- Atlas
- Philosophy
- Confirmation And Evidence
Confirmation And Evidence
Confirmation and evidence is the part of philosophy of science that asks what it is for a piece of information to support a scientific claim.
9 frameworks, from 1620 to today · Philosophy Of Science · Philosophy
Overview
Confirmation and evidence is the part of philosophy of science that asks what it is for a piece of information to support a scientific claim12. When a physicist says a new particle's existence is confirmed by a bump in a detector, or an epidemiologist says a drug trial provides evidence that a treatment works, what exactly is being claimed? The questions are old—how do we know, and when does observation justify belief?—but the modern field has a distinctive shape: it tries to state the logic of evidential support precisely enough to test it against real and imagined cases. The stakes are practical. If we can say what makes evidence good, we can distinguish science from pseudoscience, defend a verdict against a charge of wishful thinking, and understand why some experiments are worth doing and others are not.
From induction to the logic of confirmation
The earliest systematic treatments were theories of induction. Francis Bacon's Novum Organum (1620) argued that knowledge should be built by gradual ascent from observation to general axioms, with "experiments of light" designed to reveal causes rather than merely produce useful effects. John Stuart Mill's A System of Logic (1843) gave the classic statement of the methods of agreement, difference, concomitant variation, and residues—rules for inferring causation from patterns of observed circumstances. These were qualitative canons: they told the investigator what to look for, not how much a given observation raised the probability of a hypothesis.
The decisive break came with the logical positivists in the 1920s and 1930s13. Rudolf Carnap, in Logical Foundations of Probability (1950), and Carl Hempel, in "Studies in the Logic of Confirmation" (1945), tried to construct a formal, quantitative theory of confirmation. The idea was to define a measure—a "confirmation function"—that would assign a degree of support to a hypothesis given a body of evidence, much as logic assigns truth values to arguments. Hempel's work produced a famous puzzle: the hypothesis "all ravens are black" is logically equivalent to "all non-black things are non-ravens," so a white shoe, being a non-black non-raven, should confirm it4. That conclusion seemed absurd, and the "raven paradox" became a standing challenge to any purely syntactic account of confirmation.
The logical confirmation theorists also confronted the problem of irrelevant evidence. If a hypothesis is confirmed by any observation consistent with it, then almost anything confirms almost anything. Carnap's answer was to build in a prior probability distribution over possible worlds and update it by conditionalization—the rule that the probability of a hypothesis given new evidence equals its prior probability times the likelihood of the evidence given the hypothesis, divided by the overall probability of the evidence. This is Bayes' theorem, and its use in confirmation theory became the foundation of the Bayesian approach56.
Falsificationism and the problem of induction
Karl Popper offered a different diagnosis17. In The Logic of Scientific Discovery (first published in German in 1934, English translation 1959), he argued that no amount of confirming evidence can establish a universal law, because the next observation might refute it. What distinguishes science, he claimed, is not that its claims are confirmed but that they are falsifiable: a scientific hypothesis must forbid something, and the scientist's job is to try to refute it7. A theory that survives severe testing is "corroborated," but corroboration is not proof8.
Popper's falsificationism had real influence on scientific practice—it encouraged the design of experiments that could fail, and it gave a sharp criterion for demarcating science from pseudo-science. But it faced difficulties. A single anomalous observation rarely refutes a theory outright, because theories are tested in bundles with auxiliary assumptions. And the history of science shows that theories often survive apparent refutations and are later vindicated. The falsificationist picture also struggles to explain why scientists should prefer a well-corroborated theory to a new one that has not yet been tested.
Underdetermination and the Duhem–Quine thesis
The problem of auxiliary assumptions was given its classic statement by Pierre Duhem in The Aim and Structure of Physical Theory (1906) and later extended by W. V. O. Quine in "Two Dogmas of Empiricism" (1951). The Duhem–Quine thesis holds that a hypothesis is never tested in isolation: what confronts observation is a whole network of theory, auxiliary hypotheses, and background assumptions910. When a prediction fails, logic alone does not tell us which part of the network to blame10. The thesis does not say that any hypothesis can be saved from any evidence—that would be too strong—but it does say that evidential support is holistic and that the choice of what to revise is underdetermined by the data1011.
