The Essence of Response-Dependence
European Review of Philosophy 3 (1997), 31-54.
Long ago, Democritus proclaimed: ‘by convention colour, by convention bitter, by convention sweet: in reality atoms and the void’ (Barnes 1987, p.254). Ever since, philosophers have been tempted by the view that colours, flavours, values, and the like, are less objective than shape or mass or motion. It can be true that certain objects are square, or that they have a certain mass, purely in virtue of how those objects are in themselves; but it is not true, purely in virtue of how the objects are in themselves, that they are red, or bitter, or good.
In this rough and intuitive form, this thought has always seemed attractive to philosophers. However, it has proved difficult to develop the thought into a more precise theory without incorporating other, less attractive elements. In this paper, I shall explore a relatively new approach to capturing the thought that colours, or flavours, or values, are less objective than shape or mass or motion - the approach based on the idea of ‘response-dependence’. In the first section, I try to indicate an intuitive distinction, between the more objective and the less objective, which I believe that this approach should capture. In the second section, I examine the conceptions of response-dependence that have been developed by Mark Johnston, Philip Pettit and Crispin Wright, and I argue that they fail to capture this intuitive distinction. Then, in the following three sections, I propose an alternative conception that succeeds in capturing the intuitive distinction.
1. Response-Dependence: The Intuitive Distinction
Among analytic philosophers, there have been three main approaches for developing the intuitive thought, that colours and values and the like are less objective than shape or mass, into a more precise theory: non-cognitivism, eliminativism, and various broadly subjectivist forms of reductionism. Recently, each of these three approaches has come to seem deeply unattractive. In this paper, I shall not rehearse the objections that have been directed against these approaches over the years. But it is important to explain why the response-dependence approach is distinct from those older, more familiar approaches.
If colours or values are less than fully objective, then statements describing objects as red, or good, cannot be true purely in virtue of how those objects are in themselves. Non-cognitivists and eliminativists both maintain that such statements are not true at all: according to non-cognitivism, such statements, though meaningful, are neither true or false; according to eliminativism, such statements are all either false or meaningless. According to the response-dependence approach, on the other hand, some such statements are actually true. So this approach is incompatible with both non-cognitivism and eliminativism.
This approach accepts then that some statements describing objects as red or good are true. If it is to claim that colours or values are less objective than shape or mass, it must deny that these statements are true independently of us, purely in virtue of how those objects are in themselves. So this approach is committed to holding that these statements are true, not independently of us, but in virtue of those objects’ relations to us.
More specifically, this approach claims that, when an object is red or good, it is red or good in virtue of some relation to our subjective responses. However, it is not committed to giving strictly non-circular necessary and sufficient conditions for an object’s being red or good, in terms of the subjective responses that such objects evoke in us. Since the approach does not insist on such non-circular necessary and sufficient conditions, it is distinct from reductionism (including all broadly subjectivist forms of reductionism).
Although the response-dependence approach is distinct from reductive subjectivism, it too involves giving necessary and sufficient conditions, for what it is for something to be red, or good, in terms of the subjective responses that red or good objects sometimes evoke. Many philosophers have suggested that such accounts might take the following form:
x is F if, and only if, x is disposed to evoke subjective response R in subjects S in circumstances C.
In this paper, however, I shall not assume that all such accounts must employ the notion of a disposition in this way. There are numerous controversies surrounding the notion of a disposition (see Blackburn 1993 and Johnston 1992, pp.228-34). For example, if x is disposed to φ in circumstances C, does this mean merely that, if x were in C, it would φ? Or does it mean that x has the ‘higher-order property’ of having some intrinsic natural properties that would typically cause objects to φ in C? A general characterization of response-dependence need take no position on these controversies. Moreover, we should not rule out the possibility that there is some other sort of account of what it is for things to be F, in terms of some other sort of relation between objects that are F and some type of subjective response to such objects, that would also support the conclusion that being F is response-dependent. In this paper then, I shall merely assume that a response-dependence account, of what it is for things to be F, involves a claim of the following, more general form:
x is F if, and only if, A(x, R)
- where R is a type of response, on the part of thinking subjects, that essentially involves some sort of representation or recognition of something’s being F.
What has this to do with subjectivity? Gideon Rosen has objected that it is a perfectly objective matter what subjective response an object is disposed to evoke in certain circumstances (1994, pp.289-297). For example, it is a perfectly objective matter whether a certain object is disposed to evoke some mental response, such as annoyance or nausea, in normal dogs: it is a purely objective question of canine psychology. So how does a biconditional of this form show that there is anything less than fully objective about being F?
To respond to this objection, we should remember two points. First, the response-dependence approach claims that objects are red, or good (or whatever), at least partly in virtue of some relation to some type of mental response on the part of thinking subjects; and, second, the type of mental response in question involves some sort of recognition or representation of something’s being red or good. (So, for example, if nausea does not essentially involve any sort of recognition or representation of something’s being nauseating, then nausea is not a response of the appropriate kind.)
