Male-male competition and female choice are differentially affected by male call acoustics in the serrate-legged small treefrog, Kurixalus odontotarsus

Vocalization recordings

The recordings were conducted in the Mt. Diaoluo National Nature Reserve in Hainan, China (18.44°N, 109.52°E, elevation of 933 m a.s.l.) from April to May, 2014. Each male was recorded for 3 min, using a directional microphone (ME66 with K6 power module, Sennheiser, Germany) connected to a digital recorder (PMD 660, 16 bit, 44.1 kHz, Marantz, Japan), from 20:00 to 24:00 h (Temperature: 20.4 ± 0.52 °C, Relative Humidity: 94.9 ± 3.96%). The calls of 62 males were recorded in their original habitat in order to provide a large sample of communication sounds for acoustic analysis. Prior to being returned to the habitat, each male was given a unique toe-clip number to prevent repeated recording.

Acoustic stimuli

In a previous study, we found that most of the complex calls of K. odontotarsus consist of two types of notes, note A and note B, with differing note numbers per call. Note A is a harmonic sound spanning a wide frequency band while note B is a shorter broadband sound. Note B is much shorter in duration than note A and does not exhibit the distinctive pulsed structure of note A (Zhu et al., 2016a; Köhler et al., 2017). Males typically produce three kinds of calls: A-note calls, B-note calls and compound calls containing both kinds of notes (Fig. 1, also see Zhu et al., 2016a). Of interest is whether these call types serve different behavioral functions. In order to evaluate the behavioral functions and likely selection pressures acting on the evolution of these call types, three classes of stimuli were constructed: (1) A calls consisting of 1, 3 or 5 notes (1A, 3A, 5A), (2) B calls consisting of 1, 3 or 5 notes (1B, 3B, 5B), (3) compound calls consisting of 5 notes A and 2 or 5 notes B (5A2B, 5A5B). All stimulus types used in the experiments were constructed using natural calls and are listed in Table 1.

To minimize pseudoreplication effects, five exemplars of each call type, derived from five different calling males were used. Each test stimulus combination was labeled with a numerical code of the form “X^y” in which “X” represented the stimulation types (i.e., number of A and B notes in the call pair used in the test) and “y” represented the male number. For example, the number “8⁵” in the female phonotaxis tests stands for the stimulus pair “5A2B” versus “5A5B” (see Table 1), derived from the calls of the fifth calling male. To control for potential order effects, all test stimuli were randomly broadcast. Randomized call orders were generated using the Randbetween function in Excel. The test stimuli were broadcast with a 5-s interstimulus intervals (approximately equal to the mean intercall intervals in the chorus). To minimize bias, observers were blind to the experimental conditions in use during recording and analysis of all female phonotaxis and male response data. Prior to returning the frogs to their original habitat, the subjects were given a unique toe-clip number to avoid being retested.

Female phonotaxis experiments

Female phonotaxis tests were conducted in the Mt. Diaoluo National Nature Reserve in Hainan, China (18.44°N, 109.52°E, elevation of 933 m a.s.l.) from April to June, 2014. A total of 65 gravid females were utilized in the phonotaxis tests. We performed standard two-speaker phonotaxis tests offering females a choice involving eight stimuli pairs (Table 1), which were presented to females between 20:00 h to 23:00 h (Temperature: 21.3 ± 0.35 °C, Relative Humidity: 93.5 ± 3.12%). To avoid experimental fatigue, each female on average was allowed a 3-minute break between consecutive tests.

Each female was placed in the center of a sound-attenuating chamber (inside dimensions, L ×W × H: 250 ×150 × 150 cm). We broadcast the stimulus pairs using two portable field speakers (SME-AFS, Saul Mineroff Electronics, Elmont, NY, USA) placed equidistant from the opposite sides of the chamber. The distance between the two speakers was 220 cm. To control for potential side bias, we broadcast the test stimuli antiphonally such that the “fast” root-mean-square amplitude of each test call was 80 dB SPL (re 20 µPa, A-weighted, the approximate SPL of a natural call at the same distance) and alternated the presentation of each stimulus pair between the right and left loudspeaker after each test. The sound pressure levels (SPLs) of each test call were measured with a sound level meter (AWA 6291, Hangzhou Aihua Instruments Co., Zhejiang, China) at the release location of the subject. We scored a choice when the female approached within 10 cm of either speaker without simply following the wall. If a female spent more than 10 min roaming the arena without approaching a speaker, she was given a second chance at the end of the tests after a 10 min break. If the female still failed to approach a speaker, no choice was recorded. We observed the behavior of the females on a monitor using a wide-angle lens video system with an infrared light source (also see Zhu et al., 2016b). The frogs were returned that night to their original habitat after the tests were completed. Prior to being returned to their original habitat, the subjects were given a unique toe-clip number to avoid being retested.

