Vertebrates communicate through a wide variety of sounds, but few mechanisms of sound production, besides vocalization, are well understood. During high-speed dives, male trainbearer hummingbirds (Lesbia spp.) produce a repeated series of loud snaps. Hypotheses for these peculiar sounds include the birds employing their elongated tails and/or striking their wings against each other. Each snap to human ears seems like a single acoustic event, but sound recordings revealed that each snap is actually a couplet of impulsive, atonal sounds produced ∼13 ms apart. Analysis of high-speed videos refutes these previous hypotheses, and furthermore suggests that this sonation is produced by a within-wing mechanism – each instance of a sound coincided with a distinctive pair of deep wingbeats (with greater stroke amplitude, measured for one display sequence). Across many displays, we found a tight alignment between a pair of stereotyped deep wingbeats (in contrast to shallower flaps across the rest of the dive) and patterns of snap production, evidencing a 1:1 match between these sonations and stereotyped kinematics. Other birds including owls and poorwills are reported to produce similar sounds, suggesting that this mechanism of sound production could be somewhat common within birds, yet its physical acoustics remain poorly understood.
Many bird clades have evolved to produce acoustic signals with their wings (e.g. Clark and Prum, 2015). The acoustic structure of the sound may reveal clues about its source. Here, we consider a hummingbird, the black-tailed trainbearer [Lesbia victoriae (Bourcier & Mulsant 1846)], which produces loud, short, atonal, staccato sounds with a distinctive accelerating acoustic tempo during a flight display. One hypothesis for a short impulsive sound is percussion, in which two solid elements collide and induce structural vibrations that are transmitted to the air. For instance, Manacus manakins snap by rapidly colliding the radii of the opposing wings, which then briefly vibrate to radiate sound (Bodony et al., 2016; Bostwick and Prum, 2003). Other types of percussion involve striking a substrate; for example, palm cockatoos pound sticks against trees (Heinsohn et al., 2017), and woodpeckers drum their beak against a resonant surface (Miles et al., 2018). Percussive sounds in flight have been reported many times in birds, some insects, and possibly in at least one bat (Bodony et al., 2016; Clark, 2021). Alternatively, a short, impulsive sound can also be aerodynamically produced by air escaping a constriction, as in human hand-clapping (Fletcher, 2013). Here, we assessed the mechanism by which black-tailed trainbearer males produce their display sound during J-shaped dives and chases (Hilty and Brown, 1986). Some authors suggested that the sounds come from the male's elongated tail feathers (Schulenberg et al., 2010) while others hypothesized that the sounds are made with the wings (Restall et al., 2006); Clark and Prum (2015) assumed the wings collided together to produce this sound. Here, we used high-speed videography and sound recording analyses to test these hypotheses.
MATERIALS AND METHODS
We observed black-tailed trainbearer males (L. victoriae; Fig. 1A) at two localities near Bogotá, Colombia (see Supplementary Materials and Methods for further details). Both localities are farms with strips of hedges and trees in between fields, with nearby patches of Andean cloud forest. In each locality, we made observations for 29 non-consecutive days from January to July. Our observations were either close to artificial feeders or in small patches of native vegetation where the trainbearer males were seen throughout the day. On account of the number of individuals seen simultaneously in each area, and the distance between recording locations, we observed a minimum of five adult males (see Supplementary Materials and Methods for further details).
We used a hand-held Miro EX4 video camera (Vision Research, Wayne, NJ, USA) filming at 500 frames s−1 and an 80-200 Nikon lens (Nikon, Melville, NY, USA) to record dives. As dives were idiosyncratic and difficult to anticipate, we were unable to elicit dives from a predetermined direction. Instead, we followed males with the camera as they dived, to reconstruct their wing kinematics throughout the dive, and assess any components of the flapping motions that may correlate with the snapping sounds. No single video obtained a complete sequence of a dive, but by combining videos, we reconstructed a general kinematic sequence of a trainbearer's dive (Fig. 1B), following our previously established approach (Clark, 2009). Out of these videos, one video of a bird flying directly towards the camera allowed a two-dimensional measurement of the angle of each wing over the course of multiple wingbeats (Fig. 1C). Wing angle was measured between the ‘body horizontal’ (coronal plane with shoulders as reference) and the line formed by the shoulder and the wingtips, selecting the proximal wing line (see Fig. 1C) when the wingtips were blurry (e.g. Fig. 1D).
