Individual Differences in Implicit Bias Can Be Measured Reliably by Administering the Same Implicit Association Test Multiple Times

Abstract

The use of the Implicit Association Test (IAT) as a measure of individual differences is stymied by insufficient test–retest reliability for assessing trait-level constructs. We assess the degree to which the IAT measures individual differences and test a method to improve its validity as a “trait” measure: aggregating across IATs. Across three studies, participants (total n = 960) completed multiple IATs in the same session or across multiple sessions. Using latent-variable models, we found that half of the variance in IAT scores reflects individual differences. Aggregating across multiple IATs approximately doubled the variance explained with explicit measures compared with a single IAT D-score. These findings show that IAT scores contain considerable noise and that a single IAT is inadequate to estimate trait bias. However, aggregation across multiple administrations can correct this and better estimate individual differences in implicit attitudes.