In Book IV of the Essay concerning ‘Human Understanding’ Locke maintains that there are degrees of assent. In Section 2 of Book IV, Ch. 15 (‘Of Probability’) he says that as there are degrees of probability ‘from the very neighbourhood of certainty and demonstration quite down to improbability and unlikeliness, even to the confines of impossibility’, so also there are ‘degrees of assent from full confidence and assurance, quite down to conjecture, doubt and distrust’. This epistemological doctrine that assent has degrees is the presupposition of his ‘Ethics of Belief’, which he formulates later in the same chapter. ‘The mind, if it will proceed rationally, ought to examine all the grounds of probability, and see how they make more or less for or against any proposition, before it assents to or dissents from it; and, upon a due balancing of the whole, reject or receive it, with a more or less firm assent, proportionably to the preponderancy of the greater grounds of probability on one side or the other.’ (Book IV, ch. 15, Section 5.)
At the beginning of Ch. 16 he puts it thus: ‘The grounds of probability… as they are the foundation on which our assent is built, so are they also the measure whereby its several degrees are or ought to be regulated.’ He returns to the subject in the celebrated chapter Of Enthusiasm, where the disposition to regulate the degree of our assent in proportion to the strength of the evidence is connected with the love of truth. One unerring mark of a love of truth for truth's sake, he says, is ‘the not entertaining any proposition with a greater assurance than the proofs it is built upon will warrant’. (Book IV, Ch. 19, Section 1.)
The ‘grounds of probability’ which ‘make for or against a proposition’ are what we should now call the evidence for or against that proposition. And if we mean by ‘the evidence for a proposition’ what we often do mean—namely, the evidence which there is for it on the whole when due account has been taken of adverse evidence, if any1—we can simplify Locke's rather complicated formulation of his ‘Ethics of Belief’ as follows: the degree of our assent to a proposition ought to be proportioned to the strength of the evidence for that proposition.
Both these doctrines of Locke—both the doctrine that assent has degrees, and the ‘Ethics of Belief which he formulates in terms of it—may strike us as just obvious common sense. Do we not have familiar expressions in ordinary language for indicating degrees of assent, for example, ‘suspecting that’, ‘thinking that’ (the ordinary phrase for expressing an opinion), ‘being almost sure but not quite’, ‘being absolutely sure’? Another and equally familiar way of indicating the degree of our assent to a proposition is by referring to the degree of surprise we should feel if the proposition turned out to be false. I should be just a little surprised if John did not come to see me this afternoon; or I should be a good deal surprised if he did not, or I should be greatly surprised if he did not; or I should be astonished if he did not, and should even be willing to eat my hat. Alternatively, we can say that we should not be surprised if the proposition turned out to be true; and perhaps the neatest way of expressing the lowest degree of assent (traditionally called ‘suspecting’) is to say we should not be altogether surprised if the proposition assented to were true. When one hears the loud alarm-note of the blackbirds in the twilight, as they flit round a large hawthorn bush, one may say ‘I should not be altogether surprised if there were an owl roosting in the bush’.
Moreover, do we not all agree with Locke that a lower degree of assent may be justified when a higher degree would not be? Surely we all admit that one may be ‘entitled’, as we say, to think that p (to hold the opinion that p) when we are not entitled to be almost sure of it, and still less to be absolutely sure of it? Sometimes too, though only rarely, we think that a man is entitled to assent to a proposition more firmly than he actually does. If an educated Englishman were to say in 1967 ‘I suspect that Queen Victoria died a good many years ago’, we should not deny that he was entitled to suspect so, but we should point out that he was entitled to assent to the proposition a good deal more firmly than he did. And surely we all think that what ‘entities’ a person to assent in this degree or in that, or ‘justifies’ him in doing so, is just what Locke, says it is, namely the strength, greater or less, of the evidence which that person has for the proposition assented to?
Of course, we do in fact often give a firmer or stronger assent to a proposition than we are entitled to give, and occasionally a weaker one. Locke's rule that the degree of our assent ought to be proportional to the strength of the evidence is often broken or disregarded. But it is significant that when we are criticized for breaking it, we defend ourselves by claiming (rightly or wrongly) that the evidence for the proposition we assented to—or at least the evidence we ourselves had—was stronger than our critic said it was. It never occurs to us to challenge the authority of the rule itself. We do not say ‘No doubt, the evidence is rather weak, but we are entitled to be absolutely convinced all the same’. Or if we do occasionally say, in a moment of exasperation ‘Well, that is what I think, evidence or no evidence’, we are willing to admit that we are being consciously unreasonable.
Locke's two doctrines, then—that assent has degrees, and that the degree of assent ought to be proportional to the strength of the evidence—may easily seem platitudinous. One would be happy to think that they are. For if they are false, our human condition must be both more miserable and more intellectually disreputable than we commonly suppose. Nevertheless, both these doctrines are vigorously challenged by one of the most celebrated writers on our subject, Cardinal Newman, in his Grammar of Assent, Ch. VI, ‘Assent considered as unconditional’.3 In this lecture, I shall discuss Newman's criticisms of Locke; but it will also be necessary to consider the positive doctrine which they are intended to support. This is that assent is unconditional, whereas inference is conditional. It is strange that this interesting but rather difficult chapter has received so little attention from epistemologists, and still less (so far as I know) from expositors of Locke. The issues which it raises are of the greatest importance if we wish to understand what belief is, and what part it plays in human life. I shall try to show that Locke is more nearly right than Newman is on most of the points about which they differ.
Let us follow this excellent example; for no-one, and certainly no Oxford man, should criticize Newman without praising him. Newman is surely one of the masters of introspective description. No writer on belief has a keener eye for the phenomenological facts than he has, or shows more skill and delicacy of touch in disentangling them from one another; and his argument is illustrated throughout by concrete examples taken straight from life. Moreover, Newman is one of the masters of English prose. The power, and the charm, of his style are so compelling that the reader soon becomes their willing captive; and it seems ungrateful, almost ungracious, to question what has been so felicitously said.
‘Bordering Near Upon Certainty’
It must be admitted too that in the first of his criticisms (Grammar of Assent, pp. 121–124) Newman has detected an important inconsistency in Locke. In Essay, Book IV, Ch. 15, Section 2, Locke says that some propositions ‘border so near upon certainty that we make no doubt at all about them, but assent to them as firmly and act according to that assent as resolutely as if they were infallibly demonstrated’. But if such propositions only border near upon certainty, without actually reaching it, how can we be justified (on Locke's own principles) in making no doubt at all about them, and assenting to them as firmly as if they were infallibly demonstrated? Surely we should be violating the precept which Locke himself has laid down? We are here ‘entertaining a proposition with a greater assurance than the proofs it is built upon will warrant’ and thereby showing ourselves not to be lovers of truth for truth's sake. We are completely sure, when according to Locke's own Ethics of Belief we are only entitled to be very nearly sure. It is true that Locke does not say we are entitled to make no doubt at all about these propositions. He only says that we do not in fact make any doubt about them. But still, he seems to see nothing wrong with this, and to have forgotten all about the cautious precepts of his own Ethics of Belief—precepts which Newman himself emphatically rejects.
Can anything be said in Locke's defence? Something perhaps, but not much. He is surely justified in maintaining that we may (and often should) ‘act as resolutely as if the propositions were infallibly demonstrated’. There is nothing in this which is incompatible with a love of truth for truth's sake. Indeed, if we did not act so, we might well be neglecting our love for our neighbour, and also perhaps neglecting our own long-term interest. But if the propositions he has in mind do only border near upon certainty, without actually reaching it, we surely have no right to assent to them as firmly as if they were infallibly demonstrated. Perhaps, however, he was not wholly clear about the difference between propositions which are logically certified (e.g. the conclusion of a mathematical proof) and propositions which are empirically certified. The phrase ‘as if they were infallibly demonstrated’ suggests that he was not.
‘The Testimony of Psychological Facts’
As we have seen, Locke maintains two distinct, though connected, theses. One is concerned with what assent is, the other with what it ought to be. At first sight Newman seems not to distinguish clearly between the two. In some of the most brilliantly written pages of the Grammar of Assent (pp. 124–130) he offers a number of arguments to show that men do not in fact assent in the way Locke says they ought to. For example, they continue to assent to a proposition, when they have forgotten the evidence (p. 126). Assent, once given, may cease ‘while the reasons for it and the inferential act which is the recognition of those reasons, are still present and in force. Our reasons may seem to us as strong as ever, yet they do not secure our assent. Our beliefs, founded on them, were and are not; we cannot perhaps tell when they went’ (p. 126). Sometimes, in spite of strong and convincing arguments, assent is never given (p. 127). Sometimes, again, ‘good arguments, and really good as far as they go, and confessed by us to be good, nevertheless are not strong enough to incline our minds ever so little to the conclusion at which they point’. Here we do not ‘assent a little in proportion to those arguments’, which in Locke's view would be the reasonable thing to do. ‘On the contrary, we throw the full onus probandi on the side of the conclusion, and we refuse to assent to it at all until we can assent to it altogether’ (p. 127). Again, what Newman calls moral motives may ‘hinder assent to conclusions which are logically unimpeachable’. We all know that ‘a man convinced against his will is of the same opinion still’ (p. 128). And finally, even in mathematics, ‘argument is not always able to command our assent, even though it be demonstrative’—at any rate when the demonstration is long and intricate (p. 128).
We cannot dispute the facts which Newman here points out, and we cannot withhold our admiration for the psychological acumen he shows in describing them. But are they relevant? Can an ‘ought’ proposition be refuted by factual evidence? Can a doctrine about the way in which it would be reasonable to assent be disproved by pointing out that very often we do not in fact assent in that way? Surely such arguments only show that we often assent, or withhold assent, unreasonably: a conclusion which Locke never intended to dispute and one which he himself emphasises.
But this objection is not altogether fair to Newman. The arguments I have quoted must be taken in their context. What Newman wishes to show is that assent does not admit of degrees at all. And if it does not, it is absurd to lay down rules about the degree which it ought to have. ‘Ought’ implies ‘can’. Such rules would be null and void, because it would not be in our power to follow them, if assent, by its very nature, must be a matter of all or nothing, as Newman holds that it is. This, I take it, is the point of an eloquent passage on p. 124 ‘He [Locke] takes a view of the human mind, in relation to inference and assent, which to me seems theoretical and unreal… he consults his own ideal of how the mind ought to act, instead of interrogating human nature, as an existing thing, as it is found in the world. Instead of going by the testimony of psychological facts,5 and thereby determining our constitutive faculties and our proper condition, and being content with the mind as God has made it, he would form men as he thinks they ought to be formed, into something better and higher, and calls them irrational and indefensible if (so to speak) they take to the water, instead of remaining under the narrow wings of his own arbitrary theory’ (p. 124).
The testimony of psychological facts is, after all, relevant if it shows that Locke's view about the way we ought to assent would only make sense if assent were something quite different from what it actually is. As Newman puts it in the next paragraph ‘The first question which this theory leads me to consider is whether there is such an act of the mind as assent at all’ (p. 124). ‘If there is’, he goes on ‘it is plain it ought to show itself unequivocally as such, as distinct from other acts.’ But on Locke's view this is just what it could not do. For if Locke were right, assent would be nothing but inference. ‘When I assent, I am supposed, it seems, to do precisely what I do when I infer, or rather not quite so much, but something which is included in inferring; for while the disposition of my mind towards a given proposition is identical in assent and in inference, I merely drop the thought of the premisses when I assent, though not of their influence on the proposition inferred. This, then, and no more after all, is what nature prescribes; and this, and no more than this, is the conscientious use of our faculties, so to assent forsooth as to do nothing but infer’ (pp. 124–5). The arguments on pp. 125–9 which I quoted above are intended to show that assent is distinct from inference, on the ground that one can be absent when the other is present, and that the two do not vary concomitantly.
On p. 131 Newman quotes the following passage from Gambier, a writer, he says, ‘who claims our respect from the tone and drift of his work’. ‘Moral evidence may produce a variety of degrees of assent, from suspicion to moral certainty.… For a few of these degrees, though but for a few, names have been invented. Thus when the evidence on one side preponderates a very little, there is ground for suspicion or conjecture. Presumption, persuasion, belief, conclusion, conviction, moral certainty—doubt, wavering, distrust, disbelief—are words which imply an increase or decrease of this preponderancy.’
Admirable remarks, the reader may think, and a very clear expression of what all sensible men have always taken to be obvious. That is not Newman's view. This passage, he says, illustrates what he himself has been insisting upon ‘that in teaching degrees of assent, we tend to destroy assent, as an act of the mind, altogether’. Gambier's ‘assents’ are only inferences, and ‘assent’ as Gambier uses the word is ‘a name without a meaning, the needless repetition of an inference’,6 The suspicion, conjecture, persuasion, belief, etc., of which Gambier speaks ‘are not ‘assents’ at all; they are simply more or less strong inferences of a proposition’. They are ‘not variations of assent to an inference, but assents to a variation in inferences. When I assent to a doubtfulness or to a probability, my assent, as such, is as complete as if I assented to a truth; it is not a certain degree of assent.7 And in like manner, I may be certain of an uncertainty’ (p. 132). We must conclude, then, that ‘if human nature is to be its own witness, there is no medium between assenting and not assenting. Locke's theory of the duty of assenting more or less according to degrees of evidence is invalidated by the testimony of high and low, young and old, ancient and modern, as continually given in their ordinary sayings and doings’ (p. 133).
Newman'S Conception of Inference
It can now be seen that the notion of inference plays a central part in Newman's criticism of Locke. If he can show that Locke has confused assent with inference, he thinks the doctrine of degrees of assent will lose any plausibility it has; and with the collapse of that doctrine, Locke's Ethics of Belief will collapse too. This is what Newman has in mind in the passage already quoted, where he says that Locke's view of the human mind in relation to inference and assent is ‘theoretical and unreal’ and accuses him of not being content with the mind as God has made it (p. 124). It would indeed be ‘theoretical and unreal’ to lay down the rule that the degree of our assent ought to vary with the strength of the evidence, if assent does not admit of degrees at all.
At first sight it appears strange to his modern readers that Newman should think it relevant to devote so much attention to inference when he is discussing the nature of assent. And if the concept of inference is introduced into the discussion, we should expect that his argument would be quite different from the one he actually uses, when he points out, for instance, that our assents may continue ‘without the presence of the inferential acts upon which they were originally elicited’ (pp. 125–6). Instead, the modern reader would have expected him to point out that when we make an inference we have to assent to the premiss in order to draw the conclusion, and that this is indeed the crucial difference between inferring (‘because p, therefore q’) and merely noticing an implication or entailment (‘if p, then q’). Consequently—we might expect him to argue—the notion of assenting is logically prior to the notion of inferring; inferring itself has to be defined in terms of assenting, and therefore it must be wrong to define assenting in terms of inferring, as Locke, on Newman's interpretation of him, is trying to do.
But Newman is so far from using this argument, that he might rather be accused of paying too little attention to the distinction between inferring and noticing an implication. For he insists over and over again that inference is conditional. Indeed, that is why he thinks it so very important to distinguish between inference and assent. Inference is always conditional; assent, on the contrary is always unconditional, and indeed its unconditionality is the most important thing about it. That is what Newman wishes to maintain. If we fail to distinguish assent from inference, or regard it as a mere reduplication or ‘echo’ of inference (as Newman thinks Locke did) we shall be led to suppose that assent itself is conditional, and this would be tantamount to abolishing assent altogether (pp. 124–5). But once the confusion between inference and assent is removed, once we see that assent is a ‘substantive act’8 in its own right, we are freed from the obsession which prevents us from considering the actual phenomenological facts; and when we do consider them, we shall find that assent, as it actually occurs in human life, is always unconditional.
But why should anyone be even tempted to confuse assent with inference? Presumably because the two do ordinarily go together. Assent, Newman thinks, is normally ‘elicited’ by an inference. What sort of an inference can it be? It is true, of course, that before assenting to a proposition p, we do as a rule pay some attention to evidence relevant to that proposition; and, if we are reasonable, we refrain from assenting unless or until we find that p is more probable than not-p in relation to this evidence. But it is rather strange to regard this procedure as an inference. It is more naturally described as a procedure of estimation, or, metaphorically, or ‘weighing’ (we ‘weigh’ the evidence for and against the proposition). Certainly this is a rational activity if anything is. Indeed, it is perhaps the one most characteristic of a reasonable man; and certainly it could be described as ‘finding reasons for’ assenting, or at least as seeking for them. But it is not very natural to think of it as an activity of reasoning or inference.
Nevertheless, when this process of estimation has been successfully completed, its results could be expressed in the form of an inference. ‘Because there are the facts A, B and C, the proposition p has a greater probability than not-p.’ ‘Because the barometer has fallen, and the temperature has risen a little, and the wind has backed to south west, and the sky is covered with alto-stratus cloud, it is therefore much more probable than not that it will rain within a few hours.’ We do not usually put it like that. We usually say ‘my reasons for thinking that p are these’ and then proceed to state them. But still, we could use this ‘because… therefore’ language—‘because there are facts A, B and C, p is therefore more probable than not-p’—and what we express by saying this could no doubt be described as ‘an inferential act’.
It must be inferences of this kind that Newman has in mind. For these are the inferences which may be supposed to precede our assents, and if someone is tempted to confuse assent with inference, inferences of this kind must surely be the source of the temptation. This interpretation of Newman's treatment of ‘inference’ is supported by what Newman himself says on p. 126. As we have seen already, he argues that inference and assent must be distinct acts, because either may be absent when the other is present, and the two do not vary concomitantly. One of his arguments is this: ‘Sometimes assent fails, while the reasons for it and the inferential act which is the recognition of those reasons are still present and in force.’ (p. 126. My italics.)
If that is what the ‘inferential act’ is, we can now see what Newman means when he distinguishes so carefully between assent and inference. Let us suppose for the present that our reasons for thinking that p are less than conclusive, as they frequently are when p is a proposition concerning some concrete matter of fact (this is the situation in which Newman is mainly interested). Newman's point is this; it is one thing to recognize reasons for assenting to p, and another thing to assent to it. For instance, it is one thing to recognize that there are reasons for believing that it will rain, and another to believe that it will rain. Of course, the proposition that there are such and such reasons for assenting to p can itself be assented to. Newman admits this. In his language, we are then assenting to an inference.9 But the proposition which is assented to in this act of assent is not p itself, but the more complex proposition ‘There are such and such reasons for assenting to p’. Having recognized that there are reasons for assenting to p-moderately strong ones, let us suppose—we may then proceed to assent to p. But equally we may not. And if we do, it is a different ‘mental act’. The reader may dislike the language of ‘mental acts’. But still we must admit that assenting is not the same as recognizing reasons for assent. However closely they go together in a reasonable man, even in him they are still different; and we are not always reasonable.
If this is what Newman means when he insists upon the difference between inference and assent, he is clearly right and is drawing our attention to an important distinction. At any rate, he is right where the reasons for assenting to p are less than conclusive and recognized by us to be so. It does not follow that he is also right in thinking that Locke overlooked this distinction, nor that he is right in rejecting the doctrine of degrees of assent. If there is something about inference (in Newman's sense of the word) which admits of degrees, it may still be true that there is something about assent which admits of them too; and it may still be true that in a reasonable man the degree of assent varies with the strength of the reasons he recognizes for assenting.
Newman'S Use of the Word ‘Conditional’
But why should such inferences, or indeed any others, be regarded as conditional? In the terminology to which we are now accustomed, a conditional sentence is of the form ‘if… then…’. But these inferences, and indeed all others, are of the form ‘because… therefore…’. And what is called a probable inference is of this form as much as any other inference. When the conclusion is of the form ‘p has such and such a degree of probability’, this does not turn the ‘because’ into ‘if’. Again, we sometimes speak of strong and weak inferences, as Newman himself does. We might say ‘because there are the facts A, B and C, the proposition p is just appreciably more probable than not-p’; for example, you still have a quarter of a gallon in the tank, so there is just a chance that we shall get home. This would presumably be called a ‘weak’ inference. The weakness, however, does not lie in the inference itself, but only in the conclusion inferred. And this weakness, if we call it such, has no tendency to turn the ‘because’ into an ‘if’. What will turn the ‘because’ into an ‘if’ is any uncertainty we may have about the premiss. Perhaps it is not certain that you have as much as a quarter of a gallon in the tank; then I shall make a conditional statement ‘If you have as much as a quarter of a gallon, there is just a chance that we shall get home’. But in that case I am no longer making an inference. I am only noticing and pointing out an implication. At the most, I am only pointing out that q would be inferrible from p, if we could be sure that p is true. I am certainly not inferring q from p. How happy I should be if I could!
Can it be that Newman has confused ‘if… then’ with ‘because… therefore’? Unfortunately he does show some signs of doing so. There is a curious passage on pp. 137 ad fin—138 of the Grammar of Assent, where he remarks that though we may ‘include a condition in the proposition to which our assent is given’ this does not make the assent itself conditional; and obviously it does not. When we assent to ‘if p then q’ the proposition assented to is conditional, but it does not follow that the assenting is conditional. But he then proceeds to illustrate his point by two examples. In the first, the proposition assented to is ‘If this man is in a consumption, his days are numbered’. In the second it is ‘Of this consumptive patient the days are numbered’. He then says that the two propositions are equivalent, though the second is not stated in the conditional form. But plainly they are not equivalent. The first is of the form ‘if p then q’. The second is of the form ‘because p therefore q’. The second proposition (‘Of this consumptive patient…’) could only be made equivalent to the first by inserting a saving clause into it: ‘Of this consumptive patient, assuming that he is indeed a consumptive, the days are numbered.’ Nevertheless, in the latter part of the same paragraph Newman himself notices the difference I am pointing out. His example is ‘There will be a storm soon, for the mercury falls’; and he says that here ‘besides assenting to the connexion of the propositions we may assent also to “The mercury falls” and to “There will be a storm”’ (p. 138). The only objection one can make to this concerns the word ‘may’. If we say ‘for the mercury falls’ we not only may, but do, assent to both the connected propositions, as well as to the connection between them.
It is interesting that in the same paragraph Newman uses the Latin technical term inferentia to mean the connection between the antecedent and the consequent in a conditional statement. ‘We may give our assent not only to the inferentia in a complex conditional proposition but to each of the simple propositions of which it is made up, besides.’ An inferentia is what we should now call an implication or an entailment, but not what we should now call an inference. An inferentia is of the form ‘q follows from p’. And of course we may assent to ‘q follows from p’ without assenting either to p or to q themselves. That is precisely what we do when we say ‘if p then q’.
Can this be the reason why Newman holds that inferences are as such conditional? Can he be using the English word ‘inference’ as the translation of the Latin inferentia? There is indeed a rather old-fashioned English usage in which ‘p infers q’ does mean ‘q follows from p’. But in this usage ‘infer’ is an impersonal verb, like ‘entail’ or ‘have as a consequence’. Its grammatical subject is a proposition, not a person. An inferentia, likewise, is a connection between propositions. But Newman repeatedly speaks of an act of inference, or an inferential act. The subject of the verb to ‘infer’ is in his usage a person, as it is in our ordinary usage today. Much as he wishes to insist on the distinction between inference and assent, he clearly holds that they have at least this much in common, that they are both mental acts. Indeed, if they were not, it would hardly be plausible to suggest that anyone could have confused the one with the other, as Newman thinks Locke did. No one is likely to confuse an inferentia with an act of assent. The connection between the antecedent and the consequent in a conditional proposition is not an occurrence at all, still less a mental act. We must conclude, then, that though Newman may possibly have been misled by the Latin word inferentia in his account of inference, this cannot be the whole explanation of his paradoxical contention that inference is always conditional.
