The functions and evolution of graded complex calls in a treefrog

ABSTRACT Variation in signal complexity is common among different species. Understanding why some species have evolved extensive call complexity can offer insight into the evolution of complex signals. In this study, we investigated the functions of different call types in a treefrog using call network analysis, male playback experiments, and correlation analysis between call properties and body size. Our results show that the male treefrogs can produce three kinds of notes that can be combined to produce single-note call types as well as composite calls. We also identified an intermediate note type between the advertisement call and suppress call, indicating that male suppress calls may have evolved from advertisement calls. The call network revealed functional correlation and distinction between different calls. Additionally, our study revealed that the dominant frequencies of the advertisement calls and suppress calls can signal the size of calling males. Male frogs can combine different notes in different sequences to form graded complex calls, which may perform different communication functions. Our results indicate that different communication needs may drive the evolution of signal complexity, which provides new evidence to the social complexity hypothesis and deeper insights into the function and complexity of animal acoustic communication.


Introduction
Acoustic communication is important in animal aggressive, defensive, and especially reproductive behaviours (Toledo et al. 2014;Briseno-Jaramillo et al. 2017;Fan et al. 2018;Xiao et al. 2018;Matrosova et al. 2019;Brakels et al. 2023).While some animals produce simple and stereotyped calls (Chen et al. 2021), others have a vast repertoire of calls with distinct structures (Pollard and Blumstein 2012;Costa and Toledo 2013;Leighton et al. 2021;Jiang et al. 2021;Vesely and Batista 2021;Fan et al. 2022).The marsh wren (Cistothorus palustris), for instance, has an extensive repertoire of 162 different songs (Kroodsma and Verner 1987), which raises an intriguing question of why some species have such a sophisticated vocal repertoire (Leighton et al. 2021).Sexual selection has been identified as a key force driving the expansion of vocal repertoire (Robinson et al. 2019).Females, for example, have been observed to prefer novel calls to stereotyped calls (Leger et al. 2000).Furthermore, males have been observed to adjust their call complexity based on social context (Bernal et al. 2009;Nali and Prado 2014;Zhu et al. 2017a;Bhat et al. 2022).Interestingly, some animals combine different elements to create compound calls with multiple functions (Márquez et al. 2001;Zhu et al. 2017b;Flechas et al. 2018).However, there is a paucity of research on the modularity of the composite calls formed by combining different notes.According to the social complexity hypothesis, highly social taxa always produce relatively more complex vocalisations (Gustison et al. 2019).Social interactions and acoustic environments are important factors that impact communication complexity (Manser et al. 2014;Leighton et al. 2021;Fan et al. 2022), and different acoustic communication needs call for different types of vocal signals (Pollard and Blumstein 2012;Briseno-Jaramillo et al. 2017;Sigafoos and Gevarter 2019;Wang et al. 2020).Consequently, it is reasonable to assume that the increasing complexity of communication context or communication needs may result in an enlargement of animal call repertoires (Elie and Theunissen 2018).
