https://doi.org/10.5061/dryad.z8w9ghxkd

Authors: Brian C. Leavell, Hoover Pantoja-Sánchez, Viviana Vélez, Claire T. Hemingway, Kyle Wilhite, Wouter Halfwerk, Ximena E. Bernal
Dates of data collection: 2017-2019
Location of data collection: Gamboa, Panamá
Keywords: communication, male-male competition, predator-prey, foraging, vibratory cue, sexual selection

Audio Playback Files:

Audio Playback Files:

• “audio_call_airborne.wav”
• “audio_call_ripple.wav”
• “audio_swat_highintensity.wav”
• “audio_swat_lowintensity.wav”

Data  (variable descriptions shown in Scripts section below):

field_vibro_rec_data.csv

• Output from “d_wave.m” MATLAB script when processing laser doppler vibrometer audio files recorded in the field
• Measurements of call and swat ripples from the field

• Variables

  • Note: “NA” indicates that there was no value, as the measurement category did not pertain to file’s “rippletype”
  • male = unique ID for individual male túngara frog
  • rippletype = denotes cause of ripple (i.e. call- or swat-induced)
  • File = audio file name of individual sample
  • flicking_appendage = which appendage(s) the frog swatted with to create ripple in associated sample
  • midge_count_on_frog = number of midges on frog at moment swat began
  • swat_count = number of swats that generated ripple in sample

  • frog_orientation_to_laser = which side of frog faced towards the point on water surface being measured by the laser

    • A = anterior
    • L = left side (frog’s perspective)
    • R = right side (frog’s perspective)
    • P = posterior

  • swat_orientation_to_laser =  indicates whether the swatting appendage(s) faced towards (i.e. near) or away from (i.e. far) the point on the water surface being measured by the laser

  • frog_orientation_to_laser_radians = conversion of frog orientation to laser into radians (not used in analyses)

  • vel_max_mm_per_s = maximum velocity (mm/sec) of sampled ripple

  • vel_rms = RMS velocity (mm/sec) of sampled ripple

  • domin_freq_vel = dominant frequency (Hz) of sampled ripple (derived from velocity)

  • v_energy = energy of sampled ripple (derived from velocity)

  • disp_app_mm = maximum displacement of water surface (mm, peak-to-peak)

  • disp_rms = RMS displacement of water surface (mm)

  • domin_freq_disp = dominant frequency (Hz) of sampled ripple (derived from displacement)

  • disp_energy = energy of sampled ripple (derived from displacement)

experiment_calibrations.csv

• Output from “d_wave.m” MATLAB script when processing laser doppler vibrometer audio files recorded during experiment calibrations
• Measurements of call and swat ripple playbacks from experiment calibrations

• Variables  

  • date = night of experiment (trials and calibrations that extended into the early AM were still assigned to previous date)
  • Audio.file = sequence of samples
  • file = individual audio sample
  • treatment = one of 3 categories, low-intensity swat, high-intensity swat, or call rate. Threshold for "High" vs. "Low" categories of maximum velocity of swat ripple is based on the median estimate from in situ Bayesian GLMM (5.84 mm/s, 4.55-7.51 mm/s 95% CI).
  • speaker_LorR = 2 playback stations were used in the experiment, one on each side of the frog, to control for potential side biases
  • pre_post = indicates whether calibration audio was recorded pre- or post-experiment
  • vel_max_mm_per_s = maximum velocity (mm/sec) of sampled ripple
  • vel_rms = RMS velocity (mm/sec) of sampled ripple
  • domin_freq_vel = dominant frequency (Hz) of sampled ripple (derived from velocity)
  • v_energy = energy of sampled ripple (derived from velocity)
  • disp_app_mm = maximum displacement of water surface (mm, peak-to-peak)
  • disp_rms = RMS displacement of water surface (mm)
  • domin_freq_disp = dominant frequency (Hz) of sampled ripple (derived from displacement)
  • disp_energy = energy of sampled ripple (derived from displacement)
  • Heirarchical structure = date/pre_post/speaker_LorR/treatment/Audio.file/file

male_tungara_response_to_ripples.csv

• Behavioral data from rival response experiment

• Variables

  • Note: “NA” indicates that measurements were not made, either because there were no calls in the recording or if there was only 1 call in the recording and measurements that require at least 2 calls to be made (e.g. “call rate”) were not possible. In the columns “Call [#]”, “NA” is used to fill remaining cells in the row after the final call of the sequence for that recording.
  • date = night of experiment
  • frog = unique ID for individual male túngara frog

