Documentation

Linglib.Phenomena.Ergativity.Studies.CoonMateoPedroPreminger2014

Coon, Mateo Pedro & Preminger (2014) @cite{coon-mateo-pedro-preminger-2014} #

The role of Case in A-bar extraction asymmetries: Evidence from Mayan. Linguistic Variation 14(2), 179–242.

Core Claim #

Syntactic ergativity — the ban on A-bar extracting transitive subjects in languages like Q'anjob'al — is not about properties of the ergative NP. It is a locality problem in case assignment to the absolutive object.

In HIGH-ABS languages (ABS=NOM), Infl⁰ assigns absolutive case. The transitive object must raise out of vP to receive this case, passing through the single escape hatch at the vP phase edge — trapping the subject. In LOW-ABS languages (ABS=DEF), v⁰ assigns absolutive locally within vP. The escape hatch is free and the subject extracts without issue.

The Agent Focus Construction #

Q'anjob'al's AF morpheme -on is analyzed as a marked variant of Voice⁰ that assigns structural case to the internal argument (the transitive object). When Voice⁰_AF assigns case, Infl⁰ is freed to assign case to the subject. With the escape hatch unoccupied, the subject can extract. The intransitive status suffix -i surfaces because AF Voice is non-phasal (intransitive v⁰).

The Crazy Antipassive #

The same -on morpheme appears in Q'anjob'al non-finite embedded transitives — environments where Infl⁰ is absent and thus cannot assign case to objects. The -on provides the needed case source, further supporting its role as a case-assigner.

Three Factors for Syntactic Ergativity in Q'anjob'al #

  1. Transitive vP is phasal (constitutes a locality domain)
  2. The transitive subject is generated below vP (in Spec,VoiceP)
  3. There is only a single specifier available for extraction out of vP

Abstract case features assigned by functional heads. Following @cite{legate-2008}, "absolutive" is not an abstract case but a descriptive cover term for the morphological form shared by transitive objects and intransitive subjects. The actual abstract cases are:

  • NOM: assigned by Infl⁰ (to intransitive subjects universally, and to transitive objects in ABS=NOM languages)
  • ACC: assigned by v⁰ (to transitive objects in ABS=DEF languages)
  • ERG: assigned by v⁰ (to transitive subjects universally)
Instances For
    Equations
    • One or more equations did not get rendered due to their size.
    Instances For

      Which functional head assigns each abstract case.

      Instances For
        Equations
        • One or more equations did not get rendered due to their size.
        Instances For

          Transitive subjects always receive ERG from v⁰, regardless of case locus. This uniformity is the paper's key insight: the variation is in how objects are licensed, not subjects.

          Equations
          Instances For

            The object case assigner determines whether the object is licensed inside or outside vP.

            Equations
            • One or more equations did not get rendered due to their size.
            Instances For

              When the transitive object exits vP through the phase edge (Spec,vP), it occupies the single escape hatch. The subject, base-generated in Spec,VoiceP (below the vP phase boundary), cannot exit the phase domain because the escape hatch is occupied.

              This is the paper's core contribution: the ban on extracting transitive subjects follows from a locality problem in how case is assigned to objects, not from any property of the ergative subject itself.

              Equations
              Instances For

                Syntactic ergativity: the ban on A-bar extraction of transitive subjects. Predicted to occur iff the language is ABS=NOM (HIGH-ABS).

                Equations
                Instances For

                  Intransitive subjects are NEVER trapped, regardless of case locus: there is no transitive object to occupy the escape hatch. This correctly predicts that intransitive subjects extract freely in both HIGH-ABS and LOW-ABS languages.

                  Agent Focus Voice: a marked variant of Voice⁰ that assigns structural case to the internal argument (the transitive object).

                  Following @cite{ordonez-1995} on Popti', -on assigns case to the object. Unlike regular transitive Voice, AF Voice is NOT a phase head: its v⁰ is the intransitive variety (non-phasal), explaining why the intransitive status suffix -i surfaces rather than transitive -V'.

                  AF is a "last-resort" strategy, akin to English of-insertion: the marked variant of Voice⁰ is merged only when the normal derivation (with regular transitive Voice) would crash — i.e., when the subject must be A-bar extracted.

                  Equations
                  Instances For

                    AF Voice still introduces an external argument (the agent).

                    AF Voice circumvents the trapping mechanism because it assigns case to the object (so the object need not move to Spec,vP) AND because AF's v⁰ is non-phasal (so vP is not a locality domain). Either property alone would free the subject, but both hold simultaneously for AF Voice.

                    Equations
                    • One or more equations did not get rendered due to their size.
                    Instances For

                      The case-assignment property alone is what frees extraction: when Voice checks case, the object receives case inside vP and need not move to the escape hatch. This is the paper's primary explanation for why AF circumvents the extraction ban.

                      Equations
                      Instances For

                        The non-phasal status explains AF's morphology (intransitive status suffix -i) rather than the extraction facts. Since AF's v⁰ is intransitive, no ergative case is assigned to the subject.

                        In non-finite embedded clauses, Infl⁰ is absent (no aspect marking). This makes predictions depending on the case locus:

                        • ABS=NOM: transitive objects cannot be licensed (Infl⁰ absent). They require -on ("Crazy Antipassive") or detransitivization.
                        • ABS=DEF: transitive objects are licensed by v⁰ (present even without Infl⁰). They are fine in non-finite clauses.

                        Intransitive subjects lose absolutive case in BOTH types (Infl⁰ assigns NOM to intransitive subjects universally in Mayan).

                        Equations
                        Instances For

                          Intransitive subjects are unlicensed in non-finite clauses regardless of case locus, because Infl⁰ (the universal NOM assigner for intransitive S) is absent.

                          Equations
                          Instances For

                            Some objects do not require structural case: reflexive objects, extended reflexive objects, and bare (determinerless) NPs. These are licensed by pseudo-incorporation into the verb stem (§5.2).

                            Extended reflexives (possessor bound by subject, e.g., s-na? '3ERG-house') are formally identical to reflexives — the possessed nominal has lost its independent referential meaning.

                            Prediction: AF is impossible with caseless objects. Since AF exists precisely to assign case to the object, it is vacuous (and thus blocked as last-resort) when the object needs no case.

                            Instances For
                              Equations
                              • One or more equations did not get rendered due to their size.
                              Instances For

                                When the object is caseless, regular transitive Voice is used even for agent extraction: the object remains inside vP (no case-driven movement) and does not block the escape hatch. The subject is NOT trapped.

                                Equations
                                Instances For

                                  Despite sharing ergative morphology, Q'anjob'al and Chol differ in whether agent extraction is banned. The difference traces to their distinct case loci, not to properties of the ergative NP.

                                  The fragment ABSPosition values derive the correct syntactic ergativity predictions: Q'anjob'al has it, Chol does not. This grounds Tada's Generalization in fragment data rather than hard-coded table entries.

                                  Q'anjob'al's extraction data is consistent with the prediction: agent extraction is banned in regular transitives.

                                  Q'anjob'al's AF form carries the intransitive status suffix, matching the prediction that AF Voice is non-phasal (intransitive v⁰).

                                  The Crazy Antipassive form is identical to AF: same -on morpheme, same intransitive status suffix. Supports the unified analysis of -on as a case-assigner in environments where case is otherwise unavailable.

                                  Non-finite absolutive asymmetry in Chol: objects are available (v⁰ assigns case) but intransitive subjects are not (Infl⁰ absent). Follows from LOW-ABS: v⁰ handles objects, Infl⁰ handles intransitives.

                                  Tada's Generalization is now DERIVED from the case-assignment analysis rather than merely stated as a correlation. The observable parameter ABSPosition maps to CaseLocus, which determines whether syntactic ergativity arises.

                                  The morphological observation (ABSPosition) and the syntactic observation (extraction asymmetry) are connected through the case-assignment locus: for both values of the parameter, the predicted syntactic ergativity matches what Tada's table reports for the non-outlier languages.

                                  The three factors that combine to produce the ban on extracting transitive subjects in Q'anjob'al. All three are necessary; removing any one would free the subject.

                                  • vPIsPhasal : Bool

                                    I. Transitive vP is phasal (locality domain).

                                  • subjectBelowVP : Bool

                                    II. The transitive subject is generated below vP.

                                  • singleEscapeHatch : Bool

                                    III. Only a single specifier available for extraction out of vP.

                                  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

                                        Q'anjob'al instantiates all three factors.

                                        Equations
                                        Instances For
                                          theorem Phenomena.Ergativity.Studies.CoonMateoPedroPreminger2014.factor1_necessary :
                                          (have __src := qanjobalFactors; { vPIsPhasal := false, subjectBelowVP := __src.subjectBelowVP, singleEscapeHatch := __src.singleEscapeHatch }).producesTrapping = false

                                          Removing any single factor would free the subject.

                                          theorem Phenomena.Ergativity.Studies.CoonMateoPedroPreminger2014.factor2_necessary :
                                          (have __src := qanjobalFactors; { vPIsPhasal := __src.vPIsPhasal, subjectBelowVP := false, singleEscapeHatch := __src.singleEscapeHatch }).producesTrapping = false
                                          theorem Phenomena.Ergativity.Studies.CoonMateoPedroPreminger2014.factor3_necessary :
                                          (have __src := qanjobalFactors; { vPIsPhasal := __src.vPIsPhasal, subjectBelowVP := __src.subjectBelowVP, singleEscapeHatch := false }).producesTrapping = false

                                          AF removes factor I: AF Voice is not phasal. With a non-phasal vP, there is no locality boundary trapping the subject.

