Seeliger & Repp (2018): Biased Declarative Questions #
Biased declarative questions in Swedish and German: negation meets modal particles (väl and doch wohl).
Key Contributions #
- Distinguishes declarative questions (DQs) from rejecting questions (RQs) — both have declarative syntax but differ in bias profile.
- Applies [Sud13]'s two-dimensional bias scheme (evidential × epistemic) to classify four question types: PDQ, NDQ, PRQ, NRQ.
- Shows that the negation in negative RQs is non-propositional — it denotes the illocutionary modifier FALSUM, not propositional negation.
- Proposes REJECTQ as the illocutionary operator for RQs.
Bias Profiles ([Sud13]) #
Every declarative question type carries a bias profile: a pair of evidential and epistemic bias values. Evidential bias concerns contextual evidence; epistemic bias concerns the speaker's prior assumptions.
[Sud13] proposes three "plus" values for bias: [+positive], [+negative], [neutral]. Evidential bias can also take "minus" values ([-positive], [-negative]). [SR18] extend this by allowing minus values for epistemic bias as well, capturing that DQ speakers merely didn't assume the proposition (rather than assuming the opposite).
Four Question Types ([SR18], Table 1) #
| Type | Denotes | Evidential | Epistemic | Example |
|---|---|---|---|---|
| PDQ | p | +positive | -positive | Peter is coming? |
| NDQ | not-p | +negative | -negative | Peter isn't coming? |
| PRQ | p | +negative | +positive | Surely Peter is coming? |
| NRQ | not-p | +positive | +negative | Surely Peter isn't coming? |
DQs and RQs differ in that RQs are "more biased": the speaker had a specific prior assumption (epistemic bias is [+positive] or [+negative]), and the contextual evidence conflicts with that assumption.
REJECTQ Operator ([SR18] §6.2, eq. 40) #
REJECTQ takes a proposition q and an illocutionary modifier IM (FALSUM or VERUM):
⟦REJECTQ⟧ = λqλIM: [IM(¬q)]^evid & [IM(q)]^epist. {IM(q), ¬IM(q)}
The two presuppositions:
- Evidential: [IM(¬q)]^evid — IM-determined commitment to ¬q on evidential basis
- Epistemic: [IM(q)]^epist — IM-determined commitment to q on epistemic basis
The at-issue content is the question {IM(q), ¬IM(q)}. FALSUM = zero commitment (used in NRQs); VERUM = full commitment (used in PRQs).
Cross-Linguistic Findings #
German: RQs obligatorily contain doch wohl — the combination is non-compositional. doch wohl enters syntactic Agree with REJECTQ at ForceP. Without doch wohl, a German declarative cannot be a RQ.
Swedish: RQs are marked by fronted negation, the modal particle väl, or both. Unlike German, Swedish has multiple formal strategies for marking RQs. väl in positive declaratives creates PDQs; combined with negation, creates NRQs.
Experimental Evidence #
[SR18] §5.4: acceptability judgment study (24 native Swedish speakers) testing negative declaratives with fronted vs. low negation, with and without väl, in NRQ contexts. Main effect of MODAL PARTICLE (väl raises acceptability); interaction (effect only reliable with low negation). Supports fronted-negation + väl marking NRQs.
Related Work #
Semantics/Questions/Bias/Basic.lean— Romero's PQ bias framework (VERUM, FALSUM, OriginalBias, ContextualEvidence). The bridge maps Sudo's bias values to Romero's coarser three-valued scheme.Studies/RomeroHan2004.lean::Verum— detailed VERUM semantics with modal frames.Features/AnsweringSystem.lean— polar answer typology (Holmberg 2016).
Bias value for a single dimension (evidential or epistemic).
[Sud13]'s system distinguishes "plus" values (positive bias for p or not-p), "neutral" (no bias), and "minus" values (incompatibility with a given bias direction).
