Biomechanics of the peafowl ’ s crest : a potential mechanosensory role for feathers during social displays 2

22 Feathers act as vibrotactile sensors that can detect mechanical stimuli during avian flight and tactile navigation, suggesting that they may also function to detect signals during social 24 displays. We explored this novel sensory modality using the crest plumage of Indian peafowl (Pavo cristatus). We first determined whether the airborne stimuli generated by peafowl 26 courtship and social displays couple efficiently via resonance to the vibrational response of feather crests from the heads of peafowl. Peafowl crests were found to have fundamental 28 resonant modes with frequencies that could be driven near-optimally by the shaking frequencies used by peafowl performing train vibrating displays. Crests also were driven to vibrate near 30 resonance when audio recordings of sounds generated by these displays were played back in the acoustic near-field, where such displays are experienced in vivo. When peacock wing-shaking 32 courtship behaviour was simulated in the laboratory, the resulting directional airflow excited measurable vibrations of crest feathers. These results suggest that peafowl crests have properties 34 that make them suitable mechanosensors for airborne stimuli generated during social displays. Such stimuli could complement acoustic perception, thereby enhancing detection and 36 interpretation of social displays. Diverse feather crests are found in many bird species that perform similar displays, suggesting that this proposed sensory modality may be widespread, and 38 possibly derived from flow sensing in other contexts. We suggest behavioral studies to further explore these ideas and their functional implications. 40 . CC-BY-NC-ND 4.0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/197525 doi: bioRxiv preprint first posted online Oct. 2, 2017;


ABSTRACT 22
Feathers act as vibrotactile sensors that can detect mechanical stimuli during avian flight and tactile navigation, suggesting that they may also function to detect signals during social 24 displays. We explored this novel sensory modality using the crest plumage of Indian peafowl (Pavo cristatus). We first determined whether the airborne stimuli generated by peafowl 26 courtship and social displays couple efficiently via resonance to the vibrational response of feather crests from the heads of peafowl. Peafowl crests were found to have fundamental 28 resonant modes with frequencies that could be driven near-optimally by the shaking frequencies used by peafowl performing train vibrating displays. Crests also were driven to vibrate near 30 resonance when audio recordings of sounds generated by these displays were played back in the acoustic near-field, where such displays are experienced in vivo. When peacock wing-shaking 32 courtship behaviour was simulated in the laboratory, the resulting directional airflow excited measurable vibrations of crest feathers. These results suggest that peafowl crests have properties 34 that make them suitable mechanosensors for airborne stimuli generated during social displays. Such stimuli could complement acoustic perception, thereby enhancing detection and 36 interpretation of social displays. Diverse feather crests are found in many bird species that perform similar displays, suggesting that this proposed sensory modality may be widespread, and 38 possibly derived from flow sensing in other contexts. We suggest behavioral studies to further explore these ideas and their functional implications. 40 INTRODUCTION two criteria hold: when the wavelength exceeds R (R e.g., this can occur for low 88 frequencies characteristic of locomotion or motions during displays when receivers are nearby the sender), or when the source (e.g., shaking appendages such as wings, trains or tails) size is 90 comparable to or exceeds R (R A; e.g., this can occur when the receiver is close to the sender compared to the size of shaking appendages such as wings, trains or tails). In the near-field 92 regime, particle velocity dominates at very small R and the decrease in both pressure and velocity depends on exact source characteristics. While there is no hard distinction between the 94 near-and far-fields, the pressure contribution gradually increases relative to the velocity field as the far-field limit is approached. Research on low-frequency vibrational communication mostly 96 has focused on substrate-borne signals and relatively few studies have considered near-field vibrotactile reception of near-field air-borne signals. Near-field communication has been studied 98 in aquatic animals, including crustaceans, fish, and whales (Bradbury and Vehrencamp, 2011; Butler and Maruska, 2016; Mooney et al., 2016), as well as a wide variety of invertebrate taxa 100 (Markl, 1983). In arthropods, near-field airborne signals are detected via tactile as well as auditory means, and many species use filiform hairs to detect near-field air velocity for predator 102 or prey detection and, in some cases, intraspecies signaling (Barth, 2014;Santer and Hebets, 2008). It is not yet known whether birds also use non-auditory senses to detect near field 104 velocity (airflow patterns) during social displays, or what influence this may have on their social interactions. 