Particularism and Universalizability

It is commonly assumed by particularists as well as by generalists that moral particularism is incompatible with universalizability. Particularists, therefore, try to disprove universalizability, while generalists use the supposed incompatibility as an easy way to dismiss particularism. Universalizability seems far better justified than particularism and if one of them has to be given up, the odds are surely against particularism. This common assumption is wrong and the alleged conflict between particularism and universalizability is due to a misunderstanding of the universalizability thesis. Consequently, particularists need not try to disprove a thesis which is widely held to be a conceptual truth and a necessary feature of moral language. Generalists need to find substantial arguments against particularism and cannot content themselves with holding that particularism is untenable because it contradicts a conceptual truth of moral language.

1. Let us start with Roger Crisp’s recent discussion of the issue. He writes:

According to Dancy’s understanding of the universalizability thesis, if some action is judged to be right, then any relevantly similar action must likewise be judged to be right. It is clear how denying universalizability is likely to commit one to particularism about reasons, so let me briefly consider whether it should indeed be denied. (Crisp 2000: 40)

The relevantly similar action in Crisp’s formulation of the universalizability thesis must be similar in morally relevant respects. Further, it must be similar in all morally relevant respects: Two actions are relevantly similar if and only if they are similar in all the relevant respects. With these clarifications the universalizability thesis can be put as follows (for ease of discussion I speak of actions as being right instead of judging them to be right):

(U) If an action is right, then any action similar in all morally relevant respects is also right.

It is obvious that (U) is analytic and cannot possibly be denied: Two actions similar in morally relevant respects are eo ipso morally similar. Were they not morally similar, there would be a morally relevant difference between them and they would not be similar in morally relevant respects. It would be self-contradictory to admit that two actions are similar in morally relevant respects and then to judge one to be right and the other to be wrong.

Since (U) is analytic, particularism would not warrant serious attention if particularists were committed to denying (U). But why should particularists want to deny it? Which particularist doctrine is supposed to contradict it? The essential point of particularism is the rejection of moral principles along with the rejection of the claim that properties which are morally relevant in one case are likewise morally relevant in other cases. Only if (U) is thought to be at odds with these claims do particularists have reason to reject it. The idea that (U) implies moral principles is common and is held, in particular, by R. M. Hare, who is the main target of Jonathan Dancy’s criticisms of universalizability. Yet, in spite of its general acceptance, this idea is wrong. (U) does not imply moral principles. If, for example, someone judges an action to be wrong because it is a lie, all that she is committed to by (U) is that she has to judge as wrong every action similar in all morally relevant respects. This, however, does not imply, that she has to judge every lie as wrong, and there is no way to derive this (or any other) moral principle from applying (U) to the particular moral judgment. (U) implies nothing as to which actions are similar in all morally relevant respects and the point of (U) is not to apply a judgment made about one action to other actions or, as Dancy puts it, ‘to drive us in what may seem a very simple-minded way from one case [...] to another which happens to resemble the first in some limited way’ (Dancy 1993: 82). (U) is too trivial to accomplish this. Its point is that someone who passes different moral judgments on two actions which appear to be similar in morally relevant respects must be able to pinpoint a morally relevant difference between them or else revise one of her judgments. Dancy himself does rely on this pattern of argument as the following quotation shows:

The coherence of an overall outlook can be questioned in the following way without generalist motivation. ‘Here you think the fact that she was unhappy functions as a reason in one way, and there you took it to function in quite another. To me they seem to be functioning in much the same way both times, so that I can’t really see how you can distinguish in the way you do. What is the relevant difference between the two cases?’ This challenge can always be made within the constraints of particularism, and there must be an answer to it if one’s position across the two cases is to emerge as coherent. (Dancy 1993: 63f.)

In this reasoning we see a clear application of (U). Since even particularists must presuppose (U), the alleged conflict between universalizability and particularism must lie somewhere else.

2. The conflict arises only if the universalizability thesis (U) is confused with the following thesis:

(P) If some action a₁ is right, then any action which has all the properties in virtue of which a₁ is right, is also right.

(P) implies the existence of moral principles, and there would indeed be a conflict between universalizability and particularism if (P) were an adequate interpretation of the universalizability thesis. Many philosophers fail to distinguish between (U) and (P). They start talking about (U) and then interpret (U) as (P). As a consequence their talk about universalizability reflects an ambiguity between (U) and (P) and it is this ambiguity which accounts for the alleged conflict between particularism and universalizability. Some claims made about ‘the universalizability thesis’ are true only of either (U) or (P) but not of both. In particular, the claim that the universalizability thesis is analytic and a necessary feature of moral language is true only if applied to (U) but wrong if applied to (P). The claim that the universalizability thesis implies a commitment to moral principles is true only if applied to (P) and wrong if applied to (U).