[Sud13] originally restricted minus values to evidential bias. [SR18] extend the system by allowing minus values for epistemic bias as well — needed to capture the DQ pattern where the speaker merely didn't assume the proposition ([-positive]) rather than actively assuming the opposite ([+negative]).
- plusPos : BiasValue
[+positive]: bias for p
- plusNeg : BiasValue
[+negative]: bias for not-p
- neutral : BiasValue
[neutral]: no bias
- minusPos : BiasValue
[-positive]: incompatible with bias for p
- minusNeg : BiasValue
[-negative]: incompatible with bias for not-p
Instances For
Equations
- SeeligerRepp2018.instDecidableEqBiasValue x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- SeeligerRepp2018.instReprBiasValue = { reprPrec := SeeligerRepp2018.instReprBiasValue.repr }
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
Whether a bias value is a "plus" value (active bias) or not. RQs require plus values in both dimensions; DQs have a minus value in the epistemic dimension.
Equations
Instances For
Whether a bias value targets p (positive polarity) or not-p.
Equations
- SeeligerRepp2018.BiasValue.plusPos.targetsPositive = some true
- SeeligerRepp2018.BiasValue.plusNeg.targetsPositive = some false
- SeeligerRepp2018.BiasValue.minusPos.targetsPositive = some true
- SeeligerRepp2018.BiasValue.minusNeg.targetsPositive = some false
- SeeligerRepp2018.BiasValue.neutral.targetsPositive = none
Instances For
The four types of questions with declarative syntax.
- PDQ : DeclQuestionType
Positive declarative question: "Peter is coming?"
- NDQ : DeclQuestionType
Negative declarative question: "Peter isn't coming?"
- PRQ : DeclQuestionType
Positive rejecting question: "Surely Peter is coming?"
- NRQ : DeclQuestionType
Negative rejecting question: "Surely Peter isn't coming?"
Instances For
Equations
- SeeligerRepp2018.instDecidableEqDeclQuestionType x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
Equations
- One or more equations did not get rendered due to their size.
Instances For
The two classes: declarative questions vs. rejecting questions.
DQs are less biased (speaker is "prejudiced" but not committed); RQs are more biased (speaker had a specific prior assumption that conflicts with contextual evidence).
- declarative : DeclQuestionClass
Simple declarative question — speaker seeks confirmation
- rejecting : DeclQuestionClass
Rejecting question — speaker rejects what s/he sees
Instances For
Equations
- SeeligerRepp2018.instDecidableEqDeclQuestionClass x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
Classify each type into its class.
Equations
- SeeligerRepp2018.DeclQuestionType.PDQ.questionClass = SeeligerRepp2018.DeclQuestionClass.declarative
- SeeligerRepp2018.DeclQuestionType.NDQ.questionClass = SeeligerRepp2018.DeclQuestionClass.declarative
- SeeligerRepp2018.DeclQuestionType.PRQ.questionClass = SeeligerRepp2018.DeclQuestionClass.rejecting
- SeeligerRepp2018.DeclQuestionType.NRQ.questionClass = SeeligerRepp2018.DeclQuestionClass.rejecting
Instances For
What a declarative of this type denotes (positive = p, negative = not-p).
Equations
- SeeligerRepp2018.DeclQuestionType.PDQ.declPolarity = Features.Polarity.positive
- SeeligerRepp2018.DeclQuestionType.NDQ.declPolarity = Features.Polarity.negative
- SeeligerRepp2018.DeclQuestionType.PRQ.declPolarity = Features.Polarity.positive
- SeeligerRepp2018.DeclQuestionType.NRQ.declPolarity = Features.Polarity.negative
Instances For
The illocutionary modifier (IM) that occupies the ForceP specifier position in rejecting questions.
[SR18] §6.2: FALSUM and VERUM are epistemic speech-act level operators. Their structural position is:
[ForceP FALSUM/VERUM [Force' REJECTQ [TP ...]]]
