The universals problem is old and uncomfortable: when you call two different cows “cow,” what makes that right? Plato said there is a Form of Cow that both participate in. Aristotle said the universal is immanent, instantiated in each particular rather than transcendent over them. Neither answer satisfies. Participation is obscure: how does a particular cow relate to the Form without that relation requiring another universal? Immanence is mysterious: what does it mean for a universal to be wholly present in each of many numerically distinct particulars at once?
Dignāga, writing in 5th-century India, took a different approach. There are no universals. A concept picks out its instances not by a shared positive essence but by systematic exclusion: a cow is whatever is not-non-cow. Apoha (anyapoha, exclusion of the other) grounds meaning in boundary, not substance. You know “red” by where it stops being red, not by a mental template of redness. The move dissolves the participation problem immediately — there is nothing for particulars to participate in — and neutralizes immanence by removing the universal that was supposed to be immanent. The stability question does not disappear. It migrates: if concept-boundaries are defined by exclusion rather than essence, what stabilizes the boundaries? Why do different minds, carving meaning by exclusion, converge on similar divisions?
Russell arrived at exactly the same structural position from analytic philosophy of language in 1905. His theory of descriptions eliminates singular terms that purport to refer to entities (“the present King of France”) by analyzing them as quantified claims: “there is exactly one x such that x is currently king of France, and …” No positive entity is required. The meaning is constituted by a relational structure. Russell’s strategy and Dignāga’s are dual: both are anti-Platonist moves that dissolve the need for a positive entity by rerouting the semantics through relational or exclusion-based structure. Both transfer the stability question rather than answering it. Russell needs to explain what makes a description uniquely satisfied. Apoha needs to explain what makes an exclusion boundary stable across minds.
Connections
Dharmakīrti, developing apoha in the 7th century, introduced causal efficacy (arthakriyā) as his answer to the stability question: concept-boundaries track causal joints in the world. The exclusion is not arbitrary; it tracks where causal powers divide. This is proto-Kripkean externalism, 1300 years early. Kripke’s causal theory of reference holds that names refer not through descriptive content but through a causal chain anchored to an original act of ostension: “Aristotle” refers to whoever caused the appropriate uses of that name, not to whoever satisfies a cluster of descriptions. Dharmakīrti’s arthakriyā grounds exclusion in causal chains that run from concept-boundary to real causal distinction. The structures are not identical, but they converge: both hold that semantic stability is anchored by causal relations between language use and the world, rather than by mental representations or abstract entities.
Word2Vec and BERT are apoha implemented in gradient descent — representations shaped by which contexts and tokens are excluded from each other. The stability question reappears in the alignment literature as distributional shift: different training corpora carve exclusion-boundaries differently, and those differences are not always correctable. Dharmakīrti would say this is because different training distributions do not share the same causal joints. He would be right.
What lingered
Apoha dissolves Platonic participation and Aristotelian immanence. It does not close the universals problem. It relocates the problem to the stability of exclusion-boundaries, which requires either a causal answer (Dharmakīrti, Kripke) or a conventionalist one (Wittgenstein’s rule-following considerations, Putnam’s internal realism). Both horns of that choice reopen the same question at a deeper level. The remarkable fact is not that apoha fails; it is that the failure is structurally identical to the one that Kripke and Putnam ran into in the 20th century. Fifteen centuries of independent philosophy, in different conceptual vocabularies, converging on the same crossroads: how meaning gets anchored to the world without a Platonic shortcut. Nobody has a clean answer. The crossroads is still there.