I F S AND CANS'

A R E cans constitutionally iffy? Whenever, that is, we say that we can do something, or could do somethmg, or could have done something, is there an if in the offing-suppressed, it may be, but due nevertheless to appear when we set out our sentence in full or when we give an explanation of its meaning? Again, if and when there is an ifclause appended to a main clause which contains a can or could or could have, what sort or an ifis it? What is the meaning of the if, or what is the effect or the point of combining this if-clause with the main clause? These are large questions, to which philosophers, among them some whom I most respect, have given small answers: and it is two such answers, given recently by English philosophers, that I propose to consider. Both, I believe, are mistaken, yet somedung is to be learned from examining them. In sophy, there are many mistakes that it is no hsgrace to have made: to make a first-water, ground-floor mistake, so far from being easy, takes one (one) form of phtlosophical genius? Many of you will have read a short but justly admired book written by Professor G. E. Moore of Cambridge, which is called simply Ethics. In it, there is a point where Moore, who is engaged in discussing Right and Wrong, says that if we are to discuss whether any act that has been done was right or wrong then we are bound to discuss what the person concerned could have done instead of what he did in fact do. And this, he tMrs,
Reprinted from the Proceedirrgs of the British Academy, 1956, by courtesy o f the editor. Plato, Descartes, and Leibniz all had this form of genius, besides o f course others.

206

IFS AND CANS

may lead to an entanglement in the problem, so-called, of Free Will: because, though few would deny, at least expressly, that a man could have done s o m e h g other than what he did actually do he had chosen, many people would deny that he could (absolutely) have done any such other dung. Hence Moore is led to ask whether it is ever true, and if so in what sense, that a man could have done s o m e h g other than what he did actually do. And it is with his answer to dus question, not with its bearings upon the meanings of right and wrong or upon the problem of Free W d , that we are concerned. With his usual shrewdness Moore begins by insisting that there is at least one proper sense in which we can say that a man can do s o m e h g he does not do or could have done something he did not d w v e n though there may perhaps be other senses of can and could have in which we cannot say such things. This sense he illustrates by the sentence 'I could have walked a mile in 20 minutes this morning, but I certainly could not have rut1 two miles in 5 minutes': we are to take it that in fact the speaker did not do either of the two things mentioned, but h J in no way h d e r s us from drawing the very common and necessary distinction between undone acts that we could have done and undone acts that we could not have done. So it is certain that, at least in some sense, we often could have done t h g s that we c d not actually do. G Why then, Moore goes on to ask, should anyone try to deny this? And he replies that people do so (we may call them 'determinists') because they hold that everything that happens has a cause which precedes it, which is to say that once the cause has occurred the dung itself is bound to occur and nothing else could ever have happened instead. However, on examining further the 20-minute-mile example, Moore argues that there is much reason to think that 'could have' in such cases simply means 'could have if1 had chosen', or, as perhaps we had better say in order to avoid a possible complication (these are Moore's words), simply means 'should have if I had chosen'. And if this is all it means, then

if

IFS A N D C A N S

207

there is after all no conflict between our conviction that we often could have, in this sense, done things that we did not actually do and the determinist's theory: fo; he certainly holds himself that I often, and perhaps even always, should have done something different from what I cLd do i f I had chosen to do that different dung, since my choosing differentlywould constitute a change in the causal antecedenti of my subsequent act, which would therefore, on his theory, naturally itself be dfferent. If, therefore, the determinist nevertheless asserts that in some sense of 'could have' I could not ever havedone anydung different do, from what I did actu~lly this must simply be a second sense' of 'could have' different from that which it has in the 20minute-mile example. In the remainder of his chapter, Moore argues that quite possibly his first sense of 'could have', in which it simply means 'could or should have if I had chosen', is all we need to satisfy our hankerings after Free Will, or at least is so if conjoined in some way wlth yet a thlrd sense of 'could have' in which sense 'I could have done something different' means 'I might, for all anyone could know for certain beforehand, have done something different'. This third lund of 'could have' might, I thnk, be held to be a vulgarism, 'could' being used inForrectly for 'might': but in any case we shall not be concerned with it here. 1 the upshot, &en, Moore leaves us with only one important ; sense in Ghich it can be said that I could have done somedung that I cLd not do: he is not convinced that any other sense is necessary, nor has he any clear idea what such another sense would be: and he is convinced that, on his interpretation or 'could have', even t l ~ e determinist can, and indeed must, say that I could very often have done things I &d not do. To summarize his suggestions (he does not put them forward with complete conviction) once again: r. 'Could have' simply means 'could have if I had chosen'. 2. For 'could have if I had chosen' we may substitute 'should have if I had chosen'.
1

About which Moore has no more to tell us.

208

IFS AND CANS

3 . The if-clauses in these expressions state the causal con&tions upon which it would have followed that I could or should have done the thing dfferent from what I &d actually do.

Moore does not state this thud point expressly himself: but it seems clear, in view of the co&exions he alleges between his interpretation of 'could have' and the determinist theory, that he did believe it, presumably taking it as obvious. There are then three questions to be asked: Does 'could have if I had chosen' mean the same, in general or ever, as 'should have if I had chosen?' 2. In either of these expressions, is the if the if of causal condtion ? 3 . In sentences having can or could have as main verb, are we required or entitle2 always to supply an if-clause, and in particular the clause 'if I had chosen' ?
1.

It appears to me that the answer in each case is No. I . Anyone, surely, would admit that in general could is very different indeed from should or would.1 What a man could do is not at all the same as what he ~ v o u l d perhaps he could shoot do: you if you were within range, but that is not in the least to say that he would. And it seems clear to me, in our present example, that 'I could have run a mile if I had chosen' and 'I should have run a mile if I had chosen' mean quite different things, though unfortunately it is not so clear exactly what either of them, especially the latter, does mean. 'I should have run a mile in 20 minutes this morning if I had chosen' seems to me an unusual, not to say queer, specimen of English: but if I had to interpret it, I should take it to mean the same as 'If I had chosen to run a mile in 20 minutes this morning, I should (jolly well) have done so', that is, it would be an assertion of my strength
Since Moore has couched his example in the first person, he uses 'should' in the apodosis: but of course in the third person, everyone would use 'would'. For brevity, I shall in what follows generally use 'should' to do duty for both persons.

I F S AND C A N S

209

of character, in that I put my decisions into execution (an assertion which is, however, more naturally made, as 1 have now made it, with the if-clause preceding- the main clause). I should certainly not myself understand it to mean that if I had made a certain choice my making that choice would have caused me to do s o m e h g . But in whichever of these ways we understand it, it is quite different from 'I could have walked a mile in 20 minutes this morning if I had chosen', which surely says something rather about my opportunities or powers. Moore, unfortunately, does not explain why he dunks we are entitled to make this all-importan; transition from 'could' to 'should', beyond saying that by doing so we 'avoid a possible complication'. Later I shall make some suggestions which may in part explain why he was tempted to make the transition: but nothing can justify it. 2. Moore, as I pointed out above, did not discuss what sort of $'it is that we have in 'I can if I choose' or in 'I could have if 1-hadchosen' or in 'I should have if I had chosen'. Generally, philosophers, as also grammarians, have a favourite, if somewhat blurred and diffuse, idea of an +clause as a 'conditional' clause: putting our example schenlatically as 'If y, then q', then it will be said that q follows from p, typically either in the sense that p entails q or in the sense that p is a cause of q, though other important variations are possible. And it seems to be on these lines that Moore is thnking of the f i n 'I can if I choose'. But i now, it is characteristic of this general sort of J that from 'If p then q' we can draw the inference 'If not q, then not p', whereas we can not infer either 'Whether or not y, then q' or 'q' simpliciter. For example, from 'If I run, I pant' we cat1 infer 'If I do not pant, I do not run' (or, as we should rather say, 'If I am not panting, I am not running'), whereas we can not infer either 'I pant, whether I run or not' or 'I pant' (at least in the sense of '1 am panting'). If, to avoid these troubles with the English tenses, which- are unfortunately prcvalent but are not allowed to matter, we put the example in the past tense, then from 'If I ran, I panted' it does follow that 'If I did not pant,
-

210

IFS AND C A N S

I did not run', but it does not follow either that 'I panted whether or not I ran' or that 'I panted' period. These possibilities and impossibhties of inference are typical of the ijof causal condi- tion: but they are precisely reversed in the case of 'I can if 1 choose' or 'I could have if I had chosen'. For from these we .should not draw the curious inferences that 'If I cannot, I do not choose to' or that 'If I could not have, I had not chosen to' (or 'did not choose to'), whatever these sentences may be supposed to mean. But on the contrary, from 'I can if I choose' we certainly should infer that 'I can, whether I choose to or not' and indeed that 'I can' period: and from 'I could have if I had chosen' we should similarly infer that 'I could have, whether I chose to or not' and that-anyway 'I could have' period. So that, whatever this ifmeans, it is evidently not the ifof causal

This becomes even clearer when we observe that it is quite common elsewhere to find an or&nary causal conditional in connexion with a can, and that then there is no doubt about it, as for example in the sentence 'I can squeeze through if I am thm enough', which does imply that 'If I cannot squeeze through

if

I am not d m enough', and of course does not imply that 'I can squeeze through'. 'I can if I choose' is precisely different from this. Nor does can have to be a very special and ~eculiar verb for which are not causal conditional to be found in connexion with it: all kinds of ifs are found with all kinds of verbs. Consider for example th&in 'There are biscuits on the sideboard if you want them', where the verb is the highly ordinary are, but the ifis more like that in 'I can if I choose' than that in 'I panted if I ran': for we can certainly infer from it that here are biscuits on the sideboard whcther you want them or not' and that anyway 'There are biscuits on the sideboard', whereas it would be folly to infer that 'If there are no biscuits on the sideboard you do not want them', or to understand the meaning to be that you have only to want biscuits to cause them to be on the sideboard.

ifs

IFS A N D C A N S

21I

The ij; then, in 'I can if I choose' is not the causal conditional What of the ifin 'I shall if I choose'? At first glance, we see that h s is quite hfferent (one more reason for refusing to substitute shall for can or should have for could have). For from 'I shall if I choose' w e clearly cannot infer that 'I shall whether I choose to or not' or simply that 'I shall'. But on the other hand, can we infer, either, that 'If I shan't I don't choose to'? (Or should it be rather 'If I don't I don't choose to'?) I t h k not, as we shall see: but even if some such inference can be drawn, it would s t d be patently wrong to conclude that the meaning of 'I shall if I choose' is that my choosing to do the t h g is sufficient to cause me inevitably to do it or has as a consequence that I shall do it, whlch, unless I am mistaken, is what Moore was supposing it to mean. This may be seen if we compare 'I shall ruin him if I choose' with 'I shall ruin him if I am extravagant'. The latter sentence does indeed obviously state what would be the consequence of the fulfilment of a condition specified in the $-clause-but then, the first sentence has clearly cbfferent characteristics from the second. In the first, it makes good sense in general to stress the 'shall', but in the second it does not.' This is a symptom of the fact that in the first sentence 'I shall' is the present of that mysterious old verb shall, whereas in the second 'shall' is simply being used as an aux&ary, without any meaning of its own, to form the future indicative of 'ruin'. I expect you will be more than ready at h s point to hear s o m e h g a little more positive about the meanings of these curious expressions 'I can if I choose' and 'I shall if I choose'. Let us take the former first, and concentrate upon the ij: The dictionary tells us that the words from which our ifis descended expressed, or even meant, 'doubt' or 'hesitation' or 'condition' or 'stipulationy. Of these, 'condition' has been given a procbgious innings by grammarians, lexicographers, and philosophers allke: it is time for 'doubt' and 'hesitation' to

if.

In general, though of course in some contexts it does: e.g. 'I may very easily ruin him, and 1 shall if I am extravagant', where 'shall' is stressed to point the contrast with 'may'.
1

212

IFS A N D C A N S

be remembered, and these do indeed seem to be the notions present in 'I can if I choose'. W e could give, on different occasions and in hfferent contexts, many mfferent interpretations of this sentence, which is of a somewhat primitive and loose-jointed type. Here are some:
I can, quaere do I choose to? I can, but do I choose to? I can, but perhaps I don't choose to I can, but then I should have to choose to, and what about that? I can, but would it really be reasonable to choose to? I can, but whether I choose to is another question I can, I have only to choose to I can, in case I (should) choose to,

and so on. These interpretations are not, of course, all the same: which it is that we mean will usually be clear from the context (otherwise we should prefer another expression), but sometimes it can be brought out by stress, on the 'if'or the 'choose' for example. What is common to them all is simply that the asserlion, positive and complete, that 'I can', is linked to the ~aising ofthe question whether I choose to, which may be relevant in a variety of ways. I Ifs of the lund I have been trying to describe are common enough, for example the ifin our example 'There are biscuits on the sideboard if you want them'. I do not know whether you want biscuits or nit, but in case you do, I point out that there are some on the sideboard. It is tempting, I know, to 'expand' our sentence here to this: 'There are biscuits on the sideboard which you can (or may) take if you want them': but this, legitimate or not, will not make much difference, for we are still left with 'can (or may) if you want', which is (here) just llke 'can
If there were space, we should consider other germane expressions: e.g. 'I can do it or not as I choose', 'I can do whichever I choose' (quidlibet). In particular, 'I can whether I choose to or not' meails 'I can, but whether I choose to or not is an open question': it does not mean 'I can on condition that 1 choose
and likewise on condition that 1 don't', which is absurd.

IFS AND C A N S

213

if you choose' or 'can if you lke', so that the ifis still the ifof doubt or hesitation, not the if of condition.1 I will mention two further points, very briefly, about 'I can if I choose', important but not so relevant to our discussion here. Sometimes the can will be the can, and the choice the choice, of legal or other right, at other times these words will refer to practicability or feasibhty: consequently, we should sometimes interpret our sentence in some such way as 'I am entitled to do it (if I choose)', and at other times in some such way as 'I am capable of doing it (if I choose)'. We, of course, are concerned with interpretations of t h s second kind. It would be nice if we always said 'I may if I choose' when we wished to refer to our rights, as perhaps our nannies once told us to: but the interlocking histories of can and may are far too chequered .~ for there to be any such rule in p r a ~ t i c eThe second point is that choose is an important word in its own right, and needs careful interpretation: 'I call if I lke' is not the same, although the 'can' and the 'if' may be the same in both, as 'I can if I choose'. Choice is always between alternatives, that is between several courses to be weighed in the same scale against each other, the one to be prrfrred. 'You can vote whichever way you choose' is different from 'You can vote whchever way you k e '. And now for somethmg about 'I shall if I choose'-what sort of ifhave we here? The point to notice is, that 'I shall' is not an assertion ofjzct but an expression of intention, verging towards the giving of some variety of undertaking: and the $ conseAn account on these lines should probably be given also of an excellent example given to me by Mr. P. T. Geach: 'I paid you back yesterday, if you remember.' This is much the same as 'I paid you back yesterday, don't you remember?' It does not mean that your now remembering that I did so is a condition, causal or other, of my having paid you back yesterday. Formerly I believed that the meaning of 'I can if I choose' was something hke 'I can, 1 have the choice', a i d that the point of the +clause was to make clear that the 'can' in the main clause was the 'can' of right. This account, however, does not do justice to the role of the 'if', and also unduly restricts in general the meaning of 'choice'.

214

IFS A N D C A N S

quently, is the not of condition but of stipulation. In sentences Ue: I shall I marry him if I choose I intend I to marry him if I choose I promise I to marry him if he will have me the if-clause is a part of h e object phrase governed by the initial verb ('shall', 'intend', 'promise'), if this is an allowable way of putting it: or again, the ifquaMes the content of the undertaking given, or of the intention announced, it does not qualify the giving of the undertaking. Why, we may ask, is it perverse to draw from 'I intend to marry him if I choose' the inference 'If I do not intend to marry him I do not choose to'? Because 'I intend to marry hun if I choose' is not U e 'I panted if I ran' in h s important respect: 'I panted if I ran' does not assert anythmg 'categorically' about me-it does not assert that I did pant, and hence it is far from surprising to infer something. beginning 'If I did not pant': but 'I intend to marry him if I choose' (and the same goes for 'I shall marry him if I choose') is a 'categorical' expression of intention, an2 hence it is paradoxical to make an inference leading offwith 'If I do not intend'. 3 . Our third question was as to when we are entitled or required to supply iJ-clauses with can or could have as main verb. Here there is one thing to be clear about at the start. There are two quite hstinct and incompatible views that may be put forward concerning i and cans, which are fatally easy to conf s fuse with each other. One view is that wherever we have can or could have as our main verb, an if-clause must always be understood or supplied, if it is not actually present, in order to complete the sense of the sentence. The other view is that the meaning of 'can' or 'could have' can be more clearly reproduced by some other verb (notably 'shall' or 'should have') with an if-clause appended to-it. he first view is that an if is required to complete a can-sentence: the second view is that an is required in the analysis of a can-sentence. The suggestion of Moore that 'could have' means 'could have if I had chosen' is

if

iJ

IFS A N D C A N S

21 S

a suggestion of the first kind: but the suggcstion also made by Moore that it means 'should have if I had chosen' is a suggestion -of the second kind. It may be because it is so easy (apparently) to confuse these two kinds oftheorv that Moore was temvted to talk L as though 'should have' could mean the same as 'could have'. Now we are concerned at this moment solely with thefirst sort of view, namely that can-sentences are not complete without an if-clause. And if we tlunk, as Moore was for the most it part thi.dmg, about 'could have' (rather than 'canY), is easy to see why it may be tempting to allege that it always requires an if-clause with it. For it is natural to construe 'could have' as a past subjunctive or 'conditional', whch is practically as much as to say that it needs a conditional clause with it. And of course it is quite true that 'could have' m g be, and very often is, a past conditional: but it is also true that 'could have' may be and often is the past (definite) indicative of the verb can. Sonletimes 'I could have' is equivalent to the Latin 'Potui' and means 6 I was in a position to': sometimes it is equivalent to the Latin 1 Potuissem' and means 'I should have been in a position to'. Exactly similar is the double role of 'could', which is sometimes a conditional meaning 'should be able to', but also sometimes a past indicative (indefinite) meaning 'was able to': no one can doubt tlus if he considers such contrasted examples as 'I could do it 20 years ago' and 'I could do it if I had a h g u m m y . ' It is not so much that 'could' or 'could have' is ambiguous, as rather that two parts of the verb can take the same shape. Once it is realized that 'could have' can be a past indicative, the general temptation to supply ijklauses with it vanishes: at least there is no more temptation to supply them with 'could have' than with 'can'. If we ask how a Roman would have said 'I could have ruined you t h s morning (although I didn't)', it is clear that he would have used 'potui', and that h s sentence is complete without any conditional clause. But more than tlus, if he had wished to add 'if I had chosen', and however he had expressed that in Latin, he would still not have changed hls 'potui' to 'potuissem': but h s is precisely what he would have
J

216

IFS AND C A N S

done if he had bceil tacking on some other, inore 'normal' kind of if-clause, such as 'if I had had one more vote'.' That is to say, the 'could have' in 'could have if I had chosen' is a past indicative, not a past conditional, despite the fact that there is what would, I suppose, be called a 'conditional' clause, that is an $<lause, with G - ~ n thls is, of course, why we can d make the inferences that, as we saw, we can make from 'I could have if I had chosen', notably the inference to 'I could have' absolutely. Hence we see how mistaken Moore was in contrasting 'I could have if I had chosen' with the 'absolute' sense of 'I could have': we might almost go so far as to say that the addition of the 'conditional' clause 'if I had chosen' makes it certain that (in Moore's language) the sense of 'could have' is the absolute sense, or as I shouldprefer to put it, that the mood of 'could have' is indicative. It might at t h s point be worth considering in general whether it makes sense to suppose that a language could contain any verb such as can has been argued or implied to be, namely one that can never occur without an +clause appended to it. At least if the ifis the normal 'conditional' ifths would seem very difficult. For let the verb in question be to X: then we shall never say simply 'I X', but always 'I X if I Y' : but then also, according to the accepted rules, if it is true that 'I X if I Y', and nlso true (whch it must surely sometimes be) that 'I do, in fact, Y', it must surely follow that 'I X', simpliciter, without any ifabout it any longer. Perhaps t h s was the 'possible complication' that led Moore to switch from the suggestion that 'I could have' (in one sense) has always to be expanded to 'I could have if' to the suggestion that it has always to be nnalysed as 'I should have i f :
ffthe $-clause is 'if I had chosen', then I was able, was actually in a position, to ruin you: hence 'potui'. But if the +clause expresses a genuule unfufjlled condition, then plainly I was not actually in a position to ruin you, hence not 'potui' but 'potuissern'. My colleague Mr. R. M. Nisbet has pointed out to me the interesting discussion o f this point in S. A. Handford, The Latin Subjunctive, pp. 1 3 0 ff. It is interesting that although this author well appreciates the Latin usage, he still takes it for granted that in English the 'could have' is universally subjunctive or conditional.

IFS A N D CANS

217

for of course the argument I have just given does not suffice to show that there could not be some verb which has always to be annlysrd as somethng containing a conditional if-clause: suggestions that this is in fact the case with some verbs are common in philosophy, and I do not propose to argue this point, though I think that doubt might well be felt about it. The only sort of 'verb' 1 can think of that might always demand a conditional clause with it is an 'auxiliary' verb, if there is one, which is used solely to form subjunctive or conditional moods (whatever exactly they may be) of other verbs: but however this may be, it is quite clear that can, and I should be prepared also to add shall and will and may, are not in this poyition.To summarize, then, what has been here said in reply to Moore's suggestions in his book: (a) 'I could have if I had chosen' does not mean the same as should have if I had chosen'. (b) In neither of these expressions is the iJklause a 'normal conditional' clause, connecting antecedent to consequent as cause to effect. (c) T o argue that can always requires an if-clausewith it to complete the sense is totally different from arguing that can-sentences are always to be analysed into sentences containing if-clauses. (d) Neither mn nor any other verb always requires a conditional +clause after it: even 'could have', when a past indicative, does not require such a clause: and in 'I could have if I had chosen' the verb is in fact a past indicative, not a past subjunctive or conditional. Even, however, if all these contentions are true so far, we must recognize that it may nevertheless still be the case that can, could, and could have, even when used as indicatives, are to be analysed as meaning shall, should, and should have, used as auxiliaries of tense or mood with another verb (i.e. so as to make that other verb into a future or subjunctive), followed by

"

21 8

IFS AND C A N S

a conditional if-clause. There is some plausibility,' for example, in the suggestion that 'I can do X' means 'I shall succeed in doing X, if I try' and 'I could have done X' means 'I should have succeeded in doing X, if I had tried'. It is indeed odd that Moore should have plumped so simply, in giving h s account whether of the necessary supplementation or of the analysis of 'could have', for the one particular if-clause 'if I had chosen', which happens to be particularly exposed to the above objections, without even mentioning the possibility of invoking other if-clauses,at least in some cases. Perhaps thc reason was that choose (a word itself much in need of discussion) presented itself as well fitted to bridge the gulf between determinists and free-willers, whlch try might not so readily do. But as a matter of fact Moore does hmself at one point give an analysis of 'I could have done X' whlch is different in an interesting way from his usual version, although confusible with it. At a crucial point in his argument, he chioses for his example 'The ship could have gone faster', and the suggestion is made that this is equivalent to 'The s h p would have gone faster ifher oficers had chosen'. This may well seem plausible, but so far from being in line, as Moore apparently t h d s , with his general analysis, it differs from it in two important respects:
Plausibility, but no more. Consider the case where I miss a very short putt and kick myself because I could have holed it. It is not that I should have holed it if I had tried: I did try, and missed. It is not that I should have holed it if conditions had been different:. that might of course be so, but I am talking about conditions as they precisely were, and asserting that I could have holed it. There is the rub. Nor does 'I can hole it this time' mean that 1shall hole it this time if I try or if anything else: for I may try and miss, and yet not be convinced that I could not have done it; indeed, further experiments may confirm my belief that 1 could have done it that time although I did not. But if I tried my hardest, say, and missed, surely there must have been something that caused mc to fail, that made me unable to succeed? So that I could not have holed it. Well, a modern belief in science, in there being an exp!anation of everything, may make us assent to this argument. But such a belief is not in line with the traditional beliefs enshrined in the word can: according t o them, a human ability or power or capacity is inherently liable not to produce success, on occasion, and that for no reason (or are bad luck and bad form sometimes reasons ?).

'

IFS AND CANS

219

(a) the subject of the $-clause ('her officers') is different from the subject of the main clause ('the ship'), the subject of the original sentence: (b) the verb in the ifilause following 'chosen' is clfferent from the verb in the main clause, the verb in the original sentence. W e do not readily observe this because of the ellipsis after 'chosen': but plainly the verb must be, not 'to go faster', but 'to make her go faster' or, for example, 'to open the throttle'. These two features are dictated by the fact that a shtp is inanimate. W e do not wish seriouslv to ascribe free will to inanimate objects, and the 'could' of the original sentence is perhaps only justifiable (as opposed to 'might') because it is readily realized that some L verson's free will is in auestion. I If we follow up the Lnes of this new type of analysis, we should have to examine the relations between 'I could have won' and 'I could, or should, have won if I had chosen to lob' and 'I could, or should, have won if he had chosen to lob'. I will do no more here than point out that the difference between 'could' and 'should' remains as before, and that the sense of 'I could have wonY, it really is one, in which it means something if of the sort 'I should have won if he had chosen to lob' or 'to let me win' (the parallel to the s h p example), is of little importance-the 'if' here is of course the conditional It is time now to turn to a second discussion of $ and cans. Quite recently my colleague Mr. Nowell-Smith, in another little book called Ethics, also reaches a point in h s argument at whch he has to examine the sentence 'He could have acted otherwise', that is, could have done something that he did not in fact do. His reason for doing so is that, udess we can truly say t h s of people, we might find ourselves unable to blame people for things, and this would be generally regretted. This reason is not unrelated to Moore's reason for embarking on h s earlier discussion, and Nowell-Smith's views show some resemblances to Moore's: perhaps this is because Nowell-Smith,
1

if:

220

fFS A N D C A N S

hke Moore at the time he wrote his book is willing, if not

anxious, to come to terms with determinism. Nowell-Smith begins his discussion by saying (p. 274) that ' "could have" is a modal phrase, and modal phrases are not normally used to make categorical statements'. I am not myself at all sure what exactly a 'modal phrase' is, so I cannot discuss thls assertion: but I do not thnk &.IS matters, because he proceeds to give us two other examples of modal phrases, viz. 'might have' and 'would have',' and to tell us first what they are not (which I omit) and then what they are: 'Would have' and 'might have' are clearly suppressed hypothetical~, incomplete without an 'if. . .' or an 'if. ..not . . .'. Nobody would say 'Jones would have won the championship' unless (a) he believed that Jones did not win and (b) he was prepared to add 'if he had entered' or 'ifhe had not sprained his ankle' or some such clause. Here (t~) is actually incorrect-we can say 'Jones would (still) have won the championship, (even) if Hagen had enteredy-but thls does not concern us. (b),however, seems to be fairly correct, at least as far as concerns 'would have' (in the case of 'might have' it might well be doubtedz). So we have it that, when Nowell-Smith says that 'would have' is a 'suppressed hyp+
Also perhaps 'may have', for he discusses 'It might have rained last Thursday' in terms that seem really to apply to 'It may have rained last Thursday'. 1 refrain here from questioning it in the case of 'would have'. Yet 'would' is agreed to be often a past indicative of the old verb will, requiring no ifclause: and I think myself that in, say, 'X would have hanged him, but Y was against it' 'would have' is likewise a past indicative-indeed it is from this sort of example that we can see how the past tenses of will have come to be used as auxiliaries of mood for forming the conditionals of other verbs. To state what seem to be some grammaticd facts (omitting a reference to U the use of the words concerned in expressing wishes): Could have is sometimes a past indicative, sometimes a past subjunctive of the verb can. When it is the main verb and is a subjunctive, it does require a conditional clause with it. Can and its parts are not used as auxiliaries of tense or mood to form tenses or moods of other verbs. Would have, whether or not it is used as a past indicative or subjunctive of the verb will, is now commonly used (should have in the first person) as an auxiliary for forming the past subjunctive of other verbs: hence if it is the main verb it does in general require a conditiona1 clause with it.

IFS A N D C A N S

22 I

thetical' he means that it requires the addition of an $clause to complete the sense. And he goes on to say that 'could have' sentences also (though not so obviously) 'express hypothetical~',if not always at least in important cases, such as notably those where we say someone could have done something he did not actually do: in these cases 'could have' . . . is equivalent to 'would have . . . if. . .'. It will be clear at once that ~oweu-Smith, Moore, is not like distinguishing between the contention that 'could have' requires supplementation by an ij-clause and the quite different contention that its analysis contains an if-clause.1 On the whole it seems plain that it is the second (analysis) view that he wishes to argue for: but the argument he produces is that 'could have' is (in important cases) llke 'would have', the point about which is that it needs an if-clauseto complete it--& though h s , whch is an argument in favour of thefirst view, told in favour of the second view. But it cannot possibly do so: and in any event could have is liable, as we have already seen, to be in important cases a past indicative, so that the contention that it is like would have in requiring a conditional $-clause is unfounded. Nevertheless, it must be allowed that Nowell-Smith may still be right in urging that 'could have' means 'would have if' and that, as he eventually adds, 'can' means 'will if'. What has he to say in support of thts? He propounds two examples for discussion, which I tMc do not differ greatly, so I shall quote only the first. Here it is: He could have read Emma in bed last night, though he actually read Persuasion; but he could not have read Werther, because he does not know German.
It is true that he uses two different expressions: 'would have' k a (sup pressed) hypothetical, while 'could have' sentences express hypotheticals. But it does not look as if any distinction is intended, and if it is, the protracted
1

initial analogy between 'could have' and 'would have' seems irrelevant and misleading. Moreover, discussing the (unimportant) case of 'It could have been a Morris', he writes that 'it would be absurd to ask under what conditions it could or would have been a Morris' (my italics): this seems to show an indifference to the distinction that I am insisting on.

f F S AND CANS

This is evidentlv of the same kmd as Moore's 20-minute-mile example. The first thing that Nowell-Smith urges is that such a 'could have' statement is not a categorical, or a 'straightforward' categorical, statement. And his argument in favour of this view is derived from the way in w h c h we should establish its truth or falsity. No inspection of what the man actually did will, he says, verify directly that he could have done somedung else (here, read Emma) which he did not do: rather, we should, to establish this, have to show
d

(a) that

he has performed tasks of similar daculty sufficiently

often to preclude the possibility of a fluke, and (b) that nothing prevented him on this occasion. For example, we should have to establish that there was a copy of Emma in the house. T o refute it, on the other hand, we should have to show either 'that some necessary condition was absent' (there was no copy of Emma)' or 'that the capacity was absent'. That is, let us say, we have to show on the one hand that he had both the ability and the opportunity to read Emma, or on the other hand that he lacked either the ability or the opportunity. Nowell-Smith seems, at least at first, to be less interested in the matter of opportunity: for he says that we can establish 'directly', i.e. by considering what the facts at the time actually were, at least that he did not have the opportunity, that is, that somethmg did prevent h m , and he does not seem daunted by the obviously greater ddEculty of establishg, in order to establish that he could have done it, the general negative that there was nothing to prevent him. At any rate, it is at first upon our manner of establishing that he had (or had not) the ability to do this thing that he did not do that Nowell-Smith fastens in order to support his assertion that the 'could have' statement is not categorical. That the man had the ability to read Emma can not, he says, be established 'directly', i.e. by observing what happened on that past occasion, but only by considering what prowess he has displayed in the face of similar tasks in the past on other occasions, or displays now when put to the test: the

IFS AND CANS

223

argument that we have perforce to use is an 'inductive' one (and, he adds, none the worse for that). Now let us pass all thls, at least for the sake of argument.= What interests us is to discover why Nowell-Smith t M c s that these considerations show that 'He had the ability to read Emma' is not a categorical statement. I confess I fail to follow the argument: The very fact that evidence for or against 'could have' statements must be drawn from occasions other than that to which they refer is enough to show that 'He could have acted otherwise' is not a straighrfonvard categorical statement. But do we really know what is meant by a 'straightforward categorical statement'? certainly it is not the case that statements made on the strength of inductive evidence are in general not categorical-for example, the statement that the next mule born will prove sterile: this seems categorical enough. Perhaps thls example should be ruled out as not in point, on the ground that here there will some L y be 'direct' evidence relevant to the assertion, even if it is nit available at the moment. Could the same, I wonder, be said of the inductive conclusion 'All mules are sterile'? Or is that not categorical? I know that thls has been interpreted by some philosophers to mean 'If anything is a mule then it is sterile', but I see no reason to support that curious interpretation. The situation becomes still more puzzling when we remember that Nowell-Smith is about to generalize his theory, and to assert, not merely that 'could have' means 'would have . . . if', but also that 'can' means 'shall or will. .. if'. Suppose then that I assert 'I can here and now lift my finger,' and translatethis as 'I shall lift my finger if. . .': then surely h s will be 'directly' verified if the conditions are satisfied and I do proceed to lift the
1

Yet I think it is not hard to see that we cannot establish 'directly', at least

in many cases, that something 'prevented' him: he was drugged or dazzled, which prevented him from reading, which establishes that he could not have
read-but how do we know that being drugged or dazzled 'prevents' people from reading? Surely o n 'inductive' evidence? And, in short, to be prevented is to be rendered unable.

224

JFS A N D C A N S

fmger? If thls is correct, and if the theory is indeed a general one, then there seems to be no point in insisting on the nonavailability of 'direct' evidence, which is only a feature of certain cases. Incidentally, it is not in fact the case that to say d He could have done it' is always used in a way to imply that he did not in fact do it: we make a list of the suspectsin a murder case, all of whom we t h k could have done it and one ofwhom we h n k did do it. True, h s is not Nowell-Smith's case: but unless we are prepared to assert that the 'could have' in his case differs in meaning from that in the murder case, and so to rule out the latter as irrelevant, we are in danger of having to admit that even 'could have' sentences can be 'directly' verified in favourable cases. For study of the facts of that past occasion can prove to us that he did it, and hence that our original 'He could have' was c0rrect.I However, to proceed. Whether or not we should describe our conclusion here as 'categorical' it seems that it should still be a conclusion of the form 'he could have done so and so', and not in the least a conclusion concerning what he would have done. W e are interested, remember, in h s abilities: we want to know whether hc could have read Emma yesterday: we ascertain that he did read it the day before yesterday, and that he does read it today: we conclude that he could have read it yesterday. But it does not appear that t h s says anything about what he would have done yesterday or in what circumstances: certainly, we are now convinced, he could have read it yesterday, but would he have, considering that he had read it only the day before? Moreover, supposing the view is that our conclusion is not of the 'could have' but of the 'would have if' form, n o h g has yet been said to establish this, nor to tell us vhat follows the 'if'. To establish that he would have read it yesterday
.

There are, I should myself think, good reasons for not speaking of 'I can lift my finger' as being directly verified when I proceed to lift it, and likewise for not speaking of 'He could have done it' as being directly verified by the discovery that he did do it. But on Nowell-Smith's account I think that thae would count as direct verifications.

