As habitats are increasingly urbanized, developed, or modified by humans, animals are faced with novel selection pressures such as anthropogenic light and noise, different predators, altered community structures, and highly modified habitats (reviewed by Slabbekoorn and Ripmeester, 2008; Evans et al., 2011; Swaddle et al., 2015). Specifically, anthropogenic noise pollution is an obstacle for animals that communicate acoustically, because this noise is high amplitude and can overlap and mask the frequencies at which animals signal (reviewed by Slabbekoorn and Ripmeester, 2008; Barber et al., 2010; Shannon et al., 2016). For example, reduced signal transmission can affect organisms’ ability to give alarm signals warning others of predators, maintain group cohesion, attract mates, compete with rivals, and defend territories (Swaddle et al., 2015).

Many previous studies have investigated how anthropogenic noise affects communication in various vertebrates (e.g., amphibians, Grenat et al., 2019; cetaceans, Buckstaff, 2004, Melcón et al., 2012, Blair et al., 2016; and birds, Halfwerk and Slabbekoorn, 2009, Nemeth and Brumm, 2009). When exposed to traffic noise, American ground frogs (Odontophrynus americanus) increase the dominant frequency of their calls (Grenat et al., 2019). Similarly, Asian particolored bats (Vespertilio sinensis) significantly decrease call complexity and increase call amplitude in response to simulated traffic noise, which increases signal transmission (Jiang et al., 2019). Bottlenose dolphins (Tursiops truncatus) increase whistle production as sea vessels approach, which potentially increases the number of signals and thus would improve signal detection (Buckstaff, 2004). Blue whales (Balaenoptera musculus) have been shown to decrease calling rates in the presence of mid-frequency active sonar, but they appear to increase calling in relation to ship noise (Melcón et al., 2012). While an association between anthropogenic noise and acoustic communication has been shown in a wide variety of vertebrate taxa, songbirds represent two-thirds of the literature on the subject (reviewed by Shannon et al., 2016) and the majority of these studies address the response of songbirds inhabiting urban environments (Slabbekoorn and Ripmeester, 2008).

Increased anthropogenic noise is one of myriad features of urban and developed habitats that can impact bird song and behavior (Naquib, 2003; Warren et al., 2006). Many aspects of urban environments, such as the physical structure, can act in concert with anthropogenic noise to degrade or mask avian vocal signals and limit signal transmission (Warren et al., 2006). For example, impervious surfaces alter signals through reverberation off buildings and hard ground surfaces (e.g., Slabbekoorn et al., 2007; Phillips et al., 2020), similar to reverberation caused in forests (Richards and Wiley, 1980), which changes how signals propagate through the environment (Wiley, 2006). Concomitantly, loud, low-frequency anthropogenic noise masks any overlapped frequencies of birdsong, specifically at lower song frequencies, and generally lowers the signal-to-noise ratio (SNR) (reviewed by Brumm and Slabbekoorn, 2005; Warren et al., 2006; Slabbekoorn and Ripmeester, 2008). Consequently, birdsong signals may not be detectible by receivers over the anthropogenic noise as song features do not transmit well through the noisy environment (reviewed by Brumm and Slabbekoorn, 2005).

In the last two decades, numerous studies have found a variety of differences in the way birds sing in noisy habitats compared to quieter habitats. Reported strategies that birds employ include raising of minimum song frequency, singing with increased amplitude, and singing with more energy concentrated in a narrower bandwidth – all of which increase signal detectability (reviewed by Brumm and Slabbekoorn, 2005; Slabbekoorn and den Boer-Visser, 2006). Slabbekoorn and Peet (2003) first found evidence of differences in birdsong with increased anthropogenic noise in birds and showed a correlation between noise level and birds singing with higher minimum frequencies in great tits (Parus major). Raising minimum song frequency avoids the masking effects of low-frequency noise, resulting in increased signal transmission. Nemeth and Brumm (2009) found that European blackbirds (Turdus merula) in cities sing with higher minimum frequencies and with shorter intervals between bouts than forest blackbirds. They proposed that these song differences may be an adaptation to urban noise, or alternatively, that this song divergence may be a side-effect of physiological adaptation to urban habitats (Nemeth and Brumm, 2009). When experimentally broadcasting increased noise that overlaps the frequency of song, some species are able to increase the amplitude (loudness) of their song (e.g., Cynx et al., 1998; Brumm and Todt, 2002; Derryberry et al., 2017) or immediately change song features, which is known as “immediate flexibility” (reviewed by Brumm and Zollinger, 2013; Slabbekoorn, 2013). White-crowned sparrows, for example, sing with greater song amplitude, but do not raise their minimum frequencies in response to playback of urban noise (Derryberry et al., 2017). In a follow-up study on this same species, urban males responded to noise playback with songs of more narrow frequency bandwidth, while rural males showed no immediate flexibility in song frequency (Gentry et al., 2017). However, not all bird species tested show population differences in songs nor are all species able to immediately shift their songs in response to a noisy environment (Brumm and Zollinger, 2013; Slabbekoorn, 2013; Swaddle et al., 2015; Shannon et al., 2016). When these mixed results are added to the tendency for negative data to go unpublished, it is not a foregone conclusion that all Passeriform birds have vocally adapted to increase the detectability of vocal signals in noisy environments.

