1. My purpose in this essay is to examine some aspects of the debate about the 'linguistic-conventionalist' account of the status of logical principles. I do not intend either to support or to attack the conventionalist position.
2. This problem is by no means compact and self contained. It is intimately related to questions about the nature of propositions and the theory of meaning. Accordingly I would like to begin by examining what meaning is and how our answer to this question can affect the conventionalist issue. Meaning is concept which, as used normally, has a range of applications, and a variety of senses. The sorts of things which are said to have meaning, in the senses relevant to the present issue, are words, and larger linguistic units built up from words (i.e. simply consisting of sequences of words, possibly interspersed with punctuation marks), such as phrases, sentences, or even larger units. More broadly almost anything can be said to have some sort of meaning, or at least, almost any event, but we are concerned here with the meaning of pieces of language. Now, when we ask the meaning of a piece of text or an utterence, what we commonly get in reply is either something purporting to be synonymous, or if no suitable synonyms exist then we may be given an explanation of how the text or utterence is normally used, some examples perhaps of suitable contexts for its use. So the meaning of a word, phrase or sentence may be given by showing in what ways this word, phrase or sentence relates to other words, phrases and sentences. Although many elements of language can be adequately explained by giving simply their relationship to other elements of language, if this were all there was to the meanings of our languages then the languages would be adrift and without any grasp upon the world. Language gains its practical utility through words which owe their meaning primarily to their relationship with something beyond language. We may explain the meaning of a word denoting a class of physical objects by pointing out, using some sort of physical gesture, members of that class, or we may give the meaning of a colour predicate by pointing out objects which have that colour. Other words owe their meaning to correlation not with physical objects or sensory experience, but with states of mind; these words present special problems of their own which I would like to put aside for the present, so my remarks are to be understood as applying only to language not used to describe our own states of mind. This restricted portion of language consists of words and larger groupings which have meanings which either consist in their relationship to other such words or which consist in their relationship to our sensory experience. This is the way in which language refers beyond itself, it refers our to our sensory experience (and possibly also to our mental innards).
3. Now let us look at some of the different 'senses' of the word 'meaning'. The senses which interest me now are those which arise through the various context sensitivities of linguistic structures. Words, phrases, sentences, et. are context sensitive. By this I mean that their meanings, i.e. their relationships with other bits of language and their relationships (if any) with sensory experience, is not independent of the context in which they occur. For examples, personal pronouns are used to refer to people, and part at least of their meaning consists in their relationship with sensory experience, in fact they are used to pick out and say something about certain elements, or at least potential elements of our sensory experience. But their relationship with the sensory experience is not invariant, since depending on the context in which they are used they will succeed in picking our different people. Similarly sentences generally, often contain implicit references to the time at which they are uttered, and so it is reasonable to say that in some sense their meaning changes according to the time at which they are uttered. In its most complete sense then, the meaning of a piece of language depends upon wide ranging aspects of the context in which it occurs, and the meaning cannot fully be given unless we know all the relevant features of the context. However there are often many things which we can say about the meaning of something even though we have only partial knowledge, or no knowledge at all, about its context. So we can say that a sentence has meaning even if the meaning of any given utterence may vary according to the context. In accounting for the meaning of a sentence we would give other sentences which would be synonymous in all contexts (if there are any) or we might go through a list of the contexts in which the sentence can be used and perhaps give a different synonymous expressions for each of the contexts, or alternatively we could deal in the other relationships which it has with the rest of language (necessary and sufficient conditions for its use, consequences for its use, et.al.), or with our sensory experience. A complete account of the meaning of a bit of language contains in it an account of the meanings in all possible differing contexts. The meaning of a bit of language in a specified context, is simply a partial account of a more general meaning, it omits those portions of the account which refer to contexts other than that explicitly mentioned.
4. To illustrate the different senses of meaning which I have been trying to explain, I shall exemplify the following schemas, which do make sense in terms of the above account of meaning:
5. These three statements are all true if we take the following examples:
6. Now I want to look at the notion of 'proposition'. First let us consider the reasons for having some such concept so that we can evaluate to what extent the concept is justified and to what extent, if any, it goes beyond its mandate. For many purposes sentences themselves are not quite what a philosopher wants to talk about. That is to say, that they are not particularly concerned with the particular sentence used as with the 'meaning' of the sentence, and wish their remarks to be understood as applying to a whole group of sentences which for their purposes are equivalent in meaning. They are concerned not with any particular language but rather with things which are expressed in various different ways in any given language, and can equally well be expressed in different languages. Now for this purpose we could use say the word 'statement' to refer to an equivalence class of sentences, or an equivalence class of utterences (meaning by utterence here a sentence spoken in a more or less closely specified context, this differing from simply a sentence by having a more precise and determinate meaning). Talking about 'statements' in this usage would enable a philosopher to speak more generally and avoid tying his remarks to a particular member of the large class of sentences to which they are applicable. This does not however make his remarks independent of language, it makes his remarks independent of the particular language in which the 'statements' referred to are expressed, and independent of the particular way in which they are expressed in any given language, but though he now refers to a broad class of linguistic entities, he nevertheless refers to unmistakably linguistic entities.