Because of these two points, the response-dependence approach captures an intuitive thought that is suggested by the line from Hamlet (2.2.247-8): ‘There is nothing either good or bad but thinking makes it so’. According to Hamlet, it is being thought good that makes something good, and being thought bad that makes something bad: if I think that racism is bad, then my thought cannot be detecting any fact of the matter that is constituted independently of my thought; on the contrary, my thought is part of what makes it the case that racism is bad in the first place. In this way then, this view does seem to imply that goodness and badness are less than fully objective.i
There are many areas in which the idea of response-dependence might prove fruitful. Johnston suggests that for a law to be constitutional within the United States just is for the Supreme Court not to be disposed ultimately to regard the law as unconstitutional (1993, p.104). To take another example, there has been a vigorous debate in gay and lesbian studies, about whether or not sexuality is ‘socially constructed’ (Halperin 1990 and Stein 1993). This may be interpreted as the question whether it is part of what it is to be, for example, a homosexual, that one identifies oneself as a homosexual, or at least is a person of a type which is actually classified in that way in one’s society. However, the most familiar examples are the secondary qualities, such as colours. According to this sort of view, it is at least part of what it is for something to be red that it has a certain relation to human visual experience. Consider for example this account of what it is to be red:
x is red if, and only if, x has some of the intrinsic natural properties which would typically cause objects to look red when seen in suitable conditions.
This account is compatible with red things’ being extraordinarily heterogeneous with respect to the ‘intrinsic natural properties’ that cause them to look red. So it is tempting to read this account as making this relation to human visual experience, rather than any physical property, the fundamental feature that all red things have in common, in virtue of which they all count as red; and on this reading the account seems to imply that redness is response-dependent.
According to this view, redness is capable of cutting sharply across the lines of intrinsic natural dissimilarities. Plato suggests that in classifying things we should not be like clumsy butchers, who hack their way through bone; instead, we should try to carve through the joints (Phaedrus 265e). Intuitively, it seems, if redness is dependent in this way on some type of human subjective response, it would not mark any joint that the world has independently of us; this is why it is less objective than properties (such as primary qualities or natural kinds perhaps) that do mark such independent joints in the structure of the world.
So the idea of response-dependence seems to capture an important distinction, between the objective and the subjective, in an intuitively appealing way. As we shall see, however, the idea is still in need of clarification.
2. Johnston, Pettit, and Wright
The easiest way to give a general characterization of response-dependence is by means of a schema: A response-dependence account, of what it is for things to be F, accepts that it is true that some objects are F, but claims that objects are F at least partly in virtue of some relation to some type of subjective response to such objects.
This schema shows that a response-dependence account is an ontological or metaphysical account, of what it is for things to be F. It is stated entirely at the level of reference, not at the level of sense. It is not a semantical account, of the meaning of the term ‘F’; nor is it an epistemological or psychological account, of the way in which we think about, or achieve epistemic access to, an object’s being F. In that sense, it is not an account of the concept F - at least not if the concept is the meaning of the term ‘F’, or a way of thinking about the property of being F. Of course, an account of the concept F may sometimes imply an account of what it is for things to be F. Nonetheless, in the first instance, they are about two different things: one is an account of a certain way of thinking of things’ being F; the other is an account of what it is for things to be F.
If we want to regard the latter account as an account of some entity, then it is an account of the property of being F, not of the concept F. So, to characterize response-dependence by means of an explicit statement, rather than just a schema, we must quantify over properties: A property counts as response-dependent if, and only if, it is part of what it is for something to have the property that it stands in a certain relation to a certain mental response to that property. For the rest of this paper I shall speak in terms of response-dependent properties. (For the purposes of this paper, I shall prescind from nominalist qualms about the existence of properties.ii)
To defend the intuitive idea that colours or flavours or values are less objective than primary qualities or natural kinds, we must argue that the properties of being red, or bitter, or good, are response-dependent. It is not enough to show that the concepts red or good are response-dependent in some way. Even if these concepts are response-dependent in some way, it needs to be shown that the properties that these concepts stand for are also response-dependent. If those properties are not response-dependent, it would be true that some objects have those properties purely in virtue of how those objects are in themselves: those properties (redness, goodness, and so on) would be perfectly objective. Hence the core of any adequate conception of response-dependence must be an account of response-dependent properties, not of response-dependent concepts. This, I shall argue, is the basic problem with the conceptions of response-dependence that have been developed by Mark Johnston, Philip Pettit, and Crispin Wright: they are accounts of response-dependent concepts, not of response-dependent properties.
Thus, Johnston focuses on what he calls ‘response-dispositional concepts’ (Johnston 1993, pp.103-11). Assuming, for example, that our concept red is a response-dispositional concept, then the concept just is the concept of the disposition to produce experiences of the relevant type. In general, a response-dispositional concept just is the concept of a disposition to evoke responses of some specific kind. (A ‘response-dependent concept’ is a response-dispositional concept or a truth-functional combination essentially involving a response-dispositional concept.)
There are at least two reasons why this account does not provide what I am looking for. First, as I have already suggested, it is not clear why all response-dependent properties must be definable in terms of dispositions to evoke responses in suitable conditions. Perhaps an account of a property in terms of some other sort of relation, between instances of the property and some type of mental response to it, would also reveal the property in question to be response-dependent.
Second, Johnston’s distinction is a distinction between two types of concept, not between two types of property. It is possible for two different concepts to stand for the same property: for example, the concept water and the concept H2O plausibly both stand for the same property. Even if a property is in fact a disposition to evoke certain responses, there could be a concept standing for that property which is not itself a concept of such a disposition. (For example, consider the concept the property of the flag that explains why the bull got angry.) So it could be that some, but not all, of the concepts standing for a given property were response-dispositional concepts. But what I am looking for is a metaphysical distinction between two types of property - not a distinction between two types of concept.