                                          The paper's table (10): morphological ergativity and syntactic ergativity are logically independent. Morphological ergativity is shared by all Mayan languages; syntactic ergativity (the extraction ban) arises only in ABS=NOM (HIGH-ABS) languages.

                                          +morph.erg-morph.erg
                                          +syntactic ergativityQ'anjob'alunattested
                                          -syntactic ergativityCholEnglish

                                          The [-morph,+syn] cell is predicted unattested: syntactic ergativity requires Infl⁰ to assign case to the object, which only arises in morphologically ergative systems.

                                          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

                                                Morphological ergativity is necessary but not sufficient for syntactic ergativity. Q'anjob'al and Chol are both morphologically ergative but differ in syntactic ergativity.

                                                The [-morph.erg, +syn.erg] cell is predicted unattested: syntactic ergativity requires HIGH-ABS (Infl⁰ licensing objects), which entails morphological ergativity.

                                                Equations
                                                Instances For

                                                  AF is restricted to 3rd person agents in Q'anjob'al (§5.1, ex. 72). The fragment's PersonRestriction captures this. The theoretical explanation: 1st/2nd person pronouns may be base-generated in Spec,CP (following Baker 2008), so they never need to extract through the vP phase edge — the trapping problem does not arise for them.

                                                  The Crazy Antipassive, by contrast, is NOT person-restricted: it applies in ALL non-finite embedded transitives regardless of the person of the subject, because the trigger there is the absence of Infl⁰ (not extraction through a phase edge).

                                                  The paper's §5.3 prediction: in HIGH-ABS languages, not only subjects but NOTHING generated inside a transitive vP (besides the object itself) should be able to escape. The object's movement to Spec,vP for case renders the single escape hatch occupied — trapping everything inside the phase domain.

                                                  This distinguishes the case-based account from ergative-property accounts: the latter predict only subjects are banned; the former predicts a general vP-internal extraction restriction.

                                                  Instances For
                                                    Equations
                                                    • One or more equations did not get rendered due to their size.
                                                    Instances For

                                                      Double-object constructions are systematically absent in HIGH-ABS languages (Q'anjob'al, Kaqchikel) but present in LOW-ABS Chol (via applicative). This is consistent with the general vP-internal ban.

                                                      Equations
                                                      Instances For

                                                        End-to-end argumentation chain for reflexive objects (§5.2):

                                                        1. Reflexive objects are caseless (pseudo-incorporated)
                                                        2. AF exists to assign case → AF is vacuous with caseless objects
                                                        3. With no case-driven movement, the object stays in situ
                                                        4. The escape hatch is unoccupied
                                                        5. The subject is free to extract using regular transitive Voice
                                                        6. Therefore: reflexive + agent extraction = regular transitive form

                                                        Extended reflexives (possessor bound by subject) pattern with reflexives: caseless, AF impossible, subject extracts freely. Examples (75a)–(76a) of @cite{coon-mateo-pedro-preminger-2014}.

                                                        The trapping mechanism is grounded in Phase theory: regular transitive Voice is a phase head (v*), so vP constitutes a locality domain under the PIC. AF Voice is NOT a phase head, so it does not create a phase boundary — the complement remains accessible.

                                                        This connects the Boolean phaseHead on VoiceHead to the Phase module's isPhaseHead.

                                                        Phase headedness partitions Voice into {trapping, non-trapping}: a Voice head traps the subject iff it is a phase head AND does not itself check case (freeing Infl⁰ to do so).

                                                        Equations
                                                        • One or more equations did not get rendered due to their size.
                                                        Instances For

                                                          The mapping preserves the assigner structure: ERG comes from v⁰, NOM comes from Infl⁰, and the object case depends on the parameter.

                                                          Different Mayan languages circumvent syntactic ergativity through different mechanisms (@cite{coon-mateo-pedro-preminger-2014}, §4.2, §5). Q'anjob'al uses case assignment: Voice_AF assigns case to the object, freeing the escape hatch. Kaqchikel uses an anti-locality repair: AF structure avoids the too-local Spec,TP → Spec,CP movement step (SSAL), at the cost of losing Set A agreement (@cite{erlewine-2016}).

                                                          Q'anjob'al AF Voice checks case; Kaqchikel's regular Voice does NOT. This is the core parametric difference: Q'anjob'al's AF is a case-assigning repair, while Kaqchikel's AF is a locality repair.

                                                          Both languages share the underlying problem: agentive Voice is a phase head, creating a locality boundary that traps the subject.

                                                          Kaqchikel AF loses Set A agreement (the agent never enters Spec,TP). Q'anjob'al AF also loses Set A agreement. Same surface morphological effect, different underlying mechanism.