1 Preamble

In many philosophy of mind texts the Australian version of the mind–brain identity theory, the version that affirms type–type identities between mental states and brain states, is portrayed as the Hindenburg of modern materialist theories of mind – an interesting idea that went down in flames.¹ Those with identity theory predilections are advised to move to a token–token identity theory dressed in functionalist clothes (of varying hues, depending on the source of the advice). I think this is a mistake. This essay says why. This essay is not a full-blown defence of the Australian version of the identity theory. It is a defence of the view that it is the version of choice for those who favour some version or other of functionalism.

Two main reasons are offered for holding that the original version of the identity theory fails. First, the famous multiple realizability of the mental tells us, it is argued, that mental types cannot be identified with brain types. Second, even if this problem can somehow be overcome, the type–type version cannot give a satisfactory account of the causal role of mental properties – an embarrassing result, given that one of the main reasons offered in support of the theory in the first place is the need to account for the causal role of the mental.

I will spell all this out in due course but first we need to have a statement of the type–type theory in front of us, and of why many believed it. In the course of expounding the theory and a bit of the history behind its development, we will see why it had to be a type–type theory. I think an additional reason why the type–type version is unpopular is that many materialists don’t see the need to go beyond a token–token version.

2 The type–type identity theory: early expositions

Early discussions of the theory made much of the claim that it would turn out that there are close correlations between mental states and brain states. Supporters argued that one day we would find true biconditionals linking what subjects were thinking and feeling with what was going on in their brains. This was combined with the view that it is good scientific practice – and here Occam’s Razor was often invoked – to identify when we have close correlations. The right response, for example, to discovering that we have water if and only if we have H₂O, and whenever we have lightning we have an electrical discharge, is to identify water with H₂O, and lightning with an electrical discharge.² Of course, there is a use of the term ‘correlated’ which implies that if X and Y are correlated, they differ from each other, but the sense of ‘correlation’ in the discussions was the sense in which X and Y are correlated if whenever X then Y, and whenever Y then X.

It was granted by all that the principle that one should identify when one discovers close correlations was merely presumptive, and that the way to defeat it is to find differences in properties between the correlated items. This led to a series of papers debating the pros and cons of proposed candidates to be properties that differentiate brain states from mental states. We won’t review this literature.³

Obviously, any argument from correlations to identities between mental states and brain states is going to be an argument that, if successful, establishes type–type identities. For what gets correlated are types and not tokens. In addition, the examples from science cited by supporters of the identity theory are examples of type–type identities. For example, when it was discovered that lightning is an (atmospheric) electrical discharge, what was discovered was that a certain kind of happening, lightning, is an electrical discharge; likewise, water is a kind, and so the identity of water with H₂O is a type–type one. It is, therefore, not surprising that the mind–brain identity theory started life as a type–type identity theory. These discussions also prompted discussions of whether, and to what extent, information about the brain might be a good predictor of subjects’ mental states, and there was a certain amount of discussion of what attitude subjects themselves should take to cases where the best judgement based on an examination of what is going on in their brains about their current mental state clashed with their self-reports of the mental states they are in.⁴

It is also not surprising that two distinct issues became conflated around this time (here is where we return to the question left hanging in n. 1). One issue is whether mental state types are identical with brain state types. Another issue is whether there is a law-like passage from the kind of brain state a subject is in to the kind of mental state they are in, and conversely. These issues are separate. Here is an example to make the point. There is no law-like passage from time of day to the colour of the sky at that time. The reason is that conditions vary too much from day to day, and this is true for any given location and time of year, let alone for different locations and different times of year. All the same, the colour of the sky at any given time is identical with some particular colour.

Looking back on this period, two thoughts come to mind. The first is to wonder why early supporters of arguing from correlations to mind–brain identities were not more worried by the possibility that creatures in similar or indeed identical mental states might differ significantly in what goes on in their brains. They knew, for example, about the way evolution shapes our brains. It always made good sense that the difference between evolving so as to be able to live in the sea and evolving so as to be able to live on land meant that it was very possible that what happens in our brains when we perceive what is happening around us is importantly different from what happens in the brains of dolphins when they perceive what is happening around them. In this regard, the simple examples of identities from science that figured so frequently in the discussions around that time, of which lightning and an electrical discharge, and the gene and the DNA molecule, were especially prominent, were perhaps unfortunate. They gave the discussions the wrong focus.

