Felix Kaufmann, “The Nature of Scientific Method” (1945)

Felix Kaufmann’s four-part journal article locates the unity of scientific inquiry in the rules governing warranted changes to established knowledge. Methodology must clarify these rules before it can classify particular techniques of investigation. Against accounts that make science a means of discovering invariant truths, Kaufmann emphasizes the revisability of every empirical proposition. His central question is therefore not how inquiry guarantees permanent truth, but what makes accepting or rejecting a proposition correct in a given scientific situation.

Section I distinguishes two meanings of “ground.” Deductive grounds establish what premises already entail; empirical grounds warrant incorporating genuinely new propositions into science or removing previously accepted ones. Kaufmann calls both kinds of alteration “scientific decisions.” Their correctness depends on procedural standards, much as legality depends on legal norms and grammatical correctness on linguistic conventions. These standards need not already possess explicit formulations: methodology reconstructs what competent investigators acknowledge in practice.

The task of making explicit the canons of scientific procedure actually observed by competent scientists is the chief task of methodology.

This formulation connects Kaufmann’s project with Dewey’s demand that logical theory accord with scientific practice. It also makes correctness relational without making it arbitrary. A fully specified judgment must identify both the relevant rules and the body of knowledge available at the time—the “scientific situation.” Verification means warranted incorporation; invalidation means warranted elimination; falsification additionally involves accepting an incompatible proposition. These are procedural achievements, not assurances against subsequent revision.

Section II defends this account against the objection that scientific innovators succeed by breaking established rules. Kaufmann separates the circumstances of discovery from the grounds of validation. A bold conjecture may eventually succeed without having been warranted when proposed; conversely, a warranted prediction may fail. Success cannot replace procedural appraisal, because identifying success already invokes standards of observational confirmation.

A prediction may be warranted and successful, warranted but unsuccessful, unwarranted but successful, unwarranted and unsuccessful.

The distinction protects recognition of scientific genius while denying that later confirmation retroactively establishes an earlier conjecture’s warrant. Kaufmann then complicates the observational test itself. Observation does not supply an independent tribunal of uninterpreted sensations:

It is impossible to isolate perceptual content from principles of form. Every perceptual experience involves processes of selection in conformity with certain general patterns.

Instruments make the involvement of theory conspicuous, but unaided perception has the same fundamental structure. Consequently, evidence acquires significance within a wider system of scientific controls. A few observations of the predicted bending of light could strongly corroborate Einstein’s theory, whereas many positive instances might inadequately support a hypothesis disconnected from the theoretical framework. Invalidation likewise cannot always be reduced to one negative instance; Kaufmann expressly includes social laws among those not treated this way. His discussion of social statistics and the failed Harvard business-cycle barometer illustrates the costs of assuming that observed facts uniquely determine laws.

Section III addresses the apparent rigidity of a rule-based account. Particular rules can change, but judging a change appropriate presupposes further standards. The legal analogy distinguishes appraisal under existing norms from appraisal of alterations to those norms. Kaufmann nevertheless argues that scientific method must retain identifying features:

The invariant principles of scientific procedure are not embedded in single rules; they are properties of the system of procedural rules.

These properties include reliance on established propositions as grounds, the key role of observational reports, reversibility of decisions, avoidance of incompatible accepted propositions, and the exclusion of propositions undecidable in principle. They are a priori as constitutive conditions of what counts as scientific procedure, not as truths deducible from meanings. Following Hume, Kaufmann denies that empirical procedure has an ultimate rational justification.

This denial motivates his treatment of social-scientific objectivity. Social interests may shape beliefs, but appeals to realities beyond possible experience cannot defeat relativism. Criticism must instead expose tensions between relativist claims and standards their proponents themselves adopt. Permanent revisability conflicts with treating socially conditioned beliefs as unalterable; formal procedural requirements do not vary simply because interests conflict. Mannheim exemplifies, for Kaufmann, a movement from ideological perspectivism toward the possibility of synthesis. Objectivity means validity under presupposed scientific canons, not escape from every standpoint. Different causal approaches can coexist; the unjustified claim to have identified the sole cause creates their apparent incompatibility.

Section IV distinguishes “basic rules,” which define correct scientific decisions, from “preference rules,” which guide promising approaches to particular problems. Investigators usually attend to the latter, but their intelligibility depends on the former. The article closes by extending methodological clarification into liberal education and social practice. Shared standards cannot resolve every conflict, yet they make argument something other than disguised coercion. Distinguishing prejudice from warranted belief requires understanding how validity is established. Kaufmann thus preserves the philosophical ideal of clarity while grounding intellectual responsibility in revisable empirical inquiry.