Are evolutionary rates really variable?

Abstract

Langley and Fitch (1974, 1976) have shown that the pattern of nucleotide substitutions in proteins is inconsistent with a Poisson process with constant rate. From this they conclude that the rate is temporally heterogeneous. It is pointed out in this note that a process which is temporally homogeneous but not a Poisson process is compatible with the data if the coefficient of variation of the time between substitutions is around 1.63. Furthermore, theoretical analysis of samples from neutral phylogenies shows that these samples should not appear to be samples from a Poisson process, but should deviate from a Poisson process in the same direction, though perhaps not to the same extent, as do the data.