| Semantic priming researchers tend to assume that associative relatedness, as derived from word associations, and semantic similarity are equivalent constructs. Based on this observation, two methods for computing associative relatedness from word associations were mapped onto the notions of semantic distance and semantic relatedness in Collins and Loftus' (1975) spreading activation theory of semantic memory. The theory assumes that activation automatically spreads along all paths between prime and target. Thus semantic priming should depend more on the aggregate of paths between prime and target (i.e., semantic relatedness) than on the strength of a single path (i.e., semantic distance). Experiment 1 extended a common metric of semantic distance, associative probability, to examine priming for directly related targets (e.g., anchor to boat) and indirectly related targets (e.g., rake to tree via leaves). The results suggest that the associative probability metric used in direct priming studies can also be used in studies of mediated priming. Experiment 2 examined mediated priming in terms of semantic distance and semantic relatedness. The investigation was approached by mapping Deese's (1962) notion of associative meaning onto semantic relatedness, and then operationalizing semantic relatedness in terms of an existing metric of associative relatedness. The results show that priming depends on semantic relatedness. Moreover, Experiments 1 and 2 converged to suggest that semantic relatedness can override the expected semantic distance effect. These results favor semantic network models based on spreading activation theory, and raise questions about data that are presumably inconsistent with network models, as such data is usually based on the notion of semantic distance alone. |
