Adaptation is a central pillar of speciation as it is a direct outcome of natural selection. Traits that increase the fitness of individuals by enhancing growth, reproductive output, or survival capability are considered adaptive and are often under directional selection. Rates of evolutionary change in these traits may be modulated by other micro-evolutionary forces such as genetic drift and gene flow, which may constrain adaptive divergence if they are antagonistic to the direction of selection (Garant et al. 2007, Wright 1951). However, adaptive divergence of populations may occur despite significant influences of genetic drift (Funk et al. 2016) and gene flow (Fitzpatrick et al. 2015). Such strong selective forces may cause demographic differences among populations to arise and, with sufficient time, even reproductive isolation and population diversification (Carnicer et al. 2012). The evolution of traits involved in mating systems has been a subject of study since Darwin noted the striking ornamentation and mating displays that are evident in nature and that may be under sexual selection (Darwin 1871). Since then, myriad examples of traits under sexual selection have been documented in nature (see Ryan and Keddy-Hector 1992 and references therein), and many empirical examples of sexually selected attributes contributing to reproductive isolation have prompted the wide acceptance of sexual selection as having a major role in speciation (Andersson 1994, Coyne and Orr 2004, Panhuis et al. 2001). In many mating systems, the choosy sex (often the female) imposes selection on traits implicated in mating displays that are expressed in the opposite sex, resulting in it becoming more costly and elaborate (Ryan and KeddyHector 1992, Andersson 1994). This relationship between quality/intensity of the mating display trait and female choice has been hypothesized to be a consequence of natural selection on mate preference that produces a tangible benefit, such as nuptial gifts (Kirkpatrick and Ryan 1991), or may lead to offspring of better genetic quality (Zahavi 1975). Such a relationship may also be a manifestation of runaway selection, where female preference and male trait (for example) are genetically linked (Fisher 1958). Selection on the male trait imposes indirect selection on female preference for that trait, and the two co-evolve in this fashion (Fisher 1958). Last, sensory exploitation has also been proposed as an explanation for the relationship between preferences for male traits (Ryan and Rand 1990, Ryan et al. 1990). This hypothesis requires that females having a sensory preference for some facet of a mating display, where sexual selection favours male traits that exploit this preference (Ryan and Rand 1990). Under this hypothesis, male traits can only evolve within the bounds of the existing female sensory system (Ryan and Rand 1990). These alternative hypotheses are not mutually exclusive and patterns of trait evolution that are consistent with the predictions of one hypothesis may also be explicable, at least in part, by other causal factors (Ryan and Keddy-Hector 1992).

Functional Traits

In addition to traits that directly influence reproductive success and survivability, those that indirectly enhance reproductive success or chance of survival may also be under selection (Arnold 1983, Violle et al. 2007). These traits are considered functional traits and were first recognized through Arnold’s (1983) “performance paradigm”. According to this framework, traits that directly relate to fitness are considered performance traits (Arnold 1983). Traits that modulate the effectiveness of performance traits (therefore only indirectly influencing fitness) are considered functional traits (Arnold 1983). For example, colouration in some species of moths serves as camouflage and thus directly influences fitness by enabling individuals to evade predators (Endler 1984, Poulton 1980). After landing on tree bark, moths will alter their position to enhance crypticity (Kang et al. 2012). In this example, the moth’s colouration is the performance trait, and the behaviour that enhances the performance trait is the functional trait. Functional traits involved in mating displays may be under indirect sexual selection as they influence traits involved in reproduction. For example, male gliding lizards (Draco sumatranus) exhibit body-modifying behaviour in response to sun position to enhance the visibility of their colourful dewlap (Klomp et al. 2017). This behaviour increases the conspicuousness of the signal by enhancing its luminance, which may lead to greater conspecific recognition and potential mate detectability (Klomp et al. 2017). Similarly, male tree-hole frogs (Metaphrynella sundana) actively alter the pitch of their call according to the amount of water in the tree cavity they call from to enhance the attractiveness of their mating signal (Lardner and bin Lakim 2002). These studies suggest that successful mating may partially depend on traits associated with the mate recognition system that are ancillary to the main display. It is therefore important to consider the evolution of potential functional traits as well as performance traits when investigating adaptive trait evolution.


