Decoding the implicit association test: Implications for criterion prediction
Academic Article
Overview
Research
Identity
Additional Document Info
Other
View All
Overview
abstract
The implicit association test (IAT) is believed to measure implicit evaluations by assessing reaction times on two cognitive tasks, often termed "compatible" and "incompatible" tasks. A common rationale for studying the IAT is that it might improve our prediction and understanding of meaningful psychological criteria. To date, however, no clear psychometric theory has been advanced for this measure. We examine the theory, methods and analytic strategies surrounding the IAT in the context of criterion prediction to determine measurement and causal models a researcher embraces (knowingly or unknowingly) by using the test. Our analyses reveal that the IAT revolves around interpretation of two distinct relative constructs, one at the conceptual level and one at the observed level. We show that interest in relative implicit evaluations at the conceptual level imposes a causal model that is restrictive in form. We then examine measurement models of the IAT and show how computing a difference score at the observed level may lack empirical justification. These issues are highlighted in a study replicating an effect established in the literature (Study 1). We then introduce a new variant of the IAT and use it to evaluate the reasonableness of traditional IAT methods (Study 2). 2005 Elsevier Inc. All rights reserved.