This insight reshaped confirmation theory. It made clear that confirmation is not a simple relation between a hypothesis and an observation but a relation involving a body of background knowledge. It also raised the question of how scientists actually choose among competing revisions, a question that connects confirmation to the broader study of scientific change.
Hypothetico-deductivism and its limits
The hypothetico-deductive (H-D) model offered a simple answer to how evidence supports theory: propose a hypothesis, deduce observable consequences, and test them8. If the consequences hold, the hypothesis is confirmed; if they fail, it is disconfirmed8. The model was widely taught in the mid-twentieth century as a rough account of scientific method, and it captures something real about how experiments are designed.
But H-D confirmation has well-known problems. It confirms hypotheses too easily: any true observation that is deducible from a hypothesis confirms it, even if the observation would have occurred anyway. If a hypothesis entails that the sun will rise tomorrow, then tomorrow's sunrise confirms it—but so would almost any other hypothesis that entails the same thing. The model also gives no account of how strongly evidence supports a hypothesis, and it cannot distinguish between evidence that is surprising and evidence that is expected. These difficulties pushed philosophers toward probabilistic accounts.
Bayesian confirmation theory
Bayesian confirmation theory treats confirmation as a matter of probability12. The central idea is that evidence E confirms hypothesis H if E is more likely given H than given its negation—that is, if the likelihood ratio is greater than one. The degree of confirmation is then updated by Bayes' theorem512. This framework, developed in the mid-twentieth century by figures such as Harold Jeffreys, I. J. Good, and later Richard Jeffrey, Leonard Savage, and others, has become the dominant formal approach to confirmation.
The Bayesian approach handles many of the puzzles that defeated earlier theories. It explains why the white shoe does not confirm "all ravens are black" in any interesting sense: the probability of observing a white shoe is roughly the same whether or not all ravens are black, so the likelihood ratio is close to one. It explains why surprising evidence is more confirmatory than expected evidence: a result that is unlikely under the null hypothesis but likely under the alternative provides a strong likelihood ratio. And it gives a natural account of how background knowledge enters: the prior probability encodes what we already believe.
Bayesianism has its own difficulties. The choice of prior probabilities is not dictated by logic, and different priors can lead to different posterior beliefs even after the same evidence. This is the problem of prior sensitivity. There is also the problem of old evidence: if a hypothesis was formulated to explain evidence already known, the probability of that evidence given the hypothesis is one, so the likelihood ratio is one, and the evidence seems to provide no confirmation—a result that clashes with scientific practice. Various repairs have been proposed, including counterfactual priors and the idea that confirmation should be measured by the increase in probability rather than the likelihood ratio alone.
Inference to the best explanation
A different tradition, associated with Gilbert Harman's "The Inference to the Best Explanation" (1965) and developed by Peter Lipton in Inference to the Best Explanation (1991), holds that scientists often infer the truth of a hypothesis because it provides the best explanation of the evidence. This is not deduction and not straightforward probability: it is a form of abductive reasoning, where the criterion is explanatory virtue—scope, simplicity, unification, and the like1314.
The approach captures something that Bayesianism can seem to miss: the sense in which a theory is supported because it makes sense of the data in a coherent way. But it faces the challenge of specifying what counts as the "best" explanation without circularity. If explanatory virtues are themselves evidence of truth, why? And how do we compare explanations that score differently on different virtues? These questions remain live.
Error statistics and the frequentist alternative
A major rival to Bayesian confirmation theory is the error-statistical approach, developed most fully by Deborah Mayo in Error and the Growth of Experimental Knowledge (1996). On this view, evidence supports a hypothesis when the experimental procedure has a high probability of detecting errors in the hypothesis if they are present—that is, when the test is severe. Mayo's "severe testing" criterion is a refinement of Popperian ideas: a hypothesis is well supported if it has passed a test that would probably have found it false if it were false.
