Measuring frequency-dependent selection in culture

Cultural traits such as words, names, decorative styles, and technical standards often assume arbitrary values and are thought to evolve neutrally. But neutral evolution cannot explain why some traits come and go in cycles of popularity while others become entrenched. Here we study frequency-dependent selection (FDS)--where a trait's tendency to be copied depends on its current frequency regardless of the trait value itself. We develop a maximum-likelihood method to infer the precise form of FDS from time series of trait abundance, and we apply the method to data on baby names and pet dog breeds over the last century. We find that the most common names tend to decline by 2%-6% per yr on average; whereas rare names--1 in 10,000 births--tend to increase by 1%-3% per yr. This specific form of negative FDS explains patterns of diversity and replicates across the United States, France, Norway and the Netherlands, despite cultural, linguistic and demographic variation. We infer a fixed fitness offset between male and female names that implies different rates of innovation. We also find a strong selective advantage for biblical names in every frequency class, which explains their predominance among the most common names. In purebred dog registrations we infer a form of negative FDS that is consistent with a preference for novelty, in which each year's newest breeds outgrow the previous by about 1%/yr, which also recapitulates boom-bust cycles in dog fanciers. Finally, we define the concept of effective frequency-dependent selection, which enables a meaningful interpretation of inferred FDS even for complex mechanisms of evolution. Our analysis generalizes neutral evolution to incorporate pressures of conformity and anti-conformity as fundamental forces in social evolution, and our inference procedure provides a quantitative account of how these forces operate within and across cultures.