Introduction
To understand the role of adaptation in generating macroevolutionary patterns, it is necessary to understand whether and in what ways specific features of the phenotype affect subsequent phenotypic diversification. This area has been much debated by both past and present workers, some of who considered whether certain morphologies might be ‘channelled’ (e.g. Gould, 1984; Emerson, 1988; Wagner, 1996) to appear once a specific starting morphology was attained. Less radically, a number of workers have suggested that possession of certain morphological character states may reduce the ability to attain certain other character states (Lauder, 1981; Maynard-Smith et al., 1985; Emerson, 1988; Futuyma and Moreno, 1988; Wagner, 1996; Werdelin, 1996; Alroy, 2000; Donoghue and Ree, 2000; Wagner and Schwenk, 2000; Wagner, 2001; Wagner and Mueller, 2002; Porter and Crandal, 2003; Van Valkenburgh et al., 2004; Polly, 2008), implying that, in some cases, taxa may be limited in their subsequent evolutionary trajectories. Both morphological channelling and a limitation on specific character states fall into the realm of a character change bias, where certain states are more likely to appear than others (Sanderson, 1993; Wagner, 1996; Donoghue and Ree, 2000; Wagner, 2001; Goldberg and Igic, 2008; Polly, 2008). Despite ongoing theoretical debate, however, there has been relatively little empirical exploration of the possibility of bias or directionality in morphological character change, and this area remains poorly understood (Arthur, 2001, 2004; Schluter et al., 2004).