Determining Pair-Merge

Chomsky (2001) proposes that the distinction between adjuncts and phrases in other grammatical relations follows from the way in which they are each introduced into structures.¹ In particular, he distinguishes between the basic mechanisms of set-Merge and pair-Merge (see the discussion of these in section 1). A fundamental question left unanswered in his work is how the Narrow Syntax (NS) component of the derivational system can determine that pair-Merge is required in the integration of adjuncts into structures, rather than the default set-Merge. This squib proposes that the answer to this question derives from the (obligatory) existence of a functional category, Mod, in the structure of adjuncts ([Mod [_YP "Adjunct"]]) that is parallel in nature to the (obligatory) functional categories in nominals and clauses.

1 Set-Merge and Pair-Merge in Chomsky 2001

Throughout much of the development of the Minimalist Program, the fundamental notion of Merge is understood to be set-Merge. Considering how this operation works, Chomsky (2001:6) states that "NS has one operation that comes 'free,' in that it is required in some form for any recursive system: the operation Merge, which takes two elements α, β already constructed and creates a new one consisting of the two; in the simplest case, {α, β}" For example, a nominal such as the book is understood to be represented as a set, {the, book}, that was constructed by the operation Merge (set-Merge) from the two previously independent elements the and book, which themselves might be replaced by sets in other examples. Chomsky (2001:6) notes that while the set {α, β} can be informally understood "as a 'projection' of some head of α or β" (i.e., in earlier terminology it would be labeled a DP), many notions of previous conceptions of phrase structure are no longer formally invoked, including the formal primary notion of projections. Thus, we may think of structures like {the, book} as similar to the phrase structure trees of previous theories, but with much of the information about nonterminal nodes either derived or eliminated from the theory, rather than stipulated as primitive.

To deal with an apparent counterexample to strict cyclicity, Chomsky (2001:18) later introduces the notion of pair-Merge, stating that

[f]or structure building, we have so far assumed only the free symmetrical operation Merge, yielding syntactic objects that are sets, all binary: call them simple. The relations that come "free" (contain, c-command, etc.) are defined on simple structures. But it is an empirical fact that there is also an asymmetric operation of adjunction, which takes two objects β and α and forms the ordered pair <α, β>, α adjoined to β. Set-Merge and pair-Merge are descendants of substitution and adjunction in earlier theories.

For example, in a nominal like the beautiful house, beautiful is adjoined to house, which all then undergoes set-Merge with the determiner the, just as house alone might be in an example without an adjunct. We thus have the structure {the, <beautiful, house>}, in which house retains the properties it would normally possess in nonadjoined structures, and beautiful can be seen informally as occupying what Chomsky (2001:18) calls "a separate plane." As Chomsky (2001:18-19) notes, adjunction does not change the properties, including c-command, of β, "[b]ut extension of c-command to the adjoined element α would be a new operation, to be avoided unless empirically motivated. Happily, the empirical evidence disconfirms the complication."

The empirical evidence that Chomsky is invoking here is the lack of Condition C effects under reconstruction in adjuncts in examples like (1) (Chomsky 2001:(11)).²

(1) [_wh Which [[picture [of Bill_i]] [that John_j liked]]] did he_j/*i buy t_wh?

Although Bill, in the complement of picture, cannot be coindexed with he, John, embedded within an adjunct, can. These facts do not seem compatible with each other under a traditional account. That is, Condition C predicts the former fact under obligatory reconstruction, but not the latter, if adjuncts participate in c-command relations normally. Chomsky's (2001) treatment of adjunction via pair-Merge, which does not create a simple structure and therefore leaves the adjunct exempt from normal...