Male evoked vocal response experiments

Male evoked vocal response experiments were recorded at the same study site, from April to July, 2014. A total of 60 males were utilized in the playback tests. We found that male serrate-legged small treefrogs do not respond to playbacks in a sound-attenuating chamber. For this reason calling males were captured at breeding sites and brought to a field test site with an environment similar to the breeding sites where no nearby males were calling. Males used in the evoked vocal response experiments were tested with eight stimuli (Table 1), which were presented to males between 21:00 h to 24:00 h (Temperature: 20.9 ± 0.72 °C, Relative Humidity: 97.4 ± 3.58%). Each male subject was placed on the shrubbery. The playback stimuli were broadcast from a portable field speaker (SME-AFS, Saul Mineroff Electronics, Elmont, NY, USA) placed 1 m from the shrubbery (the approximate distance of two nearby males in their natural environment). Each stimulus was played for three minutes. The response calls of males before, during and after playback were recorded for three minutes respectively, using an Aigo R5518 recorder with an internal microphone (Aigo Digital Technology Co. Ltd., Beijing, China). The “fast” root-mean-square amplitude of the broadcast test stimuli at the release position was 80 dB SPL (re 20 µPa, A-weighted, the approximate SPL of a natural call at the same distance). The sound pressure levels (SPLs) of each test call were measured with a sound level meter (AWA 6291, Hangzhou Aihua Instruments Co.). The frogs were returned that night to their original habitat after the tests were completed. Prior to being returned to their original habitat, the subjects were given a unique toe-clip number to avoid being retested.

Analysis and statistics

The sonograms of calls were generated using free PRAAT software (Boersma, 2002). Data were statistically analyzed and graphs created using Sigmaplot 11.0 software (Systat Software Inc., Chicago, USA). The Chi-square test was used to evaluate the female phonotaxis results. The effects of stimulus type and playback time (before, during and after playback) on male evoked vocal responses were analyzed using Two Way Repeated Measures ANOVA (Two Factor Repetition). If a statistically significant difference was found, a multiple comparison procedure (Holm-Sidak method) was used to evaluate the differences among groups. All data were expressed as Mean ± SD, and p < 0.05 was considered to be statistically significant.

Ethics note

All applicable international, national, and/or institutional guidelines for the care and use of animals were followed. All procedures performed in studies involving animals were approved by the Animal Care and Use Committee of Chengdu Institute of Biology, CAS (CIB2014031008). This work was conducted with the permission of the Management Office of the Mt. Diaoluo Nature Reserve.

Analysis of call complexity

Through long-term field observations, we have found that contests in this species generally begin with an exchange of note A calls, then involve exchanges of note B calls and, in some cases, physical combat. Male serrate-legged small treefrogs produce calls of varied complexity. Males can produce a compound call, which consists of a series of multi-harmonic A notes with one or several B notes (Fig. 1). Males, usually, can add one to five B notes to several A notes. We referred to the measure of signal complexity in túngara frogs (Bernal et al., 2009; Akre et al., 2011) and described the call complexity of K. odontotarsus. The complexity of vocal signals is thus manifest at two levels: the number of notes and note types. Typically, calls contain 3.8 ± 0.9 (mean ± sd.) A notes. Calls with 3–5 notes are more complex than those containing only one or two notes (i.e., 5A vs. 2A). The compound calls containing two note types (such as the call shown in Fig. 1 with five A notes followed by five B notes, labeled 5A5B) may be regarded as more complex than those containing only one note type (i.e., 5A2B vs. 5A). In the field, more than half of the calls we recorded contained A notes followed by B notes. In all compound calls (n = 96), 85.6% have one or two B notes, while more than 96.7% have 1–3 B notes. In summary, male serrate-legged small treefrogs can produce calls of varying complexity consisting of varying numbers of A notes and B notes in response to potential rivals and to attract mates.

Female phonotaxis responses

The results are shown in Fig. 2. Some females did not behaviorally indicate a choice (5 females). When “5A” and silence were played back alternately, 75% of the females preferred “5A” (χ² = 25.000, p < 0.001, n = 60); when “5A” and “5B” were played back alternately, 63.3% of the females preferred “5A” (χ² = 6.760, p = 0.009, n = 60). These results demonstrate that note A is a more attractive sexual signal to females than note B. When the more complex “5A” and simpler “2A” were played back, 61.7% of the females preferred “5A” (χ² = 5.760, p = 0.016, n = 60). There was no significant difference between female responsiveness to “5B” and “silence” (χ² = 1.000, p = 0.317, n = 60) as well as to “5B” and “2B” (χ² = 0.360, p = 0.549, n = 60) consistent with the idea that females are not responsive to note B alone. In contrast 60% of the females preferred compound call “5A2B” to “5A” (χ² = 4.000, p = 0.046, n = 60). However, there were no significant differences between female responses to call “5A2B” and call “5A5B” (χ² = 0.160, p = 0.689, n = 60, Fig. 2).

Male evoked vocal responses

A total of 4,320 min of calling was recorded. We analyzed the total number of notes, notes/call, maximum number of A notes and the number of compound calls in male evoked vocal responses. There are significant differences in male evoked vocal responses between stimuli (Notes: F_7,59 = 10.923, p < 0.001; Maximum number of notes A: F_7,59 = 6.650, p < 0.001; Two Way Repeated Measures ANOVA) and playback time (Notes: F_2,59 = 41.238, p < 0.001; Maximum number of notes A: F_2,59 = 21.878, p < 0.001; Two Way Repeated Measures ANOVA). There is also a statistically significant interaction effect between playback time and stimuli (Playback time × Sound interaction effect, Notes: F_2,7 = 11.005, p < 0.001; Maximum number of notes A: F_2,7 = 8.020, p < 0.001; Two Way Repeated Measures ANOVA).

The stimuli containing note A (5A, 3A, 1A, 5A2B, 5A5B) elicited more responses during the period the stimuli were played back than before the stimuli were played back. The total note number respectively increased by 151.7% with 5A, 87.5% with 3A, 26.9% with 1A, 85.0% with 5A2B, and 129.6% with 5A5B (Fig. 3). Interestingly, stimuli containing only note B (5B, 3B, 1B) suppressed male calling responses during the period the stimuli were played, such that the number of notes produced decreased by 46.0% with 5B, 54.5% with 3B, and 39.0% with 1B (Fig. 3). More statistical details are shown in Table 2. For all stimuli, male calling responses “before” playbacks were similar and not significantly different from those produced “after” (p = 0.4, Holm-Sidak method, Fig. 3). Responses to “5A” were greater than to “3A” (p = 0.001) during the playback, and stimulation using “3A” stimulation resulted in more responses than “1A” (p = 0.013) during the playback. Responses to “5A” were greater than to “5A2B” (p = 0.015) during the playback, and “5A5B” stimulation resulted in more responses than “5A2B” (p = 0.031) during the playback (Holm-Sidak method, Fig. 3, Table 3). Stimuli containing only note B (5B, 3B, 1B) which differed in the number of B notes did not yield statistically significant differences during the playback (1B vs. 3B: p = 0.899; 3B vs. 5B: p = 0.540; 1B vs. 5B: p = 0.628; Holm-Sidak method; Fig. 3; Table 3).

The results show that the maximum number of notes A produced by males in response to all eight stimuli were similar as was the total note number. Male calling responses “during” playbacks differed significantly from those produced “before” (p < 0.001, Holm-Sidak method, Fig. 4). The stimuli containing note A elicited more complex calls during the period the stimuli were played back than before the stimuli were played back (5A: p < 0.001; 3A: p < 0.001; 1A: p = 0.014; 5A2B: p < 0.001; 5A5B: p < 0.001; Holm-Sidak method; Fig. 4; Table 2). However, stimuli containing only note B elicited fewer complex calls during the period the stimuli were played back than before the stimuli were played back (5B: p = 0.005; 3B: p < 0.001; 1B: p = 0.009; Holm-Sidak method; Fig. 4; Table 2).

