On the vocal repertoire of the monkey frog Phyllomedusa venusta: distress call and the presumed non-existing advertisement call

ABSTRACT The anuran advertisement call has been widely used in taxonomic and evolutionary studies due to its essential role in species-specific recognition. For a long time, it has been presumed that P. venusta did not emit mating calls. Here, we observed the entire vocal activity period of P. venusta over the night and recorded male calls and a female distress call. We recognized and described a single male call type, identified as the advertisement call. This call has an extremely low timbre, and it is composed of a single or a sequence of pulsed notes. The multi-noted calls were usually emitted in intrasexual agonistic contexts. Although both call duration and note number have a high coefficient of variation values, they are linearly correlated, so they should be acting as a single independent variable in the P. venusta calls. As expected, the distress call is longer and with a higher timbre than the advertisement call. Interspecific comparisons were made based on the ratio between advertisement call dominant frequency (DF) and body size. This ratio is similar to those of the species of the P. tarsius group.


Introduction
Most anuran species use vocalizations to communicate, and different kinds of calls may be emitted to respond to different ecological interactions (Köhler et al. 2017).For anurans, different call types, such as courtship calls, territorial calls, release calls, and distress calls, have been named based on their functions.These functions have usually been inferred by interpreting the context in which the call is emitted (e.g.Bogert 1960;Toledo et al. 2014;Köhler et al. 2017).Although this procedure may sometimes be speculative and oversimplified (Hepp & Pombal 2019), a first description and classification are important to characterize repertoire of calls.This allows further studies to identify homology and improve classification based on more characters and their variation (e.g.Hepp et al. 2017;Hepp & Pombal 2020).
The most reported anuran vocalization is the advertisement call, which has two main functions: attract conspecific females to breed and establish territories (Duellman & Trueb 1994; see Wells 2007 and references cited therein for alternative call classifications).This call type plays an important role in speciesspecific recognition, serving as a pre-zygotic barrier (Blair 1941;Etges 1987).Hence, it is widely used in systematic and evolutionary studies (e.g.Robillard et al. 2006;Goicoechea et al. 2010;Gingras et al. 2013;Tonini et al. 2020).
The distress call is the most common defensive vocalization.It is emitted by frogs when restrained by a potential predator and can create an opportunity for escape (Toledo et al. 2009;Forti et al. 2018).Distress calls are acoustically distinguished from the other vocalizations by being loud and having a harmonic structure with a broad frequency range, ensuring that the call can be heard by different predators (Duellman and Trueb 1994;Toledo et al. 2009;Forti et al. 2018).This broad frequency range seems to be related to openmouth emission (Gridi-Papp 2008), a behavior observed during distress call vocalizations and usually used as a diagnostic feature of this call type (Toledo et al. 2014;Köhler et al. 2017).
Several studies have sought to describe the bioacoustical repertoire of the Phyllomedusa monkey frogs (e.g.Zimmerman & Bogart 1984;Abrunhosa & Wogel 2004;Silva-Filho & Juncá 2006).The availability of bioacoustical data for almost all species of the genus has also allowed their use in comparative studies (Röhr et al. 2020b;Bezerra et al. 2021).Nonetheless, the advertisement call of Phyllomedusa venusta Duellman & Trueb 1967 is still undescribed. This South American species occurs in Colombia, Panama, and Venezuela (Infante- Rivero et al. 2006;Frost 2021) and had been presumed to lack any kind of mating call (Duellman 2001;Ballesteros-Correa et al. 2019).Only recently, the first observation of calling activity during courtship context was reported in the locality of San Juan de Nepomuceno, department of Bolivar, Colombian Caribbean (Mendoza-Roldan 2017).Until this study, only release calls had been reported and described for P. venusta (Duellman & Trueb 1967;Duellman 2001).Mendoza-Roldan (2017), while observing the courtship behavior of P. venusta, reported calls emitted during a chorus, in which males seemed to be attracting females for mating (Mendoza-Roldan 2017).The behavioral observation in that study suggested that P. venusta does have an advertisement call (sensu Wells 1977).These calls emitted during courtship were described as soft 'chuckle' calls eventually repeated in staccato series (i.e. in quick sequences with clear inter-call intervals) of nine to 13 calls in three seconds (Mendoza-Roldan 2017).
Phyllomedusa venusta was first recognized and described based on morphological differences (Duellman and Trueb 1967).In 2006, it was assigned to the Phyllomedusa tarsius group by Barrio-Amorós, although its phylogenetic placement within the genus is still unknown (Castroviejo-Fisher et al. 2017).The species of the P. tarsius group are morphologically similar, and some of them occur in sympatry (De La Riva 1999;Barrio-Amorós 2009;Frost 2021).Therefore, the status of P. venusta as a different species from some of the species of this group has been questioned (e.g.Barrio-Amorós 2009;Castroviejo-Fisher et al. 2017).Differences in the advertisement call have been used as solid evidence of full species status for some species (e.g.Hepp et al. 2015;de Andrade et al. 2018).The assessment of advertisement call differences between those species may shed light on those questions.
Recent studies have demonstrated that the dominant frequency (DF) and body size (snout-vent length; SVL) of Phyllomedusidae species reflect their phylogenetic relationships (Röhr et al. 2020b;Bezerra et al. 2021).Both variables have usually been inversely correlated for several anuran species (e.g.Wagner 1989a;Köhler et al. 2017;Tonini et al. 2020).Therefore, comparing the relationship between dominant frequency and body size (hereafter referred to as DF-SVL) of P. venusta with its congeners may be helpful to investigate whether P. venusta is similar to those species of the P. tarsius group.
In this study, we describe the advertisement and distress calls of P. venusta for the first time.We first explored the intraspecific bioacoustical variation among calls emitted by males of P. venusta during the chorus to investigate whether more than one call type is emitted in the male repertoire (e.g.isolated calls vs. staccato calls; Mendoza-Roldan 2017).Additionally, we compared P. venusta's DF-SVL with those of other phylomedusid species under a phylogenetic context to investigate their similarity, focusing on the P. tarsius species group.

