Feyerabend's Against Method: Epistemological Anarchism
Paul Feyerabend's 1975 challenge to a universal scientific method, using Galileo and other historical cases to show why rival approaches matter.
PhilosophyAll chapters Reading tools Notes & comments
The limits of a universal scientific method › What a Good Rule Can MissFull chapter · Sections in this chapter1 / 7 ↓
What a Good Rule Can Miss
Imagine being told that a new scientific theory must agree with everything well established in its field before anyone takes it seriously. That rule saves people from having to reconsider dozens of familiar observations every time someone proposes an explanation. Now think what happens if some of those observations were described in terms that only make sense within the older theory. The rule could stop a rival from being developed far enough to show which parts of what seemed established should be taken up again. The practice of checking a proposal against the evidence is still indispensable. The question is when a proposal has to pass a particular check and how to recognize a genuine disagreement with the evidence.
Paul Feyerabend takes up that question in Against Method, first published in 1975. The book follows cases in which changing the theory also changed how people could describe what they saw, which instruments they trusted, and what they took an experiment to show. It draws a broad conclusion about proposals for a single scientific method. A rule defended by appealing to the success of science has to answer to how that success was achieved. The history of a practice can show that some of its achievements depended on departures from what the rule would have required. Against Method, first-edition excerpts · Third edition, introduction and ch. 1
In the third-edition preface, Feyerabend says the phrase “anything goes” is an ironic response by someone looking for a rule that survives close contact with history. It is important to locate the remark there. He is analyzing the demand for principles that apply independently of the situation in which scientists have to make a judgment. The book invites a reader to look for the work particular departures from method performed. Quoting the phrase by itself supplies little help with that investigation. Against Method, third-edition preface
Why Keep an Alternative Around?
The consistency condition Feyerabend discusses asks that a new hypothesis agree with a theory already accepted in the field. An established theory may have solved real problems. There can be costs to reopening decisions that enabled that work. But the condition gives a special advantage to the account that arrived first. If a rival theory has to agree before people can develop it, an observation that would become intelligible on the rival account may never be made. For Feyerabend, judging the two by whether the newcomer can be expressed as an acceptable extension of the first leaves out one reason to have alternatives. Against Method, ch. 3
Consider a laboratory that treats the drift of an instrument as a nuisance and routinely subtracts it from the records. Someone proposes that under certain conditions the drift reveals a change in the object being studied. The new account has a problem. Some of the tidy results people are most proud of were calculated after the drift had already been removed. The group cannot evaluate the proposal merely by running it against the tidy series. The unadjusted records and an experiment that varies the conditions thought to produce the drift would be more informative. In that case the rival makes it possible to ask an empirical question that was hard to pose within the established treatment of the signal. The alternative can contribute to scrutiny even before anyone decides whether it is right.