IFS A N D C A N S

225

i f . . ., we shall need evidence not merely as to h s abilities and opportunities, but also as to h s charactei, motives, and so on. It may indeed be thought, and it seems that Nowell-Smith does at least partly think t h s , that what follows the 'if' should be suppliable from the consideration that to say he could have, in the full sense, is to say not merely that he had the ability, which is what we have hitherto concentrated on, but also that he had the opportunity. For to establish this, do we not have to establish that certain conditions were satisfied, as for instance that there was a copy of Emma available? Very well. But here there is surely a confusion: we allow that, in saying that he could have, I do assert or imply that certain conditions, those of opportunity, were satirfed: but t h s is totally dfferent from allowing that, in saying that he could have, I assert something conditional. It is, certainly, entirely possible to assert something conditional such as 'he could have read Emma yesterday if there had been a copy available', could being then of course a sub-junctive: but to say t h s sort of dung is precisely not to say the sort of thing that we say when we say 'He could have acted otherwise', where 'could have' is an indicative-imp1 ying, as we now do, that there was no copy available, we imply that pro tanto he could not have acted otherwise. And the same will be true if we try saying 'He would have read Emma yesterday if there had been a copy available': this too certainly implies that he could not in fact have read it, and so cannot by any means be what we mean by saying that he could have read it. In the concludmg paragraph of h s discussion, Nowell-Smith does finally undertake to give us h s analysis not merely of 'could have', but also of 'can' (whch he says means 'will if'). And this last feature is very much to be welcomed, because if an analysis is being consciously given of 'can' at least we shall at length be clear of confusions connected with the idea that 6 could haves is necessarily a subjunctive.'
.

It must, however, be pointed out once again that if w e are to discuss the assertion that somebody can (now) do something, the previous arguments that our assertions are not categorical because they are based on induction and
1

226

IFS A N D C A N S

The argument of the last ~ a r a ~ r a p h as follows. It is runs 'logically odd' to say something of this kind (I am slightly emending Nowell-Smith's formula, but only in ways that are favourable to it and demanded by his own argument): Smith has the ability to run a mile, has the opportunity to run a mile, has a preponderant motive for running a mile, but does not in fact do so. From t h s it follows directly, says ow ell-smith, that 'can' means 'will if', that is, I suppose, that 'smith can run a mile' means 'If Smith has the opportunity to run a mile and a preponderant motive for running it, he will run it'. It seems, however, plain that nothing of the kind follows. Ths may be seen first by setting the argument out formally. Nowell-Smith's premiss is of the form Not (p and q and r and not -s) Logically odd (ability+ opportunity+ motive non-action) . Now from this we can indeed infer p 2 ((q and r) that is that If he has the ability, then, if he has the opportunity and the motive, he will do it. But we cannot infer the converse
3

s),

((q and r) 3 s) 3 p, or in other words that If, when he has the opportunity and the motive, he does it, he has
the ability to do it.

(I do not say t h s last is not s o m e h l g to which we should,

cannot be verified directly, whether they were good or not, must now be abandoned: because of course it is possible to verify this 'directly' by the method Nowell-Smith has sptcified in another connexion earlier, viz. by getting the man to try and seeing him succeed.

IFS AND CANS

227

when so put into English, assent, only that it does not follow from Nowell-Smith's premiss: of course it follows merely from the premiss that he does it, that he has the abhty to do it, accordmg to ordinary English.) But unless h s second, convene implication does follow, we cannot, accordmg to the usual formal principles, infer that p is equivalent to, nor therefore that it means the same as, (q and r) 2 s, or in words that abhty means that opportunity plus motive leads to action. To put the same point non-formally. From the fact that, if three thmgs are true together a fourth must also be true, we cannot argue that one of the three thngs simply means that if the other two are true the fourth will be true. If we could argue indeed in this way, then we should establish, from NowellSmith's premiss, not merely that
'He h a s the ability to do X' simply means that 'If he has the opportunity and the motive to do X, he will do X'

but also equally that

'He has the opportunity to do X' simply means that 'If he has the ability and the motive to do X, he will do X'

and hkewise that 'He has a preponderant motive to do X' simply means that 'If he h a s the ability and the opportunity to do X, he will do X'. For clearly we can perform the same operations on q and r as on p, since the three all occupy positions in the premiss. But these are fantastic suggestions. Put shortly, ~oweU-smith is pointing out in h s premiss that if a man both can and wants to (more than he wants to do anything else), he w d : but from thls it does not follow that 'he can' simply means that 'if he wants to he wd'. Nowell-Smith is strcgglhg to effect a transition from can to will w h c h presents difficulties as great as those of the transition from coul&o would: he puts up hk show of effecting it by importing the additional, and here irrelevant, concept of motive, whch needless to say is in general very intimately connected with the question of what 'he d do. '

228

IFS A N D C A N S

When, in conclusion, Nowell-Smith finally sets out his analysis of 'Smith could have read Emma last night', it is ths:
He would have read it, if there had been a copy, if he had not been struck b h d , Gc., Gc., and if he had wanted to read it more than he had wanted to read (this should be 'doY) anything else.

But so far from this being what we mean by saying he could have read it, it actually implies that he could not have read it, for more than adequate reasons: it implies that he was b h d at the time, and so on. Here we see that Nowell-Smith actually does make the confusion I referred to above between a statementb which implies or asserts that certain conchtions were fulfilled and a conchtional statement, i.e. a statement about what would have happened if those conditions had been fulfilled. Ths is unfortunately a confusion of a general h d that is not uncommon: I need only mention the classic instance of Keynes, who confused asserting on evidence h that p is probable with asserting that on evidence h p is probable, both of which can be ambiguously expressed by 'asserting that p is probable on evidence-h', b i t only the former of whch asserts that p is (really) probable. Here similarly there is a confusion between asserting on the supposition (or premiss) that he had a copy that he could/would have read it, and asserting that on the supposition that he had a copy he could/would have read it, both of which can be ambiguously expressed by 'asserting that he could/would have read it on the supposition that he had a copy', but only the former of whch asserts that he (actually) could have read it. T o some extent, then, we learn from studying Nowellsmith's arguments lessons similar to those that we learned in the case of bore. But some are new, as for instance that many assertions about what a man would have done or will do depend, in critical cases, upon premisses about h s motives as well as, or rather than, about h s abilities or opportunities: hence these assertions c m o t be what assertions about h s abilities mean.'
I

Yet here it must be point^ ; out once more that it has not been shown that 1 all assertions about what he would have done are so dependent, so that this

