Popper's Logic of Scientific Discovery: Falsificationism
Karl Popper's account of how falsifiability distinguishes empirical scientific theories and how experiments can test the claims they make.
PhilosophyAll chapters Reading tools Notes & comments
Testing and falsification › What an Experiment Can SettleFull chapter · Sections in this chapter1 / 7 ↓
What an Experiment Can Settle
Imagine testing a claim that every valve of a certain design will open below a specified pressure under stated conditions. The first valve opens. So do the next five. A seventh stays shut when the pressure crosses the limit. You have two questions. Does an accurately described failure conflict with the original claim? And have you really established that the valve failed under the conditions the claim specifies? Karl Popper's The Logic of Scientific Discovery treats the first as a question about inference and the second as one about testing. Much of his argument concerns what happens when both questions have to be answered together. The Logic of Scientific Discovery, §§6–8, 22
The book appeared in German as Logik der Forschung, with a 1935 imprint. The first English edition appeared in 1959. Popper develops his account of empirical science by asking how claims can be exposed to tests, what evidence can do to a theory, and how experimenters decide that a result is good enough to use. In his account, testing need not begin by building a law out of the observations that were supposed to establish it. It can begin with a proposed law and the consequences that would follow if it held. The Logic of Scientific Discovery, publication page and §§1–3
Why Repeated Success Leaves a Question Open
When the first valves open as expected, you can report a series of tests. The report describes particular valves and particular runs. A law covering every valve of the design makes a wider claim. You cannot deduce it from the first few results. The premise that those valves opened is compatible with there being a valve elsewhere that fails. The difference would remain if you had tested hundreds of valves. A strictly universal claim about valves of the design reaches beyond any finite collection of cases you have observed. The Logic of Scientific Discovery, §§1, 13–15
Popper's problem of induction is what could justify the step from such reports to a universal law. It might seem that the step is warranted because relying on similar past results has often worked. But a conclusion about what reliance on such results will do in as yet unexamined cases again goes beyond the results you have. The appeal to earlier success uses the very kind of inference it is offered to justify. Popper turns the proposed order of reasoning around. A theory can be proposed first, have testable consequences drawn from it, and face experiment on that basis. He separates the logical appraisal of a proposal from the psychological question of how it occurred to the person who made it. The Logic of Scientific Discovery, §§1–3
The failed valve illustrates a different inference. Suppose you have accepted that it is a valve of the design in question, that the stipulated conditions were met, and that it failed to open at the required pressure. Those statements together are incompatible with the universal claim. They do not have to be representative of the valves in order to contradict the claim about every valve. A single accepted counterinstance does that logical work. The difference between drawing an inference from an accepted result and knowing that it is a result to accept is where the practical work begins. The Logic of Scientific Discovery, §6 · §22 teaching excerpt
Falsifiability and the Scope of the Theory
Popper proposes falsifiability as the criterion of demarcation for empirical science. A theory is falsifiable if it rules out some possible basic statements, which can figure as premises in an empirical test. The potential falsifying result need never occur. What matters for the criterion is that the theory has consequences specific enough for such a result to conflict with it. The Logic of Scientific Discovery, §§4, 6, 21
Falsifiability asks whether a theory excludes some testable outcomes. Falsification asks whether a result warrants taking a particular theory to have failed. On Popper's account, the latter involves accepting conflicting basic statements and establishing a reproducible effect. The Logic of Scientific Discovery, §§21–22, teaching excerpt
The same observations can bear differently on claims with different scopes. A valve that opens at 60 units of pressure is compatible with a rule that valves of the design open by 100. It is also compatible with a rule that they open by 75. A valve that opens at 90 distinguishes those rules. If everything else is held fixed, the more precise requirement to open by 75 excludes some possibilities left open by the other. It exposes the claim to tests that the less demanding requirement can survive. Popper gives such comparisons an explicit place in his account of degrees of testability. The comparison here depends on keeping the class of valves and the rest of the test conditions the same. The Logic of Scientific Discovery, §§31–36
