . For example, if someone wants to know the relationship between Niśa and Bilal, and uses the interrogative in 25a to pose the question, an ignorant Anu might respond with 25b or 25c.²¹ Neither response constitutes a partial answer to the CQ^E_c in the sense of answering a subquestion in the strategy defined by CQ^E_c, but both do signal a related but distinct strategy on the part of the speaker that is relevant to answering the CQ^E_c .

To explicate, we can assume uncontroversially that a felicity condition for a discourse move involving any interrogative at a context is that the questioner takes their addressee’s information state to entail a true answer to the question posed by the interrogative. Then any context in which a polar question ?p is put up for resolution is also a context in which questions such as ‘Who knows whether p is the case?’ or ‘What evidence is known to be available for/against p?’ are assumed to have been resolved appropriately. In a sense ?p presupposes these latter questions.²² Now, if these questions are not already resolved appropriately at c, the speaker may undertake a strategy of inquiry, SI^S_c, to resolve these questions. In 25b, the strongest question in SI^S_c is ‘What is the evidence for the highlighted proposition in CQ^E_c “Niśa is Bilal’s sister”?’. The prejacent resolves a subquestion of this question: ‘Do Niśa and Bilal live in the same house?’. In 25c, the strongest question in SI^S_c is ‘Who knows the answer to CQ^E_c ?’. The prejacent resolves a subquestion of this question: ‘Does Anu know the answer to CQ^E_c?’. In both responses, the CQ^E_c remains unresolved, and the implication is that Anu can provide no stronger response than the prejacent.

(25)

In either case, the question resolved by the prejacent is weak because it is not the strongest question in Q_SI^S_c, which is one of the conditions for weakness as defined in 12b. The contextual profile for underanswering cases is in 26. [End Page 737]

(26) The prejacent does not address the strongest question in Q_SI^S_c. SI^S_c is a strategy to answer a question presupposed by the explicitly raised question CQ^E_c . The prejacent provides information about the speaker’s information state with regard to resolving CQ^E_c . That is:

4.4. =to in private mental-state questions

A straightforward prediction of the analysis proposed here is that =to (with deaccented hosts) should be unacceptable in responses to interrogatives that raise questions about the speaker’s private preferences or thoughts in the utterance context. This is because, as established in §2.1, when =to attaches to deaccented hosts contributing new or given information, the implication is that the question answered by the prejacent is easily resolvable or already resolved in the common ground on the speaker’s perspective. It is difficult to see how a speaker could consistently maintain that some proposition is part of their private mental state at the utterance context while at the same time being part of their conception of the common ground, predicting that =to is infelicitous in such contexts. This prediction is borne out, as exemplified in contexts such as in 27.

(27)

4.5. =to in information-acquisition reports

Another discourse effect related to contrastive uses of =to is its use in declaratives that report the recent acquisition of some piece of knowledge. In these cases, =to-containing declaratives respond, not to an explicitly raised question in the context, but rather to information obtained directly from the larger discourse context. In examples like 28a and 28b, the presence of =to adds counter-expectational flavor; it signals that the recently acquired information expressed by the prejacent contrasts with the speaker’s prior expectations.

(28)

[End Page 738]

These uses are straightforwardly assimilable under the broader analysis of =to as commenting on the weakness of the question resolved by the prejacent. In 28a, the speaker-salient question CQ^S_c is: ‘What is the current state of this blanket?’. In 28b, the speaker-salient question CQ^S_c is: ‘What are Bilal’s current food preferences?’. Both questions are weak relative to the context they are uttered in because their answers are entailed by the publicly accessible information at the context. The mirativity effect observed in such contexts is entirely due to the fact that the prejacent content is asserted despite being publicly accessible. Any assertion made at a context is felicitous if the proposition asserted is not entailed by the common ground. If a speaker asserts a proposition that is entailed by publicly accessible information at a context c, it can only be felicitous if the proposition is not part of the preutterance common ground according to the speaker (notated (info^S_c−1)). The proposition, while publicly accessible, is new for the speaker. The relevant contextual profile is in 29.

