Cross-Linguistic Modal Typology #
Empirical modal inventories from 27 languages (17 families) mapped to the 3×3 force-flavor meaning space, following Imel, Guo, & @cite{imel-guo-steinert-threlkeld-2026}.
Mapping conventions (raw typological data → 3×3 grid) #
- Force: weak →
.possibility, strong →.necessity, weak necessity →.weakNecessity - Flavor: epistemic →
.epistemic, deontic →.deontic, circumstantial →.circumstantial, teleological →.circumstantial - Bouletic, reportative, ability, intentional flavors are outside the 2×3 space and are dropped from the mapping.
- Only positive-polarity,
can_express = 1entries are included.
Data source #
@cite{steinert-threlkeld-imel-guo-2023}. A database for modal semantic typology. https://clmbr.shane.st/modal-typology/
Abbreviations for the nine meaning points #
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
Gitksan has variable-force modals: ima('a) and gat express both weak and strong epistemic force. These satisfy SAV (varying on force only, single flavor) and IFF (since {poss, nec} × {epistemic} is a Cartesian product).
Equations
- One or more equations did not get rendered due to their size.
Instances For
Gitksan's variable-force epistemic modals satisfy both SAV and IFF: {poss, nec} × {epistemic} varies on force only (single flavor).
Greek's Prepei violates SAV: it varies on both force and flavor axes.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Greek has non-IFF modals: Prepei and Mporei express non-rectangular subsets of the meaning space. Prepei covers {(nec,e),(poss,e),(nec,d),(nec,c)} which is NOT a Cartesian product (missing (poss,d) and (poss,c)).
Equations
- One or more equations did not get rendered due to their size.
Instances For
Prepei is NOT IFF: it expresses both forces and all three flavors, but does not express (possibility, deontic) or (possibility, circumstantial).
Mandarin has many modals, extensive synonymy, but all satisfy IFF. The paper notes Mandarin has three modals all encoding strong ∧ epistemic.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Dutch has one non-IFF modal: zou/zouden...kunnen expresses {(nec,e),(poss,e),(poss,c)} which is not Cartesian-closed (missing (nec,c)).
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
English modal inventory, derived from the Fragment (single source of truth).
Uses ModalInventory.fromAuxEntries to extract modals from AuxEntry data.
Equations
- One or more equations did not get rendered due to their size.
Instances For
All ten inventories.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Eight of ten encoded languages have perfect IFF degree (1.0).
All nine languages have IFF degree ≥ 1/3 (the minimum is Greek at 1/3).
Efficient Communication (Imel, Guo, & @cite{imel-guo-steinert-threlkeld-2026}) #
Key computational results (verified over 32,301 generated + 27 natural languages):
- Every Pareto-optimal modal system consists only of IFF modals.
- IFF degree correlates positively with simplicity, informativeness, and Pareto-optimality.
- Attested modal systems are more Pareto-optimal than the vast majority of hypothetically possible systems (mean optimality: 0.937 vs 0.776).
Mean Pareto-optimality of natural languages (Table 6).
Equations
Instances For
Mean Pareto-optimality of the generated population (Table 6).