Impacto sobre la biodiversidad y zonas de vida

Faced with the common problems in the foregoing possibilities of root-level semantics, in this section I consider a third possibility: perhaps roots do not participate in model-theoretic com- position at all. This brings us back to a basic idea in DM, as Marantz states below:

[I]t is not clear that the computational system of language … must know whether a node contains “dog” or “cat.” Distributed Morphology takes the position that this difference between “dog” and “cat” is a matter of Encyclopedic knowledge and that such knowledge is used in semantic interpretation of LF, but not in grammatical computations over LF or involving LF. (Marantz 1995: 4)

The key message here is the separation of compositional/logical and noncompositional/idiosyn- cratic meanings at the syntax-semantics interface: roots by themselves are invisible not only to feature-driven syntactic operations like Agree and Move but also to syntactically based semantic composition. As a result, the semantic effect of a root is only manifested when compositional and noncompositional meanings are integrated, which is arguably not a model-theoretic process. Two consequences follow from this conception: (i) the categorizer must denote a full-fledged function with a typed variable;31(ii) the denotation of a categorizer-root merger is identical to that of its categorizer part. Thus, the composition of blackboard proceeds as in (31).

30_{This information is nontrivial, for ∗ is not the only kind. Other kinds include ∗ → ∗, ∗ → ∗ → ∗, etc., which are}

inhabited by various function types. The question mark here can be viewed as an arbitrary type constant similar to the arbitrary term constants in predicate logic (see Partee et al. 1993: 153).

31_{Here I am assuming the categorizer introduces both the variable and its type, but it could also just introduce a type,}

which is then fed into a variable-introducing function (as in Acquaviva’s 2019 model to be introduced below). Such technical variation is allowed under possibility III as long as root content is excluded from the computation.

(31) N_{[𝜆𝑥∶u . Entity(𝑥)]} N_{[𝜆𝑥∶u . Entity(𝑥)]} √board N_{⟨cat: n, pos: 1⟩} [𝜆𝑥∶u . Entity(𝑥)] Cat_[?] black Cat_{⟨cat: , pos: ⟩}

[?]

Agree

Here the compositional meaning of blackboard is merely that it denotes a particular entity cat- egory. And it is the entire SO {{Cat, black}, {N, √board}} that gets mapped to the lexicalized meaning ‘blackboard’. That is, the defective category schema merely provides a categorial skele- ton for the lexicalization of the compound but does not encode the lexicalized result. From this perspective, the conventional practice of incorporating root (or any idiosyncratic) content in model-theoretic denotations (e.g., 𝜆𝑥 . Board(𝑥), ∃𝑒[Run(𝑒)]) is at best expository and at worst completely wrong. This is reminiscent of Chomsky’s objection to the referentialist approach to natural language semantics:

[A] lexical item provides us with a certain range of perspectives for viewing what we take to be the things in the world, or what we conceive in other ways; these items are like filters or lenses, providing ways of looking at things and thinking about the products of our minds. The terms themselves do not refer …. (Chomsky 2000: 36) Chomsky’s LIs have phonological, semantic, and syntactic features all packed in, but in the above context he is mainly concerned with the semantic—or more exactly encyclopedic—dimension; namely, the dimension that brings out, for example, the “house-home difference” (ibid.). This type of information is precisely what I have attributed to the root. Hence, Chomsky’s view that LIs do not refer and my conclusion that roots do not denote convey basically the same idea. Acquaviva expresses a similar attitude regarding DM roots:

[A]n extensionalist semantic approach, where basic terms of the semantic represen- tation are ultimately defined by what they are true of, in one or more than one world … cannot possibly shed much light on those aspects of lexical semantic competence based on oppositions in conceptualization rather than in distinct extensions: con- sider again home vs. house, or broad vs. wide, or use vs. utilize, to say nothing about notorious problematic cases like time, air, or god. (Acquaviva 2014a: 281)

In a later study of nominal structure, Acquaviva (2019) explicitly denies the root any type and only introduces the first-order predicate conventionally associated with a (bare common) noun via a functional head PΣ (property of sum) merged above the nominalizer, as in (32).

(32) Syntactic object Semantic type Description

root - purely differential content

n-root e introduces an entity type

PΣ-n-root ⟨e, t⟩ introduces variable, creates lattice

(based on Acquaviva 2019: 51) In this model the nominalizer introduces an entity type which is then named by a root, and the nominalized root becomes the smallest nominal structure conceived as “an unanalyzable name, a label maximally underdetermined except for the fact of being formally distinct from other names” (Acquaviva 2019: 45).32 In other words, the root is just a modifier or name tag for the categorizer/type, though importantly it is not a modifier in the model-theoretic sense (i.e., not a predicate). As Acquaviva (2009: 4) points out, the meaning of a root is “too elusive to be pinned down,” for “something so radically underspecified cannot even convey the distinction between argument and predicate.” Rather, the semantic effect of this root-to-type modification probably only takes place in the final integration of compositional and noncompositional meanings, where various noncompositional ingredients modify the compositional structures (i.e., SOs) they are part of. This sounds like Marantz’s (1995) view just above (31), and Marantz in a later article indeed makes a remark similar to that of Acquaviva:

[T]he idiosyncrasies in use of verbal roots must be separated from the general, non- idiosyncratic connections between structure and meaning …. The little v semanti- cally introduces an eventuality, … [and] roots may modify the event introduced by the little v. (Marantz 2013: 154, 157)

So Marantz also lets the categorizer declare a type and lets the root modify that type. Interest- ingly, both Acquaviva’s proposal for the nominal domain and Marantz’s proposal for the verbal domain resemble an eclectic combination of our possibility II (where categorizers denote types) and possibility III (where roots do not denote).