Female large odorous frogs (Odorrana graminea) prefer males with higher nonlinear vocal components

Abstract In anurans, the complexity of courtship calls may affect female mate choice. The current study suggests that nonlinear phenomena (NLP) components can contribute to increasing complexity in courtship calls and attracting female attention. The results of a recent study showed that calls of large odorous frog (Odorrana graminea) contained NLP components. However, whether the nonlinear components of courtship calls in O. graminea improve male attractiveness remains unknown. We hypothesized that female O. graminea would prefer males producing calls with a higher proportion of NLP components (P‐NLP‐C). To test this hypothesis, we recorded the advertisement calls of 28 males and confirmed that the P‐NLP‐C was significantly positively related to body size. We also measured the body size of natural amplectant males and non‐amplectant males in the field and found that amplectant males had larger body sizes than non‐amplectant males, and the results of two‐choice amplexus experiments similarly revealed a female preference for males with larger body sizes. Additionally, phonotaxis experiments also revealed that females preferred male calls with a high P‐NLP‐C. The results suggest that a higher P‐NLP‐C in calls can enhance male attractiveness, and the P‐NLP‐C may provide key information about male body conditions for female O. graminea. Our study provides a new insight for better understanding the role of NLP in anuran mate selection.


| INTRODUC TI ON
Sound communication plays a vital role in female mate choice, especially in anuran species (Narins et al., 2006;Zhang et al., 2015). The characteristics of male courtship call, particularly the complexity of the call, are important factors to be considered by gravid females when choosing a mate Ryan et al., 2019). Females attend to the calls to assess males and have a stronger preference for complex calls over simple calls (Gridi-Papp et al., 2006). The acoustic complexity is often associated with variations in the spectrotemporal characteristics (e.g., frequency, duration, and the number of syllables) of the vocal signals (Fee et al., 1998;Márquez, 1995;Ryan, 1983). In some species producing nonlinear calls, improving the content of nonlinear phenomena (NLP) is an important way to increase the call complexity (Fitch et al., 2002;Rice et al., 2011;Wu et al., 2021).
NLP, as distinctive structural features of acoustic signals, are ubiquitous among the acoustic signals of vertebrates (Digby et al., 2014;Feng, Riede, et al., 2009;Fitch et al., 2002;Rice et al., 2011;Zhang et al., 2017). The occurrence and strength of NLP in calls not only increase individual vocal distinctiveness but also attract more attention from receivers in individual interactions, which was widely confirmed in fish and mammals (Fitch et al., 2002;Volodina et al., 2006;Wilden et al., 1998). Several studies have focused on the general function of NLP, for example, in alarm signals (Blumstein & Recapet, 2009;Townsend & Manser, 2011) and to convey information about direction and identity (Digby et al., 2014;Kaplan et al., 2018), but few have focused on mating choice, especially in frogs .
The large odorous frog (Odorrana graminea) can produce calls that contain four NLP components (subharmonics, deterministic chaos, frequency jumps, or biphonations), and 84.6% of the vocalizations contain one or more nonlinear components . O. graminea is an arboreal species inhabiting areas around loud streams and waterfalls in select regions of China (Chen, 1991).
Similar to that of a sympatric species, Odorrana tormota, the calls of O. graminea consist of ultrasonic and NLP components (Shen et al., 2011;Zhang et al., 2021). In O. tormota, smaller males with higher call frequencies had greater mating success than larger rivals, as their calls were more conspicuous in the species' habitat with intense but predominantly low-frequency stream noise (Zhang et al., 2020 Each night, we captured as many amplexed pairs as possible during the peak reproductive hours (between 19:30 and 22:30; Figure 1).
All captured individuals were placed in containers with water and stones in time and then brought back to the test chamber. The local nightly temperature and humidity during the study period ranged from 19 to 24°C and 90% to 96%, respectively. The test chamber was located near the sampling site and maintained a similar temperature, humidity, and dark light environment. Each individual was in the test chamber for no more than three days and released back to the sampling site immediately after testing. To avoid repeated capture, each frog was given a finger cutting mark before returning to F I G U R E 1 An amplexed pair of Odorrana graminea in the field. Photo by Jinmei Wang its habitat. All the behavioral observations conducted in this study were performed in accordance with the current laws of the China and the Animal Care and Use Committee at the Anhui Normal University (Permit # 00111).

