Anderson Conditionals and Domain Expansion #
@cite{condoravdi-2002} @cite{mizuno-2024} @cite{schlenker-2004}
Formalizes the connection between backward temporal shifts and domain expansion in conditionals. @cite{mizuno-2024} argues that Japanese Anderson conditionals use the Historical Present (@cite{schlenker-2004}) to achieve domain expansion without X-marking: Non-Past morphology shifts the evaluation time backward, and under branching time (@cite{condoravdi-2002}), earlier times have more historical alternatives, so the domain expands.
Key Results #
andersonConditional— HP-shifted antecedent evaluated against expanded domainhp_achieves_expansion— backward time + domain monotonicity yields expansiontrivialConsequent/nonTrivialConsequent— formalize the triviality puzzleexpansion_resolves_triviality— domain expansion makes conditionals non-trivial- Bridge to
BranchingTime.historicalBaseandMood.SUBJ
Connection to ContextTower #
The HP shift in an Anderson conditional is modeled as a tower push of an
hpShift: a context shift that moves time backward and expands the domain.
In @cite{schlenker-2004}'s terms, the push shifts the Context of Utterance v
while preserving the Context of Thought θ (= tower.origin).
An Anderson conditional (context-level model): the antecedent is evaluated at an HP-shifted context (backward time, expanded domain), and the consequent is evaluated at the original context.
This models the HP shift's effect on a single evaluation point.
The full Kratzer-style analysis — restricted universal quantification
over the expanded domain D⁺ — is domainRestrictedConditional:
domainRestrictedConditional D⁺ antecedent consequent
= ∀ w ∈ D⁺, antecedent(w) → consequent(w)
The domain expansion machinery here (hpShift, hp_achieves_expansion)
provides the D⁺, and trivial_domainRestricted /
nontrivial_conditional_excludes show why expansion matters.
Equations
- Semantics.Tense.ConditionalShift.andersonConditional antecedent consequent hpTime expandedDomain rc = (antecedent ((Core.Context.hpShift hpTime expandedDomain).apply rc) → consequent rc)
Instances For
The HP-shifted context in an Anderson conditional has the shifted time.
@cite{mizuno-2024}'s argument: backward time + domain monotonicity yields expansion.
If the world history is backwards-closed (worlds that agree up to t
also agree up to t' ≤ t), then the historical alternatives at an
earlier time are a superset of those at a later time. This is why
O-marking (Non-Past / HP) in Japanese Anderson conditionals expands
the domain without X-marking.
Set-level monotonicity: under backwards-closed history, the set of
historical alternatives at an earlier time is a superset of those at a
later time. This lifts hp_achieves_expansion (element-level) to
Set.Subset (set-level), connecting it to DomainExpanding.
This is the formal core of @cite{mizuno-2024}'s argument: HP shifts the evaluation time backward, and backward time yields more historical alternatives, i.e., domain expansion.
The historical base (set of situations) at an earlier time includes situations with the same worlds as the later base, plus potentially more. This is the situation-level version of domain expansion.
Domain-restricted conditional: the standard Kratzer-style analysis of conditionals as restricted universal quantification over a modal domain.
∀ w ∈ D, antecedent(w) → consequent(w)
@cite{kratzer-1986}: if-clauses restrict the modal domain rather than
functioning as binary connectives. This is the Prop-level counterpart
of Semantics.Conditionals.Restrictor.conditionalNecessity (which
operates over finite (World → Bool) for computation).
Both X-marking and O-marking strategies for Anderson conditionals work by expanding D to D⁺ ⊃ D, making this quantification non-trivial.
Equations
- Semantics.Tense.ConditionalShift.domainRestrictedConditional domain antecedent consequent = ∀ w ∈ domain, antecedent w → consequent w
Instances For
A conditional is trivial when every world in the domain satisfies the consequent. The antecedent restriction does no work — the conditional is vacuously true regardless of what the antecedent says.
@cite{condoravdi-2002}: indicative conditionals with small domains can be trivial because every accessible world already satisfies the consequent. Domain expansion (via HP/X-marking) resolves this by adding worlds where the consequent may fail.
Equations
- Semantics.Tense.ConditionalShift.trivialConsequent domain consequent = ∀ w ∈ domain, consequent w
Instances For
A conditional is non-trivial when there exists a world in the domain where the consequent fails. This is the condition under which the antecedent restriction does meaningful work.
Equations
- Semantics.Tense.ConditionalShift.nonTrivialConsequent domain consequent = ∃ w ∈ domain, ¬consequent w
Instances For
The triviality problem: when the consequent is trivial, the domain-restricted conditional is vacuously true regardless of the antecedent. This is why O-marked English Anderson conditionals are infelicitous — the consequent is an observed fact true at all worlds in D, so the conditional adds no information.
Domain expansion resolves triviality: if the original domain makes the consequent trivial, but the expanded domain contains a world where the consequent fails, then the expanded conditional is non-trivial.
This completes the HP/X-marking argument:
hp_achieves_expansion— backward time shift expands the domainexpansion_resolves_triviality— expanded domain makes the conditional non-trivial
The counterfactual "If I had left earlier, I would have caught the train" is non-trivial precisely because the expanded domain (from X-marking's backward time shift) includes worlds where I didn't catch the train.
When the domain-restricted conditional holds over an expanded domain where the consequent is non-trivial, the antecedent must exclude at least one world. The antecedent restriction is doing genuine work — it is not vacuously satisfied.
This is the formal payoff of domain expansion: the conditional becomes informative because the antecedent partitions D⁺ into worlds where the consequent holds (via the conditional) and worlds where it fails (via non-triviality).
Triviality is monotone: if a superset domain is trivial, then so is any subset domain. Expanding the domain can only resolve triviality, never introduce it.
SUBJ's situation introduction can be decomposed into two steps:
- Expand the domain (via backward time shift)
- Existentially quantify over the expanded domain
When the history is backwards-closed, SUBJ at an earlier time introduces a situation whose world is in the expanded historical alternatives.