Figure 4.

Mean estimates and 95% confidence interval of best predictors of vocal expression: number of grunts (A,B), grunt rate (C), grunt duration (D–F) and acoustic spectral score PCac. (G–I), depending on stimulus cyan for object, dark blue for human, location in the room, behavioral proximity and time during the test. Best predictors are represented for illustrating the range and directions of effects. Location (A,D,G): whether the pig was located close to the stimulus (solid lines) or away from it (dotted lines). Behavioral proximity score (B,E,H) was scaled for the statistical analysis so the Z score is represented (see composition of scores in Table 6). (A–D,G) Type of stimulus: whether the reunion was with the human (dark blue solid circles and lines) or the object light blue empty circles and dotted lines. Time during the test (F,I) is quantified by the interval index during the test and is scaled, so the Z score is represented. Plots were generated using the averaged best model resulting from the model selection (models having delta AICc below 2 and predictors having a weight of 1), for which the estimates (se) are in Table 8, the full selection model table is available in Supplementary Table 7.