| Field sampling
We captured 30 amplexed pairs that were found on top of boulders in the middle of the mountain stream and in bamboo groves next to small waterfalls in the stream. Three to six solo males (i.e., non-amplectant) in the vicinity (< 0.5 m) of an amplexed pair were randomly captured and housed in separate terrariums (22 × 17 × 15 cm). Using a caliper (Spi2000 Wiha, Germany) with an accuracy of ±0.1 mm, we measured the snout-vent lengths (SVLs) of the frogs. The SVL data of the amplectant males were later compared to those of solo males. The number of amplexed pairs we spotted and captured each night varied from 0 to 5. No amplexed pair was found in the field when the ambient temperature was below ~15 °C. A total of 123 solo males were captured, and their characteristics were compared with those of amplexed males in the field (Table 1).

| Audio recording and call analysis
To quantitatively analyze the relationship between male body size and the P-NLP-Cs in vocalizations, we recorded the vocalizations of 28 actively calling males (not in amplexus) for three evenings under similar ambient conditions (temperature: 20 ~ 22°C, humidity: 92% ~ 96%, ambient noise: 70 ~ 78 dB SPL peak) in the field and measured their SVLs. Male calls were recorded by using a digital audio recorder (Sound Devices 702, Sound Devices, WI, USA, frequency range: 10 Hz ~ 96 kHz) with a sampling rate of 96 kHz and 16-bit accuracy and a miniature omnidirectional condenser microphone with a flat frequency response over 20 ~ 20,000 Hz (AKG model C417, AKG Acoustics, Vienna, Austria) (Zhang et al., 2015).
For each male, we recorded at least 8 calls.
Calls were initially analyzed with SELENA, a custom-designed program (Feng, Riede, et al., 2009), to produce narrow-band spectrograms and determine the call durations. Next, calls comprising multiple notes, the duration of individual notes, the duration of signal breaks, and the overall call duration were measured. Then, different temporal segments of each call were identified with PRAAT based on visual inspection of the narrow-band spectrogram (Figure 2; Boersma & Weenink, 2007). After the call segments were finished, the time of occurrence of each segment was noted. In addition, the durations of the various NLP segments (i.e., subharmonics, deterministic chaos, frequency jumps, biphonations) were measured.
Finally, these durations were calculated as percentages of the total call duration, and the average NLP percentage for each male frog was calculated. The fundamental frequency (f 0 ) was tracked for each harmonic segment using the "pitch tracking" mode in PRAAT, with a 1-ms interval.

| Amplexus experiments
To determine the male body size preferred by females, we performed two different two-choice amplexus experiments. The behaviors of the males and females were recorded by a camcorder (Sony model HF M40). In the first two-choice experiment (Experiment A), we paired one gravid female with two randomly captured males with varying body sizes in a test terrarium (38 × 38 × 28 cm) for up to 15 min. We placed the female in the center of the test terrarium and the two males equidistant from the female. If the pairing did not result in amplexus within 15 min, namely, the female rejected both males, we substituted two new males and repeated the experiment (at most one repeat). Upon amplexus, we separated the pair and measured the SVLs of the amplectant male and non-amplectant male. In total, 27 out of 30 experiments resulted in amplexus.
In the second two-choice experiment (Experiment B), a female was paired with two males that had been previously paired with a different female. The goal was to determine whether a particular male phenotype was preferred by different females. To execute the second two-choice experiment, we removed the female previously paired (in Experiment A) from the test terrarium and allowed the males to remain and rest in the terrarium for at least 30 minutes.
Then, we placed another female into the terrarium and began the second two-choice experiment (Experiment B). All individuals in each experiment were randomly assigned and never tested more than once in the same experiment.