106 One possible means by which peafowl may sense potential near-field signals is the fan-like crest, 108 a planar array of feathers oriented in the sagittal plane that is found on the heads of both male and female peafowl (Dakin, 2011). Each crest feather has a long shaft (rachis) with short, sparse 110 barbs along its length, and a spatula-shaped "flag" of pennaceous vanes at the distal end ( Fig.  1). Although it has been proposed that the peafowl crest may serve as a signal of status (Dakin,112 2011), the crest feather morphology is similar to that of filoplumes, a type of feather with known mechanical sensitivity (Alibardi, 2009;Lucas and Stettenheim, 1972;Seneviratne and Jones, 114 2008) that protrude on the face and head of many bird species. This suggests that the crest may have a somatosensory function. Moreover, because the peafowl's region of most acute vision is 116 oriented laterally (Hart, 2002), when a peahen gazes at a displaying male, the maximum area of her crest feathers also points toward the peacock's moving feathers. This results in the optimal 118 orientation for responding to air motions generated by the displaying bird (Fig. 1A). 120 With this background in mind, we explored the biomechanics of the peafowl crest and its potential role as a sensor during peacock displays. We would expect bird feathers to have certain 122 biomechanical properties in order for them to function effectively as tactile airflow sensors. In particular, they should vibrate efficiently at socially salient frequencies, either to detect shaking 124 by a conspecific individual, or as a form of proprioception to provide feedback to the animal doing the shaking (Dambach et al., 1983;Kämper and Dambach, 1981). This can be 126 accomplished readily via mechanical resonance, the phenomenon whereby an object responds with maximum amplitude to a driving force that oscillates at one of its natural frequencies of 128 vibration (Smith, 2010 interactions between feathers can influence the resonant frequency and damping, we also compared the biomechanics of crests to that of isolated feathers (Cummins and Gedeon, 2012). 136 To test whether mechanical sound might cause crest motion in females located in the near-field of train-rattling peacocks, we measured deflections of crests that were exposed to audio 138 playbacks in the laboratory of train-rattling and a white noise control. 140 Next, we considered the peacock's wing-shaking display. Because avian wing flapping during flight is known to shed vortices periodically, we hypothesized that the wing-shaking display 142 would also result in periodic airflow disturbances that could drive significant crest feather motion. To test this hypothesis, we used a peacock wing-flapping robot to visualize the response 144 of peafowl crests to airflow produced during simulated wing-shaking displays. Together, these experiments provide a first step to evaluating the potential mechanosensory responses of the 146 avian crest during social signalling. We discuss how this can be followed with further behavioral experiments on live animals. 148

MATERIALS AND METHODS 152
In vitro samples All measurements and fitted values are reported as means [95% confidence interval] unless noted 154 otherwise. A total of n = 7 male and n = 8 female Indian peafowl (Pavo cristatus Linnaeus 1758) head crests with the feathers still mounted in skin were purchased from commercial 156 vendors. Length and width measurements were made by hand and from digital photographs of crests and high-resolution scans of single feathers with a ruler included in the sample plane ( Fig.  158 1). All male crest samples had feathers of uniform length (±8%), whereas the female crest samples had, on average, 7.0% [2.1, 11.8] of their feathers appreciably shorter than the 160 maximum length of the crest. Eight out of 15 crests had all feathers oriented in the same plane within ±5º; five of the crests had 7-11% of the feathers unaligned, and two male crests had 22% 162 and 50% unaligned feathers, respectively. We used photographs from a previous study (Dakin, 2011) that included a scale to compare the morphology of in vitro samples here to that of the 164 crests on live birds. The morphological traits compared included length, width and number of feathers. 166 For mechanical testing, we glued the lower side of the crest skin to a ~2.5 cm cube of balsa wood 168 using hot glue (Fig. 1C). An earlier study that compared the resonance of peacock feathers mounted using rigid balsa wood mounts versus a compliant gel found that the compliant mounts 170 resulted in only slightly lower resonant frequencies and reduced amplitudes at frequencies > 50 Hz (Dakin et al., 2016). If the crest feathers were closely clustered, the attached skin was first 172 softened in water and the crest was spread to approximate its natural configuration. To study individual, isolated feathers, we removed all but three feathers (one on each outer edge and one 174 in the middle) from two male and two female crests, and analyzed the characteristics of those remaining feathers. 176

Vibrational dynamics trials 178
