Alexander Bird, Emma Tobin (2008), Stanford Encyclopedia of Philosophy (ed. Edward N. Zalta) — https://plato.stanford.edu/entries/natural-kinds/
Key takeaways
From the reading, NOTE-769.
- Four questions, several commitments. Naturalism (some classifications are natural), strong realism (kinds are entities) and essentialism come apart; “Intermediate views tend to be more popular”.
- Usefulness is not enough. Shared natural properties, support for induction and a role in laws are each necessary but not sufficient: white things, charged things and Newton’s law are the counterexamples. Inductions “often concern relations”, but entering a natural relation does not make things a kind.
- Cluster kinds. Boyd’s homeostatic property clusters have vague boundaries and no necessary property; Slater recasts them as clusters stable enough for inference. Critics say the mechanisms are undelineated or conventional.
- Constructed but real. Hacking: whether a constructed category is real is “subject to empirical investigation”, as for child abuse, whose real instances were found, against satanic ritual abuse, whose were not. Social kinds are interactive: classified people change.
- Epistemology first, and its limit. Kinds judged by their role in scientific inference, against Lemeire’s reply that epistemic success constrains conventional classifications too, and so cannot separate them from natural ones.
- Essences. The Kripke–Putnam arguments give at most partial essences; Salmon argues that essentialism follows from the semantics only with an essentialist premise. Bird (2020) holds that the semantics carries those assumptions innocently.
Standing in the record
Filed on 2026-09-26 at the owner’s request, under the heading Relating K-complexity to minimum sufficient statistics, and read in full on 2026-10-14 in NOTE-769, superseding the seeded skim. The entry never mentions information, compression or statistics, so that link is the owner’s to make. Its weight here is on QUESTION-028: it makes the question’s qualifiers necessary and, alone, insufficient, and it supplies Slater’s stability and Hacking’s empirical test as candidate further clauses. THEORY-273, filed from the reading, holds that usefulness is necessary but not sufficient for a natural kind; it sits beside THEORY-036 on real patterns.