Field recording
Phyllomedusa venusta's vocalizations were recorded from 12 to 15 October 2017, after heavy rainfall.The species' vocal activity was observed over the entire activity period, from 20:00-4:30 h ad libitum, sampling all occurrences (Altmann 1974).The mean temperature was 25°C, and the relative humidity was 90% each night.Recordings were performed by Mendoza-Roldan J.S using a Sony PCM A10 digital recorder with a built-in microphone, at a sampling rate of 44,100 Hz and sample size of 32 bits.The recorder was placed on a tripod in front of the vocalizing males at ca. 1 m.The four recording files of the males (ca.two hours of recording) and one from the female (ca.four minutes of recording) were deposited in the Fonoteca Neotropical Jacques Vielliard (FNJV 45,547).Two recorded males were collected as vouchers and deposited in the Amphibian Collection of the Universidad de Los Andes, Bogotá, Colombia, under catalog numbers (A 4106-4107).The recordings took place in two temporal ponds on a flood plain of the Magdalena River (6°46ʹ14.54"N;74° 6ʹ36.21"O;93 m a.s.l.), surrounded by secondary wet forest cover, located in Corregimiento Bocas del Carare, municipality of Puerto Parra, Departamento de Santander, Colombia, on the Middle course of the Magdalena River.
We performed an acoustic playback exposition to see how the calling males would behaviorally react as an anecdotal observation.Since our playbacks were only exploratory and were not standardized (see Fischer et al. 2013), we did not use any calls emitted in response to it in the present study.

Bioacoustical analyses
The recordings include calling individuals from two different ponds, one with a large chorus, in which 15 individuals were counted calling, and a smallest one, in which five calling individuals were counted; a total of 20 calling males recorded.We visually examined 499 male calls from all four recordings obtained from different moments of the chorus activity.We chose 128 calls, from six males, with high recording quality to measure and analyze in detail.We also analyzed one distress call emitted by a female kept inside a plastic bag (see below).
Acoustic edits and measurements were obtained on Raven Pro 1.5 (64-bit version) (Center for Conservation Bioacoustics 2014).All the advertisement call recordings were bandpass filtered between 100 and 2,500 Hz to remove background noise and calls of other taxa (e.g.orthoptera and other frog species).The distress call recording was not filtered considering the importance of the broad spectrum of this type of call (see below) and the original low level of background sounds of the recording (this call was recorded far away from the chorus).Temporal parameters were measured on oscillograms and spectral parameters on sonograms and power spectra with a window size of 512 samples (3 dB filter bandwidth = 124 Hz) for advertisement calls and 1024 samples (bandwidth = 61.9Hz) for the distress call.We used the window type Hann and time grid overlap of 99% for both call types.Quantitative features are given in the description as the range followed by the mean, standard deviation (SD), sample size (n), and coefficient of variance (CV) in parenthesis.We made the oscillograms and sonograms presented here with the seewave and tuneR packages in R environment (Sueur et al. 2008;Ligges et al. 2016; R Core Team 2020) using the same parameters given above.
Bioacoustical terminology followed Köhler et al. (2017), using the note-centered approach.Homology assessment followed the criteria and procedure given in Hepp & Pombal (2019) based on the preliminary comparison with recordings of other Phyllomedusa species (Table S1).

