Semantics of compounds

The semantics of English modifier + head nominal composites — but especially of N + N compounds — is a recurrent topic on this blog; the array of semantic relationships exemplified in the data here is enormous, and might give the impression that things are just chaotic, though I’ve tried to pull out frequent patterns that dominate the data. One way to approach the matter in more nuanced fashion is to search for preferences for certain kinds of interpretations according to the semantics of the component elements.

And now, just appeared, we have “Systematicity in the semantics of noun compounds: The role of artifacts vs. natural kinds” by Beth Levin, Lelia Montague Glass, and Dan Jurafsky, in the De Gruyter journal Linguistics. Published online 5/16/19; I’ve found no volume, issue, and page numbers for the print version, but this is the DOI, and Lelia now reports that a pdf is freely available here. The abstract:

The nature of the relationship between the head and modifier in English noun compounds has long posed a challenge to semantic theories. We argue that the type of head-modifier relation in an English endocentric noun-headed compound depends on how its referent is categorized: specifically, on whether the referent is conceptualized as an artifact, made by humans for a purpose; or as a natural kind, existing independently of humans. We propose the Events vs. Essences Hypothesis: the modifier in an artifact-headed compound [like “water carafe”] typically refers to an event of use or creation associated with that artifact, while the modifier in a natural kind-headed compound [like “water spinach”] typically makes reference to inherent properties reflective of an abstract essence associated with the kind, such as its perceptual properties or native habitat.

(#1) The water spinach plant, Ipomoea aquatica (yes, in the same genus as morning glories)

(#2) Stir-fried water spinach

We present three studies substantiating this hypothesis. First, in a corpus of almost 1,700 attested compounds in two conceptual domains (food/cooking and precious minerals/jewelry), we find that as predicted, compound names referring to artifacts tend to evoke events, whereas compound names referring to natural kinds tend to evoke essential properties. Next, in a production experiment involving compound creation and a comprehension experiment involving compound interpretation, we find that the same tendencies also extend to novel compounds.

(Local plug: Levin and Jurafsky are professors of linguistics at Stanford, and Glass is a recent Stanford PhD in linguistics (now teaching at Georgia Tech).)

Leave a Reply

%d bloggers like this: