Commentary on Elkjær et al.’s (2020) Meta-Analysis on Expansive Versus Contractive Nonverbal Displays

Research on expansive versus contractive nonverbal displays has been and still is a highly debated topic. Despite immense attention, no one has yet tried to quantify the empirical evidence on effects of power poses and other body postures. The meta-analysis by Elkjær et al. (2020) is thus very timely. In testing a multiplicity of moderators, the authors conducted thorough analyses. Elkjær et al. advise researchers to attend the topic of demand characteristics and contextual factors in upcoming research and also to further investigate the effect of motor displays on hormones. The suggestions to design future studies with control groups, and the assumption that the experimental group performing contractive displays may drive the effect are certainly extremely important to advance the field. Yet we believe that Elkjær et al.’s meta-analysis cannot stand as a full corrective to previous research because of three limitations: (a) limited open-science practices, (b) selection of studies, and (c) the theoretical model.

First, the procedure applied in the meta-analysis did not follow recent open-science practices such as preregistering the study and publishing the data and analysis code. Of course, such practices are not compulsory, but we argue that they would be extremely helpful. Especially in the field of power posing, which has been plagued by failed replication attempts (e.g., Ranehill et al., 2015), such an approach seems important (Cesario et al., 2017). After all, the phenomenon of power posing has developed into a prime example in the replication crisis, and preregistered studies and replication projects have been conducted since 2017 (Jonas et al., 2017). Criticism on power posing has decreased trust in this topic, and it would have been a signal of trustworthiness if the researchers had specified in advance which strategies they planned to use to select the studies and analyze the data and which moderators they planned to test. Thus, we believe that a meta-analysis in this field needs to adhere to the highest standards of open-science practices and transparency.

Regarding the selection of studies in the meta-analysis, we recommend checking and potentially correcting for questionable research practices (John et al., 2012; Simonsohn et al., 2015; van Aert et al., 2016) and being more transparent regarding the corrections for publication bias. Note that p-hacking seems to be present in power-posing research (Simmons & Simonsohn, 2017) and has introduced biases in the effect sizes among published studies (Credé & Phillips, 2017). Moreover, corrections for publication bias—such as trim-and-fill, as used by Elkjær et al.—have been criticized (e.g., Carter et al., 2019). To date there is no single method correcting for bias that consistently outperforms all the others (Carter et al., 2019). Thus, we advise increased transparency regarding the impact of the bias correction procedures on the estimated effect sizes by also reporting uncorrected estimates and estimates based on additional methods.

On the basis of literature searches in electronic databases, the authors selected 25 studies for their review and 48 studies for their meta-analysis, which is the broadest database so far for meta-analytical evidence in this field. However, when considering the studies in a recent publication (Körner & Schütz, 2020) that provided a comprehensive overview of research on power poses and other body postures, Elkjær et al.’s analyses do not seem to be exhaustive. We identified 17 relevant studies that had been conducted in the time interval used by Elkjær et al. but were not included (for a list of these studies, see the Supplemental Material available online). Furthermore, because many studies in the meta-analysis dealt with power poses, we had also expected that the term “power posing” would have been used in a search string—but that was not the case. Perhaps this explains why relevant studies were missing. Note that different results may be obtained when different sets of studies are used (see Cuddy et al., 2018; Simmons & Simonsohn, 2017). Still, because many of the missing studies do not have control groups, it is possible that the observed effect is still driven by contractive nonverbal displays or movements.

Our third concern is that the authors did not distinguish among various kinds of body positions and movements. The concept of power posing was brought up in the recent literature on power. New frameworks and theories contribute to the steady evolution of this field. One of these frameworks is the dominance-prestige model, which is increasingly used to explain diverging findings in status and power research (Cheng et al., 2014).

The dominance-prestige framework was recently applied to the literature on expansive versus contractive nonverbal displays (Körner & Schütz, 2020; Witkower et al., 2020). In this context, dominance describes status attainment through threat, coercion, and intimidation; prestige refers to the ascription of status by others on the basis of one’s knowledge, expertise, and experience. The two strategies are associated with different patterns of emotions, traits, behaviors, and neuroendocrinological profiles (Cheng et al., 2014). Witkower et al. (2020) analyzed the nonverbal expression of dominance and prestige and noted that “different forms of expansiveness send different messages about power [for actors and perceivers, and so] the previously documented power pose may be associated with two largely opposing messages” (p. 114). Körner and Schütz (2020) followed this notion and distinguished upright and slumped postures from high- and low-power poses. There is initial empirical and theoretical evidence that poses can be understood as the nonverbal expression of dominance. Poses are defined by expansiveness of the whole body and correspond to the typical power pose (Carney et al., 2010). On the other hand, postures refer to the inclination of the spine (Riskind & Gotay, 1982). Upright postures can be understood as nonverbal representations of prestige. The two types of body positions carry different meanings (Witkower et al., 2020), and it is plausible that they are related to different kinds of actor reactions and should thus be distinguished (e.g., poses were found to affect power feelings whereas postures influenced emotions; Körner & Schütz, 2020). Thus, a meta-analysis on expansive versus contractive nonverbal displays might use poses versus postures as a moderator in studying the effects of body positions. Contrary to postures, it is possible that effects of poses on various outcomes are driven by the high-power-posing condition because poses are characterized by higher intensity than postures. and thus differences between high- and low-power poses compared with neutral positions should be more pronounced.

Likewise, considering cultural factors may be important in this differentiation. Dominance and prestige are two universally recognized pathways to social rank. Yet the effects of nonverbal expressions may differ between cultures. For example, Park et al. (2013) found that, with respect to poses, two dominant body positions led to identical effects in East Asian and U.S. participants, but a third dominant pose did not increase power feelings in East Asian individuals because it violated collectivistic norms. Attending to the cross-cultural and cross-situational comparability between the effects of poses and postures as nonverbal expressions of dominance and prestige may also help in clarifying whether these certain body positions lead to similar effects across cultures and situations or whether they entail different consequences independent of their physical nature (see Barrett, 2012).

We believe that considering the points above will advance the field of expansive versus contractive nonverbal displays. Thus, for further metascientific evidence of body positions, we suggest the following: (a) engagement with open-science practices, (b) clarity regarding the selection of studies, and (c) conceptual clarity in defining and distinguishing different kinds of body positions. We think these considerations would be an excellent step toward minimizing selection bias and thus increasing trust in the evidence on effects of expansive versus contractive nonverbal displays. Until then, we advise researchers not to test methodological moderators because premature conclusions may be drawn.

Clearly, it is very important to test the metascientific evidence of poses and postures. Elkjær et al. did a great job keeping the debate surrounding body-position effects alive and pointing out the methodological shortcomings of past literature. Yet we believe that additional theoretical considerations (e.g., dominance-prestige model) and methodological considerations (e.g., open-science practices and clarity on study selection) would be useful to further advance our understanding of the relevant effects in the field.