(29) The prejacent answers an implicit speaker-salient question CQ^S_c . The prejacent contributes publicly accessible information at c. That is:

4.6. Metalinguistic commentary on discourse moves

Yet another place =to occurs is in metalinguistic commentary on the discourse move itself. Suppose Bilal asks Anu whether he should give up his studies to start a job that would help support his family. Let us say that in context 1, Anu understands Bilal’s dilemma and appreciates that the resolution of the issue raised by Bilal’s question is not obvious, given the facts that need to be factored in. Anu can now respond with 30a. The use of =to here signals the obviousness of the goodness/uptake-worthiness of CQ^S_c . In contrast, in context 2, Anu has a definitive opinion: she thinks it makes absolutely no sense for Bilal to give up his education, since continuing it would ensure a much better job in a year’s time. A natural response (assuming Anu and Bilal are close) might be 30b. Anu, through her negative evaluation, rejects Bilal’s question as unworthy of deliberation.

(30)

The responses in 30a and 30b reveal the intermediate steps between the raising of a question at a context and its resolution by an interlocutor. In an articulated discourse model such as Farkas & Bruce 2010, a question, as any discourse move, is placed on the conversational table and requires uptake or acceptance from all interlocutors before they can commit to resolving the question. One might say then that at any context c, for any question CQ^E_c that is placed on the conversational table, there is raised an auxiliary metalinguistic question q regarding the goodness/uptake-worthiness/relevance of CQ^E_c at the juncture of discourse it occurs in. The prejacents in 30a and 30b offer explicit answers to the question ‘Is Bilal’s question an uptake-worthy question?’, with =to conveying, in each case, that the question has an obvious resolution (positive and negative, respectively) in c.

5.1. =to in imperatives

There are at least three facts to be accounted for with respect to =to in imperatives: (i) =to must be hosted by a constituent that carries contrastive-topic marking; (ii) =to always has the discourse function of a downtoner, translatable as at least; (iii) =to implies that the speaker believes it to be likely that expressing a preference for a proposition stronger than the prejacent may interfere with the preferences of some other agent. At an intuitive level, a speaker using =to in an imperative indicates that they have stronger preferences that they are not publicly committing to in the context given the possibility of conflict with other agents’ preferences or circumstances.²³

The phenomenon can be illustrated with the examples in 31. Suppose Niśa is visiting Anu’s family. It is mango season, and Anu’s mother has bought a big basket of mangoes. In 31a, =to is hosted by the recipient constituent, which bears contrastive-topic marking. In 31b, =to is hosted by the contrastively marked theme, and the verb is the host in 31c. In each case, =to signals that Anu’s mother takes the prejacent to express an effective preference that is weaker than expected.

(31)

[End Page 740]

Condoravdi and Lauer (2012) propose the following semantics for the imperative operator imp, where PEP_w(A, p) stands for ‘The agent A is publicly committed at w to act as though p is a maximal element of A’s effective preference structure’.²⁴

(32) ⟦imp⟧^c = λp[λw [PEP_w(S, p)]]

The imperative is understood to be an operator that applies to a proposition p and returns the proposition that is true at a world w iff the speaker is publicly committed at w to act as though p is a maximal element of their effective preference structure. In directive uses, the content of the imperative is about an addressee action. For instance, in 33a (and its Hindi counterpart), imp applies to the proposition p, which is true at exactly those worlds w in which the addressee gives Niśa the mango. The output is the proposition in 33b, abbreviated as m.

(33)

In accounting for =to in imperatives, we must take it to scope over imp, yielding the logical form as in 34a and the meaning in 34b.