Intraspecific analyses
Mendoza-Roldan (2017) reported a significant acoustic variation during the choruses, mainly in the number of notes and call duration (from isolated calls composed of a single note to staccato calls with a long sequence of notes).To verify the existence of two call types, we conducted a Principal Components Analysis (PCA) with the natural logarithmic-transformed data of P. venusta male calls to check whether there is any clustering formation.Clusters with little or no overlapping variation would indicate different call types based on their acoustic features (see Köhler et al. 2017;Folly et al. 2018).In that case, the most explanatory variables responsible for the cluster formation could be suggested as diagnostic acoustic features for different call types composing a repertoire for males (see Folly et al. 2018).Since the PCA cannot handle inapplicable or missing data, we first removed the variable note rate, which is inapplicable to calls composed of a single note.We also ran a PCA with the male calls and the distress call together, including all variables except those related to pulses (inapplicable to the distress call; see below) to compare both calls.We ran both analyses using the 'prcomp' function of the 'stats' package with argument 'scale = TRUE' (correlation matrix) and the 'ggplot2' package for graph production (Wickham 2009).
Additionally, we calculated the coefficient of variation (CV) for each acoustic variable to assess which of them are more labile (= with the highest CVs) considering our dataset (Gerhardt 1991;e.g. Gambale & Bastos 2014).Then, we looked for clear variation gaps within these traits in the PCA.For that, we considered all the 128 calls from the six males, similarly to the between-individual CV in Guerra et al. (2017).To verify whether temporal or spectral variables might be more labile, we classified the acoustic measurements into three categories: (i) spectral, (ii) temporal, and (iii) structural variables.The spectral variables are related to the frequency of the call and are better observed on the spectrogram; temporal variables are related to time and better observed on the oscillogram; and structural variables are those more related to behavioral traits and contexts, such as the number of notes/pulses emitted in the call (see Köhler et al. 2017).Spectral variables (i): BW_90, Avg_Entropy, Freq_95, Freq_75, Freq_25, Freq_5, Center_Freq, Max_Freq.Temporal variables (ii): Pulse_Period, Note_Rate, Note_Dur, Dur_50, Call_Dur.Structural variables (iii): Pulse_Numb, Note_Numb.
After finding the same two variables (= call duration and number of notes) used by Mendonza-Roldan (2017) to differentiate isolated calls from staccato calls among the most labile (highest CVs), we ran a nonparametric correlation analysis (Spearman's rank correlation coefficient) to test whether both traits act as a single independent variable.We tested the distribution normality of these variables through Shapiro-Wilk test.The correlation analysis was performed with the ggplot2's expansion 'ggstatsplot' (Patil 2018).Both analyses were performed in R environment (R Core Team 2020).

Interspecific analyses
For interspecific comparative analyses, we compiled the body size (accessed through the snout-vent length measurement -SVL) and the call dominant frequency (DF) of 14 Phyllomedusa species (including P. venusta) and other 17 close related phyllomedusid species: nine Pithecopus spp., four Callimedusa spp., three Agalychnis spp., and one Phasmahyla sp.The dominant frequency values for these species were directly measured on recordings obtained from public archives and specialists' private collections or obtained from the literature (Table S1).The male body size values of these species were directly measured from specimens deposited in zoological collections or obtained from the literature (Table S1).
We compared the SVL and DF of all species (a total of 31 species including P. venusta) through linear regression with ggplot2 package in R environment (Wickham 2009).We also used label modifications of geom_label_repel functions of the ggrepel package in R environment (Slowikowski 2018;R Core Team 2020).The body size (=SVL) and DF are highly correlated in phyllomedusid species, and their variations are significantly related to the phylogeny (Röhr et al. 2020b;Bezerra et al. 2021).These results make this relationship (DF -SVL) a good comparison proxy within a phylogenetic framework (see below).
To check for phylogenetic structuring among the species in the linear regression, we created a 'phyllomorphospace' with DF -SVL variation with the phyloScattergram function of the phytools package in R environment (Revell 2012; R Core Team 2020).We also reconstructed the ancestor character state using the functions phyloScattergram and contMap of this same package to confirm the variation pattern of these variables over the phyllomedusid tree.
Phylogenetic topology and branch lengths follow those given in Duellman et al. (2016;Figure S1).Since the P. venusta phylogenetic placement is still unknown, we treated it as incertae sedis within Phyllomedusa, placing the species in a conservative position at the base of the genus (Figure S1).

Chorus and distress observations
We observed 5-15 calling males per spot (two ponds sampled over the nights; see above).These males were found perched on vegetation between a few centimeters to six meters above the ground (usually about 1 m) (Figure 1A).Isolated advertisement calls (composed of a single note) were commonly emitted during the beginning of the chorus (20:00 hrs).At this time, males had emerged from the forest leaf litter or the canopy and moved near the pond's margins, searching for a suitable vocalizing perch.A high-density chorus was formed by midnight (00:00 hrs).Amplecting pairs were observed moving and constructing leaf nests on vegetation (see Mendoza-Roldan 2017) near the margins of the pond or higher in the forest canopy (2:00-4:00 am; Figure 1B-C).Males usually emitted staccato calls (calls composed of series of notes) after being stimulated by playback (Figure 1D).Male-male physical interactions were observed during the night (Figure 1E).The hylids Dendropsophus ebraccatus and Scinax ruber and the leptodactylids Engystomops pustulosus and Pseudopaludicola pusilla were also observed vocalizing during the same period and at the same sites.
One female (Figure 1F) emitted distress calls while temporally contained in a plastic bag under debris removed from the local vegetation.She vocalized hidden in the middle of the debris, and, unfortunately, we could not verify whether the calls were emitted with open mouth (see discussion).

