One might wonder that I used rather few stimuli as primes and targets in my experiments. The stimulus set sizes in evaluative priming studies are in general much smaller in comparison with the set sizes in semantic priming studies (see Klauer & Musch, 2003). One reason for this may lie in the rather limited number of clearly valenced stimuli. Klauer and Musch (2001) examined evaluative priming in the naming task with different set sizes of prime and target stimuli and did not report significant influences of the stimulus set size; however―as discussed earlier―they failed to find any significant priming effects in this series of experiments.
I have to admit that the magnitude and the robustness of S–S-based evaluative priming effects in my experiments were rather weak. Regarding the effect sizes of S–S-based evaluative priming effects in prior studies, the effect seems to be rather small and was associated with medium (see, e.g., Glaser & Banaji, 1999; Hermans et al., 1994; Spruyt, De Houwer et al., 2007; Spruyt et al., 2002) or even small (see, e.g., Everaert et al., 2011; De Houwer & Randell, 2004; Spruyt & Hermans, 2008) effect sizes. Only Bargh and colleagues (1996) reported large effect sizes for the positive evaluative priming effect with the naming task (but see Klauer & Musch 2001, for a failure to replicate).
One of the main and most important findings in my experiments was the prolonged response conflict between response-incompatible prime and target given evaluative congruency as compared to incongruency. Applying the rationale of the three-process model, I interpreted this result in the way that the activation of an evaluatively congruent prime is supported by the target, leading to a parallel activation of prime and target, and―as a direct consequence―resulting in a larger interference of the target response by a response-incompatible prime. Following this logic, an evaluatively incongruent prime is not sufficiently activated and does, thereby, not have the potential to disturb the target response. I must, however, admit that the finding of an increased target response interference given evaluative congruency as compared to incongruency also allows for an alternative interpretation. Taking the idea of distributed concept representation―as suggested in the distributed memory model by Masson (1991, 1995)―into account and applying this idea to my experiments, the activation of the target was necessarily
accompanied by the activation of an evaluatively congruent prime because the activation units corresponding to the evaluative connotations of prime and target overlapped. According to the rationale of the distributed memory model, however, the parallel activation of evaluatively congruent prime and target is restricted to their overlapping parts. The non-overlapping features, by contrast, may impede each other and mutually inhibit the formation of the complete activation pattern of either the prime or the target concept. Since, thus, also the response-related features of response-incompatible prime and target may impede each other, an increased response conflict given evaluative congruency in comparison with incongruency may occur.
Both interpretations, that is, the parallel activation of the prime and the target concept in case of evaluative congruency (as suggested by the three-process model) versus a parallel activation that is restricted to the overlapping, evaluative features of the prime and the target pattern with a mutual interference of the remaining prime and target features (as implemented by Masson’s [1991, 1995] distributed memory model), are compatible with the finding of increased target response interference by an evaluatively congruent compared with an incongruent prime. In order to test the explanation derived from the three-process model against the alternative interpretation, it is necessary to examine the representational status of the prime concept in case of evaluative congruency. An elegant and reasonable operationalization would be to employ prime and target stimuli from (at least) three response categories with a specific analysis of the erroneous target responses. If―in case of response-incompatibility and evaluative congruency―the response associated with the prime would interfere with the target response more frequently than expected by chance, one could conclude that evaluatively congruent prime and target representations are simultaneously activated and that the prime activation is maintained by an evaluatively congruent target. Thus, such a finding would corroborate the interpretation according to the three-process model. Otherwise, if the erroneous target responses would equally often arise due to all available responses without a significantly more frequent interference by the prime-associated response, the increased target response conflict given evaluative congruency would be caused by a mutual inhibition of the activation patterns corresponding to the prime and the target concept, impeding the formation of the entire target activation pattern. Such a result would
rather corroborate an interpretation in line with the explanation of priming effects by the distributed memory model.
Since I examined evaluative priming in sequential priming tasks with one target and a single prime per trial, my considerations concerning the interaction of evaluatively congruent concepts and the representation of the evaluative connotations in the semantic memory are restricted to two concepts. It may be interesting and a matter of future research to broaden the applicability of the three-process model on priming task settings with more than two stimuli. The memory models with mechanisms of synchronously firing activation patterns and activation units that alternate their rhythm between patterns should in principle allow for the simultaneous activation of more than two concepts, as well. Furthermore, I did not explore in how far the three-process model is able to account for S–S-based evaluative priming effects with nonconsciously perceptible primes. For this, similar S–S-based evaluative priming studies should be examined with masked prime presentations.