Many selected traits involved in reproduction may be subject to trade-offs influenced by other aspects of a species’ ecology (Ryan 1986, 1988, Simmons and Emlen 2006, Evans 2010, Engqvist 2011). Such factors may not act exclusively and many disparate factors could shape variation in the same trait. Gamete quantity and size is an iconic example of a trade-off that was first modelled by Smith and Fretwell (1974). Smith and Fretwell (1974) determined that the optimal size and number off eggs were determined by offspring fitness as it related to initial investment. Since this seminal work, studies finding evidence for antagonistic selective pressures on traits causing trade-offs have increased (e.g. Ghalambor et al. 2004, Scales and Butler 2015, Blanchard and Moreau 2016). Trade-offs associated with mating displays have also been documented, many of them leading to the notion that traits under sexual selection may be considered honest indicators of fitness. A key example of this is carotenoid-based colouration in birds where the trade-off is hypothesized to be the cost of maintaining bill colouration and immune system functioning. Carotenoids are typically obtained and sequestered from an individual’s diet and it was originally proposed that variation in diet accounted for variation in carotenoid colouration (Endler 1980, Hill 1999). However, carotenoids also serve as an immunostimulant in birds (Chew 1993) and have been found to slow the replication rates of parasites (Baeta et al. 2008). There is thus a trade-off between allocating carotenoids to colour to attract a mate versus allocating them towards combating parasitism; indeed it has been suggested that only males of high quality can both allocate carotenoids to plumage colour and sustain the heavy parasite load associated with it (Blount et al. 2003, Baeta et al. 2008).

Trade-offs in Anuran Calling

As mentioned above, some traits may be subject to multiple trade-offs acting on them at once. Anuran calling is subject to several antagonistic forces of selection leading to trade-offs. Many anuran species form large choruses where males call to attract females and mate choice is primarily determined by male call quality (reviewed in Gerhardt and Huber 2002). Producing and maintaining calls is exceptionally energetically costly for males, with some estimates for gray treefrogs (Hyla versicolor) reaching around 25 times their resting metabolic rate (Taigen and Wells 1985). Therefore, call evolution may be limited by metabolic rate and individuals may experience trade-offs between call performance and maintenance of calling (Reichert and Gerhardt 2012). Like many traits that increase conspicuousness, male calls may also increase the risk of predation or parasitism (Tuttle and Ryan 1981, Bernal et al. 2006, Page and Ryan 2008). Studies on the calls of túngara frogs (Engystomops pustulosus) have shown that predators, such as bats, and blood-sucking flies actually preferentially predate/parasitize calling individuals (Tuttle and Ryan 1981, Bernal et al. 2006, Page and Ryan 2008). The bat predators have even been shown to use the advertisement calls of the males to locate them within a chorus (Tuttle and Ryan 1981, Page and Ryan 2008). Last, signal efficacy relies on receiver’s ability to discriminate; therefore, receiver morphology/ neurophysiology may constrain the evolution of particular call traits (Akre et al. 2011, Schrode et al. 2014). These studies collectively demonstrate the need to consider multiple facets of trait evolution and the forces that may affect it (Gerhardt 1994).

