Page 140 - Emotions through the eyes of our closest living relatives- Exploring attentional and behavioral mechanisms

Chapter 6
the likelihood of yawning using a binomial GLMM, entering condition and trigger and their interaction as fixed effects, and again subject nested in trial as random effect. In the second step of the model, we were interested in how the conditions and triggers affected yawning rates in those cases that at least one yawn occurred. To investigate this, we entered condition and trigger and their interaction as fixed effects and subject nested in trial as random effect using a negative binomial GLMM.
It is possible that the likelihood of yawning in the conditions is due to the stimuli somehow being arousing to the observers, complicating the interpretation of the underpinnings of CY (see e.g., (Paukner & Anderson, 2006). For instance, yawning often involves display of the canines, which may be arousing for the orangutans (Plavcan, 2001). Therefore, as a control analysis, we looked at self-self-scratching behavior as this is indicative of arousal in primates (Troisi, 2002). In a third hurdle model, we checked whether the likelihood of self-scratching is affected by condition (fixed factor), with subject nested in trial as random factor and using a binomial GLMM. In the second step of the model using a negative binomial GLMM with subject nested in trial as random factor, we investigated whether self-scratching rate was affected by condition, trigger, and their interaction as fixed factors only in those cases when self-scratching occurred.
In all analyses, we compared the models to their respective null-models (i.e., including only the random effects) and only report on significant values if the models and null-models differ significantly from each other (Forstmeier & Schielzeth, 2011). For post-hoc contrasts of interaction effects we report corrected p-values using Tukey-adjustments. Alpha was set to 0.05.
Results
In total, we witnessed 83 yawns across 8 individuals and 289 sessions (see Tables S2.1 to S2.3). First, we investigated the likelihood of yawning in the two conditions. We found a significant effect of condition; yawning was more likely to occur in the yawn versus the control condition (b=3.45, SE=1.06, p=0.001). Next, we compared the yawning rate between the two conditions in those cases that at least one yawn occurred, but this alternative model did not deviate significantly (c2(1) = 3.09, p = 0.079) from its respective null-model.
Assessing whether familiarity affects the occurrence of CY, we found a significant interaction effect of trigger (familiar, unfamiliar, avatar) with condition.
138