Basic Relations and Stereotype Relations in the Semantics of Compound Nouns

This paper tests the hypothesis of Fanselow (1981) that the semantic relations in compound nouns are of two types: ‘basic’ and ‘stereotype’. It is shown that the probability of a compound falling into either of Fanselow’s proposed categories can be largely predicted using semantic and distributional properties of the constituent nouns, as well as the degree of lexicalisation of the compound as a whole. The so-called ‘basic’ relations, namely constitution, location, identity, resemblance and meronymy, are more likely in compounds that are not lexicalised, that have productive modifiers and/or semanticallyspecific heads, and whose constituents are perceived as representing concrete rather than abstract concepts. It is argued that such relations might be regarded as basic in several ways: they relate to states of physical entities, have a high level of generality and may be associated with semantic and phonological transparency.