Estimating the reliability and stability of cognitive processes contributing to responses on the Implicit Association Test

Abstract

Implicit measures were initially assumed to assess stable individual differences, but other perspectives posit that they instead reflect context-dependent processes. The present, pre-registered research investigates the extent to which responses on the Implicit Association Test reflect stable versus context-dependent processes using multinomial processing tree modeling to disentangle the joint contributions of multiple cognitive processes. We applied two models – the Quad model and the Process Dissociation Procedure – to six independent datasets (total N = 2,036), each collected in a repeated-measures design over two occasions, examined the retest reliability and internal consistency of model parameters, and meta-analyzed the results. Parameters reflecting control-oriented processes in both models demonstrate adequate within-measurement recoverability and fair between-measurement retest reliability, which suggests that these processes are relatively stable within individuals. In contrast, parameters reflecting evaluative associations demonstrate moderate recoverability but poor retest reliability, which suggests that these processes are not stable within individuals. These findings illustrate that the extent to which implicit attitudes are state- or trait-like depends on the process in question, which has practical implications for predicting behavior using the Implicit Association Test.