Mendes (2025): The Subordinate Future #
@cite{mendes-2025}
The Subordinate Future (SF) in Portuguese is a mood form (subjunctive with future morphology) that:
- Enables modal donkey anaphora — subjunctive binds situation variables across clause boundaries (§3.1)
- Weakens existential presuppositions of strong quantifiers in restrictors (§2.2)
- Has a future-oriented temporal interpretation that is parasitic on the modal anaphora — not stipulated by an independent temporal operator (§3.2, following @cite{crouch-1993} @cite{crouch-1994})
Organization #
- §1 The SF operator (§3.2):
subordinateFuture := dynSUBJ ∘ dynFUT, plus its conditional and relative-clause specializations. - §2 Temporal properties (§3.2): theorems showing the future shift is derived from the modal component, not stipulated.
- §3 Compositional CDRT derivations (§4.3.1): lexical entries and end-to-end derivation of "Se Maria estiver em casa, ela vai atender".
- §4 Presupposition weakening (§2.2): SF in restrictors of strong quantifiers, formalized as modal displacement.
The dynamic primitives are imported from co-located dynamic operator
files: dynSUBJ/dynIND from Theories/Semantics/Mood/Dynamic.lean
(siblings of static Mood.SUBJ/IND), dynFUT from
Theories/Semantics/Tense/Dynamic.lean, and SitContext/SVar from
Theories/Semantics/Dynamic/Core/{ContextFilter,DiscourseRef}.lean.
Subordinate Future (SF) analysis.
The SF in Portuguese conditionals: "Se Maria estiver em casa, ela vai atender." "If Maria be.SF at home, she will answer."
Structure (@cite{mendes-2025} §3.2):
- SF = SUBJ^{s₁}_{s₀} + FUT
- SUBJ introduces s₁ ∈ hist(s₀)
- FUT constrains τ(s₁) > τ(s₀)
- Main clause is anchored to τ(s₁)
This is the compositional derivation: ⟦SF⟧ = ⟦SUBJ⟧ ∘ ⟦FUT⟧
Equations
- Mendes2025.subordinateFuture history newSitVar refSitVar c = Semantics.Tense.dynFUT newSitVar refSitVar (Semantics.Mood.dynSUBJ history newSitVar c)
Instances For
Conditional with SF antecedent (dynamic version).
"Se Maria estiver em casa, ela vai atender."
- Antecedent: SF introduces s₁ ∈ hist(s₀) with τ(s₁) > τ(s₀)
- Antecedent predicate evaluated at s₁
- Consequent: temporally anchored to s₁ (future relative to s₀)
Equations
- Mendes2025.conditionalWithSF history antecedentVar speechVar antecedent consequent c = consequent (antecedent (Mendes2025.subordinateFuture history antecedentVar speechVar c))
Instances For
Relative clause with SF in restrictor.
"Cada menino [que estiver acordado] vai receber um biscoito." "Every boy [who is.SF awake] will get a cookie."
Structure:
- SF in relative clause introduces situation s₁ ∈ hist(s₀)
- Restrictor (boy ∧ awake) evaluated at s₁
- Nuclear scope (get cookie) evaluated with temporal anchor from s₁
Equations
- Mendes2025.relativeClauseSF history rcVar speechVar c = Mendes2025.subordinateFuture history rcVar speechVar c
Instances For
Strong quantifier with SF restrictor.
"Todo livro [que Maria ler.SF] será interessante" "Every book [that Maria reads.SF] will be interesting"
The SF in the restrictor:
- Introduces s₁ ∈ hist(s₀) for the relative clause
- Quantification over books is relativized to s₁
- Nuclear scope inherits temporal anchor from s₁
Equations
- Mendes2025.everyWithSFRestrictor history rcVar speechVar restrictor nuclear c = nuclear (restrictor (Mendes2025.subordinateFuture history rcVar speechVar c))
Instances For
SF introduces a future situation.
The subordinate future always introduces a situation with time ≥ current.
Temporal shift is parasitic on modal donkey anaphora.
@cite{mendes-2025} §3.2: the future-oriented interpretation of SF is not due to an independent temporal operator. Instead, it follows from:
- SUBJ introduces s₁ ∈ hist(s₀) - modal component
- hist(s₀) includes situations with τ(s₁) ≥ τ(s₀) - temporal consequence
- Main clause is evaluated at τ(s₁) via modal anaphora - binding
The temporal shift is derived from the modal semantics, not stipulated.
Modal donkey anaphora explains why subjunctive mood enables future reference in subordinate clauses.
SF in restrictor enables future reference for strong quantifiers.
With SF in the relative clause, "every" can quantify over future entities.
Restrictor and nuclear must be context filters (IsContextFilter).
