Headedness and exocentric compounding

Semantic headedness typically serves as the primary criterion for compound endocentricity, i.e. whether a compound has a head. The semantic head is often defined as the hyperonym from which the denotation of the compound is derived, with exocentric compounds being those whose denotation is not a subclass of that of their head element. Headedness, so defined, leads us to analyze every non-compositional compound as exocentric. We explore the boundaries between semantic exocentricity and non-compositionality using established diagnostics in order to decide whether a semantic characterization of headedness is valid, and to determine whether exocentricity and non-compositionality coincide. Assuming a syntactic model of morphological combinatorics we show that exocentricity must be defined configurationally, occurring when the structure of a compound modifies an external entity, frequently instantiated by an empty noun. Hence exocentricity is not the absence of a head, but the realization of the compound's head outside its internal structure. Non-compositionality, in turn, derives from how the root of each constituent member of a compound is compositionally or idiosyncratically interpreted. Finally, we put forth a new typological distribution of exocentric compounds, discriminating real exocentric compounds (bahuvrihi and dvandva) from compounds that are commonly, but wrongly, defined as exocentric (e.g. deverbal and de-prepositional compounds).