The easiest way to construct a powerful theory is simply to state that everything is possible, and the result, mirabile dictu, will be a theory which can never be disproved. If I have a theory of gravity which states that a falling object can fall in any direction, who can prove me wrong? Any way it falls will be OK with my theory, and falling down instead of up will just be a special instance of the range of phenomena allowed by my theory.
(Which is itself a quote from here; note that's an RTF file and as such I haven't looked at it.)
The context of the Amritas and Den Beste post is a discussion of the linguistic theories of Chomsky, which are much beloved by his followers because of their ability to unify all languages of the world without fail.
Oops, missed some scare quotes there, make that "the 'ability' to 'unify' all languages of the world without fail".
As they both point out, a theory that is not falsifiable is not useful. The same source I quote above goes on to say, "This example should make it clear why an empirical theory is being defined here as one which is capable of being disproved." I would go further, and say, "A theory is defined by that which it excludes."
A theory need not explain anything; when Einstein gave us Relativity, he did not necessarily know why it was true, merely that it fit the facts. I would say that relativity is not defined by the predictions it makes, but the outcomes it excludes. Specifically, it excludes relationships in the universe that do not match the equations of Relativity. It excludes any extraction of energy from matter (such as chemical reactions, fusion, fission, matter-anti-matter interactions, or anything else) where the final quantities of matter and energy do not match E=mc2.
The formulations of that part of that theory as either
- If a quantity of mass is converted to energy, the amount of energy it will be is derivable via E=mc2, or
- If a quantity of mass is converted to energy, the amount of energy derived will not be anything other then E=mc2
are logically equivalent, yet have a profoundly different human meaning.
So take the Chomsky theories explained in the post above. I'd propose a different test then Steven does. Ask one or two Chomsky followers to create a language that cannot be explained by Chomsky's theories. In other words, pull a language from the set of languages that Chomsky's theories exclude. (This will of course be a "mini-language" for practicality's sake.) Then, hand this language over to several other Chomsky followers and ask them to explain this language in terms of Chomsky. Repeat as desired.
(Of course, if the followers are incapable of coming with such a language at all, then one can immediately conclude that the theories as understood by that person have no explanatory power at all!)
If Chomsky's theories truly exclude a language from consideration, the followers should all (or largely) conclude that the manufactured is impossible. If they all come up with explanations (and I'm not sure if similar explanations or dissimilar explanations would be more damning), then while that technically does not disprove anything about Chomsky's theories since the original language creater may have made a mistake, it does provide that much evidence that Chomsky's theories do not exclude anything, and as such, aren't "theories" at all.
sciences "studies" in English, Social
Sciences, etc., find it particularly easy to fall into this trap of
theories that encompass everything and exclude nothing. It
is trivially easy to come up with such a theory, and one may even
derive some small degree of benefit from a theory by examining new
structural similarities between things previously supposed
disparate. But there are an infinite number of theories that exclude
nothing, and few of them are even that useful. It is only the theories that
correctly exclude something that are worthy of being called
"theories"; everything else is merely word games.