The display snapping sounds were recorded with a Sound Devices 702 recorder (24 bit, 96 kHz), Sennheiser MKH 20 (Sennheiser, Old Lyme, CT, USA) in a Telinga pro parabola (Telinga, Tobo, Sweden). Given the opportunistic nature of our recordings, and the unknown (and changing) distance between the bird and microphone, sound recordings were not synchronized with video. We digitized the timing of snaps directly from the waveform, as they were clearly visible as peaks, and we produced spectrograms with Avisoft-SASLab Pro© (Avisoft Bioacoustics, Glienicke, Nordbahn, Germany) (window size: 256, overlap 70%). Means are presented ±s.d.
Filming protocols were approved by the Institutional Animal Care and Use Committee at the University of Washington (UW) and at the University of Connecticut (UConn); UW IACUC 4498-01, 4498-03 and UConn Exemption Number E09-010. Visual inspection of hummingbird wings was performed at the Burke Museum of Natural History and Culture, University of Washington (UWBM) and at the Instituto de Ciencias Naturales, Universidad Nacional de Colombia (ICN).
Dive display kinematics
We obtained a total of five high-speed videos of dive displays that included the apparent kinematics of sound production. Based on these videos and our behavioral observations, we inferred that in a typical dive, (1) a male climbed almost vertically to a height of approximately 10–20 m. (2) Next, the male closed his wings, pitched head-down, and descended with its wings and tail closed. (3) Lastly, the male started beating his wings and pitched up until he was flying horizontally and no longer descending (Fig. 1B,C). During the last two portions of the display dive, during which time the snaps were produced (observation made by the researchers recording the dives and seeing the birds simultaneously), the males alternated between bouts of mostly shallow flapping and gliding. Among the shallow wing flaps were interspersed a series of discrete, stereotyped pairs of wingbeats with altered kinematics (deeper strokes), hereafter termed ‘wingbeat pairs’ (Fig. 2A).
None of the videos revealed the wings colliding with any other structure (tail, contralateral wing or body), contrary to our a priori hypothesis. Likewise, the tail was not involved in any kinematic pattern related to sound production. Our videos contained long series (≥10) of stereotyped wingbeat pairs. In most dives, the flapping plane of the wings was not aligned with the image plane of the camera. However, one video that captured this stereotyped kinematics also had the body coronal plane approximately orthogonal to the image plane (Movie 1). This video permitted two-dimensional measurement of the wing angle over the course of multiple wingbeats (Figs 1C and 2A).
The stereotyped kinematics of the wingbeat pairs were characterized by larger wing excursions (deep strokes). In the first upstroke of each pair, the wings were elevated to a higher position than in the preceding and subsequent, shallower wingbeats (Fig. 2A; Movie 1, Table S1; all analyses in R version 4.0.5 http://www.R-project.org/). The downstroke of the first wingbeat pair was arrested ‘early’ (compared with the second one; Table S1). In the second upstroke, the wings were elevated less than in the first one (Table S1), but in the second downstroke, the wings were depressed to a lower angle than in the rest of the wingbeats in the display (Figs 1D and 2A; Table S1). In total, the effect was that the first flap had an average stroke angle above the body horizontal (dihedral), while the second flap had an average stroke angle below the body horizontal (anhedral); the net stroke amplitude was similar in the two flaps of the pair (∼93–110 deg; Table S2) and greater than the stroke amplitude of the rest of the wingbeats (∼44 deg larger; Table S2).