Linguistically, predicates filter contexts without modifying assignments.
Maria — proper name.
⟦Maria⟧ = λP.P(maria)
Equations
- Mendes2025.lexMaria maria P c = P maria c
Instances For
estar em casa — "be at home".
⟦estar em casa⟧ = λxλsλc. [| at-home(x)(s)]; c
Equations
- Mendes2025.lexAtHome atHomeRel x sitVar c = {gs : Core.Assignment (Core.WorldTimeIndex W Time) × Core.WorldTimeIndex W Time | gs ∈ c ∧ atHomeRel x (gs.1 sitVar)}
Instances For
atender — "answer (the door)".
⟦atender⟧ = λxλsλc. [| answer(x)(s)]; c
Equations
- Mendes2025.lexAnswer answerRel x sitVar c = {gs : Core.Assignment (Core.WorldTimeIndex W Time) × Core.WorldTimeIndex W Time | gs ∈ c ∧ answerRel x (gs.1 sitVar)}
Instances For
SF (Subordinate Future).
⟦SF⟧ = SUBJ ∘ FUT
Equations
Instances For
ela — "she" (pronoun bound to Maria).
⟦ela⟧ = λP.P(maria)
Equations
- Mendes2025.lexShe maria P c = P maria c
Instances For
vai — future auxiliary "will".
⟦vai⟧ = λVPλsλc. VP(s)(c) — transparent; future comes from SF
via modal anaphora.
Equations
- Mendes2025.lexWill VP sitVar c = VP sitVar c
Instances For
Sequential context update: feeds antecedent output into consequent.
Note: the paper's ⟦if⟧ is dynamic implication (a test: [| P ⇒ Q]),
but the derivation theorems here require extracting properties from the
output context, which a test semantics cannot provide (tests preserve input
unchanged). Sequential composition is appropriate because the universal
force comes from SUBJ's quantification over historical alternatives, not
from the conditional operator itself.
Equations
- Mendes2025.seqUpdate antecedent consequent c = consequent (antecedent c)
Instances For
Antecedent derivation:
⟦Maria estiver em casa⟧ = SUBJ^{s₁}_{s₀}[FUT; [| at-home(maria)(s₁)]]
Introduces s₁ ∈ hist(s₀), constrains τ(s₁) > τ(s₀), asserts Maria at home.
Equations
- Mendes2025.deriveAntecedent history maria atHomeRel sfVar speechVar c = Mendes2025.lexAtHome atHomeRel maria sfVar (Mendes2025.lexSF history sfVar speechVar c)
Instances For
Consequent derivation:
⟦ela vai atender⟧ = [| answer(maria)(s₁)] — s₁ retrieved via IND.
Equations
- Mendes2025.deriveConsequent maria answerRel sfVar c = Mendes2025.lexAnswer answerRel maria sfVar (Semantics.Mood.dynIND sfVar c)
Instances For
Full sentence derivation:
⟦Se Maria estiver em casa, ela vai atender⟧
Equations
- One or more equations did not get rendered due to their size.
Instances For
The situation introduced by SF is in the historical alternatives.
The derivation enforces future ordering: τ(s₁) > τ(s₀).
If Maria is at home at s₁, she answers at s₁.
Counterfactual conditional (for comparison). "Se Maria estivesse em casa, ela atenderia." Uses SUBJ without FUT — allows past/present alternatives.
Equations
- One or more equations did not get rendered due to their size.
Instances For
SF constrains to future; counterfactual allows past/present.
Key data (Portuguese) #
With indicative, strong quantifiers presuppose existence: (17) #Cada/todo livro que a Maria ler será interessante. "Every book that Maria reads.IND will be interesting" → Presupposes Maria will read books (fails if uncertain)
With SF, the presupposition is weakened: (18) Cada/todo livro que a Maria ler será interessante. "Every book that Maria reads.SF will be interesting" → No existence presupposition (felicitous even if uncertain)
Existential presupposition: the restrictor is non-empty.
Equations
- Mendes2025.existentialPresup restrictor w = ∃ (x : E), restrictor x w
Instances For
Indicative restrictor: evaluates at the actual world. "Every book that Maria reads.IND..." → presupposes books exist that Maria reads.
Equations
- Mendes2025.indicativeRestrictor restrictor s x = restrictor x s
Instances For
SF restrictor: quantifies over historical alternatives. "Every book that Maria reads.SF..." → no categorical existence presupposition.
Equations
- Mendes2025.sfRestrictor history restrictor s₀ x = ∃ s₁ ∈ Core.Modality.HistoricalAlternatives.historicalBase history s₀, restrictor x s₁
Instances For
Indicative preserves existential presupposition.