Of course, we could base such a distinction between types of properties on a parallel distinction between types of concepts. Suppose that we had an account of what it is for a concept to present the property that it stands for as response-dependent (where, contrary to Johnston’s account, this is not limited to presenting the property as a disposition to evoke the relevant responses). Then we could define a response-dependent property as a property that can be thought of by means of any such concept. But it is not clear that anything is gained by this indirect route. Why should it be easier to give an account of what it is to think of a property as response-dependent than to give an account of what it is for a property to be response-dependent? Of course response-dependent properties are correctly thought of as response-dependent. This does not show that an account of what it is for a property to be response-dependent must be based on an account of what it is for a concept to present the property as response-dependent.
A still stronger point applies to what Pettit used to call ‘response-dependent concepts’ (1991) and now calls ‘response-privileging’ or ‘response-authorizing’ concepts (1993). According to his conception, a concept is response-authorizing if there is an a priori guarantee that (given certain suitable conditions) an object falls under the concept if, and only if, we are disposed to respond to the object in a certain way. There is, however, no reason to think that the properties that Pettit’s response-authorizing concepts stand for cannot be entirely objective in every way. He suggests that even natural kind concepts, such as our ordinary concept of water, may be response-authorizing concepts (1991, p.598 and 615). And he explicitly rejects the suggestion that the objects or properties that response-authorizing concepts stand for need be in any way dependent on our responses (1993, pp.202-3):
the assertion that a concept is response-authorizing is, precisely, an assertion about the concept, not an assertion about that of which it is a concept: not an assertion about the property or object or operation in question. It is to say that the reference of the concept is determined in such a way that our responses are privileged under certain conditions: they are not capable of leading us into ignorance or error. It is not to say anything about the property or object or operation in itself, and so a fortiori it is not to say that that entity is dependent on us....
Clearly, Pettit’s conception of response-authorizing concepts does not provide an account of the metaphysical distinction, between objective and subjective properties, that interests me here.
Wright’s ideas also centre around a distinction between two types of concept: concepts for which our ‘best opinions’ are involved in determining the concept’s extension, and concepts for which these best opinions merely reflect an independently determined extension (see his 1988, 1989, and 1992). This is supposed to be exemplified by the way in which Euthyphro’s definition of the pious, as that which the gods love, reveals the extension of ‘pious’ to be determined by the best opinions about piety - that is, by the opinions of the gods.
According to Wright, the view that our best opinions are involved in determining a concept’s extension is a denial of one sort of metaphysical realism about the property that that concept stands for. Offhand, however, his ideas seem, like Pettit’s, to have more to do with concepts, and with the way in which it is determined which property a given concept stands for, than with the nature of that property itself. Nonetheless, his ideas are at least a refinement of Pettit’s. He too considers a priori ‘provisoed biconditionals’ or ‘provisional equations’ of the form:
If subject S is in circumstances C, then: x is F if and only if S judges that x is F.
But he imposes a few more conditions on these provisional equations than Pettit does. So we should consider whether these additional conditions ensure that the property of being F really is a response-dependent property.
Wright suggests that ‘best opinions’ play this ‘extension-determining’ role if such a provisional equation meets the following four conditions. First, the equation must be a priori. Second, circumstances C must be substantially specified: they must not be specified in a trivializing fashion, as circumstances that have ‘whatever it takes’ for S to be right about whether or not x is F. Third, whether circumstances C obtain must be independent of the detail of the extension of the concept F: to ensure this, circumstances C should be specified without using the term ‘F’, except perhaps within intentional contexts attributing propositional attitudes to S. Finally (the ‘Extremal’ condition), there should no rival explanation of why the provisional explanation is true, other than the explanation in terms of the role played by best judgments in determining the extension of the concept.
It seems to me that there are provisional equations involving natural kind concepts that meet these conditions. Suppose for example that S’s circumstances are as follows. Every sample of a natural kind in S’s environment presents the type of appearance that samples of that natural kind actually typically present; and, for every type of appearance that S would regard as the typical appearance of some natural kind, nothing in S’s environment is presenting that appearance other than samples of the unique natural kind that has actually typically presented that appearance in the past. Moreover, S assumes that her perceptual circumstances are typical in this way and generally favourable; she perceives all the samples of natural kinds in her environment, registers the appearance that they present, and forms all the non-inferential judgments that it is rational for her to form on the basis of registering these appearances; and she forms no other judgments in the circumstances. Finally, S perceives x; and S possesses the concept water. Let us refer to circumstances of this kind as ‘circumstances C’. Then, plausibly, this provisional biconditional holds:
If S is in circumstances C, then S judges that x is a sample of water if and only if x is a sample of water.iii
This provisional biconditional appears to meet Wright’s four conditions. (i) According to a plausible story, the reference of the concept water is fixed as that natural kind (if any) that actually typically presents the type of appearance that those who possess the concept treat as their primary means of recognizing water. This story implies that the provisional biconditional is a priori. (ii) Circumstances C are clearly substantially specified: the provisional biconditional as a whole is far from trivial, since it implies various non-trivial theses - such as that, if water is a natural kind, then anyone who possesses the concept water will be able to recognize water when it presents its actually typical appearance. (iii) Circumstances C are also independently specified, since the specification does not use the term ‘water’ except to say that S possesses the concept water; and that is equivalent to using the term within the scope of propositional attitude ascriptions. (iv) Finally, the best explanation of why this provisional biconditional holds is by appeal to the role that our ‘best judgments’ play in fixing the reference, or determining the extension, of our concept water - best judgments here being judgments that are caused by the natural kind that actually typically causes the experiences that form our primary basis for perceptual judgments involving the concept.
