The two privative morphosyntactic features of @cite{kratzer-selkirk-2020}.
[FoC] and [G] are genuinely syntactic features: crosslinguistically they trigger displacement, agreement, and ellipsis (§2). They happen to be spelled out prosodically in Standard American and British English, but this is not their defining property.
- FoC : ISFeature
FoCus: introduces alternatives, signals contrast. Resembles [wh] — comes with obligatory ~ operator.
- G : ISFeature
Givenness: presupposes discourse salience, signals match. Contributes meaning directly (no operator needed).
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- Semantics.Focus.KratzerSelkirk2020.instBEqISFeature.beq x✝ y✝ = (x✝.ctorIdx == y✝.ctorIdx)
Instances For
Newness is NOT a grammatical feature. New material is simply unmarked — no [FoC], no [G].
Equations
- Semantics.Focus.KratzerSelkirk2020.isNew hasFoC hasG = (!hasFoC && !hasG)
Instances For
§8 (45). The Contribution of [FoC] #
[FoC] does NOT change the O-value. Its A-value is the full domain D_τ (all possible entities of the relevant semantic type). This is standard Roothian focus semantics.
⟦[α]{FoC}⟧{O,C} = ⟦α⟧{O,C} ⟦[α]{FoC}⟧_{A,C} = D_τ
Apply [FoC] to a meaning: O-value unchanged, A-value becomes full domain. K&S (45): The A-value of [α]_{FoC} is D_τ.
Equations
- Semantics.Focus.KratzerSelkirk2020.applyFoC m domain = { oValue := m.oValue, aValue := domain }
Instances For
[FoC] preserves O-value. K&S (45) first clause.
§8 (46-47). The Contribution of [G] #
[G] introduces a Givenness requirement: the expression must match a salient discourse referent. Technically:
⟦[α]{G_a}⟧{O,C} is defined iff a is a discourse referent in C, and α is Given with respect to a. If defined, ⟦[α]{G_a}⟧{O,C} = ⟦α⟧{O,C} ⟦[α]{G_a}⟧{A,C} = ⟦α⟧{A,C}
[G] contributes purely use-conditional / expressive meaning (like discourse particles German "ja", "doch"). It places a condition on the discourse context, not on truth conditions.
An expression α is Given with respect to discourse referent a iff its A-value is {a} (a singleton containing just the referent).
K&S (46): α is Given w.r.t. a in C iff ⟦α⟧_{A,C} = {a}.
Intuitively: the alternatives set has collapsed to a single salient entity, meaning there's nothing to contrast — the content is already "in the air".
Equations
- Semantics.Focus.KratzerSelkirk2020.isGiven [a] referent = (a == referent)
- Semantics.Focus.KratzerSelkirk2020.isGiven aValue referent = false
Instances For
Apply [G] to a meaning: both values unchanged, but adds a definedness condition (the expression must be Given w.r.t. some discourse referent).
Unlike [FoC], [G] does NOT change the A-value. Its contribution is purely a presupposition on the discourse context.
Equations
Instances For
[G] preserves O-value. K&S (47): if defined, O-value unchanged.
[G] preserves A-value. K&S (47): A-value unchanged.
§8 (58). [FoC] and [G] are Mutually Exclusive #
A single constituent CANNOT bear both [FoC] and [G]. The proof follows from the A-value conditions:
- [FoC] requires A-value = D_τ (the full domain, maximally large)
- [G] requires A-value = {a} (a singleton) No semantic domain is both maximal and a singleton (assuming |D_τ| > 1).
[FoC] and [G] are mutually exclusive: no constituent can satisfy both the [FoC] A-value condition (full domain) and the [G] A-value condition (singleton) simultaneously, when the domain has more than one element.
K&S (58, first part): follows from the incompatibility of A-value conditions.
§8 (45, 47). Both Features are Use-Conditional #
Neither [FoC] nor [G] changes the truth-conditional (at-issue) content of the expression it attaches to. Both contribute use-conditional / expressive meaning.
This grounds K&S's features in Potts' two-dimensional semantics, already
formalized in Expressives/Basic.lean.
[FoC] is use-conditional: at-issue content is unchanged. Grounded in TwoDimProp from Expressives/Basic.lean.
Equations
- Semantics.Focus.KratzerSelkirk2020.focAsTwoDim atIssue contrastPresup = Semantics.Lexical.Expressives.TwoDimProp.withCI atIssue contrastPresup
Instances For
[G] is use-conditional: at-issue content is unchanged. [G] resembles discourse particles (German "ja", "doch") — it places a condition on context salience without affecting truth conditions.
Equations
- Semantics.Focus.KratzerSelkirk2020.gAsTwoDim atIssue givennessPresup = Semantics.Lexical.Expressives.TwoDimProp.withCI atIssue givennessPresup
Instances For
[FoC] does not change at-issue content (grounding theorem).