Could we get some further light on this question if we turn to Newman's own chapter on Inference (Grammar of Assent, Ch. VIII)? The first sentence of that chapter repeats what he had already said in Ch. VI10 ‘Inference is the conditional acceptance of a proposition’ and a few lines later he says ‘I proceed to show how inferential exercises, as such, always must be conditional’ (p. 197).11 How does he show it? So far as I can ascertain, he shows it in just two sentences: ‘We reason when we hold this by virtue of that; whether we hold it as evident or as approximating or tending to be evident, in either case we so hold it because of holding something else to be evident or tending to be evident’ (p. 197). These statements are unexceptionable. But why should they be supposed to show that inference is as such conditional? According to our modern usage of the term ‘conditional’ they appear to show the exact opposite. For the word ‘hold’ must surely mean something like ‘assert’. And the assertion of both premiss and conclusion (as opposed to the mere entertaining or assuming of them) is just what distinguishes an inference from a conditional statement. ‘Holding this by virtue of that’ is something more than just noticing that this would follow provided that were true. When we ‘hold this by virtue of that’ we express our inferential act or reasoning process by saying ‘Because that is so, this is so’ and not just by making the conditional statement ‘If that be so, this is so’.
It seems, then, that Newman's own chapter on Inference does not help us much, if we wish to understand why he maintained that inference is conditional. But as we read it, and consider what Newman himself was doing when he wrote it, we are reminded of a distinction which may throw some light on this question. It is the distinction between inferring on the one hand, and thinking about inference on the other. The two activities are quite different. Inferring is something that every human being does, and perhaps some of the higher non-human animals do it too. But it is only logicians and philosophers who think about inference, as Newman was doing when he wrote this chapter. Thinking about inference is what is called a second-order activity; not just thinking, but thinking about thinking, or rather about one kind of thinking. Now when we think about inference, there is one very important point about it which is most naturally expressed in a conditional statement: the conclusion can only be drawn if (provided that) we assert the premisses, or on condition that we do assert them. This does not mean, of course, that the conclusion can only be true if the premisses are true, nor yet that the conclusion can only be asserted if the premisses are asserted. But it can only be inferred from these premisses if these premisses are asserted. Newman makes the point himself when he says that in an inference ‘we hold this by virtue of that’. But though we do find it natural to make use of this conditional statement when we are thinking about inference, or trying to say what inference is, it does not follow that a person who is himself inferring does or should express his thought in a conditional statement. What he says, if he puts his inference into words, is not ‘If p then q’ but ‘Because p, therefore q’.
It is not, I think, incredible that Newman should have failed to distinguish between inferring and thinking about inference. The distinction between first-order statements and second-order statements is easily overlooked even now, and could be still more easily overlooked, even by a man of genius, at a time when the terms ‘first-order’ and ‘second-order’ had not yet been invented. But when we do draw the distinction, we see that the second-order conditional statement, which we naturally use in thinking about inference, is only another way of expressing the obvious fact that to acquire knowledge by means of inference we must have some uninferred knowledge already. We could, of course, say, if we pleased, that inference is ‘conditioned’ by the possession of previous information, using the word ‘conditioned’ in a pre-Pavlovian sense. But then we must distinguish between ‘conditional’ and ‘conditioned’. From the fact that inference is in this sense ‘conditioned’ it does not follow that inference is conditional, though just possibly Newman may have thought that it did.
But though Newman may perhaps have failed to distinguish between statements expressing inferences and second-order statements about inference, and may also have failed to draw the distinction between ‘conditional’ and ‘conditioned’, neither of these suggestions, nor both together, will explain everything that he says about the conditional character of inference. For when he says that inference is conditional, he must mean by ‘conditionally’ some character which admits of degrees. Otherwise, someone who mixed up inference with assent, as Newman thinks Locke did, would not have been led thereby to hold that assent has degrees, and Newman's whole criticism of Locke's doctrine of degrees of assent would fall to the ground. The whole point of that criticism is to try to convince us that the doctrine of degrees of assent loses all its plausibility as soon as the distinction between inference and assent is clearly drawn. But if we use the word ‘conditional’ in its ordinary modern sense (the sense in which we speak of conditional sentences or conditional statements) what could possibly be meant by saying that there are degrees of conditionality? Can one conditional sentence be more, or less, conditional than another? ‘If he comes I shall slip out by the back door, if you will give me the key now, and if I can find my umbrella.’ Is this a more conditional sentence than ‘If he comes, I shall slip out by the back door’? It is a more complex conditional sentence, certainly, because there are three ‘if’ clauses in it instead of one. But what could we mean by saying that it is more conditional? How strange it would be to say ‘John has made a very conditional statement, much more conditional than yours!’ It is still more strange to say that one inference can be more conditional than another, as Newman is bound to say that it can; for otherwise he could not maintain that Locke's doctrine of degrees of assent results from confusing assent with inference.
The ‘Conditional’ Character of Inference
It certainly cannot be denied that when Newman describes inference as conditional he speaks in a way which is bound to puzzle his modern readers; and the more they reflect on this contention of his, and on the importance he obviously attaches to it, the more puzzled they are likely to be. We must try, if we can, to go behind the unfortunate terminology he uses. Perhaps we may be able to do so, if we reflect upon the argument of his chapter as a whole. The aim of Grammar of Assent, Ch. VI, is to show that assent itself is unconditional. If it can be shown that assent is unconditional, Newman thinks it will also have been shown that there are no degrees of assent; and then Locke's theory of assent, and consequently his ‘Ethics of Belief, can be consigned to the oblivion they deserve. Perhaps we shall see the point of this puzzling contention that inference is conditional if we remember that ‘conditionally’, in Newman's view, is not only a character which inference has, but also a character which assent lacks. Whatever ‘conditional’ may mean, it must at any rate be equivalent to ‘not unconditional’. To put it in another way, we should fix our attention on the contrast between the conditional character which he attributes to inference, and the unconditional character which he attributes to assent.
