Matthewson (2016) — Modality #
@cite{matthewson-2016}
Lisa Matthewson. "Modality." Chapter 18 in The Cambridge Handbook of Formal Semantics, ed. Maria Aloni and Paul Dekker. Cambridge University Press. pp. 525–559. DOI: 10.1017/CBO9781139236157.019.
A survey of three core topics in modal semantics — modal flavour (§18.2), modal force (§18.3), and modal–temporal interactions (§18.4) — presented within the Kratzerian framework with cross-linguistic data from Gitksan, St'át'imcets, Nez Perce, Niuean, and other languages.
Key contributions formalized here #
Three-way background classification (Table 18.3): factual-circumstantial, factual-evidential, content-evidential. Refines the traditional epistemic/circumstantial binary following @cite{kratzer-2012}.
Modals without duals (§18.3.2): Gitksan ima('a)/gat and Nez Perce o'qa are not specialized for necessity or possibility. Different analyses: variable force (Peterson 2010) vs. strengthened possibility (Deal 2011).
Cross-linguistic flavour–force correlation (§18.5): epistemic modals are more likely to lack duals than circumstantial modals. Gitksan and Niuean both encode force distinctions in the circumstantial domain but not the epistemic domain.
Gitksan three-way split (Table 18.1): the Gitksan modal system lexicalizes all three background classes — factual-circumstantial (da'akhlxw, anookxw, sgi), factual-evidential (ima('a)), and content-evidential (gat).
Temporal orientation and prospective aspect (§18.4.3): Gitksan requires overt prospective marking (dim) for future temporal orientation, mirroring English's requirement of have for past orientation.
The three-way classification refines the traditional binary. All three classes are distinct.
Both epistemic subtypes (factual-evidential and content-evidential) map to epistemic under the traditional classification.
Only the content-evidential class allows speaker disbelief. This is the diagnostic that separates the two epistemic subtypes: St'át'imcets k'a (factual) vs lákw7a (content).
The traditional circumstantial class is uniformly factual (Table 18.2).
Gitksan ima('a) is factual-evidential: the speaker has inferential evidence and cannot disbelieve the prejacent.
Gitksan gat is content-evidential: reportative evidence, the speaker can disbelieve.
Gitksan circumstantial modals are factual-circumstantial.
All three background classes are represented in Gitksan.
The epistemic/circumstantial split is absolute: no modal crosses the boundary. Epistemic modals are purely epistemic; circumstantial modals have no epistemic readings.
Variable-force modals (Gitksan) and strengthened possibility modals (Nez Perce) both lack duals but for different reasons.
Gitksan ima('a) and gat are both variable-force.
Nez Perce o'qa is a strengthened possibility modal.
Both analyses agree: the modals lack duals.
Despite lacking duals, both admit necessity and possibility readings.
Force analysis consistency #
Each fragment's stipulated ForceAnalysis is verified against the
observable ForcePattern derived from the meaning. The stipulation
adds explanatory content (variable-force vs strengthened), but it
must be consistent with the structural facts.
Gitksan ima('a): variable-force analysis is consistent with multiForce pattern (both possibility and necessity in meaning).
Gitksan gat: variable-force analysis is consistent.
Nez Perce o'qa: strengthened possibility is consistent with singleForce pattern (only possibility in meaning set).
St'át'imcets =ka: variable-force analysis is consistent.
Niuean liga: variable-force analysis is consistent.
Cross-linguistic tendency: epistemic modals are more likely to lack force duals than circumstantial modals. Both Gitksan and Niuean instantiate this pattern.
Niuean: epistemic domain has one modal (both forces), circumstantial has two (one per force).
All St'át'imcets and Niuean modals satisfy IFF.
In Gitksan, future temporal orientation of an epistemic modal
(imaa) requires prospective dim. The 2016 handbook chapter
presents this as the headline pattern; @cite{matthewson-2013} shows
it is part of a flavor-keyed asymmetry (circumstantials require
dim for any orientation).
For epistemics, past and present orientation do not require dim.
English–Gitksan mirror: English obligatorily marks past orientation (via have), Gitksan obligatorily marks future orientation (via dim). Both leave the remaining orientations unmarked. @cite{matthewson-2016} §18.4.3.
- obligatoryMarking : Core.Modality.TemporalOrientation
Which orientation is obligatorily marked.
- marker : String
Name of the marker.
Instances For
Equations
- Matthewson2016.english = { obligatoryMarking := Core.Modality.TemporalOrientation.past, marker := "have" }
Instances For
Equations
- Matthewson2016.gitksan = { obligatoryMarking := Core.Modality.TemporalOrientation.future, marker := "dim" }
Instances For
The marked orientations are opposite: English marks past, Gitksan future.
@cite{matthewson-2016} §18.5 discusses @cite{nauze-2008}'s proposed universal: "Modal elements can only have more than one meaning along a unique axis of the semantic space: they either vary on the horizontal axis [flavour] ... or they vary on the vertical axis [force] ... but they cannot vary on both axes."
This is exactly SAV. We verify it holds for all four new fragments.
St'át'imcets =ka satisfies SAV (varies on force, fixed deontic).
Nez Perce o'qa satisfies SAV (singleton).
Niuean: all modals satisfy SAV.
@cite{hacquard-2006} @cite{hacquard-2010}
Matthewson's three-way background classification (Table 18.3) refines the traditional epistemic/circumstantial binary. The coarse binary itself is derived, not stipulated: @cite{hacquard-2006}'s content licensing predicts that only contentful events (speech acts, attitudes) can project epistemic modal bases. VP events lack content and can only project circumstantial bases. This predicts the absolute epistemic/circumstantial split attested in Gitksan (§18.2.3).
The three-way refinement (factual-evidential vs content-evidential within epistemic) is a further subdivision of the epistemic class that content licensing does not address — it depends on the type of content (inferential vs reportative), not on whether content exists.
Content licensing correctly predicts the coarse split. Contentful events → epistemic available; contentless → not.
The three-way refinement is orthogonal to content licensing. Both factual-evidential and content-evidential are epistemic subtypes (both require content), distinguished by projection mode, not by content availability.
Gitksan's three-way split is consistent with content licensing. Epistemic modals (ima('a), gat) are high (content available); circumstantial modals (da'akhlxw, anookxw, sgi) are compatible with both high and low positions (circumstantial always available).
These inventories can be compared against the Imel, Guo & @cite{imel-guo-steinert-threlkeld-2026} typological database but are kept here because their data source is @cite{matthewson-2016}, not the Imel et al. database.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- Matthewson2016.nezPerceInventory = { language := "Nez Perce", family := "Sahaptian", source := "Deal (2011)", expressions := Fragments.NezPerce.Modals.allExpressions }
Instances For
Equations
- Matthewson2016.niueanInventory = { language := "Niuean", family := "Polynesian", source := "Matthewson et al. (2012), Seiter (1980)", expressions := Fragments.Niuean.Modals.allExpressions }
Instances For
All three Matthewson 2016 inventories satisfy IFF.