A cue-compounding model of semantic priming

Abstract

<p>Three projects on semantic priming are reported that converge on a variant of Ratcliff & McKoon's (1988) cue-compounding model. In Chapter 2, two experiments test the semantic matching mechanism of backward priming and nonword facilitation. Contrary to predictions derived from the semantic match, results from Experiment 1, in which stimulus onset asynchrony (SOA) is manipulated, indicate that backward priming and nonword facilitation are caused by independent mechanisms. Results from Experiment 2, which manipulates relatedness proportion and nonword ratio by including different proportions of "related" nonwords in the list, concur with the separate mechanism hypothesis. The cue-compounding model is presented in Chapter 3 to account for the data. In Chapter 4, two further experiments investigate semantic priming of degraded targets. Spreading activation accounts (e.g., Collins & Loftus, 1975) predict that priming should be greater if degradation slows identification of the target. However, it was found at at short SOA, degrading the target either by presenting it in mixed upper and lower case letters (e.g., DoCTor) or by masking it after a brief presentation (150 ms) decreased priming against a neutral baseline. It is asserted that this pattern of results is inconsistent with spreading activation accounts, but can be simulated with the model presented in Chapter 3. In Chapter 5, a critique of a recent article by McNamara (in press a) is presented. McNamara reported three experiments demonstrating three-step mediated priming (Experiment 1) and intertrial prime-target lag (Experiment 2 and 3). Although the results lead McNamara to conclude that spreading activation is the better account, this critique asserts that his results do very little to distinguish between current models because of questionable inferences and statistical procedures. The model in Chapter 3 is used to simulate three-step mediated priming.</p>