Just where does the contrast lie? It appears to have something to do with doubt or with the presence of mental reservations. Newman's view is that assent, if given at all, has to be given without any doubt or any mental reservations. He is perfectly explicit about this. ‘Assent is an adhesion without reserve or doubt to the proposition to which it is given’ (p. 130). Again, on p. 139 he speaks of ‘the absolute absence of all doubt or misgiving in an act of assent’; and on the very first page of the chapter, using the word ‘absolute’ in another sense, he has said that ‘assent is in its nature absolute and unconditional’ (p. 119)—absolute, presumably, because it resembles a decree of an absolute monarch.
To put it vulgarly, his view is that assent always has a ‘whole hog’ character or a ‘neck or nothing’ character. You put all your money on the proposition assented to, you commit yourself to it unreservedly, though fortunately (I think we must say fortunately) your assent may fail later or be withdrawn. To use another vulgar metaphor, if you assent you have to ‘take the plunge’ or ‘go off the deep end’, because there is no shallow one; you must assent in this all-or-nothing way if you assent at all. This or something like this is what Newman seems to mean by ‘unconditional’ when he maintains that assent is unconditional.
If this is the correct interpretation of his words, we must all hope that assent is not in his sense ‘unconditional’, or at any rate not always. There may be many cases where such an unconditional assent is perfectly justified, but surely there are many more where it is not. If assent has to have this all-or-nothing character, a reasonable person will often have to choose the alternative ‘nothing’: that is, he will have to suspend judgement and withhold assent, unless and until conclusive evidence is forthcoming, which perhaps it never will be. But however dismaying his doctrine is, Newman himself certainly does hold that assent, if given at all, his to be given undoubtingly and without any mental reservations.
If this is what the ‘unconditional’ character of assent amounts to, it is natural to suppose that the ‘conditional’ character of inference amounts to just the opposite, and that when Newman describes inference as conditional, he means by ‘conditional’ something like ‘attended with doubt’ or ‘not free from mental reservations’.
That this, or something like this, is what he means emerges fairly clearly from a passage near the beginning of Ch. VI, where he is stating the view of Locke and Locke's followers which he proposes to criticize, and sums it up in these words: ‘Thus assent becomes a sort of necessary shadow, following upon inference, which is the substance; and is never without some alloy of doubt because inference in the concrete never reaches more than probability.… Assent cannot rise higher than its source: inference in such matters is at best conditional, therefore assent is conditional also’ (p. 120).
Here Newman seems to say that if assent did have some alloy of doubt it would ipso facto be conditional, and moreover would have acquired its alloy of doubt by being the shadow of inference. Inference, or at least inference in the concrete, is intrinsically dubious in some degree; and so assent, if it were indeed the shadow of inference, would be dubious derivatively. He is still more explicit in the section on Opinion in Ch. IV, where he contrasts opinion with inference: ‘We are even obstinate in them [our opinions]—whereas inference is in its nature and by its profession conditional and uncertain’ (pp. 45 fin—46). It may be noticed that in this statement the qualifying words ‘in the concrete’ are omitted.
We are justified, I think, in concluding that the character which inference has and assent lacks is something like doubtfulness or dubiety. According to Newman, inference is always in some way doubtful or hesitant, and assent never. This is what he appears to mean by calling inference conditional and assent unconditional. But why should anyone choose to use the word ‘conditional’ when what he means is ‘doubtful’ or ‘attended with doubt’? This may well puzzle the modern reader. But it would have appeared less puzzling two generations ago, when it was not uncommon for philosophers to maintain that the function of a conditional clause was to express doubt or questioning. Cook Wilson, for instance, thought that a conditional statement states a connection between two questions. Nowadays, however, we usually think that the function of the word ‘if’, or of the ‘if… then’ form of statement, is to convey an implication or an entailment.
It has been suggested already that ‘inference’ in Newman's usage (at least when he is contrasting inference with assent) means ‘recognizing reasons for thinking that p.’ Indeed, in one passage he himself speaks of ‘the inferential act which is the recognition of those reasons’ i.e. reasons for assent (p. 126). If so, the doctrine that inference is always conditional comes to this: when we recognize reasons for thinking that p, without as yet assenting to it, we have some degree of doubt about the proposition p. However strong our evidence may be, it still leaves room for doubt, at any rate where it is empirical evidence.
Perhaps Newman would even wish to define degrees of probability in terms of degrees of dubitability, and to say that ‘p is more probable than q on evidence E’ means ‘p is less doubtful than q on evidence E’. At any rate, what impresses him most about probable propositions is their dubitability, and not the guidance which they give us (imperfect but better than nothing) in our expectations and our conduct. I shall suggest later that this somewhat depressing way of looking at probability prevented him from seeing what Locke had in mind in his doctrine of degrees of assent. But Newman does of course try to console himself, and us, by maintaining that when we not only recognize reasons for thinking that p but do actually assent to it, all our doubts are flung to the winds.
‘Conditionality’, Confidence and Doubt
As we have seen, Newman thinks that if assent can be shown to be unconditional, Locke's doctrine of degrees of assent will collapse at once, and his Ethics of Belief with it. Why is this consequence supposed to follow? What is the connection between ‘being conditional’ on the one hand, and ‘having degrees’ on the other? The modern reader is somewhat puzzled by the way in which Newman switches back and forth between these two concepts without any explanation. The title of his chapter is ‘Assent considered as unconditional’. On the second page (p. 120) he states the doctrine which he is going to criticize, and it is the doctrine of degrees of assent. But when he actually makes his criticisms, what he criticizes is the doctrine that assent is always conditional, or rather, the doctrine that it is always conditional when the proposition assented to is neither self-evident nor demonstratively proved. There are occasional references to degrees of assent later, notably in the passage where he discusses and rejects the views of Gambier (pp. 131–2). But the main target of his criticisms throughout is the contention that assent ‘in concrete matters’ is conditional, which Locke had never said it was.