Study Species

The spring peeper (Pseudacris crucifer) is a small tree frog species (Family: Hylidae) that is broadly distributed across eastern North America from the Gulf Coast north to the James Bay lowlands, and from the Atlantic coast inland to west of the Mississippi River (see Figure 1). During the breeding season, males congregate in wetlands, small vernal pools, swamps, marshes, and lake margins to begin calling to attract females (Conant and Collins 1998, Zimmitti 1999). Because of varying seasonality across latitudes, males begin chorusing in late December and early January in the southern parts of the species’ range (e.g. Florida through west Texas), to April and May in the northern extent. Calls are comprised of a single tone “peep” (Taigen et al. 1985) that are on average 87 decibels in amplitude, based on measurements in four populations from New York (Brenowitz et al. 1984). Although female attendance at spring peeper choruses is generally poorly quantified, males remain calling throughout the night and may return to call for several nights (Rosen and Lemon 1974). Male spring peepers are territorial and protect their calling site from competing males using both aggressive vocalizations and physical contact (Rosen and Lemon, 1974). Spring peepers, like many anuran species, may adopt satellite behaviour as an alternative mating tactic where they do not call, but position themselves close to calling males to intercept females (Lance and Wells 1993, Stewart et al. 2017). Figure 1. A: Picture of a calling male spring peeper. Photo taken by N. Cairns. B: Range map of the spring peeper showing the six mitochondrial lineages' range limits. Map taken from Stewart and Lougheed (2013). A B Eastern Interior West Southwest Southeast Texas Spring peeper males can vary their calling behaviour to maximize their calling performance. For example, auditory cues including conspecific advertisement calls, but also aggressive trill calls, and traffic noise, can cause males to alter aspects of the mating call (e.g. call duration), or to respond with an aggressive trill call (Rosen and Lemon 1974, Schwartz 1989, Marshall et al. 2003, Humfeld et al. 2009, Zimmitti 2009). The presence of an “intruder” on a male’s territory, tested using mating call playback, has been shown to cause individuals to respond with aggressive trill calls (Rosen and Lemon 1974, Schwartz 1989, Marshall et al. 2003, Humfeld et al. 2009). Playback calls with longer trill durations elicited the most aggressive calling behaviour (Schwart 1989). Aggressive calls themselves may be plastic, their frequency and structure determined by chorus density and inter-male spacing (Marshall et al. 2003), but whether a male switches from advertisement calling to aggressive calling depends on the social context (Humfeld et al. 2009). Humfield et al. (2009) performed a series of studies to test what intruder male behaviours would elicit aggressive calls. They found that males were more likely to respond aggressively if the intruder male began with advertisement calls, then switched to aggressive calls, rather than the calling aggressively right after the intrusion (Humfeld et al. 2009). Hanna et al. (2014) tested whether traffic noise would cause male spring peepers to modify their calling behaviour and found that background masking noise cause spring peepers to produce calls with shorter duration. They also found that, if the traffic noise masked the high frequencies of the advertisement call, males would produce lower frequency calls. This relationship has been reported for other species of frogs and may indicate a response to urbanization and anthropogenic noise (Cunnington and Fahrig 2010). These studies demonstrate that, although variation in anuran advertisement calls has a genetic basis (Welch et al. 2014), environmental and social cues can elicit plastic vocal responses and associated behaviours. The spring peeper shows marked genetic structure across its geographic range with six mitochondrial lineages (Figure 1) of varying ages (Austin et al. 2002, 2004, Stewart and Lougheed 2013). These mitochondrial lineages are thought to have diverged in geographic isolation beginning in the Pliocene, and while their distributions will have waxed and waned over the ensuing years, they are now in secondary contact in many areas across the species range (Austin et al. 2004, Stewart and Lougheed 2013). Recent work investigating the contact zone dynamics between the two most closely related mitochondrial lineages found evidence of call differences and reproductive character displacement, potentially reinforced by female choice for their natal lineage male advertisement calls (Stewart et al. 2015). There is also evidence of reduced hybrid survival at the larval stage in this contact zone (Stewart and Lougheed 2013), implying a cost to mating with non-natal individuals and an important potential role for calls in identifying appropriate mates. More broadly, advertisement call differences among all mitochondrial lineages, while subtle, are statistically significant (Cicchino et al. unpubl.), again suggesting that advertisement call might be implicated in population divergence. The male calling behaviour of this species has been well-documented (e.g. Rosen and Lemon 1974, Brenowitz et al. 1984, Taigen et al. 1985, Sullivan and Hinshaw 1990, Zimmitti 1999, Marshall et al. 2003); however, studies and field guides provide divergent accounts of preferred or typical calling sites. Some articles and field guides suggest that the species is exclusively terrestrial (e.g. IUCN 2014, Rosen and Lemon 1974, Conant and Collins 1998), calling exclusively from the ground or from floating vegetation. Others report that male spring peepers call both terrestrially and from perches above or near water bodies (e.g. Wygoda 1984, Brenowitz et al. 1984, Dodd 2013). Our combined range-wide observations suggest that this variability in the literature is a product of geographic variation in calling behaviour, with males in southern populations often calling from bushes or tree branches, while their northern counterparts call almost exclusively from the ground (Cicchino, Cairns and Lougheed pers. obs.). Previous empirical studies on spring peeper calls have demonstrated the benefits of calling from elevated perches as it increases call transmission distance thereby presumably increasing the changes of attracting a female (Brenowitz et al. 1984; Parris, 2002). As arboreal calling would appear to provide obvious benefit to males, it is reasonable to ask the question as to why all males do not undertake it.

