Mind, 1969.
VII.-CRITICAL NOTICES Moral Notions. BY Julius Kovesi. London: Routledge and Kegan Paul, 1967. Pp. vii+166. 20s.
CERTAIN ways of stating a distinction between matters of fact and matters of value have become firmly entrenched in English-speaking philosophy since at least the time of Hume and Kant. A handful of guerrillas (Miss Anscombe, Mrs. Foot, Miss Murdoch) have only counted as an Oxford resistance movement. Now Mr. Kovesi launches (from Western Australia) a lightning campaign of a mere 40,000 words which, I think, decisively and permanently alters the balance of power. No one who reads even the first few pages of Moral Notions should ever again be able to think of the yellow–good ploy except as a quaint museum piece. The book is both highly compressed, and strongly original; this makes it somewhat intoxicating. Fresh ideas crowd in so thickly that the pattern of argument is easy to lose. Time and again a startling paradox brings us to a halt, and we want a recapitulation of the steps in the argument that got us there. Nearly always we are driven back to realise that a favourite preconception has been subtly charmed away.
Kovesi is disarmingly frank about the paradoxical nature of his conclusions. (Examples: moral notions are not evaluative but descriptive; descriptive notions are evaluative.) This is the result of arguing against a theory in a terminology which itself embodies the very theory one is arguing against. Not all the Wittgensteinian ladders can be thrown away. Kovesi is also disarmingly modest about his new claims. The most devastating of all turns out to be the innocent-seeming “I would like to introduce here two technical terms borrowed from Aristotle”. With these terms – form and matter – he constructs in Chapter I a philosophical framework which enables him, in the rest of the short book, to solve or dissolve an impressive list of standard problems in moral philosophy. This framework is a general theory of concept-formation, meaning, and rules of usage. It owes (despite disclaimers) a certain amount to Wittgenstein, but there is a strongly Kantian theme: a new version of the Copernican revolution. Nature and morals are both man-made; Nature, for reasons not altogether un-Kantian; morals, for precisely the same, and therefore quite un-Kantian, reasons.
I shall give a fairly detailed account of Chapter 1, because it is so fundamental to the whole. It is entitled “Between Good and Yellow” and it begins with the deceptively familiar statement: “‘Good’ is not like ‘yellow’.” But it immediately claims that this is always asserted for the wrong reasons, to point to the wrong differences. ‘Good’ and ‘yellow’, so far from being typical representatives of their classes, are thoroughly degenerate. Almost every word, descriptive as well as evaluative, is unlike ‘yellow’. Almost every word, evaluative as well as descriptive, is unlike ‘good’. It was a cardinal error to single out a colour-word, and the most general word of commendation.
To substantiate these sweeping claims about almost all words, Kovesi needs a general theory of words. In introducing the term matter, he points out that objects like tables may vary in their characteristics without ceasing to be tables; other objects might not count as tables even though they possessed the characteristics of some which did count as tables. He calls material elements of a notion those which can vary without affecting the applicability of the notion. (“Notion” is used to cover not only object-concepts like table but also moral notions like murder.) The material elements of a table would include its wood, or its shape; the material elements of a murder would include the nature of the weapon, or the wound. But now we have the Kantian question: why are just these elements organised into a single notion? What is the principle of unification? The answer is the formal element: this is whatever gives the answer to the question, Why do we call various objects tables, or various killings murders, and others not? The formal element of tables is their use for sitting at, eating off, etc.; the formal element of murder is intentional killing of an innocent person for purposes of personal gain or satisfaction.
One might immediately suspect the exhibition here of that tedious old philosopher’s prop, the table; but here again the familiarity is deceptive. We should not dismiss the table because it is an untypical, artificial object. On the contrary, this is its virtue: we can learn a great deal from artefact-words, which are indeed a far bulkier, and perhaps more central, segment of our vocabulary than we care to think. (Compare this sequence picked from the Oxford Pocket Dictionary – not quite at random, but only so as to avoid clusters of technological and scientific vocabulary: book, boom (n.), boomerang, boor, boot, booth, booty. Of these seven successive words, five are plain artefact words, one has artefactual connections, and one is a moral notion.) Now the formal element of any artefact word is obviously a human need. And the material elements will be any features of the objects that happen to serve that need.
So one answer to Kant’s problem about the unification of the elements of a concept comes to this (though Kovesi does not put it this way): the elements that get unified are those that can be (literally) manipulated in making something. At this stage we do not necessarily have a word; we have something perhaps more exciting, a direct connection with the world of creatures without speech. But of course, as Kovesi does go on to say, tables are not just artefacts; the life of this notion includes, besides the needs of eating in a sitting posture, the need to identify, list in an inventory, keep in stock, buy, sell, and much else; and some of these needs can only be met by the verbalised notion, which can then be extended to include things that are not artefacts.