  • treatment = playback used for trial

    • "hi swat" = high-intensity swat ripple played back once every 2 seconds (no airborne playback)
    • "lo swat" = low-intensity swat ripple played back once every 2 seconds (no airborne playback)
    • "call only" = call ripple played back with airborne playback of a call (whine + 1 chuck), played back once every 2 seconds
    • "call and swat" = alternating call ripple (with airborne call; whine + 1 chuck) and high-intensity swat, once every 2 seconds (i.e. second 0 = call ripple and airborne call, second 2 = high-intensity swat, second 4 = call ripple and airborne sound, second 6 = high-intensity swat, ...)
    • "call and muted swat" = same as "call and swat" treatment, but the swat playback was muted (i.e. did not play), thus this treatment featured a call ripple with airborne call (whine + 1 chuck) once every 4 seconds

  • side = left (L) or right (R) indicates which speaker was played back during trial; two speaker locations were used to control for the effect of side biases (see methods for further details)

  • order_of_playback = denotes the # in sequence of trials; treatments were randomized across trials, so order_of_playback can be used to determine effect of treatment position across trials on male response

  • original_audio = audio file number

  • min = minute of trial (1 = silent playback, 2 = treatment playback)

  • renumbered = randomized placeholder number assigned to file to ensure that the researcher analyzing the file was blind to the condition

  • calls = number of whines in 1 minute

  • time = duration (seconds) from start of 1st call to start of last call (within 1 minute)

  • mincallrate = call rate (calls/time) within minute portion of trial

  • mincomplexity = chuckstotal/calls;  average complexity per call within minute portion of trial

  • chuckstotal = total number of chucks appended to whines within 1 minute

  • First10seccallrate = call rate within first 10 seconds of minute

  • First10seccomplexity = average complexity per call within first 10 seconds of minute

  • Last10seccallrate = call rate within first 10 seconds of minute

  • Last10seccomplexity = average complexity per call within first 10 seconds of minute

  • Maxcomplexity = greatest # of chucks appended to a whine within 1 minute

  • Call [#] = these column heads refer to the order of whines (calls) produced by a frog within the given minute; the value for a given column indicates the number of chucks appended to that whine; e.g. for a given frog, if the value for column “Call 3” is “2”, this indicates that the 3rd whine produced by the frog in the given minute had 2 chucks appended

mate_choice_experiment1.csv

• Behavioral data from female choice experiment; swatter vs. rival responding to swat ripples

• Variables

  • date: the date that a night of experiments began. (note: as these frogs are nocturnal, experiments began in the evening and would sometimes continue past midnight. in these cases, all experiments within the same night were assigned the date in which the night began.)
  • frog: unique ID of female túngara frog
  • swatter: the approximate call rate and complexity of a frog that swats every 2 seconds (i.e. 0.34 calls/second; 1.00 chucks/call) see data from Leavell et al. (2022) Am. Nat.
  • neighbor_highswat: median call rate and complexity of frogs that were exposed to a high-intensity swat ripple once every 2 seconds (i.e., 0.5 calls/second; 1.75 chucks/call)
  • latency: the duration from the start of the trial to (minutes.seconds; i.e. 1.22 = 1 minute, 22 seconds)
  • time: hour and minute in which trial took place. (hour.minute; i.e. 9.44 = 21:44 or 9:44pm)

mate_choice_experiment2.csv
• Behavioral data from female choice experiment; male not responding to swat ripples vs. male responding to swat ripples
bats.csv

