Abstract: Several commentators find in the writings of Lady Mary Shepherd a robust, realist account of natural kinds, or at least a commitment to them. Lolardo (2019) reads Shepherd as committed to "sharp natural boundaries in nature" (2). Boyle (2023) reads Shepherd as committed to natural kinds, while not explicitly arguing that they exist (66). For Bolton (2019), Shepherd thinks "kinds and their qualities and powers have metaphysical grounds in nature" (151). All three also find in Shepherd a doctrine of necessary a posteriori truths. While there is some justification for reading Shepherd as having a robust account of natural kinds, important qualifications are needed. If we are careful to avoid reading our contemporary assumptions about natural kinds into Shepherd, we find that her account is importantly different from contemporary natural kind realism. And while there is some basis for the claim that Shepherd accepts necessary a posteriori truths, the claim is misleading, since any truths about kinds will be analytic and not a posteriori.

My strategy is to read Shepherd's account of kinds in the context of Locke, given that Locke was a major influence on Shepherd, and that Shepherd downplays her disagreements with Locke. I aim to clarify Shepherd's thinking about kinds by highlighting those disagreements. Major differences include a commitment to causal principles, a rejection of mechanistic explanations, and a rejection of the prominent role of human decision in sorting the world into kinds. One similarity is their realism about causal powers and the appeal to powers in inductive inference: it is our ability to make reasonable inferences to "secret powers" that justifies our predictions about kinds. Another is that Shepherd never appeals to anything like substantial forms that divide the world according to a select set of special or essential properties. This, together with her rejection of Locke's permissive approach to sorting, leads to an account of kinds that are very fine-grained. Kinds are like in all respects that make a difference to their qualities and effects–that is, all respects that we can discern. The robustness of the account is its insistence that we follow causal powers in defining kinds rather than sort according to properties that are of interest. But the account is not nearly as realist as contemporary accounts that claim the world comes divided according to essential properties.

The best explanation for Shepherd's rejection of Locke's permissive approach to sorting, as well as for her frequent slippage between talking about particulars and kinds, is that Shepherd is doing something very different from Locke with respect abstraction and the content of general ideas. For a representation of a particular to also represent a kind, we need only abstract those aspects that it cannot share with others, such as its time and place of existence. No other discernible property can be separated at will, as these have causal implications. In this way, a representation of a particular can represent objects like it, which is a kind.