While the impact of urban noise is well documented for bird species with short, stereotyped songs and limited repertoire sizes there is an incomplete understanding of how anthropogenic noise and urban development affect the song features of more complex singers, such as vocal mimics (c.f., Brumm and Todt, 2002; Gough et al., 2014; Deoniziak and Osiejuk, 2019; Walters et al., 2019). For species with short songs (such as species in Passerellidae or Paridae), or those spanning a narrow frequency range, often only a portion of song frequencies are overlapped (Nemeth and Brumm, 2010) as peak amplitudes from anthropogenic noise are usually in the 1–2 kHz range (e.g., Luther and Derryberry, 2012; reviewed by Brumm and Slabbekoorn, 2005; Slabbekoorn and Ripmeester, 2008). For birds living in cities with lower-frequency noise (<1 kHz) that also have songs naturally at higher frequencies, shifting of song frequency upwards may not offer much release from noise-masking because minimal signal content may be masked (Nemeth and Brumm, 2010). One would expect the greater proportion of song frequency overlapped by noise would lead to a greater benefit of shifting song parameters, such as raising minimum frequencies, which would enhance signal transmission and detection under noisy conditions (reviewed by Brumm and Slabbekoorn, 2005; Slabbekoorn and Ripmeester, 2008). Alternatively, we reason that species with a wide vocal range may already vocalize with substantial signal content outside of the range usually overlapped by noise, and thus may not benefit from raising or eliminating lower frequencies. Therefore, it remains an open question whether a species with a very wide vocal range would show differences in song frequency in urban populations.

Here, we study the gray catbird (Dumetella carolinensis), a species found along rural-to-urban gradients that produces long, complex, mimicked song. They are common summer residents of areas in the urban–suburban matrix, although whether suburban habitats lead to high population growth rates seems to be highly variable (e.g., Balogh et al., 2011; Pharr et al., 2023). For catbirds, adult survival increases in urban environments (Evans et al., 2015), daily nest survival tends to increase (Ryder et al., 2010), but fledgling success decreases with increasing prevalence of impervious surface (Balogh et al., 2011). Recently, catbird survival was reported to decrease with increased anthropogenic light, but there was wide variation at the brightest sites (Pharr et al., 2023). While Breeding Bird Survey data suggest that gray catbird populations are currently stable overall (Ziolkowski et al., 2023), some catbird populations, however, are declining and the species is listed in the state of Virginia as a Tier IV Species of Moderate Conservation Need (VDGIF, 2020).

During the breeding season, catbirds regularly sing bouts of song exceeding a duration of several seconds to several minutes without a 1-second gap between song elements (Figure 1). Their songs cover a wide frequency bandwidth from about 1.2 kHz up to 9.5 kHz (as recorded in this study). To our knowledge only three studies have described catbird song in much detail (c.f., Thompson and Jane, 1969, n = 3 males; Fletcher and Smith, 1978, n = 5 males; Kroodsma et al., 1997, n = 7 lab-reared males’ songs analyzed in depth, and roughly 27 wild catbirds). Catbirds’ continuous songs include imitated, improvised, and invented elements (Thompson and Jane, 1969; Kroodsma et al., 1997), and appear to have hundreds of unique elements that are often individually distinct. Thompson and Jane (1969) suggested some song sharing in similar geographical regions while Kroodsma et al. (1997) concluded the opposite, that repertoires are individually distinct, by comparing males at similar and widely different geographic areas.

