We showed that it is possible to produce music by a Darwinian process in which consumers select particular tunes over others (1). Our objective was to isolate the role of one element of the evolutionary process: consumer selection. In our system, heritable variation was generated by a computer algorithm. Claidière et al. (2) argue that our methodology does not take account of the fact that the kinds of variants that humans generate are biased, and that such biases will influence the evolution of a population (e.g., ref. 3). We agree that such biases probably do influence the evolution of memes, just as they do the evolution of genes (4). The question, however, is the following: Can they produce music?

There are two transmission biases to be distinguished. One is a bias toward transmitting easily remembered memes. The other is a bias toward transmitting aesthetically pleasing memes. The language diffusion experiments of Kirby et al. (3) show the former but not the latter. A thought-experiment makes the distinction clear. Suppose we set up a music diffusion chain in which people were asked to reproduce, on some instrument, a song consisting of a random set of notes. We would expect that the songs would become simpler, and hence more reproducible. Just as in language-diffusion experiments, transmission fidelity would improve. However, would the songs become more musical? It is possible; perhaps the subjects would, in the course of reproducing the songs, naturally tend to transform them into more pleasing songs, say, with chords or harmonies and so on. However, it is also possible that there is a tradeoff between musicality and transmissibility. After all, nothing is more reproducible than a monotone.

Perhaps the relative importance of transmission bias and selection in cultural evolution turns on the distinction between language and music. Language consists of ordered sounds that have a stable semantic meaning; music consists of ordered sounds that are pleasing to the ear. To be sure, the evolution of language may be affected by aesthetics, and music by semantics, but the contrast remains (5). We suggest that language may evolve with little or no aesthetic selection but that music cannot.

However, the experiments remain to be done. Claridiére et al.’s letter (2) raises fascinating questions. Were we to devise an experiment in which subjects were allowed to select and transmit sounds, what would happen? Would the outcome of such an experiment be affected by whether or not the sounds have semantic meaning or not? Such questions can be answered by Darwinian experiments, such as those of Kirby et al. (3) and MacCallum et al. (1). The analytical framework that we introduced, based on the Price equation, is also ideally suited to partitioning the forces at work in such experiments.