This is a short introduction to a theoretical account of the semantics of English moral sentences: those sentences that include moral predicates like good, bad, right and wrong. I will use wrong throughout as my example, but it ought to be easy to see how the account applies to other moral predicates. I call the theory “value prescriptivism” because it is very much like the universal prescriptivism of R. M. Hare [i], except that, speaking loosely, it is values, rather than behaviors, that are prescribed. While standard prescriptivism likens “Lying is wrong” to “Don’t lie!” value prescriptivism likens it instead to “Have a value system that lying violates!” In this introduction, I will explain the theory and show how it meets certain conditions of adequacy for a theory of moral language. These are related to the Frege-Geach problem, and to moral disagreement.
A more precise explanation of value prescriptivism requires development of the idea of the characteristic schemata of moral predicates and moral sentences, and of the idea of a value system satisfying a characteristic schema. The motivation behind these ideas is that, on this account, a moral sentence prescribes to everyone (in a relevant class, such as all moral agents) that they have a value system that has a certain property or characteristic: that it meets a condition determined by the semantic content of the sentence, and represented by the characteristic schema of that sentence. (I use the term “value system” broadly to apply to systematic psychological dispositions to have certain attitudes of approval or disapproval about something given particular beliefs about it. So, for example, if honesty is part of one’s value system, one is disposed to approve of something given that one believes it is an instance of honesty, and to disapprove of something given that one believes it is an instance of dishonesty.)
To explain the characteristic schema of the moral predicate wrong we look at a sentence where it is in a “stand-alone” context, serving as the predicate of the whole sentence. If we stick with the example “Lying is wrong,” the characteristic schema would be an open sentence, “Lying violates V.” The variable V ranges over all value systems. The characteristic schema of a moral sentence S is the sentence S¢ open in V that results from uniformly replacing each instance of any moral predicate in S with the characteristic schema of that predicate. A particular value system Vi satisfies a characteristic schema S¢ just in case uniformly replacing V in S¢ with Vi yields a true sentence. (This will be a descriptive sentence about Vi, not a normative sentence.)
So the more precise explanation of the theory is as follows. A moral sentence prescribes (to everyone) acceptance of any value system that satisfies the characteristic schema of the sentence. In other words, a moral sentence S prescribes acceptance of any Vi that satisfies S¢.
The Frege-Geach problem may be thought of as the difficulty expressivism and other noncognitivist theories have meeting three natural adequacy conditions for a theory of moral language. These are (I) that the theory should account for moral predicates in embedded as well as stand-alone contexts, (II) that these accounts should be the same, and (III) that it should account for the validity of arguments comprising moral sentences, such as the following example drawn from Geach’s paper inspired by Frege [ii]. (If (II) is not met, such arguments would suffer the fallacy of equivocation, as “Lying is wrong” in (2) would not mean the same thing as (1).)
(1) Lying is wrong.
(2) If lying is wrong, getting little brother to lie is wrong.
Therefore, (3) Getting little brother to lie is wrong.
According to value prescriptivism, premise (1) prescribes acceptance of any value system that satisfies (1)¢, premise (2) prescribes acceptance of any value system that satisfies (2)¢, and the conclusion (3) prescribes acceptance of any value system that satisfies (3)¢.
(1)¢ Lying violates V.
(2)¢ If lying violates V, getting little brother to lie violates V.
(3)¢ Getting little brother to lie violates V.
Clearly adequacy conditions (I) and (II) are met by value prescriptivism, as it accounts for stand-alone context examples like (1) and embedded context examples like “Lying is wrong” in (2), and in exactly the same way. Like Gibbard’s [iii] view, the theory applies its account to whole sentences, obviating any problems related to embedded contexts or semantic compositionality.
And the prospects seem good for value prescriptivism giving an account of validity for moral arguments, showing that it meets condition (III) as well. This is because, as Hare himself emphasized, imperatives and other prescriptions, while they may not have truth values, still make use of the descriptive conditions prescribed, which are truth-apt. This, he thought, provided the resources needed to explain logical properties, including validity, in moral arguments.
