Hi Lauren,
Indeed, it is not recommended to use the IAT score computed with our IAT extension for research. The IAT extension's scoring is slightly different than the recommended, for considerations (I don't fully remember) that are related to the feedback to participants.
The IAT D score in the recommended scoring algorithm is the average of two IAT scores: one computed from Blocks 3 and 6, and one from Blocks 4 and 7. The IAT extension's scoring computes only one score, from all the blocks.
In other words, the extension's D = [(mean latency in blocks 6 and 7) - (mean latency in blocks 3 and 4) / (Standard deviation in blocks 3,4,6,7)]
The recommended D = {[(mean latency in Block 6) - (mean latency in Block 3 ) / (Standard deviation in Blocks 3&6)] + [(mean latency in Block 7) - (mean latency in Block 4) / (Standard deviation in blocks 4,7)]}/2
In addition, before computing the mean latency in the extension, we delete all trials with latency above 10000 and below 400, and we replace the latency of error trials with the [mean latency of that blocks + 600]. We do not provide feedback for participants who had at least 10% of trials with latency below 150ms.
Yoav