Factor analysis of behavioral traits
We registered 1217 behavioral events. A comprehensive check list of 20 behavioral traits displayed by the lizards during the male tournaments is presented in Table 1.
Statistical analyses revealed that the frequency of these two behaviors in the dyadic trials (i.e. 42.8% vs. 23.8%) in the lab did not significantly differ (χ2 = 1.71, P = 0.19). However, there were significant differences in the distance between opponents (U = 762, P < 0.0001 – lateral compressions were displayed at closer distance, see above), and in the likelihood of subsequently eliciting an aggressive response from the opponent (χ2 = 21.24, P < 0.0001). Lizards were significantly more likely to receive an aggressive response after performing a lateral compression than after push-ups. In contrast, there was no significant difference in their duration (U = 471.5,P = 0.70).
The factor analysis performed on the 20 behavioral traits extracted five factors that together explained 62.78% of the variance (Table 2). Factor 1 explained 22.05% of the variance in behavioral traits (eigenvalue = 4.41) and the factor loadings indicated that push-ups, lateral compression, and lunge behaviors were the most important behavioral traits in this factor. Factor 2 explained 18.34% of the variance (eigenvalue = 3.67) and climbing on top of the rock was the most important behavioral traits in this factor. However, tongue flicks (0.77) and rubbing the base of the tail onto the substrate (0.65) also had high correlation coefficients (Table 2). Factor 3 explained 8.68% (eigenvalue = 1.73) with side hop and retreat representing the most important behavioral traits in this factor. Factor 4 explained 7.58% (eigenvalue = 1.51) where tail waving was the most important behavioral traits in this factor. Finally, factor 5 explained 6.11% (eigenvalue = 1.22) where exploring the arena and trying to climb the walls was the most important behavioral trait.
The comparison of the SELF and the RIVAL models explaining Factor 1 suggested that the RIVAL model was more informative (AICc = -22.84) than the SELF model (AICc = 47.12). Opponent’s factors 1 (importance = 0.92,β = 0.14, z = 4.79, P < 0.001), 3 (importance = 1.00, β = 0.22, z = 8.26, P < 0.001), and 4 (importance = 0.98, β = 0.07, z = 3.34, P < 0.001) had positive effects on Factor 1 of the focal male (Table 3 and Fig. 1). Thus, the rivals’ behaviors summarized in Factors 1, 3, and 4 increased the intensity of the aggressive behavior of the focal males, although the behavioral traits summarized in Factor 4 may contribute very little to explain the variance in Factor 1 of focal lizards (Fig. 1d). Meanwhile rivals’ Factor 2 (importance = 0.79, β = -0.05, z = 2.33, P = 0.020) and the chroma of their blue patches (importance = 0.82, β = -0.76, z = 2.08, P = 0.037) were negatively related to Factor 1 (Table 3 and Fig. 2). Thus, rivals’ behaviors summarized in Factor 2 and the chroma of their blue patch reduced the aggression intensity of focal males, although these behavioral traits may contribute very little to explain Factor 1 of focal males.
Interestingly, the SELF model suggested that there was a significant effect of haemococcidians and tick load on Factor 1. The effect of the infection by the haemococcidian genus Lankesterella was negative (importance = 0.88, β = 0.08, z = 2.70, P = 0.007); infected lizards displayed less push-ups, lateral compressions, and lunge behaviors (Fig. 3a). The number of ticks had a positive effect on Factor 1 (importance = 0.73, β = 0.05, z = 2.18, P = 0.028) (Table 3). However, this positive relationship was only at low levels of infestation (< 4) and not when lizards had more ticks (> 6) (Fig. 3b).