Roots and Lexicality in Distributed Morphology

This paper examines the nature and content of morphological roots in relation to their syntactic context. A careful consideration of doublets, where the same root may take alternative noun - inherent features, leads to the claim that roots do not carry selectional features or class diacritics. Relying on the distinction between syntactic nodes and their exponents, central to a realizational model like Distributed Morphology, I argue that the syntactic atoms corresponding to root nodes are associated with open - class exponents but not with a specific meaning that might select a licensing syntactic context. "Lexical" meaning arises constructionally, and so do lexical properties like gender or class, which however emerge at Vocabulary insertion and may show selectional properties. Content and exponence of roots are thus dissociated, in line with the separationist character of Distributed Morphology. This predicts the existence of root - like elements with mixed status, namely open - class exponents used as grammatical morphemes (like auxiliaries or classifiers), or category - free root extensions below the innermost category - assigning head (like de- in de-struction).