Mark Timmons has recently argued that the form of broadly naturalistic moral realism put forward by Richard Boyd, Nicholas Sturgeon and David Brink - let's call this "Cornell Realism" - can be rebutted by an application in ethics of the twin earth thought experiments familiar in the philosophy of mind. In this note I argue that this argument fails. My main line of objection to the moral twin earth argument is that the disanalogies between any sensible realist proposal about the semantics of moral terms and the account of natural kind terms offered by Timmons are so great as substantially to weaken the force of his argument - which he concedes is not conclusive in any case.

Timmons's argument is this. Suppose Cornell realism is true. Then our use of central theoretical moral concepts is world guided by certain naturalistically characterisable natural kinds, specifically, functional kinds. The theory gives an illuminating theoretical description of those kinds. For the sake of his argument, Timmons assumes that on earth the relevant true theory is a form of consequentialism, which describes appropriately naturalised consequentialist moral properties. Timmons invites us to take the rocket to Twin Earth, where twin earthlings use orthographically identical terms whose use is regulated by natural properties distinct from those that regulate the moral discourse of earthlings. They are non-consequentialist properties whose functional essence is described by a deontic moral theory. But the whole of moral discourse can, taken as a whole, be described as having broadly the same function in the two scenarios. Timmons adds "the differences in causal regulation....are due at least in part to certain species wide differences in psychological temperament that distinguish Twin Earthlings from Earthlings".

As is usual in such cases, we have two competing descriptions of the scenarios, which Timmons summarises as follows: "Two hermeneutical options are available. On the one hand, we could say that the differences are analogous to those between Earth and Twin Earth in Putnam's original example, to wit: the moral terms used by Earthlings rigidly designate the natural properties that causally regulate their use on Earth; hence, moral and twin-moral terms differ in meaning and are not inter-translatable. On the other hand we could say that moral and twin moral terms do not differ in meaning and reference and, hence, that any apparent moral disagreements that might arise between Earthlings and Twin Earthlings would be genuine disagreements - that is, disagreements in moral belief and in normative theory, rather than disagreements in meaning."

Timmons continues, "I submit that by the far the more natural mode of description, when one considers the Moral Twin Earth scenario, is the second".

Now to sustain the initial analogy between the theoretical kinds in our best moral theory, as characterised by Cornell realism, and Putnam's suggestion about how natural kind terms function in the twin earth story, we need the opposite conclusion. We need the conclusion that the orthographically similar use of the terms "good" and "right" for example on earth and twin earth is just a bad pun. While the concepts are embedded in entire moral systems that seem broadly similar in their function (this is the analogy to the fact that on twin earth, in Putnam's original example, twin water behaves like water although it is a different kind) this superficial similarity conceals the fact that the words are in fact annexed to different kinds in the two worlds. How might such an argument be sustained?

Well, I think a counter-argument would need to have two parts: Timmons's argument seems to me both methodologically flawed and to ignore a significant disanalogy between the original Twin Earth scenario as Putnam developed it, and Timmons's deployment of the argument to embarrass Cornell Realism.

First, problems of methodology. Timmons argues that the intuitions governing twin earth cases are analogous to the intuitions elicited when people make judgements about syntax. Timmons calls them "semantic intuitions". This analogy seems to me incorrect. Syntactic intuitions are the product of an evolved biological capacity of the brain, which is modular in structure and modular in its typical operation. Semantic intuitions are the product of general reasoning capacities and are generated from our grasp of the role of concepts in the formation of whole judgements. The role of concepts in judgement underpins the ambiguity of the question, "what, in this scenario, would you say about x?" Speakers are unsure whether they are to supply a reason for their utterance, or to re-iterate a semantic norm governing their use of language; roughly, whether to mention an expression or to use it (or neither). This seems to me greatly to complicate the methodological proposal to elucidate semantic competence by prompting "one off" responses to scenarios. But as Timmons well knows, in TE cases we are not prompting run of the mill semantic intuitions. We are prompting modal intuitions, and it is not clear that we have any. Intra-world semantic norms have inbuilt vagueness and indeterminacy. This fact, can, I take it, be accepted by all sides. Why assume that norms that suffice for intra-world purposes deliver reliable inter-world judgements? If you step off a rocket in a different possible world, what would you say about phenomenon X? My response is that I have no idea, nor does it seem a sensible question to ask, nor does it seem sensible to expect that the semantic norms I have grasped give me an answer in such a case.

To begin to set up the next problem for Timmons - the substantive objection - let's introduce the recalcitrant Fregean. The recalcitrant Fregean - let's call him John Searle - responds that one can simply refuse to have Putnam's semantic intuition. Putnam invites one to hold the extension of the term "water" constant and vary the intension, whereas Searle holds the intension of the term "water" constant and varies the extension. There turn out to be more than one kind of water; water is an internally complex (disjoint) natural kind.The recalcitrant Fregean can form an alliance with the natural kind sceptic. The natural kind sceptic has a separate complaint about Putnam's thought experiment: armed with empirical detail about historical scientific practice, the natural kind sceptic charges Putnam with a massive over-simplification. Science does not "carve nature at the joints"; it legitimately gerrymanders boundaries and classifications for its own explanatory purposes. In short, the natural kind sceptic sees limited conventionalism as playing at least some legitimate role in our use of kind terms.

