Scientific Explanation

Scientific explanation is the part of philosophy of science that asks what it is to explain something scientifically—why the kettle boils, why the tides follow the Moon, why a mutation causes a disease—and what makes one answer to such a "why" better than another.

10 frameworks, from 350 BCE to today · Philosophy Of Science · Philosophy

still a working view ended · each leads to its entry below

Overview

Scientific explanation is the part of philosophy of science that asks what it is to explain something scientifically—why the kettle boils, why the tides follow the Moon, why a mutation causes a disease—and what makes one answer to such a "why" better than another12. The stakes are not merely academic. Scientific explanation is what science offers in place of myth and bare prediction: a theory that forecasts eclipses but cannot say why they happen seems to be missing something. The history of the subfield is a long argument about what that "something" is—a law, a cause, a mechanism, a unification of many phenomena under few principles, or simply an answer to a question someone actually asked.

Aristotle's demonstrative science

The founding text is Aristotle's Posterior Analytics (fourth century BCE). Aristotle held that scientific understanding (epistēmē) consists in demonstrative knowledge: a body of propositions organized so that conclusions follow deductively from premises that are true, primary, and explanatory of the conclusion. The premises must state the cause—the "middle term" that makes the conclusion necessary. His stock example: knowing that the planets are near the Earth explains why they do not twinkle, because nearness is the reason for the absence of twinkling. Explanation and demonstration were thus fused: to explain was to show that a fact follows necessarily from its grounds. This ideal—a deductive structure anchored in causes—set the terms of debate for two millennia, and its descendants reappear whenever philosophers demand that explanation be an argument.

The covering-law model

The modern subfield begins with logical positivism and Carl Hempel13. In 1948 Hempel and Paul Oppenheim published "Studies in the Logic of Explanation," proposing what became the deductive-nomological (D-N) model: an explanation is a deductive argument whose premises include at least one law of nature plus statements of initial conditions, and whose conclusion is the event to be explained (the explanandum)12. Why does the mercury in a thermometer rise when heated? Because mercury expands when heated (a law), and this thermometer was heated (initial conditions); therefore it rose. The model's great virtue was precision: it made explanation a matter of logical form, checkable and comparable across sciences.

Hempel soon extended it to the inductive-statistical (I-S) model for probabilistic cases: an explanation makes the explanandum highly probable given laws and conditions, rather than deducing it. He also insisted on the symmetry thesis—an explanation and a prediction have the same logical structure, differing only in whether we know the outcome in advance2. This was a bold, unifying claim, and it is where the trouble started.

The failures of the covering-law model

Three counterexamples, all well known, broke the model's grip. First, irrelevance: from the law that all men who take birth-control pills fail to get pregnant, plus the fact that a particular man took the pills, one can deduce that he did not get pregnant—yet this explains nothing2. Second, the flagpole (due to Sylvain Bromberger, discussed by Wesley Salmon): from the height of a flagpole and the angle of the Sun one can deduce the length of its shadow, and from the shadow's length and the angle one can deduce the pole's height. Both are valid D-N arguments, but only the first explains; the second merely predicts. Third, the barometer: a falling barometer reading lets us predict a storm, but does not explain it—both are effects of low pressure. The common diagnosis: the D-N model captures inferential relations but not explanatory ones. Something about causes, or about which facts are the right ones to cite, was missing.

Causation returns: Salmon and the statistical-relevance model

Wesley Salmon, in Statistical Explanation and Statistical Relevance (1971) and later Scientific Explanation and the Causal Structure of the World (1984), argued that an explanation should cite statistically relevant factors—partition the reference class so that the probability of the explanandum changes—and, more deeply, that explanation is a matter of exhibiting the causal processes and interactions that produce the event. His mark-transmission account of causality (a causal process is one able to carry a mark, such as a scratch on a moving billiard ball) was an attempt to make causation respectable for empiricists. The flagpole and barometer cases now had a principled answer: explanation traces the causal arrow, prediction need not.

The pragmatic turn

A different response came from Bas van Fraassen. In The Scientific Image (1980) he argued that explanations are not arguments at all but answers to why-questions, and a why-question has a context-dependent structure: a topic, a contrast class ("why this rather than that*?"), and a relevance relation supplied by the background theory and the questioner's interests. The same fact can be explained differently depending on the contrast. This pragmatic account dissolved the search for a single logical form: explanation is an activity, and what counts as explanatory depends on what is being asked. Van Fraassen's view also fit his constructive empiricism—explanation was not a guide to truth about unobservables4.

Unification

Michael Friedman (1974) and later Philip Kitcher ("Explanatory Unification", 1981; The Advancement of Science, 1993) offered a rival to causal accounts: science explains by unifying—by reducing the number of independent assumptions needed to derive a body of facts. A theory explains well when it covers many phenomena with few argument patterns. Newtonian mechanics explains the tides and the fall of apples because one scheme derives both. Kitcher's version made explanation a matter of instantiating a small set of "argument patterns" that the scientific community accepts. Unificationism captured a real feature of theory choice—Kepler's laws being explained by Newton's—but critics asked whether unification is genuinely explanatory or merely economical.

