This article discusses the notion of attributive compounds, aiming to shed a light on what should be called an attributive compound and how the structure and meaning of such phenomena should be characterized in linguistic and psycholinguistic theories. Controversies surround even the most basic terminology that is used to describe compounds and their classification; hence, any commentary toward a precise analysis is likely to be incompatible with somebody’s theoretical commitments. However, let us begin our account by considering some examples of what we might want to refer to as a canonical attributive compound.
(1) a. tennis ball (attributive compound)
(1) b. soda can (attributive compound)
(1) c. bookcase (attributive compound)
A first description of the attributive compounds in (1) is that the second part is the head whereas the first part (the non-head) acts as some kind of a modifier, yielding an interpretation along the lines of ‘a ball of the tennis type’, ‘a can of the soda type’, or ‘a case of the book type’, respectively. The three compounds shown above are fairly typical for English attributive compounds: First, they are all endocentric — they have a “center”, which is the head of the compound (Bloomfield 1933). The compound in each is a hyponym of the type; for example, a tennis ball is a type of a ball. Therefore, the head of the compound tennis
ball is ball. Second, the head is to the right: if tennis ball were a type of tennis, that would suggest an endocentric left-headed compound. We will for the most part focus on compounds like these, even though right-headedness is not universal and not all compounds are endocentric — some are exocentric. We briefly discuss these matters in Section 3.1 of the article. Two roots have an interpretive effect and so does a covert relation between them, such as ‘for’ in something like ‘ball for tennis’. This description, of course, needs to be elaborated, but let us first acknowledge that the characterization of attributive compounds is a theoretical classification task which most prominently involves drawing a useful and empirically meaningful contrast between attributive compounds and (at least) the other main types of compounds in natural language, synthetic compounds and coordinative compounds.
Drawing on previous work (Allen 1978; Levi 1978), attributive compounds are characterized by a covert relation R. In contrast, in
a synthetic compound, the non-head root saturates an argument requirement associated with the head root whereas the roots of a coordinative compound are combined in the semantics by Predicate Conjunction (Heim and Kratzer 1998). Semantically, we review important issues from previous work on the meaning on R, but ultimately we develop our own analysis that assumes a limited number of relations in the spirit of Levi’s work, while attributing pragmatic context dependence to the independently motivated mechanism of Contextual Domain Restriction.
attributive compounds, compounding, Contextual Domain Restriction, dissociated morphemes, Distributed Morphology, endocentricity, exocentricity, modification