ABSTRACT
After 30 years of research, the mechanisms underlying the evaluative priming effect are still a topic of debate. In this study, we tested whether the evaluative priming effect can result from (uncontrolled) associative relatedness rather than evaluative congruency. Stimuli that share the same evaluative connotation are more likely to show some degree of non-evaluative associative relatedness than stimuli that have a different evaluative connotation. Therefore, unless associative relatedness is explicitly controlled for, evaluative priming effects reported in earlier research may be driven by associative relatedness instead of evaluative relatedness. To address this possibility, we performed an evaluative priming study in which evaluative congruency and associative relatedness were manipulated independently from each other. The valent/neutral categorisation task was used to ensure evaluative stimulus processing in the absence of response priming effects. Results showed an effect of associative relatedness but no (overall) effect of evaluative congruency. Our findings highlight the importance of controlling for associative relatedness when testing for evaluative priming effects.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1. Distributed memory models (e.g. Masson, Citation1995) do not solve this problem, since valence itself accounts for a limited fraction of the activation pattern of a valent concept only (Spruyt, Hermans, De Houwer, Vandromme, & Eelen, Citation2007, Footnote 8).
2. FSAA can be also more generally applied to semantic dimensions other than valence (e.g. Spruyt et al., Citation2009; Spruyt, De Houwer, Everaert, & Hermans, Citation2012).
3. We decided to use this method in order to be able to generate a large set of associated prime/target pairs that (a) are both valent, (b) contain equal numbers of same and opposite valence pairs, and (c) are of comparable associative strengths (so that quantitative indicators for associative relatedness were required). Unfortunately, a database of information on associative strength between different word pairs does not exist for the German language. Therefore, self-report data on selective items was the most promising and straightforward approach, as it allowed us to validate and equate the selection of prime/target pairs with regard to their associative strengths.
4. “Blank screens” are always meant as “blank apart from the category labels”.
5. Using more conservative outlier criteria (exclusion of reaction times below 400 ms or above 1.5 interquartile distances above the 75th percentile of the individual response-time distributions) did not change the pattern of results. Likewise, our findings were not contingent upon the inclusion or exclusion of trials on which a target was presented that was often misclassified.