Adger (2025): Mereological Syntax: Phrase Structure, Cyclicity, and Islands #

Adger replaces set-theoretic Merge by Subjoin, which makes one syntactic object a proper part of another in one of two dimensions: the first subjunction to an object yields its 1-part, the extended-projection complement, the second its 2-part, the specifier (Parthood). Parthood is transitive within a dimension only, and Angular Locality lets a part subjoin to a target only through a single angle: it must be an n-part of some 1-part of the target (CanSubjoin). A derivation is a sequence of subjunctions licensed by Dimensionality, Angular Locality and Mereological Extension (Derives).

The book's list of consequences is verified on its own structures. The complement of a head cannot resubjoin to it (antilocality), and subjunction to an unattached object, to a specifier, or downward fails because the target must contain the mover, in lowering and sideward configurations in every derivation (parallel_blocked, sideward_blocked, lowering_blocked). A 2-part of a 2-part of an object in the target's extended projection subjoins, while a 1-part of that 2-part does not (angle_turns_once), so that a subjunction blocked from inside a specifier becomes possible once the mover has subjoined to that specifier (escape_hatch): of the objects inside a specifier, exactly its own 2-parts reach above it (edge). This is successive cyclicity without phases: who reaches the matrix C of guess who you said fell only after subjoining to the embedded C, and in every derivation it leaves the embedded clause only from its edge (said_fell, said_successive_cyclic).

Islands follow from the same mechanism with Dimensionality: a 2-part whose own 2-part slot is filled is an island in every derivation. In what you asked who bought, the embedded C has who as its 2-part, and what never reaches the matrix C (wh_island). Extraction from a nominal runs through D, so it needs D's 2-part: an indefinite leaves it free and who did you buy a statue of derives (statue_indefinite), while a definite fills it with Det and the wh-expression is trapped (statue_definite). Subjects are not islands as such: under T[uD] the wh-expression subjoins to D and on to C (subject_thetic). T[uFam] is checked by a [Fam]-bearing 2-part of T, which in English only a Det, Dem or possessor in D's 2-part supplies, so a topic subject traps the wh-expression, and once the wh-expression holds D's 2-part the subject can never be a topic (subject_topic, thetic_not_topic). These predictions match the judgments of the book's rows on extraction from clauses, nominals and subjects (clause_rows, nominal_rows). Wh-island violations, the WIRE effect, adjunct islands, *-linearization, and the cross-linguistic feature settings of chapters 5–7 are not formalized.

Implementation notes #

Each structure is the book's tree cut down to the objects its claim needs: the movement of the subject to T in (44) and (3), and the article inside the relational nominal of (32), are left out, and a filler of D is a single object. Every object with a 2-part keeps its 1-part, as Dimensionality requires.
A row's structure is read off its own columns: the type of nominal fixes the 2-part of D, by D-Interpretation for definites and by the subjunction of Dem or the possessor, and pied-piping makes P the mover. The wh-in-situ rows of (41)–(44) are read as covert subjunction to C, and the echo-only reading of (42b) is set aside.
The prediction of a row is Reaches: some licensed derivation makes the mover a part of C.

Adger (2025): Mereological Syntax: Phrase Structure, Cyclicity, and Islands #

Implementation notes #

References #

Consequences of Angular Locality #

Clausal derivations and wh-islands #

Nominal islands #

Subject islands #

Well-formedness #

The data #