The focus of this chapter will be to examine the implications of the Kripke/
Putnam theory of essences for logical possibility and thought experiments. To
return to the usual view of logical possibility, let's look at the argument A.
J. Ayer (The Central Questions of
Philosophy) makes regarding the application of laws:
A state of affairs the description of which is not self- contradictory may still be said to be impossible on causal grounds. It is not logically but causally impossible that cats should breed with mice or that steam-engines should run without fuel. But here the difference consists not in the use of different concepts of possibility but rather in an implicit reference to different sorts of laws. To say that a state of affairs is impossible is, I believe, in either case to say that there is some law which excludes it. A necessary state of affairs will then be one that is required by law, in the sense that its negation is excluded, and a possible state of affairs one that is neither required nor excluded. The difference, then, between logical and causal possibility is just that what is logically impossible is excluded by the laws of logic, and what is causally impossible is excluded by the laws of nature (p. 149).
On Ayer's account logical impossibility follows from the laws of logic what they exclude is impossible, but what they do not exclude is possible. (One wonders, though, what the laws of logic by themselves might exclude, since we also need inference rules, but we will not pursue this here.) It seems relevant to add to the laws some content, since presumably we want to rule out as logically impossible A's being taller than B and vice versa, as was discussed in Chapter One. But, in essence, Ayer's is the usual account of logical possibility outlined in Chapter One. The present chapter will show that there is a logic of essences which constrains what kinds of stories we can construct. The usual claims for logical possibility were based on a very narrow rendition of a situation: if nothing in the story contradicted any of the definitions involved, then the story was said to be logically possible. If one told a story about an elephant giving birth to kittens, we would have to check to see if any of the definitional properties of `elephant' excluded them from having kittens. And of course there is nothing in the definition of `elephant' that excludes `giving birth to kittens'. Therefore, it's a possibility since nothing in the definition of `elephant' logically excludes giving birth to any other kind of critter. But once essences are added to this story, the realm of what is possible is more limited. A good thought experiment is one in which the essential features are taken into account and nothing is incompatible with them or violates them. This means the examples adduced for testing a claim or theory should not be too farfetched viz., the method should not generate wild counterexamples to show the contingency or falsity of the claim or theory tested. The point of this chapter will be to argue that fruitful application of thought experiments requires that essences be taken into account. We will briefly examine the claim made by Strawson then we will look at a more complex thought -experiment meant to show that moral realism is untenable.
P. F. Strawson's argument was meant to show that experience is only
contingently dependent on one body. I noted three kinds of empirical facts that
a perceptual experience depends upon: (1). One's eyelids need to be open; (2).
What one sees is dependent on the orientation of his eyes, direction of head
etc. And (3) what a person sees depends on the person's field of vision. But
Strawson has us consider the following thought -experiment. What of a subject,
S, whose experience depends on three different bodies:
A (situated somewhere, but anywhere) must have eyelids open.
B (somewhere, but anywhere) determines where his eyeballs and head are turned.
C (in drawing room) determines where he sees from.
Strawson argues that S's perceptual experience can be said to rest not on one body but on three. Perhaps such a case would be one of linking the subject's brain to three different kinds of `bodies' (say a video camera, etc.) in three different places. But this sci-fi type of example would still make the subject S's brain, to which all the instruments are linked, the unifying body of the experience. This would merely be a case of physical possibility or impossibility. What Strawson is attempting to say is that it is logically possible that an experience in S is causally dependent on three distinct bodies that are not causally/ physically linked together-- yet they are all (formal?) causes of the single experience.
Now we may wonder whether this even makes sense. Certainly it would seem to be
physically impossible that A's closed eyelids would cover B's or C's eyes. A's
closing its eyes would only (physically) cause it to lose its own field of
vision. But is it even logically possible that A has its own eyelids and B has
its own eyeballs and that A's
closing its eyelids cause S to lose the sight it has through B's eyeballs? It
might seem that A's eyelids cannot affect C's vision since A's eyelids are A's
and C's are C's. Since they are numerically different there is no way that A's
eyelids can cover C's eyes unless A = C. But this makes no sense. Fortunately
for Strawson all that his example needs to assert is that A's eyelids can
affect S's sight. Again, this is a question of logical possibility: whether it
seems contradictory to imagine a situation where S has a percept that has
causes in three distinct bodies. So, either A's eyelids somehow affect the
light entering B's eyeballs, or sight need not depend on the entry of light
into eyeballs.
An essential definition of perception must include the notions of an awareness
via patterns of energy absorption by groups of receptors, but, also, an
awareness of discriminated existents (Efron
p. 147). Robert Efron states that in order for an existent to be disciminated,
it must be a segregated, isolated, cohering `thing' (p. 148). Both of these
qualities give us an essential notion of what perception is. Unfortunately for
Strawson, he has not shown that the experience which S receives is actually
coherent. It seems that Strawson has ignored the nature of human perception.
Simply because we can imagine that one body's eyelids-- which are not near
another body's eyes can affect a single person's sight is not sufficient to
claim that perception can still be perception and yet happen in a totally
different way. A perceptual experience
for a person must happen a certain way; perhaps it depends, as Strawson says,
on three kinds of empirical facts. But simply because these facts are empirical
does not mean that they are separable essentially. It seems that there is no
possible world where there are persons such as us, with our natures, and where
experiences happen as he describes it. This is because the experiences that
Strawson describes seem to rest on telepathy--- some way of having three
distinct bodies give rise to an experience in S without any physical
interaction. This is to imagine state of affairs that is logically possible,
but it is also to change the natures of the things involved.
We will now examine a contemporary and more controversial thought experiment
designed to show that Moore's open question argument applies to `synthetic'
moral realism the claim that moral facts can be explained as natural facts
that supervene upon physical facts. While not defending any particular version of
moral realism, I will attempt to show that taking the distinction between
essential necessity and logical necessity into account allows this sort of
moral realist to mount a defense against this critique.
Terence Horgan and Mark Timmons have provided a thought experiment designed to
show that moral realism which embraces naturalism faces an open question
argument. Two moral realists who hold this view are David Brink and Richard
Boyd. The sort of moral realism they advance holds that moral facts can be explained
as part of the natural world. Moore,
too, was a moral realist, since for him moral properties supervene upon
physical ones, but, for him, they are not part of the natural world. Moral
properties are non-natural
properties. Therefore, they do not require an explanation in natural terms.
They can just be said to be sui generis.
The synthetic moral realists agree that moral facts supervene upon physical
facts--are not reducible to them but they also believe that these moral facts
are part of the natural world and can be given a natural explanation. The new
moral realism of Brink et al. relies upon the plausibility of giving a
naturalistic explanation of certain supervenient qualities, such as mental
properties on physical brain states. Horgan and Timmons suggest two constraints
a naturalistic explanation must hold to:
(C1) If there are any moral properties or facts, they must be naturalistically accommodated.
(C2) If moral knowledge is possible our access to and knowledge of moral properties and facts must be explainable according to epistemological principles we use to explain our knowledge of the natural world generally (1990-91 p. 449).
In order for moral properties to be natural they must meet these requirements; otherwise we will be left with metaphysically `queer' moral entities. Of course we are not assuming that any reduction need take place; that is, the realist need not be committed to a view on which moral properties would turn out to just be physical properties. Rather, all that the moral realist need do is show how one set of natural properties causally regulates another.
Horgan and Timmons argue that moral realists claim there is nothing `queer'
about the moral supervening upon the physical because there does not seem to be
any problem with, for instance, the mental supervening upon the physical. Brink
asks why, if we can accept mental supervenience as natural, can we not accept
moral supervenience? (in Horgan and Timmons 1992 p. 225; quoted below p. 67). This is a plea of `innocence by
association', where some less controversial or established point
(naturalistically acceptable mental supervenience) is used to make more
plausible sounding a controversial point (naturalistically acceptable moral
supervenience). The point is meant to shift the burden back onto the shoulders
of the moral anti-realist. But Horgan and Timmons believe that this position is
highly problematic when spelled out in full.
Firstly, to spell out the critique of the moral realist position we will need
to look at what mental supervenience requires. Horgan and Timmons distinguish
between pure and hybrid semantic constraints. The former
are semantic constraints "governing the proper use of higher-level terms and
concepts along with concepts embodying facts about meaning that must be tacitly
mastered by anyone who employs the terms and concepts properly" (1992 p. 235).
