Law professor Stanley Fish probably knows quite a bit about evidence, but from his recent article with the above title I am led to doubt that he really understands much about science.
In particular, his main point appears to be based on a misunderstanding, for he says:
What I do assert is that with respect to a single demand — the demand that the methodological procedures of an enterprise be tethered to the world of fact in a manner unmediated by assumptions — science and religion are in the same condition of not being able to meet it (as are history, anthropology, political science, sociology, psychology and all the rest).
When a scientist expresses the criteria of an experimental test in terms of the theory being tested, that is only a shorthand for those familiar with the theoretical context and the true test can always be expressed in terms that require no theory-specific assumptions. For example the prediction that “this collision will produce an output of that particle” is just shorthand for something that could be expressed (though at much greater length) in terms of statements like “if you set those dials this way then that needle will point to this mark”.
For any statement that deserves to be called scientific there is a clear path through the literature to a testable consequence that can be described in terms of everyday objects. This may be what Richard Dawkins meant by his recent gaffe joke about citing “chapter and verse” – which, in the scientific context, refers not to the source of authority but to the first step towards finding a recipe for a test (but which Fish once again managed to take the wrong way!).