Data for "Vocal convergence during formation of cooperative relationships in vampire bats"

If you are interested in analyzing these data, please contact Gerald Carter, socialbat.org, [e-mail] R code for analyzing these data is available on github here: https://github.com/jkvrtilek/call-convergence-2025 Data are measures of contact calls from 94 individually isolated common vampire bats"acoustic data/ds_completespecan_2023-10-02.RDS" is 27 acoustic measures from the R package warblr"acoustic data/ds_fundfreq_summary_2023-11-06.RDS" is other acoustic measures of fundamental frequency"vampire_call_measures.csv" is the compiled and cleaned dataset that coverts seconds to ms and removes sounds that are not calls by selecting calls where these criteria were true: duration is > 3 and < 50 ms, peakf >10 and < 30, time.Q25, time.median, time.Q75 and time.IQR must all be greater than zero."vampire_call_measures_transformed.csv" is the same dataset with transformed measures. See github for more informationTo record contact calls with high signal-to-noise ratios, we isolated an individual bat inside a plastic storage bin (about 75 to 200 liters in volume) lined with acoustic foam and/or soft fabric to dampen echoes. To keep the bat within 10 to 30 cm of the microphone, we placed it inside a small soft-mesh butterfly cage or a tube of plastic mesh. Calls were recorded with an Avisoft CM16 ultrasound condenser microphone (frequency range 10-200kHz, Avisoft Bioacoustics, Berlin, Germany). Sound was digitized with 16-bit resolution at sampling rates ranging between 250-500 kHz through an Avisoft UltraSoundGate (one-channel or four-channel) to a laptop running the program Avisoft Recorder. To remove any effects of digitization rate on subsequent call measures, recordings digitized at a rate above 250 kHz were downsampled to 250 kHz using the seewave R package. We identified start and end times of individual calls with a custom R script that uses both amplitude and spectral density to segment audio recordings into calls. The script excludes calls which exceed the maximum amplitude of the microphone. This automated method was tested and validated with 1535 manually labeled recordings containing ~7000 call selections. Using these start and end times, we then measured 35 features of each call. We used the R package warbleR to measure 27 spectral and temporal parameters and the R package soundgen to estimate 8 additional measurements from the fundamental: maximum slope, minimum slope, and absolute minimum slope; the number of positive slopes and of negative slopes; the number of turns; the mean slope; and the number of segments. All scripts were run using the Ohio Supercomputer.