Cultural variants change in frequency through the chance of who copies whom, so popularity can shift substantially with no variant being better.
Neutral models predict the distribution of variant frequencies and turnover rates expected from random copying alone, and the predictions fit domains as different as first names, dog breeds and pottery decoration surprisingly well. The value of the model is as a null hypothesis: an explanation of why something became popular owes an account of what it adds over drift.
In practice
- A baby name rises for twenty years and falls for twenty, and no cause is needed beyond people copying what they have recently heard.
- A team's coding convention won out because two influential people happened to use it in the first month.