Two Answers for Judge Posner
1. What is my basis for thinking this? Conversations with judges. For anyone who does empirical studies of judicial decisions, and finds evidence of what seems like ideology at work in them, it is common enough to be told by a judge something like this: "that's interesting, but it doesn't square with my experience -- when I decide cases, I really feel that I'm making decisions based on legal materials, not policy preferences." I have been told that sort of thing a number of times in conversations about these same issues, and I don't think the judges who have said it are being disingenuous. (Sometimes objections like these are made public, as by Harry Edwards of the D.C. Circuit. I don't agree with him, but I don't think he is being disingenuous, either.)
The implication of the question, and of Judge Posner's earlier remarks, is that he doubts the polygraph test would be passed. I would be interested in hearing speculation about this from the other judicial participants in this panel (but I sense a danger of circularity here!). I think polygraph tests of this kind are an exciting and amusing frontier in empirical legal scholarship. If Judge Posner can persuade his colleagues to take them, I will start hunting up grant money.
2. Suppose they did pass the test; what would that show? It would show that (a) they thought those semantic arguments were important, and didn't intend them as a smoke screen; and that (b) they thought that by relying on semantic arguments, they were deciding the case on grounds independent of their policy preferences. Point (a) would fortify our claim that it's worthwhile to study how semantic arguments work, as we have been doing; for they are serving some function (but what?) in judicial decisionmaking. As for point (b), despite the polygraph results I would not conclude that the semantic grounds really were the basis for their decisions in any deep sense. On the contrary, I would point to our studies and suggest that the semantic arguments probably felt attractive because they squared with the judges' policy preferences. So perhaps the semantic arguments are best viewed an unintentional or unconscious smoke screen (or maybe as a way to reduce cognitive dissonance, to use Judge Posner's phrase).
And so I would say: beware semantic arguments; they sound unconnected from policy preferences, but easily get entwined with them and express them even if you don't realize it (arguments about language can feel very convincing, like they have nothing to do with preferences, but that feeling is unreliable). If you really do want to argue about semantics and not policy, we have found a method that might make this somewhat more likely to work, though the method may or may not be attractive in any given case for reasons we talk about in our papers (Judge Posner would probably never like this approach). If it turns out that you actually want to argue about policy and not semantics, then it would be better to stop pretending that semantics are the issue. (Here I think Judge Posner would agree.)