• Variables

• date: the date that a night of experiments began. (note: as these frogs are nocturnal, experiments began in the evening and would sometimes continue past midnight. in these cases, all experiments within the same night were assigned the date in which the night began.)
• frog: unique ID of female túngara frog (note: restarts numbering each night).
• neighbor_control: the median call rate and complexity of the rival frog before being exposed to high-intensity swat ripples (i.e., 0.45 calls/second; 1.00 chucks/call)
• neighbor_highswat: median call rate and complexity of frogs that were exposed to a high-intensity swat ripple once every 2 seconds (i.e., 0.5 calls/second; 1.75 chucks/call)
• latency: the duration from the start of the trial to (minutes.seconds; i.e. 1.22 = 1 minute, 22 seconds)
• time: hour and minute in which trial took place. (hour.minute; i.e. 9.44 = 21:44 or 9:44pm)

• Behavioral data from bat foraging experiments

• Variables

  • bat: unique ID of Trachops cirrhosus individual

  • experiment: two-choice tests

    • 1v1: not used for these analysis. same as 1v2, but the swatter playback was 0.34 call/second, whine only (no chuck).

    • 1v2: same as "experiment 1" playback in mate_choice trials

      • i.   swatter: the approximate call rate and complexity of a frog that swats every 2 seconds (i.e. 0.34                     calls/second; 1.00 chucks/call) see Leavell et al. (2022) Am. Nat.
      • ii.  neighbor_highswat: median call rate and complexity of frogs that were exposed to a high-intensity swat ripple once every 2 seconds (i.e., 0.5 calls/second; 1.75 chucks/call)

    • 2: same as "experiment 2" playback in mate_choice trials

      • i.   neighbor_control: the median call rate and complexity of the rival frog before being exposed to high-intensity swat ripples (i.e., 0.45 calls/second; 1.00 chucks/call)
      • ii.  neighbor_highswat: median call rate and complexity of frogs that were exposed to a high-intensity swat ripple once every 2 seconds (i.e., 0.5 calls/second; 1.75 chucks/call)

  • choice: which speaker was attacked by bat. 0 = simple (lower call rate and complexity); 1 = complex (greater call rate and complexity)

  • feeding: experiments were spread across a maximum of 3 feeding sessions per night. 1, 2, 3 = 1st, 2nd, and 3rd feeding sessions of the night.

  • date: the date that a night of experiments began. (note: experiments began in the evening and would sometimes continue past midnight. in these cases, all experiments within the same night were assigned the date in which the night began.)

Scripts:

1_Ripple_analysis_insitu.Rmd

• Script for analysis of field recordings
• Requires “field_vibro_rec_data.csv”

2_Ripple_analysis_experiment_calibrations.Rmd

• Script for analysis of ripple calibration recordings
• Requires “experiment_calibrations.csv”

3_rival_males.Rmd

• Script for analysis of behavioral data from rival response experiment
• Requires “male_tungara_response_to_ripples.csv”

4_mate_choice.Rmd

• Script for analysis of behavioral data from female choice experiments
• Requires “mate_choice_experiment1.csv” and “mate_choice_experiment2.csv”

5_bat_foraging.Rmd

• Script for analysis of behavioral data from bat foraging experiments
• Requires “bats.csv”

6_audio_file_visualizations.Rmd

• Visualizations of ripple and airborne stimuli used for playbacks; shown in Figure 1 of manuscript
• Requires “audio_call_airborne.wav”, “audio_call_ripple.wav”, “audio_swat_highintensity.wav”, “audio_swat_lowintensity.wav”

d_wave.m

• Script for analysing field and calibration audio files recorded as output from laser doppler vibrometer

Methods (from manuscript):