Objectives and Hypotheses

In this study, I first quantify perching behaviour across the spring peeper’s range and test for differences in prevalence and proportion of arboreal calling behaviour among populations. Finding marked population-level differences, I then investigate factors that may cause or limit this potentially adaptive behaviour. Although past studies have indicated a sound transmission advantage to calling from elevated perches in this species (Brenowitz et al. 1984, Parris 2002), this benefit may not be ubiquitous across its range. For example, perching behaviour may be a way to circumvent increased call degradation in habitats with dense leaf litter, but in habitats with low leaf litter density or sparse vegetation, sound propagation from elevated perches may not differ from ground-level propagation. I thus hypothesize that perching behaviour arises because of local selection for enhanced sound transmission. I predict that, in habitats with greater leaf litter and understory density, perching behaviour will be more prevalent; but in habitats with less leaf litter and sparse understory, this behaviour will not be common. To test this, I performed playback experiments at peak breeding times for local populations across the spring peeper’s range with source representative calls projected at different heights to quantify sound degradation in each habitat. Environmental conditions, such as temperature and humidity, influence many aspects of anuran ecology, including physiological processes, locomotor performance, and behaviour (Rome et al. 1992, Shoemaker et al. 1992). Temperature is also implicated in anuran reproductive success as it is positively correlated with many temporal characteristics of advertisement calls, which females often use to discriminate between males (Gerhardt and Huber 2002). Maximizing call transmission efficacy is critical for individual males realizing mating opportunities, and fine-scale temperature differences between arboreal and terrestrial micro-habitats can drive calling site selection (Höbel and Barta 2014). However, arboreal behaviour may expose individuals to higher wind levels as well as decrease contact with moist vegetation or water, therefore increasing the risk of desiccation. I thus also hypothesize that there can be a physiological cost to arboreal calling in some environments, which may be mitigated by higher air temperatures and humidity in some parts of the range or at different parts of the breeding season, or some combination thereof. I thus expect to find increased prevalence in perching behaviour in populations that typically breed at higher air temperatures and humidity. To address desiccation risk, I performed a study measuring the effect of perch height on evaporative water loss and body temperatures in different environmental conditions using model frogs. In addition to testing for spatial variation in arboreality, I investigate morphological variation that may also be related to arboreal behaviour. Over time, morphological adaptations to enhance arboreal behaviour may have been selected for, which would result in morphological differences among populations. Certain traits may be better suited to arboreal habitats, such as longer forelimbs typical of hylids, although these have been asserted to be for increased dexterity in manipulating prey (Gray et al. 1997). Alternatively, morphological differences may have arisen and thereafter allowed populations to exploit an arboreal niche, which also would result in morphological differences between populations. As the evolution of ecological niche and morphology is often linked (Scharf et al. 2000, Ortiz-Medrano et al. 2016), I hypothesize that there will be observable phenotypic differences among populations that have diverged in their arboreal behaviour.