For Popper this requirement marks out what a claim has to do to enter a particular kind of inquiry. It does not measure the importance of the question the claim concerns. A proposal can guide people toward a new investigation before they develop a way to test it. A scientist might take guidance from a general picture about how nature hangs together, while the claims drawn from that picture have to meet the demands of testing separately. The demarcation distinguishes roles within inquiry and does not determine their exact value in advance. The Logic of Scientific Discovery, §4
The scope of a claim matters in the other direction as well. Compare “there is a valve that has failed somewhere” with “there is a valve that has failed on this test bench today.” A sufficiently careful inspection of the bench could count against the second statement. It cannot settle the first if the search has no definite limit. Popper makes this difference with statements about ravens. The claim that all ravens are black has an incompatible case, a raven that is not black. The existence of a black raven is established by a genuine instance. The absence of one in the next place you look leaves the open-ended existential claim untouched. The structure of a statement tells you which kinds of observations could conflict with it, as well as which kinds could support it. The Logic of Scientific Discovery, §§13–15
You can also alter the scope of the valve rule by limiting it to the five devices on your bench. If you check all five under the specified conditions, you have done work that can settle a bounded claim about those devices for the conditions you checked. Popper distinguishes statements of this kind from a law with no such limit. Counting to the end of a list may put you in a position to say something about everything on it. It gives you no way of counting to the end of all the possible instances the strictly universal law covers. The Logic of Scientific Discovery, §§13–15
How a Result Earns a Place in a Test
A gauge that jumps suddenly might give you your first reason to think the valve stayed closed above the pressure limit. It could also be a bad gauge, an indication that the pressure was applied in an unanticipated way, or a sign that this valve is different from the model named in the claim. To continue, you would want to know whether you could describe what happened well enough for somebody else to check it. A statement that identifies the apparatus, relevant conditions, and observed result gives other people a point at which to challenge or extend the test. The Logic of Scientific Discovery, §§7–8, 22
Popper calls certain singular test statements basic statements. A basic statement may function as a premise from which a theoretical claim can be challenged. Its role as a premise does not give it permanent authority. Other people can dispute which gauge was used, repeat the test, or ask whether the apparent failure can be checked by a different measurement. In §22 Popper says that accepting some basic statements incompatible with a theory is necessary for regarding it as falsified. He also asks for a reproducible effect described by a testable, corroborated falsifying hypothesis. An unusual reading that cannot be reproduced can prompt a search without settling how that search will end. The Logic of Scientific Discovery, §§22, 25–30 · §22 teaching excerpt
This leaves a practical issue. If every statement in the test can itself be questioned, do people have to settle all such questions before they can do anything with the result? Popper says that a particular test stops where investigators decide their accepted statements are sufficient for its purposes. They can later ask more of them. He compares the situation to supporting a structure with piles driven into a swamp. The piles can support the building without having encountered an absolutely firm bottom. Finding a good enough point to stop is part of conducting a test, and no decision about it needs to preclude further testing. The Logic of Scientific Discovery, §30 and 1972 addendum
A result can therefore bear different weights at different stages of inquiry. Suppose the first run gives a puzzling pressure reading. A second lab measures the valve travel using a separate device and reproduces the effect. That tells you something more than the second lab's willingness to trust the first result. It gives the proposed description a chance to fail in a new setting. It might also show where the failure can be made to appear reliably. Working out which conditions produce the effect is already a small empirical inquiry of its own. The Logic of Scientific Discovery, §§22, 29–30
Changing a Theory After a Failure
An experiment usually draws on more than one claim. The stated law about the valves leads to a prediction together with assumptions about the equipment and this particular specimen. A failed prediction challenges those claims taken together. You can ask which one should be reconsidered. Redoing the measurements after replacing an incorrectly calibrated gauge can be a step forward. Declaring that a valve stays within the law because any specimen that fails is by definition of a different design would defeat the test unless there were an independently checkable way to identify that difference. The Logic of Scientific Discovery, §§19–20
