Table 4. Model selection statistics and relative influence of the predictive variables on the variation of the observed behaviours.

Response variables	GLM	Subsets of models	AICc	ΔAICc	VIF
Rate of social behaviours	Best model 1	Juvenile Social Exposure + Sex	143.27	0.00	≤1.04
	Full model 1	Sex + Life Experience + Species + Rearing + Infancy Social Exposure + Juvenile Social Exposure + Estimated Current Age + Estimated Age at Arrival	187.82	44.52	≥10
Rate of anxiety behaviours	Best model 2	Estimated Age Arrival + Species + Sex	142.29	0.00	≤1.33
	Full model 2	Sex + Life Experience + Species + Rearing + Infancy Social Exposure + Juvenile Social Exposure + Estimated Current Age + Estimated Age at Arrival	184.63	42.34	≥10
Rate of abnormal	Best model 3	Rearing + Sex	146.43	0.00	≤1.03
	Full model 3	Sex + Life Experience + Species + Rearing + Infancy Social Exposure + Juvenile Social Exposure + Estimated Current Age + Estimated Age at Arrival	194.25	47.82	≥10
EloRating	Best model 4	Sex	144.60	0.00	0.00
	Full model 4	Sex + Life Experience + Species + Rearing + Infancy Social Exposure + Juvenile Social Exposure + Estimated Current Age + Estimated Age at Arrival	191.36	46.76	≥10

Note. The Best Models in bold have some significant results at p < 0.05. AICc = Akaike’s information criterion corrected for small sample sizes. VIF = variance inflation factor. We only select models with a VIF < 5.