Let us say we hope to understand some particular belief system. We shall find ourselves confronting a mass of propositions, couched perhaps in terms that we find exotic or puzzling (or, more treacherously, in terms that seem easy and self-evident but that are being used in an unfamiliar sense), and with no obvious clue as to what any of it means—let alone why anyone would believe it, or how it might feel to do so. Believers themselves cannot be relied upon as a guide: they are of course the absolute authority on what they believe, but their insight into why they believe what they do is complicated by the fact that belief systems often include their own doctrines about what makes a good reason for believing them (the conviction of sin, or the evidence, or whatever it may be). These doctrines form part of the belief system in question, so its adherents will tend to believe that they are true—and, by extension, that these are the very reasons that they themselves find persuasive (Towards a Science of Belief Systems, §44).
There is, however, no such thing as a wholly illogical belief system. If there were, it could not be distinguished from a set of doctrines assembled at random: and, however much time we spent eliciting believers’ views on one point after another, we would never get any better at predicting their answers to questions we had not yet asked. Belief systems are not, in fact, like this—nor should we expect them to be. The basic postulate of descriptive logic is simply that people do not, all other things being equal, like to feel that they hold incoherent or mutually incompatible beliefs on matters about which they care strongly. If I am powerfully committed to the proposition that A, and I find myself inclined also to believe that A', and it seems to me that A and A' are contradictory, then I will probably begin to feel uncomfortable. Several remedies for this discomfort are open to me. If I can, I may well dismiss A' outright; if not, I shall have to come up with ways in which A and A' do not contradict one another after all. Whatever I do, my conviction that A will predispose me to accept the analogous propositions that B, C, D,—and to reject the propositions that B', C', D', which, though they may not quite contradict A, all the same sit jarringly with it.
It is the task of the descriptive logician to reconstruct these processes—to dispose the belief system’s propositional content into the most coherent logical arrangement we can. The result will not necessarily resemble a succession of well-tempered Aristotelian syllogisms: it is more likely to involve steps such as ‘in a world where A, it seems plausible that B’ or ‘when I say that A, everyone thinks I mean that B’ (Towards a Science, §31). But, if the propositions that interest us actually are logically dependent on one another, then the relationships involved are capable of being reconstructed. This procedure relates to existing philosophical logic much as descriptive grammar relates to normative grammar. Normative logic is interested in how we must reason: given the propositions that all men are mortal, and that Socrates is a man, we must conclude that Socrates is mortal—and, if there are people who accept both premises and nonetheless go on to exclaim, in voices ringing with certainty, ‘yet Socrates can never die!’, then the normative logician’s concern begins and ends with pointing out that these enthusiasts have apparently made some kind of slip. Descriptive logic is interested in the enthusiasts.
There is no guarantee whatsoever that the reconstruction we produce will be an accurate one. On the contrary: descriptive logic’s claim to be scientific rests precisely on the ease with which a proposed analysis can be challenged, criticized, and amended, once it is set out in proper descriptive logical form.
To the extent that our reconstruction is accurate, however, it will enable us to see which are the central and determining propositions—like A in the imaginary example given above,—and which are merely secondary. For the first time, we shall be able to form a sense of what each proposition is doing and what the terms that occur in it mean. We shall also, by the same token, have identified which propositions are the most heavily charged with affect (emotion). It is sometimes not too difficult to assess quite precisely the relative magnitude of affect associated with individual propositions: in cases like that of A and A', where believers themselves feel two propositions to be only awkwardly compatible, the existence of a group of dependent propositions serving to soothe the contradiction and to render it bearable is a sure sign that neither proposition could be easily dispensed with. We shall thus be in a position where we can make informed use of the abundant evidence for the quality and nature of this affect that can be derived from artistic works created by and for adherents of the belief system we are studying.
Descriptive logic paves the way for a fundamentally new kind of comparative method in the study of belief systems—one where what is compared is not their content but their logical form. Comparison of content was never very reliable, even in its own terms: it was too hard to be confident that the elements being compared were really comparable. And descriptive logic makes clear that the same propositional doctrine can be doing quite unrelated jobs in different belief systems. On the other hand, two belief systems may have entirely different content, address entirely different questions, and yet have the same logical form (Towards a Science, §§72, 88). Eventually, the comparative application of descriptive logic promises to lead to a general classification of belief systems on the basis of their logic—which, in turn, will make it possible to frame theories about the kinds of logical steps that characterize belief systems of different types. The science of belief systems is still in its infancy, however, and we would be ill-advised to rush into constructing overall schemas before the necessary evidence is in place.