Philosophy must be of some use and we must take it seriously; it must clear our thoughts and so our actions. Or else it is a disposition we have to check, and an inquiry to see that this is so; i.e. the chief proposition of philosophy is that philosophy is nonsense. And again we must then take seriously that it is nonsense, and not pretend, as Wittgenstein does, that it is important nonsense !
In philosophy we take the propositions we make in science and everyday life, and try to exhibit them in a logical system with primitive terms and definitions, etc. Essentially a philosophy is a system of definitions or, only too often, a system of descriptions of how definitions might be given.
I do not think it is necessary to say with Moore that the definitions explain what we have hitherto meant by our propositions, but rather that they show how we intend to use them in future. Moore would say they were the same, that philosophy does not change what anyone meant by 'This is a table'. It seems to me that it might; for meaning is mainly potential, and a change might therefore only be manifested on rare and critical occasions. Also sometimes philosophy should clarify and distinguish notions previously vague and confused, and clearly this is meant to fix our future meaning only1. But this is clear, that the definitions are to give at least our future meaning, and not merely to give any pretty way of obtaining a certain structure.
I used to worry myself about the nature of philosophy through excessive scholasticism. I could not see how we could understand a word and not be able to recognize whether a proposed definition of it was or was not correct. I did not realize the vagueness of the whole idea of understanding, the reference it involves to a multitude of performances any of which may fail and require to be restored. Logic issues in tautologies, mathematics in identities, philosophy in definitions; all trivial but all part of the vital work of clarifying and organizing our thought.
If we regard philosophy as a system of definitions (and elucidations of the use of words which cannot be nominally defined), the things that seem to me problems about it are these:

(1) What definitions do we feel it up to philosophy to provide, and what do we leave to the sciences or feel it unnecessary to give at all ?

(2) When and how can we be content without a definition but merely with a description of how a definition might be given ? [This point is mentioned above.]

(3) How can philosophical enquiry be conducted without a perpetual petitio principii ?