An example of this is as follows:
(P1) For any physical-stuff natural kind term t, and any physico-chemical property P, if S refers at our actual world to physical stuff with a distinctive physical essence, and this stuff's physical essence is its possession of P, then t refers rigidly (i.e. at every possible world) to stuff that possess P (1992 p. 235).
This is the same view of the rigid designation of natural kind terms that was defended in Chapter Two. The latter constraints are the product of pure semantic constraints and certain additional actual- world facts— "empirical facts that need not be known or tacitly mastered in order to employ the terms and concepts properly" (ibid). So, the first is a formal requirement of the correct use of words; the second constraint combines with relevant empirical facts to govern natural kind terms. For instance, `water' has the following hybrid semantic constraint:
(H1) For any possible world w, a quantity of liquid (in w) belongs to the w-extension of `water' iff it is composed of H2O molecules (in w) (ibid).
They argue, then, that one kind of naturalistic explanation of objective moral facts would "consist in showing that they are a consequence of relevant lower-level facts in conjunction with semantic constraints governing the relevant terms and concepts" ( 1992 p. 235). They call these kinds of explanations semantic constraint satisfaction explanations (or SCS).
They note that functionalism is popular in the philosophy of mind. There are
two functionalist theories one could adopt; the first is analytic
functionalism. I will quote their succinct explanation:
Analytic functionalism ties a mental state's functional role quite directly to the very meaning of the propositional-attitude terms and the concepts expressed by those terms, and construes this functional role as captured by the fundamental generalizations of common-sense psychology (so-called folk psychology). The leading idea is that these generalizations operate collectively as pure semantic constraints. As a matter solely of the meaning of folk psychological terms and concepts, a correct assignment of extensions to these terms and concepts, at a given possible world, must be one under which the generalizations of folk psychology come out by and large true (1992 p. 237).
Analytic functionalism does not lend itself to naturalistic explanations of supervenience relations, since such relations are based on meanings of terms, not scientific facts.
Psychofunctionalism asserts that propositional attitude terms rigidly refer to
certain functional properties whose essence is not fully captured by folk
psychology. So on this view a more detailed psychological theory is required;
but even though we may not know what this theory is, we can say that the
content of our assertions and beliefs are causally regulated by such
properties. Every folk psychological term t refers rigidly to the specific
causal property that causally regulates people's use of t; this is the same
view we outlined in the last chapter, where the truth or falsity of claims
about kinds (water) or persons (Aristotle) depends on facts about H2O
or a particular person independently of whether the essence was known.
Psychofunctionalism asserts the following pure semantic constraint:
(P2) For any psychological theory T, if (i) there is some unique family of interconnected functional properties that causally regulate (in the actual world) the attributions by humans of propositional attitudes to one another and to themselves, and (ii) the generalizations of T collectively characterize the functional essence of these properties, then each propositional -attitude term rigidly refers to the T-characterizable functional property that regulates it (1992 p. 237).
As they note, psychofunctionalism claims the following: (1) the clause (i) of (P2) is true; (2) that the complete psychological theory that is true of humans performs the role characterized by (ii). Also, (3) this theory itself will posit propositional attitude properties. So, if one psychological theory (T*) turns out to be true, then this empirical fact, together with empirical claims (1)-(3), will combine with (P2) to yield the following putative hybrid semantic constraint:
(H2) For any possible world w, the correct assignment of w- extensions to propositional attitude terms is an assignment which renders true (at w) all the generalizations of T* (1992 p. 238).
Horgan and Timmons argue that these sorts of mental functionalisms lend
themselves to the SCS mode of explanation. If asked why a mental state M
supervenes upon a given physicochemical state P, the explanation will be as
follows:
P is a member of a system of physicochemical properties which together more or less realize the pattern of causal relations characterized by the psychological generalizations which semantically constrain mental terms and concepts. Furthermore, P itself occupies the role in that system which, according to those psychological generalizations, constitutes the M-role. Hence, whenever someone instantiates P, he must instantiate M as well (1992 p. 238).