Female mate preference often favours males of a specific size, as body size is correlated with offspring fitness (Kolliker 1999;Cox and Calsbeek 2010;Calsbeek et al. 2015;Nali et al. 2023).However, detecting a potential mate's body size based on visual cues at night is challenging, which leads to reliance on vocal communication (Sun et al. 2022).Although there are numerous studies on the correlation between male body size and the properties of advertisement calls, the associations between body size and the acoustic properties of multiple call types have been neglected (Barbadillo and Castanet 2002;Castellano et al. 2002;Gingras et al. 2012;Zhu et al. 2016aZhu et al. , 2016b;;Escalona Sulbaran et al. 2019).During breeding seasons, many frogs form large choruses and call loudly to attract females and repel rival males (Aguiar et al. 2022).Therefore, it is reasonable to assume that the different call parameters of multiple call types may be associated with male body size, which could enhance the accuracy of both males and females in accessing competitors and potential mates through acoustic signals (Burke and Murphy 2007;Elie and Theunissen 2018;Fan et al. 2019).Furthermore, although most females prefer large males, some species prefer medium-sized males (Charlton et al. 2007;Deb et al. 2012;Zhu et al. 2016b).Despite the association between call frequency and body size having been studied extensively (Zhu et al. 2016b;Zhang et al. 2020;Kozhevnikova et al. 2021), whether call complexity could reflect male body size is still unknown.
Combining different call types with varied sequences is considered common in primates and birds (Bohn et al. 2009;Xiao et al. 2018;Engesser and Townsend 2019;Fan et al. 2022).These various call sequences may express different meanings to receivers (Braaten et al. 2006;Suzuki et al. 2016).In contrast, the calling behaviour of anurans is often thought to be simple and stereotyped (Gerhardt and Huber 2002).Studies on vocal sequences of birds and mammals have been conducted, while anuran call sequences have been neglected (Bernal et al. 2009;Suzuki et al. 2016;Xiao et al. 2018).The call components network is a new method that can reflect the transition of vocal signals and provides an intuitive way to visualise and quantify the associations between different call types (Bhat et al. 2022).Meanwhile, the call network can reflect social stability and call complexity (Gustison et al. 2019;Morino et al. 2021).Although previous studies have mainly focused on studying the communication function of different calls from the level of a single type, the transition among multiple call types has been ignored.
The Hainan frilled treefrog (Kurixalus hainanus) is a tropical anuran species.It emits specific note types that can be combined to produce single-note call types as well as composite calls (Zhu et al. 2017b).These calls are predominantly composed of two notes: note A acts as an advertisement call, while note B is an aggressive call used to suppress competitors' advertisement calls (Zhu et al. 2017b).Compound calls containing note A and note B, 'call AB', play a vital role in male-male competition and female choice (Zhu et al. 2017b).Males produce more compound calls as competition escalates (Zhu et al. 2017a).Additionally, we discovered a new note type (named note C) in field recordings, but have yet to determine its acoustic structure and function.Since males produce call C both alone or in combination with note A or/and note B to form diverse compound calls (e.g.call AC, call BC, call ABC), investigating the function of call C and the compound calls composed of different note types is important for uncovering the evolutionary mechanisms of the graded complex calls.
We are curious about whether the mechanism behind the complex ways of note combination in male Hainan frilled treefrog has evolved in response to the primary functions of calls, which are attracting females and competing with males.Therefore, we investigated the functions of different call types using multiple methods.In this study, we analysed the temporal and spectral structures of graded calls.Using call components network analysis, we investigated the functional relationships between graded calls and male call complexity and verified the function of call C via male playback experiments.We also examined the correlation between male body size and the properties of multiple call types.Finally, we compared the subtle functional differences between different graded calls to determine whether increasing communication needs may expand the communication network.By conducting these analyses, we aimed to gain a better understanding of the functions and complexities of anuran acoustic communication.

