Elkins, Torrence and Brown (2026): Wh-movement paths and adjunct extraction in Mam #

The movement enclitic =(y)a' of San Juan Ostuncalco Mam optionally appears on the predicate, and on any directional auxiliary, when an instrument, benefactive, dative, locative, reason, purpose or manner adjunct is extracted, but not with absolutive or ergative arguments, Aissen's Ergative Extraction Constraint sending the agent through an antipassive, and not with temporals. The enclitic may occur once per Voice⁰ and Dir⁰ of a clause and, in long-distance extraction, once per clause along the dependency: in the embedded clause exactly when that clause is at least VoiceP-sized, so on both predicates over a full CP or an aspectless VoiceP complement, on the matrix predicate only over a nonfinite VP complement, and on the embedded predicate only in an embedded question. Elkins, Torrence and Brown analyse the enclitic as the spellout of Ā-agreement between the extracted adjunct and the [Ā]-bearing heads Voice⁰ and Dir⁰ it passes through under the Relativized-Minimality successive cyclicity of Rizzi and Abels: Attract Closest forces the mover through the specifier of each intervening [Ā]-bearer, each Voice⁰ or Dir⁰ copies the [obl] Case feature the mover receives from its relational noun, and the bundle [Ā, obl] is realized as =(y)a' or as ∅. Mendes and Ranero's rival Chain Reduction via Substitution for the K'ichean fronting particle wi, which spells out lower copies at the base position and at each Spec,CP stopover and so derives the Fronting Particle Generalization, predicts a single reflex within a clause, none in the matrix clause over an aspectless complement, and one inside a nonfinite complement, all contrary to the Mam data; Keine and Zeijlstra's DP-intervention leapfrogging predicts at most one reflex per intervening argument, contrary to a clause with three; and an Agent-Focus-like analysis fails because the enclitic co-occurs with the passive and appears in every clause of the path while the antipassive is confined to the clause of origin. The two systems also differ in which adjuncts trigger the reflex: reasons, purposes and manners do in Mam but are high adjuncts without [appl] in K'ichean, and temporals trigger neither.

Implementation notes #

Clause sizes are ClauseSpines, the Mam finite, aspectless and nonfinite clauses of (8) and §3.3–3.4, with the directional head Head.dir, a Mayan-specific category above VoiceP, spliced in by spine. A dependency is the list of clauses the mover crosses, bottom-up; its path is the [Ā]-bearers of those clauses and its sites the Voice and Dir heads among them, so the reflex in a clause is decided by whether the clause projects Voice (licensed_iff_mem_voice). The rival mechanisms are predicates on the same dependencies, and the rows discriminate them.
The Agree-and-insertion step is the substrate's applyAgree and subsetPrinciple over the valued features of a Voice head with an unvalued [oblique] probe and the vocabulary item (46a); the null exponent (46b) is the independent optionality of each site, patterns.
[obl] on the movers is the article's featural hypothesis (§1.3, §4.2, §5.3): relational nouns assign it, the locative and manner wh-words are assumed to acquire it, temporals lack it.
The examples are Data.Examples.ElkinsTorrenceBrown2026; the K'ichean rows are Mendes and Ranero's as reported there, and the article's (64), headed K'iche' in the preprint, is Patzún Kaqchikel (Mendes and Ranero's (14a), as the article's own (68) records). The variety is SJO Mam; Scott's San Juan Atitán Mam is a distinct variety.

Elkins, Torrence and Brown (2026): Wh-movement paths and adjunct extraction in Mam #

Implementation notes #

References #

The extended verbal domain (§1.3) #

The movement path (§4.1–4.3) #

Multiple exponence (§3.1, §5.2) #

The rival mechanisms (§3.6, §5.1) #

Which movers trigger the reflex (§2, §5.3) #

Agree and insertion (§4.2) #

The rows (§2, §3, §5) #