Rival male experiment
In August 2017, we collected calling male túngara frogs 1 to 4 hours after sunset from breeding sites within 1.5 km of Smithsonian Tropical Research Institute facilities (Gamboa, Panamá; 9°07.0’N, 79°41.9’W). Frogs took part in behavioural experiments in a large outdoor flight cage (5×5×2.5 m) within 3 hours following capture and before midnight. Following protocols established by the American Society of Ichthyologists and Herpetologists (https://asih.org/animal-care-guidelines), after each experiment we weighed, measured and toe-clipped the frog (to prevent testing the same individual multiple times), and released him the following night at his collection location. The research reported here complied with University of Texas at Austin #s AUP-2015-00048, AUP-2017-00051 and AUP-2020-00034; STRI IACUC protocols 2017-0102-2020-A8, 2018-0411-2021; and República de Panamá permits SE/AO-1-17, SE/A-44-18, SE/A-40-19, SE/A 69-15 and SE/AH-2-6.
Field recordings. To sample the natural variation of swat and call ripples from calling male túngara frogs, we recorded ripples from six males (n = 2 in situ; n = 4 in a planter tray filled with water from the breeding site, placed by the original calling site). We measured the frequency, velocity and displacement of the water ripples with a digital laser vibrometer (LDV; Polytech PDV-100; Velocity = 20 mm/s, Low Pass = 22 kHz, High Pass = none). The laser was focused on a reflective marker floating on the water surface; its reflection is detected by the LDV, allowing measurement of the surface vibrational velocities. We recorded these measurements from the digital output of the LDV using a Marantz audio recorder (48 kHz sample rate, 24 bit). See Supplementary Materials for a full description of the recording setup. We analysed call ripples (52 samples from 6 males; range: min = 5, max = 13 per male) and swat ripples (66 samples from 6 males; range: min = 5, max = 15 per male) using the velocity measurements recorded from the LDV digital output. We also converted velocity data to displacement values for comparison with related studies that have used this variable to characterize call ripples in this species (e.g. [32,36]). All ripple characterization analyses were performed in MATLAB R2016a (https://www.mathworks.com) using custom code (Supplementary Materials).
The vertical velocity of a swat ripple differed depending on the location of the sensor relative to the side of the body that swats (Figure S1, Supplementary Materials and R scripts for details of the analysis). We therefore tested calling behaviour in response to both high- and low-velocity swat ripples to determine male responses to the range of swat ripple intensities encountered in the wild. We used the point estimate of the median maximum velocity of all recorded swat ripples (5.84 mm/sec. peak-to-peak), which was between the 95% credible intervals for the median low- and high-intensity swat values, as a guide to calibrate low- and high-velocity swat playbacks. While there was variation across playbacks, the vast majority of calibration measurements were within the range measured from the field recordings (see Figures S2-3 and associated R scripts for further details).
Experimental setup. For our ripple playback files, we used a total of three field recordings. These files represented the median dominant frequency and median maximum velocity of either a i) high-velocity swat, ii) low-velocity swat, or iii) call ripple. As playback of call ripples always co-occurred with an airborne call playback, we ensured that the call ripple accurately coincided with the airborne call component by synchronizing their playback to match the temporal profile of the field recording. For the airborne component of the multimodal call playback, we generated a call comprised of a whine with one chuck, as this is the most common call complexity observed in natural settings (see Supplementary Materials for details) [37,38].
Prior to an experiment, a male (n = 29 frogs; no males were from field recordings) was placed in a pool (2.6 × 2.1 m) of dechlorinated tap water. The male was contained within a ripple cage but could otherwise call freely. This ripple cage, used in a previous study of call-induced ripples, is acoustically transparent to low frequency ripples [33]. The cylindrical ripple cage (10.5 cm total diameter) consisted of a plastic, circular base and top held in place by three evenly spaced, vertical screws (~0.4 cm diam.). Monofilament fishing line (~0.36 mm thick) was strung vertically, encircling all sides of the plastic base and top (spaced ~0.25 cm apart creating an acoustically transparent cylinder 8 cm. in diameter) to prevent the frog from escaping. Once the male was consistently calling, he was presented with the first of four trials, which were presented in random order and separated by at least 1 minute of silence. The order of presentation did not affect male responses (Tables S1-2). Each trial consisted of a 1-minute silent control playback and 1-minute playback of one of the following treatments: i) high-intensity swats, ii) low-intensity swats, iii) multimodal calls (airborne calls and call ripples), and iv) multimodal calls alternating with high-intensity swats (Figure 1). Each stimulus was played back every 2 seconds, which is consistent with the average natural call rate [37] and within the natural range of swat rates [19]. For a subset of the males (n = 15) we performed an additional trial at the end of the experiment, playing back the multimodal call stimulus once every four seconds (i.e. “slow multimodal call”, Figure 1) to control for the reduced call rate of the treatment in which the multimodal call alternated with a high-intensity swat ripple.
For each frog (n = 29 frogs; # trials/frog: range = 1 to 5, median = 4; 106 trials total), we compared call rate (average # calls/sec) and complexity (average # chucks/whine) between control conditions (i.e., silent playback) and treatment playback. Males called during all silent control conditions (106/106 trials), but sometimes ceased calling (9/106 trials) or only produced a single call (2/106 trials) in response to a treatment playback. All 11 of these latter trials were excluded from call rate analysis, whereas only the 9 trials in which the male ceased calling were excluded from complexity analysis. We also excluded 2 treatment playbacks due to playback error or a sudden, loud stimulus from the environment. All acoustic analyses were performed in Adobe Audition CS6.