The second is to ask about the impact of the realization – a realization that largely post-dated early presentations of the identity theory – that there is an important distinction between what’s contingent and what’s a posteriori.⁵ Early presentations of the identity theory make much of the claim that the identities between mental states and brain states are contingent, like, it was said, the identity of lightning with electrical discharges, and the morning star and the evening star.

The reasons given for holding that the identities were contingent were, first, that the case for believing in them comes from science and not conceptual analysis, and, second, that we knew about the mind long before we knew about the brain. But these are, as we now know only too well, reasons for holding that any mind–brain identities are a posteriori; it is quite another question whether or not they are contingent. And, of course, some of the examples given of contingent identity are nowadays held by most philosophers to be examples of the necessary a posteriori. What is more, one can ask about the very coherence of the notion of contingent identity: isn’t everything necessarily self-identical? This leaves us with the question, should mind–brain identity theorists have said that the identities between mental states and brain states are a posteriori, rather than contingent?

Interestingly, it turns out, first, that there is no problem with possible variations, e.g., between species or for a given subject over time, in which brain states line up against which mental states; second, that the mind–brain theorists were right to claim that the identities are contingent; and, finally, that the notion of contingent identity, in the sense they need to give it, is perfectly coherent. But to see this, we need to note an important change in the way the case for the identity theory was made. Here the contributions of, e.g., Brian Medlin (Reference Medlin and Presley1967), David Armstrong (Reference Armstrong1968) and David Lewis (Reference Lewis1966) were especially important.

3 Realism about mental states and functionalism

The early versions of the identity theory, in the hands of Place and Smart,⁶ were directed at sensations and consciousness. The claim was that sensations and conscious experiences are identical to brain states. Place and Smart were behaviourists about, e.g., belief. They saw the identity theory as helping with the problem of accounting for the ‘raw feels’ aspect of the mind. The sensational side of psychology was to be accounted for by the fact that, e.g., pains and colour experiences are states of the brains whose nature is accessible to us – not as brain states of course, but in a way that accounts for the phenomenal nature of pain and colour experience. By contrast, Medlin, Armstrong and Lewis were realists about mental states in general (and Smart soon followed them).

They were realists about mental states in general because they held that it was a general truth about mental states that they play causal roles. Beliefs, every bit as much as pain, are caused and do some causing. The same goes for desires, decisions and intentions, depression, fear, hopes of glory and all the rest. What is more, what causes them and what they cause include other mental states as well as happenings around us and to our bodies. Perception, for example, causes belief and the forming of intentions to act: the perception as of a tiger may cause one to believe that there is a tiger and to form the intention to run at the first opportunity. What is more, different mental states work in tandem, the most obvious example being the way belief and desire combine to cause behaviour that tends to satisfy the desire if the belief is true. In short, they accepted a version of what we now call functionalism but was then called a central state or causal theory of mind.⁷

Nowadays, the idea that mental states play distinctive causal roles seems to many of us pretty much a commonplace, albeit a commonplace articulated in different ways by different philosophers of mind. However, in the days when the identity theory was being developed, the influence of Ryle (Reference Ryle1949), and arguably of Wittgenstein (Reference Wittgenstein1953), obscured what now seems obvious. Interestingly, Medlin’s exposition of the identity theory in his (Reference Medlin and Presley1967) is presented as a reaction to Ryle.

Once we have a functional picture of mental states in place, we can give an argument for the identity theory that makes no reference to correlations or Occam’s Razor, and this is what Medlin et al. did. Schematically it runs as follows:

1. Premise 1. Mental state M = the state that plays causal role C.

2. Premise 2. The state that plays causal role C = brain state B.

3. Conclusion. Mental state M = brain state B.

The argument is a simple application of the transitivity of identity. Premise 1 was offered as a kind of conceptual truth. This is the sense in which this way of arguing for the identity theory can be seen as a precursor to analytic functionalism, but for our purposes here what matters is that it is plausibly true that any given mental state has a distinctive causal profile, and we will be taking this for granted;⁸ what follows is, focused on what functionalists should say about the type–type identity theory. Premise 2 was offered as a discovery of neuroscience, and this also is something we will be taking for granted.

A question that might well be asked about premise 1 is whether it should, or should not, be read in a way that implies that the mental state a subject is in fails to supervene on how the subject is from the skin in. Is the kind of causal role in question one that implies that a difference in environment, unaccompanied by an internal difference, may make a difference in mental nature? However, the issue of narrow versus wide content was not an item on the agenda back then.

I said earlier that the argument for an identity theory from alleged mind–brain correlations leads inexorably to a type–type identity theory. What about the functionalist style of argument just given, the style of argument we can all agree is an improvement? Here the point is not so obvious. Mental state tokens play functional roles. This means we aren’t forced to read the mention of mental state M in the argument as the mention of a type. All the same, the right way to read the argument is as leading to a type–type identity theory.

