Documentation

Linglib.Theories.Syntax.Minimalism.Core.Voice

Voice Head Flavors #

@cite{chomsky-2001} @cite{coon-2019} @cite{cuervo-2003} @cite{harley-2014} @cite{kratzer-1996} @cite{legate-2003} @cite{schaefer-2008} @cite{wood-2015}

Voice heads introduce (or fail to introduce) external arguments. The key typology from @cite{schaefer-2008}:

Key Claim #

Non-thematic Voice contributes no semantics. SE is a PF marker of absent external argument, not a semantic operator. Evidence: SE is optional in Chilean Spanish when Fission produces a syncretic clitic.

Voice–VerbHead Bridge (§§ 5–6) #

Voice and VerbHead are both "little-v" but encode different dimensions: Voice determines whether an external argument is introduced; VerbHead decomposes the event structure into subevents.

Following @cite{cuervo-2003} and @cite{pylkkanen-2008}, the CAUSE relation is modeled here as an independent VerbHead in the root's event decomposition — present in both causative and anticausative alternants. Voice contributes vDO (the agent's activity subevent) when it assigns a θ-role, but does NOT contribute the causal relation.

Note: @cite{wood-2015} uses a SINGLE v head whose interpretation introduces CAUSE, rather than @cite{cuervo-2003}'s multi-headed decomposition. The VerbHead decomposition used here is a linglib modeling choice that captures the same Voice–CAUSE independence using @cite{cuervo-2003}'s notation. See Fragments.Icelandic.Predicates for the Icelandic fragment.

Typology of Voice head flavors.

Agentive Voice introduces an agent; causer Voice introduces a causer; non-thematic Voice has no semantics (anticausative SE); expletive Voice has neither specifier nor semantics (middles); passive Voice checks Case without assigning θ (@cite{collins-2005}: by); reflexive Voice introduces agent that binds internal argument (@cite{wood-2015}); experiencer Voice introduces experiencer external argument (@cite{wood-2015}).

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

      A Voice head with its properties.

      • flavor : VoiceFlavor

        The flavor determining argument introduction and semantics

      • hasD : Bool

        [D] subcategorization feature: requires a specifier at PF

      • phaseHead : Bool

        Is this Voice head a phase head? (v* = agentive Voice)

      • checksCase : Bool

        Does this Voice head check Case? In active, v checks accusative; in passive, Voice/by checks it (@cite{collins-2005}, p. 96: feature dissociation). Default false — only passive Voice checks Case.

      • features : FeatureBundle

        Agree-relevant features on Voice (e.g., [uOblique] for Mam =(y)a'). Default empty — most Voice heads carry no probe features.

      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

              Does this Voice head introduce a θ-role?

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

                Does this Voice head have semantic content?

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

                  Agentive Voice (transitive/unergative): introduces agent, is a phase head.

                  Equations
                  Instances For

                    Causer Voice: introduces causer, is a phase head.

                    Equations
                    Instances For

                      Non-thematic Voice (anticausative): no θ-role, [D] for PF marking.

                      Equations
                      Instances For

                        Expletive Voice (middle): no specifier, no semantics.

                        Equations
                        Instances For

                          Impersonal Voice (Finnish "passive"): demotes agent to an implicit generic human referent. Has semantics (existential closure over agent) but does not assign a θ-role to a syntactic specifier.

                          Equations
                          Instances For

                            Passive Voice: headed by by, checks Case but does not assign a θ-role — v assigns the θ-role to the external argument in Spec,vP. Passive v is NOT a phase head: the Case-checking feature that makes v* a strong phase head has been dissociated onto Voice/by.

                            This is why PartP (complement of v) remains accessible for smuggling: passive v is a defective v, not v*. Cf. @cite{chomsky-2001}: "only v* (transitive) is a strong phase."

                            Contested: @cite{legate-2003} argues passive v IS a phase head based on reconstruction and parasitic gap data. The current formalization follows @cite{collins-2005} and @cite{chomsky-2001}.