| Phonotaxis experiments
Three acoustic-stimulus pairs were synthesized using Cool Edit Pro 2.1 software based on the advertisement calls of different males that have been shown to attract females. We first chose some typical calls that fits the criteria (specific frequency or P-NLP-C), then synthesized stimulus from different call components in a certain ratio, and finally standardized some parameters including the call duration and amplitude. The characteristics of stimulus were based on the f 0 and P-NLP-C of the calls observed in this and previous studies . In phonotaxis test 1, the goal was to test the still failed to approach a speaker, no choice was recorded . Six females were found to be about to oviposit before the experiment and released back to their natural habitat in time. In total, 24 gravid females were used in the phonotaxis tests. The order of test 1 and test 2 was random, but test 3 was performed after test 2.
The order of all individual in each experiment was random.

| Female mate choice in the field
The field observations revealed that on average, amplectant males had a larger SVL (mean SVL ± SD = 50.74 ± 2.01 mm, N = 30) than non-amplectant males (mean SVL ± SD = 49.07 ± 2.48 mm, N = 123), as shown in Figure 4a. The difference was statistically significant (Mann-Whitney U test, p < .05). The pooled data were consistent with the data collected nightly. Table 1 shows that the average SVL of amplectant males was larger than that of solo non-amplectant males for all 10 nights for which we located amplexed pairs. The SVL of males (mean SVL ± SD =50.74 ± 2.01 mm, N = 30) and females (mean SVL ± SD = 97.72 ± 3.52 mm, N = 30) in amplexed pairs was not significantly correlated (linear-regression analysis, R 2 = 0.000, p = 1.000, Figure 5). Thus, O. graminea showed no size-assortative mating.

| Correlation between male body size and call characteristics
Analysis of the advertisement calls of 28 male frogs revealed that body size was significantly positively related to the P-NLP-C (linearregression analysis, R 2 = 0.120, p = .040, Figure 6a), but not significantly correlated with the fundamental frequency (linear-regression analysis, R 2 = 0.028, p = .216, Figure 6b); in other words, calls of larger males showed a higher P-NLP-C, and P-NLP-C might be a potential predictor of male body size.

| Female phonotaxis
A total of 24 gravid female O. graminea were subjected to the phonotaxis experiments, and some females did not make a choice (N = 4); however, most female frogs preferred advertisement calls with a low f 0 over calls with a high f 0 (binomial test, p = .003, Figure 7b). Female frogs preferred advertisement calls with a high P-NLP-C (50%) compared to the advertisement calls with a low P-NLP-C (10%) (binomial test, p = .003, Figure 7c). Twenty gravid female frogs that responded in phonotaxis test 2 were subjected to phonotaxis test 3. The results of the experiment showed that calls with a high P-NLP-C (75%) were chosen by 18 females, while 2 females chose calls with a low P-NLP-C (20%) (binomial test, p < .001, Figure 7d).