To arrive at moral notions, there is no gap to bridge, only differences to mark. Blaming, praising, finding fault, changing attitudes, are more sophisticated activities; they generate moral notions which contain as material elements features of the life of the very same class of agents who entertain the ends which are the formal element. Moral notions provide not only rules for thinking, but also rules for behaving. To call some act murder, or another act inadvertent, is not to point to any empirical similarities in a class of acts. It is to group certain features from the point of view of fault-finding. Only if we grasp this point will we be able to pick out further instances of murder, or inadvertence.
It now becomes clear why both ‘yellow’ and ‘good’ are degenerate. ‘Yellow’ lacks a formal element. A thing is described as yellow simply in virtue of its having that very property. There is no point to calling something yellow except, trivially, to describe what colour it is. There is no principle of selection of material elements when there is only one material element.
By contrast, ‘good’ lacks material elements. It indicates all too clearly what the point is – to evaluate – but there is no limit to the number of characteristics in virtue of which something can be good.
A question which Kovesi seems deliberately to ignore is this. Are colour-words really in a class by themselves, or are there other ‘simple’ property-words at the same end of the spectrum? On page 19 there is an allusion, impossibly condensed and apocryphal, to a general theory about “the world of raw data” whereof “one really cannot speak” for the odd, or at least oddly stated, reason that “the very description of it . . . lies outside it”. But colour words are “the nearest analogy to the words which would mirror the world of data”. Kovesi’s point here is to deny that there is any more of a gap between “raw data” and evaluation than there is between them and description. Schematically we have
Al. Statements about raw data.
A2. Statements about empirical characteristics.
B. Statements about inanimate objects.
C. Statements about animate objects, involving moral notions.
Kovesi argues that if, following Hume, we tried to say that the moral notions in C-statements have no counterparts in the real world, we have exactly the same reasons for saying that the object-notions in B-statements have none either; if there are no murders, there are no tables. Both notions are constituted by empirical properties collected from different points of view. But colours are only one small sub-class of empirical properties; and ‘wooden’, ‘having four legs’, ‘sharp steel moving into place occupied by body’, though empirical, are a very long way removed from “raw data”.
However, ignoring this problem about the difference between Al and A2, I should like to applaud Kovesi’s move (pp. 11–12) against those who, impressed by the Hume gap (B–C) seek to bridge the gap by means of an imperative general premiss.
As if we could solve the problem of epistemology by the following way: ‘There are two sorts of statements that we can make about a piece of furniture: examples of the first sort are, “This piece of furniture has four legs and a flat surface”, “This has such-and-such a height”: examples of the second sort are, “This is a table”, “This is a desk”, the first sort of remark is usually given as a reason for making the second sort of remark, but the first sort does not by itself entail the second sort, nor vice versa. Yet there seems to be some close logical connection between them. Our problem is: what is this connection? What we need is a major premise: whatever has four legs and a flat surface and has such-and-such a height is a table. Then we can proceed: this has . . ., therefore this is a table.’
Two other ways of making the point are important. (1) There is a familiar argument to the effect that two things can be exactly alike in every respect except that one is X and the other is not – except when X is an evaluative term like ‘good’. A thing can only be good because of certain other characteristics it has. It is now claimed that almost every term in the language is also an exception. Something is a table only because of certain “other” characteristics it has. (2) “Decisions and attitudes, insights, wants, needs, aspirations and standards do enter into our moral notions. But they do not enter from the top” (pp. 24–25). They enter in as one of many ways, equally rational, of organising features of situations into unified notions.
Emphasis on the place of decisions and attitudes in moral language could foster a subjectivist interpretation. This is firmly excluded. Tables could not exist if they served only unshared private purposes. This is not because there would be insufficient business for carpenters, but because the formal element of the notion would be lacking. And without this we could not follow a rule for using the term. The now standard argument against private language is extended to moral language.