Intraspecific comparative analyses
We found all the male calls from all recordings clustered in the PCA (Figure 2).The first two principal components (PC) represent 61.97% of the call variability (PC1 = 39.46% and PC2 = 22.51%).Eigenvalues of 3.0 and 1.6 for the PC1 and PC2, respectively.The biplot of both components did not show any cluster discrimination between the staccato calls, composed of several notes, and isolated calls, composed of a single note.For the first component, the most explanatory variables were BW_90%, Freq_95%, and Freq_5%, being the latter negatively related to the first two.For the second component, the most explanatory variables were Freq_25% and Center_Freq, which are positively related to each other (Figure 2).
The female call differs from the male calls mainly by having higher values for the spectral variables (Figure S2).Note duration (Note_Dur) also explains part of the discrimination.
Regarding the CV, we found the spectral variables as having the lowest coefficient values (Figure 3), varying from 9 to 22.4 (� x = 13.2 ± 6.0; n = 8 variables).The CV of the temporal variables varied from 11.4 to 158.0 (� x = 74.4± 72.7; n = 5 variables), while that of the structural variables varied from 19.1 to 128.5 (� x = 73.8± 77.4; n = 2 variables).Among these variables, those related to pulses, the note duration, and the repetition rate had the lowest values (Figure 3).In contrast, call duration and those related to it, i.e.Dur_50 and Note_Numb (see below), had the highest CV values (Figure 3).
The variances in the call duration and the number of notes used to differentiate isolated calls from staccato calls are highly correlated to each other (Figure 4; Spearman's R = 0.84, p ≤ 0.001).These variables vary continuously without any clear gap separating clusters that could indicate more than one call type with different call duration and number of notes (Figure 4).
In summary, our exploratory analysis on intraspecific variation did not show any clustering among male calls that could justify the recognition of more than one call type.Our field observations indicate that this call type has a primary mating function and possibly an additional male-male distance maintenance function (see behavioral observations above).Both functions have been attributed to the so-called advertisement call (Wells 2007;Köhler et al. 2017).Therefore, we recognize and describe a single call type emitted by males in the chorus, considered the advertisement call.Additionally, we describe the distress call emitted by a female.

Advertisement call
The advertisement call (Figure 5A) is composed of a single pulsed note or a sequence of up to 22 pulsed notes (� x = 3.4 ± 4.4; Mo = 1; n = 128; CV = 128.5),with the call duration ranging from 0.018 to 3.416 s (� x = 0.428 ± 0.677; Mo = 0.038; n = 128; CV = 158.0).When the call has several notes, the amplitude peak of the call is usually at the first note.There is then a significant reduction of amplitude in the second note, after which the amplitude gradually increases until a second amplitude peak near the end of the call (Figure 5A).Half of the call energy (50%) is distributed in a period that varies from 0.007 to 0.882 s (� x = 0.181 ± 0.270; Mo = 0.019; n = 128; CV = 148.6),corresponding to between 6.7 and 83.0% of the call duration (� x = 41.2 ± 14.3; Mo = 38.5;n = 128), which means that the amplitude peak can be more sharp or flat depending on the call being observed.When the call is composed of a single note, this note has amplitude similar to the first note of the multinoted calls.Note duration varies from 0.009 to 0.052 s (� x = 0.027 ± 0.010; Mo = 0.019; n = 128; CV = 36.5)and note rate varies from 5.3 to 9.6 notes per second (� x = 6.9 ± 0.8; Mo = 7.0; n = 43; CV = 11.4).There is no increase or decrease in note duration throughout the multi-noted call.The note envelope is elliptic, usually with note rise time similar to fall time.Each note is composed of a sequence of 2 to 6 pulses (� x = 3.9 ± 0.

Distress call
The single distress call analyzed has 14 non-pulsed notes and has a duration of 5.894 s (Figure 5B).The amplitude peak is near the end of the call, with 50% of the call energy distributed in 2.690 s, which corresponds to 45.6% of the call duration.Note duration ranges from 0.098 to 0.301 s (� x = 0.249 ± 0.050; n = 14; CV = 20.2), with note rate of 2.4 notes per second (see Figure 5B).There is no increase or decrease pattern in note duration throughout the call.The first notes have an elliptic envelope, with similar note rise and fall times, while the last notes of the calls have a shorter rise time and tend to have an envelope with a triangular shape.The dominant frequency is at 843.8 Hz and matches the fundamental frequency (ca.800 Hz).A significant amount of energy is distributed in the higher band of the spectrum, giving a high timbre to the call.The bandwidth with 90% of the call energy is 9,000 Hz, with the center frequency at 2343.8 Hz.There are several clear visible harmonics, although the notes sound noisy due to deterministic chaos, mainly among the highest harmonics.The average entropy is 4.655 bits, a value higher than the maximum value of the advertisement call for this variable.Notes have an up-downward frequency modulation pattern (arch shape on the spectrogram).Moreover, there is a slight periodic frequency modulation within the notes and additional abrupt and irregular frequency modulations in some notes (Figure 5B).

