The selection intensity for codon bias and the synonymous diversity ha
ve been used in the recent literature to estimate the effective popula
tion size of Escherichia coli. The results have varied between 10(5) a
nd 10(8). It is suggested here that most of this disparity can be expl
ained by a model that accounts for the population structure of the spe
cies. Thus it is assumed that weakly selected characters, like synonym
ous substitutions, are selectively fixed within individual lines or co
lonies but spread throughout tile population in an essentially neutral
way when colonies replace one another. In this way, the effective pop
ulation size that enters expressions for the codon bias will be that o
f an individual colony, which, if hitchhiking effects are considered,
can be a very small number. The effective population size that appears
together with the mutation rate in expressions for the synonymous div
ersity, on the other hand, will be related to the total number of colo
nies that make up the species and can be a very large number.