This paper uses world running records by age to estimate a biological frontier of decline rates. Two models are compared: a linear/ quadratic (LQ) model and a nonparametric model. Two estimation methods are used: (a) minimizing the squared difference between the observed records and the modeled biological frontier and (b) using extreme value theory to estimate the biological frontier that maximizes the probability of observing the existing world records by age. The results support the LQ model and suggest a linear percentage decline up to the late 70s and quadratic decline after that.