Farkas and Bruce (2010): On Reacting to Assertions and Polar Questions #

This file formalizes [FB10]'s account of why reactions to assertions and to polar questions overlap only in part. The context structure is Commitment.Table: the participants' discourse commitments, the common ground, and the Table, whose items project the common grounds that would settle them. A default assertion commits its author, places the declarative on the Table and projects confirmation alone (9); a default polar question places the interrogative on the Table and projects both resolutions (12). So an assertion leaves the common ground as it was, unlike [Sta78]'s update (Table.commonGround_assert, assert_not_narrowing); the two moves differ in whether the author is committed and in whether the projected set is inquisitive (projectedSet_assert, projectedSet_polarQuestion), and agree in deciding the sentence radical in every projected common ground (Table.mem_of_mem_projectedSet_assert, Table.mem_or_compl_mem_of_mem_projectedSet_polarQuestion). Confirmation (16) followed by the common-ground increase M' (17) settles an assertion (shared_assert_confirm, isStable_settle_assert, mem_commonGround_settle); a total denial (22) leaves nothing consistent projected and the conversation in crisis (21, Table.inCrisis_assert_assert_compl), from which agreeing to disagree (23) recovers with the commitments intact (not_inCrisis_agreeToDisagree, discourseCommitments_agreeToDisagree). A reverse answer to a polar question (28) is no crisis (not_inCrisis_polarQuestion_assert_compl), a confirming answer (24) projects the single common ground with the answer (projectedSet_polarQuestion_assert), and the questioner's confirmation of it settles the question (isStable_settle_polarQuestion).

§4.3 and §5 classify responding assertions by two features: the relative polarity [same] or [reverse] of the response to the sentence radical on the Table (31) and the absolute polarity [+] or [−] of the asserted sentence. Sentence is a radical with its polarity and Confirming/Reversing the move types (26) and (29) on sentences; the same [reverse] response is a denial after an assertion and a reverse answer after a question (inCrisis_reversing_assert, not_inCrisis_reversing_polarQuestion); on the responses of Discourse.Response the relative polarity is the quotient of the two polarities (Reversing.iff_relative_eq_negative).

§5.2 says what the polarity particles mark, as sets of responses: the English yes [same] or [+] and no [reverse] or [−], their double duty (english); the Romanian da [+], nu [−] and ba [reverse] except in a [reverse, −] answer to a question (romanian); and French si and German doch the marked combination [reverse, +] (reversePositive). Both English particles can confirm a negative question, while in Romanian only nu does (english_confirmsNegativeQuestion, romanian_confirmsNegativeQuestion_iff). rows are the responding assertions of (4), (5) and (35)–(50), whose particles are the fragments' polarity particles; every acceptable response is one its English and Romanian particles mark (english_mem, romanian_mem), si and doch occur in [reverse, +] responses only (si_doch_mem_reversePositive), and ba is possible in every denial (ba_denial) and required to be absent from a [reverse, −] answer (ba_not_reverse_answer_neg).

Implementation notes #

The Table records the issue a sentence raises, not its syntactic object, so a negative polar question raises the same issue as its positive counterpart (15), Question.polar_compl, and the relative polarity of a response is computed from the sentences of the exchange rather than read off the Table.
M' is Table.settle, which strips the shared proposition from the individual commitment lists as (17) prescribes; the projected set is derived from the Table rather than stored, as the paper notes it can be.
Rows record the polarities of the initiating and responding sentences as a Discourse.Response; since a responding assertion shares its radical with the initiating sentence, [same] is agreement of polarity and [reverse] its reversal.
The paper discusses only si of the French particles and only doch of the German ones, so what they mark is a named set rather than an interpretation of the whole fragment.

TODO #

(5) accepts a bare nu denial of a positive assertion that (42) stars, so §5.2's claim that ba is required in a [reverse, −] denial is left as the weaker ba_denial.
Example and section numbers follow the 2009 preprint; UNVERIFIED against the numbering of the journal version.

Farkas and Bruce (2010): On Reacting to Assertions and Polar Questions #

Implementation notes #

TODO #

References #

Default assertions and default polar questions #

Confirmation, denial, and agreeing to disagree #

Reacting to a polar question #

Responding assertions and polarity features #

Polarity particles #

The examples #