The competition among relations in nominals (CARIN) theory of conceptual combination (C. L. Gagné & E. J. Shoben, 1997) proposes that people interpret nominal compounds by selecting a relation from a pool of competing alternatives and that relation availability is influenced by the frequency with which relations have been previously associated with the modifying concept. The current authors derived relation frequencies by using a sample of compounds occurring in the British National Corpus and compared them with those derived by Gagné and Shoben. The authors demonstrated that the original relation frequencies are unrepresentative and that Gagne' and Shoben's technique for dichotomizing them into high and low is unreliable. In addition, the authors revealed anomalies in the mathematical instantiation of the CARIN model and showed that it does not provide evidence for competition among relations.