Song measure averages of allopatric and sympatric Black-capped and Mountain Chickadee individuals

Character displacement theory predicts that closely related co-occurring species should diverge in relevant traits to reduce costly interspecific interactions such as competition or hybridization. While many studies document character shifts in sympatry, few provide corresponding evidence that these shifts are driven by the costs of co-occurrence. Black-capped (Poecile atricapillus) and Mountain Chickadees (P. gambeli) are closely related, ecologically similar, and broadly distributed songbirds with both allopatric and sympatric populations. In sympatry, both species appear to suffer the costs of their co-occurrence: 1) both species are in worse body condition compared to allopatry and 2) hybridization sometimes yields sterile offspring. Here, we explored character displacement in the songs of Black-capped and Mountain Chickadees by characterizing variation in male songs from sympatric and allopatric populations. We found that Mountain Chickadees sing differently in sympatry versus allopatry. Specifically, they produced more notes per song, were more likely to include an extra introductory note, and produced a smaller glissando in their first notes compared to all other populations. Combined with previous research on social dominance and maladaptive hybridization between Black-capped and Mountain Chickadees, we posit that differences in sympatric Mountain Chickadee songs are population-wide shifts to reduce aggression from dominant Black-capped Chickadees and/or prevent maladaptive hybridization.

Methods
Field sampling

During April and May of 2020, we recorded breeding Black-capped and Mountain Chickadees in both sympatry and allopatry. We recorded sympatric populations in Boulder County, Colorado, an area where hybridization is known to occur and we sampled along both species’ elevational distributions (Black-capped Chickadee n = 21; Mountain Chickadee n = 11). We recorded allopatric Black-capped Chickadees in Ithaca, New York (n = 21), and allopatric Mountain Chickadees near Truckee, California (n = 20). In both allopatric populations, neither species overlaps with any other Chickadee species. To avoid repeat sampling, we walked at least 500 meters between recordings and never sampled in the same location twice. We collected a total of 2,080 songs and an average of 29 songs per individual. All individuals produced a minimum of six songs. We recorded songs with a PMD660 or PMD661 Marantz digital recorder and a Sennheiser ME-66 unidirectional microphone with a sampling rate of 44,100 Hz and a 16-bit resolution. 

Measuring song characteristics

We analyzed songs using the bioacoustics software Raven Pro 1.6.3. We formatted spectrograms in the color scheme “Jet” with the x-axis scaled at 0.1 second intervals and the y-axis scaled at 0.5 kHz intervals. To control for variation in background noise levels and recording distance from birds, we adjusted contrast and brightness on each recording to establish consistency among all recordings. To control for researcher subjectivity, only one researcher took all song measures. As such, brightness and contrast were adjusted somewhat subjectively, but consistently by the observer to optimize visualization of the song. For all songs we measured song duration, frequency ratio, and internote interval, which was calculated post hoc (Table 1). For each note we measured duration, glissando, frequency 5%, and frequency 95%. To avoid pseudoreplication, we calculated the mean of each song measure per individual across all of their songs, such that each individual is represented only once in downstream statistical analyses. For each song from an individual, we annotated song number, note number, and whether or not the song included introductory notes. These song measures are based on previous research on Black-capped and Mountain Chickadee song. No recordings contained songs with more than three introductory notes. We omitted faint noise or “echoes” at the beginnings and ends of notes, which resemble smears of a consistent, recording-specific, length in the spectrograms. As such, notes that appeared to be connected by an amplitude dip rather than separated by a true break were selected as two separate notes; we treated the amplitude dips as omittable echoes and selected them as different notes. While our analysis methods differ from traditional chickadee literature (48), which treat notes connected by an amplitude dip as a single note, we treated all populations the same so results are not affected. Importantly, this aspect of song does not drive any of the differences we report. 

Statistical analysis

We performed all statistical analyses in R version 4.0.3 using the packages tidyverse and ggplot2. To assess whether chickadee songs exhibit more divergence in sympatry, as predicted from character displacement theory, we used a principal component analysis (PCA) to assess if songs cluster together by population. We used individual means for all song measures in the principal component analysis. Then, to assess what features of song explained the most variation between and within populations, we ran linear models (command lm()) for each PC axis that had a loading stronger than 0.3, with population (Allopatric BCCH, Sympatric BCCH, Allopatric MOCH, and Sympatric MOCH) as the dependent variable. To assess variation between populations, we ran ANOVAs (command aov()) on each linear model followed by a Tukey HSD test to quantify differences between populations. Statistical significance was set at p < 0.05.