There are two notions in the philosophy of probability that are often used interchangeably: that of subjective probabilities and that of epistemic probabilities. This paper suggests they should be kept apart. Specifically, it suggests that the distinction between subjective and objective probabilities refers to what probabilities are, while the distinction between epistemic and ontic probabilities refers to what probabilities are about. After arguing that there are bona fide examples of subjective ontic probabilities and of epistemic objective probabilities, I propose a systematic way of drawing these distinctions in order to take this into account. In doing so, I modify Lewis's notion of chances, and extend his Principal Principle in what I argue is a very natural way (which in fact makes chances fundamentally conditional). I conclude with some remarks on time symmetry, on the quantum state, and with some more general remarks about how this proposal fits into an overall Humean (but not quite neo-Humean) framework. This second notion of 'epistemic probabilities', which is perhaps most common in the philosophy of probability, belongs somewhere along the axis between subjective and objective probabilities. It is often contrasted with subjective probabilities when these are understood as merely 'prudential' or 'pragmatic' or 'instrumental'. This distinction is challenged by Berkovitz (2019), who convincingly argues that de Finetti's instrumental probabilities are also epistemic in this sense. Indeed, when talking about subjective probabilities in Sections 4, 6 and 10, I shall always implicitly take them to be so. This second sense of 'epistemic probabilities' is not the topic of this paper, and will be discussed no further. Epistemic probabilities in the sense of ignorance-interpretable ones instead are liable to be assimilated to subjective probabilities, but, as I shall argue, they ought not to, because the distinction between probabilities that are epistemic in this sense and those that are not (we shall call them 'ontic') is conceptually orthogonal to the subjective-objective distinction.