What counts as scientific knowledge, and how does it change?

Rationalist in logical structure — certainty derives from the necessity of demonstration; induction merely brings the intellect to the point where it can perform it.

Shifts weight toward experience decisively. Reason's role is methodological rather than foundational — a bridle, not a source.

Radicalizes the experiential commitment: reason cannot cross the gap between observed instances and universal law. Experience stands alone — and alone, it cannot justify itself.

Refuses the opposition itself. Reason structures experience; the categories are the very medium through which coherent experience becomes possible. The tension is relocated — not dissolved.

Sets the standard at its highest: episteme is of what cannot be otherwise. Anything short of this is opinion, not science.

Lowers the bar in practice, though without announcing it philosophically. Reformed induction yields reliable, revisable knowledge — not necessary truth.

Makes fallibility explicit and irreducible. No empirical claim has logical certainty. The appearance of certainty in our causal beliefs is a psychological illusion.

Partial rehabilitation — causal judgments within possible experience are necessarily and universally true — but only within that domain. The certainty is transcendental, not metaphysical.

The relation between knower and nature is direct and, in principle, transparent. The intellect is made to know being. The subject is not a problem — it is the solution.

Introduces the knowing subject as a problem for the first time. The Idols systematically distort perception — the individual mind must be disciplined before it can function as a reliable instrument.

Deepens this: causality is not in the world but projected by the mind, and the subject cannot step outside its own habits of association to verify that nature corresponds to its expectations.

The subject's constitutive role is not merely an obstacle but the condition of knowledge. The question of what stands beyond that constitution is left permanently open.

The syllogism, applied to known first principles, guarantees necessity. Method is the formal expression of the intellect's proper activity.

Method's task is purgation and discipline. It guarantees reliability, not certainty. Its product is corrigible rather than final.

Method can guarantee nothing of logical consequence. However rigorous the inductive procedure, it cannot bridge the gap between observed instances and universal law.

Method must be supplemented by a prior inquiry: what the cognitive apparatus brings to experience — the transcendental conditions of any knowledge. Method is a refinement of activities whose preconditions philosophy must first uncover.

Reconceives the opposition as conjecture vs. test: reason generates bold hypotheses that go beyond available evidence, and experience — in the form of critical tests — provides the sole legitimate means of eliminating them. Neither alone generates scientific knowledge. The aspiration to a priori certainty is simply abandoned.

Historicizes both sides. The paradigm determines what counts as a legitimate observational test and what counts as reasonable inference — so neither "experience" nor "reason" is theory-neutral. The opposition between them is internal to a framework, not a neutral vantage point above all frameworks.

Attempts to rehabilitate reason in the form of rational appraisal — retrospective judgment that a programme is progressive or degenerating. This is reason exercised over time, accountable to the historical record of confirmed novel predictions, rather than operating a priori.

Neither reason nor experience, in any philosophically tidy form, accounts for the most creative episodes in the history of science. What drove those episodes was less tractable: the willingness to hold onto a framework in defiance of current evidence. Whether this is a form of practical reason or the dissolution of the reason-experience framework altogether remains genuinely open.

Makes the embrace of fallibility explicit and methodological. Corroboration is not certainty — it is the record of a conjecture's survival, which leaves it perpetually exposed to future refutation. Fallibilism is not a deficiency of science but its defining virtue: a science that risked nothing could not be a science.

Introduces a different dimension: not the fallibility of individual theories but the fallibility of the standards of assessment themselves. Each paradigm generates its own criteria of success, and those criteria change in revolutions. The framework within which fallibility is assessed is itself fallible.

Tries to recover a fallibilist but rational standard: theories, programmes, and even the retrospective judgment that a programme was progressive are all subject to revision. But the progressive-degenerating criterion itself is presented as more stable than any particular scientific theory.

Extends fallibilism to the standards of appraisal themselves, including Lakatos's. No methodological criterion has proven immune to historical refutation. Fallibilism has consumed not only the content of science but the meta-level criteria by which scientific content is assessed — leaving a radical intellectual humility that is difficult to make practically useful.

Redirects the tension: objectivity belongs not to individual knowers but to the critical community. Scientific knowledge is objective because claims are subjected to public, intersubjective testing. Objectivity is a social and institutional achievement, not a feature of any individual's cognitive apparatus.

Complicates the community account: the scientific community is not a neutral testing ground but a tradition-constituted body sharing a paradigm. In revolutionary periods the community splits, with adherents of competing paradigms arguing from incommensurable starting points. The knowing community and nature are no longer transparent to each other even when inquiry is communal.

Makes the community's judgments explicitly evaluable through rational reconstruction. The individual knower and the community are assessed together against the historical record — but the standard of assessment remains external to the community being assessed.

Introduces the political dimension: the scientific community is a social institution with power over other institutions. The question of the knowing community's relationship to nature is inseparable from the question of science's relationship to the societies it operates in. The individual knower — the citizen told to defer to scientific authority — reappears as a political actor whose epistemic autonomy is at stake.

Method must guarantee testability, not truth. A method that produces only unfalsifiable claims is no scientific method at all. What falsificationism guarantees is not the truth of surviving theories but the scientific character of the enterprise — that the claims being made are ones that could, in principle, be shown to be wrong.

Method cannot guarantee the conditions for its own application. Without a paradigm specifying what a legitimate test looks like, falsificationism has no purchase. What normal scientific method guarantees is not paradigm-falsifiability but the productivity of puzzle-solving within an established framework.

Splits method into two functions: real-time guidance of individual decisions (which no method can reliably provide) and retrospective rational reconstruction of programmes (which his methodology can provide). Method now guarantees intelligibility — the capacity to show that the history of science is largely rational — rather than certainty or testability.

Method guarantees nothing — not truth, not progress, not even the retrospective intelligibility of history. Every proposed methodological norm has been productively violated by the scientists we most admire. Method is the institutional formalization of past successes — a conservative force masquerading as a rational criterion. Whether this means method is dispensable or that we need a richer account of methodological reasoning remains, at the close of this arc, genuinely and productively unresolved.