Although laboratory dependence is an acknowledged problem in microbiology, it is seldom intensively studied or discussed. We demonstrate that laboratory dependence is real and quantifiable even in the popular model Escherichia coli. Here laboratory effects alter the equilibrium composition of a simple community composed of two strains of E. coli. Our data rule out changes in the bacterial strains, chemical batches, and human handling but implicate differences in growth medium, especially the water component.