  • Cicchino, Amanda S.


Understanding the trade-offs that affect potentially adaptive traits is fundamental to our understanding of evolutionary diversification and speciation. Heterogeneous landscapes may lead to spatial variation in such traits among populations as they may experience different selective pressures. In this study, I investigate the spatial variation associated with arboreal calling behaviour in the spring peeper, Pseudacris crucifer. Finding marked variation in the median height of perching and the proportion of males perching within a population, I investigated the factors that might mediate this behaviour. Spring peeper males at the northern extent of their range almost exclusively call from ground-level, whereas individuals in populations towards the central and southern part of the species’ range are largely arboreal callers. Results of a playback transmission study using male advertisement calls at different perch heights suggest that there is a benefit to arboreal calling: less call degradation across longer distances. To understand some potential physiological costs associated with calling from elevated perches, where individuals would experience higher wind and evaporative water loss, I performed a desiccation experiment using plaster model frogs. Model frogs at high perches lose water at a markedly faster rate than those on the ground, suggesting an increased risk of desiccation for arboreal individuals. Arboreal calling behaviour may therefore be partially governed by a trade-off between reproductive attractiveness and physiological limitations. In addition to spatial variation in arboreality, I investigated morphological variation associated with this behaviour. I found longer limb length and larger body size in arboreal individuals, suggesting possible local morphological adaptation for populations with increased propensities for calling from elevated perches. Whether arboreal calling behaviour is a behaviourally-plastic, context-dependent trait or is fixed within populations remains to be studied, although the trade-offs mediating its prevalence in a population offers a unique opportunity to evaluate the genetic underpinnings of a potentially adaptive trait that is found in only some populations of a species with a dynamic evolutionary history.


Field Sampling

I sampled 357 spring peepers from 43 sites across the species range from 2016- 2017 (see Figure 2 for sampling map). Sites were chosen based on mitochondrial lineages representation, attempting to maximize representation across all lineages. Sampling in the southwestern part of the range allowed me to collect data on three mitochondrial lineages within less than 800km. Upon locating a calling male, I measured its height off the ground and recorded surface body temperature using a laser thermometer. For potential use in playback studies (see below), I recorded each male from approximately one metre away from its snout using a Marantz (Marantz America, NJ) PMD-660 digital recorder and Sennheiser (Point Claire, PQ, Canada) ME67 directional microphone. I also measured the amplitude of the call from about 30cm from its snout using a Galaxy Audio (Wichita, KS, USA) CM-130 SPL-meter. I used 30cm as the distance for SPL as distances further away were less reliable as background noise would be reflected in the decibel reading. For each site, I measured air temperature (°C), humidity (%RH), average wind speed at the site (km/h), and barometric pressure (hPa) using a Kestrel (Kestrel Pocket Weather Meters, Boothwyn, PA, USA). Each individual was hand-captured after recording for subsequent processing. I measured six morphological variables: snout-urostyle length (SUL), femur length, tibia length, tarsus length, radio-ulnar length, and foot length (from joint to tip of the longest toe). Each measurement was taken twice to the nearest hundredth of a centimetre using digital calipers and averaged for all analyses. Figure 2. Sampling distribution map of Pseudacris crucifer populations. The size of the circle corresponds to the number of individuals caught and assessed in each population. 