There is some difficulty here because Kovesi has not formally laid out relations between his Aristotelian terminology and terms such as meaning, rules, and definition. In Chapter I he says that a thing cannot be defined in terms of its material elements (“and this is perhaps one of the secrets the anti-naturalists are after”, p. 8) but only in terms of the formal element, which determines which material elements can and which cannot be accepted as constituting the thing. But in Chapter II (“Following Rules and Giving Reasons”) he speaks of “the defining characteristics of (= among?) the material elements of a thing” (p. 41). This seems inconsistent, for he is denying here that these characteristics, which he calls “recognitors”, constitute the meaning of the term, or the rules of its use. He has staged the introduction of an artificial term, “tak”, which is supposed to be the name for a certain sort of figure; but to be able to recognise the figures that are called taks is not to have a rule for the use of this term, to understand its meaning. To have these, it is necessary to show what is the point of calling things taks; which is to show the need for the term in a “way of life” – to provide the formal element. Recognitors, we might say, provide necessary conditions for the use of a term, but a full understanding requires sufficient conditions as well, which only a “way of life” can supply. To be able to follow a rule is to be in possession of the formal element, which in turn is to know what the material elements, in different cases, “amount to”:
Supposes someone is trying to buy flowers, but cannot find any. If he then buys a packet of paper streamers we can say that he was trying to buy decorations, but if he comes home with a box of chocolates, we can say that he was looking for a present (p. 60).
We could never learn the use of moral words if they expressed only our decisions, attitudes, etc. It is one of the many seeming paradoxes of Moral Notions that the fact that our interests and attitudes enter twice into our moral notions (as material elements as well as formal ones) makes them less, not more, subjective.
Only those features of our lives can be incorporated into these notions that are shared by any of us and are recognizable by any of us, and in turn the formation of the notion must itself be done from the point of view of anyone (p. 57).
One reason for saying that this is a gain in objectivity is that inanimate objects, which non-moral notions are about, “cannot put up any resistance” if changing conceptual interests of ours alter the selection of features – if, for instance, we come to speak of galaxies instead of constellations. But we, as material of our own moral notions, would indeed resist such a change. Again, when arguing against Hume (p. 72) Kovesi claims that feelings enter into moral judgment not as expressed by the speaker towards a happening, but as one of the “raw materials which are the object of our reason”. It is not my feeling, but anyone’s (not, of course, everyone’s) feeling that is an actual constituent of the notion of parricide. If I did not disapprove of this, I would not understand the notion. Again, a notion does not reflect my wants but anybody’s wants. “The tension between my preferences and what is good, or between what I would like to do and what I ought to do, enters into our life already with our language” (p. 57).
The exaggerated respect accorded to moral universalisability in Kantian and neo-Kantian moral philosophy is, according to Kovesi, nothing more than an unclear recognition of the publicity of moral language. “A large number and surprising variety of moral philosophers seem to talk about our moral life and language as if each of us spoke a private language and yet paradoxically lived in society, as if our moral notions were private notions that we try to make universal” (p. 57); and it is “not because we happen to be such nice people that we formulate our notions from the point of view of anyone, but because our language is public” (p. 111).
There has been much discussion recently whether there is a real, or only seeming, opposition between neo-Kantian and Existentialist moral philosophy. Professor Hare showed (Freedom and Reason p. 38) that he could deal with what has become a rather boringly familiar conflict-situation from Sartre. But Kovesi has an easy way with both of them. His quarrel with Hare I shall not spell out; his disposal of Existentialist “uniqueness” will have to wait until we have looked at a further extension of his conceptual framework: the distinction between complete and incomplete moral notions.
Certain types of act have already had their features completely selected from the moral point of view: examples are murder, theft, treachery. To say that the features are completely selected from a point of view is to say that we possess a rule for identifying instances of the act in question that are (in the case of the moral point of view) all of them right, or all wrong. We do not need to add a remark that an instance of such an act is wrong. This is already decided for us by the formal element of the notion itself. In such cases, where the terms ‘right’ and ‘wrong’ serve only as reminders, we have what Kovesi calls a complete notion. We also have a moral principle. A moral principle is just a complete moral notion, formulated with emphasis on the formal element; or (in the case of a newly propounded principle) the formal element of a possible complete moral notion which has not actually been formed.
An incomplete notion (such as killing, entering a building, etc.) does not enable us to follow a rule for picking out right and wrong acts. This is precisely why we need moral judgments. A moral judgment attaches to an incomplete moral notion a moral “discriminator”: ‘right’ and ‘wrong’, which are now doing an original job, instead of just reminding us of a job done by those who formed a complete moral notion.
Now for the alleged problem about unique situations:
It is only by the help of moral judgments, or at least by the help of complete or incomplete moral notions, that an existentialist can produce his examples of extraordinary situations where no principle can help the moral agent to make his decision. What these extraordinary examples show is only that we have not got a single term to sum up the whole situation in which one ought to do one thing rather than another. We need a whole novel to state all the relevant facts (p. 112).
