Oblique Extraction in Mayan #
@cite{elkins-torrence-brown-2026} @cite{mendes-ranero-2021} @cite{imanishi-2020}
Part I: Cross-Linguistic Comparison #
Cross-linguistic comparison of extraction morphology in two Mayan language groups: SJO Mam (=(y)a') and K'ichean (wi). Both mark oblique extraction with a dedicated morpheme, but the underlying mechanisms and distributional properties differ.
Shared Properties #
- Both mark oblique extraction (spatial, instrumental)
- Both exempt temporal obliques ('when')
- Neither marks subject extraction (Agent Focus instead)
- Neither marks object extraction
Parametric Differences #
| Property | Mam =(y)a' | K'ichean wi |
|---|---|---|
| Locus | On probe (Voice⁰) | At extraction site |
| Mechanism | Agree reflex | Copy spellout |
| Reason obliques ('why') | =(y)a' ✓ | wi ✗ |
| FPG (matrix wi ↔ embedded comp) | Does not hold | Holds |
| Conditioned by clause size | Yes (Voice project.) | No |
| Multiple spellout in LD | Yes (per Voice/Dir) | Unclear |
Part II: Minimalist Analysis #
Connects three Minimalist abstractions — ClauseSpine, Agree/feature-valuation, and Spellout — to the empirical data on =(y)a' distribution in SJO Mam.
- Voice⁰ (and Dir⁰) in Mam carry [uOblique] (an unvalued probe feature).
- When an oblique DP undergoes successive-cyclic Ā-movement through Spec,VoiceP, Agree values [uOblique] as [+oblique] on Voice⁰.
- At Spellout (PF), [+oblique] on Voice⁰ is realized as =(y)a'.
- In infinitival complements, Voice is not projected (VP-sized), so there is no [uOblique] probe — =(y)a' cannot appear.
- In long-distance extraction, each Voice⁰/Dir⁰ along the movement path independently Agrees, yielding multiple =(y)a' (one per Voice/Dir).
Both Mam and K'ichean use dedicated morphemes for oblique extraction.
Both exempt temporal obliques from extraction marking.
Neither marks subject extraction (Agent Focus instead).
Neither marks object extraction.
KEY CONTRAST — Reason obliques ('why'): Mam =(y)a' IS licensed with reason extraction; K'ichean wi is NOT.
Mam =(y)a' is conditioned by clause size (Voice must project); K'ichean wi is conditioned by complementizer presence (FPG).
The FPG holds for K'ichean: matrix wi tracks overt complementizer.
Genuinely different mechanisms producing superficially similar patterns.
- agreeReflex : ExtractionMorphologyMechanism
- copySpellout : ExtractionMorphologyMechanism
Instances For
Equations
- Phenomena.FillerGap.Studies.ElkinsTorrenceBrown2026.instDecidableEqExtractionMorphologyMechanism x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
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.
Mam Voice substrate (Minimalist) #
This subsection houses the Minimalist VoiceHead, ClauseSpine, and
MamDirHead definitions for Mam, formerly in
Linglib/Fragments/Mayan/Mam/VoiceSystem.lean. Per CLAUDE.md
"Per-language paper-specific apparatus lives in Studies, not
Fragments," these belong with the paper that anchors them
(@cite{elkins-torrence-brown-2026} for the =(y)a' analysis;
@cite{scott-2023} for the antipassive). The Fragment file retains only
the theory-neutral mamVoiceSystem : VoiceSystemProfile.
Phenomena/Agreement/Studies/Scott2023.lean consumes mamVoice and
eqYaVocab from this file via cross-Studies import.
Mam agentive Voice head with [uOblique] probe.
In Mam, Voice⁰ probes for an oblique feature on a passing Ā-moved constituent. When an oblique DP moves through Spec,VoiceP, Agree values [uOblique] as [+oblique], which is then spelled out as =(y)a' at PF.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Mam transitive clause spine: full CP with Voice.
Equations
Instances For
Mam aspectless complement spine: VoiceP-sized. Still has Voice → =(y)a' possible.
Equations
Instances For
Mam infinitival complement spine: VP-sized. No Voice → =(y)a' impossible.
Equations
Instances For
Mam directional auxiliary head (Dir⁰).
Dir is NOT a universal category — it is specific to Mayan languages.
Modeled as a language-specific type rather than added to Cat.
In Elkins et al.'s analysis, Dir⁰ occupies V1 position in the verbal
template (Voice > V1(Dir) > Appl > V2(root)). Like Voice⁰, Dir⁰
bears [uOblique] and can host =(y)a'.
- cislocative : Bool
Cislocative (toward speaker) vs translocative (away).
- hasUOblique : Bool
Whether this Dir head carries [uOblique].
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
Dir⁰'s probe features when it carries [uOblique].
Equations
- d.features = if d.hasUOblique = true then [Minimalist.GramFeature.unvalued (Minimalist.FeatureVal.oblique false)] else []
Instances For
Cislocative directional with [uOblique].