These responses have both been much discussed, and I do not regard them as correct response to Putnam's original TE thought experiment. But there are obviously significant similarities between these responses and Timmons's counter-example to Cornell realism. Why? Well, Timmons, like the recalcitrant Fregean, takes the similarity of how the two moral systems function - the "nominal essence" of the whole practice - to be prior to how the earth and twin earth theories carve moral nature at its joints. It is the similarity rather than the differences between the two practices that seems to him to be salient. One could bolster the argument further by adding an analogous form of natural kind scepticism: the sort of "kind" we are talking about when we morally theorise is a kind like "goodness", "well being" - surely very soft and highly conventional "kinds"? Doesn't natural kind scepticism seem intuitively plausible in these cases of "thin", highly abstract ethical concepts?

How might one rebut these objections? Well, first by noting how they fail in the original twin earth case and noting how Putnam and others have rebutted them. The recalcitrant Fregean ignores the fact that our theory of people's linguistic competence is embedded in all those theories we hold to be true and it is covering law that tells us that chemical composition is a more fundamental explanatory practice than superficial appearance. Of course we can continue to have a "common or garden" use of the term "water" to refer indiscriminately to the stuff that is drunk on both earth and twin earth, and on earth covers both water and heavy water. But is internal to the use of such a "common or garden" term that we know that our practice has twin allegiances. It is, on the one hand, marking off a quotidian distinction, while on the other hand we take it also to be shaped by the underlying nature of the stuff. That is why deference to experts is built into the practice. Typically, in the kind of case Searle focuses on, we linguistically innovate to mark this twin allegiance. We might mark off "twin water" from "water" in the way that we marked off "heavy water" from "water". But it is internal to our practice with natural kind terms that the prior allegiance is to how our terms actually succeed or fail in picking out that underlying nature that gives the most fundamental explanation of all uses of the term. We defer in our usage to experts when the issue of how the world is actually structured enters into our practice.

As for conventionality, well, we had better not deny the obvious facts. The old Pete and Dud joke about "a whale is not a fish - it is an insect" pokes fun at our classificatory practices. But once again covering law picks out the salient kind, even when such a kind contains a high degree of conventionality. Putnam allowed for this in his original example.

It is covering law that identifies the salient description of a phenomenon - that picks out, for example, microstructure or chemical composition as more fundamental than quotidian classificatory interest. It is this point that supplies an answer to Searle's question - why vary intension and not extension - and meets the natural kind sceptic half way. Perhaps there is a degree of legitimate conventionality in the classifications enjoined by explanatory law; so be it. But covering law is the basis for picking out the relevant description of the phenomenon to be explained

So Timmons's response now has to be: ok, but you are committed to the view that the "punning" use of the terms "right" and "good" on earth and twin earth is sustained by two different kinds of law. This is so in spite of the broad functional similarity across the two moral systems. And that is a very implausible claim. Same functional kind, same underlying law. The Cornell realist and I, by contrast, have to prove the opposite: same functional kind but different law.

I do think that is what the Cornell realist (and his or her "fellow travellers" such as myself) has to prove, but I do not think it is implausible and Timmons himself has explained why. A central moral of the Putnam case (pardon the pun) was that similarity of classificatory practice - functional use of the term - was not conclusive if the semantic category of the term is that it is a natural (or social/conventional) kind and it is internal to that very same practice that it is guided by that intrinsic nature that underpins the function. Furthermore, different laws hold on moral earth and moral twin earth because different boundary conditions obtain, as Timmons himself explained. To ensure the similarity across the two cases that made his argument plausible, he varied the psychology of the two species. It is an acknowledged feature of law in the social sciences that it only holds relative to various boundary conditions. This is, of course, a feature of law generally, but the narrower case is the one directly applicable here. Timmons explicitly varies a boundary condition across his two cases: he changes the psychology of human beings and the beings on twin earth. So there cannot be any direct mapping of the laws across the two scenarios on to one another. So the use of the central moral concepts remains a "punning" use.

I am happy to concede that this rebuttal of Timmons rests on a controversial premiss. However,the premiss does seem to me to be very plausible. The idea is that generalisations in psychology, sociology and history depend essentially on both ceteris paribus clauses and on boundary conditions. Take this example from Phillip Pettit: the generalisation that unemployment causes alienation depends on certain presupposed boundary conditions; for example, that the unemployed do not live near to the beach and that the weather is constantly fine, such that unemployment leads to tans, some level of contentment and to a lack of alienation, not to mention a high level of ability to surf.

Let's go back to the original example. A key premiss for Timmons is that taking the practices of earthlings and twin earthlings as a whole, at the level of the whole system, their moral concepts are embedded in a system that plays broadly the same role. That is what inclines us to translate the words in the same way. But now to make this premiss plausible, Timmons makes a compensatory adjustment in the psychologies of earthlings and twin earthlings. But this adjustment begs the question. The interpreter's response should now be: "well, they seem to be using the same words as us, 'right', 'good' and so on. They seem to be theorising as we do. But they have a fundamentally different psychology from ours, and the variation in the background theories affects the boundary conditions of the 'laws' they are trying to formulate in their moral theories. So it is now unclear that their theoretical concepts are translatable into ours, even if the whole system in which they are embedded is broadly similar to ours". Different laws, different embedded concepts.

A final point. In the specific example Timmons gives, it seems to me more likely that a deontological theory would map constraints directly on to the space of possible reasons - it is not about outcomes at all - so the kind of contrast envisioned in this particular case does not arise. This offers an alternative diagnosis: that the two groups on earth and twin earth both have incomplete parts of the one true theory. No problem for Cornell realism there: Putnam's theory of natural kind terms has to begin with theories which are true. If it is true that water is H2O and twin water is XYZ, then we have an example to worry about. But a more plausible description of Timmons's actual case in his apparent counter-example is that it is a case where that condition is not met.

I conclude that Cornell realism escapes this particular attempted refutation.