How (exactly) do theories constrain the generation of various, potential hypotheses to explain something? In particular, what are the different implications of different accounts of the structure of scientific theories (syntactic, semantic, or pragmatic) in this regard?

(Unfortunately, the SEP article on The Strucutre of Scientific Theories explicitly avoids to discuss the matter.)

Thanks for any response!

  • 1
    Hypotheses are constrained by the "hard core" of established theories (at least, under "normal science"). Some of it is discussed in the Lakatos's entry:"The negative heuristic of the programme forbids us to direct the modus tollens at this “hard core”. Instead, we must use our ingenuity to articulate or even invent “auxiliary hypotheses”, which form a protective belt around this core, and we must redirect the modus tollens to these."
    – Conifold
    Oct 18, 2023 at 12:38
  • 2
    Start with Feynman's lectures on the Character of Physical Law. They require no training beyond a high school education.
    – g s
    Oct 18, 2023 at 22:28
  • I don't know, isn't there more to be said on this than it's a matter of consistency? Anything consistent with our theories is allowed, though otherwise, what we come up with to explain something in particular depends entirely on our imagination? ...
    – Turtur
    Oct 19, 2023 at 19:03
  • It's the semiotic consisting of all the three parts as syntactic, semantic, and pragmatic constraining the search of a hypothesis space defined by its underlying representation such as logical descriptions or linear functions or decision trees appropriate for learning different kinds of target functions... Oct 31, 2023 at 6:27

5 Answers 5


Your question has quite a technical answer that lies in the body of working surrounding theory-ladenness:

In the philosophy of science, observations are said to be "theory-laden" when they are affected by the theoretical presuppositions held by the investigator. The thesis of theory-ladenness is most strongly associated with the late 1950s and early 1960s work of Norwood Russell Hanson, Thomas Kuhn, and Paul Feyerabend, and was probably first put forth (at least implicitly) by Pierre Duhem about 50 years earlier.

The mechanics of it can be seen primarily as linguistic, and a short example might clarify.

Consider that you are doing research in Austria, and the popular paradigm (see Kuhn's thesis) is, say, that diseases are transmitted by a thing called miasmas. Well, when you do you research, and someone comes along and claims that germs are a better explanatory device, you might reject any conclusion that uses the term 'germ' at all, because you simply don't believe they exist. Therefore, a miasma proponent and a germ proponent don't have just a different theory of disease, they have a fundamental disagreement about what is physically real and therefore have an ontological dispute using language. In philosophical language, they have different ontological theories, and perhaps that even indicates a dispute in meta-ontological methods.

So, yes, there is a dispute about semantics superficially, but the important lesson is that there is the effect that such a semantic dispute actually is indicative of differences in ontological commitments and epistemological methods. In fact, the example above is pulled from the life of Austrian Ignaz Semmelweis who faced skepticism and hostility from the medical community who didn't seem to believe that washing your hands after working on cadavers to deliver babies was impactful. This is why some philosophers of science have gone so far post-Quine to insist that science is best viewed through the lens of persuasion and rhetoric, and logos plays but a minor role.

  • If you want clarificatory language on the aspects of philosophy of language, cook up some questions and I'll see if I can't field them.
    – J D
    Oct 19, 2023 at 19:44

Not sure about structure of scientific theories, but I think the Durheim Thesis is applicable here: it is always possible to avoid refuting a theory T by adding auxiliary hypotheses to "save" it.

Good example is epicycles vs heliocentric theory to explain the apparent motion of stars.

As @Confold mentioned in comment that echoes a very similar idea from Lakatos.

Apart from how theories can constrain how we respond to new data, hypotheses also constrain where we think we need to ask questions: a Newtonian will not see a need to test the idea that time is affected by Gravity, or that time is not universal (as a simple example), but will test if the trajectory of a celestial body around a planet confirms a hypothesis about the planet's density.

  • Epicycles in Mercury's orbit showed Newtonian gravity couldn't provide a complete account. Gravitational lensing might be a better example
    – CriglCragl
    Oct 22, 2023 at 0:53

Karl Popper considered the method by which we arrive at hypotheses to be innately uncharacterizable -- as creativity and imagination are not algorithmic. What he considered important was not HOW we derived our hypotheses, but that we do test case checks against our hypotheses.

