Thursday, February 26, 2009

Laws, Counterfactuals, and Essential Properties

I find it curious that nobody seems to be particularly bothered by the fact that the following three commonly-held and seemingly plausible theses seem to be somewhat at odds:
  1. Unlike accidental generalizations, nomic generalizations support counterfactual conditionals. (So, for example, if it is a law that copper is a good conductor, then, if this piece of wood was made of copper, it would be a good conductor.)
  2. Some properties are essential to their bearers (So, for example, it is metaphysically impossible for this piece of wood to be made of anything other than wood and a fortiori to be made of copper).
  3. Counterfactuals whose antecedent is necessarily false are vacuously true.
The conflict seems to arise from the fact that, since laws of nature often involve essential properties, if (2) and (3) are true, (1) would not seem to be generally true--many accidental generalization would seem to support (vacuously true) counterfactuals just like nomic generalizations do.

Now, I'd be curious to hear which one(s) of the above theses (if any) the readers of this blog think should be amended/rejected in order to resolve the conflict and why. (I do have a main suspect, but, in order to avoid skewing my little survey, I'm not going to reveal its identity for the moment).


  1. Thesis 3 clearly has to go of course, but I would reject thesis 2 as well, at least for that sort of property. (Well, sometimes I'm more concessive about essences, but when I am I don't think they hold of the kind of necessity that is at stake in thesis 3.)

    I'm not seeing the conflict, though - the counterfactuals supported are true, on the view you describe, so what is the problem with saying they are supported by the laws? Of course various counter-nomics are also true according to this view (if this piece of wood was made of granite, it would conduct electricity), but it needn't be supported by a law of nature. Nor by an accidental generalisation, either. I haven't seen any reason to think the counterpossibles are _supported_ by accidental generalisations. My guess is the people who believe 1-3 won't see a conflict here yet either.

  2. Would someone who's committed to 2 think that there's something nomic to be said about the potential copper-like properties of pieces of wood? If being made of wood is an essential property, then surely "if this piece of wood was made of copper" makes no more sense than "if this object was another object".

    The version of 1 that makes sense for a follower of 2 can't refer to the object specifically. What could be held is something like this: Nomic generalizations support counterfactual conditionals such as "If the cogs in this wheel were made of copper, they would be good conductors" (and it so happens that the cogs in this wheel are wooden).
    The reference to the cogs must be de dicto. And in that case, the antecedent is not necessarily false so 3 doesn't create any trouble. The version "this piece of wood" OTOH implies a de re reference which results in trouble.

  3. There are probably ways to sharpen (1) into something that is true, but it won't look much like the way it is stated here.

    Jim Woodward has defended that view that many special science generalisations support a variety of counterfactuals while lacking most of the traditional features of laws (Making Things Happen, especially Chapter 6). I think he is right. If "nomic generalizations" means "generalizations having the properties traditionally attributed to laws of nature", and "accidental generalizations" means all the rest, then (1) is false.

    Moreover as he points out there are examples that suggest even weaker generalisations can support counterfactuals. Here is a paraphrased passage from the book (Woodward credits the example to Aardon Lyon):

    "A museum has adopted a policy such that all of the Sisleys in its possession are hung in room 18.
    You are ignorant of this policy and ask, regarding some painting in room 17, whether it is a Sisley. You are told in response that if this painting were a Sisley, then it would be in room 18. There is a natural reading of this counterfactual according to which it is true and according to which it is supported by the generalization specified in the policy. Nonetheless, that is no law and is a dubious candidate for a causal or explanatory generalization".

    Elliott Sober has made much the same point.

  4. Daniel,

    I suspected that was going to be your reaction and I'm with you (although for different reasons), but I'm quite surprised to hear that you think that vacuously true counterfactuals can do the job. Clearly, those who believe (1) have genuinely true counterfactuals in mind. Moreover, the fact that (1) can be satisfied by any true universal generalization involving essential properties (whether accidental or nomic) would seem to be worrisome for the supporter of (1), for (1) explicitly says that accidental generalizations *do not* support counterfactuals.


    I agree but the de dicto reading does not seem to capture the idea of (1), if as it is likely if you believe (2) take laws to be relations between properties. If you do so, you think that if it is a law that all As are B then there is some sort of modal tie between being an A and being a B and that all it would take for this non-A to be a B is its possibly being an A not that if some other object that happens to be an A was in its place it would be a B (the latter is simply guaranteed by the fact that all A are B and seem to have no modal force).


    I completely agree that the one between nomic and accidental generalizations is a false dichotomy and usually prefer to use nomic/non-nomic and accidental/non-accidental myself (here I was only following the more-or-less standard formulation of (1)), but I'm not sure I see how this would solve the problem at hand anyway.

  5. A quick question for Daniel (or anyone else who wishes to enlighten me)!

    You said that thesis (2) seemed to be "less of a necessity" to you and that you were fine with rejecting that. Even if we did not know the essential properties of an object it seems very difficult for me to understand how we could attribute names or identity to any object if it did not retain some sort of essential property or essential "feature".

    If I was to reject thesis (2) as you suggested, what would you propose in its place that would afford objects the ability to remain distinct from others?

    I am a philosophy student who is very interested in the subject matter of this blog. I cannot apologize enough if these are elementary questions. Thanks for taking the time to read this!

  6. Hi Steven: If you are generous enough with features, then there will be features that are essential to objects: I am identical to Daniel, for example, and if that attributes a "feature" then I guess I couldn't lose it without ceasing to be Daniel.

    I took Gabriele to be talking about more robust essential features than that though: being wooden, being human, and others that might show up in the laws. I don't think we need to suppose anything has those properties essentially in order to name or identify those objects - isn't it enough to locate the humans, for example, whether or not the entities located are essentially human?

    I also doubt that objects need anything else (essences or otherwise) in order to have the ability to remain distinct from others: objects seem to be able to do that perfectly well on their own.

    I hope that helps...

  7. Maybe when we think that "if this piece of wood was made of copper, it would be a good conductor" we're not thinking about the counterpossible involving this piece of wood being made of copper, but rather that if there were a piece of copper here shaped like the piece of wood, it would be a good conductor?

  8. Gabriele,

    I don't think you have addressed Daniel's point concerning support, which was that it is unclear in what sense your (2) and (3) show that accidental generalisations involving essential properties support vacuously true counterfactuals.

    Here is a way of making that point, as I understood it: (2) and (3) do not require the truth of any generalisations; so the counterfactuals they license are not supported by generalisations; so they provide no reason to disbelieve (1).

  9. Can you please recommend an account (article, book) of accidental vs. nomic as pertains to generalizations? Thanks.