This explanation does not trade on the meanings of terms only; we need to know empirical facts about the relevant physicochemical properties and causal generalizations regarding these properties. So, we can see how semantic constraints along with lower-level facts `conspire' to explain how M supervenes upon P. This is a naturalistic explanation based of empirical evidence that is not open to Moore's open question argument (see below).
Horgan and Timmons now go on to spell out how synthetic moral realism, if it is
to work, must mirror psychofunctionalism in the philosophy of mind. They note
that Brink is not as specific as needed in his claim that moral supervenience
can be explained in the same way as mental supervenience. What does seem
apparent from the long quote of Brink's (in Horgan and Timmons 1992 pp. 240
-241; quoted below p. 68) is that he accepts a synthetic view, not an analytic
one. He claims that moral properties are functional properties. Moreover, we
will need a moral theory that best explains all of our moral beliefs (in Horgan
and Timmons 1992 p. 241). This theory must provide generalizations that
constrain our language— hence providing a basis for explaining objective
supervenience relations. This theory will not be an analytic one, though, since
analytic theories depend only on the meaning of moral terms and concepts. They
call the theory that he implicitly advocates synthetic
moral functionalism. On this view, moral terms pick out certain
functional properties "whose essence is revealed by the generalizations of a
synthetic moral theory "(1992 p. 242). There are facts purportedly built into the
meaning of moral terms and concepts: (i) that they express certain abstract
(multiply realizable) properties; (ii) that the essential nature of these
properties involves their interconnected functional roles; and (iii) that it is
an empirical question what these functional roles are (ibid). Also, we need a
moral theory T "whose generalizations, as a matter of empirical fact, collectively capture the
essential nature of moral properties" (ibid). So, for the synthetic moral
realist, even though persons often disagree and are mistaken about moral
beliefs, we can still see that moral beliefs are causally regulated by some
unique set of functional properties captured by T.
Synthetic moral functionalism can be said to assert the following putative pure
general semantic constraint on moral terms:
(P3) For any normative moral theory T, if (i) there is some unique family of interconnected functional properties that causally regulates the actual- world moral judgements and moral statements of humans, and (ii) the generalizations of T collectively characterize the functional essence of these properties, then each moral term refers rigidly to the T- characterizable functional property that regulates it (1992 p. 243).
There are also the following two empirical claims:
(E.i) There is indeed a unique family of functional properties that causally regulates the moral judgements and moral statements of human beings in general, despite the fact that humans widely disagree among themselves about matters of morality (ibid).
(E.ii) There is a single normative moral theory whose generalizations would correctly capture the functional essence of properties that causally regulates human moral judgements and statements (ibid).
The empirical claims in combination with (P3) give us the following putative general hybrid semantic constraint:
(H3) For any possible world w, a correct assignment of w- extensions to moral terms is an assignment which renders true (at w) all the generalizations of T* (a synthetic normative moral theory)(ibid).
So, if (P3), (Ei) and (Eii) are true, then SCS explanations of putative moral supervenience facts is in principle possible. Hence, moral synthetic functionalism can seem to give the same sort of explanation of supervenience relations as psychofunctionalism.
The authors propose to use a thought experiment to test the claim of a pure
semantic constraint on moral terms; this is a valid test, they argue, since
such constraints are about linguistic competence that any language users should
be able to possess in order to use the terms correctly. They propose to use a
Putnam Twin Earth experiment since this functions as a test for (P1)-- the
putative pure general semantic constraint on physical-stuff natural kind terms
(1992 p. 244). This experiment showed, as we saw, that (1) the term `water' in
Twin English does not mean what it does in English; and (2) that the Twin Earth
word `water' is not translatable by the English word `water'; finally, (3) the
beliefs that Twin Earthlings express with their word `water' are not about
`water' (H2O) at all, but a
substance distinct from it (XYZ).
Horgan and Timmons propose that their version of Twin Earth is pretty much
identical to this Earth: all of the relevant features are the same; the
geography is similar; many Twin Earthlings even visit the grave site of Twin
Elvis once a year, etc (1992 p. 245). But there are some differences. While
everyone's moral beliefs on Earth are causally regulated by a certain unique
family of functional properties, Twin Earthling's moral beliefs are causally
regulated by a different set of functional properties. (Here we are assuming
(Ei) and (Eii) to be true. And these functional properties can be discovered
through empirical investigation.) So, we can see that the two moral vocabularies
act almost the same way: Twin Earthers use the terms `good' and `bad' to
evaluate persons and actions just as we do. The main difference is that the
moral properties the Twin Moral terms `track' are nonconsequentialist moral
properties, "whose functional essence is captured by some specific
deontological theory" (1992 p. 245). This theory they call Td, in
contradistinction to the consequentialist moral theory of Earth Tc.