| DISCUSS ION
The results of our study in O. graminea support the working hypothesis that females would prefer males producing calls with a higher P-NLP-C. We found that in the field, females of O. graminea preferentially mated with larger males whose calls had a higher P-NLP-C.
In manipulative experiments, females also showed a preference for males with larger body sizes and calls with a high P-NLP-C. The present study highlights the important role of NLP in anuran mate selection.
Anuran mating patterns are often nonrandom in terms of body size (Mansouri et al., 2020), which was also demonstrated in this study. Females of O. graminea preferentially mate with larger males in the field, which was validated in two two-choice experiments.
For most anuran species, females show a preference for larger males (Andersson, 1994;Rausch et al., 2014). Mating with larger males could result in indirect genetic benefits in terms of high offspring fitness. Márquez reported that midwife toads (Alytes obstetricans and A. cisternasii) who mate with larger males produce offspring with faster growth, a larger adult size, and superior survival ability (Márquez, 1995). But there are a few exceptions, for example, female serrate-legged small treefrogs (Philautus odontotarsus) favored the intermediate size of males to reduce energy consumption for carrying male . In addition, the body size ratio of amplexed pairs could affect the fertilization rate, for example, effective fertilization requires that cloaca to be properly juxtaposed during amplexus (Bastos & Haddad, 1996;Friedl & Klump, 2005). In O. graminea, there is greater female-biased sexual size dimorphism between sexes (sexual dimorphism index = 1.93, N = 30; Figure S1) compared to most anuran species (Monnet & Cherry, 2002;Zhang & Lu, 2013); thus, choosing a large mate is conducive to improving the efficiency of amplexus and fertilization in O. graminea.
Due to individual differences in vocal apparatus, body size variation can affect the occurrence or intensity of NLP components in acoustic signals and increase the complexity of vocalizations (Cazau et al., 2016;Serrano et al., 2020). In O. graminea, a larger body size was associated with a higher P-NLP-C. The phonotaxis experiments indicated that the females preferred male calls with a higher P-NLP-C, so it is reasonable to speculate that a higher P-NLP-C in calls might enhance male attractiveness, affecting female mate choice and contributing to female preference for large males. It is likely that call features such as nonlinearities encode relevant information, such as body size (Juola & Searcy, 2011;Wu et al., 2021).
The relationship between body size and NLP has been previously reported to occur in mammals (Cazau et al., 2016;Fitch et al., 2002), but it has not been corroborated by a significant correlation in any living organism. Serrano et al. (2020) first revealed that body size had influences on NLP components at the intra-and interpopulation levels, for example, in Darwin´s frogs (Rhinoderma darwinii), smaller individuals had higher proportions of relative duration of chaos. This study demonstrated that there was a significantly positive correlation between the P-NLP-C in calls and body size in male O. graminea.
An inverse relationship between body size and fundamental frequency occurs typically in many anurans, but this study did not find the significant correlation in male O. graminea. It may be limited to the sample size of our study and require further experimental verification. Meanwhile, the evolution of female preference for male calls with different frequencies needs to take into account the multiple effects of auditory sensitivity and background context . We speculate that males of O. graminea with bigger body size may not always have the calls with lower f 0 in the field, otherwise, their opportunities of being detected by females would be reduced due to the masking of background stream noise. This suggests that for female O. graminea, the P-NLP-C may be a more reliable clue of male body size than the fundamental frequency. Further, this study also revealed that a higher P-NLP-C in courtship calls can enhance female attention to calling males, and larger males with a higher P-NLP-C had greater mating success than smaller rivals.
The complex structure and irregular frequency spectrum of NLP can increase the specificity of calls, making it easy to attract the attention of receivers and prevent habituation (Blumstein & Recapet, 2009). The results of phonotaxis experiments indicated that females of O. graminea had a strong preference for the calls with a high P-NLP-C or a low f 0 . However, the time of phonotaxis in the high-low P-NLP-C tests was significantly less than that in the high-low f 0 test.
This suggests that a high P-NLP-C might play a more important role in female phonotactic behavior than a low f 0 , and a higher P-NLP-C in calls has stronger effect on mate choice by females. These results are in accordance with recent findings in mammals, in which females displayed the strongest preference for calls with NLP over those without NLP (Charlton et al., 2017(Charlton et al., , 2018Reby & Charlton, 2012).
Therefore, the hypothesis that calls with nonlinear components are more attractive to females is not limited to mammals, as it also applies to other animals, including frogs.
The communicative significance of NLP has been examined in anurans , reptiles (Labra et al., 2013), birds (Digby et al., 2014), and mammals (Tyson et al., 2007). Different functions have been assigned to NLP, including individual recognition (Feng, Riede, et al., 2009;Fitch et al., 2002), prevention of habituation (Karp et al., 2014), indicators of fitness (Fitch et al., 2002;Riede et al., 2004), and so on. However, the role of NLP in social interactions has rarely been evaluated (Digby et al., 2014). The study in O. tormota showed that male frogs with a high P-NLP-C in calls had a greater advantage in mate selection , consistent with the results of this study, indicating that the P-NLP-C in calls might play an important role in mate selection in frogs that can produce calls with NLP.

ACK N OWLED G M ENTS
We thank Albert S Feng for his constructive suggestions on the field and laboratory study, and always cherish the memory of this sincere mentor and friend. We also thank anonymous referees for their valuable comments which help us greatly improve the manuscript.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflicts of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The raw data used to perform analyses and generate figures for this manuscript are available at https://doi.org/10.5061/dryad.sqv9s 4n5h; https://doi.org/10.17632/ xtrf2 dxv6j.1.