This is a problem for Wright’s account because, as Wright himself recognizes (1992, p.131), it is intuitively plausible that natural kinds are not response-dependent properties. If any properties mark joints that the world has independently of us, then natural kinds do. So Wright does not give an adequate account of the metaphysical distinction that interests me any more than Johnston or Pettit do.
3. ‘Constitutive Accounts’: What Are They?
I propose that a property is response-dependent just in case any adequate constitutive account of what it is for something to have the property must mention some type of mental response to that property. The relevant sort of dependence then is what might be called ‘constitutive’ dependence. But what is a ‘constitutive account’ of a property?
The idea of a constitutive account of some object or property is expressed by many phrases that are commonly used by philosophers, though their meaning is rarely adequately explained. For instance, if we take as our example the property of value or being valuable, then one way to express this idea is to speak of an account of what constitutes a thing’s value, or of what its being valuable consists in. As with almost all phrases that express this idea, this way of speaking can be transposed into the formal mode; we can speak of what the truth of the statement that a thing is valuable consists in, or of what constitutes the truth of that statement. Another phrase that expresses the same idea is a special non-causal use of ‘makes it the case that...’. Thus, we speak of what makes it the case that a thing is valuable, or simply of what makes it valuable, or (in the formal mode) what makes it true that the thing is valuable. In a similar way, we also speak of the ‘truth-maker’ for statements about value. Other explanatory idioms can also express this idea. We can speak of that in virtue of which a thing is valuable, or of what the statement that it is valuable is true in virtue of. More simply, we can just ask why a thing is valuable, or why it counts as valuable, or what explains why it is valuable, so long as it is clear that ‘why’ and ‘explanation’ are not to be given a specifically causal interpretation. Finally, we may also speak of what it is for a thing to be valuable, or simply of what value fundamentally is. A statement of what value is, in this sense, may be called a constitutive account or explanation of value, and perhaps also an analysis or definition of value.
Though commonly used in philosophy, these phrases are hard to understand; and we need to understand them better if my proposal is to shed any light on the nature of response-dependence. I shall continue to assume here that constitutive accounts can be stated in the form of necessary, universally quantified biconditionals. I shall also assume that these biconditionals do not have to be non-circular; that is, the right-hand side of the biconditional need not avoid using any simple term that stands for the property that is being accounted for (nor need it avoid using any simple term that stands for some other entity a full constitutive account of which would have to speak of the property that is being accounted for). If a constitutive account of a property is fully non-circular in this way, then it amounts to a reduction of the property in question. Indeed, we could even define reductions in this way, as completely non-circular constitutive accounts: if a necessary biconditional is to give a reduction of the property ascribed on its left-hand side, then it must not only be non-circular, but the right-hand side must also give a constitutive account of what it is for the left-hand side to be true.iv
These biconditionals then do not have to be reductive or non-circular in order to be genuine constitutive accounts. But clearly not all necessary biconditionals are genuine constitutive accounts: otherwise, there would never be any reason to read such a biconditional as giving a constitutive account of the property ascribed by the left-hand side of the biconditional, rather than the property ascribed by the right-hand side. The biconditional would simply reveal its two sides to be necessarily equivalent: we would have no reason to regard it as a constitutive account of the property ascribed by one side rather than the other. In many cases of constitutive accounts, however, we do have a reason to treat the two sides of the biconditional in this asymmetric fashion.
Someone might propose that such a biconditional is a constitutive account of the property ascribed by its left-hand side just in case it is a conceptual truth, guaranteed to be true by the nature of the concept that is used to ascribe the property in question. But this proposal can be shown to be inadequate by appeal to the example of water that we considered earlier. The following seems to be a conceptual truth, guaranteed to be true by the nature of the concept water:
x is a sample of water if and only if x is a sample of the underlying natural kind that actually typically causes the sort of experience that we use as our primary means of recognizing water.
But intuitively this is not a constitutive account of the nature of water. This connection to human experience is not part of what water is; it is incidental to what makes something water.
This point has an important consequence for the theory of response-dependence. The fact that there is a biconditional conceptual truth, where the left-hand side ascribes a property to some arbitrary object, and the right-hand side speaks of some relation between the object and some type of mental response to the property, is not enough to show that the property in question is response-dependent. Otherwise, the property of being made of water would be response-dependent. We must impose a further condition: the biconditional must also be a constitutive account of the property in question. But we have still made no progress towards understanding what a constitutive account of a property is.
Moreover, it seems that being a conceptual truth is not always even necessary to make a biconditional into a constitutive account. Consider the following biconditional:
x is a sample of water if and only if x is a sample of H2O.
This is obviously not a conceptual truth. Still, it seems to me to give a constitutive account of water: it explains what it is about something that makes it a bit of water; it tells us what water fundamentally is. Admittedly, it may be that all the constitutive accounts that reveal properties to be response-dependent are conceptual truths: the accounts of what it is to be red, or constitutional, that I considered above would seem to be conceptual truths, if they are truths at all. But the general point still holds: constitutive accounts are not always conceptual truths, and conceptually true biconditionals are not always constitutive accounts.
As the example of water makes clear, a constitutive account of a property tells us something about the property itself, and need not elucidate or give the meaning of the concepts that may be used to think about that property. Of course, we could give a constitutive account of a concept too: we could grapple, not with the question ‘What is water?’, but with the question ‘What is the concept water?’ - that is, ‘What is it to think about water?’ But that would be an account of something entirely different: it would not be an account of water, of a type of material stuff; it would be an account of a concept or type of thought.v Perhaps our concept water - the way in which we typically think of water - depends on our perceptual experiences of water; it does not follow that water itself has any such dependence on our experiences.