For vibrational dynamics measurements, the feather assembly was mounted on a model SF-9324 mechanical shaker (Pasco Scientific, Roseville, CA, USA) driven by an Agilent 33120A function 180 generator (Agilent Technologies, Wilmington, DE, USA) (Fig. S1). Two orthogonal directions of the driving force were used: "out-of-plane", oriented normal to the plane of the crest; and "in-182 plane", oriented parallel to the plane of the crest and in the posterior-anterior axis of the head (Fig. 1C). The first orientation (out-of-plane) corresponds to the geometry when a peafowl either 184 visually fixates the display by orienting one side of the head towards it, or else drives its own crest into vibrations by performing a train-or tail-rattling display (Dakin et al., 2016). The 186 second (in-plane) orientation recreates the geometry when the front of the head is oriented towards the display or when the bird bobs its head during feeding or walking. The vibrational 188 response spectrum was measured using three linear frequency sweeps with rates validated by an earlier study of peafowl displays and feather vibrational properties ( Hanning filter. This yielded the magnitude, A, of the fast Fourier transform (FFT) at each vibrational drive frequency, fd, which was divided by the shaker drive magnitude, Ad, at that 206 frequency to give the drive transfer function (A/Ad). Finally, the drive transfer function was smoothed over a 1.3 Hz window using a cubic Savitzky-Golay filter and all peaks in the response 208 were fit to a Lorentzian function using nonlinear-least squares fitting to obtain the resonant frequency, fr, and full-width-half-maximum, Δf, of the spectral power (Smith, 2010). These fits 210 were performed in Origin 8.6 (Originlab, Northhampton, MA, USA). The quality factor, Q (a measure of how sharply defined in frequency the resonance is), was computed from Q = fr/Δf. 212

Audio playback experiments and analysis 214
To determine if peafowl crests move detectably due to the near-field airflow of train-rattling vibrations, we filmed high-speed video of one female and one male feather crest in the near-field 216 of a speaker playing audio recordings of peacock train-rattling displays. Two types of playback stimuli were used: (1) three different train-rattling sequences recorded in the field in a previous 218 study ( Dakin et al., 2016) from three different displaying peacocks, with mean rattle frequencies of 25.0 ± 1.0 Hz; 25.5 ± 0.6 Hz; and 24.6 ± 0.8 Hz; and (2) white noise generated by Audacity 220 (audacityteam.org; downloaded June, 2017). To ensure that playbacks were in phase over several seconds, sequences lasting approximately 1.2 s long were edited to contain an integer 222 number of rattling periods, and combined to make up a 30 s long audio file. 224 Audio recordings were played on a personal computer and amplified using a model 402-VLZ3 mixer (Mackie, Woodinville, WA, USA) and model 120 servo amplifier (Samson Technologies,  226 Hicksville, NY, USA). An earlier analysis of peacock train-rattling mechanical sound indicated that these noises were broadband rattles emitted at a rate of ~26 Hz, as opposed to sound waves 228 with spectral density predominantly in the infrasound (Dakin et al., 2016). We consequently used a model MR922 speaker (JBL Professional Products, Northridge, California) with a 230 broadband response that had a 30 cm diameter low frequency driver mounted in an acoustically absorbing enclosure (± 10 dB over 60 Hz-17 kHz). Two crest samples (one male and one 232 female) with resonant responses determined in the vibrational dynamics trials were studied by remounting each crest on a 0.64 cm thick square of plywood attached to a force plate. These 234 samples were positioned 30 cm away from the 30 cm diameter speaker face to ensure that the samples were in the near-field. To confirm that the broadband nature of the audio resulted in no 236 variation in intensity due to near-field interference, we measured average sound pressure levels (SPL) near the speaker using a model JTS1357 sound level meter (range: 31.5 Hz-8.5 kHz; ± 2 238 dB accuracy; A-weighting) (Sinometer, Shenzhen, China). No variation was found within measurement error (± 0.3 dB SPL) at five locations across the speaker's face vertically and 240 horizontally and perpendicular to the speaker face. Measurements taken when no audio was playing found audible frequency background SPL values of 54 ± 0.1 dB. 242 Relatively few values of the sound intensity generated by bird wing-flapping have been reported 244 in the literature to use as references for this experiment. Peacock wing-shaking sound levels for frequencies ≤ 20 Hz were reported as 73-79 dB SPL at 4 m, which extrapolates to approx. 79-85 246 dB SPL at 1 m using far-field scaling (Freeman and Hare, 2015), whereas audible bird wing-beat sound levels for much smaller species were reported at 64-66 dB SPL and 54-60 dB SPL at 1 248 kHz and 25 kHz, respectively, at 1.2 m for Eastern phoebes (Sayornis phoebe) and chickadees values with previous work, we Fourier-analyzed a recording of the playback made with a Sennheiser ME-62 microphone (±2.5 dB: 20 Hz to 20 kHz; Sennheiser, Wedemark, Germany). 254 This indicated that the component of the power spectrum of the playback near the crest resonance was only 3.5-11% of the total playback power. Thus, while sound levels measured for 256 the audio playbacks in the human audible range were approx. 