Quine argues that "our statements about the external world face the the tribunal of sense experience not individually, but only as a corporate body."(p. 41.) Individual sentences are part of, or implied by, whole theories, and the theories themselves are malleable. Quine cites Duhem as holding the same view. No sentence can be empirically falsified simply because other sentences in the theory can be altered to keep the sentence as 'true.' Let us suppose that we have a theory about how the planets move around the sun, and let us suppose that we predict to see planet x at a particular point in the sky at some time t. We then run the experiment, and discover that the planet does not appear where we thought it would. Is the sentence "planet x at a particular point in the sky at some time t" false? If the positivists are right, the answer must be 'yes.' But maybe the problem is not with the claim about where the planet will appear, but about how light behaves. Then the planet really is there, but we are seeing it in a different place. Science is actually full of just these kinds of examples. Indeed, maybe the observation itself is poorly made. We can even throw out the apparent observation when it conflicts with deeply held theories. In fact, we do so all the time.
On quines view, our theories about the world are radically underdetermined by the data, so that multiple theories explain and predict the same data, whether the theory is about science or history. As he puts it:
"The totality of our so-called knowledge or beliefs, from the most casual matters of geography and history to the profoundest laws of atomic physics and pure mathematics and logic, is a man-made fabric which impinges on experience only along the edges.... the total field is so underdetermined by its boundary conditions, experience, that there is much latitude of choice as to what statements to re-evaluate in light of any single contrary experience. No particular experiences are linked with any particular statements..."He continues:
"Even a statement...can be held to be true in the face of recalcitrant experience by pleading hallucination or by amending certain statements of the kind called logical laws."
Not only is any statement not falsified by any particular observation, no statement is immune from revision. Even the laws of logic are revisable. We can, for example, give up the law of excluded middle. Some think we should give up the law of non-contradiction. We play with the axioms of geometry, of set theory, and these changes effect the theorems that are then true or false. In short, just as there is no pure synthetic claim, there is no pure analytic claim , either, at least not in the way Kant and others imagines, claims that are true no matter what.
We posit metaphysical entities to account for what we see, like physical objects, and these are not reducible to experiences. The same is true of God and gods. Which metaphysics we adopt is a question for which theoretical framework accounts for our experiences better. But in principle, there is no difference between ontology and natural science.
If asked whether or not sets or numbers exist, Carnap writes that it depends on the linguistic framework we choose, that there is no matter of fact which determines which framework we ought to use. For Quine, the exact same is true of questions in natural science. And we choose our linguistic framework not on the basis of many different reasons, not merely experiential. Hence, conservatism, simplicity, explanatory power, etc., all help choose which theory, and hence which ontology we accept.