In defence of Autonomous Syntax: Toward a formal theory of features at the syntax-semantics interface
With the rise of the Minimalist Program, the focus of syntactic investigation has shifted toward features and their bundles. In the very core of our theorizing, we assume that a larger structure is represented by a label of its maximal projection or a phase (e.g., Chomsky 2013, 2015) but we do not have a good theory of what features form a label and what happens if there is more than one feature of the same type present in the search domain of a label. This talk directly addresses the question of labeling by investigating narrow-syntax features at the syntax-semantics interface. I entertain the idea that narrow-syntax features are computed by the syntax-semantics interface in a manner parallel to the computations of overt syntax features at the syntax-morphology interface. The empirical core of the talk concerns syntactic properties of associative constructions in Czech. As we will see, these constructions exhibit unusual correlations of four grammatical properties that do not form an obvious class, namely, variable agreement in the domain that otherwise does not allow a variable agreement, an obviation of biding Condition A in an environment that otherwise displays strict Condition A properties, blocked wh-movement from a domain that otherwise allow wh-extraction, and Person-Case Constraint (PCC) violation in the domain that otherwise does not display sensitivity to PCC. I argue that these unexpected correlations arise via a structural relationship with the hypothesized interface-bundle formation at Spell-Out. The proposal not only provides a principled explanation of unexpected correlations discussed in the paper, but it also restricts when and where in the derivation such correlations arise. The talk thus has two goals: (a) to provide an empirical argument for the syntax-semantics interface bundle formation and (b) to build a corresponding fragment of a grammar as a proof of concept.