Common attempts to define science each capture something real, but each excludes things we’d intuitively count as science, or sets a standard actual science doesn’t meet.
- Science as a mastery of the environment The Baconian view (Francis-bacon, 17th century): the point of natural philosophy is practical dominion over nature. Knowledge is valuable only insofar as it gives power to control and manipulate the environment — science should be operative and beneficial, not merely contemplative.
Failure: defines science by its products, collapsing science into technology. As Ziman puts it, this “confuses the ideas with things” — penicillin is not science, any more than a meal is a recipe. It credits the application while discarding the questions, methods, and theories that constitute the actual intellectual practice.
- Science as the study of the material world Rooted in logical positivism where for Carnap, any claim not in-principle verifiable through observation was not just unscientific but meaningless. This distinguishes science from inquiry into abstract or metaphysical claims.
Failure: depends on a definition of “the material world” that’s hard to draw consistently. Does psychology count? Sociology? And if we stretch the definition to include these, what about pure mathematics — excluded by this definition, yet the very language science is written in.
- Science is the Experimental method Experiment as the symbol of rigour: you test the world to know how it works. Refined by Popper’s falsificationism — the point of science isn’t confirming theories but ruling them out. A good theory makes predictions that could in principle be proven wrong and empirically tested.
Failure: excludes mathematics (again), and excludes major sciences that don’t run experiments — evolution, general relativity, astronomy, geology. These observe consequences of events they cannot manipulate or recreate in a lab.
- Science as logical inferences from empirical observations The most intuitive definition, and the one most practicing scientists implicitly hold: observe the world, draw logical inferences, arrive at the truth. Rests on the principle of induction — the sun rose yesterday and the day before, so it will rise tomorrow.
Failure: runs directly into problem-of-induction. These inferences are highly plausible but never strictly true — they presuppose the future conforms to the past, which cannot itself be justified. If this “logico-inductive scheme” is the definition, no actual lab work could ever count as scientific, since our conclusions are always partial. Worse, perfectly reasonable scientific behaviour gets classified as unscientific by the definition’s own standard:
“Surely, it is going too far to say, for example, that it was ‘unscientific’ to continue to believe in Newtonian Dynamics as soon as it had been observed and calculated that the rotation of the perihelion of Mercury did not conform to its predictions” (Ziman, 1968, p. 6)
Problem with defining science
Each definition captures something real about science — practical power, empirical grounding, experimental rigour, logical structure — but fails by either excluding things we’d intuitively count as science, or setting a standard no actual science consistently meets.
This may not be a failure of definitions found so far, but a feature of the concept itself: science may resist clean definition. Practicing scientists aren’t taught a definition and then apply it — they’re initiated into a community through imitation, the way one learns most complex human practices.
This doesn’t make the question worthless — courts deciding whether intelligent design counts as science, or funding bodies deciding what counts as legitimate research, show the stakes are real. But this may indicate that science is recognized more like a family resemblance concept than something with necessary and sufficient conditions — we know it when we see it, even without being able to articulate why.