We obtained 10 audio recordings of snapping sound sequences (e.g. Audio S1). Each snap sounds (to humans) like a single acoustic event. However, analysis of the sound recordings revealed that each ‘single’ snap actually always comprised two consecutive sounds in very quick succession, which we hereafter refer to as a ‘couplet’ (Fig. 2B). The number of couplets in individual displays varied between five and nine. The interval between sounds within each couplet, or intra-couplet interval, was nearly constant (13.1±1.1 ms; Fig. 2C; Table S3). The interval between couplets (specifically between the second sound of one couplet and the first one of the next), or inter-couplet interval, decreased over the course of the first half of the sound sequence and, late in the display, stabilized at an approximately constant interval of 29.9±1.6 ms (Fig. 2C; Table S4).
Comparison between kinematics and sounds
The stereotyped wingbeats and the snaps exhibited cadenced sequences that were tightly correlated (Fig. 2C). We used as a reference the moment when the inter-couplet interval became constant in each of the 10 sound sequences; similarly, there was a point in the five video sequences in which the inter-wingbeat pair interval became constant (Fig. 2C; Table S4). These reference points within the sounds and videos aligned and unveiled a 1:1 match (Fig. 2D; Fig. S1) between the timing of videos (wingbeat pairs) and the sound recordings (couplets). The intra-couplet intervals (13.1±1.1 ms) were similar to the intra-wingbeat pair intervals (14.9±1.0 ms) across all of our data (Table S3).
As our videos were neither synchronized with nor taken simultaneously to the audio recordings, the only video suitable for describing the wing excursion angle quantitatively (and thus potentially idiosyncratic), was not sufficient to conclusively locate the point of the wing cycle where the sound is produced. Our main conclusion (that sonations are associated with stereotyped wingbeats) is reinforced by the wing angle analysis of that single video (showing significant differences in wing kinematics; Tables S1 and S2), but it is mainly supported by the fact that the analyses of all of our recordings exhibited temporal spacing and shifts, yielding a sequence of events (strokes with larger amplitude) that matched the snap sequences (Fig. 2; Fig. S1).
High-speed video of the dive display of the black-tailed trainbearer shows that the timing of altered wing kinematics matches the timing of sound production (Fig. 2), implying that these deeper wingbeats are the origin of the sound. Clark (2018) developed criteria for distinguishing between locomotion-induced sounds that are signals and cues: of these, this sound appears to satisfy the ‘modified behavior’ criterion, and is therefore likely a signal. Our data also refute all a priori hypotheses about the mechanism: the high-speed videos show that neither is the tail involved nor is there physical contact between the wings and other bodily structures (contraClark and Prum, 2015), unlike the percussive sonations of manakins, which are produced by striking the wings together (Bostwick and Prum, 2003; Prum, 1998). Instead, our analyses show that the snaps are produced through a within-wing mechanism. The physical acoustics of this mechanism are not clear. One hypothesis is that air could be expelled from between two primary feathers, similar to human hand-clapping, where the sound is produced by air escaping a concavity in one hand as the other rapidly approaching hand forces the air out of the way (Fletcher, 2013). Snapping one's fingers produces sound via a similar mechanism. We examined the wings of adult male and female museum specimens (see Acknowledgments) looking for any modified morphology (sensuClark, 2018), i.e. differences from other hummingbirds or sexual dimorphism, but we failed to find any obvious morphological feature that was potentially related to sound production.