The argument as stated has a type–type identity conclusion. One reason is that the word ‘state’ is most naturally read as a word for a kind. When the newspapers report on ‘the state of the economy’, they report on the way it currently is. However, the more philosophically important reason turns on the point that neither premise says that anyone is in fact in any specified mental or brain state. Saying what states play what roles doesn’t say anything about whether or not the roles are in fact being played. (The NAIRU in economics is the rate of unemployment that plays a certain role in balancing unemployment against inflation. Saying that it is such and such a rate of unemployment, 5 per cent or whatever, isn’t saying that any economy in fact has that rate of unemployment.) It might be thought a simple matter to rewrite the argument so that it is framed in terms of token states, and so as having premises and a conclusion that imply that the roles are in fact being played. However, this rewriting is not possible. The reason is the nature of the causal-functional roles played by mental states. On any at all plausible account of the roles that mark out one mental state from another, they pertain as much to relations to states that subjects are not in at the time in question, as they do to states subjects are in at that time. What gives a desire of mine a certain object and intensity is not just what it in fact does when I have it but what it would have done had my mental states been different in various ways. Its being a desire for, say, beer is in part a function of how, had I had a different belief from the one I in fact have about the location of the beer, I would have changed my behaviour in a way that brought me nearer to the changed location. Its being a desire for beer rather than for any kind of alcohol is in part a function of the fact that had I learned that there was wine available, I would still have reached for the beer. Its intensity is in part a function of the extent to which I would still have acted on the desire had I learned of this or that cost of my consuming beer. And so on.

Here is a simple example to reinforce the point. Suppose we have a gauge, G, wired to a tank, T, in such a way that the following identities are true:

1. G’s pointer being in the far left position = the state that indicates that T is empty;

2. G’s pointer being in the far right position = the state that indicates that T is full;

3. G’s pointer being in the 2/3 right position = the state that indicates that T is 2/3 full;

and so on.As stated, these are type-type identities. Now suppose that, at some time t, G’s pointer is in the 2/3 right position, then the following will be true:

However, this statement doesn’t say what it needs to say, namely, which feature of the token state of G’s pointer is the one that indicates that T is 2/3 full. The token state will have many features in addition to being in the 2/3 right position. Suppose, for example, that the token is a token of being at temperature K. This will mean that

is true. Moreover, we cannot recover the needed information by speaking solely in terms of tokens. What makes G’s pointer being in the 2/3 right position the feature of the token state that carries the information that T is 2/3 full concerns how a whole range of kinds of states of G and its pointer relate to how T is. What matters are the inter-connections between various kinds of states of the pointer and the tank – especially and obviously, those to do with causal co-variation relations between the kinds of states of the two items.

We can sum things up as follows. Any functionalist-inspired approach to the mind–brain identity theory needs to have a tripartite structure. One part gives the roles for each mental state, roles that will involve relations between mental states as well as roles linking mental states to the world around the subject and the subject’s responses to that world. A second part says that each role is occupied by the appropriate kind of brain state. The third part says what it takes to be in some given mental state M at t: it is to be in the state that occupies the M role at t, and that will, as a matter of fact, be being in the appropriate kind of brain state. The first two parts are about state types. Only the last is about tokens in the sense that it implies that a subject is in M at t just if a token state of the subject plays the M role at t.⁹

4 Back to our three questions

We can now answer the three questions we placed on notice.

First, does it matter that it is very likely that any correlations between mental states and brain states will not hold across species, may not hold across individuals that belong to some given species and may change over time for some given individual? Well, of course it matters for the version of the identity theory that rests on an argument from mind–brain correlations. But it doesn’t matter for the improved version lately described that rests on a broadly functionalist picture of mental states. The variations will simply be variations in what plays the roles. Indeed, the reasons for holding that certain variations are likely or possible are precisely the reasons for thinking that, in some cases, we should expect the same job to be done different ways, and that in other cases we should think it at least possible that the same job is done different ways. What is true, however, is that the rather bald statement of the argument for the identity theory we gave above needs to be embellished, as follows:

1. Premise 1. Mental state M = the state that plays causal role C.

2. Premise 2. In creatures of kind K, the state that plays causal role C = brain state B.

3. Conclusion. In creatures of kind K, mental state M = brain state B.

The message is that the type–type identities of the mind–brain identity theory should be restricted to creatures of kind K. The nature of this restriction depends on how the science turns out. Indeed, in the very unlikely event that the state that plays the role distinctive of any given mental state is the same for all creatures with minds, we can drop the restriction altogether. More plausibly, what will be true are statements like ‘In creatures of species S, the state that plays causal role C = brain state B’, leading to the conclusion ‘In creatures of species S, mental state M = brain state B’, or ‘In creatures of species S, at so and so stage of their development, the state that plays causal role C = brain state B’, leading to the conclusion ‘In creatures of species S, at so and so stage of their development, mental state M = brain state B.’