Equations
- Phenomena.FillerGap.Studies.ElkinsTorrenceBrown2026.dirCis = { cislocative := true, hasUOblique := true }
Instances For
Translocative directional with [uOblique].
Equations
- Phenomena.FillerGap.Studies.ElkinsTorrenceBrown2026.dirTrans = { cislocative := false, hasUOblique := true }
Instances For
Vocabulary entry for =(y)a': maps [+oblique] on Voice⁰ to "=(y)a'".
Equations
- One or more equations did not get rendered due to their size.
Instances For
The Mam Voice vocabulary: just the =(y)a' entry.
Equations
Instances For
Mam passive Voice head: carries [uOblique] just like agentive Voice. @cite{elkins-torrence-brown-2026} §7.2: =(y)a' co-occurs with passive -njtz.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Mam antipassive Voice head (@cite{scott-2023} §2.5.4.1). Subject gets ABS not ERG; not a phase head.
Equations
- Phenomena.FillerGap.Studies.ElkinsTorrenceBrown2026.mamAntipassiveVoice = { flavor := Minimalist.VoiceFlavor.antipassive, hasD := true }
Instances For
Mam Voice head carries [uOblique].
Mam Voice is a phase head.
Mam Voice assigns a θ-role (agentive).
Mam transitive spine projects Voice.
Mam aspectless spine projects Voice.
Mam infinitival spine does NOT project Voice.
Cislocative Dir carries [uOblique].
Translocative Dir carries [uOblique].
Passive and agentive Voice differ in flavor but share the same oblique probe features.
Antipassive Voice assigns a θ-role (the agent is present).
Antipassive Voice is NOT a phase head.
Antipassive and agentive Voice differ in phase-head status but both assign θ-roles.
Valued [+oblique] on Voice spells out as =(y)a'.
Without [+oblique], Voice has no exponent from this vocabulary.
[+oblique] on a non-Voice head does not yield =(y)a' (context restriction blocks insertion).
Predict whether =(y)a' is licensed from a clause spine and extraction type.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Transitive clause + oblique extraction → =(y)a' licensed.
Transitive + subject extraction → no =(y)a'.
Transitive + object extraction → no =(y)a'.
Passive + oblique → =(y)a' licensed.
Temporal oblique + full clause → =(y)a' BLOCKED.
Map MamClauseType to the corresponding ClauseSpine.
Equations
- Phenomena.FillerGap.Studies.ElkinsTorrenceBrown2026.spineOf Fragments.Mayan.Mam.MamClauseType.fullCP = Phenomena.FillerGap.Studies.ElkinsTorrenceBrown2026.mamTransitiveSpine
- Phenomena.FillerGap.Studies.ElkinsTorrenceBrown2026.spineOf Fragments.Mayan.Mam.MamClauseType.aspectless = Phenomena.FillerGap.Studies.ElkinsTorrenceBrown2026.mamAspectlessSpine
- Phenomena.FillerGap.Studies.ElkinsTorrenceBrown2026.spineOf Fragments.Mayan.Mam.MamClauseType.infinitival = Phenomena.FillerGap.Studies.ElkinsTorrenceBrown2026.mamInfinitivalSpine
Instances For
LD from full CP → =(y)a' licensed on both predicates.
LD from aspectless → =(y)a' licensed on both.
LD from infinitival → =(y)a' on matrix only.
EQ → =(y)a' on embedded only.
All monoclausal predictions match.
All LD embedded predictions match.
Against resumptive-pronoun analysis: =(y)a' is island-sensitive.
Against Agent Focus analysis: =(y)a' co-occurs with passive voice.
Against copy spellout: =(y)a' disappears when Voice is absent.
ClauseSpine is finer than ComplementSize: it can distinguish infinitival (VP) from aspectless (VoiceP).
Dir⁰ with valued [+oblique] also spells out as =(y)a'.
Agree at Voice: [uOblique] valued by oblique DP's [+oblique].
Spellout: valued [+oblique] on Voice maps to "=(y)a'".
Full derivation pipeline: Agree then Spellout → "=(y)a'".
Voice⁰ in Mam carries two independent probes:
- φ-probe [uPerson, uNumber] (analyzed by @cite{scott-2023}): Agrees with agent in Spec,VoiceP, yielding Set A morphology.
- Oblique probe [uOblique] (analyzed by @cite{elkins-torrence-brown-2026}, this file): Agrees with a passing Ā-moved oblique, yielding =(y)a' on Voice⁰.
Both are instances of the same abstract Agree operation: probe searches c-command domain, finds closest matching goal, copies features, and the valued features are spelled out by Vocabulary Insertion. They differ only in which features they probe for and what vocabulary entries match.
Both Voice probes are unvalued features.
φ-Agree (Scott 2023) and oblique-Agree (this paper) are parallel instances of the same operation, differing only in which features are probed and which vocabulary entries match.