Bat foraging & female choice experiments
The effectiveness and consequences of calling strategies are ultimately defined by the responses of the receivers. We therefore investigated how frog-biting midges impact decision-making in foraging bats and female túngara frogs by modulating the calling behaviours of male frogs and their rivals. Bats (n = 5; 10-20 trials/bat per experiment) and female túngara frogs (n = 24; 1 trial/female per experiment) were presented with two-choice tests of airborne (no ripples) frog call playbacks to assess preferences (see Supplementary Materials for detailed descriptions of methods).
To evaluate relative preferences between a calling, swatting male and a rival responding to his swat ripples (i.e. the scenario examined in the male behavioural experiment), we broadcast calls of a male who intermittently swats – i.e. representative of the call rate and complexity of a frog that swats every 2 seconds; 0.34 calls/second, 1.00 chucks/call [19] – versus a calling rival male responding to high-intensity swat ripples (i.e. representative of the call rate and complexity of frogs that were exposed to a high-intensity swat ripple once every 2 seconds; 0.5 calls/second, 1.75 chucks/call). In a second two-choice test, to investigate if and to what extent responding to swat ripples alters receiver preferences relative to a typical, non-swatting male, we broadcast calls of a characteristic non-swatting, calling male (i.e. representative of the call rate and complexity of a rival frog before being exposed to high-intensity swat ripples; 0.45 calls/second, 1.00 chucks/call) and calling male responding to high-intensity swat ripples (same as in previously mentioned two-choice test). A third two-choice test was given to a single bat (n = 1; 20 trials), in which we broadcast calls of a male who intermittently swats, but does not append ornamental chucks (i.e. approximate call rate of a frog that swats every 2 seconds, but without ornamental chucks; 0.34 calls/second, 0 chucks/call) versus a calling rival male responding to high-intensity swat ripples (same as above).

Statistical analyses
For our analyses, we formulated generalized linear mixed models fit under a Bayesian framework using the brms package version 2.14.4 [39] and R version 4.0.3 [40]. We used weakly informative priors to ensure that draws from the prior could be from any hypothetical dataset [41]. In this way, these priors include probability mass around potential, but not implausible extreme values. We ran 4 chains with 3000 iterations each and discarded the first 500, resulting in 10,000 posterior samples. To ensure model convergence, we visually examined traceplots for good mixing of chains and confirmed that R ̂ < 1.1 for all parameters. We performed posterior predictive checks to assess model fit with the bayesplot package version 1.8.1 [42]. Models were compared by assessing expected log pointwise predictive densities with the loo package [43]. See Table S3 for details specific to each model. Effects were determined using the hypothesis function in brms, first with two-side tests to determine an effect and directionality and subsequently a post-hoc one-sided test. Comparisons of effects between treatments were determined by subtracting the posterior probability densities from one another.