Interspecific comparative analyses
The average male body size of P. venusta is similar to that of P. trinitatis and is only smaller than P. tarsius and P. bicolor.The dominant frequency of P. venusta has the lowest value on average for all the species included in the analyses (Figure 6A-D).The position of P. venusta in the phylomorphospace is close to that of P. tarsius, P. camba, and P. neildi (Figure 6C), all of which belong to the P. tarsius species group (Barrio-Amorós 2006;Castroviejo-Fisher et al. 2017).Phyllomedusa trinitatis is the only species of the group found in a distant position due to its higher dominant frequency (ca.1800 Hz, second harmonic, vs. 730-760 Hz, fundamental/first harmonics in the other species of the P. tarsius group).The comparisons between the ancestral state reconstructions of SVL and DF suggest that the species with larger body sizes tend to have lower dominant frequency (Figure 6A and B), and the phylomorphospace illustrates that the DF -SVL relationship has phylogenetic structuring (Figure 6C).In general, Pithecopus species have the smallest body sizes (SVL) and the highest dominant frequencies (DF), whereas the Phyllomedusa species have the largest SVLs and the lowest DFs (Figure 6).and DF resulted in a significant inverse correlation (y = 2330 −17.4 x, R 2 = 0.45, p ≤ 001; Figure 5D).The phylogenetic structure is particularly reflected by the SVL, whereas the DF pattern is less clear considering some unexpected dominant frequency values such as those of A. lemur, Pi. centralis, C. ecuatoriana, and Phy.trinitatis (Figure 6C and D).