The error-statistical approach is closely tied to frequentist statistics, where probabilities are understood as long-run frequencies rather than degrees of belief. It emphasizes the design of experiments and the control of error rates, and it has been influential in the philosophy of statistics and in methodological debates in the social and biomedical sciences. Its critics argue that it struggles to give an account of evidence for singular hypotheses and that it does not capture the cumulative nature of confirmation as naturally as Bayesianism does.
Causal inference and the evidential turn
In recent decades, the philosophy of confirmation has increasingly engaged with causal inference, especially in the biomedical and social sciences. The work of Donald Rubin and Judea Pearl on causal models, and the "credibility revolution" in econometrics and epidemiology, has raised philosophical questions about what it means for a study to provide evidence for a causal claim. The randomized controlled trial became a gold standard, and philosophers have examined why: randomization ensures that treatment and control groups are comparable in expectation, so differences in outcomes can be attributed to the treatment. But randomization is not always possible or ethical, and observational studies require assumptions—no unmeasured confounding, correct model specification—that are difficult to verify.
This has led to a richer picture of evidence as something that comes in degrees and from multiple sources, and to the study of how different kinds of evidence can be combined. It has also connected confirmation theory to the social epistemology of science: how communities aggregate evidence, how peer review and replication function, and how evidential standards vary across fields.
The present
Today the field is plural. Bayesian confirmation theory remains the most developed formal framework, with active work on priors, old evidence, and the relationship between confirmation and explanation. Error-statistical and frequentist approaches continue to offer a serious alternative, especially in contexts where experimental design and error control are central. Inference to the best explanation remains a live option, particularly in the philosophy of the special sciences. Causal inference has brought new energy and new problems, as philosophers try to understand the evidential logic of modern empirical research.
The questions that remain open are the ones that have driven the field from the start: What makes evidence relevant? How should background knowledge be represented? Can there be a single logic of confirmation, or does evidential reasoning vary by domain? And how should we understand the relationship between confirmation and truth—between what the evidence supports and what is actually the case? These are not merely technical puzzles. They bear on how science is defended, how it is taught, and how its results are used.
References
- Philosophy of science, Wikipedia
- Empirical evidence, Wikipedia
- Logical positivism, Wikipedia
- Raven paradox, Wikipedia
- Bayesian inference, Wikipedia
- Bayes' theorem, Wikipedia
- Falsifiability, Wikipedia
- Hypothetico-deductive model, Wikipedia
- Confirmation holism, Wikipedia
- Duhem–Quine thesis, Wikipedia
- Underdetermination, Wikipedia
- Bayesian epistemology, Wikipedia
- Abductive reasoning, Wikipedia
- Inference, Wikipedia
The frameworks of Confirmation And Evidence
Each framework is a way of working the field has used: when it held, and the question it answers.
- 1620-1930
- 1906-Present
- 1930-1970
Hypothetico-deductive model
Does evidence confirm a hypothesis by its consequences?
About Hypothetico-deductive model
Hypothetico-Deductivism emerged as a dominant framework in philosophy of science during the mid-20th century, offering a direct answer to a pressing question: how can a hypothesis be tested when it cannot be directly observed? Instead of building confirmation from accumulated instances, as Inductivism did, or from logical probability, as Logical Empiricism attempted, Hypothetico-Deductivism placed the emphasis on deducing testable consequences from a hypothesis and then checking them against observation. Its core move was simple and powerful: a hypothesis is confirmed when its observable predictions turn out to be true, and disconfirmed when they are false. This switched the focus from finding evidence that fits a hypothesis to designing tests that could potentially falsify it.Part 1: Core CurriculumA first course on Hypothetico-Deductivism in confirmation and evidence would center on the following concepts, taught in logical sequence:- Hypothesis: The explanatory conjecture under test. It must be general enough to yield specific, testable predictions.
- Deduction: The logical process that derives a singular, observable prediction from the hypothesis, often in conjunction with initial or background conditions.
- Prediction: A specific, observable claim that follows deductively from the hypothesis. If the hypothesis is true, the prediction must hold.
- Observation: The act of checking the prediction against empirical evidence. Observations are the final arbiter of the test.
- Confirmation: When the observation matches the prediction, the hypothesis receives provisional support. It is not proven true, but it passes a test.
- Disconfirmation: When the observation contradicts the prediction, the hypothesis is refuted and must be modified or rejected.
- Auxiliary Hypotheses: Additional assumptions needed to derive a prediction. They play a crucial role, because a failed prediction may indicate a problem with an auxiliary hypothesis rather than the main hypothesis.
- Initial Conditions: The specific circumstances or assumed states that, together with the hypothesis, allow a prediction to be deduced.
- The Duhem-Quine Thesis: The recognition that it is impossible to test a hypothesis in complete isolation; a conjunction of hypotheses is actually tested, complicating straightforward disconfirmation.
- Modus Tollens: The formal logical structure underlying disconfirmation: if a hypothesis implies a prediction, and the prediction is false, the hypothesis is logically negated. This gives falsification its deductive force.
- Falsifiability: The requirement that a scientific hypothesis must be capable of being shown false by possible observations. A hypothesis that makes no risky predictions is not testable.
- Risky Prediction: A prediction that is unlikely to be true if the hypothesis is false. The riskier the prediction, the stronger the confirmation if it holds.
A representative exemplar might be: If Newton's law of gravitation is true, then the orbit of Uranus should match a computed path. When observed deviations could not be explained by known initial conditions, Auxiliary Hypotheses (the existence of Neptune) were introduced and later confirmed, preserving the core hypothesis.Part 2: Intellectual HistoryThe historical roots of Hypothetico-Deductivism lie in the work of William Whewell (1794–1866), who emphasized the role of hypothesis in science and the consilience of inductions. However, the framework became a dominant force in the mid-20th century, largely through the work of Karl Popper (1902–1994). His 1934 work Logik der Forschung (translated as The Logic of Scientific Discovery) laid out a comprehensive account of science as a hypothetico-deductive enterprise. Popper argued that science progresses not by confirming theories, but by subjecting them to severe tests and eliminating false ones. This falsificationist stance was the sharpest expression of the Hypothetico-Deductivist framework.Popper's core ideas were taken up and reformulated by other figures within the local dominance window (1930–1970). Carl Hempel (1905–1997), in his 1966 work Philosophy of Natural Science, systematized the hypothetico-deductive account of scientific explanation. Hempel's 'Deductive-Nomological Model' showed how scientific explanation could be understood as a deductive argument from laws and initial conditions. While Hempel was also a key Logical Empiricist, his later work helped make the hypothetico-deductive logic a standard part of the philosophy of science curriculum.The framework's local dominance coexisted with and was later challenged by other approaches. The Duhem-Quine Thesis, independently developed by Pierre Duhem (1861–1916) and W. V. O. Quine (1908–2000), pointed out a critical limitation: because the deduction of a prediction requires Auxiliary Hypotheses, a failed test logically refutes only the whole conjunction, not any specific hypothesis. This insight, canonically formulated by Quine in his 1951 paper 'Two Dogmas of Empiricism', complicated the clean falsification picture.As a successor and parallel framework, Bayesian Confirmation Theory (beginning around 1950) offered a more nuanced, probabilistic account of how evidence changes our confidence in hypotheses. Inference to the Best Explanation (starting around 1965) argued that scientists choose hypotheses based on explanatory virtues like simplicity and coherence, not just on pass/fail test results. Error Statistics (from 1980 onward) refined frequentist testing procedures. Hypothetico-Deductivism did not disappear but was largely absorbed into these richer frameworks. Its lasting legacy is the emphasis on testing via prediction, falsification as a scientific ideal, and the logical structure of scientific inference.Key figures within the local period: Karl Popper (1902–1994), who made falsifiability the demarcation criterion; Carl Hempel (1905–1997), who systematized the deductive-nomological model of explanation. Seminal works: The Logic of Scientific Discovery (1934/1959) by Popper; Philosophy of Natural Science (1966) by Hempel. The Duhem-Quine Thesis, while introduced by earlier thinkers, was sharply articulated by Quine in his 1951 paper and is an essential critical check on the naive version of the framework. Later codifiers and transmitters: Imre Lakatos (1922–1974), who attempted to rescue the project by treating research programmes with a 'hard core' of hypotheses protected by a 'protective belt' of Auxiliary Hypotheses. His work belongs to a post-1970 period where Hypothetico-Deductivism was being reconciled with its own difficulties.The 12 concepts its tower teaches
- Auxiliary Hypotheses
- Confirmation
- Deduction
- Disconfirmation
- Duhem-Quine Thesis
- Falsifiability
- Hypothesis
- Initial Conditions
- Modus Tollens
- Observation
- Prediction
- Risky Prediction
- 1934-1980
- 1945-1975
Logical confirmation theory
How much does evidence support a hypothesis?
- 1950-Present
Bayesian confirmation theory
How should belief change with new evidence?
About Bayesian confirmation theory
In the long search for a logic of evidential support, a central tension has been whether confirmation can be captured by a single formal rule or whether it inevitably involves judgment that no simple algorithm can capture. Bayesian Confirmation Theory offers a striking resolution: it treats confirmation as a matter of rational belief-updating, governed by the laws of probability. Instead of asking whether evidence absolutely proves or disproves a hypothesis, Bayesianism asks how much the evidence should change our degree of belief in that hypothesis should change. This shift—from all-or-nothing confirmation to graded, probabilistic degrees of belief—is the framework’s core move. It provides a unified, mathematically precise answer to Hume's problem of induction, while also addressing paradoxes that plagued its predecessors.Unlike the earlier Inductivism, which relied on simple enumeration of positive instances, or the Hypothetico-Deductivist, which demanded successful predictions, Bayesianism offers a formal calculus for weighing both prior plausibility and new evidence. It also avoids the quest of Logical Empiricism for a purely syntactic inductive logic, instead welcoming subjective prior probabilities as unavoidable elements of scientific reasoning. This makes Bayesian Confirmation Theory less a single algorithm and more a flexible framework for modeling rational inference.Part 1 of this article orients the reader to the core concepts that form the first teaching spine of this framework. Part 2 traces how these concepts emerged through canonical figures and works within the local dominance window of 1950-2026.Bayesian Confirmation Theory is built around a handful of interlocking concepts that any first course must master. In logical learning order, these are: (1) the probability calculus itself, with its axioms of additivity and normalization; (2) conditional probability, written as , which measures the probability of hypothesis given evidence ; (3) the notion of prior probability (), representing the plausibility of before seeing ; and (4) likelihood, , the prior, orlikelihood, or , which captures how probable the evidence would be if the hypothesis true. These combine into (5) Bayes' Theorem, the canonical formula for computing the posterior probability of after is observed:From these formal foundations flow several key distinctions. (6) The likelihood ratio compares the probability of the evidence under competing hypotheses, a common tool for weighing which hypothesis is better supported. (7) The difference between confirmation as the increase of belief is captured by the confirmation ordering: a piece of evidence confirms a hypothesis if > , and disconfirms if the inequality reverses. This leads to (8) measures of confirmation, such as the difference or ratio of posterior to prior odds, which quantify the amount of evidential support.A crucial non-mathematical concept is (9) subjective probability, the interpretation of probabilities as degrees of belief held by a rational agent, as opposed to objective chances. This opens the door to (10) prior probability as a legitimate, unavoidable, and often subjectively determined input to Bayesian inference. (11) Conditionalization is the rule by which a rational agent updates their degrees of belief after receiving new evidence, essentially treating Bayes' Theorem as a dynamic constraint on belief change. (12) The notion of coherence—that a set of degrees of belief must obey the probability axioms to avoid a Dutch Book—provides a normative justification for the framework. Finally, (13) Bayesianism's handling of the Raven Paradox shows how it dissolves the problem by showing that observing a non-black non-raven does confirm the hypothesis x ((x) (x). Together, these concepts form a complete toolkit for framing evidential reasoning.Part 2: Intellectual HistoryBayesian Confirmation Theory as a named, self-conscious framework within philosophy of science emerged in the mid-twentieth century, but its roots reach back to Thomas Bayes' 1763 essay and Pierre-Simon Laplace's massive work in the early 1800s. These precursors used the probability calculus to solve problems of inference, but they did not articulate a logic of confirmation specifically. The crucial step for the local subfield was taken by Rudolf Carnap in his 1950 book Logical Foundations of Probability. Carnap, a central figure in Logical Empiricism, sought to build a formal inductive logic that assigned a precise degree of confirmation to a hypothesis given a body of evidence, using the concept of logical probability. His work provided the key bridge from probability theory to the philosophy of science.However, Carnap's project was not yet Bayesian in the modern sense; he aimed for a purely logical, objective notion of confirmation, downplaying the role of subjective priors. The full Bayesian turn came with the work of L. J. (Leonard Jimmie) Savage in 1954, whose Foundations of Statistics provided a rigorous axiomatic foundation for subjective probability and decision theory. Savage showed that coherent degrees of belief, as measured by betting behavior, could be represented probabilistically and updated via Bayes' Theorem. This offered a powerful normative framework for any rational agent.Within philosophy of science, the seminal works that established Bayesian Confirmation Theory as a distinct subfield core are Richard Jeffrey's 1965 The Logic of Decision, which extended Bayesianism to cases where evidence itself is uncertain (Jeffrey conditionalization), and later essays by authors like Wesley Salmon and Clark Glymour, which debated the role of explanatory virtues. The framework's local dominance period runs from roughly 1950 to the present, but its mature codification came in the 1970s and 1980s, with works like Colin Howson and Peter Urbach's Scientific Reasoning: The Bayesian Approach (1989) providing a textbook synthesis. Key figures in this local period include Carnap (as precursor-transitioner), Savage (as axiomatizer), Jeffrey (as extensionist), and later Howson and Urbach as textbook codifiers.Bayesian Confirmation Theory did not replace its predecessor frameworks by refutation; instead, it absorbed and reinterpreted them. It offered a unified solution to the paradoxes of confirmation that bedeviled Logical Empiricism (the Raven Paradox, Goodman's New Riddle of Induction), a formal grounding for the intuitions of Inductivism (by showing how repeated positive instances raise posterior probability), and a more nuanced account of evidence than Hypothetico-Deductivism could provide. Its main successor framework, Inference to the Best Explanation (IBE), emerged around 1965 with Gilbert Harman and later Peter Lipton, and coexists alongside Bayesianism. Many philosophers of science now treat IBE as a pragmatic heuristic that Bayesianism can in principle model, while others see them as competing accounts of what justifies belief. Error Statistics, developed by Deborah Mayo in the 1980s, offers a frequentist alternative that criticizes Bayesianism's reliance on prior probabilities.In sum, Bayesian Confirmation Theory transformed confirmation and evidence from a search for a purely logical rule into a normative theory of rational belief change. It made the agent's subjective prior a legitimate, indeed unavoidable, component of scientific inference, and provided a flexible, powerful language for debating confirmation in real scientific contexts. The framework remains central to the subfield today, often intertwined with IBE and debated against error-statistical approaches.The 13 concepts its tower teaches
- Bayes' Theorem
- Bayesian Resolution of the Raven Paradox
- Coherence
- Conditional Probability
- Conditionalization
- Confirmation Ordering
- Dutch Book
- Likelihood
- Likelihood Ratio
- Measures of Confirmation
- Prior Probability
- Probability Calculus
- Subjective Probability
- 1965-Present
- 1990-Present
- 1996-Present
How the ideas connect
Each arrow runs from an earlier idea to the later one it shaped. The list below says what passed between them.
Bayesian confirmation theory grew out of Logical confirmation theory
Bayesian confirmation theory grew directly out of logical confirmation theory, adopting Carnap's use of conditionalization and prior probability distributions while generalizing the framework.
Carried by Rudolf Carnap.
Logical confirmation theory grew out of Inductivism
Logical confirmation theory grew out of the inductivist tradition, replacing qualitative canons of induction with formal, quantitative measures of evidential support. Carnap and Hempel sought to make precise the idea that observation confirms general hypotheses.
Carried by Rudolf Carnap; Carl Hempel.
Falsificationism influenced by Duhem–Quine thesis
Popper's falsificationism was shaped by the Duhem–Quine problem: he acknowledged that auxiliary assumptions complicate refutation and proposed methodological rules to decide which hypotheses to reject.
Carried by Karl Popper.
Logical confirmation theory reacted against Hypothetico-deductive model
Logical confirmation theorists rejected the hypothetico-deductive model's inability to quantify support or handle irrelevant evidence, seeking a formal probabilistic account instead.
Carried by Carl Hempel; Rudolf Carnap.
Bayesian confirmation theory reacted against Hypothetico-deductive model
Bayesian confirmation theory developed in response to the shortcomings of the hypothetico-deductive model, particularly its failure to distinguish strong from weak evidence and to handle surprising evidence.
Carried by Harold Jeffreys; I. J. Good.
Bayesian confirmation theory reacted against Falsificationism
Bayesian confirmation theory offered an alternative to Popper's falsificationism, arguing that evidence can support hypotheses probabilistically even if it cannot prove them, and that corroboration can be understood as probabilistic confirmation.
Carried by Harold Jeffreys.
Error statistics influenced by Falsificationism
Error statistics, especially Deborah Mayo's severe testing account, built on Popper's idea that hypotheses are tested by attempts to falsify them, formalizing severity as a probabilistic notion.
Carried by Deborah Mayo.
Abductive reasoning reacted against Logical confirmation theory
Inference to the best explanation was proposed as an alternative to formal confirmation theories, arguing that explanatory considerations, not just probabilistic relations, drive scientific inference.
Carried by Gilbert Harman.
Abductive reasoning reacted against Bayesian confirmation theory
Proponents of inference to the best explanation argued that Bayesianism cannot capture the role of explanatory virtues like simplicity and unification in theory choice.
Carried by Gilbert Harman; Peter Lipton.
Error statistics reacted against Bayesian confirmation theory
Error statistics arose as a criticism of Bayesian confirmation theory, arguing that it cannot account for the severity of tests and the role of error probabilities in experimental reasoning.
Carried by Deborah Mayo.
Abductive reasoning reacted against Inductivism
Abductive reasoning revived an inductive idea that hypotheses can be inferred from evidence, but shifted the focus from enumerative induction to explanatory virtue.
Carried by Gilbert Harman.
Causal inference influenced by Bayesian confirmation theory
Bayesian networks and probabilistic graphical models provided a framework for causal inference, linking probabilistic confirmation with causal structure.
Carried by Judea Pearl.
Beyond the field
Where Confirmation And Evidence's ideas came from in other fields, and what they led to there: the strongest 12 of 43 links.
Inductivism influenced Inductive Logic in Logic
Mill's inductive logic systematised the Baconian inductivist programme, taking over its aim of discovering causes by eliminative methods and recasting it as rules of experimental inquiry.
Carried by Francis Bacon; John Stuart Mill; A System of Logic (1843).
Hypothetico-deductive model influenced Hypothesis-testing theory of concept learning in Learning And Concept Formation
The hypothesis-testing account of concept learning borrowed the hypothetico-deductive schema from philosophy of science, treating the learner as generating and confirming hypotheses. Bruner's experiments applied that model to categorisation.
Carried by Jerome Bruner.
Falsificationism reacted against Logical positivism in Scientific Change And Theory Choice
Popper's falsificationism was formulated as a direct attack on the verifiability criterion of the Vienna Circle, replacing confirmation by verification with refutability as the mark of empirical science.
Carried by Karl Popper; Logik der Forschung (1934).
Logical confirmation theory replaced by Bayesian Confirmation Theory in Philosophy Of Science
Bayesian confirmation theory superseded Carnap's logical confirmation theory, retaining the probabilistic measure of support but replacing the logical prior with subjective degrees of belief.
Logical confirmation theory replaced by Bayesian confirmation theory in Philosophy Of Science
Bayesian confirmation theory replaced the Carnapian logical confirmation theory, keeping its probabilistic measure of evidential support but recasting it as subjective prior probability updated by Bayes' theorem.
Logical confirmation theory grew out of Logical positivism in Epistemology
Logical confirmation theory grew directly out of logical positivism: Carnap and Hempel turned the verificationist programme into a quantitative logic of confirmation, measuring the support evidence gives to hypotheses.
Carried by Rudolf Carnap; Carl Hempel.
Logical confirmation theory provoked Confirmation theory in Formal Epistemology
Confirmation theory arose largely as a response to Hempel's and Carnap's logical confirmation theory, replacing qualitative instance-confirmation and degree of confirmation with probabilistic and later Bayesian analyses.
Carried by Carl Hempel; Rudolf Carnap.
Bayesian confirmation theory influenced by Bayesian decision theory in Formal Epistemology
Bayesian confirmation theory took its formal core from Bayesian decision theory: the subjective-probability and conditionalisation apparatus of Ramsey, de Finetti and Savage was carried over to the confirmation of hypotheses.
Carried by Frank Ramsey; Bruno de Finetti; Leonard Savage.
Inductivism led to Mill's methods in Philosophy Of Science
Mill's methods are a refinement within the inductivist programme, specifying the canons by which inductive generalisations about causes are to be established. They grew out of the Baconian inductivist tradition as its systematic codification.
Carried by John Stuart Mill; A System of Logic (1843).
Inductivism provoked Whewellian colligation and consilience in Philosophy Of Science
Whewell's colligation and consilience were formulated against the Baconian and Millian inductivism that treated discovery as mechanical collection and generalisation from facts. He argued that induction requires an active colligating concept supplied by the mind, defining his position by rejecting the pure inductivist account.
Carried by William Whewell.
Inductivism reacted against Aristotelian demonstrative science in Epistemology
Bacon's Novum Organum defined the new inductive method against the Aristotelian demonstrative science of the schools, rejecting syllogistic derivation from first principles in favour of gradual ascent from particulars.
Carried by Francis Bacon; Novum Organum (1620).
Inductivism led to Confirmation theory in Formal Epistemology
Confirmation theory defined itself against the inductivism of Bacon and Mill, replacing simple enumerative induction with probabilistic accounts of how evidence supports hypotheses.
Carried by Francis Bacon; John Stuart Mill.

Confirmation And Evidence is a planet in Nousfarer
Its city has a tower for each of these 9 frameworks. Land, walk in, and climb one concept at a time.
Land on Confirmation And EvidenceRelated fields
Fields its ideas are linked to
- Philosophy Of Science 11 links
- Formal Epistemology 4 links
- Justification Theory 4 links
- Scientific Change And Theory Choice 3 links
- Epistemology 2 links
- Scientific Realism 2 links
- Learning And Concept Formation 1 link
- Philosophy Of History 1 link
- Evidence Based Medicine 1 link
- Epidemiology 1 link
- Nutritional Epidemiology 1 link
- Analytic Philosophy 1 link
More in Philosophy Of Science
- Models And Representation
- Philosophy Of Biology
- Philosophy Of Chemistry
- Philosophy Of Economics
- Philosophy Of Medicine
- Philosophy Of Physics
- Philosophy Of Science
- Philosophy Of Social Science
- Scientific Change And Theory Choice
- Scientific Explanation
- Scientific Realism
See where Confirmation And Evidence sits on the Map of Human Knowledge