90-97 dB SPL, we estimate that the component due to frequencies near resonance were much lower, approx. 75-87 dB SPL (-10 log 258 (3.5 to 11%)). 260 To minimize direct mechanical coupling via the substrate between the speaker and the samples, we mounted the speaker separately on the floor and used a Sorbothane™ vibration-isolation pad 262 under the optical breadboard holding the crest samples. Insertion of an acoustic foam tile between the feather crests and speaker to block airflow reduced the FFT spectral power at the 264 resonant frequency of the crests to 6.5% of its value without the tile; the remaining background vibrations are due to background caused by substrate vibrations and reverberation, as well as any 266 pressure waves transmitted through the foam. To find the background noise level due to environmentally driven vibrations for use in the Fourier analysis, we also measured crest 268 vibrations in the absence of audio playbacks. The background FFT power spectrum peak showed a single peak at the resonant frequency with the same power either when measured with 270 no audio playing or when measured during lower intensity playbacks (≤ 75 dB at distances ≥ 30 cm). 272

Simulated wing-shaking experiments 274
High-speed videos from a previous study were used to determine the frequency and amplitude of wing motions during the peacock's wing-shaking display (Dakin et al., 2016); we used four 276 videos with the correct perspective that also showed tail feathers with known lengths to estimate the mean diameter of wing motion circumscribed by the tips of the partly-unfurled wings during 278 this display as 7.6 cm (range 5.5 to 10 cm; Fig. 4). To simulate the resulting air motions in the laboratory, we used a robotic mechanism that caused an actual peacock wing to move such that 280 its plane remained in the same orientation while its distal end circumscribed a circle with the same rotational circulation as found in living birds (movie 1 and Fig. S2). The peacock wing 282 was mounted on a carbon fiber rod using a balsa wood base that was attached to the wing via adhesive at the shoulder; this rod pivoted about a clevis joint, which allowed the wing axis to 284 move in a vertical circle while the wingspan remained in the vertical plane. At the end opposite the wing, the rod was attached to a circular crank by a universal joint. The crank and attached 286 wing assembly was driven at 4.95 ± 0.05 Hz by a DC motor. While actual wing-shaking involves motions of two wings toward each other, each with diameter 10 cm, which presumably 288 displaces more air than a single wing, this apparatus used a single flapping wing moving in a slightly larger diameter (14 cm) circle at the wingtips. 290 To determine how wing-shaking influences the crest of an observing bird, we first determined 292 the location of maximal airflow speed during robotic wing-shaking. Airflow speeds were measured by a model 405i Wireless Hot-wire Anemometer (Testo, Sparta, NJ, USA) oriented 294 with its sensor facing in the same direction as the crest samples; this device has a resolution of 0.01 m/s, accuracy of 0.1 m/s, 1 Hz measurement rate, and approx. 5 s equilibration time. To 296 define the airflow pattern around the flapping wing, air speed was sampled at every point on a 5 cm grid 5-7 times per location. Using to these results, wood-mounted peahen feather crests were 298 positioned using a tripod at the vertical midline of the wing located at varying distances from the wing-tips as shown in Fig. S2. The resulting motions of the crests were then filmed using high-300 speed video as described above in "Video analysis" to quantify the vibrational response of three different peahen crests. To verify that substrate vibrations did not drive the crest motion, we 302 performed a control by inserting a 3 x 4 ft foamboard in between the crest and wing to block the airflow from the wing motion; this reduced the root-mean-squared crest motion to 14% of its 304 value with wing motion-induced airflow present. For comparison with the wing-shaking frequency during displays, flapping frequencies during ascending and level flight were also 306 measured for 9 peacocks from 6 online videos (Table S1). 308 Air vortex experiments To understand further the response of crests to individual airflow impulses, we used a Zero 310 Blaster vortex gun (Zero Toys, Concord, MA, USA) to generate single air vortex rings of artificial fog (2-4 cm in diameter, 1 cm diameter cross-section, speed 1.8 m/s [95% CI 1.7, 2.0 312 m/s, range 1.5 -2.1 m/s]), aimed so as to impact whole crests (n = 2 peacocks and 1 peahen crests) in the out-of-plane orientation. The motion of crest feathers struck by the vortices was 314 measured by tracking the crest position on high-speed video when an intact vortex impacted the crest oriented with its widest cross-section facing the source at 0.5 m from the point of creation. 316

Force measurements 318