Vocalization recordings
Natural vocalisations were recorded from 62 male Hainan frilled treefrogs on Mt.Diaoluo in Hainan, China (18.44°N, 109.52°E, with an elevation of 933 m a.s.l.) from April to May.The vocalisations of each male frog were recorded for 3-5 min, using a directional microphone (ME66, Sennheiser, Germany) connected to a digital recorder (PMD 661,16 bit,44.1 kHz,Marantz,Japan).The microphone was placed approximately 1 m away from the frog, and calls were recorded in their natural habitat from 20:00 to 24:00 h (temperature: 20.4 ± 0.52°C, relative humidity: 94.9 ± 3.96%).Before being returned to the habitat, each male was given a unique toe-clip number to prevent recording or testing repeatedly, and their body mass (BM), snout-vent length (SVL), and head width (HW) were measured.We analysed several acoustic properties of the calls, including note duration (ND), fundamental frequency (FF), dominant frequency (DF), and inter-note intervals (INV).Calls were classified into types by their difference in acoustic properties, sonograms and spectrums.

Call components network analysis
We analysed 62 male natural recordings and categorised each call component by acoustic properties.To determine the relationship between different call types, we organised the preceding and following calls into a transition probability matrix.Each cell in the matrix represented the total instances of the preceding call type and the following call type.To calculate the expected value for each transition, we multiplied the column frequency with its corresponding total row observation value.We used a modified chi-square value with one degree of freedom per cell (Clark 1994) to determine which of the dyads in a row were significant.To calculate the transition probabilities, we divided the total instances of each call type by the corresponding row total (Bernal et al. 2009).

Male-evoked vocal response experiments
We conducted male-evoked vocal response experiments from April to July.A total of 60 males were tested in the playback experiments.We placed male frogs in a natural environment similar to their breeding sites and where no nearby males were calling.Each male was tested with eight stimuli (1A, 3A, 5A, 1B, 3B, 5B, 5A2B, 5A5B) during 21:00 h to 24:00 h (temperature: 20.9 ± 0.72°C, relative humidity 97.4 ± 3.58%).Every test stimulus was broadcast at a 5-s interstimulus interval (approximately equal to the mean inter-call intervals in natural recordings) and repeated for 3 minutes.A portable field speaker (SME-AFS, Saul Mineroff Electronics, Elmont, NY, USA) located 1 m from the male was used to broadcast each stimulus.Male response calls before, during, and after each stimulus were recorded for three minutes respectively, using an Aigo R5518 recorder (Aigo Digital Technology Co. Ltd., Beijing, China).We measured the sound pressure levels (SPLs) of each test call with a sound level meter (AWA 6291, Hangzhou Aihua Instruments Co.) and the 'fast root-mean-square' amplitude of the stimulus was 80 dB SPL (re 20 mPa, Z-weighted).Additional details can be found in Zhu et al. (2017b).

Analysis and statistics
We use Adobe Audition 3.0 (Adobe, USA) to confirm the differences in the sonograms and spectrums of different call types and analysed several acoustic properties of the calls.We classified the call properties as static or dynamic based on interindividual variability (coefficient of variation, CV) during bouts of calling.Static properties had mean CVs of 8.7% or less, while dynamic properties had mean CVs of 9.7% or greater (Gerhardt 1991).We created Venn diagrams using jvenn (Bardou et al. 2014).Nonparametric tests were used to detect the differences between notes (IBM SPSS Statistics 25).Data were statistically analysed and graphs were created using Origin 2022b software.Correlation analysis was used to determine the associations between male body sizes and call properties or network diagram properties.The effects of stimulus type and playback time (before, during, and after playback) on male evoked vocal responses were analysed by constructing generalised linear mixed models (GLMM) using the package lme4 (v.1.1-19) in the program R (v. 4.0.3).We used Cytoscape 3.9.1 to represent graphically the network of associations between different call types, including different note types and note numbers, and quantify the transitions between them.Data were expressed as Mean ± SE, and p < 0.05 was considered statistically significant.

Ethics note
We followed all relevant international, national, and institutional guidelines for the care and use of animals.Our experimental procedures involving animals were approved by the Animal Care and Use Committee of Chengdu Institute of Biology, CAS (CIB2014031008).

