This paper uses techniques from theoretical biology and population genetics to investigate how likely different scenarios for the biological evolution of language are. Specifically, it investigates whether, given what we know about the time frame of human evolution and the estimated early human population sizes, a scenario depending on a single large mutation or on a series of small mutations is more likely. It uses extreme value theory and diffusion models to estimate the probability of mutations occurring and spreading in the population. It finds that in the end, scenarios consisting of an intermediate number of medium-sized mutations is most likely.