Written by Carol Nickerson Power Posing Redux Lately Amy Cuddy has been asserting that there have been multiple statistically significant replications of the effect of "power posing" on "feelings of power." Brief historical review: Remember that the original power-posing article (Carney, Cuddy, and Yap, PSYCHOLOGICAL SCIENCE, 2010) claimed that power posing --- adopting an expansive rather than a contractive posture --- produced elevations in testosterone, decreases in cortisol, increased feelings of power, and increased risk tolerance. A subsequent article by Ranehill et al. (PSYCHOLOGICAL SCIENCE, 2015) with a much larger sample size (n = 200 vs. n = 42) failed to replicate any of these effects except for power posing increasing feelings of power; the supplemental material for this article indicated that this effect held only for men. In response to this failure, Cuddy declared that the most important effect of power posing is that "adopting expansive postures causes people to feel more powerful," downplaying interest in other outcomes of power posing: https://ideas.ted.com/inside-the-debate-about-power-posing-a-q-a-with -amy-cuddy/ In response to Ranehill et al.'s (2015) failed replication attempt, Carney, Cuddy, and Yap (PSYCHOLOGICAL SCIENCE, 2015) reviewed 33 studies of the effects of expansive versus contractive postures on various outcomes, attempting to explain why Ranehill et al. (2015) failed to replicate three of their four findings. Using a p-curve analysis of the studies in Carney, Cuddy, and Yap's (2015) review, Simmons and Simonsohn (PSYCHOLOGICAL SCIENCE, 2017) then demonstated that, once selective reporting is accounted for, power posing has no significant effect. But Simmons and Simsonsohn (2017) excluded from their p-curve analysis studies that examined only the relation between power poses and feelings of power because a few studies had described this relation as a manipulation check. Update: Cuddy, Schultz, and Fosse (PSYCHOLOGICAL SCIENCE, 2018) have just published a reply to Simmons and Simsonsohn (2017). Oddly, given Cuddy's statement on Facebook that p-curve analysis is a flawed methodology ("a new statistical approach, that we now know has all kinds of problems"), Cuddy, Schultz, and Fosse (2018) used p-curve analysis to examine the effects of power poses (now renamed "postural feedback") on (1) all outcomes, (2) feelings of power, (3) emotion, affect, and self- evaluation (EASE) outcomes (a subset of all outcomes but excluding feelings of power), and (4) outcomes other than EASE outcomes. Of particular importance to Cuddy now is the effect of power poses on feelings of power. From her twitter account (@amyjccuddy): * January 24, 2017, in response to a tweet from the University of Pennsylvania that "It turns out that power poses don't boost confidence after all" "This is absolutely false. Effect of adopting expansive postures on feelings of power replicated > than 10 times. Please correct, @Penn." * October 19, 2017, in response to Susan Dominus's NEW YORK TIMES article https://www.nytimes.com/2017/10/18/magazine/when-the-revolution -came-for-amy-cuddy.html) "The effect of power posing on feelings of power has replicated 17 times, which @nytimes piece didn't convey. Pls correct @susandominus Thnx!" * November 30, 2017 "The effect of posture on feelings of power has replicated 17 times." * December 17, 2017 "Effect on feeling power has replicated 17 times, declared a real effect in published Bayesian meta-analysis, & strong support in p-curve analyses. (Also, original author (singular) [CUDDY], who has since published paper [THE 2018 PSYCHOLOGICAL SCIENCE ARTICLE] stating effect on feelings is real.) This tweet contained a link to a draft version of the 2018 PSYCHOLOGICAL SCIENCE article. The claim of 17 replications also appeared in the January 21, 2018, BOSTON GLOBE article written by Pardis Sabeti but clearly greatly influenced by Cuddy: https://medium.com/@OmnesRes/who-wrote-this-article-feac245345cb and in a recent "CBS This Morning" podcast: https://soundcloud.com/cbsthismorning/social-psychologist-amy -cuddy-on-body-mind-effects-the-science-behind-it-and-critics at about 11:45 minutes in. Jordan Anaya and I became curious about this claim of 17 replications of the effect of power posing on feelings of power. Are there really so many? Cuddy, Schultz, and Fosse (2018) reported 15 cases examining the effect of power posing on feelings of power. (Note that the text indicates 14 cases; Figure 1b and the database indicate 15 cases.) In 11 of these 15 cases, the effect was significant; in 4 it was not. The 15 cases are listed in an Excel file ("Disclosure_Table_CSF_2017.xlsx") available on the Open Science Framework (https://osf.io/pfh6r/). Examination of these 15 cases shows that there are indeed 4 that do not have significant effects (Cuddy et al., 2015; Huang et al., 2011, experiment 2; Smith and Apicella, 2017 --- this article was the basis for the University of Pennsylvania press release mentioned above; Turan, 2015). What about the 11 cases that supposedly show a significant effect of power posing on feelings of power? Let's take a look: (1) In the original article by Carney, Cuddy, and Yap (2010), power posing showed a significant effect on feelings of power; high-power posers reported greater feelings of power than did low-power posers. Although Cuddy, Schultz, and Fosse (2018) including this case in their p-curve analysis is legitimate, the case itself cannot be considered a *replication*; it is the original finding to be replicated (or not) by other studies. (Cuddy may be misusing the word "replications" to mean "instances.") (2) The original article by Carney, Cuddy, and Yap (2010) also reported with minimal detail a case from "another study" (p. 1367) that showed a similar significant