Despite the fact that the number of notes/call which were evoked differed between stimuli (F_7,59 = 2.561, p = 0.013, Two Way Repeated Measures ANOVA), the only stimuli containing note A yielding statistically significant differences during playback were “5A5B” and “1A”, and the only note B containing stimuli yielding significant differences during playback were “1B” and “3B” (5A5B vs. 1A: p < 0.001; 1B vs. 3B: p = 0.002; Holm-Sidak method; Fig. 5). The number of notes/call differed among playback periods (F_2,59 = 36.307, p < 0.001, Two Way Repeated Measures ANOVA). Male calling responses “during” playbacks were significantly different from those produced “before” (p < 0.001) except for the stimulus “5B” (p = 0.064) and the stimulus “3B” (p = 0.178, Holm-Sidak method, Fig. 5, Table 2). For all stimuli, male calling responses “before” playbacks were similar and not significantly different from those produced “after” (p = 0.4, Holm-Sidak method, Fig. 5).

Twenty six (26) males (43.3%) emitted 96 compound calls which consisted of 4.33 ± 0.96 (mean ± sd.) A notes and 1.66 ± 0.93 (mean ± sd.) B notes. Males emitted more compound calls during the periods the stimuli containing note A (i.e., 5A, 3A, 1A, 5A2B, 5A5B) were played back (n = 84) than during the period before the stimuli were played back (n = 2). Notably males produced no compound calls or fewer compound calls during the period stimuli containing only note B (i.e., 5B, 3B, 1B) were played back (n = 1) compared to the period before these stimuli were played back (n = 6) (Fig. 6). In other words, playback stimuli containing note A but not stimuli only containing note B elicited increased numbers of compound calls.

In order to further investigate male call response patterns elicited by different playback stimuli, we counted the total number of notes A and notes B produced in all response calls. As can be seen in Figs. 7 and 8, total numbers of note A in calls produced and total numbers of note B in calls produced differed between stimuli (Note A: F_7,59 = 2.484, p = 0.016; Note B: F_7,59 = 19.540, p < 0.001; Two Way Repeated Measures ANOVA) and playback time (Note A: F_2,59 = 11.564, p < 0.001; Note B: F_2,59 = 108.358, p < 0.001; Two Way Repeated Measures ANOVA). We also found a statistically significant interaction effect between playback time and stimuli (Playback time × Sound interaction effect, Note A: F_2,7 = 2.323, p = 0.004; Note B: F_2,7 = 12.591, p < 0.001; Two Way Repeated Measures ANOVA). Stimuli containing only note B elicited fewer male calls containing note A during the period the stimuli were played back than before the stimuli were played back, such that the total numbers of note A produced decreased by 54.3% with 5B, 59.7% with 3B, and 56.0% with 1B (Holm-Sidak method, Fig. 7, Table 2). For all stimuli containing note A, male calls containing note A “before” playbacks were not significantly different from those produced “during” (5A: p = 0.619; 3A: p = 0.427; 1A: p = 0.943; 5A2B: p = 0.905; 5A5B: p = 0.760; Holm-Sidak method; Fig. 7; Table 2).

Stimuli containing A notes and stimuli containing only B notes did not yield statistically significant differences in the total numbers of note A produced during the playback (Fig. 7; Table 3). Furthermore, stimuli containing note A elicited more male calls containing note B during the period the stimuli were played back than before the stimuli were played back, such that the total number of B notes increased by 856.6% with 5A, 758.0% with 3A, 225.4% with 1A, 732.6% with 5A2B, and 753.7% with 5A5B (Holm-Sidak method, Fig. 8, Table 2). For all stimuli containing only note B, the numbers of male calls produced containing note B “before” playbacks were not significantly different from those produced “during” (5B: p = 0.78textit; 3B: p = 0.935; 1B: p = 0.508; Holm-Sidak method; Fig. 8; Table 2). As shown in Fig. 8, responses to “5A” were greater than to “3A” (p < 0.001) during the playback, and stimulation using “3A” resulted in more responses than “1A” (p = 0.003) during the playback (Table 3). Responses to “5A” were greater than to “5A2B” (p < 0.001) during the playback, and “5A5B” stimulation resulted in more responses than “5A2B” (p < 0.001) during the playback (Holm-Sidak method, Fig. 8, Table 3). Stimuli containing only note B did not yield statistically significant differences in the total numbers of note B produced during the playback (1B vs. 3B: p = 0.419; 3B vs. 5B:p = 0.716; 1B vs. 5B: p = 0.656; Holm-Sidak method ; Fig. 8; Table 3).

During the 6 min periods “before” and “after” the playback stimuli, male serrate-legged small treefrogs produced 15.60 ± 2.97 (mean ± sd.) calls containing note A, but produced 2.38 ± 1.30 (mean ± sd.) calls containing note B (Figs. 7 and 8).