Docs / Studies / Studies.Fagen2026

Order Selection: The Logical Foundations of Exhaustivity

Fagen, L. (2026). PhD thesis, The University of Chicago

Fagen derives the exhaustive interpretation of an answer from its meaning and the question, with no theory of alternative sentences: following Schulz and van Rooij, an answer holds in the best worlds where it is true, and what varies is the order that ranks the worlds. The default order minimizes the question-property, which gives a positive answer such as Agnes its negative direction and unrestricted extent; under it exhaustive interpretation is the minimal-worlds exhaustifier over the question's propositions. A negative or mixed answer is only about its restriction, the intersection of the sets it lives on, which the positive and negative restrictions make up; the hybrid order minimizes the question-property within the first and maximizes it within the second, so that nothing outside the restriction is concluded. Positive skew, a restriction equal to the positive restriction, selects the default order. Some of the linguists but not all of the computer scientists is not positively skewed, and the hybrid order makes exactly one linguist and all but one computer scientist attend. Positive skew is exactly Spector's quasi-positivity, which Spector stated over formulae of a propositional language.

Implementation notes

  • An answer is a quantifier T over individuals and the question a property P : W → Set E of worlds; exhaustive interpretation exh P T r takes the order r as a parameter, as (29) does.
  • The worked examples take the worlds to be all extensions of the question-property, as Fagen's diagrams do.
  • A proposition of Spector's question-relative language, a set of valuations of the atoms, is a quantifier over the atoms, quantifierOf.

Errata

  • The hybrid order (31) is not transitive once an answer's positive and negative restrictions each have three members, as in some of the linguists but not all of the computer scientists with three of each (not_transitive_ltHybrid); Fagen calls it an order and notes only that it is antisymmetric. Exhaustive interpretation is defined from the relation itself, so his predictions stand.
  • The proof that a quantifier lives on its restriction (A4) ends by claiming the result for any countable family of live-on sets; it holds only for finitely many, and fails for the cofinite sets of numbers (not_livesOn_restriction_cofinite). The appendix assumes a finite model, where the result holds (livesOn_restriction).

TODO

  • Preserve Restriction (26), from which §3.2 derives order selection (12), is not formalized; nor are answers to second-order questions (§2.4) or the conditions under which the dual order is chosen (§3.3).
  • The answers of Table 3.1 are not recorded as data.

References

  • Fagen 2026, Order Selection: The Logical Foundations of Exhaustivity
  • Fagen 2026, Order Selection: The Logical Foundations of Exhaustivity
  • Fagen 2026, Order Selection: The Logical Foundations of Exhaustivity

Restriction operators (§2.2, Appendix)

L78–79
L81–83
L85–87
L96–98

Orders over worlds (§2.1, §2.3, §3.3)

L253–254

Under the default order, exhaustive interpretation is the minimal-worlds exhaustifier over the propositions that the question-property holds of each individual ((83)).

theorem exh_ltDefaultL267–277
{E : Type u_1} {W : Type u_2} (P : W → Set E) (T : Quantifier E) :
exh P T (ltDefault P) = Exhaustification.exhMW (Set.range fun (x : E) => {w : W | x ∈ P w}) {w : W | T (P w)}

Built from ,

Worked examples (§2.1.2, §2.3)

L328–339
L342–352
L389–394

Some of the L but not all of the C takes the hybrid order and says that exactly one L and all but one C have the property ((40)).

theorem exh_selectedOrder_someButNotAllL396–458
{E : Type u_1} {L C : Set E} (hLC : Disjoint L C) (hL : L.Nonempty) (hC : C.Nonempty) :
exh id (someButNotAll L C) (selectedOrder id (someButNotAll L C)) = {S : Set E | (∃ (l : E), S ∩ L = {l}) ∧ ∃ (c : E), C \ S = {c}}

Built from , ,

L461–471

Spector's favoring (§1.3.2, §2.2.2)

A proposition is quasi-positive in Spector's sense, strongly favoring no negative literal, just when the quantifier it determines is positively skewed (§2.1.1, after Spector's (77b)).

theorem quasiPositive_iff_positivelySkewedL519–524
{Atom : Type u_2} [DecidableEq Atom] {P : Spector2007.Proposition Atom} :

Built from ,