In this regard the situation parallels that for the gauge and tank we talked about earlier. We imagined the setup was such thatG ‘s pointer being in the far left position = the state that indicates that T is empty.But of course the set up might have been such that

1. G ‘s pointer being in the far right position = the state that indicates that T is empty.

If the first way of setting things up is S1 and the second way is S2, then we can say it this way

whereasIn set up of kind S2, G ‘s pointer being in the far right position = the state that indicates that T is empty.Moreover, maybe at some time in the future, supporters of the mind–brain identity theory will need to allow that it is false for certain computers or certain kinds of robots. For something like ‘In machines of kind J the state that plays causal role C= silicon state S’ is true, in which case identity theorists should conclude that, in machines of kind J, mental state M = silicon state S. And it may be that, sometime in the future, there exist machines of kind J. It is clear that anyone who approaches the identity theory from a functionalist perspective on mental states must allow the possibility (likelihood?) that (silicon) machines of the future will be in mental states.

We said earlier that affirming the identity of mental state M with brain state B implies nothing about whether or not some subject is in fact in mental state M. Any statement of the identity theory needs to have appended to it a clause about what it takes to actually be in mental state M at t. Now that we have noted the need to think in terms of possible restrictions of the type–type identities, we need to give the clause in a form that takes account of this. We do this when we say it as follows: a subject S is in mental state M at t if and only if (i) there is a state that plays the role distinctive of M in creatures of kind K, where S is of kind K at t, and (ii) S is in that state at t. The identity theorist can then add that, as a matter of fact, the states that play the role distinctive of M in any creatures of a kind to be found on earth up to now are one or another state of the brain. The theorist might well add that maybe in the twenty-third century things will be different. There will be ‘silicon’ states that play the role distinctive of M in some creatures that exist at that time. In this case the mind–brain identity theory will be false for these creatures; for them a silicon state will be identical with M.

Our second question on notice was whether or not mind–brain identity theorists should drop the insistence that the type–type identities they affirm are contingent. Should they say instead that they are a posteriori? The answer to that question is now clear. The causal-functional role played by a kind of state is a contingent property of that state. Indeed we have lately been reflecting on the fact that it is very likely that the kinds of roles distinctive of a given mental state will in fact be played by different kinds of states in different kinds of creatures, and maybe by silicon states in creatures of the future. This means that any identity between a mental state and a brain state can at best be contingent.

The last question on notice was the coherence of the notion of contingent identity. In one sense we have just answered this question. What the identity theorist is saying is that brain state kinds get to be the mental state kinds that they are because of what they do, and that what they do is a contingent property of them, the brain states. You may or may not believe this, but it isn’t in any way incoherent. It is akin to saying that what makes heart disease the biggest killer in our society is the role it plays, and that it is a contingent fact that it plays that role. However, it is true that sometimes (but only sometimes) the way identity theorists talked about the contingent identity of mental states and brain states invited the thought that they were committing themselves to the existence of types that were contingently identical to themselves. The better way to say it is that mental state types are not essentially the mental state types that they are in fact. Any mental state type, that very property, might not have been the mental type that it is. Belief is not essentially belief; pain is not essentially pain; and so on. Analogy: arsenic is in fact the chemical type which is Agatha Christie’s favourite poison in her stories; that is, the type–type identity ‘arsenic = Agatha Christie’s favourite poison’ is true. But arsenic isn’t essentially her favourite; that very type might not have been her favourite (and would not have been had she used cyanide more often in her stories).

5 Multiple realizability and the identity theory: first pass

It is time to address the famous multiple realizability objection to the type–type identity theory. I think it is important to distinguish two versions. The simple version argues that anyone with functionalist sympathies in the philosophy of mind must deny any putative identity between mental state types and brain state types. For functionalism tells us that mental state types and brain state types can vary independently. What matters for the mental state one is in turns on what one’s internal states are doing, not their nature qua brain states. Mental types are functional types.

