The English noun-noun construct: a morphological and syntactic object

In English, debates about the boundary between morphology and syntax have often focussed on combinations of two nouns (NNs) in which the first modifies the second, e.g. ‘coffee cup’ or ‘silk shirt’. These are widely regarded as falling into two classes: morphological compound nouns on the one hand, and syntactic noun phrases on the other (e.g. Payne & Huddleston 2002: 449). But although various tests have been proposed by which the two types might potentially be distinguished from one another, the results of these tests do not converge, and their reliability has been questioned (e.g. Bauer 1998). Either the distinction between morphological and syntactic types is purely a matter of definition, and depends on the test that is chosen, or there is actually no clear boundary. This paper investigates one of the most widely accepted tests for phrasal status, namely the possibility of independent modification: in cases where either noun can be adjectivally modified independently of the other, proponents of this test take the NN to be syntactic (e.g. Payne & Huddleston ibid., Lieber & Štekauer 2009: 11). But what properties of a particular NN determine whether such modification is possible? The present study attempts to answer this question by examining a large database of constructions of the form [AdjN]N or N[AdjN] randomly extracted from the British National Corpus. It is shown that, except for a small number of cases where the second noun is appositive, the possibility of modifying the first noun (N1) independently of the second (N2) depends on whether there is a combination of adjective plus N1 that is lexicalised, institutionalised, or at least more frequent than the NN itself. In the case of N2, the possibility of independent modification seems to depend largely on the nature of N1. In nearly all N[AdjN] constructions, the first noun is either a proper noun, a noun with an incorporated numeral such as ‘one way’, a material noun such as ‘silk’, or another noun that occurs very frequently in attributive position compared to its occurrence elsewhere. Overall, the results suggest that, for a given NN, the probability of either noun being modified independently of the other depends largely on the relative frequencies with which the two nouns occur in various positions. If we accept that such modification distinguishes between objects usually viewed as compounds and those usually viewed as phrases, then a possible conclusion is that the distinction between morphological and syntactic objects is itself based on relative frequencies: as such, it is gradient and usage-based, and the lack of a clear boundary is expected