                            Equations
                            Instances For

                              Reflexive Voice (@cite{wood-2015}): introduces agent that is coreferent with the internal argument. [+θ, +D], phase head (assigns θ). Icelandic -st spells out this Voice in reflexive constructions.

                              Equations
                              Instances For

                                Experiencer Voice (@cite{wood-2015}): introduces experiencer external argument. [+θ, +D], phase head. Icelandic subject-experiencer -st verbs (e.g., leiðast 'be bored').

                                Equations
                                Instances For

                                  Agentive Voice assigns a θ-role.

                                  Non-thematic Voice does NOT assign a θ-role (Muñoz Pérez's key claim).

                                  Non-thematic Voice has no semantic contribution. This is the core claim of Muñoz @cite{munoz-perez-2026}: SE is a PF phenomenon.

                                  Agentive Voice is a phase head (v* = Voice_AG).

                                  Non-thematic Voice is NOT a phase head.

                                  Impersonal Voice does NOT assign a θ-role (agent is existentially closed, not projected to a syntactic specifier).

                                  Impersonal Voice HAS semantics: it contributes an existential closure over the agent variable, unlike non-thematic Voice which is vacuous.

                                  Passive Voice does NOT assign a θ-role (v does).

                                  Passive Voice IS NOT a phase head.

                                  Passive Voice HAS semantic content (by mediates Case-checking).

                                  Passive Voice checks Case (@cite{collins-2005}, p. 96: feature dissociation).

                                  Reflexive Voice assigns a θ-role (@cite{wood-2015}).

                                  Experiencer Voice assigns a θ-role (@cite{wood-2015}).

                                  Build the full verbal decomposition by combining Voice's contribution with the root-determined lower event structure.

                                  Voice contributes vDO when it assigns a θ-role (agentive, causer, antipassive). The root supplies the lower structure, which now includes vCAUSE for change-of-state roots (@cite{wood-2015}, @cite{pylkkanen-2008}). CAUSE is independent of Voice:

                                  • Voice_AG + [vCAUSE, vGO, vBE] → [vDO, vCAUSE, vGO, vBE] (causative)
                                  • Voice_nonTh + [vCAUSE, vGO, vBE] → [vCAUSE, vGO, vBE] (anticausative)
                                  • Voice_AG + [] → [vDO] (unergative activity)
                                  • Voice_nonTh + [vBE] → [vBE] (stative)
                                  Equations
                                  Instances For

                                    θ-assigning Voice prepends vDO to the root structure.

                                    Non-θ Voice leaves the root structure unchanged.

                                    Causative pattern: agentive Voice + [vCAUSE, vGO, vBE] yields a causative decomposition.

                                    Activity pattern: agentive Voice + [] yields an activity.

                                    State pattern: non-thematic Voice + [vBE] yields a state.

                                    The causative alternation: same root structure [vCAUSE, vGO, vBE] is causative under agentive Voice but inchoative under non-thematic Voice. CAUSE is shared across both alternants — only vDO (from Voice) differs. This formalizes the @cite{wood-2015}/@cite{pylkkanen-2008} insight: CAUSE is independent of Voice.

                                    Voice determines causativity: if the root structure is [vCAUSE, vGO, vBE], then whether the result is causative tracks exactly whether Voice assigns a θ-role.

                                    CAUSE is present in both causative and anticausative decompositions. This is the independence claim: CAUSE is part of the root, not Voice.

                                    In active, v (= agentive Voice) assigns θ AND controls Case-checking (Case is checked by v, not by Voice). In passive, these functions dissociate: v assigns θ (external argument in Spec,vP), while Voice/by checks Case.

                                    Passive: θ-assignment and Case-checking are dissociated. Voice does NOT assign θ (v does), but Voice DOES check Case.

                                    UTAH compliance: the external argument is structurally present (hasD = true) in BOTH active and passive. The external argument occupies the same position (Spec,vP) regardless of voice — satisfying the Uniformity of Theta Assignment Hypothesis.