In exploring the nature of response-dependence then, we must bear in mind this vital distinction between a constitutive account of a property and a constitutive account of a concept that can be used to ascribe that property. We must understand what ‘constitutive accounts’ are in general, regardless of whether they are accounts of concepts in human thought or language, or accounts of properties that are real features of external things.
Once it is clear that we are considering constitutive accounts of properties themselves, not just constitutive accounts of the concepts that we use to ascribe those properties, then we may well wonder how the biconditional linking water with H2O can be a constitutive account of water. It is highly plausible that the property of being made of water is exactly the same property as the property of being made of H2O. So why should we take one side of the biconditional as giving an account or analysis of the other, rather than vice versa? We simply have the same property twice over, once on each side of the biconditional.
Part of the answer to this challenge must be that, even if the property ascribed on each side of the biconditional is the same, the analysing side speaks of certain other objects or properties that are not spoken of on the analysed side. Obviously, if we analyse water as H2O, we speak, not only of the property being analysed - namely, water or H2O -, but also of hydrogen and oxygen, which are not spoken of on the analysed side of our biconditional. This is why the analysing side of these biconditionals is usually more complex than the analysed side.
This, however, is only part of the answer. If we assume that any two necessarily equivalent predicates stand for the same property, then it is quite trivial to generate biconditionals where both sides ascribe the same property to some arbitrary object, but where one side also speaks of certain further objects or properties that the other side does not. For example:
x is red iff x is red and a member of the set that contains x and all the natural numbers.
But obviously this does not count as a proper constitutive account of what it is to be red.
The point must be that, even if a given property can be ‘constructed’ (by means of operations that are analogues of conjunction, negation, existential generalization, and so on)vi out of various other objects or properties, it does not follow that (to put the point crudely) the property itself involves those objects or properties as constituents. For example, even if the property of being red can be constructed out of various arithmetical and set-theoretical objects in this way, it need not itself involve any of them as constituents. On the other hand, if any adequate constitutive account of what it is for something to be red must mention some type of visual response, then the property of being red would involve that type of visual response as a constituent. However, we still need to understand what exactly a ‘constitutive’ account amounts to. I shall attempt to develop such an understanding in what follows.
4. The Essentialist Interpretation of Constitutive Accounts
Aristotle clearly believed that when Socrates asked Euthyphro ‘What is piety?’, he was seeking a real definition of piety.vii A real definition of some object (whether a particular or a universal) is a formula that states the essence of that object. Perhaps then the feature that makes a necessary biconditional into a constitutive account is that it should be a real definition, or state the essence, of a particular or universal that is spoken of on the left-hand side. This proposal seems the more plausible since Aristotle’s favourite formula for essence - phrases of the form ‘what it is for something to be a human being’ - is also, as we have seen, one of the phrases that contemporary philosophers employ to indicate that they are giving a constitutive account. It is also plausible that it is essential to redness that something is red if and only if it has one of the intrinsic natural properties that would typically cause objects to look red in suitable conditions.
As I claimed in the previous section, there is no reason to demand that all constitutive accounts should be both non-circular and individuating. To impose this demand is to insist that these accounts should be reductive. But real definitions do not have to be reductive or non-circular. Assuming that one object ‘constitutively depends’ on another if the second object must be mentioned in a real definition of the first, then we can make the same point by saying that constitutive dependence is not asymmetric: it is possible for two objects each to depend constitutively on the other. This, anyway, is the proposal that I shall try to develop here: that the distinctive feature of constitutive accounts is that they are real definitions or statements of essence.
This proposal faces a serious obstacle, however. This obstacle is the widespread belief that the notions of real definition and of essence, if they are intelligible at all, are reducible to more simple modal terms. Those contemporary philosophers who tolerate the notion of essence simply define the essence of an object as those properties which it is impossible for the object to exist without. But then the distinction between a necessary biconditional, and one that states the essence, or gives a real definition, of something spoken of on the left-hand side, will simply collapse. It is impossible for water to exist without being the basic kind of stuff that is actually dominantly causally responsible for the experiences that we use as our primary means of recognizing water. So it would follow that this connection to our experience is an essential feature of water, according to this modal view of essence.
There are however strong reasons against this modal conception of essence: as Kit Fine has pointed out, it has several quite unacceptable consequences. It is impossible for Socrates to exist without being a member of the set that includes Socrates and all the natural numbers. So according to the modal conception of essence, it is part of the essence of Socrates that he is a member of this set. But intuitively, this is not part of Socrates’ essence or nature. As Fine puts it (1994, p.5): ‘Strange as the literature on personal identity may be, it has never been suggested that in order to understand the nature of a person one must know which sets he belongs to.’viii
Rather than repeating Fine’s attack on the modal conception of essence, I shall simply assume that we need a new conception of essence. But I shall not attempt any reductive definition of essentialist notions in other, non-essentialist terms. It seems plausible to me that Fine is right: these essentialist notions are fundamental metaphysical concepts, which are incapable of any non-circular definition. Still, these essentialist notions can be explicated by characterizing their logical properties, and their connections to other metaphysical concepts. Indeed, I suspect that these essentialist notions are of the greatest importance in clarifying many other metaphysical concepts - including the concepts of supervenience and ontological dependence, of the relations of part to whole, and of determinate to determinable, of the way in which a belief (for example) can be ‘realized’ by a certain neurophysiological state, and so on. However, I cannot pursue these questions here. The only way in which I will try to explicate the concept of essence is by giving a rough sketch of what is arguably its most fundamental connection - that is, its connection to modality.