So one strategy we might pursue is to develop a general theory of validity for arguments comprising imperatives, and then use this as the basis of a theory of validity for moral arguments, given the identification value prescriptivism makes between moral sentences and prescriptions, and the close connections between imperatives and prescriptions in general.
However, there are serious difficulties in developing a theory of validity for imperatives, in part because of problems such as the one called “Ross’s paradox.” It may seem that, for imperative arguments, all we have to do is to adopt the criterion that such an argument is valid just in case it is impossible for the conditions of the premises to be complied with while the condition of the conclusion is not complied with. Validity for imperative arguments, in other words, simply mirrors validity for descriptive arguments using the conditions prescribed. But, as Ross pointed out, if I say “Mail this letter!” this criterion would suggest that I should also be willing to say “Mail this letter or burn it!” as this mirrors the validity of descriptive arguments of the form “P, therefore: P or Q.” I would not be willing to say that, however, because then you could comply with the second imperative by burning the letter that I wanted you to mail for me.
Luckily, we do not have to develop a general theory of imperative validity to give an account of validity for moral arguments. All we have to do is account for a proper class of arguments comprising prescriptions of the sort called for in our theory, all of which are of the form “Have a value system V satisfying S¢.” In the case of our example, all we need establish is that any value system that satisfies (1)¢ and (2)¢ must satisfy (3)¢ as well. In general, for any moral argument f1,f2,…fnㅏj, we say that the argument is valid just in case it is impossible for any value system to satisfy the characteristic schemata f1¢,f2¢,…fn¢ but not satisfy the characteristic schema j¢.
Besides the three conditions of adequacy related to the Frege-Geach problem, there is one more condition to look at here, and it is related to indexicalism and moral disagreement. As an example of disagreement, consider my own humanistic value system Vh in contrast to Sarah Palin’s bibliolatrous value system Vb. On an indexical account, (4) would mean (5) when Sarah says it while, if I were – for some reason – to say (4), I would not mean (5), but (6) instead.
(4) Gay marriage is wrong.
(5) Gay marriage violates Vb.
(6) Gay marriage violates Vh.
Note that (5) and (6) are examples of the kind of descriptive statements referred to in the definition of satisfaction for characteristic schemata. While (6) is false, it should also be noted that (5) is a true statement. That is to say, Vb satisfies (4)¢ whereas Vh does not satisfy (4)¢.
The obvious problem with this is that (5) and (6) are about different things, Vb and Vh respectively, and so when I say (4) I do not mean the same thing Sarah means when she says (4). This, in turn, means that she and I do not really disagree about (4) in a way in which it certainly seems we do disagree. Problems of this sort have been extensively discussed in the literature [iv]. They suggest the following additional condition of adequacy for a theory of moral language.
(IV) It should account for shared semantic content in moral disagreement.
For Sarah and me to genuinely disagree, in a way that it seems we do, we have to share the same semantic content in saying (4), so that we are disagreeing about the same thing. But indexicalism gives us no “neutral” account of the meaning of (4), and provides only (5) and (6), so that instead of disagreeing about the same thing, we are saying different things. It appears that the necessary semantic content shared by Sarah and me is simply missing in an indexical theory.
Value prescriptivism, however, does seem to easily satisfy (IV). For both Sarah and me, saying (4) means exactly the same thing: prescribing any value system that gay marriage violates. Since gay marriage does not violate my value system, I am not inclined to make such a prescription. But since it does violate her value system, she is inclined to make the prescription.
Obviously, (I) – (IV) do not exhaust the conditions of adequacy for a theory of moral language. A full evaluation of value prescriptivism would require consideration of other conditions, and philosophers will disagree about what conditions should apply for such theories. However, this introduction to value prescriptivism shows that it does meet adequacy conditions (I) – (IV), something other theories, including expressivism and indexicalism, seem unable to do.
[i] Hare, R. M. (1952) The Language of Morals (Oxford)
[ii] Geach, P. T. (1965) “Assertion” (Philosophical Review 74 (4) pp. 449-65)
[iii] Gibbard, A. (2003) Thinking How to Live (Harvard)
[iv] e.g., Dreier, James (2009) “Relativism (and Expressivism) and the Problem of Disagreement” (Philosophical Perspectives 23, ed. by John Hawhtorne)