Even if I concede that it's a damn fine start, it is not research sufficient to jump to the conclusion that (a) CSS layouts are overcompensating (based on a "hypothesis" taken out of thin air) and (b) people who believe using tables for layouts is bad are wrong based on said research.

Edit: He also managed to not even properly conduct this simplistic research considering that many of the tables used on the sites were used correctly, e.g. for tabular data.