Call repertoire of P. venusta
In the present study, we recorded P. venusta vocalizations from dusk to early morning (20:00-4:30 hrs) and noted changes in vocal activity along this period.This species emits only short and sporadic single-noted calls (= isolated or 'chack' calls; see Mendoza-Roldan 2017) earlier in the night, as noted by Duellman (2001), Mendoza-Roldan (2017), andBallesteros-Correa et al. (2019).After midnight, it exhibits an increase in vocal activity, often emitting advertisement calls composed of several notes (= 'staccato' calls in Mendoza-Roldan 2017).Other authors have described similar patterns for phyllomedusids, in which vocalizations between isolated males or at the beginning of the chorus are markedly different from those produced during intense chorus activity (Nali et al. 2015).Most anuran species reduce their vocal activity after midnight (Wells 2007), making the observation before 24:00 h the most productive period to invest effort (e.g.Wogel et al. 2005).Indeed, in P. burmeisteri, the vocal activity peak is considerably earlier, at 21:45 h (Abrunhosa & Wogel 2004).The presumption of the absence of complex calls or even of mating calls (Duellman 2001;Barrio-Amorós 2009) in P. venusta might be a consequence of its unusual vocal activity peak after midnight.Moreover, as in other phyllomedusids, P. venusta male call has low amplitude and may be masked by other species calling syntopically.
Here we have observed a call variation over the night, with shorter and isolated calls emitted at the beginning of the night.Our results suggest the occurrence of a single graded call type emitted by males  during the night.Since this call type seems to have mating and male-male distance maintenance functions, we identified this call as being the advertisement call (Duelman & Trueb 1994;Wells 2007;Köhler et al. 2017).Our call delimitation approach was based on looking for gaps within the acoustic trait variations, i.e. we looked for evidence in the acoustic data that could justify classifying the observed variation into different categories (= calls; see Hepp & Pombal 2019).Gaps in continuously varying traits should allow recognition of different biologically meaningful categories (e.g.discrete calls in Wells 2007).Otherwise, categorization would include arbitrary thresholds (Thiele 1993;Hepp & Pombal 2019).Even the most variable features (e.g.note number and call duration) vary continuously among male calls.
Still, these findings do not rule out the hypothesis of a graded call with different communication functions (e.g.Acris crepitans, Ololygon centralis, and Boana goiana ;Wagner 1989a;Bastos et al. 2011;Dias et al. 2017).Some studies have reported variation of a single call type throughout the night, possibly associated with a trade-off between establishing calling sites and female attraction (Wells 2007;Bastos et al. 2011;Dias et al. 2017;Köhler et al. 2017).In P. venusta, calls composed of one note can function primarily as advertisement calls, and multi-noted calls can progressively act more as a signal of aggression as more notes are uttered (Wagner 1989b;Grafe 1995).Our behavioral observation indicates that these latter calls composed of more than one note are usually emitted near the vocal activity peak during high density choruses and after exposing calling males to acoustic playback.It suggests that longer advertisement calls composed of several notes might be more related to male-male agonistic contexts as reported for other anuran species (Sinsch & Joerman 1989;Wagner 1989b;Grafe 1995;Folly et al. 2018).Moreover, Mendoza-Roldan (2017) observed that these multi-noted calls (= staccato calls) cause substantial vibration in the branches as a result of the rapid contraction of the body wall (see also Duellman 2001), which might suggest a multimodal signal with a substrate vibration, a component often emitted during intrasexual contests (De Luna et al. 2010;see Folly & Hepp 2019).
In P. tetraploidea, differentiable call types have mating and aggressive functions called advertisement and interacting calls, respectively (Pombal & Haddad 1992).In this species, the calls are distinguished by qualitative features such as grouped pulses in the advertisement call (vs.isolated pulses in the interacting call; Pombal & Haddad 1992).Other Phyllomedusa species that have advertisement calls varying considerably in the number of notes (e.g.P. bicolor, P. neildi, and P. vailanti; Zimmerman & Bogart 1984;Barrio-Amorós 2006) might have similar correspondences to the continuum of mating-agonistic function found in P. venusta.For instance, in P. bicolor and P. tarsius, the advertisement calls are usually composed of a single or a few notes, and calls emitted in agonistic contexts have similar notes but in greater numbers (Zimmerman & Bogart 1984).
Our acoustic playback expositions were only exploratory and were not standardized (see Fischer et al. 2013).Further experimental studies should be conducted to test the aggressive component of multinoted calls observed here (e.g.Bastos et al. 2011) as well as whether receivers might perceive branch vibration as part of a compound signal or it is only a meaningless side-effect of the vocalization (e.g.Lopez et al. 1988).
We calculated the coefficient of variation (CV) to verify which variables in our dataset are more labile and might have acoustic differences to diagnose different male call types.We did not classify variables as static or dynamic (sensu Gerhardt 1991) since our sample was not designed for such a task (cf.Gambale & Bastos 2014;Guerra et al. 2017;Röhr et al. 2020a).There is no clear threshold between these categories for Phyllomedusidae as the CV values for the advertisement calls vary continuously (Röhr et al. 2020b).Our results indicate that the CVs of the spectral traits showed considerably lower values than those found for the structural and temporal traits, indicating that the spectral traits are the least labile.These spectral variables, such as dominant frequency, have been found to be static for several anuran species, including those in Phyllomedusidae (Köhler et al. 2017;Röhr et al. 2020b).This may be due to stabilizing selection considering the role of these variables in species recognition (Wilczynski & Ryan 1999;Gerhardt & Huber 2002).
Among structural and temporal variables, the number of pulses, pulse period, note rate, and note duration also showed comparatively low CVs, suggesting that some of these traits may also be static.In fact, pulse rate seems to be species-specific in the family (Röhr et al. 2020a).It has been demonstrated as an important variable to discriminate advertisement calls of sympatric species (Loftus-Hills & Littlejohn, 1971).The variables related to the number of notes and call duration (i.e.call duration, duration of 50% of the call, and note number) had the highest CV values and are therefore the most labile variables in our dataset.These variables might allow males to adjust their acoustic signals accordingly to the demands imposed by the social context (e.g.Bastos et al. 2011;Dias et al. 2017).Mendoza-Roldan (2017) pointed to call duration and note number as the main differences between calls over the calling activity.Here, we found that both variables are highly correlated, indicating that males of P. venusta increase the advertisement call duration mainly by emitting additional notes instead of increasing note or interval durations.Therefore, call duration and note number should be acting as a single variable set in the communication system of this species.Considering the high CV values, these traits might be under directional selection by female preference (intersexual) or by giving advantages during male-male bioacoustical contests (intrasexual selection; Gerhardt and Huber 2002).
We analyzed a single female distress call which prevented us from evaluating intra-and inter-individual variation.This type of call is known to be very conservative in its features (Toledo et al. 2009;Forti et al. 2018).Therefore, the features described here should be representative of the distress call of the species, and the differences found between distress and advertisement calls should be reliable (Figure S2).The P. venusta distress call differs from the advertisment call in having an exceptionally long duration, long harmonic notes, and a high-pitched timbre.This timbre is expected for distress calls, and it is usually a result of a combination of spectral features such as a wide frequency range, a significant amount of energy in the higher harmonics, and a high dominant frequency (Gridi-Papp 2008;Forti et al. 2018).Indeed, the distress call of P. venusta has all these features compared to its advertisement call.
The distress call may function as a defensive signal to avoid predation by directly frightening the predator or attracting other animals that could scare or attack the predator (Toledo et al. 2014).Many taxa prey upon anuran species and are capable of hearing (Toledo et al. 2007).Still, the frequency hearing range can be significantly different depending on the predator in question (Bradbury & Vehrencamp 2011).Therefore, the wide frequency range of the distress call should be advantageous by increasing the chance of overlapping the frequency range of the signal with that of the predator's senses.As well as other anuran distress calls, the evolution of the P. venusta distress call is likely being driven to benefit communication with a wide range of taxonomic groups (Hödl & Gollmann 1986;Forti et al. 2018).
Despite the absence of a vocal sac (Duellman 1970), we observed that the gular region is filled with air during male vocal activity.Although we could not observe the female while emitting the distress call (see Material and Methods), we presume she vocalized with an open mouth.In most species that have distress calls reported, specimens were observed vocalizing with open mouth (Toledo et al. 2009(Toledo et al. , 2014)).By emitting the call in this way, the distress call is not filtered by the vocal sacs or gular tissues as would occur with the advertisement calls.This results in a wide frequency range typical of distress calls (Gridi-Papp 2008).This seems to be the most crucial mechanical difference that causes the remarkable timbre difference between distress and advertisement calls (Gridi-Papp 2008), as observed in P. venusta.
Apart from the present study, P. bicolor and P. trinitatis are the only species of the genus that had their distress calls described (Kenny 1966;Zimmerman & Bogart 1984).Although distress calls have been recorded from males, females, and juveniles for many anuran taxa (Hödl & Gollmann 1986;Toledo et al. 2009), this is the first description of a distress call emitted by a female of the genus Phyllomedusa.During fieldwork, one of us (JSMR) observed that females remained in the nest in a lethargic state after oviposition, which can last hours even during dawn (4:30 h).This lethargic condition could make P. venusta females significantly more vulnerable to predation and more prone to emit distress calls, a hypothesis that should be investigated in further behavioral studies.