(34)

a. =to(imp(A-give-N-the-mango))

b. ⟦=to(imp(A-give-N-the-mango))⟧^c

  = λpλw : weak_c(CQ_c). [p(w)](λwʹ. m(wʹ))

  = λw : weak_c(CQ_c). [m(w)]

[End Page 741]

The idea is that an imperative utterance answers the question ‘What are the speaker’s effective preferences at context w?’.²⁵ As discussed in §3.5, an utterance containing a contrastive topic (obligatory in =to-containing imperatives) must indicate a salient strategy of inquiry determined by the contrastive topic. =to conventionally signals that its prejacent (the imperative proposition) answers a weak question relative to the context. So, a =to-containing imperative, with a contrastively marked =to-host, always signals that the prejacent does not answer the strongest question in a strategy of inquiry determined by the contrastive-topic constituent given the rules in 15. At any context, the question answered by the imperative pertains to the speaker’s effective preferences—the realistic, consistent, action-guiding preferences a speaker has at that context. By using =to in an imperative, the speaker signals that this question is weak—there is a salient strategy of inquiry, viz. one whose strongest question is ‘What are the speaker’s preferences at context w?’, and the question ‘What are the speaker’s effective preferences at context w?’ is a strict subquestion of that question. The effect is that the speaker conveys that their broader preferences are strictly stronger than the effective preference they commit to using the imperative.

Example 31a implies that according to Anu’s mother, the strongest question in the salient strategy is ‘What are Anu’s mother’s preferences regarding mango giving in the context?’. The question answered by the imperative is ‘What are Anu’s mother’s effective preferences regarding mango giving in the context?’. Example 31b implies that the strongest question in the salient strategy is ‘What are Anu’s mother’s preferences regarding how Niśa, a guest, should be treated in the context?’. The imperative answers the question only with respect to her effective preferences. Example 31c implies that the strongest question in the salient strategy is ‘What are Anu’s mother’s preferences regarding Niśa’s mango-consumption experience?’. The imperative answers the question only with respect to her effective preferences.

If this is correct, then we can take =to in imperatives to allow speakers to signal that their real preference is strictly stronger than what they commit to as their effective preference given other information in the context (including conflict with addressee preferences). A clear prediction of this way of treating the interaction between =to and imperative meaning is that =to should be unacceptable in uses of imperatives for giving permission or making concessions, such as those seen in 35.

(35)

  a. OK, go out and play.   (Condoravdi & Lauer 2012:39)

  b. OK, go to Paris since you want it so much!   (Condoravdi & Lauer 2012:43)

  c. Sure, come see me at half-past three if you want.

In these uses, the speaker communicates that they recognize an addressee desire and have no objection to the realization of that desire. According to Condoravdi and Lauer, permissions arise when the following preconditions are in place: (i) the addressee has a preference for the content p, and (ii) there is some q that is incompatible with p such that the speaker is committed to an effective preference for q. The imperative utterance indicates a change in the speaker’s effective preferences, such that p is now ranked above q. In concession uses, the speaker retains their previous (noneffective) preference against the realization of the content of the imperative, even though their effective preferences have changed, possibly under pressure from the addressee (Condoravdi & Lauer 2012:50). In either context type, if =to were used, it would convey that the speaker has preferences that are strictly stronger than the effective preference they have [End Page 742] expressed using the imperative. However, any speaker, in making concessions and giving permissions, has had to alter their preutterance effective preference structure in order to accommodate the addressee’s preferences despite their own preference for something incompatible with this. It follows that the speaker’s preferences do not include anything strictly stronger than the effective preference they commit themselves to using the imperative. =to is correctly ruled out in these uses of imperatives.

5.2. =to in interrogative clauses

There are two facts to be accounted for with respect to =to in interrogatives: (i) in polar interrogatives, =to functions as a downtoner and is uniformly translatable as at least in positive interrogatives; (ii) =to cannot appear in wh-interrogatives. We look at both observations in turn, but I offer a possible solution only to the first one.