FALSUM signals zero commitment to the proposition (the speaker is essentially not committed to adding q to the CommonGround). Used in NRQs. VERUM signals full commitment (the speaker is sure q should be in the CommonGround). Used in PRQs. In PRQs, Swedish visst/nog or an evidential version of VERUM may appear.
These correspond to the operators defined in
Semantics.Questions.Bias (verum, mkFalsum).
- falsum : IllocutionaryModifier
FALSUM: zero commitment to q (non-propositional negation). [Rep13]: speaker is not committed to q at issue.
- verum : IllocutionaryModifier
VERUM: full commitment to q (q should be in the CommonGround). [RH04]: for-sure-CommonGround that q should be added.
Instances For
Equations
- SeeligerRepp2018.instDecidableEqIllocutionaryModifier x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
REJECTQ — the illocutionary operator for rejecting questions.
[SR18] eq. 40: ⟦REJECTQ⟧ = λqλIM: [IM(¬q)]^evid & [IM(q)]^epist. {IM(q), ¬IM(q)}
REJECTQ takes a proposition q and an illocutionary modifier IM. In German, IM is determined by syntactic Agree with doch wohl. In Swedish, it is determined by the presence of FALSUM (fronted negation) or VERUM (modal particles like visst/nog).
- modifier : IllocutionaryModifier
The illocutionary modifier (FALSUM or VERUM)
- evidentialPresupposition : Bool
Evidential presupposition: [IM(¬q)]^evid — on an evidential basis, the IM-determined degree of commitment to ¬q holds.
- epistemicPresupposition : Bool
Epistemic presupposition: [IM(q)]^epist — on an epistemic basis, the IM-determined degree of commitment to q holds.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- SeeligerRepp2018.instReprRejectQ = { reprPrec := SeeligerRepp2018.instReprRejectQ.repr }
Equations
- One or more equations did not get rendered due to their size.
Instances For
Construct a well-formed REJECTQ with both presuppositions satisfied.
Equations
- SeeligerRepp2018.mkRejectQ im = { modifier := im, evidentialPresupposition := true, epistemicPresupposition := true }
Instances For
Derive the evidential bias from the IM choice in REJECTQ.
The evidential presupposition is [IM(¬q)]^evid:
- IM = VERUM: evidence strongly supports ¬q → evidential [+negative] (VERUM(¬q) = full commitment to ¬q on evidential basis)
- IM = FALSUM: evidence yields zero commitment to ¬q → evidential [+positive] (FALSUM(¬q) = not committed to ¬q → by contrast, evidence for q)
Equations
Instances For
Derive the epistemic bias from the IM choice in REJECTQ.
The epistemic presupposition is [IM(q)]^epist:
- IM = VERUM: speaker is epistemically sure q should be in CommonGround → epistemic [+positive] (speaker assumed q)
- IM = FALSUM: speaker has zero epistemic commitment to q → epistemic [+negative] (by pragmatic strengthening in the RQ context, zero commitment to q implies belief in ¬q)
Equations
Instances For
The bias profile derived from REJECTQ's presuppositions.
Equations
- SeeligerRepp2018.rejectQBiasProfile im = { evidential := im.evidentialBias, epistemic := im.epistemicBias }
Instances For
DQ bias profiles are observational — DQs are not marked by REJECTQ, so their profiles don't derive from the IM parameter. Instead, DQs require contextual evidence matching the declarative polarity, and the speaker must not have already assumed the declarative's content.
[SR18]: "DQs pattern with each other" (p. 136).
Equations
- SeeligerRepp2018.dqBiasProfile Features.Polarity.positive = { evidential := SeeligerRepp2018.BiasValue.plusPos, epistemic := SeeligerRepp2018.BiasValue.minusPos }
- SeeligerRepp2018.dqBiasProfile Features.Polarity.negative = { evidential := SeeligerRepp2018.BiasValue.plusNeg, epistemic := SeeligerRepp2018.BiasValue.minusNeg }
Instances For
The bias profile for each declarative question type.