But perhaps we can now see why Newman should think that there is a very close connection between the two statements ‘assent admits of degrees’ and ‘assent in concrete matters is conditional’, or even that they are two ways of saying the same thing. It is because ‘conditional’ in his sense of the term means something like ‘attended with doubt’. Doubt does admit of degrees, even though conditionality in our sense of the term does not. If in assenting to a proposition concerning matters of fact we do often have doubts, great or small, about that proposition, and if we always ought to have some unless the reasons for assenting are conclusive and recognized by ourselves to be so, it will follow that assent admits of degrees.
It is true that Locke and his followers usually think of assent as having degrees of confidence, rather than degrees of doubtfulness. Locke's usual word is ‘assurance’. For example, in the passage about loving truth for truth's sake which Newman finds so disagreeable, Locke speaks of ‘not entertaining a proposition with greater assurance than the proofs it is built upon will warrant’.12 But elsewhere, in the passage which Newman commends for its inconsistency with this one, the word ‘assurance’ is used to mean the highest degree of assent.13 To avoid this ambiguity it is better to use the word ‘confidence’ in stating Locke's view. (He himself speaks once of ‘full assurance and confidence’.14)
Confidence and Doubt
What is the relation between Locke's degrees of confidence and Newman's degrees of doubt? I now wish to suggest that what Locke thinks of as confidence and Newman as doubt are not two different states of mind, but the same one looked at in different ways and measured, as it were, on different scales. When someone entertains a proposition with just a little confidence (‘suspecting’ in the terminology of Locke and his followers) we can equally well say that he has a good deal of doubt about it. When he entertains a proposition with a good deal of confidence but less than the maximum (opinion, but not conviction) we can equally well say that he has some doubt about the proposition, though not very much. When he is completely sure, we can equally well say that he entertains the proposition with no doubt at all.
Here, as in his treatment of probability, Newman prefers to look at the dark side of the picture, whereas Locke prefers to look at the bright side. Let us imagine a conversation between them. Locke says, ‘I have considerable confidence that it will be a fine day to-morrow’ and considers this a satisfactory state of affairs as far as it goes. In the same circumstances, and on the same evidence, Newman says, ‘I have some doubt whether it will be a fine day to-morrow’, and considers this an unsatisfactory state of affairs, because it does not go far enough. But Locke has to admit that he has some doubt about the proposition, and Newman has to admit that he has less doubt about it than he has about the proposition ‘It will rain continuously to-morrow all day, from dawn to dusk’. And as for satisfactoriness or unsatisfactoriness, they do not altogether disagree about this either. At any rate both agree that certainty would be better if they could get it, and both agree that they have got something which does at least take them beyond sheer suspense of judgement.
We may draw two conclusions from this. The first is that if Newman can show that assent is always ‘unconditional’ (undoubting), he really will have shown that Locke's doctrine of degrees of assent is false. For the degrees Locke is speaking of are degrees of confidence, and these can equally well be described as degrees of doubt. ‘Being more confident that…’ can equally well be described as ‘Being less doubtful whether…’, and ‘Being less confident that…’ as ‘Being more doubtful whether’. If Newman can show that whenever we assent, we have to assent with no doubt at all, he will have shown that assent does not admit of degrees in the sense in which Locke thought it did.
The second conclusion which we may draw is more surprising. It is that Newman and Locke are in much closer agreement than Newman thinks. They both agree that there is a common and familiar propositional attitude which does admit of degrees. Locke thinks of it as an attitude of greater or lesser confidence about a proposition and Newman as an attitude of lesser or greater doubt. The difference between them is at least partly one of emphasis. Locke emphasizes the positive—we might almost say the ‘life-enhancing’—aspect of this attitude. When he describes it as an attitude of confidence, greater or less, he is pointing out that when we have this attitude to a proposition, we rely upon the proposition to a greater or lesser degree in our subsequent thoughts and actions, and this is why he holds that assent admits of degrees. If I have some confidence that it will be a fine day to-morrow, I can consider various alternative ways of spending to-morrow afternoon, consult maps, get the tank of my car filled up with petrol, spend a little time looking for the field-glasses which I have mislaid somewhere, or a little money buying a new walking-stick if I have lost my old one. These activities and these plans do of course have something tentative and provisional about them. I engage in them ‘on the assumption that’ to-morrow will be fine, not in the certain knowledge that it will. We might even apply Newman's favourite term ‘conditional’ to these thinkings and doings which I embark upon. I am well aware, all the time, that to-morrow may not be a fine day after all. I never thought that the weather forecast was infallible, or that the weather-signs I can observe for myself were conslusive. But still, the proposition ‘It will be fine to-morrow’, though not accepted ‘unconditionally’, does give me some guidance as to what I should do and what I should think about to-day. I do not just have to wait and see, and fold my hands meantime in a state of inert agnosticism.
But if we prefer to emphasize the ‘doubting’ aspect of this propositional attitude, as Newman does, we are liable to overlook the important facts which I have just mentioned. We may easily fail to notice that a proposition about which we have some doubt does nevertheless give us some guidance both in thought and in action; that we do nevertheless rely on it in some degree, even though we also have some doubt of it; and moreover, we may fail to notice that this reliance is reasonable, provided it is no greater than the evidence available to us justifies. If we think of this attitude as just a doubting one, we are emphasizing the negative side of it, its shortcomings, so to speak, rather than its advantages, its depressing or devitalizing features rather than its ‘life-enhancing’ features. It does have these depressing features. It is true that half a loaf is worse than a whole one. If a proposition is of importance to us, whether practically or theoretically, to have some doubt about this proposition is a less desirable state than having no doubt would be—provided that it were reasonable to have no doubt. But it is also true that half a loaf, or even a quarter, is better than nothing.