I performed playbacks at 15 sites across the P. crucifer range to test whether perching behaviour enhances sound propagation or decreases call degradation. I created a playback loop consisting of five single calls (i.e. “peeps”) from 18 males. I included males from across the entire range, being sure to represent three males from each of the six mitochondrial lineages described by Austin et al. (2002, 2004) and Stewart and Lougheed (2013). The males selected from these mitochondrial lineage had call attributes closest to the lineage mean for six call variables (call duration, call rate, call interval, rise time, fall time, peak frequency) based on 64 total calls from across the spring peeper’s range. The calls were standardized to an amplitude of 87 dB using Audacity (Version 2.1.1; and were broadcasted at 87dB. The playback was broadcasted using a Braven (Irvine, CA, USA) BRV-1 speaker at ground level and 1.2m above ground, and re-recorded using a Marantz (Marantz America, NJ) PMD-660 with a Sennheiser (Point Claire, PQ, Canada) ME67 directional microphone at two metre and ten metre distances to investigate possible differences between short-range propagation and long-range propagation. 1.2m was chosen as the height treatment as, it was the highest I had observed an individual perching when I was deciding on experimental design. The playbacks were always recorded at ground-level to reflect that I have only observed females on the ground and thus this is the most relevant vantage point. I quantified aspects of the environment that might be relevant to sound transmission at each site using an array of approaches. First, I recorded the maximum amplitude (dB) of background noise at the time of playback using a Galaxy Audio (Wichita, KS, USA) CM-130 SPL-meter. To determine vegetation density at each location, an opaque white sheet with a black 10cm2 grid on it was held at two metres from the point of broadcast behind emergent vegetation and photographed from eye level (approximately 1.75m off the ground and parallel to the speaker). The photos were digitally cropped to only include the middle 100 squares of the sheet (cropping out squares to the left and right because wind would often cause the ends to bend) and were analyzed in random order. Based on these photographs, I subjectively binned each square on the sheet into 0%, 5%, 15%, 25%, 50%, 75%, and 100% cover categories. The values were then averaged for all 100 squares to estimate vegetative density for each site. I estimated the percent of ground cover by leaf litter to the nearest 10% (versus water) in a two metre by two metre quadrat centered on the point of playback. Canopy cover (%) at the point of broadcast was estimated using a Convex Spherical Densiometer. I also noted whether the playbacks were done primarily over water or land as this may have implications for sound transmission.

Playback Analysis

Call degradation was quantified by using cross-correlation tests in RavenPro 1.4 ( as described in Malone et al. (2014), such that higher cross-correlation coefficients signify lower degradation. Correlations were performed for each “peep” call, and averaged for each male.

Evaporative Water Loss Study


Using plaster models (see below), I tested for differences in evaporative water loss and surface temperatures at different perching heights. I made models approximately the same mass as a sexually mature male spring peeper (1.5g based on data from 489 frogs; Cicchino and Lougheed, unpubl. data) and surface area (8.3cm2 ) using the formula described in McClanahan and Baldwin (1969). I used Smooth-On (Macungie, PA) OOMOO-30 to create a mold of the model shape that fit these parameters (length: 2cm, width: 1.03cm, height: 0.7cm). Past studies have shown that Plaster of Paris is suitable for creating models of wet-skinned amphibians as these models experience similar rates of evaporative water loss and surface temperature changes to live animals (Peterman et al. 2013; Tracy et al. 2007). Following methods outlined in Peterman et al. (2013), I mixed four parts Plaster of Paris with three parts of water, and poured the solution into the mold. The plaster dried for at least one hour before I removed it from the mold. I did not paint the plaster as colour would not affect light absorption at night when I conducted trials, but paint might alter the evaporative properties of the model (Peterman et al. 2013). I cured the plaster models in an incubator at 65° C for at least four hours, then labelled them with individual identifiers using a permanent marker. I weighed each model twice to the nearest 0.01g using a portable electronic scale (Durascale, MyWeigh, Phoenix, AZ) and averaged the measurements to determine the dry mass. Before deployment in the field, I soaked the models water for at least four hours and weighed them twice again to determine the average starting wet mass. Trials were completed at the same site (Curtis Marsh) at the Queen’s University Biological Station (Chaffey’s Lock, ON, Canada; 44.571121, -76. 330742) over two nights with similar environmental conditions. This site is a palustrine marsh dominated by Typha and Salix sp. and known to have recurring P. crucifer breeding assemblages (Klaus 2012), as well as other anuran species including Hyla versicolor, Lithobates catesbeianus, and Lithobates clamitans (Klaus 2012). To understand evaporative water loss and temperature differences between perching and non-perching individuals, I initially included three treatments for our models: 1.2m above the ground, 5cm above the ground, and on the ground with 30% of the model in water, mimicking a calling male spring peeper. However, the treatment with the models placed partially in water were removed from subsequent analysis as they disintegrated over the course of the night as a result of oversaturation, therefore skewing the results of their net loss of mass. I set up twenty-five 2m tall stakes along the periphery of the water’s edge where the density and stature of vegetation was approximately consistent among stakes. I used small finishing nails driven into the stakes parallel to the ground to serve as ‘stands’ for the models. I deployed the models at 8pm local time (GMT-4) and collected them at midnight, similar to the timing of chorusing activity observed at the field station. Throughout the night, air temperature at the different treatment heights and water temperature was logged every minute using an Arduino and four 10K Precision Epoxy Thermistor (Elmwood Electronics, Toronto, ON, Canada). At the end of the trial, models were immediately weighed to determine their ending wet mass, and their surface temperature was taken before removal from the perch using a laser thermometer Raytek Minitemp MT6, Santa Cruz, CA, USA). Net water loss was calculated following Peterman et al. (2013): Net water loss = (starting wet mass – end wet mass) / (starting wet mass - dry mass)