Interspecific comparison
The phyllomedusid advertisement calls remarkably vary in call duration, note repetition rate, dominant frequency, and period of pulses (Röhr et al. 2020b;Bezerra et al. 2021).The so-called 'chack' call (see Duellman 2001; Mendonza-Roldan 2017) usually has notes with long pulses and short inter-pulse intervals (= high pulse repetition rate) and is the call type of several Phyllomedusa species (e.g.P. bicolor, P. boliviana, and P. tarsius; see Zimmerman & Bogart 1984;Köhler 2000).On the other hand, some phyllomedusid species have calls with short pulses clearly separated by long intervals and usually organized into groups (e.g.P. distincta, P. iheringii, and P. tetraploidea; see Haddad et al. 1994;Forti et al. 2019).These latter calls sound like a rattle to human ears.The advertisement call of P. venusta corresponds to the former type ('chack') with long pulses and a high pulse repetition rate.Other Phyllomedusa species with the same call type include most of the species of the P. tarsius group with described advertisement calls (Zimmerman & Bogart 1984;Duellman 2005;Barrio-Amorós 2006).This species group currently comprises six species (Castroviejo-Fisher et al. 2017): P. camba, P. chaparroi, P. neildi, P. tarsius, P. trinitatis, and P. venusta.Among these species, P. camba, P. tarsius, and P. venusta have 'chack' calls (Zimmerman & Bogart 1984;Duellman 2005;Bezerra et al. 2021; present study).The advertisement call of P. chaparroi is unknown in literature so far, and that of P. neildi could not be classified based on its original description (Barrio-Amorós 2006).We also could not access recordings P. neildi to examine its call directly (Table S1).
Some authors have justifiably questioned the taxonomic status of P. venusta from P. trinitatis due to the striking morphological similarity between these species (Barrio-Amorós 2009;Castroviejo-Fisher et al. 2017).We obtained a single recording of P. trinitatis from Caracas, Venezuela (FNJV 0031163).Based on this file, the advertisement call of P. trinitatis is the only one among the species of the P. tarsius group that we examined that sounds like a rattle, i.e. with short pulses, long interpulse intervals, and pulses grouped near the end of the call.These features allow us to distinguish P. trinitatis from every species we have examined in this group (i.e.except P. chaparroi and P. neildi), including P. venusta.
Besides the structural and temporal differences between the advertisement calls of P. venusta and P. trinitatis, they also have vastly different values for the spectral variable dominant frequency (DF).Although Barrio-Amorós (2006) suspected a value of about 800 Hz of DF (based on a spectrogram in Rivero & Esteves 1969).Our measured DF value of P. trinitatis varies from ca. 1700 to 1900 Hz (corresponding to the second harmonic), and it is more than twice that of P. venusta, which ranges from ca. 500 to 1000 Hz (corresponding to the first or the second harmonic respectively; see call descriptions above and Table S1).This spectral difference is still more remarkable considering that both species have similar body sizes (SVLs of 76.3 and 77.4 mm for P. trinitatis and P. venusta, respectively; Table S1).
The tight correlation between SVL and DF is wellknown in Anura (see Gerhardt & Huber 2002;Wells 2007;Köhler et al. 2017).Our DF-SVL analyses indicate that P. trinitatis has an unexpectedly higher DF (on the second harmonic) for its size relative to the other phylomedusids.This species has a large Phyllomedusa-like body size but a high Pithecopus-like dominant frequency, which seems to represent an allometric scape within the Phyllomedusidae family.This unexpected DF may be linked to anatomical modifications of the P. trinitatis' vocal apparatus.Dominant frequency values higher than predicted might result from allometric shifts between the vocal apparatus and body size, where the former is smaller than predicted by the latter (Tonini et al. 2020).Those shifts may occur in structures such as the larynx and vocal membranes (see De La Vega et al. 2021), or even in structures responsible for post-source filtering, such as the vocal sac (Gridi-Papp 2008).Analysis of new recordings of P. trinitatis and further comparative studies on this species' vocal apparatus and vocal sac must be conducted to investigate those questions.
Phyllomedusa neildi and P. tarsius are also very similar to P. venusta morphologically (Castroviejo-Fisher et al. 2017).The DF of P. venusta is similar to those of the P. camba, P. neildi, and P. tarsius, all species of the P. tarsius group (Castroviejo-Fisher et al. 2017).The FD -SVL of the advertisement call of P. venusta is particularly similar to those of P. tarsius and P. camba (Figure 6).This bioacoustical similarity reinforces the placement of P. venusta in the P. tarsius species group, as previously proposed based on morphological similarities (Barrio-Amorós 2006; Castroviejo-Fisher et al. 2017).Nonetheless, it is important to point out that a phylogenetic analysis is still needed to adequately address the relationship of P. venusta within the genus as well as a taxonomic revision of the P. tarsius species group (Faivovich et al. 2010;Castroviejo-Fisher et al. 2017;Barrio-Amorós et al. 2019).We encourage that bioacoustical characters, such as DF, are included in these studies.