With such a diverse repertoire and wide frequency range, catbirds may have an increased ability to vary songs in response to noise, given that syllables and elements with frequencies outside of the range masked by anthropogenic noise are common in the repertoire. In our study region, anthropogenic noise can overlap a substantial portion of their lower frequencies and, while noise amplitudes peak between 1–2 kHz, anthropogenic noise often occurs up to 3 kHz or higher. Alternatively, catbirds may not show differences between noisy and quieter habitats since many of their high-frequency syllables and elements would likely transmit over background noise. Studying how urban and modified habitats affect multiple aspects of gray catbird song behavior would provide insights into related species or other species with similarly complex singing styles.

However, to address the topic of urban impacts to song in continuous singers with a wide vocal range poses methodological hurdles and constraints. First for consideration is the question of song duration sample size. While 3–10 songs per individual may appropriately reflect repertoires for short, stereotyped singers, a longer duration of song sampled is likely necessary for mimics and continuous singers (Figure 1C). Accurately measuring frequency can be difficult for song-bouts spanning minutes as even the recordings with the highest SNR have the occasional sound from another bird species or sudden anthropogenic noise in the background. We aim to tackle these hurdles by processing large sample sizes of song per individual and using peak frequencies as calculated per time bin for each song element and deriving minimums and maxima from these peak frequencies. In part, our study was an investigation into whether or not catbird songs vary across an urban gradient, and, if so, what features of urban habitats drive any song differences.

We hypothesized that catbird song features would be impacted by various aspects of urban and modified habitats such as anthropogenic noise and built-up landscape features. We determined the degree of urbanization of sampling sites on a continuous basis by measuring spatially varying aspects such as impervious surface and canopy cover as well as ambient noise levels. Specifically, we predicted that song characteristics such as minimum and peak frequencies will increase with increased levels of urbanization and noise, as increasing these parameters would likely lead to enhanced signal transmission. We predicted that increased impervious surface would predict increases in song minimum frequency but decreases in song maximum frequency, frequency bandwidth (Gentry et al., 2017), and song entropy. Song entropy is generally higher for more complex and widely frequency dispersed (less tonal) songs (Briefer et al., 2010), which we predict higher song entropy in less urban habitats with less noise. We aim to expand previous findings through analyzing a large dataset of over 18,000 elements of catbird songs and using an expanded urban-to-rural gradient from rural Virginia to the Washington, DC metro region including city centers, urban and suburban parks, agricultural land, and National Forest habitats. Examining how various levels of urbanization and noise affect avian communication, especially mating signals, is essential to better understanding how songbird species may persist or thrive in urban habitats.

Our study finds that male gray catbirds in noisier habitats sing at higher minimum, peak, and maximum frequencies, and marks one of the few studies on the topic of noise impacting a complex and continuous singer. Generally, our results support the hypothesis that catbird songs differ structurally along an urban gradient and are in line with our predictions that males in louder, more urban habitats would have higher minimum song frequencies as do many species studied to date (e.g., Halfwerk and Slabbekoorn, 2009; Nemeth and Brumm, 2009; Hu and Cardoso, 2010; Dowling et al., 2011). In contrast to many previous studies, we studied a species with a singing style of long, continuous, complex song bouts with a wide vocal range that also imitates the sounds of other species. Only two other studies, to our knowledge, have focused on anthropogenic noise impacting the song of vocal mimics, the northern mockingbird, Mimus polyglottos (Walters et al., 2019), and the song thrush Turdus philomelos (Deoniziak and Osiejuk, 2019). Walters et al. (2019) similarly found increases in average peak frequency overall and peak frequency of the lowest syllable types under increasing noise levels. Urban song thrushes were shown to sing higher minimum and peak frequencies as well as to sing with a greater syllable repertoire (Deoniziak and Osiejuk, 2019). Previously, few species that do not deliver short, stereotyped songs had been examined for the impacts of noise on song, likely because of the difficulty in song recording and analyses (c.f., Brumm and Todt, 2002; Gough et al., 2014).

Results that catbirds sing higher minimum and peak frequencies are in line with the noise-masking avoiding hypothesis, as higher minimum and peak frequencies would escape masking effects and increase signal transmission in the presence of low-frequency anthropogenic noise. Avoidance of low-frequency noise by shifting song minimum frequency may help to limit signal masking and enhance signal transmission (Nemeth and Brumm, 2009; Hu and Cardoso, 2010; Bermúdez-Cuamatzin et al., 2011). Although, it is possible that raised minimum frequencies could be a by-product of birds singing with higher amplitude as described by the Lombard effect (Brumm and Zollinger, 2013), in our study, birds in noisier habitats sang with considerably higher minimum and, notably, peak frequencies differing by about 400–500 Hz between habitats with less noise (minimums ~2300–2430 Hz) and noisier sites (minimums ~2700–2870 Hz) (Figure 2).

