It is a commonplace of philosophical logic that there are, or appear to be, divergences in meaning between, on the one hand, at least some of what I shall call the formal devices-~, A, V, D, (Vx), (3x), (vx) (when these are given a standard two-valued interpretation)-and, on the other, what are taken to be their analogues or counterparts in natural language-such expressions as not, and, or, if, all, some (or at least one), the. Some logicians may at some time have wanted to claim that there are in fact no such divergences; but such claims, if made at all, have been somewhat rashly made, and those suspected of making them have been subjected to some pretty rough handling.

Those who concede that such divergences exist adhere, in the main, to one or the other of two rival groups, which I shall call the formalist and the informalist groups. An outline of a not uncharacteristic formalist position may be given as follows: Insofar as logicians are concerned with the formulation of very general patterns of valid inference, the formal devices possess a decisive advantage over their natural counterparts. For it will be possible to construct in terms of the formal devices a system of very general formulas, a considerable number of which can be regarded as, or are closely related to, patterns of inferences the expression of which involves some or all of the devices: Such a system may consist of a certain set of simple formulas that must be acceptable if the devices have the meaning that has been assigned to them, and an indefinite number of further formulas, many of which are less obviously acceptable and each of which can be shown to be acceptable if the members of the original set are accept-able. We have, thus, a way of handling dubiously acceptable patterns of inference, and if, as is sometimes possible, we can apply a decisionprocedure, we have an even better way. Furthermore, from a philosophical point of view, the possession by the natural counterparts of those elements in their meaning, which they do not share with the corresponding formal devices, is to be regarded as an imperfection of natural languages; the elements in question are undesirable excres-cences. For the presence of these elements has the result both that the concepts within which they appear cannot be precisely or clearly de-fined, and that at least some statements involving them cannot, in some circumstances, be assigned a definite truth value; and the indef-initeness of these concepts not only is objectionable in itself but also leaves open the way to metaphysics-we cannot be certain that none of these natural language expressions is metaphysically "loaded." For these reasons, the expressions, as used in natural speech, cannot be regarded as finally acceptable, and may turn out to be, finally, not fully intelligible. The proper course is to conceive and begin to construct an ideal language, incorporating the formal devices, the sentences of which will be clear, determinate in truth value, and certifiably free from metaphysical implications; the foundations of science will now be philosophically secure, since the statements of the scientist will be expressible (though not necessarily actually expressed) within this ideal language. (I do not wish to suggest that all formalists would accept the whole of this outline, but I think that all would accept at least some part of it.)

To this, an informalist might reply in the following vein. The philosophical demand for an ideal language rests on certain assumptions that should not be conceded; these are, that the primary yardstick by which to judge the adequacy of a language is its ability to serve the needs of science, that an expression cannot be guaranteed as fully intelligible unless an explication or analysis of its meaning has been provided, and that every explication or analysis must take the form of a precise definition that is the expression or assertion of a logical equivalence. Language serves many important purposes besides those of scientific inquiry; we can know perfectly well what an expression means (and so a fortiori that it is intelligible) without knowing its analysis, and the provision of an analysis may (and usually does) consist in the specification, as generalized as possible, of the conditions that count for or against the applicability of the expression being ana-lyzed. Moreover, while it is no doubt true that the formal devices are especially amenable to systematic treatment by the logician, it remains the case that there are very many inferences and arguments, expressed in natural language and not in terms of these devices, which are nevertheless recognizably valid. So there must be a place for an unsimplified, and so more or less unsystematic, logic of the natural counterparts of these devices; this logic may be aided and guided by the simplified logic of the formal devices but cannot be supplanted by it. Indeed, not only do the two logics differ, but sometimes they come into conflict; rules that hold for a formal device may not hold for its natural counterpart.

On the general question of the place in philosophy of the reformation of natural language, I shall, in this essay, have nothing to say. I shall confine myself to the dispute in its relation to the alleged diver-gences. I have, morcover, no intention of entering the fray on behalf of either contestant. I wish, rather, to maintain that the common assumption of the contestants that the divergences do in fact exist is (broadly speaking) a common mistake, and that the mistake arises from inadequate attention to the nature and importance of the conditions governing conversation. I shall, therefore, inquire into the gen-cral conditions that, in one way or another, apply to conversation as such, irrespective of its subject matter. I begin with a characterization of the notion of "implicature."