Hypothesis Testing In Evolutionary Inference

The comparative method, amongst other things, searches for correlations between evolutionary variables. These can be used to test null hypotheses. Here I consider, in the context of binary variables, the bases of such tests. I examine grounds upon which evolutionary traits and events can be regarded as statistically independent of each other. I argue that no description of observations as independent or non-independent makes sense except in the context of a population of possible observations from which they are regarded as having being sampled. Significant correlations between traits or changes in traits in comparative tests have been taken by some to imply causal links between traits. However, the statistical significance of an observed correlation between traits is neither necessary nor sufficient for the inference of a causal connection between them.