7. The word 'proposition' as commonly used (and indeed the more usual use of 'statement') seems to be an attempt to go one step beyond the notion of statement which I suggested above. It is an attempt to get right away from language to some 'meaning' which differs from the notion of meaning which I have expounded above in being completely extralingual. This forms a part of a theory of meaning which treats meanings as mysterious ethereal entities associated with sentences, in virtue of which sentences fulfil their roles in communication. In this scheme the equivalence of sentences is established by comparing these completely non-linguistic entities called meanings which are attached in some way to them. Two sentences are equivalent or two words synonymous if the meanings attached to them are identical. This conjures up a picture of a philosopher engaged in gazing upon these meanings rather like that of Plato contemplating the forms.
8. This attempt to escape entirely from language seems to me misguided and futile. All cognitive processes consist in the manipulation of information structures. If 'propositions' do not appear in the mind as information structures then there is no way in which we can distinguish between them, and they can be of no utility in cognitive processes. If they do appear as information structures then they are not extralingual, for any way of representing anything at all as an information structure constitutes a language (not necessarily a spoken language). If we are to get away from any given language then all we can do is translate into another language, we can only abandon language if we abandon discourse, and even thought, altogether. I am here using 'language' in a very broad sense. I do not wish to suggest (though some have) that all cognitive processes take place in the language of everyday discourse. Such languages are tailored to a specific task, that of communicating between people, and they probably bear little relationship to some of the information structures which are utilised in the myriad activities of the mind.
9. How does this debate about 'meaning' and 'proposition' relate to the linguistic-conventionalist controversy? It turns out to be of central importance to one of the most plausible arguments against the conventionalist position, for the issue is simply that of whether we allow that extralingual entities such as these play any role in determining the truth or falsity of sentences. In my account of meaning, meanings consist in relationships between bits of language and sensory experience. Truth is then determined sometimes by reference to sensory experience, and sometimes by observing that the constituents are combined in such a way as to guarantee truth without reference to experience. This is possible because under this scheme it is a part of the meaning of the word 'not' that any proposition of the form 'p or not p' is true whatever our sensory experience turns out to be. So the laws of logic are embedded in the meanings of our words and sentences. On the other account meanings are not relationships between linguistic entities which sometimes are sufficient to guarantee truth. Meanings take us out of language from the realm of sentences into that of propositions. We first step out of language by identifying the proposition which our sentence expresses and then we establish, either by reference to experience, or by reference to the laws of logic, whether or not the proposition we have lighted upon is true. But this account I find most unsatisfactory. For once we have stepped out of language then we can do nothing until we step back in again. There may be great value in stepping out of some sentences by producing a synonymous sentence which has a more transparently tautologous form, but to step out of language altogether could never be anything than a step into a very short cul-de-sac.
10. If we abandon the attempt to give 'propositions' an extralingual status then they lose much of their mystery and they are likely to become merely useful, rather than vacuously essential. One possibility would be that of adopting some special language as the language of propositions. In the latter case a suitable language might be somethings similar to that of Principia Mathematica in which the logical relationships are much more obvious and well defined. If a 'proposition' were than simply a translation of a sentence into a certain special language, then this might well play a useful role in establishing the truth of sentences, but this would be at most a matter of convenience, it would never be an essential step (no information processing task is tied to any particular language, except where it interfaces with other tasks already given; any language will do provided that it gives all the necessary information, i.e. provided that it has equivalent expressive power and precision. Some languages, however, are easier to use than others for a given purpose, more efficient and economical.
11. What I have done so far is try to cast doubt upon one sort of argument against the conventionalist position by showing that it rests upon ideas of truth and meaning which are at best suspect, which effectively prejudge the issue at hand. What I have said however, though hopefully illuminating, does not amount to a very conclusive argument in favour of conventionalism. For just as the theory of meaning of the abolutists prejudges the issue, so does mine, this time in favour of conventionalism. What I have done is to adopt a usage of 'meaning' which embeds into the meaning of terms in a language all sorts of things, including the sort of logical relationships which the terms satisfy. This hints back to the way Leibniz rendered all true statements except those asserting existence analytic, by taking to be part of the essence of substance all the predicates which that substance justified. This procedure itself is almost as suspect as the mysterious entities it is intended to avoid.