The confusion between (U) and (P) can be traced back to Hare’s introduction of the universalizability thesis in his Freedom and Reason. There he writes:

If a person says that a thing is red, he is committed to the view that anything which was like it in the relevant respects would likewise be red. The relevant respects are those which, he thought, entitled him to call the first thing red; in this particular case, they amount to one respect only: its red colour. [...] ‘This is red’ entails ‘Everything like this in the relevant respects is red’ simply because to say that something is red while denying that some other thing which resembles it in the relevant respects is red is to misuse the word ‘red’; [...].
[...] And so if a person who says ‘This is red’ is committed also to the proposition ‘Everything like this in the relevant respects is red’, then he is, further, committed to the proposition that there is a property such that this has it and such that everything which has it is red. And the second part of this proposition contains no singular terms, and can therefore be called properly universal. (Hare 1963: 11)

Universalizability in this sense is, according to Hare (1963: 15ff., 39), a feature not only of descriptive judgments but of all judgments with descriptive meaning. Because moral judgments too have descriptive meaning they are universalizable in the same sense as descriptive judgments. The quoted passages, therefore, hold equally if we replace the descriptive judgment “This is red” with the moral judgment “This action is right”. Applied to moral judgments, then, Hare is moving here from (U) to (P). But his reasoning is fallacious. Hare first assumes that if an action a₁ is right, then any action similar in morally relevant respects is also right. Then he claims that the morally relevant respects are those properties of a₁ in virtue of which it is (judged to be) right. Therefore, any action with those properties in virtue of which a₁ is right, is also right – which is just what (P) says. As it stands, however, this argument is incomplete and contains as a suppressed premise the following thesis:

(R) Two actions a₁ and a₂ are similar in all morally relevant respects if a₂ shares with a₁ all the properties which are the reasons why a₁ is right.

Dancy has effectively shown that (R) is untenable. Besides the properties in virtue of which a₁ is right, a₂ may have some additional morally relevant properties which may result in a₂ not being similar in all morally relevant respects to a₁. From the fact that a₂ shares with a₁ all the properties in virtue of which a₁ is right, one cannot infer that a₂ does not have any further morally relevant properties which make it wrong. Therefore, one cannot infer that a₁ and a₂ are similar in all morally relevant respects. Consequently, (R) is wrong.

How could anyone have thought otherwise? Perhaps the reason for holding (R) to be true was this: First, one holds that if a₂ has those properties in virtue of which a₁ is right, then a₂ and a₁ have the same morally relevant properties. Then it is assumed that if two actions have the same morally relevant properties, they are similar in all morally relevant respects. These assumptions naturally lead to (R) if an ambiguity in talking about two actions having the same morally relevant properties is neglected. That two actions have the same morally relevant properties could mean either that those properties which are morally relevant for a₁ (i.  e. in virtue of which a₁ is right) do occur in a₂ as well, or that a₂ shares with a₁ the properties in virtue of which a₁ is right and does not have any further morally relevant properties (which do not occur in a₁). That two actions have the same morally relevant properties according to the first interpretation does not exclude the possibility that a₂ has some further morally relevant properties which do not occur in a₁. Therefore, one cannot infer that a₁ and a₂ are similar in all morally relevant respects and have the same moral property. If two actions have the same morally relevant properties according to the second interpretation, they are similar in all morally relevant respects and have, therefore, the same moral property. But based simply on the fact that a₂ has those properties in virtue of which a₁ is right, one cannot infer that a₁ and a₂ have the same moral properties as interpreted in this second sense. That is, one cannot infer that a₁ and a₂ are similar in all morally relevant respects.

Another way to see the fallacy in Hare’s reasoning is this: Hare claims that the expression ‘like this’ can be substituted by a term which describes the respects in which two actions are alike. This claim, however, holds only in one case: If two actions are both judged to be right, then – in retrospect – one can find (or invent) a term to describe the respects in which the two actions are alike and in virtue of which they are both right. Yet, from this it does not follow that it is possible to judge one action to be right and then to find a term to describe the respects in virtue of which this action is right, such that it further holds: every other action which has the properties to which this term refers is right too. Neglecting this results in neglecting the difference between (U) and (P).

Hare’s fallacy is repeated by Dancy:

Hare holds that moral judgements are universalizable in this sense: a person who makes a moral judgement is committed to making the same judgement of any relevantly similar situation. A situation is relevantly similar to the first if it shares with the first all the properties that were the person’s reasons for his original judgement. So if we come across a case which resembles the first one in those limited respects, we are compelled to make the same judgement or to retract our first judgement. In this sense each judgement creates a moral principle. Where our reasons for approval were features F1-Fn, our judgement establishes for us the principle ‘All actions that have F1-Fn are right’. (Dancy 1993: 80)

Here Dancy is moving from (U) via (R) to (P). In a recent article he holds that (U) and (R) together make up the universalizabilty thesis:

