There is danger in stressing commonalities among methods because the differences matter in fixing the meaning of our claims. Different methods can, and often do, test the same claim. But it takes a strong network of theory and empirical results to ensure that. Failing that, we are likely to fall into inference by pun. We use one set of methods to establish a claim and then draw inferences licensed by a similar-sounding claim that calls for different methods of testing. Our inferences fail, and bridges we build (or policies we set) depending on them fall down.