Feyerabend's term counterinduction includes developing hypotheses that conflict with established theories or with observations as currently understood. Such a proposal can make assumptions visible that were taken for granted in setting up the contrast between theory and observation. Having brought them into view, a researcher can investigate what measured features or patterns each account would expect. Feyerabend also cautions against presenting counterinduction as a replacement method. Its value in a particular case has to do with how the rival proposal challenges something that the established approach leaves unquestioned. Against Method, chs. 2–3
Developing a proposal that conflicts with an established theory or with what is taken to be a fact can help uncover the assumptions by which the facts were identified. Feyerabend uses counterinduction as a way to investigate those assumptions in concrete cases. Against Method, ch. 2
There is a limit to what the laboratory illustration establishes. Its familiar way of handling drift may turn out to be sound. The alternative may fail to identify any circumstances in which the drift varies as it predicts. Working through that possibility can still matter because it gives a basis for deciding what the correction does and when to use it. If the alternative only repeats that the researchers might be wrong, it has yet to offer the comparison that would bring a hidden assumption into reach. Developing a rival requires specifying enough of it to say where the disagreement is. Against Method, chs. 2–3
Feyerabend also uses Brownian motion to give the argument a different scientific setting. The small, irregular movements of particles suspended in a fluid needed more than a report that motion was visible. Within the kinetic theory of heat, a calculation could connect features of the movement to a model of the medium and make definite statistical results worth investigating. In Feyerabend's telling, that work changed what experiments on the motion could show about the older phenomenological account. He doubts that investigators could have reached the same result just by trying to measure every slight change in the particle's motion and the surrounding transfer of heat directly. Against Method, ch. 3, discussion of Brownian motion
There are several questions one could press about this history. One is whether a scientist using only the established framework might eventually have decided the motions deserved attention. Feyerabend treats that as uncertain. Another is what the kinetic theory gave people a reason to measure that was previously hard to connect to the older discussion. The example is more specific on that point. It asks what extra evidence becomes available when a theory tells investigators how to compare many observations, instead of requiring them to read the relevant point directly from a record of any one movement. Against Method, ch. 3
The Stone Dropped from a Tower
Feyerabend considers how Galileo treats an argument against the motion of the Earth. Drop a stone from a tower. It lands near the bottom of the tower. If the Earth were moving, an opponent might ask, why does the base of the tower not move a noticeable distance away before the stone lands? The stone's path seems to favor a stationary Earth when the observer takes the apparent vertical motion to tell them how the object moves without further qualification. Against Method, ch. 6, tower argument
In Feyerabend's reading, the force of the objection depends on a natural interpretation. The observer is accustomed to describing a seen movement as the movement of the thing itself, without saying what it is being measured relative to. That use of language appears straightforward until someone asks about a motion the observer, tower, and stone already share. Galileo's defense allows an observer at the tower to describe the stone as falling vertically relative to them while the stone also participates in a common motion with the Earth. The description changes what counts as a contradiction between what was seen and what the Copernican proposal says. Against Method, chs. 6–7
A passenger can drop a key on a smoothly moving vehicle and catch it below the point where it was first held. Looking at the path within the vehicle gives one useful description. A person at the side of the road is in a position to describe a different path relative to the road. The example does not require a judgment about every aspect of the Earth's motion. It helps separate the description of a path in a familiar setting from what would be needed to rule out a background motion shared by the people making that description. The ordinary wording can continue to serve while its scope becomes clearer. Against Method, chs. 6–7, movement and observation
This also makes the argument an example of criticising evidence without discarding an observation. The sight of the stone landing near the tower had been taken to exclude a moving Earth. Trying out the rival picture reveals an assumption needed for that inference. On the new description, the result of the drop still needs explaining. What changes is the set of motions with which the result is compatible. For Feyerabend, someone who required the Copernican account to agree with the accepted description before allowing it to be developed could block the very work by which that compatibility becomes visible. Against Method, ch. 6, inversion of the tower argument
Learning to Look Through an Instrument
The discussion of Galileo's telescope raises a related problem. A telescope that seems faithful when pointed at a tower or an inscription can produce disputed appearances when pointed at stars. A terrestrial mark can be inspected from nearby to check whether the instrument shows what is there. The observer does not have that kind of access to a distant celestial object. Feyerabend cites reports in which people who had looked through an instrument did not agree on whether they saw the features Galileo wanted them to see. He analyzes how the use of the instrument earned a place in the inquiry amid those uncertainties. Against Method, chs. 8–9
The difficulty is more specific than asking whether any instrument could be trusted. An optical device can have characteristic effects in particular conditions. Judging a claim about what it shows involves working out whether a feature recurs with the object, with the arrangement of the lenses, or with the circumstances in which somebody looked. People can improve an instrument and the ways to check its results while also learning to recognize new objects with it. Requiring all that work to be finished before facing a new astronomical question can leave no point at which an instrument that first has uncertain standing begins to be useful. Against Method, chs. 8–10
The choice of examples matters here. Feyerabend offers a reading of Galileo's argument and the reports he selects, and the account of a particular observer's reasoning is open to historical disagreement. To assess a case, one would have to ask how the instrument was actually used, which of its effects could be checked at the time, and which supporting explanations became available later. The argumentative point concerns a possible mismatch between those stages and the method imposed on them retrospectively. Explaining why a test is accepted after its results become familiar can leave out the decisions that made the test possible in the first place. Against Method, chs. 8–10 and third-edition preface
How Much Work Should a Rival Get?
