Figure 7
The structure of vocal interactions mirrors that of the imposed network. (a) In an asymmetric network, the interactions tend to be asymmetric, and (b) in a symmetric network, they tend to be symmetric. (a) In pairs of asymmetrically connected birds, cross-covariance (CCV) functions (black lines, see “Technical details”) indicate unidirectional vocal exchanges revealed by unimodal peaks. Stacks of example spectrograms are shown (right) with the auditory stimulus (pb) presented in chamber T (top), T’s response broadcast to bird L (middle), and L’s response broadcast to R (bottom); corresponding rows in the 3 sub-panels are from simultaneous recordings. (b) In a symmetric hierarchical network, CCV functions reveal bidirectional vocal interactions. (a,b) The gray areas represent 3 standard deviations of a random shuffle predictor (see “Technical details”). (c) Normalized CCV functions in unidirectional (left), bidirectional (middle), and non-connected (right) bird pairs, n = 3 bird groups (blue in all 3 subpanels, orange in the middle and right subpanel, and yellow in the left subpanel) across a total of 5 different network configurations (two asymmetric, three symmetric). To allow for comparison among experiments, CCV functions were normalized by the shuffle predictor (gray areas; see “Technical details”), which sets the threshold for statistical significance at a fixed unit distance along the y axis.
