Jason Rosenhouse draws an analogy that illustrates the equivocation:
Think of fish swimming in a small tank in someone's living room. For the fish there are clear laws that can not be violated. Their whole world is bounded by walls that are impenetrable for them, for example. They have very limited abilities to alter their physical environment. As a human being I am not subject to those restrictions. I can reach into their world and make all kinds of changes to their environment. Any fish scientist seeing the extraordinary changes I can make would be right to conclude that there is nothing from within the tank that can explain such wondrous changes. The best conclusion is that their are entities in the world not subject to the restrictions of the tank.Rosenhouse labels such entities as "supernatural"; this label, however, especially when talking about methodological naturalism, equivocates the concepts of materialism and naturalism in just the sort of way I object to. Instead, I propose that the fish scientists' "inside the tank" theories are natural materialist theories, and their "outside the tank" theories are natural non-materialist theories.
In contrast, fish philosophers and theologians might ask such questions as: where did the tank come from? What is the purpose of the tank? What are our moral obligations to the creators of the tank? These are substantively different kinds of questions than those that might be answered by "outside the tank" scientific theories. Since these are a different kind of question than both the questions that lead to inside and outside the tank theories, it make sense to draw a different kind of distinction: I propose natural and supernatural.
When we are talking about methodology, we want to ask: how — i.e. by what method — do we know? It seems obvious that to talk about methodological naturalism we are talking about an epistemology. But to say that methodological naturalism entails an a priori commitment to finding forces and causes in the material world — i.e. the world inside the tank — is to ascribe an ontological characteristic to an epistemology. But this doesn't make sense: Why should we constrain how we know, our method, by a specific a priori commitment to what we know? And since this ontological commitment is supposedly baked into our epistemic method, the commitment is not therefore know by virtue of that method: we are saying we know something about the world (all explanations are material/inside-the-tank) without saying how we know it. But if we can know something without an epistemic method, why bother with an epistemic method at all?
Consider two fish tanks. One is a standard salt-water tank that requires constant monitoring and intervention by its human (outside-the-tank) owner. The other is one of those "ecological" self-contained tanks that requires no intervention whatsoever.
In both tanks, the fish scientists will develop scientific theories about what happens inside the tank, using the ordinary method of coming up with the simplest theory that accounts for perceivable experimental outcomes. But our salt-water fish scientists will see something unusual: There are perceivable changes to the tank that cannot be explained by inside-the-tank mechanisms. They observe that the temperature of the tank sometimes changes (as the owner adjusts the thermostat) without any correlation to anything observable inside the tank. Likewise the chemical composition of the water stays in equilibrium again without any correlation to anything observable inside the tank; and when they isolate water within the tank, they notice that the chemical composition is not in equilibrium.
In the self-contained tank, however, everything that happens has an inside-the-tank explanation. The O2/CO2 composition, for example, is in equilibrium because the plants change CO2 to O2 and the animals do the reverse.
The key point is that the fish in the salt-water tank know there's something outside the tank in precisely the same way that they know about their inside-the-tank physics, and in precisely the same way that the fish in the self-contained tank know there's nothing causally effective outside their tank. This knowledge is publicly available to all the fish, it is the result of repeatable, shareable, public perceptual evidence, dependent only on the operation of the fishes' ordinary brains and sense. Thus, the epistemic method deserves the same name, i.e. naturalism. Since the evidence differs between the salt-water and self-contained tanks, they draw different conclusions about the world, and this ontological difference deserves its own name, different from the epistemic distinction, i.e. materialism and non-materialism.
The primary controversy between science and religion is not about what conclusions we draw about the world, it is between how we draw conclusions about the world. The controversy is not primarily ontological, it is an epistemic controversy.
The religious try to confuse the issue. They want to say that religion talks about what's outside the material world; science can discuss only what's inside the material world, therefore science a priori excludes religion. But this is nonsense: As Rosenhouse, Coyne and others have pointed out, the actual methodology of science does not exclude a priori non-materialism; it just happens to be the case that material explanations are so far — and we have gone far indeed — entirely successful: it is a matter of empirical — not metaphysical or philosophical — truth that the best explanation is that we're in a self-contained tank. For every phenomenon — life, the stability of the solar system, evolution — although superficially suggesting a non-materialist outside-the-tank explanation, we have found a materialist explanation; where we have not yet found a materialist explanation, we know we have not yet looked thoroughly, much less exhaustively.
The religious try to shift the issue to an ontological basis to disguise the sad truth that they do not have an alternative epistemological method to talk about a particular ontological domain; the religious have no epistemological method whatsoever.
To counter this obfuscation, I suggest we always keep the distinction clear between natural and supernatural epistemology and materialist and non-materialist ontology, and make it clear that a materialist ontology is the result, not an a priori commitment, of natural scientific epistemology.
Second, the question that must be sharply posed to the religious is just this: how do you know? We must keep holding their feet to the fire on this epistemic question. To the extent that the religious say that a natural epistemology is inapplicable to questions about a god, we should first ask why? We must call bullshit on the answer that science can discuss only the material: this assertion is simply not true. Having undermined this excuse, we can ask again: how do you know? They have no answer, and can only retreat into the obvious bullshit that it is rude and offensive to ask such a question.