                                    Passive Voice does not prepend vDO: it does not assign θ, so buildDecomposition passes the root structure through unchanged.

                                    Agentive Voice corresponds to traditional v* (phase head). In the @cite{kratzer-1996}/Schäfer framework, agentive Voice replaces v*. Both agentive and causer Voice are phase heads.

                                    Non-thematic and expletive Voice are NOT phase heads. Only θ-role-assigning Voice heads (agentive, causer) are phases.

                                    Phase-head-ness correlates with θ-role assignment: Voice is a phase head iff it assigns a θ-role.

                                    How a Voice head introduces (or fails to introduce) an external argument semantically.

                                    • thematicArgument: [+λx] — introduces agent/causer via λ-abstraction; the external argument occupies Spec,VoiceP
                                    • thematicExistential: [+∃x] — introduces agent/causer via ∃-binding; agent is semantically present but syntactically implicit (@cite{schaefer-2017} (31b)/(31e): "medio-passive Voice" {λe∃x[agent(e,x)]})
                                    • expletive: [−λx] — no semantic contribution; Voice is semantically vacuous (anticausative SE, middles)
                                    Instances For
                                      Equations
                                      • One or more equations did not get rendered due to their size.
                                      Instances For

                                        The ±D / ±λx parametric decomposition of Voice heads.

                                        From @cite{alexiadou-schaefer-2015} (p. 109, ex. (12)), extended by @cite{schaefer-2017}. Two binary parameters generate the core cross-linguistic typology of Voice:

                                        none values represent underspecification: the morpheme is compatible with multiple parameter settings, with the actual setting determined by independent factors (argument realization, verb class, pragmatics). Indonesian ber- is fully underspecified ⟨none, none⟩ (@cite{beavers-udayana-2022}); Spanish se is underspecified for ±D.

                                        • selectsSpecifier : Option Bool

                                          Does Voice select a syntactic specifier (DP)? some true = [+D], some false = [−D], none = underspecified

                                        • extArgSemantics : Option ExternalArgSemantics

                                          Does Voice introduce semantic agentivity/causation? none = underspecified

                                        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

                                                Map each named VoiceFlavor to its position in the ±D / ±λx parameter space.

                                                Flavor±D±λxExample
                                                agentive+D+λx (arg)English active
                                                causer+D+λx (arg)Psych causative
                                                reflexive+D+λx (arg)Icelandic -st reflexive
                                                experiencer+D+λx (arg)Icelandic subject-exp -st
                                                nonThematic+D−λxRomance anticausative SE
                                                expletive−D−λxEnglish dispositional middle
                                                impersonal−D+∃xFinnish impersonal
                                                passive+D−λxEnglish passive (by)

                                                Note: nonThematic and passive occupy the same cell [+D, −λx]. They differ in Case-checking (VoiceHead.checksCase), which is a property of the full VoiceHead, not of the parametric decomposition.

                                                Equations
                                                Instances For

                                                  The parametric decomposition of a VoiceHead, derived from its flavor.

                                                  Equations
                                                  Instances For

                                                    Does this parameter setting assign a theta role? Returns none when underspecified.

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

                                                      Are two VoiceParams settings compatible? Two settings are compatible if they agree on all specified dimensions. An underspecified dimension (none) is compatible with anything.

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

                                                        Is this parameter setting fully specified (no underspecification)?

                                                        Equations
                                                        Instances For

                                                          All named VoiceFlavors produce fully specified params.

                                                          Compatibility is reflexive.

                                                          theorem Minimalism.underspecified_compatible_with_all (f : VoiceFlavor) :
                                                          have ber := { selectsSpecifier := none, extArgSemantics := none }; ber.isCompatibleWith f.toParams = true

                                                          A fully underspecified VoiceParams is compatible with every named VoiceFlavor — the key property for Indonesian ber-.