Exclusion features and X/O-marking strategies #
@cite{iatridou-2000} @cite{condoravdi-2002} @cite{deal-2020} @cite{anderson-1951} @cite{schlenker-2004} @cite{von-fintel-iatridou-2023}
Framework primitives for the Exclusion Feature analysis of past morphology (@cite{iatridou-2000}) and the X-marking / O-marking typology of counterfactuality (@cite{von-fintel-iatridou-2023}).
This file provides the framework primitives — types, predicates, and
bridge theorems usable by any study of counterfactuals or
crosslinguistic conditional morphology. Paper-specific typologies and
empirical claims live in the corresponding Studies/ files:
Phenomena/Conditionals/Studies/Iatridou2000.lean—IatridouPredType/CounterfactualType/classifyCounterfactual(FLV/PresCF/PastCF taxonomy)Phenomena/Conditionals/Studies/Mizuno2024.lean— Anderson-conditional empirical observations (English X-marking, Japanese O-marking, FLV correlation)
Core claims #
Past morphology encodes exclusion:
- Temporal: T(topic) ≠ T(speaker)
- Modal: w(topic) ≠ w(speaker)
This maps onto the ContextTower's origin / innermost distinction:
ExclF dim tower holds iff the relevant coordinate of tower.innermost
differs from that of tower.origin.
The X-marking / O-marking distinction (per @cite{von-fintel-iatridou-2023}) is the typological label for whether a language uses counterfactual morphology (X-marking, produces ExclF) or some other strategy (O-marking, e.g., Japanese Historical Present) to grammatically distinguish live from non-live possibilities.
Bicontextual semantics #
@cite{schlenker-2004} distinguishes the Context of Thought θ
(speech-act context, anchors temporal indexicals) from the Context of
Utterance v (can be shifted by HP). The ContextTower implements
this distinction: tower.origin = θ, tower.innermost = v. The
origin_stable theorem captures Schlenker's claim that θ-anchored
indexicals are unaffected by HP/CF shifts.
The two dimensions along which past morphology can exclude.
@cite{iatridou-2000}'s key insight: past morphology has a single semantic contribution (exclusion) that applies to different dimensions. The temporal/modal ambiguity of "past" is not lexical ambiguity — it is the same feature targeting different coordinates.
- temporal : ExclDimension
Temporal: T(topic) ≠ T(speaker)
- modal : ExclDimension
Modal: w(topic) ≠ w(speaker)
Instances For
Equations
- Semantics.Modality.Exclusion.instDecidableEqExclDimension 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
The Exclusion Feature: past morphology signals that the topic coordinate differs from the speaker coordinate on the given dimension.
This is a predicate over context towers: ExclF dim tower holds iff
the relevant coordinate of the innermost context (topic) differs from
the origin context (speaker).
Equations
- Semantics.Modality.Exclusion.ExclF Semantics.Modality.Exclusion.ExclDimension.temporal tower = (tower.innermost.time ≠ tower.origin.time)
- Semantics.Modality.Exclusion.ExclF Semantics.Modality.Exclusion.ExclDimension.modal tower = (tower.innermost.world ≠ tower.origin.world)
Instances For
ExclF temporal unfolds to time inequality.
ExclF modal unfolds to world inequality.
Map ExclDimension to @cite{deal-2020}'s PastMorphologyUse.
This connects @cite{iatridou-2000}'s exclusion analysis to @cite{deal-2020}'s tense typology: temporal exclusion corresponds to temporal tense, modal exclusion corresponds to counterfactual tense.
Equations
Instances For
subjShift changes world → produces modal ExclF.
When a subjunctive clause introduces a new world that differs from the origin, the resulting tower has modal ExclF. This is the tower-level formalization of @cite{iatridou-2000}'s claim that counterfactual morphology signals world exclusion.
temporalShift changes time → produces temporal ExclF.
When an embedding shifts the evaluation time away from the speech time, the resulting tower has temporal ExclF. This is ordinary temporal past.
Two shifts → both ExclFs (the PastCF configuration).
X-marking shift produces modal ExclF on RichContext towers.
xMarkingShift (from Core.Context.Rich) changes both world and
time. When the counterfactual world differs from the origin, the
resulting tower has modal ExclF.