It is unclear if urban catbirds have adapted across generations to sing higher minimum frequency or if catbirds exhibit vocal flexibility to immediately shift their song frequency in response to low-frequency noise (Bermúdez-Cuamatzin et al., 2011). Moseley et al. (2018) found that nestling white-crowned sparrows (Zonotrichia leucophrys) tutored with noise learned less-masked songs significantly more often, suggesting that cultural evolution may be a potential mechanism for acoustic adaptation. The possibility of cultural selection may also be the case for gray catbirds, but a similar experiment would need to be conducted.

In addition to noise levels, canopy cover received model support for minimum, peak, and frequency bandwidth, as well as entropy (discussed below), though canopy cover was not always significant within the model as a fixed effect and parameter estimates were smaller than for noise (Supplementary Figure S1). Surprisingly, minimum and peak frequencies increased with increasing canopy cover, perhaps because some urban parks can be heavily forested (e.g., Rock Creek Park, and the arboretum at JMU), but reside within an urban matrix including high anthropogenic noise and large cover of impervious surfaces. Because all top models for song frequency included anthropogenic noise and noise estimates were significant within the top and second-to-top models (based on lowest AIC_C), we opted not to perform multi-model averaging, as interpretation of averaged estimates can be complex (Cade, 2015).

Few studies have examined bird species with complex, long, or versatile songs in terms of how anthropogenic noise affects song structure. Of species that have been studied to date, those with the most complex songs include nightingales (Luscinia megarhynchos, Brumm and Todt, 2002; Brumm, 2004), Pacific wrens (Troglodytes pacificus, Gough et al., 2014), and song thrushes (Deoniziak and Osiejuk, 2019). A previous study by Dowling et al. (2011) modeled the songs of six backyard bird species including recordings from eight catbirds. They found that gray catbird minimum frequency increased with higher noise levels for eight individuals, while song bandwidth and maximum frequency decreased with increased urbanization for only four catbirds at low-noise sites. The decrease in maximum frequency in Dowling et al. (2011) could have been a response to the urban structural environment and would make the song more tonal and better able to transmit without reverberation, while the increase in minimum frequency avoids low-frequency masking noise. However, their study analyzed few song structural features and had a sample size of 4–8 individual catbirds, as the goal of their study focused on trends across several backyard bird species. Lowering maximum frequency and the narrowing frequency bandwidth may increase signal transmission but may trade-off with attracting mates if singing broad frequency bandwidths is preferred by females (e.g., Halfwerk et al., 2011).

We found the opposite effect on maximum frequency in that catbirds in our study sang with significantly higher maxima in noisier habitats, as compared to previous studies, (e.g., Slabbekoorn et al., 2007; Dowling et al., 2011; Gentry et al., 2017). Gentry et al. (2017) found that urban white-crowned sparrows sang with lower maximum frequencies and narrower bandwidths than rural birds in a noise-playback experiment, and Phillips et al. (2020) found similar results for naturally varying song at more urban sites, but also at one rural site. While maximum frequencies increased with increasing anthropogenic noise for gray catbirds, maxima decreased with increasing impervious surface cover. Decreasing maximum frequencies in habitats with greater amounts of hard surfaces could be a response to the tendency for hard surfaces to increase reverberation, which attenuates signals (Phillips et al., 2020). These findings support the acoustic adaptation hypothesis (AAH), as songs with narrower bandwidths may transmit better in urban areas. The AAH states that signals are endowed with optimal characteristics for transmission to overcome environmental constraints (Morton, 1975; Endler, 1992). With an expanded sample size, we found that gray catbirds sang with a higher maximum frequency as noise level increased, but that frequency bandwidth only marginally differed. Even though urban catbirds have higher minimum song frequencies, it appears that there is compensation through higher maximum frequencies as well, resulting in a similar or slightly higher bandwidth. This compensation may be a result of sexual selection acting on male body size in these urban populations (Ryder et al., 2012). Males singing at higher minimum frequencies may be less appealing to females. However, through increasing maximum frequency, urban males are able to maintain a similar bandwidth, which has been shown to be attractive to females in other bird species (e.g., Ballentine et al., 2004).

