Compounding theories come in four different flavors. Transformational theories claim that compounds are derived from relative clauses. Reductionist theories have proposed a limited number of primitive relations between the compounding constituents. But most contemporary theories are so-called slot-filler theories, in which the constituents are conceptualized as bundles of features, and the modifying constituent simply "adds" a feature to the other constituent. Finally, pragmatic compounding theories claim that there is only one relation between the constituents of a compound, but that its meaning is solely derived from pragmatic knowledge about the world.
I argue that all such theories are inadequate. More interestingly, I claim that slot-filler theories face serious challenges so as to account for the fact that modifying constituents most often add more than one feature to the other constituent. And also, that exocentric compounding seems to be productive in many languages.
Instead I propose a conceptual typology of nominal compounding. The typology is based on a distinction between literal, metonymical, and metaphorical relatedness. A two-constituent compound consists of at least one pointer and at most one modifier. The notion of head is misguiding, since it does not allow for gradualness between literal and metonymic relatedness, and since it does not distinguish between grammatical and conceptual percolation.
This typology enables one to describe important differences between languages. Further, it deals with a variety of cases that most traditional theories cannot account for, such as the occurrence of both "left-headed" and "right-headed" compounding in one language. Also, the typology accounts for "headless compounding." I introduce analogy bases as a vehicle for compounding. Irregular analogy bases produce "second-order" compounds that are not captured by the conceptual typology itself. Such compounds are likely to be metaphorical extensions of existing, more or less specific, constructions (analogy bases).
Finally, a promising slot-filler theory, formulated within the Generative Lexicon program, is examined in the light of the proposed typology, and it is seen that this theory is inadequate in many respects. Being a slot-filler theory, it faces the problems of multiple feature exchange and grammatical and conceptual exocentricity, but it also provides us with insights about how analogy bases specify for autonomous components. Interestingly, it seems compounding constructions specify for qualia roles, taxonomic levels, and other such features. I briefly sketch the architecture of a formal theory of compounding which is able to incorporate such specifications.
| Workshop Homepage | Workshop Program | Previous Abstract | Next Abstract |