(1) Philosophy is not concerned with special problems of definition but only with general ones: it does not propose to define particular terms of art or science, but to settle e.g. problems which arise in the definition of any such term or in the relation of any term in the physical world to the terms of experience.
Terms of art and science, however, must be defined, but not necessarily nominally; e.g. we define mass by explaining how to measure it, but this is not a nominal definition; it merely gives the term 'mass' in a theoretical structure a clear relation to certain experimental facts. The terms we do not need to define are those which we know we could define if need arose, like ' chair ', or those which like ' clubs ' (the suit of cards) we can translate easily into visual or some other language, but cannot conveniently expand in words.
(2) The solution to what we called in (1) a 'general problem of definition ' is naturally a description of definitions, from which we learn how to form the actual definition in any particular case. That we so often seem to get no actual definitions, is because the solution of the problem is often that nominal definition is inappropriate, and that what is wanted is an explanation of the use of the symbol.
But this does not touch what may be supposed to be the real difficulty under this head (2); for what we have said applies only to the case in which the word to be defined being merely described (because treated as one of a class), its definition or explanation is also, of course, merely described, but described in such a way that when the actual word is given its actual definition can be derived. But there are other cases in which the word to be defined being given, we are given in return no definition of it but a statement that its meaning involves entities of such-and-such sorts in such-and-such ways, i.e. a statement which would give us a definition if we had names for these entities.
As to the use of this, it is plainly to fit the term in connection with variables, to put it as a value of the right complex variable; and it presupposes that we can have variables without names for all their values. Difficult questions arise as to whether we must always be able to name all the values, and if so what kind of ability this means, but clearly the phenomenon is in some way possible in connection with sensations for which our language is so fragmentary. For instance, ' Jane's voice' is a description of a characteristic of sensations for which we have no name. We could perhaps name it, but can we identify and name the different inflexions of which it consists ?
An objection often made to these descriptions of definitions of sensory characteristics is that they express what we should find on analysis, but that this kind of analysis changes the sensation analysed by developing the conplexity which it pretends merely to discover. That attention can change our experience is indubitable, but it seems to me possible that sometimes it reveals a pre-existing complexity (i.e. enables us to symbolize this adequately), for this is compatible with any change in incidental facts, anything even except a creation of the complexity.
Another difficulty with regard to descriptions of definitions is that if we content ourselves with them we may get simply nonsense by introducing nonsensical variables, e.g. described variables such as 'particular' or theoretical ideas such as ' point'. We might for instance say that by 'patch' we mean an infinite class of points; if so we should be giving up philosophy for theoretical psychology. For in philosophy we analyse our thought, in which patch could not be replaced by infinite class of points: we could not determine a particular infinite class extensionally; 'This patch is red ' is not short for ' a is red and b is red etc....' where a, b, etc., are points. (How would it be if just a were not red ?) Infinite classes of points could only come in when we look at the mind from outside and construct a theory of it, in which its sensory field consists of classes of coloured points about which it thinks.
Now if we made this theory about our own mind we should have to regard it as accounting for certain facts, e.g. that this patch is red; but when we are thinking of other people's minds we have no facts, but are altogether in the realm of theory, and can persuade ourselves that these theoretical constructions exhaust the field. We then turn back on our own minds, and say that what are really happening there are simply these theoretical processes. The clearest instance of this is, of course, materialism. But many other philosophies, e.g. Carnap's, make just the same mistake.
(3) Our third question was how we could avoid Petitio principii, the danger of which arises somewhat as follows:?
In order to. clarify my thought the proper method seems to be simply to think out with myself ' What do I mean by that ? ' ' What are the separate notions involved in this term ? ' ' Does this really follow from that ? ' etc., and to test identity of meaning of a proposed definiens and the definiendum by real and hypothetical examples. This we can often do without thinking about the nature of meaning itself; we can tell whether we mean the same or different things by ' horse ' and ' pig ' without thinking at all about meaning in general. But in order to settle more complicated questions of the sort we obviously need a logical structure, a system of logic, into which to bring them. These we may hope to obtain by a relatively easy previous application of the same methods; for instance, it should not be difficult to see that for either not-p or not-q to be true is just the same thing as for not both p and q to be true. In this case we construct a logic, and do all our philosophical analysis entirely unself-consciously, thinking all the time of the facts and not about our thinking about them, deciding what we mean without any reference to the nature of meanings. [Of course we could also think about the nature of meaning in an unselfconscious way; i.e. think of a case of meaning before us without reference to our meaning it.] This is one method and it may be the right one; but I think it is wrong and leads to an impasse, and I part company from it in the following way.
It seems to me that in the process of clarifying our thought we come to terms and sentences which we cannot elucidate in the obvious manner by defining their meaning. For instance, variable hypotheticals and theoretical terms we cannot define, but we can explain the way in which they are used, and in this explanation we are forced to look not only at the objects which we are talking about, but at our own mental states. As Johnson would say, in this part of logic we cannot neglect the epistemic or subjective side.
Now this means that we cannot get clear about these terms and sentences without getting clear about meaning, and we seem to get into the situation that we cannot understand e.g. what we say about time and the external world without first understanding meaning and yet we cannot understand meaning without first understanding certainly time and probably the external world which are involved in it. So we cannot make our philosophy into an ordered progress to a goal, but have to take our problems as a whole and jump to a simultaneous solution; which will have something of the nature of a hypothesis, for we shall accept it not as the consequence of direct argument, but as the only one we can think of which satisfies our several requirements.
Of course, we should not strictly speak of argument, but there is in philosophy a process analogous to ' linear inference ' in which things become successively clear; and since, for the above reason, we cannot carry this through to the end, we are in the ordinary position of scientists of having to be content with piecemeal improvements: we can make several things clearer, but we cannot make anything clear.
I find this self-consciousness inevitable in philosophy except in a very limited field. We are driven to philosophize because we do not know clearly what we mean; the question is always' What do I mean by x ? ' And only very occasionally can we settle this without reflecting on meaning. But it is not only an obstacle, this necessity of dealing with meaning; it is doubtless an essential clue to the truth. If we neglect it I feel we may get into the absurd position of the child in the following dialogue: ' Say breakfast.' ' Can't.' ' What can't you say ? ' ' Can't say breakfast,'
But the necessity of self-consciousness must not be used as a justification for nonsensical hypotheses; we are doing philosophy not theoretical psychology, and our analyses of our statements, whether about meaning or anything else, must be such as we can understand.
The chief danger to our philosophy, apart from laziness and woolliness, is scholasticism, the essence of which is treating what is vague as if it were precise and trying to fit it into an exact logical category. A typical piece of scholasticism is Wittgenstein's view that all our everyday propositions are completely in order and that it is impossible to think illogically. (This last is like saying that it is impossible to break the rules of bridge because if you break them you are not playing bridge but, as Mrs C. says, not-bridge.) Another is the argumentation about acquaintance with before leading to the conclusion that we perceive the past. A simple consideration of the automatic telephone shows that we could react differently to AB and BA without perceiving the past, so that the argument is substantially unsound. It turns on a play with ' acquaintance ' which means, first, capacity to symbolize and, secondly, sensory perception. Wittgenstein seems to equivocate in just the same way with his notion of ' given.'

1 But in so far as our past meaning was not utterly confused, philosophy will naturally give that, too. E.g. that paradigm of philosophy, Russell's theory of descriptions.

In: Ramsey, Frank Plumpton: Philosophical Papers (ed. by D.H.Mellor). Cambridge 1990, pp.1-7