We studied the static mechanical response of peafowl crests in the single cantilever bending geometry by measuring the relationship between flag displacement and restoring force of the 320 crest in the out-of-plane orientation (Fig. 1C). Force measurements were made using a Model DFS-BTA force sensor (accuracy ± 0.01 N) connected to a LabQuest2 datalogger (Vernier 322 Software & Technology, Beaverton, OR, USA), which was calibrated using known masses. The force sensor was attached to a thin rectangular plastic blade oriented in the horizontal plane. The 324 edge of the blade was pressed against the midpoint of the flags of the vertically oriented crest to measure the restoring force exerted by the bent crests. The crests were mounted on a micrometer 326 which moved them toward the force sensor and enabled measurement of crest displacement relative to the location at which restoring force first became non-zero. The resulting force vs 328 displacement data were fit to a linear model to determine the elastic bending constant, k, for three trials each for three male and three female crest samples. 330 For the air vortex and audio playback studies, the crest samples were mounted on a balsa wood 332 block that was mounted on two vertically-oriented force probes at either end separated by 3.0 ± 0.1 cm. Since the flags are approximately 5 cm from the base, the mean force on the crest 334 feather flag, F, can be computed from ΔF (the difference between the downward forces recorded by each sensor) using F = (5/1.5) ΔF = 3.3 ΔF. 336

Statistical analysis 338
To analyze sources of variation in whole crest fr and Q, we fit Gaussian linear mixed effects models with a random effect of crest ID to account for repeated measures of each bird's crest 340 using the nlme package (Pinheiro et al., 2017) in R (R Core Team, 2017). We first verified that trial order and frequency sweep rate, two aspects of the experimental design, did not have 342 significant effects on either fr or Q (all p > 0.28). The next step was to evaluate the potential effects of morphological traits that could influence crest resonance. Because we had only 15 344 crests, we considered models with one morphological trait fixed effect at a time, selected from the following list of traits: length, width, number of feathers, percent of unaligned feathers, and 346 percent of short feathers. All models also included fixed effects of the vibration orientation (in or out-of-plane), as well as sex. We used AICc to select the best-fit model (Bartoń, 2015) and 348 evaluated significance of the fixed effects in that model using Wald tests. We report R 2 LMM(m) as a measure of the total variance explained by the fixed effects (Bartoń, 2015;Nakagawa and 350 Schielzeth, 2013). We used the variance components of the best-fit model to calculate the repeatability of measurements after accounting for variation explained by the fixed effects 352 (Nakagawa and Schielzeth, 2010). Inspection of the data and model residuals revealed that variance in fr differed among crests, so when modelling fr, we specified this heteroscedasticity 354 using the weights argument (Pinheiro et al., 2017).

358
Morphology Fig. 1D shows that the range of lengths of dried feather crests measured in this study agreed with 360 that of live peafowl, indicating that the crest samples used in the experiments were fully grown (Dakin, 2011). However, the widths of the mounted crests were approx. 20% (female) to 27% 362 (male) smaller than those found on live birds (Fig. 1D). This difference could be due to the crest ornament being spread 1-2 cm in the sagittal plane by muscle action in the live bird, similar to 364 erectile crest plumage in many other species (Hagelin, 2002), in addition to the effect of skin drying. 366 We also studied the morphology of individual crest feathers to understand their unusual structure 368 (Fig. 1E) Analysis of the sources of variation in fr indicated that 28% of the variation in the resonant 386 frequency could be explained by sex, crest orientation, and the total area of the pennaceous flags (Fig. 2). The effect of crest orientation was strong and significant, such that out-of-plane 388 vibrations have approx. 2.4 Hz higher fr on average (p < 0.0001), whereas the sex difference was not significant (p = 0.87) and crests with reduced flag area have a slight but non-significant 390 tendency to have higher fr values (p = 0.10). Although length, width, number of feathers, and percent of unaligned and short feathers did not explain variation in crest fr, the repeatability of 392 crest fr was very high at 94%, suggesting that other characteristics may contribute to the consistent differences among individual crests. 394 The sharpness of the crest's resonant frequency is indicated by the quality factor, Q. The mean Q 396 for peafowl crests vibrated in the out-of-plane orientation (grand mean 6.2 for males, 4.8 for females) was intermediate between those of peafowl eyespot feathers (Q = 3.6-4.5 ± 0.4 and 1.8 398 ± 0.3, for individual feathers and feather arrays, respectively) and the tail feathers that drive the shaking, for which Q1 = 7.8 ± 0.5 (Dakin et al., 2016), indicating that peafowl crests are 400 moderately-tuned resonators. The quality factor values also have implications for undriven vibrations, such as those caused by single gusts of air. These undriven vibrations take place at 402 the crest's natural frequency, fo = fr sqrt(1 -½ Q -2 ); this results in an undetectably small shift (≤ 1.2%) relative to measurement errors for our measured Q values of peafowl crests. We discuss 404 the effect of this level of damping on the time behavior of feather vibrations below. 