The Power of Pictures: Vertical Picture Angles in Power Pictures

Abstract

Conventional wisdom suggests that variations in vertical picture angle cause the subject to appear more powerful when depicted from below and less powerful when depicted from above. However, do the media actually use such associations to represent individual differences in power? We argue that the diverse perspectives of evolutionary, social learning, and embodiment theories all suggest that the association between verticality and power is relatively automatic and should, therefore, be visible in the portrayal of powerful and powerless individuals in the media. Four archival studies (with six samples) provide empirical evidence for this hypothesis and indicate that a salience power context reinforces this effect. In addition, two experimental studies confirm these effects for individuals producing media content. We discuss potential implications of this effect.

Notes

The research reported in this article was supported by a fellowship from the Erasmus Research Institute of Management, Erasmus University Rotterdam, The Netherlands, awarded to the first author and a Feodor-Lynen-Fellowship from the Humboldt Foundation awarded to the third author.

1. Please note that Cohen's Kappa would not be a good estimator of agreement, because we have used a continuous scale and not a categorical scale. Further, if we would treat the seven answer possibilities as categories, Kappa values will be relatively low, because Kappa becomes lower with more categories.

2. We controlled for this fact, because children are smaller and might, therefore, naturally more often being photographed from an above camera angle.

3. We had to restrict our search by these means, because the maximum picture results was 10,000 (at the time of our study). Increasing the database (e.g., by looking for 10 instead 1 year) would have resulted in search results of >10,000—not enabling us to calculate usable statistics.

4. Interestingly, many of the jobs outlined in Table 2 might actually be stereotypically gendered—especially some of the ones with the strongest and lowest percentage of pictures from below (e.g., priest, politician, president, boss, leader vs. tea lady, cleaning lady, secretary etc.).

5. We could not conduct a log-linear regression for the different pictures to test, for instance, whether target gender significantly impact the found interaction between target power and picture angle, because some expected cells are zero (i.e., violating assumptions of the chi-square and log-linear tests). Therefore, we used Fisher's exact probability test for the female and male target persons. Both times the test was significant (male: p = .001; female p = .009). Distributions are very similar.