We question the claim that the largest US cities are Pareto distributed. We show that results of multiple tests on real data are similar to those obtained when the true distribution is lognormal, and largely depend on sample sizes.