Kratzer-level counterpart: in @cite{kratzer-2012}'s modal
semantics, the same X-marking operation corresponds to ∗-revision of
the modal base (Semantics.Modality.Kratzer.XMarking.IsStarRevision):
the expandedDomain parameter here maps to the widened accessible
world set ∩f'(w) ⊇ ∩f(w). See @cite{ferreira-2023} for the full 2×2
square of necessities generated by X-marking on both modal base (Xf)
and ordering source (Xg).
The two crosslinguistic strategies for grammatically distinguishing live from non-live possibilities (@cite{von-fintel-iatridou-2023}'s typological label).
Used by @cite{mizuno-2024} to characterize how different languages express Anderson conditionals: English uses X-marking (counterfactual morphology), Japanese uses O-marking (Non-Past + Historical Present).
- xMarking : MarkingStrategy
X-marking: counterfactual morphology expands the domain from D to D⁺. "Actually" in the consequent recovers the actual world via Kaplanian origin access. English: "If Jones had taken arsenic, he would have shown exactly the symptoms he is actually showing."
- oMarking : MarkingStrategy
O-marking: Non-Past morphology triggers a perspectival shift analogous to the Historical Present (@cite{schlenker-2004}). The backward time shift expands the domain under branching time, avoiding triviality without counterfactual morphology. The actual world is directly accessible (no world shift, so no need for "actually"). Japanese: "Jones-si-ga... nom-eba,... mise-ru (hazuda)."
Instances For
Equations
- Semantics.Modality.Exclusion.instDecidableEqMarkingStrategy 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
X-marking strategy uses counterfactual morphology; O-marking does
not. The single primitive property of marking strategies; all
others (producesExclF, requiresActuallyOperator) derive from
it.
Equations
Instances For
X-marking produces ExclF; O-marking does not.
X-marking is counterfactual morphology: subjShift changes the
evaluation world, creating modal ExclF. O-marking leaves the tower at
the root, so no ExclF arises. Definitionally equal to hasXMarking.
Equations
- s.producesExclF = s.hasXMarking
Instances For
X-marking requires "actually" to recover the actual world; O-marking
does not. When X-marking produces ExclF, the actual world is excluded
from the shifted evaluation; "actually" (Kaplanian origin access)
reaches back through the shift. With O-marking, the evaluation world
IS the actual world. Definitionally equal to hasXMarking.
Equations
Instances For
Map marking strategies to the ExclDimension they produce.
Equations
Instances For
Map marking strategies to @cite{deal-2020}'s PastMorphologyUse.
Equations
Instances For
X-marking produces modal ExclF: subjunctive shift changes the
world, creating world exclusion. Wraps subjShift_produces_modal_exclF.
"Actually" recovers the origin world even after a tower shift.
In an X-marked Anderson conditional, the CF morphology pushes the
tower (shifting the evaluation world). But "actually" — being a
Kaplanian indexical with depth = .origin — resolves to the
speech-act world regardless. Wraps opActually_shift_invariant.
O-marking has no modal ExclF: without CF morphology, the tower
stays at the root. Wraps root_no_modal_exclF.
Both X-marking and O-marking avoid the triviality problem by expanding the domain. General pattern: if a domain expansion produces a world where the consequent fails, the conditional is non-trivial.
- X-marking achieves this via
xMarkingShift(CF morphology → D⁺) - O-marking achieves this via
hpShift(HP → backward time → D⁺)
Wraps ConditionalShift.expansion_resolves_triviality.
End-to-end O-marking domain expansion: backwards-closed history + backward time shift → the HP-shifted domain contains the original.
Connects three layers:
BranchingTime.WorldHistory.backwardsClosed(semantic property)ConditionalShift.history_monotone_set(set-level monotonicity)hpShiftinstalls the expanded domain
The O-marking triviality problem: without domain expansion, the
Anderson conditional is vacuously true. The domainRestrictedConditional
over the non-expanded domain D is trivially satisfied because the
consequent (an observed fact) holds at every world in D.
The X/O-marking payoff: when domain expansion makes the consequent
non-trivial, the antecedent of a true conditional must exclude at
least one world in D⁺. The Anderson conditional is informative —
the antecedent restriction partitions D⁺ into consequent-satisfying
worlds (selected by the conditional) and consequent-failing worlds
(excluded by the antecedent).
In O-marking (HP) strategy, temporal indexicals still evaluate against the speech-act context (θ = origin), not the shifted context (v = innermost).
This explains why Japanese sakuya 'last night' in the antecedent of an HP-shifted O-marked conditional evaluates against the utterance time, paralleling the behavior of seventy-eight years ago in @cite{schlenker-2004}'s HP example.