Abstract

Abstract This research investigates the process of metaphor comprehension. We compare two classes of processing algorithms: those that begin by accessing or deriving an abstraction from the base (or vehicle ) and then project it to the target (or topic ) and those that begin by aligning the representations of the terms and then project further inferences from the base to the target. To decide between these accounts, we recorded subjects’ time to interpret metaphors primed by either the base term or the target term (or both or neither). Abstraction-based models predict that priming by the base should lead to faster metaphor comprehension than priming by the target; alignment-based models predict no such advantage. Across a series of experiments, the results were most consistent with alignment-first processing. No base advantage was found, with the single exception of metaphors having highly conventional meanings and low (metaphorical) similarity. Further, high-similarity metaphors were interpreted faster than low-similarity metaphors, consistent with the alignment view. We conjecture that this pattern may result from a shift in processing with conventionalization: novel metaphors may be understood by alignment and conventional metaphors by abstraction.