This paper characterizes two distinct philosophies underlying previous work on how Bayesian computations are linked to neural data, highlighting how different theories may be motivated by different tacit assumptions and thereby explain different data.