Nevertheless, in Newman's exposition of his own theory of assent there are two points at which he might quite well have admitted the existence and importance of this ‘degree-having’ attitude to a proposition. He himself holds that what he calls ‘inference’—the recognition of evidence, strong or weak, for a proposition p-is accompanied by an inclination to assent to p, and moreover that this inclination is ‘greater or less according as the particular act of inference expresses a stronger or weaker probability’ (Grammar of Assent, p. 129). He does, of course, insist that we can have such an inclination without actually assenting in his ‘unconditional’ sense of the word. Nevertheless, he could quite well have admitted that this inclination often results in something else: namely, the taking up of a new attitude to the proposition which we are inclined to assent to, an attitude of relying upon the proposition to some degree, and a disposition to use it, though not without reservations, as a guide to our subsequent thoughts and actions. In the language he himself uses, this attitude could not have been called assent, because it is not in his sense ‘unconditional’. But he could quite well have admitted its existence and its importance. In that case, his disagreement with Locke would have been largely a difference of terminology.
Moreover, as we have seen, Newman describes the recognition of reasons for assenting as ‘an inferential act’ (p. 126). He also says at the beginning of his chapter on Inference that ‘Inference is the conditional acceptance of a proposition’ (p. 197, my italics). Unfortunately, he is more interested in the conditionality than the acceptance. But he might quite well have added that when such acceptance occurs (as it does when we recognize that a proposition p is supported in some degree by the evidence we have), it has effects upon our subsequent thoughts and actions which are almost as important as the effects of unconditional acceptance would have been; indeed more important, since we do not so very often have evidence which justifies us in accepting a proposition in this ‘unconditional’ way, whereas we do quite often have evidence which justifies us in accepting a proposition ‘conditionally’ (with some degree of doubt or mental reservation, but also with some degree of confidence). And then again his disagreement with Locke would have been, at least partly, a difference of terminology.
The fact remains, however, that Newman did not actually take either of these opportunities of bringing his own views a little nearer to Locke's. Newman's doctrine, as he himself sees it, is totally opposed to Locke's. Locke maintains that there are degrees of assent, and Newman maintains that there are none.
It was suggested earlier that if Newman were right and Locke wrong on the main point at issue between them (does assent admit of degrees?) our human condition would be at once more miserable and more intellectually-disreputable than we commonly suppose.16 It would be more miserable, because we so often need to be able to assent to propositions on evidence which is far less than conclusive; and therefore we need to be able to assent to them with something far less than total or unreserved self-commitment, if we are to have any guidance for our subsequent thoughts and actions. But Newman tells us that we cannot do it. According to him, the very nature of assent makes it impossible for anyone to assent in this conditional and tentative way. If Newman were right, our situation would also be more intellectually-disreputable than we commonly suppose. In such circumstances, where we have evidence which is less than conclusive, only two alternatives would be open to us: either complete suspense of judgement, or else an assent of the all-or-nothing (‘unconditional’) sort, which would be unreasonable, because nothing short of conclusive evidence could justify it. We could not be content with the first alternative, which gives us no guidance at all for our subsequent thoughts and actions. We should just have to assent unreasonably, with the clear knowledge, at least sometimes, that our assent is unreasonable: for everyone can see, at least sometimes, that the evidence he has for a proposition is less than conclusive, even though it is perfectly good evidence so far as it goes.
But there is a third alternative. We do not always have to choose between an inert agnositicsm—a helpless ‘wait and see’ attitude—and a total and unreserved self-commitment. When our evidence for a proposition, though not conclusive, is favourable, or favourable on balance when any unfavourable evidence there may be is taken into account, we can assent to that proposition with a limited degree of confidence; and we can then conduct our intellectual and practical activities ‘in the light of the proposition, though not without some degree of doubt or mental reservation. This, surely, is what we find oursleves and our neighbours continually doing. This is what we find when we follow Newman's own plan of ‘interrogating human nature, as an existing thing, as it is found in the world… and being content with the mind as God has made it’. (Grammar of Assent, p. 124.) And on this point, it is surely Newman's view, not Locke's, which must be called ‘theoretical and unreal’.
This is what Locke himself means by ‘the proofs it is built upon’ in the passage quoted from Ch. 19, Section 1. Nowadays we should reserve the word ‘proofs’ for cases where the evidence is conclusive.
The word ‘appear’ here has its ‘present to the mind’ sense, not its ‘seeming’ sense. Cf. ‘the moon has just appeared above the horizon’, as contrasted with ‘it appears larger at the horizon than it does at the zenith’.
The Grammar of Assent was first published in 1870. I shall quote from Longmans Edition of 1947. (There is an earlier edition, also published by Longmans, dated 1930. In this the page numbers are different and there is no index.)
Grammar of Assent (Longmans, 1947), p. 122.
Cf. also p. 120, ‘Abstract argument is always dangerous,… I prefer to go by facts.’ Newman is after all an Empiricist philosopher, and in this chapter he claims to be more of an Empiricist than Locke is.
J. E. Gambier (Rector of Langley, Kent) An Introduction to the Study of Moral Evidence (Ed, 3, 1824) p. 6.
‘A certain degree’ I think means gradus quidam, not gradus certus. But in the next sentence ‘certain’ does mean certus.
‘If it be only the echo of an inference, do not treat it as a substantive act’ (p. 125).
Cf. the phrase he uses in his criticism of Gambier, quoted on p. 137–138 above. Gambier's ‘suspicion, conjecture, persuasion’ etc., Newman says, ‘are not variations of assent to an inference, but assents to a variation in inferences’ (Grammar of Assent, p. 132).
On p. 119 (cf. p. 130).
‘An inferential exercise’, I think, is any actualization of our capacity for inferring, and is not limited to cases where we are training ourselves to infer correctly or being trained by others to do so.
Essay, Book IV, Ch. 19, section 1.
Essay, Book IV, Ch. 16, section 6. ‘Our belief thus grounded rises to assurance.’ Newman remarks that ‘assurance’ is here equivalent to ‘certitude’ (Grammar of Assent, p. 122).
Essay, Book IV, Ch. 15, section a.
Essay, Book IV, Ch. 19, section 1.
P. 133, above.