Testing the automaticity features of the Affect Misattribution Procedure: The roles of awareness and intentionality

Abstract

The Affect Misattribution Procedure (AMP) is a measure of implicit evaluations, designed to index the automatic (unintentional) retrieval of evaluative knowledge. The AMP effect consists in participants evaluating neutral target stimuli more positively when preceded by positive primes and more negatively when preceded by negative primes. Hughes et al. (2022) questioned the automaticity of the AMP based on an awareness effect, i.e., the finding that AMP effects were larger when participants indicated that their response was influenced by the prime than when they did not. Here we present seven experiments (six preregistered; N = 2,350) investigating the nature of the awareness effect and, more broadly, the automaticity features of the AMP. In Experiments 1–4, the awareness effect was robust but highly variable as a function of procedural changes and stimulus valence. Critically, Experiments 5–7 provided evidence for an alternative explanation of the awareness effect, namely that awareness can be the outcome, rather than the cause, of evaluative congruency between primes and responses: Awareness effects emerged even under conditions that made it impossible for awareness to contribute to AMP effects, including when participants judged influence awareness for third parties or primes were presented post-hoc, following influence judgments. Finally, increasing evaluative strength of the primes increased participants’ tendency to misattribute AMP effects to the influence of target stimuli. Together, these findings support the construct validity of the AMP as a measure of implicit evaluations by suggesting that AMP effects are unintentional and can create awareness effects rather than vice versa.

Evidence against effects on the Affect Misattribution Procedure being unaware: AMP effects involve construct-irrelevant individual differences

“…In four experiments, Kurdi et al (2022) successfully replicated and extended Hughes et al's (2022, experiment 2) procedure and effect. However, they dismiss the results of these replications as "uninformative with respect to the underlying theoretical mechanism [i.e., awareness in the AMP effect]".…”

“…We replicated and extended this effect in a larger second dataset, showing that the effect is replicable and that these problematic correlations are larger for the AMP than five other common implicit measures. Based on the line of reasoning of Kurdi et al (2022), we therefore found evidence against the AMP effect being unaware. Finally, the observed heterogeneity in tendency to demonstrate AMP effects implies the task is measurement non-invariant between participants.…”

“…Effects on the Affect Misattribution Procedure (AMP) have been claimed to be implicit in the sense of being unaware but the evidence for this remains to be weak. Recently, Kurdi et al (2022) provided a new method of testing this: they argue the absence of a construct-irrelevant individual difference factor would represent supportive evidence for the AMP effect being unaware (Kurdi et al account), whereas the presence of a construct-irrelevant individual difference factor would represent evidence against the unaware account. They test for this by correlating AMP effects between multiple domains, where a non-zero correlation would support the latter position.…”

“…However, Hughes et al (2022) pointed out that this evidence came from a small study that presented a nonsignificant p-value as evidence for the null effect, which is a statistical fallacy (Lakens et al, 2018). In a recent preprint, authors including the AMP's creator (Payne) have (1) acknowledged that Payne et al (2013, Experiment 3) does not provide credible evidence for the AMP's unawareness; (2) stated that they successfully replicated and extended Hughes et al's (2022) effect in seven studies; (3) but also argue against Hughes et al's (2022) account (Kurdi et al, 2022). While we disagree with Kurdi et al's (2022) analyses and conclusions in multiple ways, even if we take their arguments and evidence at face value it would mean that there is no supportive evidence for Payne et al's (2013, Experiment 3) original claim that the AMP effect is unaware.…”

How do explicit and implicit evaluations shift? A preregistered meta-analysis of the effects of co-occurrence and relational information

“…We do not share these concerns about the validity of the AMP. We believe that in its totality, available evidence overwhelmingly supports the construct validity of the AMP as a measure of implicit evaluations (for a review, see Payne & Lundberg, 2014), and the specific concerns raised by Bar-Anan and Nosek (2011) and Hughes et al (2022) have been laid to rest in followup work by Payne et al (2012) and Kurdi, Melnikoff, et al (2022). Moreover, we are generally of the view that susceptibility to relational information should not be used to judge the construct validity of measures of implicit evaluation.…”