General Motivation
My primary interest is in how genetic and cultural evolution interact. Research over the last 35 years has established culture as an interesting parallel evolutionary mechanism to genetical evolution, even constituting the most recent major transition in evolution (Maynard Smith & Szathmary, 1998)... at least in theory. While many rigorous theoretical models have provided a well-rounded conceptual backbone to the field, empirical support is comparatively spartan. Although notable exceptions exist (e.g. Durham, 1990), there is a pressing need for more established empirical systems of cultural evolution to test the models. Bird song is potentially just such an example of cultural transmission. It is ubiquitous and diverse: most songbirds seem to require a cultural input to develop a normal song, but there is great variation between species in the form and function of songs and song repertoires. It is also comparatively easily to quantify through the bioacoustical analysis of song recordings. Perhaps most importantly, bird song is already one of the best-developed empirical research fields within animal behavior. This has provided a wealth of knowledge across all of Tinbergen’s four ethological domains, including function, where song has been a critical testing ground for many theories of sexual selection and signaling. Within the bird song research field, the cultural evolution of song is widely acknowledged. Nevertheless, in my opinion, the consequences of cultural evolution have not been taken into consideration. Cultural evolution is often treated as a curious side-effect of song learning, epiphenomenal, and of no great importance. This contrasts with the perspective initially developed by Feldman & Cavalli-Sforza (1973), who constructed the first general theoretical model of gene-culture coevolution. They realized that because culture changed the behavior of organisms in a predictable way, it might be expected to change selection pressures in a stable way, and thus the direction of genetical evolution. From this, one might think that most if not all aspects of the evolution of song, and the communication system underlying it, would be shaped to some degree by cultural evolution. The focus of my research is to attempt to test this general hypothesis.
Results up to the Present
To develop a research program to investigate the role of cultural evolution in song evolution, I have employed three general methodological approaches:
1) Theoretical. I have attempted to develop theoretical models that are explicitly cultural evolutionary to explore its affect on song evolution. Since alternative song-types are often regarded as equivalent from a population-level perspective, this research involves modeling the effects of drift more explicitly than most other research into cultural evolution. I have published four conceptual theoretical studies, and have one close to submission that indicate that cultural evolution may indeed influence many aspects of song evolution. These cover:
• the evolution of song versatility and song learning itself (Lachlan & Slater, 1999, Lachlan & Feldman, 2003), where we found that cultural evolution might create a selection pressure for more versatile song learning. This could explain the evolutionary maintenance of song learning.
• the rate of speciation (Lachlan & Servedio, 2005), where we found that cultural evolution might increase the rate of allopatric speciation, due to a shielding effect generated by cultural transmission for rare alleles.
• the evolution of communicative function attached to song sharing (Lachlan et al, 2004) and song learning accuracy (Lachlan & Nowicki, In Prep.). In the former, we found that song conformity might be enforced by multi-player interactions in spatially explicit games. In the latter, we found that song categorization, structure and repertoires might evolve as a way to facilitate song-learning assessment as predicted by the developmental stress hypothesis.
2) Bioacoustical analysis. Although sound spectrograms have been employed for many years to visualize bird song, progress has been slower in developing quantitative tools to compare songs. This has hampered our ability to measure cultural evolution. I have developed a bioacoustical software package, Luscinia, (http://luscinia.sourceforge.net) that uses developments in the fields of computational linguistics, statistics and data-mining to provide a solution to this problem for field recordings.
The computational song comparison method implemented in Luscinia has been tested on three species, where it generated similarity scores that corresponded very closely to those of human observations of spectrograms (Lachlan et al, Submitted). We also used the method to test the degree to which phonological and syntactical units fell into discrete “universal” categories. We found only very limited evidence for such categories, suggesting that “learning by instruction“ is a more important developmental mechanism than ”learning by selection“ (Marler, 1997).
Luscinia has developed into a generally useful research and pedagogical tool. It is currently used in around 10 research projects around the world, and interest in it is growing.
3) Fieldwork. I have integrated the two methods above with the standard field tools of bird song research: song recording and playback experiments.
A specific goal has been to fit geographical patterns of song variation to models of cultural evolution. Lachlan & Slater (2003) represents the first step in that process. A simulation of song variation was fitted to geographic variation in song-type diversity in chaffinches (Fringilla coelebs). This enabled us to estimate two important parameters of learning behavior: the error rate, and the dispersal distance.
A second project has examined the evolution of song in isolated oceanic island populations of chaffinches (three studies, in prep.). Using the analytical techniques developed in Luscinia, these demonstrate how song structure and learning behavior have evolved along a colonization chain. We have found that learning evolved from accurately learning whole song-types on the mainland to accurately learning constituent syllables, but not whole song-types on the islands. Concurrently, syntactical and phonological constraints on song structure have progressively weakened along the chain. These changes are reflected in responses to playback of song form different islands.