Now we must imagine some explorers from Earth who visit Twin Earth. They find
that the inhabitants' terms `good' and `right' have all the same formal marks
that characterize those terms in their own
moral vocabulary. Twin moral terms are used to reason about Twin Earthly well being; Twin Earthlings
believe that it would be incoherent to use `good' and `right' in ways that
would violate supervenience. But the question is, even though the explorers
would desire to translate the Twin Earthers' moral terms into their own, would
they be correct in thinking that the twin moral terms have the same meaning as their own moral terms?
Conflicts arise between the explorers and the Twin Earthlings in the same cases
where deontology and consequentialism clash on Earth. Both theories offer
conflicting prescriptions in cases concerning individual rights, respect for
persons, etc. A more concrete case is as follows: In a case where two persons
will both die, can we intentionally cause the death of one of the persons in
order to save the other's life? A woman cannot deliver her child since the
child's head is too large to pass through the birth canal. As Horgan and
Timmons argue, the physician's options are to (a) let the mother die, whereupon
the child will also die; or (b) to save the mother by crushing the child's
skull (1992 p. 246). This sort of case ought to cause disagreement among the
explorers and the Twin Earthlings. The former, as consequentialists, will
choose the option where at least one person lives. The rules about not killing
are waived in this case since it is better consequence that one person line
than neither. The latter, as deontologists, will choose the option whereby no
rules are broken and no one is intentionally killed.
If synthetic moral functionalism is correct, then the Twin Earthlings' moral
terms must rigidly designate different properties than the properties that
Earthling moral terms rigidly designate. This follows from the pure general
semantic constraint (P3). Twin Earth moral terms must differ in meaning from
those of Earth. In Putnam's example our `intuitions' drive us to conclude that
the Twin Earth term `water' has a different meaning than does the term on
Earth. Since (i) twin Earthling moral judgements appeal to a different moral
theory than Earthling moral judgements, and (ii) since the twin moral judgements
are causally sensitive to different functional properties than the moral
judgements of Earthlings, moral realists (the explorers) should conclude that
the Twin Earthlings just do not mean by `good' and `right' what Earthlings mean
by them (1992 p. 247).
But, Horgan and Timmons argue, contrary to synthetic moral realism there is no
`hermeneutic pressure' to draw these conclusions at all (ibid). While there
does not seem to be an open question regarding who is right about the meaning
of `water', there does seem to be one regarding the use of moral terms.
Earthlings and Twin Earthlings differ in their respective moral beliefs, which
is fundamentally a disagreement about moral theories. The moral theories, Td
and Tc, hold that there are different right-making properties they are causally
sensitive to; but Horgan and Timmons claim that to put it this way is to say
that the differences are differences in moral
belief, not differences in the meaning of `good' and `bad' (1992 p. 248). If
(P3) were correct,
then recognition of these differences ought to result in its seeming rather silly, to members of each group, to engage in inter-group debate about goodness— about whether it conforms to normative theory Tc or Td. ( If, in Putnam's original scenario, the two groups learn that their respective uses of `water' are causally regulated by different physical properties, it would be silly for them to think they have differing views about the real nature of water.) But such inter-group debate would surely strike both groups not as silly but as quite inappropriate, because they would regard one another as differing in moral beliefs and moral theory, not in meaning (1990 -91 pp. 460-461).
The moral that Horgan and Timmons draw from this thought experiment is that
synthetic moral functionalism cannot account for moral terms— that is, for
ascribing truth values to statements involving them. Such moral facts end up
being open to question. They point out that Moore's open question argument
focused on analytical definitions of moral terms in terms of natural
properties. A definition for Moore would (ideally) be analytic found to be
true simply by an analysis of the meanings involved. But if a moral term M were
analytically defined as N, then questions of the following sort ought to be
easy to resolve (`Entity x is N, but is x M?') simply by looking at the
meaning of the terms involved. But, of
course, it seems that such questions remain open. The new moral realism suffers
the same fate, even though it eschews analytic definitions. If synthetic moral
realism is correct then the follows two questions ought to be closed:
Q1 Given that the use of `good' by humans is causally regulated by natural property N, is entity e, which has N, good?