Even if we reject the idea that essence can be defined in purely modal terms, there is clearly a strong link between essentialist and modal notions. Fine proposes that we should reverse the order of priority between the concepts of essence and modality. Instead of defining essence in terms of modality, we should define modality in terms of essence, in something like the following way.
We may think of the essence of an object (whether an individual or a universal) as given by the real definition of that object - that is, by the basic necessary principle which, together with the essences of other things, is the source of all the object’s modal properties. If the object is an individual, then this basic necessary principle will concern what it is for something to be that individual; it will be the principle that determines which individual (if any), in any possible world, is identical to the individual in question. In this way then, the essence of an individual is what explains its identity: it is its principium individuationis, what individuates it, or makes it different from all other things.
On the other hand, if we are concerned, not with an individual, but with a property or relation, then the basic necessary principle about the property or relation will concern what it is for a sequence of objects to exemplify this property or relation: it will be the principle that determines which sequences of objects (if any), in any possible world, are instances of that property or relation. In this way, the principle will reveal what ‘unifies’ the property, or constitutes the ‘real similarity’ shared by all its instances, both actual and counterfactual.
Suppose for example that the essence of being red is to have some intrinsic natural property that would typically cause objects to look red when seen in suitable conditions. This then would be the source of various modal properties of redness, such as the impossibility of ‘fool’s red’, or the possibility of red objects that are never seen, or of red objects that are physically quite unlike the red things of the actual world. Redness has an essence because these modal truths involving redness are not chaotic: they all have their source in, and are explained by, the fundamental necessary principle concerning what it is for something to be red - that is, the essence of redness.
But what is meant by saying that these principles are the source of the modal properties of objects? I can only suggest the rough outline of an answer to the question; and I will not be able to defend this answer in detail. But very roughly, we may say that a proposition is necessary if it is a logical consequence of these fundamental necessary principles concerning things; and a proposition is possible if it is logically consistent with these fundamental necessary principles.ix But what sort of ‘logical consequence’ is in question here?
Just as I am treating properties as real features of things, and not just as predicates or concepts in human thought or language, so, in the same way, I shall assume that propositions are conditions, or possible states of affairs; that is, they are real ways the world might be, or real conditions that the world might be in, not just thoughts or sentences in any human thought or language. Many of these conditions or propositions will be conditions that actually obtain - that is, ways the world actually is, or facts. So it is important here that we do not interpret ‘logical consequence’ in metalinguistic terms; we need a conception of logical consequence that applies directly to propositions, conceived purely as ways the world might be, not as sentences in any language. Equally, however, we should not understand logical consequence simply as ‘necessary consequence’, conceived in purely modal terms; we should avoid presupposing the modal notions that we are trying to explain.
I have assumed that propositions and properties (more precisely, universals of any kind) can be constructed out of objects (whether individuals or universals) by means of various operations, which are analogues of predication, negation, conjunction, and so on. The required conception of logical consequence can be defined in terms of these operations. Just as there will be some fundamental necessity concerning each individual or universal, so too there will be a fundamental necessity concerning each of these operations. We may assume that the fundamental necessity associated with each of the operations will in effect be a familiar logical principle, such as: ‘Necessarily, for any conditions (or propositions) A and B, any condition that can be constructed by conjunction out of A and B obtains iff A and B both obtain.’ This principle defines the nature of the conjunction operation; and principles of this kind enable the required notion of logical consequence to be defined in a familiar way. Roughly, we can imagine these principles as constraining the construction of sets of propositions: for example, if such a set of propositions contains a proposition that can be constructed by conjunction out of A and B, then it must also contain A and B as well. Then we can say that one proposition is a logical consequence of a second if any of these sets that contains the second proposition also contains the first as well.x My tentative suggestion then is that it is in this sense that all necessary propositions are logical consequences of the essences of things; and this is how it is that the essences of things constitute the source from which all metaphysical necessity and possibility flow.
5. Response-Dependent Essences
This then is the essentialist conception of ‘constitutive’ accounts that I propose to appeal to in elucidating the idea of response-dependence. A property is response-dependent just in case it is an essential part of something’s being an instance of the property that it stands in some relation to some sort of mental response to that property. If redness is response-dependent, then it must be essential to something’s being red that it stands in some relation to a certain type of visual response to redness. If ethical qualities, such as the property of being morally wrong, are response-dependent, then it must be essential to something’s being morally wrong that it stands in some relation to some type of disapproval or opposition that is a response to such moral wrongness.
According to this proposal, such statements of essence, or real definitions, provide the most fundamental constitutive explanations that can be found. There is no room for any further question about why ethical qualities are response-dependent while primary qualities and natural kinds are not. That is simply a basic constitutive fact about these various types of property. However, although this fundamental metaphysical fact neither requires nor admits of further explanation, we can certainly still ask the epistemological or methodological question: What justifies us in believing that ethical qualities are unlike primary qualities and natural kinds in this respect?
The answer to this question, I believe, is that such beliefs can be justified by reflection on the nature of the concepts that we use to ascribe these qualities. In some cases, for example, we can simply introduce a concept by the stipulation that it is to stand for a property that has a certain essence. In other cases, analysis of some concept that we already possess may reveal that the concept can only stand for a property that has a certain sort of essence. In yet other cases, conceptual analysis may reveal that some concept in our repertoire does not determinately stand for any unique property; and one may propose a conceptual reform determining that the concept is to stand for a property that has a certain essence.
