A realization assigns to an index r and a context c the finite set realize r c of forms
that realize r in c. The index is opaque, and the frameworks that separate an item from its
form choose it differently: a root of Distributed Morphology, given its form at Vocabulary
Insertion; one of Borer's roots, a phonological index whose form its categorial context selects,
as grow and growth; a lexeme of Paradigm Function Morphology, realized at a cell; Spencer's
lexemic index. An empty set says that r is not licensed in c, and a set of two forms is an
overabundant cell. An interpreted realization adds a second contextual map, to readings, which
is Marantz's separation of the forms of List 2 from the meanings of List 3.
A homomorphism of realizations merges indices and translates contexts so that realization is
preserved. The strict homomorphisms of interpreted realizations translate contexts independently
of the index and preserve readings too, so the indices they merge have the same forms and
readings in every context, and passing to these profiles, Interpreted.reduce, merges as much
as any strict homomorphism can. The disputes over individuation, such as fat roots against sister
roots, are mergers at the lax tier, which include forms and readings rather than preserving them.
Sign-based lexicalism, Construction Morphology and Haspelmath's morph-based comparative root
are rival ontologies that the parameters of a realization measure, not instances of it. The
interface does not say whether indices are listed or constructed.
A strict homomorphism cannot consult the index when it translates contexts, so categorial
information has to sit in the context, which quietly encodes the Categorization Assumption.
Extensional profiles underdetermine individuation. The criteria the literature uses, such as
locality of first categorization, are further structure on admissible mergers that studies
supply.
A root realization realizes an opaque index type in context. The fiber
realize r c is empty where r is unlicensed and non-singleton at an
overabundant cell.
realize r c is the set of forms realizing the index r in the context c, computed by Vocabulary Insertion (DM), the paradigm function (PFM) or spellout (nanosyntax).
IsSuppletive r says r has two distinct forms, in contexts that need not
differ, as √GO has go and went, the case behind Harley 2014's argument
that indices, not forms, individuate. When the forms are competing Vocabulary
Items this is suppletive allomorphy, of a root or of an affix alike
(Gouskova & Bobaljik 2020); when they are stems it is variance in Corbett 2007's
lexical material, which is suppletion in Stump 2016's sense where no inflection
class determines it (IsClassDetermined).
An index is overabundant when some cell offers two forms, as dived and dove
share one cell (Thornton 2011's cell-mates). This is cell-internal variance,
disjoint in kind from suppletion.
IsProperlySuppletive r says two licensed contexts of r are realized by
different fibers, as √GO is by go and went. An unlicensed cell never
witnesses it.
The forms of r are class-determined by the patterns A, the paradigms of stems that the
inflection classes of a language generate, when they are drawn from one pattern, so that they
alternate as kindred stems; an alternation between stems no pattern determines is suppletive
(Stump 2016).
{R : Type u_1}{Ctx : Type u_2}{F : Type u_3}(S : RealizationRCtxF)(A : Set(Ctx → F))(r : R) :
The transpose of S realizes each context at each index as S realizes the index in the
context, so that comparing one context across indices is comparing one index of the transpose
across contexts.
The pipeline composition S.comp T realizes through an intermediate
inventory, the Kleisli composition of Finset over a shared context.
Late-insertion architectures factor the grammar's realization as such a
composite; PFM's stem-choice-then-blocks cascade is another.
A hom is an index merger that tracks spellout, in which onRoot may merge
indices and onCtx translates contexts and may consult the source index. It is
the transport tier; individuation disputes are adjudicated with
Realization.Interpreted.Hom, whose root-independent context translation blocks
re-encoding the source index in the target context.
A strict hom, the tier on which individuation disputes are stated,
translates contexts independently of the root, so that the target context
cannot re-encode the source index, and preserves interpretation alongside
spellout.
At the lax tier, a hom includes realization and interpretation rather
than matching them. Where a strict Hom merger asserts identity, a lax merger
asserts subsumption, so that each source index's forms and readings are among
its image's, as when a lexeme's listed properties are among its root's
Encyclopedia entry (Arad 2005). Pattern-bound lexemes lax-merge into a total
root without ever strict-merging.
The reduced presentation takes its indices to be
realization-and-interpretation profiles. Every system presents its reduction;
the strict hom tier is change of presentation, and only the lax tier merges
beyond profiles.
Mergers along any strict hom refine profile equality, so reduce merges
as much as the strict tier ever can. Mergers beyond profiles, such as the
hammer carvings or pattern-bound lexemes into an Arad 2005 root, are
necessarily lax.