In the presentation she specifically talks about how "power posing"--standing in a way that denotes confidence (hands on hips)--impacts testosterone and cortisol levels and ultimately behavior.
In an upcoming issue of the Journal of Applied Psychology, Cuddy and her colleagues describe the results of an experiment where university students were told to adopt either a high-power (shown on left) or low-power (shown on right) posture:
They were instructed to stand in this pose for 5-6 minutes while preparing for a job interview speech. They then delivered said speech, which was videotaped, to two evaluators/interviewers. These evaluators subsequently rated the participants on (a) performance, and (b) hirability. They also judged the participants on their verbal and nonverbal behavior.
Results (N=61)? Those who prepared for their speech while in a high-power pose were rated significantly higher (p<.01) on both factors than those who prepared in a low-power pose. And the results could not be explained by nonverbal behavior in the interview itself. Per the authors, "Compared to low-power posers, high-power posers appeared to better maintain their composure, project more confidence, and present more captivating and enthusiastic speeches, which led to higher overall performance evaluations."
This suggest, in the words of the authors, "By nonverbally manipulating their own sense of power, the high-power posers were effectively imbued with the psychological and physiological advantages typically associated with high power, despite their low-power position relative to the evaluators."
So what does this mean? I believe it has several important implications. First, it indicates a potential source of "error", akin to test anxiety, that may impact assessment performance. Second, it suggests a potential avenue that those plagued by test anxiety may pursue to increase their chances of success. This includes those who experience stereotype threat. Third, it may help explain why even a "perfect" measurement of job-related KSAs does not yield a perfect correlation with performance (I'm thinking of similar confounding effects like mood or physical appearance).
You can read what appears to be the submitted version here. I recommend reading it as well as the excellent TED talk if this subject interests you.