What I am proposing now is quite compatible with my earlier claim, that it is a mistake to explain what it is for a property to be response-dependent in terms of the nature of the concepts that can be used to ascribe that property. I am now suggesting that we must appeal to the nature of our concepts to explain how we can know that a property is response-dependent. That is, I am proposing, not that essentialist notions can be reduced to the idea of conceptual truths, but that such essentialist notions actually figure in the content of certain conceptual truths.
Why should we expect that reflection on the nature of a concept should tell us anything about the essence of the property that the concept stands for? The key point is that a concept cannot succeed in determinately singling out any unique property unless the concept, along with the thinker’s circumstances, also singles out the essence of that property as well. To single out a unique property, out of all the countless properties that there are, a concept must not only have the right extension in the actual world: it must also have the right extension in all other possible worlds as well; and I have suggested above that it is precisely the essence of a property that determines which objects (if any), in any possible world, are instances of that property. So singling out the essence of a property is an essential part of singling out the property itself; and there must be something about the nature of the concept that enables the concept, together with the thinker’s circumstances, to fix on the essence of the property that the concept is to stand for.
For example, it seems somehow built into the concept water that, if the concept stands for anything at all, then it stands for a kind of stuff whose essence consists in its underlying nature, rather than its superficial appearances. So the nature of the concept water determines that the concept must stand for a natural kind, not a response-dependent property. On the other hand, the nature of the concept red determines that it must stand for a response-dependent property: it is built into the concept red that the concept stands for a property that essentially consists in having any feature that will present the right sort of superficial appearance in appropriate conditions.
I cannot undertake an investigation of the nature of concepts here, or of what it is for a certain truth to be ‘built into a concept’. But the claim that an essentialist truth is built into the concept red need not imply that everyone who possesses the concept red must possess the concept of essence. On the part of the thinker himself, there need be no more than a vague grasp of what makes conditions for perceiving red things unfavourable, and of what it is for an unperceived object to be red. Even here there is a sharp contrast between the concepts water and red. Conditions are unfavourable for perceiving water wherever there is anything superficially resembling water whose underlying nature is different from that of most samples of water; whereas conditions are unfavourable for perceiving redness only if there is some abnormality in one’s perceptual function, or if the lighting and atmospheric conditions differ too much from a certain familiar paradigm. It is plausible that these differences between the concepts red and water support the claim that there is an important metaphysical difference between redness and water. The superficial appearances familiar to those who possess the concept water are merely involved in ‘fixing the reference’ of the concept; whereas the superficial appearances familiar to those who possess the concept red are involved, not merely in fixing the reference of the concept, but in the essence of the property that the concept stands for.
In several respects, I concede, my proposed conception of response-dependence remains programmatic. In particular, it relies on a concept of essence that many philosophers will regard as dubiously intelligible. I am confident that further investigation will dispel many of these suspicions. But until such further investigation is carried out, we will not know whether there is anything behind the intuitive distinction, between more and less objective properties, that the response-dependence approach attempts to clarify.xi
iRosen argues that the response-dependence theorist must accept that, if one side of any such biconditional is objective, then so is the other side; indeed, it is plausible that both sides of the biconditional state exactly the same fact. So he would respond to my claim about Hamlet’s view of goodness by claiming that it is an entirely objective (psychological) matter whether racism is thought bad; hence it must be an equally objective matter whether racism actually is bad. - But, if x’s being thought bad is exactly the same fact as x’s being bad, then it is a fact with a very curious feature: it essentially involves a representation of itself. For the fact to obtain it must be thought to obtain. So, even accepting Rosen’s assumptions, it seems plausible to me to deny that this fact (that x is thought bad) is quite as objective as it first appears.
iiWe should also note that a response-dependence account of what it is for things to be F is not an account of what it is for the property of F-ness - conceived as a universal, or as an abstract object of a certain sort - to exist. A response-dependence account of what it is for things to be F need take no stand on the ontological status of properties. Suppose that we believed that the existence of the property of F-ness, conceived as an abstract object of a certain kind, was dependent on the concept or concepts that can be used to ascribe this property. This does not commit us to any particular view of what it is for something to be F; in particular, it does not commit us to the view that nothing can be F completely independently of our concepts.
iiiFor example, consider some unusual artifact that presents exactly the type of appearance that S would regard as the typical appearance for water. This is obviously liable to prompt S to form false judgments of the form: x is a sample of water. But circumstances C will not contain any such thing. If anything in S’s environment is presenting that type of appearance, then it is a sample of the unique natural kind that has typically presented that type of appearance in the past - viz. water itself. (If there is no unique natural kind that has typically presented that appearance in the past, then nothing in S’s environment will present that type of appearance.) Because S assumes that her circumstances are typical in this way, and forms only rational, non-inferential judgments on the basis of her registering these appearances, there is also no chance that she will think that x is some unusual sample of water that presents an appearance other than that which water typically presents.
ivThis is the view of Michael Dummett (1993, p.57), who states that a reduction must give an account of what the reduced sentence is true in virtue of, if it is true. (An alternative understanding of reductions is simply as property identities, involving the identification of the property ascribed by a ‘suspect’ concept, with the property ascribed by some ‘kosher’ concept. If one is sceptical of the coherence or interest of this distinction between ‘suspect’ and ‘kosher’ concepts, then one should be equally sceptical of the coherence or interest of this conception of reductions.)
vMany philosophers assume that an account of the concept water will be of the form: x is a sample of water iff A(x). But that is the form of an account of water itself, not an account of the concept water. An account of the concept water should take the form: A concept C is the concept water iff A(C). Compare Peacocke 1992, chapter 1.