Mechanisms

From the 1990s, philosophers of biology and neuroscience pressed a mechanist account. Peter Machamer, Lindley Darden, and Carl Craver's "Thinking about Mechanisms" (2000) defined a mechanism as a system of entities and activities organized so as to produce a phenomenon. To explain a phenomenon is to describe the mechanism responsible: the ion channels and neurotransmitter release behind a neuron's firing, the molecular steps of DNA replication. This was a deliberate move away from laws—much of biology has few exceptionless laws—and toward the detailed, hierarchical, often multi-level descriptions working scientists actually give. The mechanist programme became the dominant account in philosophy of the life sciences, with Craver's Explaining the Brain (2007) as a landmark.

Interventionism

A parallel development sharpened the causal idea. James Woodward's Making Things Happen (2003) advanced an interventionist theory: a causal claim is a claim about what would happen under an idealized intervention—if we could set a variable to a value while holding others fixed, how would the outcome change? This gave a precise, manipulability-based semantics for causation that meshed with experimental practice and with the causal-modeling literature in statistics and computer science (Pearl's work on causal graphs). It also supplied a criterion for explanatory relevance: a factor explains an outcome if intervening on it would change the outcome.

Explanation in practice and the present

The field today is pluralist in temper. Causal and mechanistic accounts dominate philosophy of biology and neuroscience; interventionism connects explanation to causal inference and to the sciences of experiment; unificationism survives as an account of theory-driven explanation in physics; and the pragmatic tradition reminds everyone that "why" questions are asked by people with purposes. A recurring live question is whether there is one thing—explanation—at all, or a family of explanatory practices (deductive, causal, mechanistic, statistical, functional) suited to different sciences. Another is the relation between explanation and understanding: does an explanation produce a distinct cognitive state, or is it just a well-supported account1? And the old Aristotelian demand for necessity—that a genuine explanation show why the outcome had to happen—still divides those who think explanation is about causes from those who think it is about laws, and both from those who think it is about answering questions. The argument Aristotle started is not over.

References

  1. Models of scientific inquiry (introduction, Logical empiricism), Wikipedia
  2. Deductive-nomological model, Wikipedia
  3. Carl Gustav Hempel, Wikipedia
  4. Scientific realism (History), Wikipedia

The frameworks of Scientific Explanation

Each framework is a way of working the field has used: when it held, and the question it answers.

  1. 350 BCE-1600

    Posterior Analytics

    What makes a demonstration truly explanatory?

    Fly to its tower in Nousfarer

  2. 1948-1975

    Deductive-nomological model

    When does a law plus conditions explain an event?

    Fly to its tower in Nousfarer

  3. 1962-1980

    Inductive-statistical model

    How do probabilistic laws explain an outcome?

    Fly to its tower in Nousfarer

  4. 1971-1990

    Statistical-relevance model

    Which statistical factors are explanatory rather than irrelevant?

    Fly to its tower in Nousfarer

  5. 1974-Present

    Explanatory unification

    Does explaining mean unifying many facts under few principles?

    Fly to its tower in Nousfarer

  6. 1980-Present

    Pragmatic account of explanation

    How does the question asked shape what explains?

    Fly to its tower in Nousfarer

  7. 1984-2005

    Causal-process account

    Does explanation require tracing the causal process?

    Fly to its tower in Nousfarer

  8. 1990-Present

    Explanatory pluralism

    Are there many legitimate kinds of explanation?

    Fly to its tower in Nousfarer

  9. 2000-Present

    Mechanist account

    How do organized parts and activities produce a phenomenon?

    Fly to its tower in Nousfarer

  10. 2003-Present

    Interventionist theory of explanation

    Does a factor explain if intervening on it changes the outcome?

    Fly to its tower in Nousfarer

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.

Posterior AnalyticsDeductive-nomological modelInductive-statistical modelStatistical-relevance modelExplanatory unificationPragmatic account of explanationCausal-process accountExplanatory pluralismMechanist accountInterventionist theory of explanation

Beyond the field

Where Scientific Explanation's ideas came from in other fields, and what they led to there: the strongest 12 of 37 links.

A city in Nousfarer: a red-carpeted stair climbing terraces to a crowned tower on an acropolis, more towers round it, forest and snowy mountains beyond.
A city in Nousfarer, the game played in the browser.

Scientific Explanation is a planet in Nousfarer

Its city has a tower for each of these 10 frameworks. Land, walk in, and climb one concept at a time.

Land on Scientific Explanation

Related fields

Fields its ideas are linked to

More in Philosophy Of Science

See where Scientific Explanation sits on the Map of Human Knowledge