Eye cues increase cooperation in the dictator game under physical attendance of a recipient, but not for all

In our experimental study, we investigated whether cues of being watched positively affect prosocial behavior in the dictator game when a recipient is physically present. Our final sample consisted of 211 female and 107 male students aged 18-35 years (mean=22.93 years, sd=3.35 years). Subjects were asked to play a one-shot dictator game in the role of the dictator with one confederate being present in the same laboratory room serving as the recipient in this economic game. For our study, the 318 participants were randomly assigned to one of three experimental conditions (eye image, image of flowers, control). After playing the dictator game, subjects were asked to complete a post-experimental paper-pencil questionnaire. The questionnaire covered socio-demographic variables, emotional ratings of eye images and flower images as well as a 3-item-scale on attitudes towards trust and voluntary cooperation. With respect to data analysis, we first tested for an increased probability of donation effect using a Chi2-test for nominal data with sharing money (0=nothing, 1=something) as the outcome variable and experimental group (eyes vs. flowers vs. control) as independent variable. As there were too few cases, we could neither fit a Generalized Linear Model nor examine the proportions of students sharing something rather than nothing for each sex of participants separately. We then tested for an increased mean donation effect using a General Linear Model with shared amount of money as the outcome variable and experimental group (eyes vs. flowers vs. control), gender of participants, and their interaction as predictors. As shared amounts of money were not normally distributed within experimental groups, which is typical for the dictator game, we also used a non-parametric Kruskal-Wallis H-test as well as Mann-Whitney U-tests for post hoc analysis to confirm the results from the General Linear Model analysis. In a next step, we incorporated attitudes towards trust and voluntary cooperation of participants (GSS-Prosocial-Index) as a covariate into our General Linear Model to analyze whether this confounding variable accounts for the observed effect. With respect to data from the questionnaire, we conducted independent samples t-tests to analyze whether participants exposed to the eye image during the experiment reported experiencing different positive or negative affect levels on the PANAS compared to students who were exposed to an image of flowers. As positive and negative affect scores were not normally distributed within the two experimental groups, we also applied a non-parametric Mann-Whitney U-test to confirm our findings. Finally, we compared the measures from the Self-Assessment Manikin (SAM) on valence, arousal, and dominance associated with the eye image and the image of flowers by participants applying dependent samples t-Tests and non-parametric Wilcoxon-tests