How Understanding Novel Compounds is Facilitated by Priming from Similar, Known Compounds

We examine how the understanding of novel nounnoun compounds is facilitated by priming from similar, previously-encountered compounds. Similarity between a set of familiar primes and two sets of target phrases was defined using a computational model. This similarity classification was further validated with a rating study. Two reaction time experiments showed that people find it easier to understand target compounds similar to their primes than target compounds dissimilar to their primes, irrespectively of whether the compounds invite relational or property interpretations. Taken together, these findings have important implications: The use of an ‘objective’ similarity measure improves methodologically on previous work in the field. The empirical substantiation of the similarity metric provides us with some extra comfort for its use as a control tool for experimental studies and as a component of concept combination models. The picture of the similarity priming becomes more complete given the evidence collected for both major compound types (property and relational). The findings also suggest that those models providing mechanisms for producing property and relational interpretations utilizing previous knowledge are