Receiver response to high-intensity courtship differs with courter status in spotted bowerbirds Ptilonorhynchus maculatus

Abstract

Understanding sexual communication requires assessing the behaviour of both the sender and the receiver. Receiver responses to sexual displays carry relevant information, but such signals or cues may be subtle and therefore technically challenging to investigate. Here, we focus on receiver body movements in response to high-intensity courtship in spotted bowerbirds (Ptilonorhynchus maculatus). Male bowerbirds perform a vigorous courtship choreography on dedicated display structures-bowers. Bower owners tolerate other non-territorial males at their bowers, yet the courtship displays of these so-called 'subordinate' males rarely result in successful copulations. Males that display at high intensity are preferred by females in this species, yet excessively aggressive displays may be threatening, hence scaring prospective mates away. In this study, we hypothesized that bower owners are better able to exhibit high-intensity movements without startling their audience compared with subordinate males. To address this question, we used a combination of behavioural coding and AI-based tracking of body movements, which allows precise spatial and temporal resolution for the study of subtle behavioural responses. Contrary to our predictions, we found that bower owners evoked stronger startle responses than subordinate males. We discuss these unexpected results and suggest further experimental approaches for future investigations.

Keywords: courtship; machine learning; signal receiver; startle response; visual display.

Figure 1.

Automatic detection of keypoints on the bird’s body. (a) Screenshot of example output video showing a bird inside the bower walls, with automatically detected keypoints marked with individual labels. Only a subset of the keypoints shown in the picture was used for the analyses presented in this study. (b) Graphical representation of the keypoints of interest for this study. The dashed silhouette of the bird indicates the change in its body position between two subsequent frames (‘Beak tip 1’ and ‘Beak tip 2’). The Euclidean distance between beak coordinates in subsequent frames is used to calculate the variable ‘Relative displacement’ (RD) (see text). ITJ, inter-tarsal joint.

Figure 2.

(a) Visual representation of the variable ‘relative displacement’ (red line) as a function of frame number in one example video of the duration of approximately 500 frames. A steep increase in relative displacement appears as a spike on the plot and indicates a startle response. Vertical black lines show the occurrence of display elements by the courting male. The position of the vertical line indicates the start frame of each behaviour. (b) Example of a window (light blue box) of a set number of frames that was used to isolate relative displacement values before (baseline) and after (post) the occurrence of display elements; the frame of occurrence was included in the calculation of both ‘baseline’ and ‘post’ segments.

Figure 3.

Density plots for bower owners (orange) and subordinate males (green) depicting the sampling distributions of 1000 Δ values calculated by bootstrapping random frames (window size = 10 frames) in the original dataset, compared with the observed mean Δ values (vertical dashed lines) after high-intensity (top) and low-intensity (bottom) display elements. Vertical solid lines with diamonds indicate the individual means. The mean Δ value for high-intensity elements in bower owners (dashed orange line) falls outside the 95% of the sampling distribution (p < 0.001).

Figure 4.

Results of the analysis of Δ values calculated using different window sizes (4–50) in subordinate males (green) and bower owners (orange). Results are shown separately for low-intensity elements (left panel) and high-intensity elements (right panel). Error bars depict for each window size the 2.5 and 97.5% limits of the distribution of Δ values calculated from randomly sampled frames. Observed mean Δ values (filled points) that fall outside of the 95% of the random distribution significantly differ from random. Solid lines connect the mean values of the random distributions.