This thesis focuses on three aspects of evolutionary contingency: (i) the characterisation of evolutionary contingency, (ii) the mechanisms of evolutionary contingency, and (iii) the evidence for and against evolutionary contingency.

Part I of the thesis, which focuses on characterisation, challenges previous proposals of contingency by revealing their failure to entail the biological evitability' of outcomes. I diagnose the issue under a dynamical systems framework to make explicit an often-overlooked fact. This is that any evolutionary outcome is preceded by a series of antecedents, which may have varying probabilities of occurrence. In response to the failure of these accounts, I advance an original conception of evolutionary contingency as non-trivial objective probability that overcomes the shortcomings of some previous proposals. According to this outcome-based' account, contingency claims are probabilistic statements about an evolutionary outcome's objective probability of evolution. But contingency claims also ought to be indexed to a specifically defined `modal range', which, under the proposed account, is tantamount to the reference class. This additional apparatus allows for claims to be made more conceptually precise with respect to extent of modality, which, in turn, enables greater empirical tractability.

Part II centres around the `sources of contingency'. There has been little systematic investigation into the possible sources of contingency. Contingency-theorists have repeatedly gestured to certain processes as being sources of contingency but have failed to provide explicit reasons. Chapters 4 and 5 investigate key candidate sources of contingency: random mutations, recombination, genetic drift, species sorting, and mass extinctions. I conclude that random mutations fail to be a strongsource of contingency as they are often biased towards particular mutational outcomes on account of certain physical dispositions. On the other hand, mass extinctions and some forms of genetic drift are paradigmatic sources of contingency.

Just as contingency theses can vary in degree, so must sources of contingency. I provide an analysis of how the strength of a source of contingency varies; in short, it depends on the degree of bias towards any particular outcome(s). More technically, I import a statistical measure with which to evaluate the strength of a source of contingency, namely Shannon's information entropy.

Part III focuses on the evidence for and against contingency and attempts to dissolve the methodological pessimism associated with evolutionary contingency in the literature. Previous discussions pertaining to evidence have narrowly focused on convergent evolution'. I investigate its converse: evolutionary idiosyncrasies' or the singular evolution of certain forms, and I argue that they are supportive of evolutionary contingency. Armed with convergent evolution's counterpart (i.e. idiosyncrasies), I formulate a likelihood framework (whilst making explicit certain idealisations) for evaluating the relative significance and/or strength of contingency in any particular domain. In this way, the number of convergences vis-à-vis the number of idiosyncrasies in a modal range can be informative of the degree of contingency in the modal range, thereby dissolving some of the methodological pessimism.