Imanishi (2020): Parameterizing Split Ergativity in Mayan #
@cite{imanishi-2020}
Explains the alignment puzzle in the accusative (non-perfective) side of Mayan split-ergative systems.
The Alignment Puzzle #
Kaqchikel, Chol, and Q'anjob'al all have (nearly) identical biclausal
structures for non-perfective clauses — an aspectual predicate embedding
a nominalized clause [Asp ... [vP_NMLZ]]. Yet:
- Kaqchikel: S/A = ABS (set B), O = ERG/GEN (set A)
- Chol/Q'anjob'al: S/A = ERG/GEN (set A), O = ABS (set B)
The Analysis #
Two parameters explain the contrast:
Restriction on Nominalization (RON): The nominalizing head n in Kaqchikel obligatorily selects a vP lacking an external argument. In Chol and Q'anjob'al, n does not impose this restriction.
Mayan Absolutive Parameter (@cite{coon-mateo-pedro-preminger-2014}): High absolutive languages (Kaqchikel, Q'anjob'al) have Infl as the locus of absolutive Case; low absolutive languages (Chol) have Voice.
The RON alone determines the alignment type: it controls which argument is the highest DP inside the nominalized clause and thus which receives genitive Case from D.
Intransitivization Strategies #
The RON forces nominalized verbs in Kaqchikel to be intransitive. Three strategies satisfy this:
- Passivization: Voice[PASSIVE] suppresses the external argument
- Antipassivization: internal argument demoted to oblique
- Pseudo noun incorporation: object Case-licensed by adjacency
The Restriction on Nominalization (RON): "Nominalized verbs must lack a syntactically projected external argument."
A property of the nominalizing head n in a given language. When active, n obligatorily selects for a vP without an external argument (i.e., no specifier of VoiceP projected inside the nominalized clause).
- active : Bool
Instances For
Equations
- Phenomena.Ergativity.Studies.Imanishi2020.instBEqRON.beq { active := a } { active := b } = (a == b)
- Phenomena.Ergativity.Studies.Imanishi2020.instBEqRON.beq x✝¹ x✝ = false
Instances For
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
Equations
- One or more equations did not get rendered due to their size.
Instances For
Kaqchikel: RON active, high absolutive.
Equations
- Phenomena.Ergativity.Studies.Imanishi2020.kaqchikelParams = { ron := { active := true }, absPos := Fragments.Mayan.ABSPosition.high }
Instances For
Chol: RON inactive, low absolutive.
Equations
- Phenomena.Ergativity.Studies.Imanishi2020.cholParams = { ron := { active := false }, absPos := Fragments.Mayan.ABSPosition.low }
Instances For
Q'anjob'al: RON inactive, high absolutive.
Equations
- Phenomena.Ergativity.Studies.Imanishi2020.qanjobalParams = { ron := { active := false }, absPos := Fragments.Mayan.ABSPosition.high }
Instances For
Structure of the nominalized clause embedded under the aspectual
predicate on the accusative side. The clause is
[DP [nP [vP [VoiceP [VP]]]]] — verbal projections dominated by
nominal projections.
The key structural variable: whether an external argument is syntactically projected inside the nominalized clause. Determined by the RON.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Build the nominalized clause from the RON and transitivity.
Equations
Instances For
Derive the accusative-side alignment pattern from language parameters.
The D head of the nominalized clause assigns genitive Case to the structurally closest (highest) DP. Since ERG and GEN are homophonous in Mayan (both realized as set A markers), the DP receiving GEN is cross-referenced by set A.
- RON active → no external arg → internal arg is highest DP → set A on O
- RON inactive → external arg present → external arg is highest → set A on S
Equations
- One or more equations did not get rendered due to their size.
Instances For
Kaqchikel's parameters derive the Kaqchikel alignment pattern.
Chol's parameters derive the Chol alignment pattern.
Q'anjob'al's parameters derive the Chol pattern (same as Chol).
The RON alone determines the alignment type, regardless of ABSPosition. This is the paper's central result: the alignment puzzle reduces to a single binary parameter on the nominalizing head.
Q'anjob'al and Kaqchikel share ABSPosition but differ in alignment — confirming that ABSPosition alone does not determine the accusative-side alignment.
Passive Voice satisfies the RON: no θ-role assignment means no external argument is projected.
Agentive Voice violates the RON: it projects an external argument.
A Voice head is compatible with the RON iff it does not assign a θ-role (and hence does not project an external argument).
Instances For
Kaqchikel's perfective (ergative) case assignment from the existing fragment: A = ERG, S = P = ABS. Confirms the ergative side is shared across all three languages.
Chol's perfective alignment matches the same ergative pattern.
Q'anjob'al's perfective alignment matches the same ergative pattern.
Map Minimalist case values to the Mayan marker set that realizes them. ERG and GEN are both realized by set A (they are homophonous in Mayan). ABS is realized by set B. The wildcard maps non-Mayan cases (NOM, ACC, DAT, etc.) to set A — these should never appear in Mayan fragments.
Equations
- Phenomena.Ergativity.Studies.Imanishi2020.caseToMarker Minimalism.CaseVal.erg = Fragments.Mayan.MarkerSet.setA
- Phenomena.Ergativity.Studies.Imanishi2020.caseToMarker Minimalism.CaseVal.gen = Fragments.Mayan.MarkerSet.setA
- Phenomena.Ergativity.Studies.Imanishi2020.caseToMarker Minimalism.CaseVal.abs = Fragments.Mayan.MarkerSet.setB
- Phenomena.Ergativity.Studies.Imanishi2020.caseToMarker x✝ = Fragments.Mayan.MarkerSet.setA
Instances For
ERG/GEN homophony: both map to set A.
Chol's fragment case values, mapped through the Mayan marker bridge, yield the predicted accusative-side pattern.
Q'anjob'al's fragment case values yield the same pattern as Chol.
Kaqchikel's fragment accusative-side case values, mapped through the Mayan marker bridge, yield the predicted Kaqchikel alignment pattern.
The accusative-side case contrast between Kaqchikel and Chol is a true mirror image: agent and patient cases are swapped.
End-to-end: for all three languages, the fragment case data (mapped through the marker bridge) matches the parametrically derived pattern. This closes the argumentation chain from parameters → alignment → case → markers.
The Mayan split is aspect-conditioned: perfective → ergative,
non-perfective → accusative. Instantiates the same SplitErgativity
infrastructure as the Hindi example in Core.Case.SplitConditions.
Equations
- Phenomena.Ergativity.Studies.Imanishi2020.mayanSplit = { ergCondition := fun (a : Core.Aspect) => a == Core.Aspect.perfective }
Instances For
Mayan and Hindi have the same aspect-conditioned split direction: perfective triggers ergativity in both language families.
Imanishi's RON determines the accusative-side alignment (this study), while CMP2014's CaseLocus determines syntactic ergativity (ergative side). Together they form the full Mayan parameterization: RON for the accusative side, ABSPosition→CaseLocus for the ergative side.
The two studies agree on Kaqchikel: RON active + HIGH-ABS = syntactic ergativity + Kaqchikel-type accusative alignment.
The two studies agree on Chol: RON inactive + LOW-ABS = no syntactic ergativity + Chol-type accusative alignment.
Q'anjob'al shows that the two dimensions are independent: HIGH-ABS (like Kaqchikel) but RON inactive (like Chol). Syntactic ergativity yes, but Chol-type accusative alignment.