Two lexical decision experiments were conducted to further investigate the notion that metaphor comprehension involves the formation of a new association between topic and vehicle. Experiment 1 was essentially a replication and extension of Camac and Glucksberg (1984) demonstrating that known word associate pairs show a significant lexical decision latency advantage over their randomly paired counterparts, while topic/vehicle word pairs drawn from apt metaphors do not. The results of Experiment 1 confirm their initial findings even when printed word frequency of the two pair types is held equivalent (a factor not controlled for in the original Camac and Glucksberg study). This result suggests that preexisting topic/vehicle similarity is not an important factor in metaphor comprehension. Experiment 2 was an attempt to detect the hypothesized shift in attribute salience that results in the formation of a new association between topic and vehicle during metaphor comprehension. In Experiment 2, subjects made lexical decisions on topic/vehicle word pairs that were preceded by a paragraph designed to induce either a metaphorical or literal interpretation. For many subjects, a latency advantage was observed for topic/vehicle pairs preceded by a metaphorical context as compared to a mismatching literal context. This finding suggests that metaphor comprehension is a dynamic process which modifies preexisting topic/vehicle similarity, and that metaphorical interpretations are facilitated by extended context.