406 Approximately 49% of the variation in crest Q could be explained by sex, crest orientation, and the total area of the pennaceous flags (Fig. 2). Male crests were significantly more sharply-tuned 408 than those of females (p < 0.0001), and crests that had less flag area tended to be more sharplytuned as well (p = 0.03). Peafowl crests also have more sharply-tuned resonance when they are 410 vibrated out-of-plane (p < 0.0001) as compared to the in-plane orientation. The repeatability of Q was moderate, at 47%. 412 To determine the response of individual crest feathers, we removed all but three feathers (the two 414 outermost and middle) for three of the crests studied above (two male and one female An example power spectrum for the vibrational response of a peahen crest during audio playback is shown in Fig. 3. For audio files played back at in the near-field of the speaker, the vibrational 434 power spectra of the peafowl crests had a peak well above noise near the resonant frequency for all but the white noise signal, for which there was no measurable response above noise. In each 436 case measured, the peak frequency of crest vibrations exceeded the resonant frequency of the crest by 4 ± 2 (male) to 4.5 ± 0.2 (female) Hz (mean ± SE). This shift toward higher frequencies 438 was greater than the ≤ 0.3 Hz shift expected from the playback system's low frequency roll-off, but it was smaller on average than the width, Δf, of the crest's vibrational resonant response 440 (90% Δf for female and 66-125% Δf for the male crest). 442

Wing-shaking experiments
The simulated wing-shaking experiment resulted in an airflow pattern with speeds ≤ 0.3 m/s.

444
We used the measured positions of maximum airflow speed to determine the locations for three female crests for vibrational motion studies. Up to a maximum distance of approx. 90 cm (one 446 sample) and 80 cm (two samples) from the mean wingtip position, the FFT power spectra of the crest flag vibrational motion resulted in a peak (Fig. 4) that agreed with the wing-shaking 448 frequency within 95% CI. One crest had a lower power peak at the resonant frequency (29.0 ± 0.1 Hz vs. fr = 27.1 ± 0.2 Hz) for the greatest distances measured; a few samples also showed 450 peaks with weak power at the first harmonic of the shaking frequency.

452
The average peacock wing-flapping frequency during ascending and level flight was 5.53 ± 0.30 Hz (mean ± SE) (Table S1). This frequency agrees with the average frequency of 5.4 Hz (range 454 4.5-6.9 Hz) found for wing-shaking display frequencies measured in the field (Dakin et al., 2016). 456

Air vortex experiments 458
When ring-shaped air vortices traveling at approx. 1.4 m/s impacted the crests, the barbs responded with clearly visible motion on video with the average amplitude of motion at the flags 460 of 9.4 [4.3, 14.4] mm (Fig. 5A). Analysis of the free vibrational displacement of the crests over time revealed an exponentially decaying sinusoidal response with a frequency that agreed closely 462 with the measured resonant frequency of each crest (Fig. 5B). Thus, vortices cause the feather crest to vibrate at its natural frequency, with a decrease in amplitude of approx. 13% after 0.2 s 464 (the approx. period of peafowl wing-shaking displays). 466 We were unable to detect any forces above noise during either audio playbacks or vortex impact experiments, indicating that the forces exerted on the crests were ≤ 0.03 N. 468

Mechanical bending properties 470
All feather crests exhibited an elastic response in the bending experiments: force and displacement were linearly related for displacements up to 10.  (Fig. 6). The mean static bending spring 474 constants for the individual crests ranged from 0.0022 to 0.0054 N/mm with a measurement repeatability of 47%. Two out of three male crests had values of k that were higher than any of 476 the measurements for the peahen crests, but the difference between sexes in this small sample of 3 male and 3 female crests was not statistically significant (p = 0.09). 478 The findings from this study on peafowl point to a possible role of their feather crests for sensing airborne signals generated by their social motor displays. Morphometrics confirmed that the 484 crests of different individual peafowl are relatively uniform in length and area, as previously found in live birds (Dakin, 2011). This structural uniformity helps explain their well-defined and 486 narrow vibrational resonances (Fig. 2). We performed several different biomechanical experiments to understand whether the vibrational mechanics of peafowl crests were consistent 488 with a sensory role. The fundamental resonant vibrational frequencies of peafowl crests agreed closely with the frequencies used during male train-rattling and female tail-rattling displays. 490 This similarity seems unlikely to be due to coincidence, given the wide range of vibrational fundamental frequencies found for feathers of various lengths and structures in prior studies: for . This finding also indicates that peafowl crests can be driven efficiently by 494 stimuli caused by their social displays. To further test this hypothesis, we examined the response of crests to audio playback of train-rattling sounds, and verified that train-rattling caused the 496 crests to vibrate detectably near-resonance, whereas white noise resulted in no measurable vibrations above background noise levels. 