effect of power posing on feelings of power. The review by Carney, Cuddy, and Yap (2015) indicated that this was a "pilot experiment" (p. 658). Although Cuddy, Schultz, and Fosse (2018) including this case in their p-curve analysis may be legitimate, I would consider the case itself not to be a *replication*, but, as a pilot experiment, part of the Carney, Cuddy, and Yap (2010) original findings to be replicated (or not) by other studies. Note that the document by Dana Carney disavowing power posing http://faculty.haas.berkeley.edu/dana_carney/pdf_my% 20position%20on%20power%20poses.pdf stated that "The self-report DV [FEELINGS OF POWER] was p-hacked in that many different power questions were asked and those chosen [FOR INCLUSION IN CARNEY, CUDDY, AND YAP (2010)] were the ones that 'worked.'" (3) The study by Ceunen et al. (2014) should not have been included in the sample of 15 cases. The expansive pose is an unusual one, and the authors explicitly warned that it may not be associated with dominance or pride (p. 3). Moreover, the effect, although significant, is in the wrong direction. (4) The effect in the study by Garrison, Tang, and Schmeichel (2016) is in the wrong direction, so it cannot be counted as a replication. (5) The effect in the study by Huang et al. (2011), experiment 1, is significant in the right direction. (6) The effect in the study by Park et al. (2013), study 2a, is significant in the right direction. (7) The effect in the study by Park et al. (2013), study 2b, is significant in the right direction. (8) The effect in the study by Park et al. (2013), study 3, is significant in the right direction but only for participants from the USA, not for participants from East Asia. This case should be considered a mixed or ambiguous result, not a replication. (9) The effect in the study by Ranehill et al. (2015) is significant in the right direction, but an analysis in the supplemental material showed that this effect held only for men, not for women. This case should be considered a mixed or ambiguous result, not a replication. (10) Rotella and Richeson (2013) examined shame/guilt/pride rather than power. In my opinion, this case should not have been included in the sample of cases for the p-curve analysis of "feelings of power" but in the sample of cases for the p-curve analysis of "emotion, affect, and self-evaluation outcomes." The only other case in the database examining pride (Stepper and Strack, 1993) is not included in the sample of cases for feelings of power, but this seems to be due to an objection to this case in the disclosure table for the article by Simmons and Simonsohn (2017). (11) The effect in the study by Teh et al. (2016) is significant in the right direction. There has been no reply to my e-mail message to Cuddy, Schultz, and Fosse about problematical cases #3, #4, and #10. So, for the purpose of counting replications (not for the purpose of doing a formal p-curve analysis), there are 11 (not 15) relevant cases in the Cuddy, Schultz, and Fosse (2018) article (omitting the 2 cases from Carney, Cuddy, and Yap, 2010; Ceunen et al., 2014; Rotella and Richeson, 2013). Of these 11 cases, 5 are non-replications (either because p not < .05, or because p < .05 in the wrong direction); 4 are replications (p < .05); and 2 are mixed or ambiguous. (Perhaps these 2 might be considered partial replications.) Where Are the 17 Cases? But Cuddy repeatedly has claimed that there are 17 (significant) replications. Eleven of those (supposedly) are included in the new p-curve analysis by Cuddy, Shultz, and Fosse (2018). Where are the remaining 6? The article by Cuddy, Schultz, and Fosse (2018) mentions (endnote 5) that the publication end date for inclusion in their p-curve analysis was December 20, 2016, and that several additional studies from 2017 are listed in their supplemental materials archived on the Open Science Framework. These articles are listed in a file named "Additional Studies." The "Additional Studies" file contains 8 articles, 7 published in the 2017 special issue of COMPREHENSIVE RESULTS IN SOCIAL PSYCHOLOGY (CRSP) devoted to power poses, plus an additional article published in a computer-science journal. Of the 7 CRSP articles, 6 examined the effect of power posing on feelings of power. Another file ("Systematic Literature Review Methods") on the Open Science Framework notes that one of these articles (Ronay et al., 2017) did not obtain significant results. So, 11 (supposed) cases from the Cuddy, Schulz, and Fosse (2018) p-curve analysis article, plus 5 cases from CRSP, plus 1 case from a computer-science journal, and we have 17 cases. But Cuddy's tally assumes that the 5 cases (excluding Ronay et al., 2017) from CRSP all gave significant results. They did not. Two of the 5 did not find a significant effect of power poses on feelings of power (Keller et al., 2017; Latu et al., 2017); 1 did find a significant effect (Klaschinski et al., 2017); and 2 yielded mixed or ambiguous results in which the findings for the overall group of participants differed from those for subsamples of the participants (Bailey et al., 2017; Bombari et al., 2017). The article published in a computer-science journal did find a significant effect of power posing on feelings of power (Pena and Chen, 2017). So, combining this "Additional Studies" tally with my earlier tally yields 6 replications, 8 non-replications, and 4 mixed or ambiguous cases. Six replications is substantially fewer than the 17 repeatedly reported by Cuddy. Thus, it seems to me that the effect of power posing on feelings of power is being overstated. My analysis assumes that all the relevant cases are included in the various files available on the Open Science Framework website. This may not be so, although there is no indication otherwise. Carol