Evolution of Phenotypic Robustness Joachim Hermisson and G¨ unter P. Wagner Department of Ecology and Evolutionary Biology, Yale University, New Haven, CT 06520, USA 1 Introduction Evolutionary biology, in the neo-Darwinian tradition, is based on the study of genetic and phenotypic variation and its fate in populations. Thus the observation that genetic variability of a trait is itself influenced by the genotype has obvious theoretical implications (Waddington, 1957; Stearns, 1994). It is in this context that the robustness of phenotypic traits was first conceptualized as canalization and became the focus of a significant research effort (reviewed in Scharloo, 1991). With increasing awareness of the intricate molecular mechanisms maintaining the life of cells, the ubiquity of buffering and compensatory mechanisms came into focus (Wilkins, 1997; Gerhart and Kirschner, 1997; Rutherford, 2000). Recent years have seen a confluence of the classical concept of canalization and new research in molecular biology that resulted in a sharp increase in the interest in canalization and related phenomena. One can speak of an emerging field of biological robustness research that is able to draw on a sophisticated arsenal of technical and theoretical tools that were developed over the last ten years (de Visser et al., 2003). The present chapter aims at summarizing the principal findings of the classical and the newer literature on robustness of phenotypes and to identify the issues that require attention in future research. We will start with an attempt to formalize the notion of phenotypic robustness to clarify the criteria that need to be met in order to experimentally demonstrate phenotypic robustness. Then we review the experimental evidence about environmental and genetic robustness and find that genetic robustness is particularly difficult to demonstrate. Finally an overview of theoretical models for the evolution of canalization is provided which shows that the existing literature is strongly biased towards a few scenarios. We conclude that there are major unsolved questions both in the experimental demonstration of canalization as well as in understanding the evolutionary dynamics of canalization. 2 Defining phenotypic robustness Phenotypic robustness is about the sensitivity of the phenotype (e.g. some quantitative trait) with respect to changes in the underlying variables (genotype and environment) which determine its expression in an individual. Intuitively, this seems to be a clear notion. However, although the concept originated more than forty years ago there is a conspicuous lack of a formal definition. Such a definition should entail a clear criterion for the detection and characterization of robustness in empirical and theoretical study. For empirical work, robustness must be formalized as a operational concept. In order to be of conceptual use for an understanding of evolutionary processes, robustness should be classified into types that provide information about its evolutionary role and the circumstances of its origination. In this section, we will point out some key elements for such a formal definition and discuss demarcations to related concepts. Our starting point is the formalization of robustness as a state of reduced impact of a given source of variation (such as mutations or environmental change) on the trait, i.e. as reduced variability due to that source (cf Wagner and Altenberg, 1996; Wagner et al., 1997). Let us state this first in an informal way: A character state that has evolved under natural selection is phenotypically robust if the variability of the character under a given source of variation is significantly reduced in this state as compared to a 1