In a previous post, I argued that Earman and Roberts, in their 2005 PPR articles on Humean Supervenience (HS) about laws of nature, were implicitly relying on the view that nomic facts (such as singular causal facts) are unobservable.  There followed an interesting discussion.  But the question that none of us could answer was, what do Earman and Roberts think about HS vis a vis the observability of nomic facts?

John Roberts wrote me as follows (reproduced by permission):

You're right that the argument of that paper depends on the assumption that nomic facts are not observable.  You call attention to one possible counterexample:  facts about causal relations.  I assume that you mean facts about token causal relations, rather than causal generalizations — like, "taking that pill caused my headache to go away this afternoon," as opposed to "aspirins cause headaches to go away (in general)", right?

Another, similar potential counterexample comes from dispositions.  The possession of a dispositional property by a given object at a given time may be a nomic fact, and it seems it may be observable as well.

There are two lines of defense here.  The first is for us to simply deny the counterexamples.  Why should we think that singular causal relations are observable?  I may observe that one token event is followed by another, and there may be many other observable features that lead me to infer that the first caused the second, but why should we think of this as a case of observation?

A lot depends here on how observation is conceived of.  In the paper, we adopt an account of observability (or measurability) that requires that in genuine cases of observation, a detection procedure is used whose reliability is guaranteed by the laws of nature.  That is, roughly:  the laws have to imply that when observation O is made with result R, fact F (the one getting observed) obtains (or that some statistical version of this).  Now, if the fact F that is getting observed is itself a fact about singular causal relations, then this means there has to be a law that says roughly that whenever two events stand in some causal relation, a third event happens.  The simplest version would be:

It is a law of nature that:  An event of type D (the detection) occurs only when an event of type C causes an event of type E.

But this is strange.  I don't know of any plausible candidates for laws of nature in which the fact that two events (not only occur but also) stand in a certain casual relation itself causes some further event to occur.

Now, it may be that we don't disagree — that the sense in which the psychological tradition you are referring to says that we can observe causal facts does not require that observations involve nomologically reliable detection procedures.  Maybe it's enough to call something "observable" if we are able to come to a judgment about it pretty quickly and without any conscious inference.  In that case, I don't think we have a non-terminological disagreement.  But it may be that you are using the same notion of observation that John E. and I are.  In that case we really do have a disagreement.  And your view commits you to the existence of laws of nature of a kind that we think don't exist.

Suppose you don't buy this first line of defense, and you convince me you're right.  Then there's the second line of defense:  If we grant for the sake of argument that singular causal facts are nomic facts, then some nomic facts are observable.  Since the thesis we defend under the name "HS" is really the thesis that the laws supervene on the observable facts, this means that all we can really claim to have argued for is the thesis that the laws supervene on a base of facts that includes singular causal facts in addition to the more paradigmatic sorts of non-nomic facts.

You might say that this would trivialize our thesis.  But I don't think it does.  If it turns out that the laws of nature supervene on a base of facts that includes the instantiations of basic occurrent physical properties and relations by physical particulars, and also the singular causal relations, then that's surprising, and I think it is incompatible with most of the leading anti-Humean accounts of laws.