Both features project their use-conditional content through negation, just like conventional implicatures.
"It's not the case that [ELIZA]_{FoC} mailed the caramels" still contrasts Eliza with alternatives.
§8 (49). Contrast Representation #
An expression α represents a contrast with discourse referent a iff: (i) a ∈ ⟦α⟧{A,C} — the referent is among the alternatives (ii) a ≠ ⟦α⟧{O,C} — the referent differs from the actual value (iii) There is no FoC/G-variant β of α with ⟦β⟧{A,C} ⊂ ⟦α⟧{A,C} and a ∈ ⟦β⟧_{A,C} — no smaller alternatives set also captures a
Condition (iii) prevents over-FoCusing.
Conditions (i) and (ii) of Contrast (K&S 49). Condition (iii) — the minimality condition — is structural and requires checking FoC/G-variants, which we leave to the prosodic spellout layer.
- aValue : List α
The expression's A-value (alternatives)
- oValue : α
The expression's O-value (ordinary denotation)
- referent : α
The contrasting discourse referent
(i): referent is among the alternatives
(ii): referent differs from the O-value
Instances For
§8 (53-54). The ~ Operator #
[FoC]-marked constituents must be c-commanded by a operator.
The operator:
- Takes a set of discourse referents 𝔠 as its contextual variable
- Requires α to represent a contrast with each member of 𝔠
- Stops the propagation of alternatives (consumes them)
- Contributes expressive meaning: the contrast is signaled
Unlike Rooth's original (which allows questions as antecedents),
K&S's always signals contrast. Questions do NOT have a special
direct relation to FoCus.
The ~ operator (K&S version, allowing multiple antecedents).
K&S (54): ⟦~𝔠 α⟧{O,C} is defined iff 𝔠 is a set of discourse referents in C, and α represents a contrast with each member of 𝔠.
If defined, ⟦𝔠 α⟧{O,C} = ⟦α⟧_{O,C}
A-values: ⟦𝔠 α⟧{A,C} = {⟦α⟧_{O,C}} (singleton — alternatives consumed).
- meaning : Core.InformationStructure.AltMeaning α
The expression's meaning
- antecedents : List α
The contrasting discourse referent(s)
- antecedents_in_alts (a : α) : a ∈ self.antecedents → a ∈ self.meaning.aValue
Each antecedent is in the alternatives
Each antecedent differs from the O-value
Instances For
The ~ operator consumes alternatives: result A-value is singleton.
Instances For
~ preserves O-value.
§8 (56). The Semantics of only #
K&S's only directly takes a contextual variable 𝔠 (the contrast set), rather than accessing focus alternatives indirectly:
⟦only_𝔠⟧ = λp λw. ∀q. (q ∈ 𝔠 ∧ q(w)) → q = p
The contrast set 𝔠 is supplied by the operator that comes with [FoC].
Since stops alternative propagation, only associates with [FoC]
indirectly via a second occurrence of 𝔠.
Semantics of only with explicit contrast set (K&S 56). Takes a contrast set 𝔠 and a prejacent proposition p. True at w iff every true member of 𝔠 equals p.
Equations
Instances For
§8 (58). [G] Containing [FoC] Requires Alternatives Consumption #
A constituent α containing [FoC]-marked β can be [G]-marked only if α also contains an operator that consumes the alternatives generated by β.
Proof: For α to be [G], its A-value must be a singleton {a}. But [FoC] on β would make α's A-value non-singleton (alternatives propagate upward) UNLESS some operator inside α (like ~ or only) has consumed them.
This explains Second Occurrence Focus: in "the fáculty only quote [the faculty]_{FoC}", the second "the faculty" is [FoC]-marked but sits inside a [G]-marked VP. This is possible because only + ~ consume the alternatives before they reach the VP level.
§7. Prosodic Spellout #
In Standard American and British English, [FoC] and [G] are spelled out prosodically at the syntax-phonology interface (MSO → PI mapping).
The architecture has three levels:
- MSO: Morphosyntactic Output (syntactic structure with [FoC]/[G])
- PI: Phonological Input (prosodic constituency)
- PO: Phonological Output (tones, prominence)
Match constraints (MatchWord, MatchPhrase, MatchClause) generate prosodic constituency in PI from syntactic constituency in MSO. Then spellout constraints map [FoC] and [G] to prosodic properties.
Spellout of [FoC]: maps to head at a prosodic level. K&S (34, 43): [FoC] = {ω, φ, ι}-Level-Head.
A [FoC]-marked constituent in MSO is spelled out as a head at the corresponding prosodic level in PI. Being a head in a chain ending at ι means being the MOST PROMINENT constituent in the sentence.
- ω_level_head : FoCSpellout
[FoC] = ω-Level-Head: head of prosodic word
- φ_level_head : FoCSpellout
[FoC] = φ-Level-Head: head of phonological phrase
- ι_level_head : FoCSpellout
[FoC] = ι-Level-Head: head of intonational phrase (highest prominence)
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Spellout of [G]: removes φ constituency (dephrasing). K&S (38): [G] = No-φ.
A [G]-marked constituent in MSO corresponds to a prosodic constituent in PI that is NOT a φ and contains no φ. The phonological consequences:
- No obligatory H accent tone (which requires φ-head status)
- No L edge tone (which requires φ-final position)
This replaces the traditional "destressing" analysis with a structural one.
- no_phi : Bool
A [G]-marked constituent has no φ in PI
Instances For
K&S (41, 44): When [G] and [FoC] spellout conflict, [G] wins.
Ranking in Standard American and British English: [G]=No-φ >> MatchPhrase >> [FoC]=φ-Level-Head
This means: dephrasing a [G]-marked constituent takes priority over giving a [FoC]-marked constituent φ-level prominence.
Consequence: Second Occurrence Focus [FoC] inside [G] gets only ω-level head status, not φ-level. Hence reduced prosody for SOF.
- g_over_match : SpelloutRanking
[G]=No-φ outranks MatchPhrase
- match_over_foc_phi : SpelloutRanking
MatchPhrase outranks [FoC]=φ-Level-Head
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
The ranking is fixed for Standard American and British English.
Equations
- One or more equations did not get rendered due to their size.
Instances For
§4, §7.3. Second Occurrence Focus #
SOF is the strongest empirical argument for the two-feature system.
Example (@cite{beaver-2007}, K&S 42): "Both Sid and his accomplices should have been named in this morning's court session. But the defendant only named [Síd]_{FoC} in court today."
MSO: Even [the prosecutor]{FoC} [only named [Sid]{FoC} in court today]_{G}
The second "Sid" is [FoC]-marked (it associates with only) but sits inside a [G]-marked constituent. The ranking [G]=No-φ >> [FoC]=φ-Level-Head predicts: Sid gets ω-level head status but NOT φ-level prominence. Result: an H accent but no phrase-level pitch scaling — exactly what @cite{beaver-2007} @cite{selkirk-2008} found experimentally.
A Second Occurrence Focus datum: [FoC] inside [G].
- sentence : String
The full sentence
- sofWord : String
The SOF word
- consumingOperator : String
The operator that consumes SOF's alternatives
- hasHAccent : Bool
Whether H accent present (yes for SOF)
- hasPhiProminence : Bool
Whether φ-level prominence present (no for SOF)
- source : String
Source
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
@cite{beaver-2007} SOF example.
Equations
- One or more equations did not get rendered due to their size.
Instances For
@cite{katz-selkirk-2011} FoC-New vs New-FoC vs New-New triples. K&S (36): Phonetic evidence distinguishing [FoC] from newness.
- firstStatus : Core.InformationStructure.DiscourseStatus
First post-verbal phrase status
- secondStatus : Core.InformationStructure.DiscourseStatus
Second post-verbal phrase status
- pitchPattern : String
Description of the pitch pattern
- source : String
Source
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
- 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
§8 (61, 66). Pressure for [G]-Marking and [FoC]-Marking #
[G]-marking and [FoC]-marking are obligatory under certain discourse conditions in Standard American and British English.
(61) Pressure for [G]-Marking: [G]-mark a constituent if it is Given w.r.t. a salient discourse referent.
(66) Pressure for [FoC]-Marking: Represent non-trivial contrasts with salient discourse referents.
These are not semantic/syntactic constraints but PRAGMATIC pressures, possibly reducible to Maximize Presuppositions.
Bridge: K&S vs @cite{schwarzschild-1999} #
Schwarzschild's "A-Givenness" (within Rooth's Alternatives Semantics) falls out as a special case of K&S's [G]-feature.
A-Givenness: α is A-Given in C iff there is a salient discourse referent that is a member of ⟦α⟧_{A,C}.
K&S's Givenness (46): α is Given w.r.t. a iff ⟦α⟧_{A,C} = {a}.
K&S's condition is STRONGER (singleton vs membership). The old A-Givenness condition was too weak — Schwarzschild noted it was trivially satisfiable for universal quantifiers (every cat is a complainer → trivially A-Given).
Schwarzschild's A-Givenness: some referent is in the alternatives set.
Equations
- Semantics.Focus.KratzerSelkirk2020.isAGiven aValue referent = aValue.any fun (x : α) => x == referent
Instances For
K&S Givenness entails Schwarzschild A-Givenness. If the alternatives set is a singleton {a}, then certainly a ∈ alternatives.
The converse fails: A-Givenness does NOT entail K&S Givenness. A non-singleton alternatives set can satisfy A-Givenness but not Givenness.
This is the Schwarzschild overgeneration problem (K&S fn. 14): "Every cat is a complainer" is trivially A-Given because ∃P[every P is a complainer] is always true. K&S's singleton condition avoids this.