Popper calls attention to ways a theory can be protected by changes that prevent a possible failure from ever counting. His rule for auxiliary hypotheses that change the theoretical system asks them to increase testability. A revised account could predict that failures occur only after a specified contaminant has entered the valve, for example. Tests could check for contamination, deliberately introduce it, and see whether valves fail in the location and manner the revision predicts. If no possible finding about contamination could count against the revision, the adjustment would perform a very different role. The comparison shows why changing an assumption in response to a failed prediction is insufficient by itself to tell us whether a theory has been protected from test. The Logic of Scientific Discovery, §§19–20
Even an initially reasonable adjustment can lose its force. The effect could remain when a second measurement shows the pressure to be correct and when inspection excludes the proposed contaminant. At that point investigators have learned something about how the original law and the proposed explanations fall short. They would need to formulate a new proposal or specify a narrower domain in which the original claim can still be defended. Stating that domain clearly can make the remaining commitments easier to check. What matters to Popper is whether the next proposal makes claims whose failure can be recognized. The Logic of Scientific Discovery, §§19–22
There is also a difference between revising a general law and checking whether the setup used in one run matches its description. A record that the gauge was defective on Tuesday can be a testable singular claim. It need not serve as a new law about the conditions under which every valve will operate. Popper qualifies his treatment of auxiliary hypotheses to leave room for such cases. Taking that qualification seriously lets the investigator test an explanation of one run while keeping sight of the broader claim that prompted the experiment. The Logic of Scientific Discovery, §20
The Point of Looking for a Failure
Popper uses corroboration to describe how a theory has stood up to testing. To see what that can mean, compare testing a valve well below the range in which its opening behavior is disputed with testing under conditions where different accounts give different predictions. In the second case, a rival result had a real chance to occur. The tests a theory has survived can enter an assessment of it, but that assessment does not give a deductive warrant for extending the result to every future case. It tells us what the theory has had to withstand so far. The Logic of Scientific Discovery, ch. 10 · “Science as Falsification,” companion excerpt
Popper illustrates the difference in a later essay by describing the 1919 eclipse observations associated with Arthur Eddington. A calculation from Einstein's theory of gravitation specified a displacement in the apparent positions of stars whose light passed close to the Sun. An eclipse offered a way to observe the relevant stars. Popper's interest is the possibility of a definite result that did not show the predicted effect. Even in his telling, asking what follows if the predicted effect is absent leaves the experimental question of whether the observations establish its absence. The episode has a clear job in the example. It shows the relation he wants between a theory, a stated prediction, and conditions under which getting the prediction wrong would be recognizable. “Science as Falsification,” Einstein and the eclipse example
One way to understand this achievement is to compare two accounts of the same experiment. The first says that the valve behaved normally. The second describes what was expected, specifies how a different outcome would have appeared, and gives enough evidence to determine which occurred. The second account can help an investigator with a different theory say exactly where their prediction would have departed. That makes the result available for later criticism and further work. A report that shows why it was possible to be wrong can teach more than a collection of reports assembled because they all agree. “Science as Falsification,” companion excerpt on risky predictions and testability
Return to the valve that stayed closed. One investigator thinks the gauge caused the problem. Another thinks the valve design has a weakness that has just become visible. It takes a new test to separate the accounts. Making that test possible can improve the inquiry even when nobody is certain what it will show. In Popper's view, a theory is useful in part for specifying what its defenders are willing to take as a challenge. That can give people who disagree about it a piece of work they can perform together. The Logic of Scientific Discovery, §§6–8, 20–22
Further Reading
- Karl Popper, The Logic of Scientific Discovery. Start with §§1, 4, and 6, then compare §§19–22 and 28–30 for how tests are conducted and provisionally assessed.
- Karl Popper, selections from The Logic of Scientific Discovery. This reading includes the distinction between falsifiability and falsification.
- Karl Popper, “Science as Falsification”, excerpted from Conjectures and Refutations. This separate work illustrates severe testing with examples of risky prediction.
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!