498 Since there were no resonant modes of peafowl crests close to the 5 Hz frequency of peacock 500 wing-shaking displays, we were interested to find that the wing-flapping motions used to simulate this display in the laboratory also resulted in crest deflections of several mm at a 502 distance approx. 50 cm from the wing-tips. This implies that airflow impulses generated by the wing-shaking display can also stimulate the feather crests of nearby females. To understand the 504 crest response at a frequency so far from resonance, we measured the deflection of peafowl crests when they were struck by individual air vortices. As expected from their relatively low 506 values of quality factor, Q, we found that the crests vibrated near resonance only briefly and returned close to equilibrium after a time comparable to the period of peacock wing-shaking 508 displays. Thus, the airflow due to wing-shaking constitutes a series of essentially distinct impulses that can drive detectable crest responses when air flow disturbances are of sufficient 510 magnitude. One implication of this result is that hybrid biomimetic structures using a combination of feathers and resistance-based flex sensors provide a novel approach to making 512 sensitive detectors for sensing impulsive or periodic airflows. Such devices are required for proposed robotic applications of air vortex rings and other airflow signals as a communication 514 channel (Russell, 2011). Interestingly, peacocks also tilt their trains fore-and-aft during trainrattling at approx. 1 Hz, although we have not yet tested whether the airflows generated by these 516 slower maneuvers can also influence the crest. 518 Static mechanical tests also showed that the peafowl feather crest flags deflected linearly with bending force. This indicates that the results measured for high magnitude airflows, sound 520 pressure waves, and shaking forces can be extrapolated to the regime of lower-amplitude driving forces that might correspond to actual peafowl displays. This suggests that the magnitude of 522 deflections found when feather crests were exposed to either wing-shaking or audio playbacks are not inconsistent with a sensory role, given that the combined effects of amplification via 524 mechanical resonance and neural processes result in exquisitely small thresholds for animals that detect flows using mechanosensors. For example, in mammals, hair cells are sensitive to sub-526 nanometer displacements and 0.01 deg rotations (Crawford and Fettiplace, 1985), tactile receptors in human skin are sensitive to vibrational amplitudes well under a micron (Lo¨fvenberg 528 and Johansson, 1984), pigeons can detect submicron threshold vibrational amplitudes applied to flight feathers (Hörster, 1990), and insect filiform hairs are sensitive to airspeeds as low as 0.03 530 mm s −1 (Shimozawa et al., 2003). Further histological and electrophysiological studies of the receptors at the base of avian crest feathers are needed to determine their sensitivity to the types 532 of stimuli studied here. 534 The hypothesis that feathers might help detect airborne signals also suggests a new way to conceptualize behavioral experiments on birds that produce low frequency sound. Such studies 536 have tended to assume that these signals can be reproduced suitably by a distant source. Indeed, greater distances often have been emphasized in experiments on signals emitted at low-538 frequencies because of their potential as an effective means of long-distance communication. For example, capercaillie males produce mechanical sound when they perform 540 "flutter-jump" wing-shaking displays (Lieser et al., 2005). Experiments designed to study this behavior found no behavioral response when females were exposed to playbacks of the 542 infrasound (< 20 Hz) component of flutter-jump recordings produced by speakers located 5 m away (Freeman and Hare, 2011;Lieser et al., 2006;Manley et al., 2011). In another study, 544 peafowl were observed to perceive and respond behaviorally to playbacks of the infrasound components of train-rattling and wing-shaking recordings using rotary subwoofers located 5-20 546 m away from the birds studied (Freeman and Hare, 2015). Behavioral studies of auditory thresholds indicate that that some bird species (chickens and pigeons, but not budgerigars or 548 ducks) can detect low frequency sounds < 20 Hz with their ears (Heffner et al., 2013;Heffner et al., 2016;Hill, 2017;Hill et al., 2014); these studies also argue that eardrum perforation 550 experiments prove that these birds lack the ability to detect sound by mechanosensory means. However, all of these studies probed only the far field given the experimental design. Thus, they 552 were designed appropriately for determining the detection by hearing of the pressure-wave components of the acoustic signal, but not the effect of near-field airflow component on nearby 554 receivers. It would be of great interest to study birds that produce sound with a low frequency component by vocal and mechanical means using near-field study designs. For example, 556 cassowaries produce sound with fundamental frequencies of 23 or 32 Hz using vocalizations that make their entire bodies vibrate; humans are reported to both hear and feel these vibrations, 558 suggesting that this sound likely produces tactile airflow or substrate vibration signals in conspecifics as well (Mack et al., 2003). In the future, audio playback experiments should be 560 conducted in both the far-and near-fields to explore the possibility of such signals being transmitted via airflow rather than far-field pressure waves. 