Threatening joy: Approach and avoidance reactions to emotions are influenced by the group membership of the expresser

Abstract

It has been repeatedly stated that approach and avoidance reactions to emotional faces are triggered by the intention signalled by the emotion. This line of thought suggests that each emotion signals a specific intention triggering a specific behavioural reaction. However, empirical results examining this assumption are inconsistent, suggesting that it might be too short-sighted. We hypothesise that the same emotional expression can signal different social messages and, therefore, trigger different reactions; which social message is signalled by an emotional expression should be influenced by moderating variables, such as the group membership of the expresser. In two experiments, we show that group membership influences approach and avoidance reactions to emotional expressions: Emotions (fear and happiness) expressed by in-group members elicited concordant behaviour, whereas emotions expressed by out-group members activated the reverse pattern. A third experiment, in which participants directly evaluated smiling and fearful individuals resembling in-group and out-group members supported this result.

Keywords:

• Approach and avoidance reaction
• Facial expressions
• Group membership
• Emotion

The authors thank Leon Makelis for help with the data collection and Juliane Degner for helpful comments on an earlier version of this manuscript.

The research reported in this article was supported by grant [WE 2284/3] from the Deutsche Forschungsgemeinschaft to Dirk Wentura.

The authors thank Leon Makelis for help with the data collection and Juliane Degner for helpful comments on an earlier version of this manuscript.

The research reported in this article was supported by grant [WE 2284/3] from the Deutsche Forschungsgemeinschaft to Dirk Wentura.

Notes

¹ Different outlier criteria (i.e., lower and higher general cut-offs, individual cut-offs) led to the same results. We decided to use general cut-offs since previous research (Krieglmeyer & Deutsch, 2010) showed that this revealed the strongest effects.

² Please note that any result of this analysis is identical to a corresponding result of a 2 (emotion) × 2 (group membership) × 2 (movement) analysis with the raw reaction times as dependent variable (e.g., the two-way interaction of ethnicity and emotion is identical to the three-way interaction of ethnicity, emotion, and movement).

³ We conducted a pilot study with a more traditional minimal group design, in which participants were anonymously assigned to one of two arbitrary groups without receiving further information. Participants then learned the group membership of 16 white Caucasian men and women. In a subsequent approach/avoidance task using emotional pictures of these persons, a main effect of group emerged: Photographs of in-group members elicited stronger approach than avoidance behavior, whereas photographs of out-group members elicited stronger avoidance than approach behavior, F(1, 65) = 4.46, p < .05. However, no influence of emotional expression occurred, F < 1. We therefore hypothesized that a strong negative evaluation of the out-group might be necessary for the interaction between group membership and emotional expression to occur. Note that the in-group favoritism found in traditional minimal group designs is typically based on a positive evaluation of the in-group and not a negative evaluation of the out-group (e.g., Brewer, 1979; Otten & Wentura, 1999). Therefore, the result of this pilot study should be taken as an indication of discriminate validity of our approach (see also the General Discussion with regard to the main effect of group membership).

⁴ Essentially, the same results reported here were found with other general cutoffs higher than 1200 ms as well as with individual cutoffs.

⁵ Note that this F-test corresponds to a t-test for paired samples. A one-tailed test is therefore allowed (e.g., Maxwell & Delaney, 1990).

⁶ We did not specify a three-way interaction (i.e., a moderation of the two-way interaction by relevance) here. Besides the point that the three-way interaction is not essential for our argument, one can only expect a small effect for this interaction since it would be based on a significant interaction for the other-relevant traits and a non-significant interaction for the possessor-relevant traits. Power calculations for a small effect (f =.1), given α = .05 and 1 - β = .80 yielded N = 199.

⁷ Obviously, the following argument also applies to happy facial expressions. However, since happiness is the only facial expression clearly associated with positive valence, this could not be avoided.