Do birds vocalize at higher pitch in noise, or is it a matter of measurement?

Studies have found that some birds use vocalizations with higher minimum frequency in noisy areas. Minimum frequency is often measured by visual inspection of spectrograms (“by-eye practice” (BEP)), which is prone to bias, e.g., if low-frequency components are masked by noise. We tested for this bias by comparing measurements of minimum frequency obtained with the BEP for the same set of red-winged blackbird vocalizations (songs and two call types “checks” and “cheers”) played back under ambient, medium, and high noise conditions using a dual playback experiment where both vocalizations and noise were introduced. We compared BEP measurements to those obtained from power spectrum analyses using a preset amplitude threshold (“threshold method” (TM)). The BEP was biased when measuring the minimum frequencies of songs and checks, which are masked by noise, but not when measuring cheers, which are higher pitched and thus not masked. Measures using the TM were not affected by noise, but this method may fail to identify the vocalizations’ lowest frequency if noise necessitates a low (i.e., conservative) threshold. Using the BEP, we also found a bias toward shorter-duration measurements for songs in increasing noise, and for checks, a bias toward increased measures of an energy distribution parameter (Freq5%), likely in correlation with increased measured minimum frequency. Measures taken from the unmasked cheers were similar regardless of the technique used. We discuss limitations of each approach and encourage the use of the TM, as studies using the BEP may lead to spurious results. Significance statement Noise from human activities is ubiquitous. Researchers have found that some birds vocalize at higher frequency (pitch) in noise, hypothesizing that this may improve signal detection in low-frequency noise. Noise may also hinder detection of signal components by researchers using the most common measurement technique (the BEP), which may be mistaken for increasing frequency. To examine this bias, we conducted a dual playback experiment, in which we broadcast the same vocalizations at three background noise levels. We found that BEP measures of minimum frequency increased with increasing noise even though the vocalizations did not change. We recommend the TM which yielded similar measures across noise levels, although it excluded some lower-frequency elements included by the BEP. We encourage researchers to use the TM over the BEP and to validate their methods across noise levels of interest.