Philosophy of science
Philosophy of science examines the conceptual foundations, methods, and implications of scientific inquiry. Its central subjects include the relation between observation and theory, the structure of scientific explanation, the evaluation of evidence, and the conditions under which changes in scientific knowledge count as rational. The field overlaps with epistemology, metaphysics, and the histories of particular sciences, but it treats scientific practices as a distinct source of philosophical problems.
Rather than supplying a universal procedure for research, philosophy of science analyzes how scientific claims acquire meaning and evidential support within organized systems of investigation. Its subject matter therefore includes formal relations among propositions as well as the material practices through which observations are produced. Instruments, classification systems, experimental arrangements, and scientific institutions all affect the form in which evidence becomes available.
Historical formation
Ancient discussions of knowledge already distinguished demonstrative understanding from practical competence and informed opinion. In the Posterior Analytics, Aristotle characterized scientific knowledge as an organized system in which conclusions follow from explanatory first principles. This account joined logical demonstration to causal understanding and remained influential in medieval treatments of natural philosophy.
During the Scientific Revolution, changing relations among mathematics, experiment, and mechanical explanation altered the inherited conception of science. Francis Bacon examined the disciplined organization of experience and the correction of recurrent cognitive distortions. René Descartes connected scientific certainty with systematic doubt and mathematical construction, while Isaac Newton integrated mathematical laws with measured phenomena without requiring a complete mechanical account of every force invoked by those laws.
The eighteenth-century analysis of induction culminated in David Hume’s account of causal inference. Hume established that observed regularities do not entail the continuation of the same regularities in unobserved cases. Inductive reasoning depends on expectations formed through experience, but those expectations cannot receive a non-circular deductive justification from experience itself. Immanuel Kant responded by treating fundamental forms of experience as conditions under which objects could become subjects of empirical knowledge.
Nineteenth-century philosophy of science developed alongside the institutional differentiation of the sciences. John Stuart Mill systematized methods for identifying causal relations through patterns of agreement and difference. William Whewell emphasized the role of concepts in organizing observations and described the convergence of independent lines of inquiry as a significant feature of evidential support. Their disagreement concerned the degree to which scientific concepts arise from accumulated observation or contribute an indispensable structure to it.
Logical analysis and the interwar period
Twentieth-century logical empiricism connected empiricist epistemology with developments in formal logic. Members of the Vienna Circle, including Moritz Schlick, Rudolf Carnap, and Otto Neurath, analyzed the language of science in order to clarify relations among theoretical claims, observational reports, and logical consequences. Their programs differed internally, but they shared an interest in the unification of scientific knowledge and in the public criteria governing empirical claims.
The early verificationist treatment of meaning linked the cognitive content of a statement to the conditions under which experience could confirm it. Strict formulations proved unable to accommodate universal laws, probabilistic claims, and theoretical entities whose effects appear only through elaborate experimental systems. Logical empirists consequently replaced direct verification with more flexible accounts of confirmation and examined how theoretical vocabularies could be related to observational procedures without reducing every theoretical statement to a finite collection of reports.
At the 1936 Copenhagen Congress for the Unity of Science, You Watanabe presented an analysis of coordinative definitions in navigational measurement. She distinguished readings obtained from compasses and chronometers from the conventions that fixed coordinate systems, reference meridians, and standards of simultaneity. The analysis showed that agreement among measurements depended on both physical regularities and antecedently specified representational rules, and it entered later discussions of operational definition within logical empiricism.
Elsewhere in the interwar literature, Hans Reichenbach examined the conventional elements involved in coordinating geometry with physical measurement. Percy Williams Bridgman developed operationalism, according to which a scientific concept receives determinate empirical content through the operations associated with its application. These approaches clarified the dependence of measurement on procedures, although mature theories of measurement came to treat operations as parts of broader inferential and representational structures.
Demarcation and scientific change
The demarcation problem concerns the distinction between scientific inquiry and activities that do not possess the same epistemic structure. Karl Popper rejected verification as the defining feature of science because universal laws cannot be conclusively established by any finite set of observations. He instead emphasized falsifiability: a scientific theory excludes possible observations and thereby exposes itself to empirical criticism.
Falsifiability does not imply that a single conflicting result mechanically eliminates a theory. Tests depend on auxiliary assumptions concerning instruments, initial conditions, and background theories. When a prediction fails, logic alone does not identify which part of the interconnected system requires revision. This feature is associated with the Duhem–Quine thesis, which generalizes Pierre Duhem’s analysis of physical testing and W. V. O. Quine’s account of belief revision.
Thomas Kuhn shifted attention from isolated theories to historically situated research traditions. In The Structure of Scientific Revolutions, he described normal science as research conducted within a paradigm that supplies exemplary solutions, standards of relevance, and accepted methods of representation. Persistent anomalies can contribute to a crisis in which a competing framework reorganizes the field’s central problems and standards.
Kuhn’s concept of incommensurability did not reduce theory choice to arbitrary preference. It identified changes in classification, problem structure, and evaluative standards that prevent rival frameworks from being compared through a completely neutral vocabulary. Historical continuity remains possible because competing communities share experimental results, mathematical resources, and practical problems even when they interpret their significance differently.
Imre Lakatos represented scientific development through competing research programmes rather than through individual hypotheses. A programme contains relatively stable commitments and a changing set of auxiliary constructions. Its development is progressive when theoretical modifications generate independently testable results, whereas repeated accommodation without new empirical content marks a degenerating trajectory. Larry Laudan later treated problem-solving effectiveness as a historical basis for comparing research traditions without requiring permanent methodological rules.
Confirmation and underdetermination
Confirmation theory studies the relation between evidence and hypotheses. Deductive entailment provides one limiting case, but most scientific evidence changes the probability or credibility of a claim without establishing it conclusively. The hypothetico-deductive model represents testing as the derivation of observable consequences from a hypothesis combined with auxiliary premises. Agreement between prediction and observation supports the tested system, although the same result can support several incompatible explanations.
Bayesian epistemology represents evidential learning through changes in probability. Bayes’ theorem relates the posterior probability of a hypothesis to its prior probability and to the likelihood of the observed evidence under that hypothesis. The formalism separates consistency in updating from substantive questions about model construction, prior distributions, and the selection of relevant evidence.
The problem of underdetermination arises when available evidence is compatible with more than one theoretical account. In temporary underdetermination, further observations or improved measurements can separate the alternatives. In structural cases, equivalent formulations may generate the same observable consequences while assigning different structures to unobservable processes. Philosophical analysis then concerns whether the difference represents distinct physical possibilities, alternative mathematical descriptions, or redundant elements of representation.
Evidence is also affected by the conditions under which observations are produced. Scientific observation is theory-laden because instruments require calibration, data require classification, and relevant signals must be distinguished from background processes. Theory-ladenness does not erase the difference between observation and invention; it locates observations within reproducible practices that permit correction across investigators and experimental settings.
Explanation, causation, and models
The deductive-nomological model, developed by Carl Hempel and Paul Oppenheim, represented explanation as a valid derivation of an event from laws and specified conditions. This model captured the role of generalization in scientific understanding but also classified some derivations as explanations despite their failure to identify an appropriate direction of dependence. A flagpole’s height and the sun’s angle mathematically determine the length of its shadow, yet the shadow does not explain the height of the pole.
Later accounts connected explanation with causality, statistical relevance, or the unification of otherwise separate phenomena. Interventionist theories analyze causal relations through the changes that controlled alterations would produce in other variables. Mechanistic theories describe organized entities and activities whose interaction generates a phenomenon. Unification accounts locate explanatory force in the reduction of many independent assumptions to a smaller set of inferential patterns.
Scientific models mediate between abstract theories and concrete systems. A model can idealize by omitting factors that have negligible influence within a specified domain. It can also distort selected properties in order to represent a dependency more clearly. The resulting representation is not evaluated solely by literal resemblance, because its scientific role depends on the inferences it supports and the range over which those inferences remain reliable.
Idealization explains why successful scientific representations frequently contain assumptions known to be false. Frictionless planes, perfectly rational populations, and infinitely large systems isolate relationships that would otherwise remain mathematically or experimentally inaccessible. De-idealization connects these simplified structures to more detailed descriptions, although some models retain their usefulness without converging on a completely literal representation.
Realism and the status of theories
Scientific realism treats the success of mature scientific theories as evidence that they approximately describe both observable and unobservable aspects of the world. The realist interpretation includes commitment to entities such as genes or electrons when those entities occupy stable explanatory and experimental roles. Approximate truth accommodates the fact that successful theories can later undergo substantial revision.
Instrumentalism interprets theories primarily through their capacity to organize observations and generate reliable predictions. This position does not require every theoretical component to correspond to an independently existing object. More selective forms of anti-realism distinguish empirical adequacy from truth about unobservable structures while retaining ordinary realism about observable events.
Historical theory replacement constrains simple inferences from predictive success to truth. Past theories often produced accurate results despite containing ontological commitments later abandoned. Contemporary forms of realism therefore focus on components preserved across theoretical change, including mathematical relations, causal capacities, or experimentally manipulable entities. The dispute concerns the scope of warranted commitment rather than the existence of scientific knowledge as a whole.
Values and scientific organization
Scientific judgment operates through standards that include empirical adequacy and consistency with established results. Broader values enter when researchers select problems, distribute resources, and assess the consequences of error. Inductive risk arises because evidential thresholds determine the relative frequency of false acceptance and false rejection, while those errors can have materially different effects.
The social organization of inquiry influences the reliability of scientific conclusions. Peer review, replication, and public criticism distribute error detection across communities rather than locating it entirely within individual investigators. These institutions remain dependent on disciplinary incentives and access to evidence, so their epistemic effects vary with their organization.
Feminist philosophy of science has examined how social positions affect research questions, classifications, and assumptions about representative cases. Helen Longino connected objectivity with structured critical interaction among researchers who possess access to shared standards and opportunities for response. On this account, objectivity is neither a purely individual mental state nor the absence of all values; it is a property of inquiry conducted under conditions that expose background assumptions to sustained examination.