Fig 1.
Drum length predicts number of residents involved in defense.
Number of resident woodpeckers who engage in territorial defense varies based on the length of the simulated drum. Fisher’s exact test revealed that the proportion of solo responses varied based on drum length. Post-hoc tests showed that a resident is more likely to respond alone (solo defense) if the playback drum is short compared to both average length (** p < 0.01) and long (* p < 0.05) stimuli.
Fig 2.
Behavioral correlations of population level paired response to an average threat simulated intruder.
Population-wide correlations between defense behaviors performed during the 10-min defense bout in response to an average threat simulated intruder. A) Ranked behavior counts for an exemplar negative correlation. In this population and context, female flyovers and pik calls are generally not performed together in defense (rho = −0.43, p < 0.001). B) Ranked behavior counts for an exemplar positive correlation. In this population and context, female pik calls and male crest raises are generally performed together during a 10-min defense bout (rho = 0.57, p < 0.001). C) All colored squares indicate significant behavior correlations, where dotted green is a positive correlation and dotted red is a negative correlation. For exact correlation coefficients and p-values see supplement (S1 Table). Starred (*) boxes indicate a behavior pair correlation that is sometimes positive and sometimes negative, depending on the threat context (see Fig 4 for other contexts). Grey hatched squares represent non-significant correlations. Strikethrough behavior names indicate that behavior was not performed in response to an average simulated intruder. The orange box encompasses correlations that include behaviors performed by both the female and male resident.
Fig 3.
Behavioral network shows population level paired response to an average threat simulated intruder.
Network analysis and clustering pipeline reveals distinct patterns in how resident pairs organize their defense. A) Territorial network graph where each shape (node) represents a behavior and the lines (edges) connecting nodes depict significant correlations. Solid green lines indicate positive correlations and dashed pink lines indicate negative correlations. Line width (edge weight) indicates the relative strength of the correlation (Spearman’s rho). B) Same territorial network plotted along eigenvector 1 (x-axis) and eigenvector 2 (y-axis) of the signed Laplacian matrix. Distance between nodes in eigenspace is informative, such that closer nodes are more similar in performance and more distant nodes are less similar. Nodes are jittered slightly for visualization. C) Motifs of related behaviors determined through spectral clustering of the full network. Each motif is composed of behaviors that are likely to be deployed together during a single territorial defense bout. Arrows indicate significant sequential transitions between behaviors, indicating that one behavior is likely to follow another x% of the time. Dashed arrows indicate that the transition likelihood was relatively low (<10%), but still significantly different than the null expectation. Distance in the motif panel is not informative.
Fig 4.
Behavioral correlations show contextual variation in paired defense.
Patterns of behavioral correlations in paired territory defense vary across multiple threat contexts. All colored squares indicate significant behavioral correlations, where dotted green is a positive correlation and dotted red is a negative correlation. For exact Spearman’s rho correlation coefficients and p-values see supplement (S1 Table). Starred (*) boxes indicate a behavior pair correlation that is sometimes positive and sometimes negative, depending on the threat context. Grey hatched squares represent non-significant correlations. Strikethrough behavior names indicate the behavior was not performed in response to that simulated threat context. The orange box encompasses correlations that include behaviors performed by both the female and male resident. A) Population-level correlations for pair responses to simulated High Threat intruder (long/fast drum). B) Population-level correlations for pair responses to simulated Mixed Threat intruder (short/fast drum). C) Population-level correlations for pair responses to simulated Mixed Threat intruder (long/slow drum). D) Population-level correlations for pair responses to simulated Low Threat intruder (short/slow drum).
Fig 5.
Paired territorial response networks vary across simulated threat contexts.
Network graphs capture population-level behavioral patterns for paired defense in response to simulated territorial intruders. All edges (lines) indicate significant, non-random correlations between behaviors (nodes/shapes) where solid green is positive and dashed pink is negative and line width indicates the strength of the correlation (Spearman's rho). Paired defense networks are plotted along eigenvector 1 (x-axis) and eigenvector 2 (y-axis) of the signed Laplacian matrix and distance in these planes is informative such that closer nodes are more similar and more distant nodes are less similar. Nodes are jittered slightly for visualization. We have plotted each network on just two eigenvectors for readability. However, each of these paired defense networks was clustered using 3 or 4 eigenvectors for defining motifs (see Methods for more information and see S1 Fig for eigengap plots for each context). Population level networks for paired defense in response to simulated intruders of: A) High Threat (long/fast drum), B) Mixed Threat (short/fast drum), C) Mixed Threat (long/slow drum), and D) Low Threat (short/slow drum).
Fig 6.
Paired territorial response networks cluster into diverse behavioral motifs.
Within the larger network graph, motifs are subgraphs of mostly positively correlated behaviors, with few negative correlations within motifs (as determined by spectral clustering). Line width is indicative of the strength of the correlation between two behaviors; however, line length and relative position of nodes are NOT informative in this figure. Arrows show significant sequential transitions between two behaviors, with the number indicating the transitional probability. Dashed arrows are significant transitions that occur less than 10% of the time. Clustered motifs for population-wide paired defense responses to simulated intruders of: A) High Threat (long/fast drum), B) Mixed Threat (short/fast drum), C) Mixed Threat (long/slow drum), and D) Low Threat (short/slow drum).
Fig 7.
Behaviors can fill a role in multiple defense motifs.
Heat map showing how many motifs each behavior clusters into within the population-level paired response networks. Behaviors that cluster into more than one motif may show functional variability within the broader defense landscape. The color corresponds to the number of motifs that a given behavior (column) clusters into for each threat context (row).
Fig 8.
Population level solo responses to short drum stimuli.
Population-wide correlations between defense behaviors performed during 10-min defense bout in response to short drum stimuli reveal distinct patterns in how residents organize solo defense. The top row of panels is solo response to Mixed Threat (short/fast) stimuli, the bottom row is solo response to Low Threat (short/slow) stimuli. A) All colored squares indicate significant behavior correlations, where dotted green is a positive correlation and dotted red is a negative correlation. For exact correlation coefficients and p-values see supplement (S1 Table). Starred (*) boxes indicate a behavior pair correlation that is sometimes positive and sometimes negative, depending on the threat context. Grey hatched squares represent non-significant correlations. Strikethrough behavior names indicate the behavior was not performed in that context. B) Solo defense network graphs capture population-level behavioral patterns. All edges (lines) indicate significant, non-random correlations between behaviors (nodes/shapes) where solid green is positive and dashed pink is negative, and line width indicates the strength of the correlation. Solo networks are plotted on eigenvector 1 (x-axis) of the signed Laplacian matrix because they were only clustered on eigenvector 1. Distance between nodes is informative only in the horizontal (x) direction, position in the vertical (y) direction is solely for visualization. Nodes are purple because these represent female and male responders. Nodes are jittered slightly for visualization. C) Motifs of related behaviors determined through spectral clustering of the full network. Each motif is composed of behaviors that are likely to be deployed together during a single territorial defense bout. Arrows represent significant sequential transitions between behaviors, indicating that one behavior is likely to follow another x% of the time. Distance in the motif panel is not informative.
Table 1.
Ethogram summary of observed downy woodpecker response behaviors.