SF weakens existential presupposition: even without actual existence at s₀, the SF restrictor can be satisfied in alternative situations.
SF makes strong quantifiers felicitous under uncertainty.
Relative clause with SF weakens strong quantifier presupposition. This is the formal version of the indicative-vs-SF contrast in restrictors.
Equations
- Mendes2025.relClauseSF history noun relClause s₀ x = ∃ s₁ ∈ Core.Modality.HistoricalAlternatives.historicalBase history s₀, noun x s₁ ∧ relClause x s₁
Instances For
Modal displacement: SF introduces quantification over situations, "displacing" the existential presupposition to be local within each situation.
Equations
- One or more equations did not get rendered due to their size.
Instances For
SF semantics is equivalent to modal displacement.
Modal displacement is weaker than global accommodation.
The central theoretical insight of @cite{mendes-2025} §3.1: SF enables modal donkey anaphora — subjunctive binds situation variables across clause boundaries, just like indefinites bind individual variables in classic donkey sentences.
Classic donkey anaphora: "If a farmer owns a donkey, he beats it."
- "a donkey" introduces individual dref x
- "it" retrieves x outside the syntactic scope of "a"
Modal donkey anaphora: "Se Maria estiver em casa, ela vai atender."
- SF introduces situation dref s₁
- Main clause retrieves s₁ for temporal anchoring
Correspondence with the dynamic primitives in
Theories/Semantics/Mood/Dynamic.lean:
- SUBJ introduces:
dynSUBJ history v = dynIntroduce (historicalBase history) v - IND retrieves:
dynIND v = dynRelationOn (·.2) (·.1 v) sameWorld
Cross-clausal situation binding: a situation introduced in one clause can be retrieved in another clause via modal donkey anaphora.
Example: "Se Maria estiver em casa, ela vai atender." ↑ SUBJ introduces s₁ ↑ IND retrieves s₁
Equations
- Mendes2025.crossClausalBinding history antecedentVar _consequentVar c = Semantics.Mood.dynIND antecedentVar (Semantics.Mood.dynSUBJ history antecedentVar c)
Instances For
Cross-clausal binding preserves world identity: when a situation is introduced in the antecedent and retrieved in the consequent, the two clauses are evaluated at the same world.
The SUBJ-IND anaphoric chain: SUBJ introduces s₁, the antecedent
predicate filters at s₁, IND retrieves s₁ (same-world check), and
the consequent inherits the temporal anchor from s₁.
Equations
- Mendes2025.subjIndChain history v antecedentPred consequentPred c = consequentPred (Semantics.Mood.dynIND v (antecedentPred (Semantics.Mood.dynSUBJ history v c)))
Instances For
The SUBJ-IND chain establishes modal donkey anaphora: the consequent is evaluated at a world that agrees with the bound situation's world.
Q must be a context filter — predicates filter contexts without
modifying assignments.
Unselective binding gives universal force. When SUBJ introduces a situation in a conditional antecedent, the conditional quantifies universally over situations satisfying the antecedent — the modal analog of donkey universals.
Pipeline characterization #
The full pipeline SUBJ → filter(P) → IND → filter(Q) on a singleton
context is equivalent to a static existential conjunction over
historical alternatives. The dynamic pipeline gives conjunction
(P ∧ Q), not implication (P → Q); sequential composition gives
the stronger conjunctive reading, while the static conditionalSF
uses implication.
The SUBJ-IND chain with predication filters on a singleton context
characterizes the static existential conjunction
∃ s₁ ∈ historicalBase(s₀), P(s₁) ∧ Q(s₁).
The dynamic pipeline entails the static conditional (conditionalSF).
Conjunction is stronger than implication: if the dynamic pipeline finds
an s₁ satisfying both P and Q, then P(s₁) → Q(s₁) holds trivially.
Bridge to Hofmann (2025) accessibility #
The same-world constraint enforced by dynIND (via the sameWorld
kernel in Mood/Basic.lean) parallels @cite{hofmann-2025}'s
veridicality-based accessibility for individual drefs:
- Situation level (this file, @cite{mendes-2025}):
dynINDretrievess₁vias₂.world = s₁.world. Governs cross-clausal situation binding (modal donkey anaphora). - Propositional dref level (
Phenomena/Anaphora/Studies/Hofmann2025.lean, @cite{hofmann-2025}): a dref is accessible iff it has a referent in all worlds of the local context, plus a discourse-consistency condition. Governs individual dref accessibility across negation, disjunction, and attitude contexts.
Both enforce the structural pattern that the retrieval context must be compatible with the introduction context. For situations, this is world identity; for individual drefs, this is the subset-plus- existence condition (Hofmann 2025 Definition 39).