Reader Comments

Post a new comment on this article

This study is very interesting, but the effect sizes are pretty small...

Posted by Tong on 01 Sep 2011 at 20:56 GMT

... at least in the case of Experiment 1, where the sensitivity measure, d', was only ~0.2. d' is not always easy to interpret, but note that a d' of 0 represents chance level classification performance.

No competing interests declared.

RE: This study is very interesting, but the effect sizes are pretty small...

nickrule replied to Tong on 09 Sep 2011 at 14:12 GMT

The calculation of d' from the reported mean hits and false-alarms gives only a rough estimate of its size. Better is to use the reported signal detection statistics and companion effect sizes reported in the paper. In Study 1, the effect size for perceivers' capacity to distinguish political affiliation is r = .64, which is moderate to large.

Yet and still, the point about the mean accuracy as compared to chance is quite valid. Indeed, if political affiliation was very obvious from the face (as it is in most cases for race or sex, for example), the results of this work would be far less interesting or meaningful. The fact that the effect is so consistently above chance guessing for a distinction that is so subtle is what is most compelling about the work. The purpose of the research is to demonstrate and understand how it is that we detect and utilize subtle and ambiguous information from appearance and nonverbal behavior, not to claim that this is a blockbuster effect, as it is not.

N. Rule

No competing interests declared.