IFS A N D C A N S

229

On one point I may perhaps elaborate a little further. It has been maintained that sometimes when we sav 'He could have done X' ths is a conditional: it requires completion by an ifclause, typically 'if he had had the opportunity', and so does not require us, if we are to establish its truth, to establish that he did infact have the opportunity. Sometimes on the other hand it is a past indicative, implying that he did have the opportunity: in which case we do, to establish its truth, have to establish that certain conditions were satisfied, but the assertion is not to be described as a conditional assertion. Now while I have no wish to retract ths account in general or in all cases, I doubt whether it is the whole story. Consider the case where what we wish to assert is that somebody had the opportunity to do s o m e t h g but lacked the ability-'He could have smashed that lob, if he had been any good at the smash': here the i$-clause, whch may of course be suppressed and understood, relates not to opportunity but to ability. Now although we might describe thewhole sentence as 'conditional', it nevertheless manages to assert, by means of its main clause, s o m e h g 'categorical' enough, viz. that he did have a certain opportunity. ~ n in the same way Nowell-Smith's 'He could d have read Emma, if he had had a copy', does seem to assert 'categorically7 that he had a certain ability, although he lacked the opport&ity to exercise it. Looking at it in this way, there is a temptation to say that 'could have' has, besides its 'all-in' sense several more restricted senses: t h s would be brought out if we said 'He could have smashed it, only he is no good at the smash' or 'He could have read Emma but he had no copy', where, we should say, 'could have' is being used in the restricted senses of opportunity or of ability' only, and is a past indicative, not a past conditional.
d

particular argument against the analysis o f 'could have' as 'would have if' is not conclusive: in particular, it does not dispose of the possible suggestion that 'could have' means 'would have if he had tried', for here considerations of motive may be irrelevant. I talk here and throughout of 'ability' and 'opportunity' only: but I realize that other abstract nouns like 'capacity', 'skill', and even 'right' arc

230

IFS A N D C A N S

Ths view might be reinforced by considering examples with

the simple 'can' itself. W e are tempted to say that 'He can' sometimes means just that he has the ability, with nothing said about opportunity, sometimesjust that he has the chance, with nothing said about ability, sometimes, however, that he really actuallyfully can here and now, having both ability and opportunity. Now nobody, I thmk, would be tempted to say that 'can,' where it means one ofthe two lesser dungs, for example, 'has the opportunity', i.e. 'can in the full sense if he has the ability', is grammatically a subjunctive or conditional. Perhaps, then, it was not correct to describe 'He could have', either, as always a conditional where it asserts ability or opportunity only, wi& nothng said about the other, or even where the other is denied to have existed. The verb can is a peculiar one. Let us compare it for a moment with another peculiar verb, know, with G h c h it shares some grammatical peculiarities, such as lack of a continuous present tense. When I say that soinebody kr~olvs what the thng in my hand is, I may mean merely that he has the ability to identify it given the opportunity, or that he has the opportunity to identitjr it if he has the ability, or that he has both. What do we say about know here? Certainly we are not prone to invoke the idea of a conditional, but rather that of different senses, or perhaps the still obscure idea of the dispositional. I must be content here merely to say that I do not t h d that the old armoury of terms, such as 'mood' and 'sense', is altogether adequate for handling such awkward cases. The only point of w h c h I feel certain is that such verbs as can and know have each an all-in, paradigm use, around whch cluster and from which hvagate, little by little and along different paths, a whole series of other uses, for many of which, though perhaps not for all, a synonymous expression ('opportunity', 'realize', and so on) can be found. It is not unusual for ail audience at a lecture to include some
equally involved. All these terms need listing and elucidating before w e really get to grips with 'can'.

IFS AND CANS

231

who prefer t h g s to be important, and to them now, in case there are any such present, there is owed a peroration. Why, in short, does all t h s matter? First, then, it needs no emphasizing that both ifand con are highly prevalent and protean words, perplexing both grammatically and pMosophically: it is not merely worth wMe, but essential, in these studes to discover and cans, and to remove the confusions they the facts about engender. In pMosophy it is con in particular that we seem so often to uncover, just when we had thought some problem the settled, grinning residually up at us l ~ k e frog at the bottom of the beer mug. Furthermore and secondly, we have not here been dissecting these two words in general or completely, but in a special connexion which perhaps no one will hold trivial. It has been alleged by very serious phllosophen (not only the two I have mentioned) that the things we ordinarily say about what we can do and could have done may actually be consistent with determinism. It is hard to evade all attempt to decide whether t h s allegation is true-hard even for those who, llke myself, are inclined to h k that determinism itself is still a name for n o t h g clear, that has been argued for oilly incoherently. At least I should like to claim that the arguments considered tonight fail to show that it is true, and indeed in failing go some way to show that it is not. Determinism, whatever it may be, may yet be the case, but at least it appears not consistent with what we ordinarily say and presumably think. And finally there is a thlrd point. Reflecting on the arguments in t h s lecture, we may well ask ourselves whether they might not be as well assigned to grammar as to philosophy: and this, I think, is a salutary question to end on. There are constant references in contemporary philosophy, w h c h notoriously is much concerned with language, to a 'logical grammar' and a 'logical syntax' as though these were things distinct from ordinary grammarian's grammar and syntax: and certainly they do seem, whatever exactly they may be, different from traditional grammar. But grammar today is itself in a state of dux; for fifty years or more it has been questioned on all hands and counts

ifs

IFS AND C A N S

whether what Dionysius Thrax once thought was thc truth about Greek is the trith and the whole truthibout all language and all languages. Do we know, then, that there will prove to be any ultimate boundary between 'logical gammar' and a revised and enlarged Gronlnlor ? In the history ofhuman inquiry, ph~losophy the place of the initial central sun, seminal a i d has tunlultuous: from time to time it throws off some portion of itself to take station as a science, a planet, cool and well regulated, progressing steadily towards a distant final state. This happened long ago at the birth of mathematics, and again at the birth of physics: only in the last century we have witnessed the same process once again, slow and at the time almost inlperceptible, in the birth-of the science of mathematical logic, through the joint labours of pl~losophers and mathematicians. Is it not possible that the next century may see the birth, through the joint labours of philosophers, grammarians, and numerous othkr students of language, b f a true and comprehensive srieme of language? Then we shall have rid ourselves of one more part of philosophy (there will still be plenty left) in the only way we ever can get rid of philosophy, by kiclung it upstairs.
-