DQ profiles are from dqBiasProfile (evidence-based);
RQ profiles are from rejectQBiasProfile (REJECTQ-derived).
Equations
- SeeligerRepp2018.DeclQuestionType.PDQ.biasProfile = SeeligerRepp2018.dqBiasProfile Features.Polarity.positive
- SeeligerRepp2018.DeclQuestionType.NDQ.biasProfile = SeeligerRepp2018.dqBiasProfile Features.Polarity.negative
- SeeligerRepp2018.DeclQuestionType.PRQ.biasProfile = SeeligerRepp2018.rejectQBiasProfile SeeligerRepp2018.IllocutionaryModifier.verum
- SeeligerRepp2018.DeclQuestionType.NRQ.biasProfile = SeeligerRepp2018.rejectQBiasProfile SeeligerRepp2018.IllocutionaryModifier.falsum
Instances For
DQs and RQs are distinct classes.
RQ bias profiles are fully derived from the IM choice —
they come out of rejectQBiasProfile, not independent stipulation.
RQs have conflicting biases: evidential and epistemic target opposite polarities. This is what makes them "rejecting" — the speaker sees evidence against what s/he believed.
Follows from the REJECTQ structure: IM(¬q) and IM(q) target opposite polarities by construction.
DQs have compatible biases: evidential matches declarative polarity, epistemic is merely "minus" (speaker didn't assume, rather than actively assuming the opposite).
RQ epistemic bias is always "plus" (active commitment); DQ epistemic bias is always "minus" (non-commitment). This is the defining difference between the two classes.
NRQ is a subset of PDQ contexts: both require +positive evidential bias, but NRQs additionally require the speaker to have assumed ¬p ([SR18] p. 138: "a NRQ is used in a subset of the situations where a PDQ can be used").
PRQ is a subset of NDQ contexts: both require +negative evidential bias, but PRQs additionally require the speaker to have assumed p.
VERUM and FALSUM produce opposite evidential biases — they interpret the evidence in opposite directions.
VERUM and FALSUM produce opposite epistemic biases.
Map Sudo's evidential bias values to Romero's contextual evidence.
The [+positive]/[+negative] values map directly. [neutral] maps to neutral. The "minus" values have no direct Romero counterpart — they encode incompatibility constraints rather than positive evidence.
Equations
- SeeligerRepp2018.evidentialToContextualEvidence SeeligerRepp2018.BiasValue.plusPos = some Semantics.Questions.Bias.ContextualEvidence.forP
- SeeligerRepp2018.evidentialToContextualEvidence SeeligerRepp2018.BiasValue.plusNeg = some Semantics.Questions.Bias.ContextualEvidence.againstP
- SeeligerRepp2018.evidentialToContextualEvidence SeeligerRepp2018.BiasValue.neutral = some Semantics.Questions.Bias.ContextualEvidence.neutral
- SeeligerRepp2018.evidentialToContextualEvidence SeeligerRepp2018.BiasValue.minusPos = none
- SeeligerRepp2018.evidentialToContextualEvidence SeeligerRepp2018.BiasValue.minusNeg = none
Instances For
Map Sudo's epistemic bias values to Romero's original speaker bias.
[+positive] maps to forP (speaker expected p). [+negative] maps to againstP. [neutral] maps to neutral. The "minus" values are not directly representable in Romero's three-valued system.
Equations
- SeeligerRepp2018.epistemicToOriginalBias SeeligerRepp2018.BiasValue.plusPos = some Semantics.Questions.Bias.OriginalBias.forP
- SeeligerRepp2018.epistemicToOriginalBias SeeligerRepp2018.BiasValue.plusNeg = some Semantics.Questions.Bias.OriginalBias.againstP
- SeeligerRepp2018.epistemicToOriginalBias SeeligerRepp2018.BiasValue.neutral = some Semantics.Questions.Bias.OriginalBias.neutral
- SeeligerRepp2018.epistemicToOriginalBias SeeligerRepp2018.BiasValue.minusPos = none
- SeeligerRepp2018.epistemicToOriginalBias SeeligerRepp2018.BiasValue.minusNeg = none
Instances For
NRQ evidential bias maps to Romero's "evidence for p" — the same contextual evidence configuration that licenses HiNQs.