Unlike previous studies on gray catbirds, we also measured entropy and phrase rate across an urban-to-rural gradient. We found no effects on phrase rate of element durations per bout durations. From our observations in the field, catbirds appear to sing longer and with a more rapid delivery of elements either early in the season, between nesting bouts, or in aggressive encounters (DLM pers. obs), but we do not have data to quantitatively assess this observation. Dolby et al. (2005) found phrase rate predicted feeding rates of nestlings, and thus phrase rate may indicate paternal quality to females. Average entropy was not predicted by noise level but was significantly higher in habitats with greater canopy cover. Notably, canopy cover did not change evenly across our urban gradient as rural farm sites had very low canopy cover and some very urban parks (Rock Creek Nat. Park in DC) had higher canopy cover especially at close radius distances from site centroids. The result of higher entropy with increasing canopy cover indicates that catbirds in more forested areas sing with less tonality and more disorder across a wider frequency range. Perhaps the high entropy in the songs of catbirds in more forested sites could be influenced by mimicry of forest heterospecifics (i.e., greater species and song diversity). Generally, entropy can potentially indicate vocal function, the effects of habitual conditions, and cognitive or song complexity of individuals (Briefer et al., 2010). For instance, song entropy decreases between early and late developmental periods of zebra finches, Taeniopygia guttata (Saar et al., 2008). Contrastingly, higher entropy has been associated with the trills of java sparrows that are signaled during aggressive encounters (Furutani et al., 2018). Call entropy of female vermilion flycatchers, Pyrocephalus obscurus, is significantly greater than call entropy of males, and potentially functions as a method for sex differentiation (Ríos-Chelén et al., 2020). For gray catbirds, a future direction would be to assess entropy across age or in the context of song repertoire and mimicry.

Further study of gray catbird song and what factors impact structural song differences is warranted to better understand what aspects of their invented, improvised, or imitated song repertoire differ. As vocal mimics, gray catbirds are unique in their ability to copy the sounds of other species; future studies could investigate interplay of urbanization, noise, and the incidence of song mimicry. It is well-established that many bird species in urban environments sing with a higher minimum frequency than rural birds, but the impact on female choice (Halfwerk et al., 2011) or mating success has received considerably less attention. Additionally, the effect of these song differences on reproductive success in urban catbirds is unknown. While shifting song minimum frequency may assist in signal transmission, certain aspects that make song attractive to females may be negatively impacted (e.g., Halfwerk et al., 2011). For example, lower frequency vocalizations are correlated with larger body size, which may mean that urban males are perceived as smaller and less threatening by rural males and may be less appealing to females. A study conducted by Francis et al. (2011) suggested that masking of lower-frequency signals, which are associated with higher quality males, may result in maladaptive mating decisions made by females. Alternatively for gray catbirds, if males are able to exhibit immediate flexibility in their songs, large-bodied males could potentially shift song frequency in real time to better transmit their signal acoustically, but still gain benefits through visual assessment for body size at close range by females. Since our results show that urban males do sing with higher minimum, peak, and maximum frequencies, there may be increased sexual selection for other traits such as body size as found by Ryder et al. (2012).

Examining the impacts of urbanization on birds is vital from a conservation perspective, as populations have decreased by 29% since 1970, resulting in a net loss of approximately 3 billion individuals (Rosenberg et al., 2019). Urbanization and its consequences, such as increased traffic noise and large-scale artificial surfaces, influence the vocalizations and behavior of multiple bird species (e.g., Halfwerk and Slabbekoorn, 2009; Nemeth and Brumm, 2009; Hu and Cardoso, 2010; Bermúdez-Cuamatzin et al., 2011). Determining how increased urbanization affects avian communication, reproductive success, and behavior is essential to future conservation efforts of vulnerable songbird species as well as understanding how humans influence the natural world.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The animal study was approved by James Madison University Institute of Animal Care and Use Committee Smithsonian Migratory Bird Center IACUC Bird Banding Lab. The study was conducted in accordance with the local legislation and institutional requirements.

DM and MR designed the hypotheses and song sampling techniques. DM, TR, and BE designed the study in terms of the urban gradient and choosing sites in Washington, DC and DM chose sites in Virginia. DM, MR, JT, CW, and field technicians made song recordings, MR, JT, LO, EN, CW, and DM processed song recordings. BE computed the geospatial metrics, managed banding data, and contributed code for model selection. MR and DM performed statistical analyses and led the writing, while all authors contributed to revising versions of the manuscript.