Acoustic analysis
We analysed a total of 3326 calls from 62 male individuals and identified three types of notes in male calls, note A, note B, and note C, which differed in note duration, dominant frequency, and number of notes per call type (Figure 1, Table 1 and Table S1).The duration of note A is significantly longer than that of note B or note C, and the duration of note C is shortest (Table 1 and Table S1).The dominant frequency of note B is the highest, and the dominant frequency of note C is remarkably lower than the dominant frequency of note A and note B (Table 1).There were significant differences between the number of notes in the three call types, with call A being composed of 2 notes on average, call B 1.4 notes on average, and call C 5.4 notes on average (Table 1 and Figure 2).Call B was the most common call type (53.9% of total calls), and the total number of call C was significantly less than the number of call A or call B (4.3% of total calls; Figure 3).
Male frogs produced diverse compound calls with two or three types of notes (Figures 2 and 3).The most common compound call is call AB (56.4% of total calls), composed of 3.88 A notes and 2.85 B notes on average (Figure 3).Call BC is less observed than call AB (18.1% of total calls), composed of 2.2 B notes and 3.47 C notes on average.While the call AC and call ABC appear less frequently in our recordings (Figure 3).Some males produced a compound call AB that had a transitional note type (i.e.note A-B) from note A to note B (Figure 4).Meanwhile, we found that most call properties were dynamic except for the dominant frequency of three note types (Table 1).

Call components network analysis
Call components network analysis from 62 individuals showed that their vocalisations were highly complex, which could be classified into several primary types with 33 transitioning ways (Figure S1).Call 1B, consisting of one note B, was the most common.Call 1A and call 1B tend to be followed by the same type of calls, but this did not occur in call 1C.Call C tended to be produced before or after call B. When producing advertisement calls, males tended to add notes one by one, with switching from 1A to 2A and from 2A to 3A being more frequent than the opposite sequence (Figure 5).
The sequence is significantly different between call type transitions (χ 2 = 2476.38,n = 62, p < 0.01, Tables S2-S3).Using chi-square values to estimate each  cell of the matrix, we found that repetitions of the same call type appeared significantly higher than expected in call A, call B, call C, and call AB (χ2 ranged from 20.39 to 754.47, p < 0.01; Table S3).However, the number of transitions from call C to call A (χ 2 = 20.39,p < 0.01), and from call B to call A (χ 2 = 474.49,p < 0.01) occurred significantly less than expected.Interestingly, males tended to produce calls according to a certain sequence of call A-call B-call C. The number of transitions from call A to call AB is 2.25 times that from call AB to call A, the number of transitions from call AB to call B is 1.3 times that from call B to call AB, and the number of transitions from call A to call C is 1.55 times of that from call C to call A (Table S2).In addition, the call AB is usually produced before or after call A and call B (73.6% of the total), with males tending to produce call A (34.6% of the total) before call AB and to produce call B (53.7% of the total) after call AB.The transition between call BC and call B was more frequent than the transition between call BC and call A, call BC, and call C (Figure 5).In summary, our network analysis revealed complex and diverse vocalisations with distinct patterns in call types and their transitions.

Male evoked vocal responses
We recorded a total of 4320 minutes of calls and analysed male-evoked vocal responses for the number of calls C, notes C, and notes per call.In all recordings, a total of 60 frogs collectively produced 5418 call A, 3240 call B, and 1222 call C (Figure 6).Compared with nature recordings, males produced more call C when stimuli were played back (percentage of total call C in nature recordings: 4.5%; percentage of total call C during stimuli playback: 12.4%; χ 2 = 162.65,p < 0.01).
There is a statistically significant interaction effect between playback time and stimuli (total number of call C: χ 2 = 148.75,p < 0.01; the total number of note C: χ 2 = 805.2,p < 0.01).There are also significant differences in male evoked vocal responses between stimuli (total number of call C: χ 2 = 414.46,p < 0.01, n = 60; the total number of note C: χ 2 = 2757, p < 0.01, n = 60; GLMM) and playback time (total number of call C: χ 2 = 756.52,p < 0.001; the total number of note C: χ 2 = 5112.5,p < 0.001; Table 2).The stimuli containing note A (5A, 3A, 1A, 5A2B, 5A5B) elicited more call C during the period the stimuli were played back than before the stimuli were played back, the total number of call C respectively increased by 7.83 with 5A, 5.57 with 3A, 2.8 with 1A, 4.43 with 5A2B, and 4.91 with 5A5B.While stimuli that only contain note B (5B, 3B, 1B) have no significant effect on male calling responses during the period the stimuli were played back (Figure S2).These results are similar to the total number of note C in experiments, except the stimuli 1B which elicits more note C during the playback (Figure S2).For total numbers of call C, compound calls with more note B (5A5B) elicit more responses in call C than with less note B (5A2B) (Figure 6).During playback, responses to call 5A were greater than to 5A2B (p < 0.01), and 5A5B stimulation resulted in more responses than 5A2B (p = 0.02, Tukey's method; Figure 6, Table S4).Males tended to produce more call B and call C in response to call A and call AB (Figure 6).

Relationship between call properties and body size
We used correlation analysis to determine if male call properties are associated with body size (Figure 7).While the dominant frequency of note C was not significantly correlated, both the dominant frequencies of note A and note B were significantly negatively correlated with body size (Table S2 and Figure 7).Regarding the fundamental frequency, only note A was positively correlated with SVL and HW (Table S5 and Figure 7).The duration of note A was significantly positively correlated with body size (Table S2 and Figure 7), and the inter-note interval of call C was positively correlated with BM (Table S2 and Figure 7).There are also some links between call properties.The numbers of note B per call and note C per call, maximum note numbers of call B and call C, were positively intercorrelated.The total number of notes, calls, and compound calls per minute were significantly correlated with the number of note B per call, the maximum note number of call B, and the inter-call interval of call C. Some males produced calls with only one type of note, while some produced calls with different types of notes (Figure S3).Interestingly, we did not find a significant correlation between body size and call complexity (Figure 7).Likewise, there was no significant correlation between the dots (call type) and lines (transition between call types) of the call components network with the body size (Figure 7).

Discussion
Using multiple methods, we investigated the functions of graded calls.Our study identified five main findings: (1) male Hainan frilled treefrog can produce three kinds of call notes that are combined in many different ways; (2) note C is a different note type from note A and note B in structure and function; (3) there was an intermediate note type between the advertisement call and aggressive call; (4) the call network showed interesting functional correlations between different call types; (5) some call properties are correlated with male body size, while there is no significant correlation between call complexity and male body size.
Our previous study revealed that call A plays a major role in female attraction (Zhu et al. 2017b), while males produce more call B than call A to suppress competitors.Males may produce more aggressive calls but reduce advertisement calls in male competition because of energetical constraints (Schwartz 1989).Note C was found to be distinctly different in temporal and spectral structures compared to note A and note B. Our field observations revealed that when two males were at a close distance, they exhibited a higher frequency of call C. The result of the male-evoked vocal responses shows that call C is similar to call B, but not totally the same (Figure 6).Based on these results, we define call C as an encounter call, which performs a warning function to an intruder who is calling too close to a resident male to resolve disputes without physical combats (Wells 1977;Toledo et al. 2014).Male cricket frog, Acris crepitans blanchardi, produced similar graded aggressive calls.Male calling appeared to be graded in response to intruder distance, increasing in aggressiveness with decreasing distance (Wagner 1989).
Male Hainan frilled treefrogs tend to repeat simple calls, such as call A or call B, as observed in call network analysis, which may reduce the masking effect of background noise (Freeberg et al. 2003;Moors and Terhune 2004;Grace and Noss 2018), and help receivers localise signallers more accurately (Fan et al. 2022).Competitive pressure is an important factor to generate graded calls (Bernal et al. 2009;Zhu et al. 2017a).Male Hainan frilled treefrogs tend to add the note number of advertisement calls one by one when competing with other males.This adaptive strategy helps maintain attractiveness to females in a competitive environment (Gerhardt et al. 2000;Bernal et al. 2009).Our analysis further revealed that males tend to produce call B after call A and produce call C after call B. Compound calls AB and BC were the most common, while call BA and call CB were less frequently observed.Compound calls with note A and note C only appear in the sequence of AC (i.e.call AC).The progression of call types suggests potential evolutionary processes, such as the aggressive call (call B) evolving from the advertisement call (call A).This progression is not always a sudden switch but may exist as a transitional form, as seen in an intermediate call produced by dendrobatid frogs (Wells 1980).The presence of a transitional note type A-B reveals a possible evolutionary pathway in Hainan frilled treefrogs, where aggressive calls may have evolved from advertisement calls.Overall, our findings suggest that the evolution of male call types is influenced by increasing competition and communication needs, and highlights the importance of considering the evolutionary and ecological context of vocal communication in understanding animal calling behaviour and social dynamics.Male Hainan frilled treefrogs can produce diverse compound calls with 3 note types that are combined in different ways with different communicative significance.The compound call AB appears to be a transitional call type between call A and call B, similar in function to an advertisement call (Table S2; Zhu et al. 2017b).Similarly, call AC is mainly sequential to call A and call C (61.5% of the total), and may act as a transitional call type.Call BC and call ABC are primarily associated with call B in the call network (65.6%, 66.7%), and may function as an aggressive call.The monotonous calls may lead to habituation in receivers and negatively impact mating choices (Reichert 2009;Howell et al. 2019;Fuss 2021;Berz et al. 2021).Inversely, diverse compound calls can reduce habituation (Leger et al. 2000;LaBarbera et al. 2020).Previous studies also showed that compound calls were favoured by females over simple calls (Bernal et al. 2009;Zhu et al. 2017b).Thus, producing diverse compound calls may help male Hainan frilled treefrogs keep the novelty of calls and reduce female habituation.The complexity of compound calls may increase working memory for mating signals, also explaining the preference for such calls in females (Akre and Ryan 2010).
Species recognition and individual recognition are critical in social animal communication (Fang et al. 2015;Elie and Theunissen 2018).Previous studies have suggested that static call properties may play a major role in conspecific identification, while dynamic parameters may play a major role in individual identification (Gerhardt 1991;Yu et al. 2011;Morais et al. 2012;Fang et al. 2018).Our result shows that only the dominant frequencies of the three types of notes were static, indicating that dominant frequencies may be crucial in identifying conspecific males.Similar to our previous study, both dominant frequencies of advertisement calls and aggressive calls are negatively correlated with body size, allowing males to deliver messages about their body size to potential mates and competitors.Females prefer medium male frogs, thus the highly stereotyped dominant frequency of note A may result in the female selection that is stabilising static properties (Zhu et al. 2016a).Dynamic call properties such as note number and call duration can deliver messages about different environmental and social conditions (Elie and Theunissen 2018), and animals may use multiple call properties to transmit various types of important information (multiple-messages hypothesis) (Koren and Geffen 2008;Morais et al. 2012).However, female mate choices may be constrained by hearing sensitivity or deception from calling males (Wiley 2006;Tumulty et al. 2022).Additionally, physiological and biomechanical constraints may explain the correlation between body sizes and call properties, such as an increase in larynx size changing the dominant frequency of callers (Gridi-Papp et al. 2006;Reichert and Hobel 2018).Interestingly, we found that the call properties of note A and note B are less correlated with each other than those of note B and note C, suggesting that note B and note C share similar communication functions and have a similar pattern of variation between individuals (Reichert 2013).Previous studies have shown that females may use multiple call properties to assess the quality of potential mates (Candolin 2003;Zhu et al. 2016a;Fan et al. 2022), and males may use multiple call properties to assess the strength of competitors (Velez and Guajardo 2021).Thus, the use of multiple call properties in both sexes may improve the accuracy of mate choice and reduce intrasexual physical conflict (Elie and Theunissen 2018).
In conclusion, our study highlights the diverse call repertoire of Hainan frilled treefrogs, with different call types serving distinct functions towards different receivers in complex contexts.Call A is produced to attract females, while calls B and C are used to suppress competitors and keep them away from calling sites.Compound calls allow for graded call complexity, providing an advantage over simple calls in male competition and female mate choice.The social complexity hypothesis argues that groups with complex social systems require sophisticated communication systems to regulate interactions among group members (Freeberg et al. 2012).In the breeding season, males need to interact with different conspecific individuals (Vilaça et al. 2011;Leverett et al. 2022).Male Hainan frilled treefrogs evolved three kinds of notes, which can be emitted isolated or combined, which has helped to meet the complex communication needs.Our findings suggest that males can use a graded communication system by combining different note types in different sequences to communicate in varied contexts (Zhu et al. 2017a(Zhu et al. , 2017b)), providing clear cues about signallers' state and the surrounding conditions for receivers (Rauber et al. 2020;Fan et al. 2022).Our results indicate that the social needs of both signaller and receiver exert selective pressure to promote the evolution of graded calls.Importantly, our study suggests that call complexity does not reflect signaller body size, which can change according to social context.In summary, we have discussed the possible functions of the graded call system of Hainan frilled treefrogs, providing deeper insights into the complexity of anuran acoustic communication and new evidence for the social complexity hypothesis.

