If Olive Oil Is Made of Olives, then What’s Baby Oil Made of?

Abstract

Noun+noun constructions (NNs) are becoming increasingly frequent in the English language. From a semantic perspective, there is a wide range of meaning relationships that are possible between the two nouns. Recent research on NNs has focused broadly on variation in the use of NNs, entirely ignoring variation and diachronic change in the use of these different semantic categories. Moreover, there has been no research on the psycholinguistic reality of these semantic categories. In this study we attempt to fill these gaps by having native English speakers classify a large set of highly frequent NNs into semantic categories and then describing diachronic variation in NN use across those categories. The 400-million-word Corpus of Historical American English (COHA) was used to generate a list of the 1,535 most frequent NNs. We developed a list of 12 major semantic categories for NNs and created a classification instrument that was used by Mechanical Turk workers (N = 255) to classify each of the 1,535 NNs. We then used COHA to produce diachronic frequency data for each of 12 semantic categories. We also show that changes in the use of NNs over time varies dramatically according to the meaning relationship between two nouns. This study, which begins with a corpus-based frequency list, relies heavily on user-based classification data, and culminates with diachronic corpus data, demonstrates the power of triangulating use-based and user-based linguistic data.