Quine, in critiquing Popper, noted that any hypothesis, if it is sufficiently kluged with additional caveats, can be compatible with any evidence or test case.

Popper's response is to argue that one should then choose a hypothesis based on its falsifiability -- that kluges that are untestable in principle, have vitiated a hypothesis such that it is no longer predictively useful.

The cleanest articulation of this response to Quine comes from Imre Lakatos: https://bertie.ccsu.edu/naturesci/PhilSci/Lakatos.html Hypotheses are formed within a Research Programme framework. A Research Programme may have active refutations, but so long as they are under investigation, that is not sufficient to dismiss the Programme. And there can be multiple concurrent Research Programmes that can support competing suites of hypotheses and concept explorations.

The primary alternatives to Lakatos include Kuhn's "paradigm shifts are not rational", and Feyerabend's "Science is whatever scientists who do it say it is". Quine and Wittgenstein's "science is just a coherent game, and we could have all sorts of such coherent games" fails the falsification test cases that also refute Popper's falsificationism (IE the "game" can't actually pass full coherency tests), but Lakatos research Programmes can accommodate.

  • "A Research Programme may have active refutations, but so long as they are under investigation, that is not sufficient to dismiss the Programme." If you're implying that the linked page states this: I could not find anything to that effect. Oct 21, 2023 at 19:25
  • @DennisHackethal The link is just to a summary of Lakatos' thinking. My preferred summary link disappeared a few months ag, and that is one of several I have used instead. Check these alternatives out, and dig into Lakatos' writings your self, to get the full content of his ideas: linkedin.com/pulse/… mantleplumes.org/Lakatos.html qualityresearchinternational.com/socialresearch/…
    – Dcleve
    Oct 23, 2023 at 20:43

Obvious ways in which theories tend to constrain new ideas are socio-economic. Once a theory gains a critical mass of adherents, the barriers to gaining acceptance, or even just attention, for a competing hypothesis can become almost insurmountable. Theories that have wide acceptance become embedded in academia and schools. People base their careers on elaborating the established ideas, publish books about them, and so on. Consider, for example, the influence of Marxism and the hold it has had on political and social thinking. The challenger with a new idea has to overcome a huge amount of inertia and vested interests to have any chance of making an impact.


Like several others before I would like to refer to Thomas Kuhn and his book “The Structure of Scientific Revolutions”.

Kuhn discriminates between two phases of scientific work:

  • Normal science: Working within a given theory, applying the theory to elaborate on phenomena covered by the theory.

As examples I consider: Calculating the motions of the planets by solving Newton’s equations. Solving questions about electromagnetism by applying Maxwell’s equation. Calculation the spectra of atoms by solving the Schroedinger or Dirac equation. Scientific research is guided by the theory which is accepted in this field.

  • Scientific revolutions: Changing the paradigm, breaking with the previously accepted theory. Introducing new concepts.

As examples I consider: Breaking with the geocentric model, breaking with the concept of absolute space and absolute time (theory of relativity), breaking with the concept of action at a distance by introducing the concept of waves, breaking with the concept of determinism for the microcosmos (quantum theory).

Developing a new theory based on a different fundamental principle, because important problems cannot be explained on the basis of the accepted theory.

Note. My answer only deals with the first part of your question. I am not able to contribute to answers about the syntax, semantics and pragmatics of scientific theories.

Added: During the phase of normal science the dominant paradigm can restrict the generation of hypotheses. Maxwell’s theory of electromagnetism, based on his fundamental equations, predicts that accelerated charges generate electromagnetic radiation. And the latter reduces the energy of the charged sources.

Applied to the model of atoms with their circulating electrons, the electrons should radiate on their circular orbits, lose energy and fall into the nucleus of the atom within a short period. As a consequence, matter cannot be stable according to Maxwell theory.

Nobody could create a hypothesis to explain stability of matter as long as he/she follows the Maxwell theory.

  • I understand why Kuhn comes up here, but then again, I do not really see how this goes beyond the statement that theories constrain hypothesis generation to specify how they do this ...
    – Turtur
    Oct 27, 2023 at 4:46
  • @Turtur I made some addition at the end of my answer.
    – Jo Wehler
    Oct 27, 2023 at 5:44

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .