Page 214 - Emotions through the eyes of our closest living relatives- Exploring attentional and behavioral mechanisms

conditions and triggers. The model including the interaction between condition and trigger did not significantly improve the null model (c2(3) = 3.50, p = .321).
Testing the link between contagious yawning, condition and familiarity using self- scratching as covariate
To control for self-scratching in our models that investigate the presence and strength of contagious yawning, we perform two Hurdle models, each containing two analyses. First, using a binomial GLMM with Subject nested in Trial as random factors and condition and self-scratch occurrence as fixed factors, we find a main effect of condition on the likelihood of yawning (b = 3.52, SE = 1.05, Z = 3.35, p = .0008), but no effect of self-scratch occurrence (b = 1.33, SE = .96, Z = 1.39, p = .163). In those cases that at least one yawn occurred, is the yawn response larger in the yawn condition versus the control condition? To answer this question, we perform a negative binomial GLMM with Subject nested in Trial as random factors and condition and self-scratch rate as fixed factors and compare this model to the null model (without fixed factors). The result shows that the alternative model cannot explain the data better than the null model: c22(2) = 3.32, p = .191; there is no main effect of condition (b = .47, SE = .26, Z = 1.80, p = .071), nor a main effect of self-scratch rate (b = .03, SE = .06, Z = .48, p = .634) on yawn rate.
In the second hurdle model, we look at effects of familiarity on the occurrence and strength of contagious yawning. First, using a binomial GLMM with Subject nested in Trial as random factors and condition, trigger, condition*trigger, and self-scratch occurrence as fixed factors, we find a main effect of condition and condition*trigger on the likelihood of yawning, but no effect of self-scratch occurrence (b = 1.33, SE = .96, Z = 1.39, p = .163). Specifically, yawning is more likely to occur in the yawn versus control condition in case of a familiar trigger (b = -6.50, SE = 1.64, Z = -3.97, p = .0001) and an unfamiliar trigger (b = -3.93, SE = 1.57, Z = -2.50, p = .012). Next, in those cases that at least one yawn occurred, is the yawn response larger in the yawn condition versus the control condition and does familiarity affect this result? To answer these questions, we perform a negative binomial GLMM with Subject nested in Trial as random factors and condition, trigger, condition*trigger, and self-scratch rate as fixed factors and compare this model to its respective null model. The result shows that the alternative model cannot explain the data better than the null model: c22 (6) = 5.30, p = .505; there is no main effect of condition (b = .18, SE = .62, Z = .28, p = .776), nor a main effect of self-scratch rate (b = .04, SE = .07, Z = .58, p = .560), nor an interaction between condition*trigger (b = .19, SE = .76, Z = .26, p = .798) on yawn rate.
212