Implicit Association Test (IAT)

1 minute read

The Implicit Association Test (IAT: Greenwald et al., 1998) was designed to capture implicit associations.

Despite its popularity, there have been relatively few attempts to implement the IAT in a free and open source way, such as using PsychoPy. A version is available from a team at the University of Nottingham (the “openIAT”). However, IMHO, this implementation differs in important ways from the implementations of the IAT available in Inquisit from Millisecond, and the versions employed on the Project Implicit website.

First, the openIAT it employs both free responding and also fixed duration corrective feedback (i.e., if you get a trial wrong you are presented with “oops” for one second but do not have to emit a correct response). In contrast, the vast majority of implementations of the IAT as described in the literature use either free responding (i.e., whether you get trials correct or wrong you continue to the next trial) or contingent responding (i.e., if you get a trial wrong you are presented with a red “X” and must emit the correct response to continue). I think this difference is non-trivial because the two algorithms that are most frequently used to quantify effects on the IAT (the D1 and D600: Greenwald et al., 2003) map on to whether contingent or free responding designs, respectively, are used within the task. Second, the block layout of the openIAT doesn’t conform to the original one described in Greenwald el al. (1998) or the updated version described in Nosek et al. (2007). Both of these describe 7 blocks, whereas the openIAT includes 5.

As such, if one is looking for an “off the shelf” open-source implementation of the IAT that is equivalent to that employed on Project Implicit or that distributed by millisecond, existing solutions aren’t quite up to scratch.

I’ve therefore put this one together. To the best of my knowledge, it provides high fidelity to Millisecond’s Inquisit implementation, and the task description by Nosek et al., (2007).

Edit: I’ve now used this code in two experiments and am satisfied that it’s sufficiently bug free for more general use.

Download the current release or check out the project page if you’d like to see the most recent code or contribute to the project.