What do scientists mean by 'theory'?

In my basic concepts post on theory testing, I set out what I take to be a fairly standard understanding of "theory" in philosophy of science discussions:

... a theory is a group of hypotheses that make claims about what kind of entities there are and how those entities change over time and interact with each other. If you like, the theory contains claims about ontology and laws. If you prefer, the theory tells you what kind of stuff there is and how that stuff behaves.

In a comment, Lab Lemming opined:

That's a really odd way to think of theories. I reckon a theory is something that is both sufficiently broad and sufficiently tested that a scientist can use it to figure out how to frame a question in a testable manner.
For example, if you're looking into an outbreak of a drug resistant germ, you can use the theory of evolution to design an experiment that will test how this resistance arose, and how effective it is against other drugs.

Now, I don't think these two descriptions of a theory are necessarily contradictory, although the theory-as-set-of-hypotheses I'm describing might not be well tested (i.e., it could be a new theory). Indeed, it seems like having some picture of what kind of stuff you're dealing with and some of the ways that stuff behaves it at least extremely helpful, if not absolutely necessary, to the framing of questions in a testable manner.
The thing that interests me here is the possibility that scientists think of "theory" in a different way than the philosophers of science who are trying to make sense of what scientists are doing. So, a question to the scientists: How would you define (or at least describe) a theory? If you think it's relevant (and it may well be), indicate your scientific field.
Once we have some "empirical data" on this, I'll give you my hunches as to what may be behind some of the differences in perspective.

Comments are off for this post

  • Sean Carroll says:

    Here was my take from a while back:
    Basically, if what you care about is how scientists use the word theory, rather than some prescriptive notion of how it should be used, you have to face up to the fact that the usages are wildly varying and inconsistent. Sometimes it just means "a model," other times it means "a subject area," other times it means "a well established set of rules and ideas." I worry about pretending that it means something more specific and well-defined than it really does.

  • Sean Carroll says:

    As a follow-up, here is some empirical data, in the form of papers in the SPIRES high-energy physics database that use the word "theory" in the title:
    It's easy to see all of the above meanings, plus several more.

  • Capella says:

    As a theorist (of soft matter; polymers and the like), I think an important element of a theory is that it is a (somewhat) comprehensive model of a class of phenomena. An equation is not, by itself, a theory. A prediction is not, by itself, a theory. In order to qualify as a theory, a model has to provide an explanation for some already-observed phenomenology and a prediction for some other (already observed, or capable of being observed, at least technically) phenomenology. Your definition makes sense, but it's not really how I would think of a theory while in the process of developing one.

  • gengar says:

    I tend to regard science as an excercise in building (and testing) explanatory links between disparate observations and data; so I think of a theory as the description of all the links between a particular set of data, as well as an assessment of how well-supported those links are (tested to death, seemingly well-supported, speculative).
    I'd guess Lemming is just making the point that theories are not solely an end in themselves, but that they are also used as tools - people use the present pattern of links to propose new connections, or transplant them into new fields.

  • Lab Lemming says:

    I take back the well-tested requirement for theories.
    Or, rather, I admit that it was a value judgement on my part- that I think that scientists are generally better off using those sorts of theories.
    As Sean points out in his blog, some theories, like phlogiston, or lamarkian evolution, don't hold up very well at all. So there are good theories and bad theories.
    However, I don't think you can tell the difference between them by any sort of logical proof or analysis. I think the most appropriate way to evaluate theories is on the basis of their utility. There are useful theories, and theories that are a waste of time. And judging theories by their utility gives us a handy way to differentiate science from philosophy, logic or religion, which deal in non-scientific matters like truth and correctness.
    I have a half-baked explanation of how this applies to evolution written somewhere, which I'll try to find, finish, and post.

  • John B says:

    I'm in comparative religions... do social scientists count?
    We use theory to focus in on the giant pile of data in our field, to tease out specific elements we want to examine. It's a necessary tool for methodological reduction.

  • To me a theory is a set of experiments incorporated into a story which lets you make predictions about experiments to come. Thus you have a theory in your head, and in that framework you construct hypotheses which you can test.
    I think a lot of divergence comes in the barriers you set yourself in how you can tell stories from a theory, and what the relevant objects are (if you're worrying about binary effects of having genes, the structure looks rather different than if you're fighting with the 23rd decimal place of a measurement of a physical constant).

  • Bob O'H says:

    I agree with Sean, that the problem is that theory has several, closely related, meanings in science. I think you need a lexicographer to tease out these meanings.

  • We don't need no stinkin' lexicographer!
    That is to say: What I'm really asking for is not the officially way Science defines Theory. Rather, I'm interested in how actual scientists in various scientific fields and contexts use the word theory. (Lexicographers need to attend to usage, I reckon.) So, I'm under no illusion that all the things people mean by "theory", or expect should come along with a "thoery", or would identify in contrast to "theory" will line up neatly.
    I just want some good reports from the field, as it were.

  • RBH says:

    I tend to think of a theory first as a system for establishing correspondences between the terms and operators of some formalism on the one hand, and objects, processes, and relationships of phenomena in the world on the other hand. That is, theories are interpretations of formalisms.
    Testing theories then resides in messing with the transition rules of the formalism to generate predictions of what will happen if the corresponding objects, processes, and relationships are altered in a similar way. For example, interpreted in terms of objects and operations performed on them, F=MA suggests that if I pare some M off an object and accelerate it, measured F will be a specified amount less than it was before I did the paring at the same acceleration. Lots of tests tend to more strongly corroborate the theory (formalism plus correspondence rules).
    The formalism doesn't necessarily have to be expressed in explicit equations. I do applied work with evolutionary algorithms -- simulations -- and theory tells me about correspondences between the processes in the model running in the computer and the phenomena of interest in complex adaptive systems. Multiple runs of the simulators under plausible perturbations provide ranges of outcomes to be expected in the world of phenomena we model, and the veridicality of those predicted outcomes are tested with real money every day in various derivatives markets. Trading in negative-sum markets is much less forgiving of invalid or plain wrong theories than are one's (former) academic colleagues. 🙂

  • Bob O'H says:

    Lexicographers need to attend to usage, I reckon.

    I was assuming they did! At least the approach the OED takes is descriptive - i.e. they describe the way words are used. I agree that having a prescriptive lexigographer involved would be awful.
    I think a survey of usage would be useful, and may have been done. But searching lexicographal journals for "theory" and "science" may not be the wisest way of finding it.

  • wolfwalker says:

    My $.02 worth on this topic:
    Science starts with an observation. You devise a hypothesis to explain the observation. Then you use the hypothesis to make predictions. Then you test the predictions. If the predictions are wrong, then you go back and try a different hypothesis. If the predictions are correct, you try to expand the hypothesis to be more general and more powerful. A very general, very complex hypothesis that makes lots of verifiable predictions and explains lots of different observations is called a theory.
    I still call things like phlogiston and Ptolemaic astronomy by the term "theory" because they were well-meant attempts to form theories, using the best evidence then available. They failed because they didn't agree with the facts, not because they were internally flawed or illogical.