Figure 3 .
Figure 3.A Venn diagram showing the number of three main calls in natural recordings.Where the circles intersect, we show the number of calls with two or three types of notes.The size of each list represents the total number of calls A, B, or C. The bottom diagram represents the number of calls with one, two, or three types of notes.

Figure 4 .
Figure 4.The amplitude-modulated waveform and spectrogram of compound call AB (a), with the transitional note type A-B (b).

Figure 5 .
Figure 5.Call components network showing the transition of specific call types.The size of each circle represents the quantity of each type of call.Arrows indicate the direction of call transition and the thickness of lines represents the number of call transitions.

Figure 6 .
Figure 6.Results of male-evoked vocal responses showing the total number of call A, call B and call C produced during the stimuli were played back.The different playback stimuli are labelled according to the numbers of each note type contained in each stimulus and include: 5A, 5B, 3A, 3B, 1A, 1B, 5A2B, and 5A5B.The Generalized Linear Mixed Model (GLMM) was used, and different superscript letters indicate significant differences among different treatments (p < 0.05).The data for call a and call B comes from Zhu et al. (2017b).

Figure 7 .
Figure 7. Correlation heatmap between call characteristics and body size.Redder colours indicate that r values are positive, and bluer colours indicate that r values are negative.The plot * indicates a significant correlation (p < 0.05).NNA, NNB, NNC: number of note a per call, note B per call, note C per call, respectively; ICIA, ICIB, ICIC: inter-call intervals of call A, call B, call C; NNA max , NNB max , NNC max : maximum note number of call A, call B, call C; DFA, DFB, DFC: dominant frequency of note A, note B, note C; FFA, FFB, FFC: fundamental frequency of note A, note B, note C; NDA, NDB, NDC: duration of note A, note B, note C; INIA, INIB, INIC: inter-note intervals of call A, call B, call C; TNC: total number of calls per minute; TNN: total number of notes; notes/call: number of notes per call; compound calls: number of compound calls per minute; DCN: dots of call network; LCN: lines of call network.

Table 1 .
Call properties of male Hainan frilled treefrogs. of note A per call; NNB: number of note B per call; NNC: number of note C per call; NDA: duration of note A; NDB: duration of note B; NDC: duration of note C; DFA: dominant frequency of note A; DFB: dominant frequency of note B; DFC: dominant frequency of note C; FFA: fundamental frequency of note A; FFB: fundamental frequency of note B; FFC: fundamental frequency of note C; INIA: inter-note intervals of call A; INIB: inter-note intervals of call B; INIC: internote intervals of call C; CV: coefficient of variation.

Table 2 .
The statistical results of male call responses produced in the periods before, during, and after the stimuli were played back.
'bef', 'dur', 'aft' represents before, during, and after respectively.5A, call containing five A notes; 5B, call containing five B notes; 5A2B, call containing five A notes followed by two B notes.Other abbreviations are labelled accordingly.The numbers in each row are p values.