Measuring automatic associations: Validation of algorithms for the Implicit Association Test (IAT) in a laboratory setting

Background and objectivesIn their paper, “Understanding and using the Implicit Association Test: I. An improved scoring algorithm”, Greenwald, Nosek, and Banaji (2003) investigated different ways to calculate the IAT-effect. However, up to now, it remained unclear whether these findings – based on internet data – also generalize to laboratory settings. Therefore, the main goal of the present study was to cross-validate scoring algorithms for the IAT in a laboratory setting, specifically in the domain of psychopathology.MethodsFour known IAT algorithms and seven alternative IAT algorithms were evaluated on several performance criteria in the large-scale laboratory sample of the Netherlands Study of Depression and Anxiety (N = 2981) in which two IATs were included to obtain measurements of automatic self-anxious and automatic self-depressed associations.Results and conclusionsResults clearly demonstrated that the D2SD-measure and the D600-measure as well as an alternative algorithm based on the correct trials only (DnoEP-measure) are suitable to be used in a laboratory setting for IATs with a fixed order of category combinations. It remains important to further replicate these findings, especially in studies that include outcome measures of more spontaneous kinds of behaviors.

► D2SD, D600 and D-measure without error penalty appeared suitable for laboratory settings. ► D-measures showed highest correlation with explicit equivalent. ► D-measures showed lowest correlation with general response speed. ► D-measures showed highest predictive validity.