viThe fundamental ontological conception that I shall be assuming, at least in order to fix ideas, is the theory outlined in George Bealer 1982, chapters 1-4. (I shall not be assuming the views that Bealer defends in later chapters.)
viiSee Irwin 1995, §15, pp.25-6. Irwin refers there to Aristotle’s Metaphysics 987b1-4 and 1078b23-30.
viiiSee also Fine 1995a. Fine’s two other main objections are as follows. (i) Where ‘p’ can be replaced by any necessary proposition, then, for any object x, it is necessary that if x exists then p; so the essence of anything involves all modal facts (which is absurd). (ii) It seems possible for two philosophers to agree on all the modal facts while disagreeing about essence: e.g. the philosopher who thinks that persons are embodied minds, so ontologically dependent on minds, versus the one who thinks that minds are abstractions from persons.
ixFine holds that the essence of an object consists of those propositions that are ‘true in virtue of the object’s identity’, and that necessary propositions are ones which are ‘true in virtue of the identity of all objects’ (see Fine 1994, p. 9, or 1995b, p.56). Fine also distinguishes between ‘constitutive essence’ - the properties of an object that are strictly definitive of its nature - and ‘consequential essence’ - the essential properties of the object closed under logical consequence. When he says that necessary propositions are ‘true in virtue of the identity of all objects’ he is referring to the consequential sense of ‘true in virtue of x’s identity’. When I speak of essence I shall always mean constitutive essence: this is why I say that a necessary proposition is one that follows from the essences of objects.
xThis conception of logical consequence between propositions (conceived simply as ways the world might be) need not rely on the assumption that propositions have a unique intrinsic structure: a proposition has consequences, not in virtue of its intrinsic structure, but in virtue of all the ways in which it can be ‘constructed’, by means of various operations (analogous to predication, negation, conjunction, existential generalization, etc.), out of objects such as individuals, properties, relations, and propositions. It is the nature of these operations that is the source of this relation of logical consequence. (Of course, if we assume that each proposition has a unique intrinsic metaphysical structure, then the notion of consequence is more straightforward: it could simply parallel the metalinguistic notion for an ideal language that mirrors this basic structure of reality.)
Kit Fine writes, ‘we should think of the nature of the logical concepts ... as being given, not by certain logical truths, but by certain logical inferences. ... The concept of consequence is not presupposed but is already built into the rules’ (1995b, p.58). But what is an inference if not a linguistic structure or a psychological process? So I doubt whether Fine’s understanding of logical consequence is ideal for elucidating the ideas of essence and modality. Let me stress however that I have no objections to Fine’s formal ‘logic of essence’, which is developed in detail in his 1995c.
xiAn earlier version of this paper was presented to the University of St Andrews Philosophy Club, and to my former colleagues in the Department of Philosophy at the University of Stirling. I am grateful to both groups, and also to Alex Byrne, Timothy Williamson, Crispin Wright, and two anonymous referees, for helpful comments. I should especially like to record my gratitude to the late Murray MacBeath.
Barnes, Jonathan, 1987. Early Greek Philosophy (Harmondsworth: Penguin).
Bealer, George, 1982. Quality and Concept (Oxford: Clarendon Press).
Blackburn, Simon, 1993. ‘Circles, Smells, Finks and Biconditionals’, Philosophical Perspectives 7, 259-79.
Dummett, Michael, 1993. The Seas of Language (Oxford: Clarendon Press).
Fine, Kit, 1994. ‘Essence and Modality’, Philosophical Perspectives 8, 1-16.
Fine, Kit, 1995a. ‘Ontological Dependence’, Proceedings of the Aristotelian Society 95, 269-90.
Fine, Kit, 1995b. ‘Senses of Essence’, in Walter Sinnott-Armstrong, ed., Modality, Morality, and Belief (New York: Cambridge University Press).
Fine, Kit, 1995c. ‘The Logic of Essence’, Journal of Philosophical Logic 24, 241-73.
Halperin, David, 1990. A Hundred Years of Homosexuality (New York: Routledge).
Irwin, Terence, 1995. Plato’s Ethics (Oxford: Clarendon Press).
Johnston, Mark, 1992. ‘How to Speak of the Colors’, Philosophical Studies 68, 221-63.
Johnston, Mark, 1993. ‘Objectivity Refigured: Pragmatism without Verificationism’, in John Haldane and Crispin Wright (eds.), Reality, Representation and Projection (New York: Oxford University Press).
Peacocke, Christopher, 1992. A Study of Concepts (Cambridge, Mass.: MIT Press).
Pettit, Philip, 1991. ‘Realism and Response-Dependence’, Mind 100, 587-626.
Pettit , Philip, 1993. The Common Mind: An Essay on Psychology, Society and Politics (New York: Oxford University Press).
Rosen, Gideon, 1994. ‘Objectivity and Modern Idealism: What is the Question?’, in Michaelis Michael and John O’Leary-Hawthorne (eds.), Philosophy in Mind (Dordrecht: Kluwer Academic Publishers).
Stein, Edward, 1993. Forms of Desire (New York: Routledge).
Wright, Crispin, 1988. ‘Moral Values, Projection and Secondary Qualities’, Proceedings of the Aristotelian Society Suppl. Vol., 1-26.
Wright, Crispin, 1989. ‘Wittgenstein’s Rule-Following Considerations and the Central Project of Theoretical Linguistics’, in A. George (ed.), Reflections on Chomsky (Oxford: Oxford University Press).
Wright, Crispin, 1992. Truth and Objectivity (Cambridge, Mass.: Harvard University Press).