Figure 1 .
Figure 1.Recording site at a temporary pool in Bocas del Carare, Santander, Colombia.(a).Amplexus occurring at the breeding site (b).Amplectant couple reaching the canopy for nesting (c).Vocalizing male emitting the multi-noted call after playback (d).Interacting males during the chorus (e).The recorded female that emitted the distress call (e).

Figure 2 .
Figure 2. Distribution of 128 male calls of Phyllomedusa venusta along the first and second components of a Principal Component Analysis (PCA) based on eight bioacoustical variables.Colors correspond to the isolated call and staccato call sensu Mendoza-Roldan (2017).Ellipses are the 95% confidence interval of them.Note that calls were found in a single cluster.See text for variables.
Agalychnis, Callimedusa, and Phasmahyla species have intermediate values.The linear regression between SVL

Figure 3 .
Figure 3. Coefficients of variance (CV) for the bioacoustical variables measured from male calls emitted during the chorus (interpreted as being the advertisement calls; see text) of Phyllomedusa venusta.Dashed red lines limit different types of variables, with the spectral variables at the top, followed by the temporal variables in the middle, and the structural variables at the bottom.

Figure 4 .
Figure 4. Variation of call duration and note number of 128 calls emitted by males during the chorus of Phyllomedusa venusta.Spearman's rank correlation between male call duration and note number (log-transformed values).Note that call duration and number of notes are linearly correlated (R = 0.84, p ≤ 0.001), and both variables range gradually (i.e.without a gap separating short and single-noted calls from long and multi-noted calls).Histograms for the variables Note Number (green) and Call Duration (Orange).Note the skewed distribution; most of the calls are short and composed of a single note.

Figure 5 .
Figure 5. Spectrogram and oscillogram of one advertisement call with eight pulsed notes (A) and one distress call with 14 nonpulsed notes (B) of Phyllomedusa venusta.Note that the first two harmonics are more energetic and the typical call envelope with two amplitude peaks, one in the first note and the other in the end of the call (A).Note the periodic frequency modulation within the notes and the deterministic chaos mainly among the higher harmonics, making the notes slightly noisy (B).

Figure 6 .
Figure 6.Comparative analyses among 31 phyllomedusid species, including Phyllomedusa venusta.Reconstruction of ancestral states of male body size, accessed through snout-vent-length (SVL; A) and dominant frequency (DF) of the advertisement call (B).Phylomorphospace between these variables (C) connected according to the Duellman et al.'s (2016) phylogeny (see text and Figure S1).Linear regression trade line with all species labeled (D); gray shade indicated the confidence interval of 95%.Species of the P. tarsius group are indicated in (C).Note the inverse relationship between SVL and DF (A-D), the existence of a phylogenetic structuring pattern with Phyllomedusa species as having larger body sizes and lower call dominant frequencies (A-D), and the unexpected dominant frequency of P. trinitatis for the P. tarsius species group.