NRQ epistemic bias maps to Romero's "original bias against p".
PRQ epistemic bias maps to Romero's "original bias for p".
All RQ bias values have Romero counterparts (they are all "plus" values). DQ epistemic "minus" values do not — this is precisely where Sudo's system extends Romero's.
Swedish väl is question-inducing — declaratives with väl are questions, not assertions ([SR18] §5.2). Derived from the fragment's distribution facet.
S&R's bias classification of väl (formerly fragment fields; a particle's bias requirement is the analysis, so it lives here): felicitous only in contexts with contextual evidence for the proposition, matching the evidential bias of PDQs and NRQs.
Equations
Instances For
S&R's classification, epistemic dimension: väl signals epistemic
uncertainty — the speaker suspects p but is not certain,
corresponding to the [-positive] epistemic bias of PDQs — so it
imposes no original-bias requirement (contrast dochWohlOriginalBias).
Equations
Instances For
S&R's bias classification of doch wohl: both dimensions active —
contextual evidence for p (prototypical NRQ reading) and prior
speaker bias against p — consistent with RQs having both evidential
and epistemic presuppositions in the REJECTQ definition (eq. 40).
Shares its evidential value with valContextualEvidence; the
epistemic dimension is where German is stricter than Swedish.
Equations
Instances For
doch wohl is not usable in assertions — it marks questions. Derived from the fragment's distribution facet.
The derived RQ property behind dochWohlOriginalBias: both DQ types
doch wohl can mark have active ("plus") epistemic bias — the
theory-level theorem rq_epistemic_is_plus in fragment-free form.
Both Swedish väl and German doch wohl require evidential bias
(valContextualEvidence = dochWohlContextualEvidence = some .forP),
reflecting the shared property that both languages require contextual
evidence for DQs/RQs. Where they differ is the epistemic dimension:
doch wohl marks RQs (prior commitment against p,
dochWohlOriginalBias), väl marks DQs (uncertainty, no
original-bias requirement, valOriginalBias) — German is stricter.
German denn differs from both: it imposes no bias requirement at
all; its felicity condition is the highlighting/precondition relation
(see Theiler2021).
doch wohl is a two-particle complex with conventionalized meaning. [SR18] §4.2: the combination does not receive a compositional interpretation. If it were compositional, doch wohl should have a reading combining conflict (doch) + reportativity (wohl), but this reading is unavailable in RQs.
The doch in doch wohl has a different meaning from polarity-
reversal doch (as in German.PolarityMarking.dochPreUtterance).
In RQs, doch has a "conflict" meaning — it signals surprise or
realization — rather than the "reminding" function of assertive doch.
German doch is formally ambiguous between two distinct roles:
- Polarity-reversal doch: pre-utterance correction particle that contradicts a negative antecedent ([Hol16])
- RQ-marking doch: part of the doch wohl complex that enters Agree with REJECTQ at ForceP ([SR18])
The two share the surface form "doch" but differ in:
- Syntactic position: polarity doch is pre-utterance; RQ doch wohl is in the middle field
- Function: polarity doch reverses polarity; RQ doch signals conflict between evidence and prior belief
- Obligatoriness: polarity doch is optional (Verum focus available); RQ doch wohl is obligatory for German RQs
RQ bias values all have Romero counterparts (they are all "plus" values); DQ epistemic "minus" values do not — this is precisely where [Sud13]'s system (as extended by [SR18]) goes beyond [Rom24].
Instances For
Both modal-particle complexes that mark RQs/DQs in this study sit at PerspP — the layer for biased, matrix-only question particles.