Universalisability enters because a moral principle has collected, and a moral judgment collects, features of situations which are not unique but would be relevant to anyone else in situations of that type. Moral judgments are universalisable not because they are moral but because they are rational. Reason, not ‘moral’, carries the weight.
The paradox “Moral notions do not evaluate the world of description but describe the world of evaluation” (p. 119, repeated on dust jacket) has already been half explained. (“Evaluation is not an icing on a cake of hard facts”, p. 25.) The other half is explained when we note that human situations and predicaments cannot be described except by the use of already-formed moral notions. “Situations are not like puddles that we can step in and out of; to be in a situation is to be related to other human beings in a certain way” (p. 119). Not all human relations are moral; but those that are cannot be described in an evaluatively neutral idiom.
In the last chapter, “Evaluation and Moral Notions”, Kovesi aims yet another blow at the evaluative/ descriptive dichotomy. Not only is it not the case that descriptive statements state how things are and evaluative judgments judge how things ought to be – this “is not so much a misleading picture as not the picture of anything ” – but “without moral notions there would be nothing to express an attitude to, nothing to make a decision about, nothing to evaluate” (p. 148). So evaluation has no place in morality. Here I think Kovesi over-reaches himself. He has already stated several times that moral notions are formed not only (like all notions) by ourselves, but also (unlike other notions) about ourselves. But the fact that our evaluations can themselves be evaluated does not throw any more doubt on the evaluative function than the duplicate role of interests, wants, etc., does on them. However it turns out that Kovesi has a stronger reason for denying that in our moral life we are evaluating: a new version of the thesis that ‘right’ is a moral term but ‘good’ is not. He starts with the familiar point that we never judge something to be good, but only to be a good such-and-such. We always evaluate under a certain description; and what we are doing is to decide about the qualities of particulars falling under one and the same description. In moral decisions, however, we are faced with acts of different descriptions. We do not choose the best available instance of an act falling under one description, but rather we choose the right one out of several possible descriptions. Hence the difference between evaluative judgments and judgments of obligation. Both are exercises of “the same rules of rationality”, but in different fields. Practical syllogisms cannot have (merely) evaluative premisses.
I have already noted several issues in moral philosophy to which Kovesi supplies an easy, and almost incidental, resolution. I shall notice one more. In half a page (133) the three-cornered fight between Intuitionists, Deontologists, and Teleologists, is quietly broken up. They are all perfectly right – given a suitable sample of moral notions. Intuitionism is a perfectly correct account of complete moral notions; if a term already specifies all we need to know in order to judge the act in question wrong, we do not need any additional relevant facts. Intuitionism errs, of course, in claiming that all moral notions are complete notions. Deontologists, or whatever we call those who claim that acts are justified by reference to intentions, could be perfectly right about a certain class of incomplete moral notions; if a term leaves the field of intentions unspecified, then an act referred to by such a term could be justified only by adducing additional relevant information, which could well be about intentions. Such a theory errs in claiming that all moral notions are incomplete in this one respect. Teleology could be perfectly right about notions which are incomplete in respect of consequences, but in error if it claimed that all moral notions were incomplete in this respect.
In order to discuss Kovesi’s central thesis, and its more striking applications, I have been obliged to omit much that is equally striking; in particular, his sustained criticism of Hare’s views about imperatives; most of his critique of Hume; his discussion of Aristotle, both on ‘good’ and on the Mean; the distinction between acting according to a principle and acting on principle; between rules and regulations. I have mentioned only one of several passages of constructive analysis, where Kovesi invites us to consider how certain terms, of which he is endlessly inventive (taks, wousins, mistickets, savingdeceit) might have been introduced and given a use. It may seem odd that a long review of a short book must ignore so much. The reason is, once again, that so much is packed into so little space; and much of that is tantalisingly brief and often enigmatic.
One or two small grumbles. The jargon-phrase “open texture” on pages 8–10 adds nothing and should have been avoided. Next, Professor J. L. Austin. Kovesi attributes the genesis of some of his ideas to Oxford companions in 1956. Some, in fact, were being put about by Austin in 1946. This would not be worth mentioning, were it not that Austin is here acknowledged only as the source of a somewhat inept illustration about levers (p. 14). Austin was indeed (like Wittgenstein) fond of illustrations from ironmongery; but it happens to be a subject on which (very unlike Wittgenstein) he was remarkably ill-informed.
This is a thoroughly disturbing book, and should lead to some agonising reappraisals. Moral philosophy courses are due for a new look; Kovesi’s challenge is timely, and there is material here for several dozen prolific seminars.
University of St. Andrews BERNARD MAYO