In genetics, as in many other fields, theoretical models rarely provide good substitutes for empirical evidence; after all, who would continue to uphold a model that has been contradicted by experiments? However, in a reversal of this situation, the authors of two studies published in Genetics now successfully defend the primacy of theoretical models, and force us to review our current views — that are largely derived from experiments — on developmental evolution.

In the late 1950s, Waddington laid the conceptual foundations for a growing body of research into the ability of a wild-type genetic system to withstand the effects of mutations — a property that he named 'canalization'. According to this idea, a wild-type character is so well adapted to its environment that any deviation from this condition — caused by a mutation, for example — would inevitably be deleterious. As a result, the wild-type trait would evolve to be buffered from any insult that threatened to deviate it from this fitness optimum. The fact that organisms show increased phenotypic variability after a genetic or environmental 'stress' provides good evidence that wild-type organisms suppress a certain degree of hidden genetic variation (also called cryptic genetic variation (CGV)), which is released only under altered circumstances.

Hermisson and Wagner have now probed this concept further. Is the release of CGV intrinsically linked to a robust (that is, canalized) wild-type character? Their theoretical model indicates otherwise. The authors devised a mathematical expression that would follow the impact of change (in the form of a genetic mutation or altered environmental conditions) on several statistical properties of a quantitative character. They found that CGV is always unleashed when a trait is subjected to change — this occurs when any genetic background is perturbed, not only a canalized, wild-type one. In fact, the observed increase in genetic variance seems to be a generic property of any system that has epistasis and genotype × environment interactions. The message that goes out to the growing number of biologists that are interested in CGV is that canalization is not a prerequisite for accumulating CGV and, conversely, that CGV does not constitute sufficient evidence for canalization.

In the second paper, Bagheri and Wagner focus on the evolution of phenotypic robustness itself, and tackle the thorny issue of how dominant phenotypes — which themselves represent a form of robustness to mutations — have evolved. Biologists have dismissed the need for an evolutionary explanation for dominant phenotypes on the grounds that selection for dominance can only arise under special circumstances. Alternatively, in some situations (such as metabolic pathways), it has been maintained that dominance is simply a default property of the system. The theoretical model that is presented in this paper, which is based on the interaction between two metabolic enzymes, now punches a few holes into this long-standing theoretical argument. Not only does the model predict that dominance can evolve as both a direct or indirect result of selection, but that the predictions of the model can be applied more generally to explain the evolution of any form of phenotypic robustness to mutations.

The results of these two studies, and the revisionist approach to the literature that they suggest, are intriguing. But what is equally admirable is the authors' effort to reconcile three areas of research — population genetics, quantitative genetics and phenotypic robustness — that have largely enjoyed independent research histories.