The data were collected during August 2018 and 2019 in the eastern gulf of Shark Bay, Western Australia, where our research on Indo-Pacific bottlenose dolphins (Tursiops aduncus) has been carried out on a near-annual, seasonal basis (typically austral winter and spring) since 1982. We conducted playback experiments to male dolphins from three second-order alliances; the KS alliance (6 males), the PD alliance (5 males in 2018, 4 males in 2019) and the RR alliance (3 males). We utilised a within-subject experimental design, where male dolphins were subject to signature whistle playbacks of males from their second-order alliance and males from their third-order alliance.

We used a UAV (DJI Phantom 4 Pro+), with an integrated, gimbal-controlled camera (with a focal length of 8.8 mm and maximum resolution of 3,840 × 2,160 pixels and 23 frames/s), to obtain high quality visual records of how individual dolphins responded to the playback experiments. During each playback, the UAV was flown at altitudes (x̅: 34 m, range: 30-42 m, as determined by onboard sensors) to maximise our ability to detect behavioural changes but also maintain a relatively broad field of view. On playback of each stimulus, every effort was made to hold the UAV stationary above the animals in order to best assess behavioural responses.

Using QuickTime player (v10.5), we scored the following variables from the videos of each playback experiment: (i) time to response (seconds) – time from the start of the first whistle stimulus to any visible response by the subject animal(s) to the playback; (ii) response duration (seconds) - time from the subject animal(s) first response to the playback until returning to pre-playback behaviour; (iii) approach distance (metres) – distance from first response until returning to pre-playback behaviour, estimated using adult dolphin body lengths (2 m) and corroborated with measurements and notes taken in real-time from the research vessel; (iv) delta orientation (degrees) - relative change in heading on response to playback, estimated in degrees; and (v) orient to source (yes/no) - a binary measure of whether the subject animal oriented directly toward the playback source.

One observer (SJA) coded all 40 playback videos from the UAV footage, as well as the time-synchronised audio recordings from the hydrophones and voice notes (using Adobe Audition 13.0.3.60). A second observer (SLK), blind to playback treatments and the first observer’s scores, independently coded variables (i) through (iv) for all 40 playback videos: time to response; response duration; approach distance; and delta orientation. We then conducted an inter-observer reliability analysis using the intraclass correlation coefficient (ICC) for two-way models in the irr package in R 3.6.1 (R project for statistical computing; GNU project), and found strong agreement between observers (ICC = 0.976, P = < 0.0001, CI = 0.967 - 0.983). Given this parity, we used the data scores by observer one in all further analyses and provide all playback videos here.

The videos are all approximately two minutes in length with 60 seconds pre-playback and 60 second post-playback, with the playback occurring ~ 60 seconds into the video. The DJI Phantom 4 Pro+ stores videos in 4 GB chunks, if the two-minute playback period spanned two of these chunks then the playback will span two video files (totaling two minutes in duration). The playback number is visible in the file name, e.g., "PB1", and the corresponding data for that playback (focal ID, context, response measurements etc.) can be found in the Source Data file alongside the original article; King et al (2021) Cooperation-based concept formation in male bottlenose dolphins. Nature Communications.

Please note, there is no video for PB17, this is also noted in the Source Data file mentioned above. Acoustic files are stored with the first author and can be made available upon request.