An inverse relationship between a eukaryotic gene's level of methylation and its level of expression has long been recognized, and generally believed to result from reduced binding of postulated activator proteins to methylated target DNAs. There are, however, some genes where there is no apparent correlation between levels of methylation and gene expression, and even a small class where gene activation is correlated with increased methylation of the DNA. I propose a unifying hypothesis to explain these apparently divergent cases: methylation acts to reduce or abolish binding of regulatory proteins to their DNA target sites. In the majority of genes, methylation acts to block binding of activating factors; "indifferent" genes lack such methylation sites, while the minority class, which is more active when methylated, contains methylation sites which block binding of repressor proteins.