The principle [of universalizability] is in two parts:
i. If we judge one action right, we must judge any other relevantly simi­lar action right.
ii. An action is said to be relevantly similar if, roughly, it shares with the first action all the properties which were reasons why the first action was right. (Dancy 1999: 28)¹

Dancy’s (i) and (ii) correspond to my (U) and (R). (U) and (R) together entail (P). The view that the universalizability thesis consists of these two parts was, to my knowledge, first explicitly stated by Sinnott-Armstrong (1999: 5) in paraphrasing the passage from Dancy (1993) quoted above. Traditionally the universalizability thesis referred only to (U) and almost all philosophers who talk about universalizability formulate principles similar to (U) (as does Crisp in the quotation at the beginning of this paper). What is more, all the known arguments to the effect that the universalizability thesis is analytic and a necessary feature of moral language have been advanced with regard to (U) only. (U) and (R) are two distinct theses, each of which requires its own justification. (R) is no explication of (U) and cannot be inferred from it. Yet, many philosophers moved from (U) to (R) and (P) by way of the fallacious reasoning exposed above and then put forward claims about (P) which they have, in fact, established only with regard to (U). As a consequence of neglecting the difference between (U) and (P) and the resulting ambiguity in the use of the term ‘universalizability thesis’, it often happens that defenders and opponents of universalizability talk about different things. The debate about particularism and universalizability is fruitless as long as some of the debaters conceive of the universalizability thesis as (U) and others conceive of it as (P) (or, which is the same, as composed of (U) and (R)). Whereas Dancy’s attack on universalizability is directed only at (R) and (P), supporters of universalizability respond by defending (U).

3. A case in point is Crisp’s recent defense of universalizability. He renders the universalizability thesis as (U), claims that the denial of this thesis commits one to particularism and then successfully defends it against criticisms put forward by Peter Winch. This defense of (U), however, is completely beside the point with respect to Dancy’s arguments against (R) and (P). After having rebutted Winch’s arguments Crisp concludes that the plausibility of the universalizability thesis provides a further argument against particularism:

Particularism about reasons implies the falsity of the universalizability thesis. Since, even if it is not practically important, that thesis seems plausible, we have here a further argument against particularism. (Crisp 2000: 42)

As should be clear by now, this conclusion is wrong. Particularism implies the falsity of (R) and (P). What Crisp has shown to be plausible is (U). (U) is compatible with particularism and its plausibility provides no further argument against particularism.

The difference between the theses which Crisp defends and Dancy attacks can be further illustrated by taking a quick look at what Winch actually argues against. In Winch’s article moral principles are not in dispute. In fact, there is no mention of moral principles. The only thing Winch takes issue with is what he calls ‘Sidgwick’s conclusion’ in the following passage:

I decide that a certain action is right for myself and act accordingly; another agent is then confronted with a situation not relevantly different and decides that for him the action I had regarded as right would be wrong; then, according to Sidgwick’s conclusion, unless I have changed my mind about the rightness of the action for me in the earlier situation, I am committed to saying that the other agent decided wrongly. (Winch 1965: 152)

Disputing this conclusion is an entirely different matter than disputing the existence of moral principles. To undermine the case for moral principles Dancy adduces cases whereby, for example, in one situation the fact that an action produces pleasure is a right-making property of the action whereas in another situation the very same fact is a wrong-making property. Winch’s case of Captain Vere and Billy Budd is of a different kind. He maintains that it would have been wrong for him to act as Captain Vere did but that, nevertheless, he is not committed to saying that Captain Vere acted wrongly. If it was right for Captain Vere to hang Billy Budd for killing his superior, then it is one thing to ask whether it would have been right for me in this situation to hang Billy Budd and another thing to ask whether it is always right to hang someone for killing his superior. The second question is about moral principles and may be answered negatively by particularists without being committed to answering the first question negatively. The first question is about (U)² and the second about (P).

Distinguishing between (U) and (P) and being aware of the ambiguity in the use of the term ‘universalizability thesis’ would advance the discussion by directing attention to the real issues between generalism and particularism.

References

Crisp, R. 2000. Particularizing particularism. In Moral Particularism, ed. B. Hooker and M. Little, 23–47. Oxford: Clarendon Press.
Dancy, J. 1993. Moral Reasons. Oxford: Blackwell.
Dancy, J. 1999. Defending particularism. Metaphilosophy 30: 25–32.
Hare, R. M. 1963. Freedom and Reason. Oxford: Clarendon Press.
Sinnott-Armstrong, W. 1999. Some varieties of particularism. Metaphilosophy 30: 1–12.
Winch, P. 1965. The univeralizability of moral judgments. In his Ethics and Action, 151–70. Oxford: Blackwell 1972.

Notes

1 He adds that ‘I took my (ii) more or less directly from Hare (1963: 11); [...].’ The reference is to the passage I quoted above.

2 In fact, the question is about the following variant of (U): If some action is right for one person then it is right for any similar person in similar circumstances.