Although many hummingbirds are known to perform flight displays that include close-range shuttle displays and high-speed dives (e.g. Clark, 2009; Skutch, 1973; Stiles, 1982), only a few species of hummingbird have been documented to produce atonal, impulsive sounds during their displays. Close relatives of the black-tailed trainbearer make similar snapping sounds, including green-tailed trainbearers (Lesbia nuna: Hilty and Brown, 1986; Schulenberg et al., 2010), purple-backed thornbills (Ramphomicron microrhynchum: Hilty and Brown, 1986) and red-tailed comets (Sappho sparganurus: N. Areta, personal communication), representing a single phylogenetic origin of these sonations. However, a number of distant relatives also produce somewhat similar sounds (Table S5), suggesting that percussive mechanisms are widespread in hummingbirds.
Certain other birds produce wing-snapping sounds similar to those described here (e.g. ‘clicks’; Bostwick and Prum, 2003). A few species have morphology that strongly implies a within-wing mechanism: guans (Cracidae), for example, have a series of modified wing feathers, and produce a distinctive, repeated knocking sound during displays. The available descriptions of their displays do not imply that the wings are snapped together (Delacour and Amadon, 1973); rather, the mechanism is likely some sort of interaction between these neighboring highly modified wing feathers. But guans appear to be an outlier, and their sonations sound very different from the trainbearers' snaps we describe here. A number of nightjars (Caprimulgidae) produce wing snapping or wing clapping (e.g. Eddowes and Lea, 2021), and there is disagreement as to whether the wings touch during the sound production in species including the European nightjar (Caprimulgus europaeus) (Coward, 1928), chuck-will's widow (Caprimulgus carolinensis) and common poor-will (Phalaenoptilus nuttallii) (Mengel et al., 1972). Mengel et al. (1972) suggested that ‘the primaries were somehow snapped like fingers, or, more precisely, like fans’. Likewise, Chaetura swifts (Apodidae) make snapping sounds, and some descriptions of how the sound is produced imply the sound could be generated by a within-wing mechanism (Collins, 1968; Sick, 1959). Most of the aforementioned species do not have distinctively modified feathers, similar to the trainbearer we describe here.
Notably, the function of the display dives we observed in black-tailed trainbearers is not entirely clear. As this display is apparently performed only by males and we observed females in the surroundings, our hypothesis is that it is a courtship display. However we did not confirm breeding, and in many cases the recipient of the display was unclear (and it is even possible that this display is sometimes undirected). We highlight the need to continue studying behavioral and morphological adaptations associated with the pugnacious nature of hummingbirds (e.g. Rico-Guevara and Araya-Salas, 2015; Rico-Guevara et al., 2019), as well as deciphering the physical mechanism of sound production of trainbearers' snaps and, more generally, delving into the unexplored acoustic diversity in flying animals (Clark, 2021).
We thank Kristiina Hurme and the Familia Peña-Rico for field support at Te Faruru (El Rosal) and Victoria Lizarralde at the Observatorio de Colibríes (La Calera), Nacho Areta for observations on the red-tailed comet, Rosalee Elting for help with the manuscript and figures, and the members of the Behavioral Ecophysics Lab and Gary Stiles for helpful discussions and access to the Ornithological Collection at the ICN, as well as the staff of the Burke Ornithology Collection (UWBM).
Conceptualization: A.R.-G., L.E.-M., C.J.C.; Methodology: A.R.-G., L.E.-M., C.J.C.; Validation: A.R.-G., L.E.-M.; Formal analysis: A.R.-G., L.E.-M.; Investigation: A.R.-G., L.E.-M., C.J.C.; Resources: C.J.C.; Data curation: A.R.-G., L.E.-M.; Writing - original draft: A.R.-G.; Writing - review & editing: A.R.-G., L.E.-M., C.J.C.; Visualization: A.R.-G., L.E.-M.; Supervision: A.R.-G.; Funding acquisition: C.J.C., A.R.-G.
This work was supported by a National Geographic Society Explorer's grant [#WW-047R-17]. A.R.-G. is supported by the Walt Halperin Endowed Professorship and the Washington Research Foundation as Distinguished Investigator. L.E.-M. was supported by Idea Wild.
The authors declare no competing or financial interests.