Tense Theories: Cross-Cutting Comparison #
@cite{abusch-1997} @cite{von-stechow-2009} @cite{kratzer-1998} @cite{ogihara-1996} @cite{klecha-2016} @cite{deal-2020} @cite{sharvit-2003} @cite{tsilia-zhao-sharvit-2026} @cite{zeijlstra-2012} @cite{wurmbrand-2014}
Comparison matrix for ten tense theories:
- Semantic: @cite{abusch-1997}, @cite{von-stechow-2009}, @cite{heim-kratzer-1998}, @cite{ogihara-1996}, @cite{klecha-2016}, @cite{deal-2020}, @cite{sharvit-2003}, @cite{tsilia-zhao-sharvit-2026}
- Syntactic: @cite{zeijlstra-2012}, @cite{wurmbrand-2014}
Verdicts are assembled from derivation theorems proved in each theory's file, not stipulated here. This file proves cross-cutting theorems about where theories converge, diverge, and where each is silent.
Structure #
- Convergence: phenomena all theories handle (core SOT)
- Mechanism comparison: same phenomenon, different derivations
- Divergence / Limitations: what distinguishes theories
- No complete theory: every theory has gaps
- Bridges to existing infrastructure: grounding theorems
- Extended coverage: Agree-based SOT, indirect question SOT, infinitival tense
Scope #
This file compares tense theory identity cards — what mechanism each
theory uses for SOT, where it is silent, and how theories relate
structurally. For @cite{partee-1973}'s tense–pronoun structural analogy
(indexical/anaphoric/bound-variable interpretations of tense morphemes),
see Comparisons/Partee1973.lean.
All six theories derive the core SOT phenomena: shifted and simultaneous readings of past-under-past, and double-access for present-under-past. The mechanism differs (binding vs deletion vs zero tense vs feature checking), but the predicted Reichenbach frames are identical.
All six theories derive the shifted reading (R < P in embedded frame).
All theories that handle core SOT derive the simultaneous reading, but via distinct mechanisms:
- Abusch: bound variable receives matrix E
- Kratzer: SOT deletion removes embedded past
- Ogihara: zero tense reading of past
- Zeijlstra: [uPAST] Agrees upward with [iPAST]
- Sharvit: simultaneous tense with its own semantics All produce R = P (present relative to embedded perspective).
The same surface phenomenon (simultaneous reading) has three completely different derivations across the theories.
Three mechanisms for the simultaneous reading, all producing the same Reichenbach relation (R = P).
Von Stechow's mechanism is orthogonal to all three: feature checking doesn't require binding, deletion, OR zero tense.
Four distinct simultaneous mechanisms across the nine theories: binding (Abusch) ≠ deletion (Kratzer) ≠ zero tense (Ogihara) ≠ Agree (Zeijlstra) ≠ simultaneous tense (Sharvit).
Only Abusch has temporal de re.
Klecha is the only theory designed specifically for modal-tense interaction. Deal also addresses modals but through the counterfactual lens.
Deal is the only theory that distinguishes temporal from counterfactual uses of past morphology.
Sharvit's challenge to Abusch: the binding/de re mechanism does not extend straightforwardly to relative clauses, where the tense takes the perspective of the modified NP, not of an attitude holder.
Von Stechow handles this via feature checking (uniform mechanism). Abusch requires attitude semantics or res movement, which relative clauses don't obviously provide.
The deletion-vs-ambiguity debate: Kratzer and Ogihara make the same SOT predictions but differ on what "past" means.
- Kratzer: past is never ambiguous; simultaneous = deletion
- Ogihara: past is ambiguous; simultaneous = zero tense reading
The empirical predictions are identical for core SOT data. The debate is about linguistic ontology, not about truth conditions.
Every theory is silent or problematic on at least one phenomenon.
Semantic theories:
- Abusch: silent on relative clause tense, modal-tense, counterfactual
- Von Stechow: silent on temporal de re, modal-tense, counterfactual
- Kratzer: silent on temporal de re, modal-tense, counterfactual, RC tense
- Ogihara: silent on temporal de re, modal-tense, counterfactual, RC tense
- Klecha: silent on temporal de re, counterfactual, RC tense
- Deal: silent on temporal de re, RC tense
- Sharvit: silent on temporal de re, modal-tense, counterfactual
Syntactic theories:
- Zeijlstra: silent on temporal de re, modal-tense, counterfactual, RC tense
- Wurmbrand: silent on temporal de re, modal-tense, counterfactual, RC tense
Formalized via the identity cards: no single theory has all capabilities.
Minimal cover: Abusch + Von Stechow + Klecha + Deal + Sharvit covers all phenomena.
- Abusch: temporal de re, ULC, shifted/simultaneous/double-access
- Von Stechow: relative clause tense, feature checking
- Klecha: modal-tense interaction
- Deal: counterfactual tense, ULC refinement
- Sharvit: indirect question SOT, embedded present puzzle
The existing embeddedFrame in IS/Tense/Basic IS Von Stechow's
perspective shift mechanism.
The existing zeroTense_receives_binder_time IS Ogihara's
zero tense mechanism.
The evalTimeIndex field on TensePronoun IS Klecha's insight:
modals and attitudes shift the eval time by updating this index.
All theories interpret tense differently. The semantic theories
share the TensePronoun type but give it different interpretations;
@cite{zeijlstra-2012} operates at narrow syntax (Agree);
@cite{wurmbrand-2014} classifies infinitival tense orthogonally.
Sharvit is the only theory with a dedicated simultaneous tense. Other theories derive the simultaneous reading from binding (Abusch), deletion (Kratzer), zero tense (Ogihara), Agree (Zeijlstra), or feature checking (Von Stechow). None of those posit a separate tense morpheme whose semantics IS simultaneity.
The semantic vs syntactic divide.
Eight theories operate at LF (semantic denotation); only @cite{zeijlstra-2012} uses syntactic Agree for SOT. @cite{wurmbrand-2014} is a syntactic approach to infinitival tense classification but does not present an Agree-based SOT mechanism — it is compatible with both semantic and syntactic SOT.
- Only Zeijlstra uses Agree-based SOT
- None of the non-Agree theories posit temporal de re or ULC as part of an Agree system
- Wurmbrand's contribution is orthogonal: infinitival tense classification, not SOT mechanism
Only @cite{tsilia-zhao-sharvit-2026} treats tenses as presupposition
triggers. All other theories have hasPresuppositionalTense = false.
All ten theories are distinct.
Only Zeijlstra's Agree-based account predicts size-sensitive SOT. All semantic theories (Abusch, Von Stechow, Kratzer, Ogihara, etc.) treat SOT as a whole-language parameter, not as structurally conditioned. @cite{egressy-2026} argues this is evidence for the syntactic (Agree-based) approach over purely semantic ones.
Only Zeijlstra predicts size-sensitive SOT.
@cite{egressy-2026}'s core argument: among the ten implemented theories, the only theory that predicts size-sensitive SOT is one that uses Agree-based SOT. All purely semantic theories (which lack structurally- local SOT mechanisms) predict uniform SOT within a language.
This captures the paper's argument structure:
- Semantic SOT is not locality-constrained → predicts uniform SOT
- Agree-based SOT is locality-constrained → interacts with PIC
- Hungarian shows non-uniform (size-sensitive) SOT
- Therefore: data favors Agree-based (syntactic) approach