Q2 Given that the use of `good' by humans is causally regulated by natural property N, does entity e, which is good, have N? (1990-91 p. 461)
But if the Twin Earth thought experiment is correct, then our intuitions tell us that the questions are open, not closed. If the right-making properties are in dispute, then neither question— whether things with N are good, and whether things that are good must have N— is closed.
Now I will argue that the synthetic moral realist can cut off Horgan and
Timmons' argument before reaching the open question argument. If there is a
functional essence of morality for human beings, then this must be the same for
human-like persons on a Twin Earth. The thought experiment that they adduce to
show how moral realism fails, fails itself. Their story can be shown to be
incoherent since, if functional moral realism is true, two worlds with the same
essential properties must have the same functional properties regulating a
moral theory; their claim that there could be two worlds alike in all natural
properties, but with differing accounts of functional moral properties is
essentially incoherent.
The view that moral realism was supposed to mirror in order to have a chance at
providing a SCS explanation was psychofunctionalism. The general point to be
shown regarding any kind of naturalistic supervenience claim was that,
the physical facts at any possible world w, along with semantic constraints governing the correct use of the terms in one's language and the concepts those terms express, together conspire to fix uniquely all the facts about w statable in the language (1992 p. 234)
And the view of moral realism that Brink and others espouse follows psychofunctionalism. That view is that propositional attitude terms rigidly refer to certain functional properties whose essence is fully captured, not by folk psychology, but by a detailed psychological theory. This is akin to Putnam's claim that even though we do not always know what the meanings of natural kind terms are, our terms are causally regulated by them. This allows a SCS explanation of supervenience claims such as why a mental state M supervenes upon a physicochemical state P. Semantic constraints and lower level facts `conspire' to explain why M supervenes upon P.
The question is, then, are there semantic constraints that together with lower
level facts conspire to fix objective supervenience relations in any possible
world? According to Brink there is such a synthetic theory that captures the
functional essence of morality. There are factual components built into moral
terms. But also, to quote Horgan and Timmons again, there is:
the normative moral theory T whose generalizations, as a matter of empirical fact, collectively capture the essential nature of moral properties. The idea is that even though people often disagree morally, and even though mistaken moral beliefs commonly occur, nevertheless, human moral terms and moral beliefs are causally regulated by some unique set of functional properties whose essence is captured by theory T (1992 p. 242).
As we saw above, synthetic moral realism gives rise to a pure general semantic constraint (P3). Together with the two empirical claims (Ei) and (Eii), it gives us the pure hybrid semantic constraint (H3). But (H3) only follows if (P3) and (Ei) and (Eii) are true. And Horgan and Timmon's main contention is that (P3) does not work because of their Twin Earth thought experiment. If their thought experiment works, then the moral realist is not entitled to (P3), viz., that moral terms rigidly designate certain properties. As we have seen, the ability to imagine a Twin Earth where different functional properties regulate moral properties is enough to defeat the moral realist aspirations to rigid designation.
But one wonders if the story they tell of consequentialist and deontological
theories butting heads is as plausible as it first seems. There are problems
with pure forms of both theories. The oft rehearsed challenge to the
deontologist is that he must hold the counterintuitive view that we should not
break a rule now even if doing so would cause us to break fewer rules later.
Often it seems right to break a rule now if better consequences follow. Also
familiar is the charge against consequentialism that there is not always a way of weighing consequences to make
them interpersonally measurable (a la utilitarianism). So, if there are
problems with either theory, then there is a good reason to think that neither
theory can work in either world. This helps to point the way to a solution. We
are assuming, along with Horgan and Timmons, that our investigations are
converging on the truth. But it seems unlikely that the truth in any world
remotely like ours will be one either like Td or Tc.
