Replicability and Models of Priming: What a Resource Computation Framework can Tell us About Expectations of Replicability

Abstract

In this article, we argue that whether or not a replication attempt is informative is dependent on the accuracy of one’s underlying model to explain the effect, as it is the explanatory model that enumerates the contingencies necessary for producing the effect. If the model is incorrect, then a researcher may unknowingly change variables that the model says are irrelevant but which are really essential, rendering the replication results ambiguous. The expectation that effects of priming on social behavior should be widely invariant makes sense only under the assumptions of strict direct expression and spreading activation models, yet it has been shown that these models cannot adequately explain findings from the priming literature. We describe one model of priming that predicts variability across experimental contexts and populations: the resource computation model. We highlight variables that have been uncovered under the assumptions of this model that cannot be accounted for by direct expression models and which can explain replication failures. The model is also consistent with evolutionary understandings of the mind, in which information from multiple sources beyond just stimulus information is incorporated into behavioral decisions. To the degree that anything other than a strict, direct expression, spreading activation model is correct, the expectation that priming of social behaviors should be widely invariant is unreasonable.