Of all types of behaviour, mating signals are the most likely traits to function as important isolating mechanisms. They are used by animals to attract mates, and changes in a signal used by one population may prevent interbreeding with another simply because individuals no longer recognize each other. In this way, speciation—the process by which a parent taxon splits into two reproductively isolated populations—could be driven by behavioural factors, as much as by divergent ecologies or mutated genes.

Of all mating signals, from the subtle to the spectacular, birdsong is perhaps the best studied. We know that songs are crucial to mate choice, species recognition and the prevention of hybridisation, and we suspect that they play an integral role in the speciation process. However, there is little convincing evidence to support this notion, perhaps because previous research has been constrained by two major biases.

First, the focus on temperate regions has limited our ability to understand the origins of biodiversity, which is largely a tropical phenomenon. Second, most studies of song have focused on oscine birds, the common songsters of the temperate zone, which produce complex variable songs that are learnt at an early stage of development, and modified during adulthood. While learning may promote rapid change in song structure when populations become separated (because anomalies accumulate, forming dialects), it probably hinders speciation because closely related forms copy each other's songs when contact is resumed, leading to hybridisation.

My research addresses these problems by focusing on suboscine birds in tropical Amazonia, where the diversity and complexity of life reaches its utmost expression. This is true for all major groups of plants and animals, but perhaps most obviously for birds: a 100 ha plot of woodland in temperate North America supports 70 species at most, and in Europe fewer still, but the same area of Amazon rainforest contains upwards of 500 species. This astonishing total is made up largely of suboscine birds, whose songs are fixed, unlearnt and relatively simple, making them better subjects for studies of song and speciation.

Under the broad question of how species form, I am tackling two key issues. One is the assumption that new behaviours only arise in the latter stages of speciation, as a by-product of genetic and ecological changes accumulated over long periods of time. The other concerns the 'allopatric' and 'sympatric' models of animal speciation, which, despite being established ideas, have rarely been experimentally tested. According to the former, populations must be geographically isolated from one another for new species to form; according to the latter, new species can arise without physical barriers to gene flow. I aim to determine whether changes in behaviour in allopatry can act to catalyse speciation, rather than merely finalise it when sympatry is resumed, and whether this process has driven the radiation of suboscine birds in Amazonia.

But given that speciation cannot be observed directly, how can it be studied? The only viable approach is to map current patterns of variation and use these patterns to make inferences about changes in evolutionary time. Focusing on suboscine birds, I use comparative analyses in conjunction with experimental fieldwork and genetic studies to examine the role of song in the early stages of divergence. To determine how geographic variation in birdsong impinges on speciation, I use 'playback experiments'. These involve simulating contact between geographically isolated populations by playing songs from one population to members of another and measuring the response. This can then be compared with the response to local songs, allowing the likelihood of interbreeding to be assessed. As such, playbacks circumvent the need for translocating territorial birds (almost impossible) or for actual contact to be resumed between populations (this can take a long time).

The suboscine radiation is numerically important, accounting for over 20% of all passerine birds, yet suboscines themselves are virtually unstudied by evolutionary biologists. Because their unlearnt songs are likely to prevent breeding between divergent populations, my research has the potential to reveal a key role for vocal communication in animal speciation, evidence for which is lacking.