The ongoing spread of urban areas, highways and airports throughout the world makes anthropogenic noise almost omnipresent. We have found that urban great tits (Parus major) at noisy locations sing with a higher minimum frequency, thereby preventing their songs from being masked to some extent by the predominantly low-frequency noise. They have presumably learned selectively from a restricted range of their repertoire — a behavioural plasticity that may be critical for breeding success in a noisy world.

Cars, planes and all sorts of machinery create a new selection pressure on wildlife species that use acoustic signals to achieve reproductive success. This might create two groups of species: one that can adapt their signals to the competing noise, and another that cannot. Although there is a decline in species density and diversity associated with sprawling cities and highways1,2,3,4, there is no evidence yet for a direct role of sound pollution5,6, nor is there much insight into how successful urban species cope under noisy circumstances7,8.

We investigated an urban population of great tits in the Dutch city of Leiden. Noise-amplitude measurements, taken with a sound-pressure meter, varied markedly between territories. Mean amplitude levels per territory ranged from 42 to 63 decibels, from very quiet in residential areas to extremely noisy near a highway or a busy crossing. We used a highly directional microphone for song recordings and an omnidirectional microphone for independent noise recordings at a height of 5 m. The spectral composition of ambient noise was generally characterized by loud, low-frequency sounds.

We compared noise amplitude with the spectral distribution of sound energy within the range of the minimum frequency of great-tit song and found that in noisy territories there is a greater proportion of sound energy in the lower half of this range than in quiet territories (Pearson's r = 0.78, P < 0.001).

We measured the acoustic characteristics of 32 male great tits, each of which had a repertoire of between three and nine distinct song types. Mean song frequencies varied considerably between individual birds. The average minimum frequency ranged from 2.82 to 3.77 kHz and was significantly correlated with ambient noise (multiple regression: n = 32, d.f. = 2, F = 4.74, P = 0.017), with regard to both amplitude level (t = 3.02, P = 0.005) and spectral distribution (t = −2.0, P = 0.055). Noisy territories were home to great-tit males whose songs had a high average minimum frequency. Birds in quiet territories sang more notes that reached the lowest frequencies measured for the population (Fig. 1).

Figure 1: Correlation between song frequency and ambient noise in urban great tits.
figure 1

a, Two song types, one with low minimum frequency (Fmin) from a low-noise territory (left) and one with a high minimum frequency from a high-noise territory (right), recorded at a quiet moment. Centre, typical spectrum of urban noise. Recordings were made between 3 April and 13 May 2002. The sampling area covered 12 km2 and we recorded no direct neighbours. b, Relationship between average noise amplitude in a territory and average minimum frequency of the territory owner's song (n = 32, r2 = 0.11, t = 3.02, P = 0.005). Data were taken on eight rainless days (10 June to 14 July 2002). Each territory was visited on three days, each day at a different time: before, during and after rush hour. On each visit we took 20 repeated noise measurements in five directions. Average amplitude per territory was pooled over the three days.

It is possible that individuals with genetically predetermined song spectra could end up in matching territories with regard to noise spectra through a process of trial and error. But we know that great tits learn their song and that major adjustments occur in their breeding territory during interactions with neighbours9. Hence it is more likely that they learn to use a restricted range of their spectral capacity in response to frequency-dependent interference from local noise conditions — adjusting song to territory instead of territory to song.

Frequency use by great tits is also known to vary with sound-transmission properties in different habitats10, and correlations with natural ambient-noise spectra are found in other birds11,12. Local song adjustment within the heterogeneous urban habitat, as we find here, might indicate that habitat-wide adjustment through song learning may also contribute to acoustic divergence at the population level12,13.

Our findings show, to our knowledge for the first time, that human-altered environments might change the communication signals of a wild bird species5. The apparent song plasticity of great tits may represent a general behavioural mechanism that allows more bird species to reproduce despite high noise levels. Species that lack such learning plasticity after dispersal to the breeding territory, or have no room for variation within the conspecific frequency range, might suffer from auditory masking. For those species, anthropogenic noise could affect breeding opportunities and contribute to a decline in species density and diversity.