Noun-noun combinations like "dog scarf" are common in everyday discourse but often have more than one interpretation. How do language users arrive at an interpretation of the relationship between the two nouns? This paper reports three expression-picture matching experiments that used a priming paradigm to investigate the influence of modifier and head constituents on the comprehension of novel ambiguous noun-noun combinations. Experiment 1 examined the effects of lexical repetition and semantic relation. We found reliable relation priming, regardless of whether the modifier or head was repeated between prime and target: Participants tended to choose target pictures involving the same relation as a preceding prime picture. Experiment 2 demonstrated significant relation priming when neither constituent was repeated. Experiment 3 showed significant relation priming when each picture contained both possible semantic relations, arguing against a possible visual-priming account of the effect. We interpret the findings in light of competing models of conceptual combination.