A reader could agree that a consistency rule privileges established theories and still ask why a particular rival deserves time. An invitation to reconsider a result is easy to issue. Repeating the work or developing a second instrument can be expensive. Feyerabend gives a reason to take such possibilities seriously when alternatives can make otherwise inaccessible assumptions easier to test. That is only one consideration in an actual decision. The questions of which alternatives to explore, how much to invest, and how long to wait can turn on what each proposal makes possible and the cost to other work. There is a difference between keeping open a means of criticism and knowing in advance how much effort it should receive. Against Method, chs. 2–4
Suppose the rival story about drift fails on the first experiment. A failure that can be characterized accurately may show the team what to change in its next proposal. A change that predicts a new pattern can be judged by what happens in a new run. A proposal whose defenders simply promise an explanation after every result will give the investigators much less to compare. Looking at what the alternative makes observable supplies a way to discuss the issue without expecting one fixed policy to settle every question about how long research should continue. The difference is visible only after somebody has said what the rival would do with the first failure.
A further issue is who gets to identify the alternatives worth pursuing. In the preface, Feyerabend links his argument about method to a defense of public control over scientific institutions in democratic societies. This is a political proposal with its own questions, including how people can weigh expert judgment, check the use of public funds, and participate in decisions about the purposes of research. Those questions arise even if all the relevant specialists agree about what result an experiment established. The evidence needed to settle an empirical disagreement differs from the reasons people have for assigning a scientific institution particular authority. Against Method, preface
A town choosing what to investigate about its water supply illustrates one distinction without needing to decide what Feyerabend's entire political proposal would require. People who know the instruments can judge which measurements are reliable and what they can show. The people who depend on the water can have reasons for asking about particular neighborhoods or about periods the first study excluded. Giving those questions due weight might change the scope of the investigation. It leaves the question of which measurements answer them to the work that is appropriate for settling that matter. The example makes clear why it matters to distinguish the standing of a result from the standing of an institution to decide what gets investigated.
What the History Tests
There are several possible responses to the tower case. A defender of a scientific method might allow for temporary conflicts between theories and observations. They could say that a theory and the tools for testing it sometimes have to develop together. That would bring the method closer to what the history shows, but would also make it harder to formulate a rule that tells researchers exactly when to insist on agreement. Feyerabend's criticism puts pressure on the details of the rule. If it is flexible enough to accommodate the work that generated the new evidence, a reader can ask what actual decisions it would have ruled out. Against Method, chs. 1–2, 6–10
Return to the person who drops a stone and watches it fall. They have learned something about the motion, and they still have to ask what description the result supports. Considering another account of movement helps someone see the difference between an observation that has been made and the role it has been assigned. The case gives the general argument of Against Method a place to start. To evaluate a rule by the history of inquiry, look at the occasion on which it would have been used, the alternatives the investigators had, and the further observations those alternatives taught them how to make. Against Method, chs. 2 and 6–7
Further Reading
- Paul Feyerabend, Against Method, third edition. Read the third-edition preface alongside Chapters 2 and 3, then turn to Chapters 6–10 for the stone and telescope cases.
- Paul Feyerabend, selected passages from the first edition. Compare the analytical contents and concluding material with the later preface.
Join the discussionAsk a question or share what you learned.
Keep exploring
Reference
Citation details
Cite or share this article.

Comments
Add to the conversation
A question, a correction, or an example from your own work.
No comments yet. Be the first to share your thoughts!