According to life-history theory, any organism that maximizes fitness will face a trade-off between female fertility and offspring survivorship. This trade-off has been demonstrated in a variety of species, but explicit tests in humans have found a positive linear relationship between fitness and fertility. The failure to demonstrate a maximum beyond which additional births cease to enhance fitness is potentially at odds with the view that human fertility behaviour is currently adaptive. Here we report, to our knowledge, the first clear evidence for the predicted nonlinear relationship between female fertility and reproductive success in a human population, the Dogon of Mali, West Africa. The predicted maximum reproductive success of 4.1+/-0.3 surviving offspring was attained at a fertility of 10.5 births. Eighty-three per cent of the women achieved a lifetime fertility level (7-13 births) for which the predicted mean reproductive success was within the confidence limits (3.4 to 4.8) for reproductive success at the optimal fertility level. Child mortality, rather than fertility, was the primary determinant of fitness. Since the Dogon people are farmers, our results do not support the assumptions that: (i) contemporary foragers behave more adaptively than agriculturalists, and (ii) that adaptive fertility behaviour ceased with the Neolithic revolution some 9000 years ago. We also present a new method that avoids common biases in measures of reproductive success.