All analyses were performed in R (R Core Team 2016). All plots were created using the packages ggplot2 (Wickham 2009) and ggmap (Kahle and Wickham 2013) in R. To test whether site conditions and arboreality are related, I performed a multiple regression analysis on perch height as a response variable and air temperature, humidity, longitude, latitude, and possible interactions among these variables as predictor variables. I also performed multiple regressions with the percent of perching individuals in a population as the response variable and the same predictors as the first model. Top model selection was based on the Akaike Information Criterion and an Information Theoretic approach (Burnham et al. 2011) to determine the best predictors of arboreality. R packages used for these analyses and subsequent ones include arm (Gelman and Su 2016), MuMIn (Barton 2016), gvlma (Pena and Slate 2014), car (Fox and Weisberg 2011), and dplyr (Wickam et al. 2017). To visualize potential trends in morphology with regards to arboreality, I first performed a correlation-based Principal Components Analysis on all morphological variables using the vegan (Oksanen et al. 2017) package in R. I then performed a series of logistic regressions, performing Bonferroni corrections to minimize false discovery, using morphological variables as the independent variables and whether the individual was perching (defined as a perch height > 1cm) or terrestrial as the dependent variable. To understand the effects of arboreality on the transmission of calls, I tested for differences in transmission between the two treatments (ground-level broadcast and 1.2m broadcast), as well as for differences in possible environmental variables that affected the transmission of the calls. I used a two-tailed t-test on the cross-correlation values from the two broadcast treatments to test for a direct difference in transmission. I then performed a series of multiple regression analyses on both the spectral characters and temporal attributes of calls, with correlation coefficients as the dependent variable and height (cm), vegetation density (estimated %), ground cover estimate (%), latitude, longitude, background noise level (dB), and wind speed (m/s) as predictors. These analyses were carried out for the entire dataset (both transmission distances combined), and for each transmission distance treatment separately. I also performed a logistic regression (binary dependent variable: perched versus not perched) with vegetation density, background noise, and canopy cover as predictors to test whether the correlation coefficients of the two playback heights were different while accounting for environmental differences. To test for differences in percent water loss in the plaster models, I performed a t-test with net water loss as the response variable and position (5cm, 120cm) as the predictor variable. A t-test was also used to determine whether there were differences in surface temperature at the end of the night between the plaster models in the two treatments. To determine whether position of the model would be a significant predictor of water loss, I performed a multiple regression with water loss as the response variables and position, air temperature, model surface temperature, and humidity as predictor variables.