Let us consider David Brink's claims about moral realism. Brink explains his moral functionalism as follows:
[T]he moral realist might claim that moral properties are functional properties. He might claim that what is essential to moral properties is the causal role which they play in the characteristic activities of human organisms. In particular, the realist might claim that moral properties are those which bear upon the maintenance and flourishing of human organisms. Maintenance and flourishing presumably consist in necessary conditions for survival, other needs associated with basic well-being, wants of various sorts, and distinctively human capacities. People, actions, policies, states of affairs, etc. will bear good-making properties just insofar as they contribute to the satisfaction of these needs, wants, and capacities... [and] will bear bad-making properties just insofar as they fail to promote or interfere with the satisfaction of these needs, wants, and capacities. The physical states which contribute to or interfere with these needs, wants, and capacities are the physical states upon which, on this functionalist theory, moral properties ultimately supervene (quoted in Horgan and Timmons 1992 pp. 240-41).
If there is a functional essence of morality for creatures such as us, then our
empirical investigations ought to be converging on this essence. This is an
assumption that Horgan and Timmons allow for psychofunctionalism, in their
illustration of how one could have naturalistic supervenience. If their
argument is to have force they must allow it for syenthtic moral realism too.
If there is a property (or properties) that regulate moral discourse in this
world, and these properties are empirically discoverable, then it is
essentially impossible that there be another world (Twin Earth) just like this
one where moral terms are regulated by different base/ subvenient properties.
The discovery of functional essences would allow us to say that Twin Earthers
are wrong about their moral claims if their moral theory did not conform to the
actual functional properties. But more importantly it allows us to say that
Horgan and Timmons are mistaken about their claim that no pure semantic
constraints are possible. What they have described is a logically possible
state of affairs and not an essentially possible one.
Brink could defend his view of moral realism by taking essential necessity and
possibility into account. If he is granted that our investigations are
converging on the truth of moral theory T (in this world) as explaining
supervenenient properties, and their respective regulative base properties,
then any other world essentially like this one has the same properties and
relations. It would be an essential impossibility that Twin Earth could have a
different group of base properties regulating moral terms and still be relevantly
the same as Earth (where moral theory T is true). An essential impossibility
follows from taking into account the empirically discoverable essences
(functional properties) with explanations and definitions of such properties
which, together with a set of claims about such properties (such as the Twin
Earth experiment), land us in a contradiction. The contradiction, namely,
between two essentially similar worlds with different moral theories (Td and
Tc). So, Brink could claim that our intuitions are driven to conclude that it
is just not possible that there could be two essentially similar worlds which could support two contrary theories;
such an experiment, he could argue, cannot get off the ground, hence it does not
lend itself to shoring up an open question argument.
It could be logically possible, of course, that there is a world just like this
one which has different functional properties. This seems to be Horgan and
Timmon's claim; we can have another world like this one which has different functional
properties regulating moral terms. But this claim would seem to work against
psychofunctionalism as well. Could we not use Horgan and Timmons' Twin Earth
experiment to show that another world could have the same supervenient
properties but different subvenient regulating properties? This claim could be
proposed to cast into doubt any putative pure semantic constraints regarding
psychofunctionalism. But it does not cast them in doubt if we take the
essentialist analysis proposed earlier; we can say that any world the same as
this one in essential respects must have the same discoverable functional
essences.
So, we have seen how Horgan and Timmons attempt to rule out synthetic moral
realism with the wrong type of thought experiment; rather than describing a
world that is merely logically possible, they ought to have looked at essential
possibility. Their claim cannot work, since logically possible worlds have no
force in an argument of this sort; but when we turn to the relevant essentially
possible worlds, we see that there cannot be such worlds as they have in mind.
This was a subtle example which has helped to demonstrate the usefulness of
essences in thought experiments. When we test philosophical theories, we need
to imagine certain states of affairs meant to point out what seems contingent
or alterable about the theory and what seems necessary and unalterable about
the theory. If we look only at the logical structure of the claim under
investigation, we will find that it rules out few counterexamples— it will rule
out only what does not contradict the narrow definitions of the terms. On the
other hand taking essences into consideration we will rule out more
counterexamples. Given that essences are empirically discoverable, when we
utilize them we are closer to the way things really are. This might not have
the virtue of the a priori necessity that philosophers desire for their
theories, but it has the virtue of a certain kind of a posteriori necessity,
and this is all the necessity that is required.