Fondon and Garner have recently offered a hypothesis that gene-associated tandem repeat expansions and contractions in the protein coding regions of developmental genes are a major source for rapid morphological variation in dog breeds . Repeat expansions and contractions can occur at rates up to 100, 000 times higher than point mutations [2, 3], so this class of mutation potentially has a much greater effect on morphological evolution than point mutations. Wallace Arthur has recently proposed that developmental bias, the tendency for developmental systems to produce variant trajectories in some directions more readily than others, is as important as natural selection in driving morphological evolution . In this review, we present arguments that repeat expansions and contractions, because they affect morphological features in a specific and graded manner, are examples of developmental bias, and therefore support Arthurs “biased embryos” hypothesis of morphological evolution. We also extend the “biased embryos” model by exploring the possibility that expansions, contractions, and retrotransposon mobilizations are epigenetically upregulated during times of stress, possibly through a genome scanning process that utilizes Hsp90 in germ cells. In support of this idea, we found that the incidence of CpG dinucleotides is much higher in vertebrate trinucleotide repeats than in other protein coding regions, thus suggesting that CpG methylation is under stabilizing selection. Based on these and other observations, we propose a model whereby the regulation of the CpG methylation status of repetitive sequences in germ cells could be a powerful means to increase the rate of morphological variation, and thereby the rate of morphological evolution, during times of stress.