562 Peafowl are not the only bird species that have crests and perform shaking displays; for example, 564 we have compiled a list of at least 35 species distributed over 10 avian orders (Table S2). Given that feathers function as airflow sensors during flight, it is easy to imagine how they could be co-566 opted to function as sensors during social signaling. Birds from diverse species spanning several orders are known to have filoplumes on their heads that extend past the contour feathers 568 (Childress and Bennun, 2002; Clark and de Cruz, 1989;Imber, 1971;James, 1986). Many external stimuli and animal motions produce incidental sounds and airflows that could stimulate 570 these feathers, and eventually be adapted to communicative uses. The congruence between peacock wingbeat frequencies in flight and during wing-shaking displays is consistent with this 572 scenario. The sound and air-flow signals associated with such motions can be associated with kinematic parameters that may serve as signals of muscular performance, such as wingflap 574 frequency, amplitude, and/or duration (Clark, 2016). One can also imagine hitherto unsuspected functions in addition to signal detection. For example, if feather crests allow birds to detect wind 576 speed and direction, this could be useful for stabilization during flight or during roosting when the eyes are closed. Because predators that hunt by scent tend to approach prey from downwind, 578 the ability to sense wind direction may be a useful anti-predator adaptation (Conover, 2007). Such a wind sensor should be flexible enough to provide sensitivity via detectable 580 deformations for airflows in the range of interest, but rigid enough so its maximum bending occurs outside the range of typical airflow magnitudes. Many feather crests of birds meet these 582 criteria. Of course, the fact that a particular species' feather crest is used for some function doesn't mean that its shape has evolved primarily in response to that function. 584 Our results raise the important question of whether peafowl respond behaviorally to near-field 586 airflow signals detected by the crest. Further experiments could test this by removing or altering the crest and examining the effect on response to courtship displays, and/or performance during 588 flight, roosting, and/or predator avoidance. One way to do this would be to paint the crest feathers with a clear varnish, because this should shift the resonant responses of the crest without 590 changing its appearance visually or adding substantially to its mass. Other experimental approaches might involve measuring the response of peafowl to puffs of air directed toward 592 specific regions of their plumage, to see how this influences attention and body orientation. Also, the airflow patterns generated by wing-shaking peacocks could be determined and 594 compared to the movement of females during this display, to test whether specific female movements are induced by male actions. 596 Thus far, the elaborate shape and size of bird feather crests has led to an emphasis on their visual 598 appearance. Many avian courtship displays also involve wing-shaking, tail-fanning and mechanical sound production that may be detected by nearby females in the vibrotactile channel 600 (  and was also associated with the sex of the bird and the area of pennaceous flags at the top of the crest (although the association with flag area was not statistically significant). (C) The quality 634 factor, Q, was also influenced by the vibrational orientation, and was associated with the sex of the bird and the area of pennaceous flags. The average 95% CI for each Q estimate spanned 636 0.233. Black horizontal lines in (B-C) are means. Fig. 3. Effect of audio playback on crests. Vibrational response of a peahen crest exposed to peacock train-rattling audio playbacks in the near-field of the speaker. The FFT spectral power 640 for the peahen crest during playbacks peaked near the resonant frequency of the crest. 642 Fig. 4. Effects of simulated wing-shaking displays. Vibrational response of a female peahen crest exposed to airflow from a robot that simulated 5.0 Hz peacock wing-shaking displays at a 644 distance 50 cm from the moving wingtip (Fig. S2). When peafowl crests were impacted by air ring vortices, they deflected measurably, oscillating at their resonant frequency with an amplitude that decayed to a few percent of the initial value over 650 the period of the peacock's wing-shaking display. (B) Resonant frequencies (fr) and vortex response frequencies (± 95% CI) for three crests in the vortex experiment. 652