We have in effect already replied to the simple version of the multiple realizability objection. Our reply is contained in what we said when we noted that the right way to frame the conclusion of the functionalist argument for the identity theory contains a restriction clause. The conclusion is ‘In creatures of kind K, mental state M = brain state B.’ M and B may vary independently but that no more shows that they aren’t identical in creatures of kind K than observing that the favourite poison varies between crime writers shows that it is false that, in Agatha Christie stories, the favourite poison is arsenic. Of course, the identities between mental states and brain states posited by identity theorists may – almost certainly do – vary as we go from one kind of creature to another, and the reason is, as we said before, essentially a functionalist one. On this way of looking at matters, identity theorists and functionalists are singing from the same songbook.

Identity theorists have said things like the immediately above before.¹⁰ And they have been puzzled that so many supporters of the multiple realizability objection to the type–type identity theory have failed to take the point. As one of those puzzled, I now think the reason for resistance is that proponents of the multiple realizability objection have really had in mind a rather different argument against the identity theory. They know perfectly well, I am sure, that a type like the poison that is the favourite of mystery writer X is multiply realizable, and yet, for any given value of X, it is identical with some particular chemical type. Their objection is best understood as one about the right question to ask when asking about the identity of mental types. In their view, the right question to ask is, ‘What unifies different creatures in mental state M, what’s the relevant commonality, what do creatures in M share that’s to the point?’ And the answer to that question, from a perspective sympathetic to functionalism, is, they urge, the functional roles occupied. This is the real reason that they want to say that mental types are functional types and not brain types.

If this is indeed the real argument behind the famous multiple realizability objection, the reply to it is that any functionalist approach to the mind has to acknowledge a fundamental distinction between being in pain and pain, between belief that snow is white and believing that snow is white and, in general, between mental state M and being in mental state M. The identity theory is a claim about mental state M, not about being in M. We spell this out in the next section.

6 Why functionalists have to acknowledge a fundamental distinction between M and being in M

At any given time, a subject S will typically be in many different mental states. These states will stand in various relationships to the subject’s surroundings and to each other, just as functionalists say. This means that one thing to say about S’s mental nature at t will be something of the form: S is in M₁, M₂ and M₁₈, and we need to distinguish between the functional role that S, the subject, plays, which will be the product of the roles played by M₁, M₂ and M₁₈, and other stuff besides, and the roles the states themselves play. This means that what makes it true that S is in, e.g., M₁₈ will not be the role S as such plays. It will be the role a state of S plays, namely the role M₁₈ plays. But of course it is S, not M₁₈, that is in M₁₈. How come? The answer is that for S to be in M₁₈ is for S to have a state playing the M₁₈ role. The subject isn’t playing the role, a state of the subject is.

We suggested above that proponents of the multiple realizability objection to the identity theory were drawing our attention to the question, ‘What is it that unifies all creatures in M?’ The answer, or the answer any functionalist has to give, is not the role of the subject as such plays, but the role a state of the subject plays. A subject has to have a state playing the M role, and what else might that state be other than M itself for a functionalist?

We can spell it out as follows. For S to be in mental state M is for one of S’s states to play the M role in S. A person in pain is a person one of whose states is playing the pain role in that person. A person who believes that snow is white is a person one of whose states is playing the belief that snow is white role in that person. And so on. Being in M is one thing; M is another. Having M is what it is to be in M.

We suggested above that proponents of the multiple realizability objection were focusing on the question of what unities all those in, say, pain. What we have seen is that it would be a mistake to say that what unifies them is a functional property of them, the subjects that are in pain. What unifies them is a functional property of a state they are in. What the type–type identity theorist says is that the state that plays the functional role is a state of the brain, not necessarily the same state for different creatures, or indeed for the same creature at different times.

7 How types can be inside us

We can now answer a worry that I know some have.

The type–type identity theory is a version of internalism. It says that mental states are literally inside us. This is offered as a contingent truth. Maybe sometime in the future we will ‘sub-contract out’ some of the key work our brains now do to especially efficient computers in a way that means that what goes on in them counts as part of one or another mental state. But, as things are currently, the key work is done inside us. Saying this is consistent with some versions of externalism – not, obviously, with versions that deny that mental states are inside us, but with versions that allow that, although mental states are inside us, their being such and such a mental state depends in part on how things are around the subject in ways that mean that internally identical subjects can be in different mental states. The worry is how to make sense of talk of mental states being inside a subject. ‘States’ in this essay means, as we have already said, state types, that is, properties, and properties as such aren’t located. It is their instances – the things that have the properties, the tokens – that are located. How then can supporters of the type–type identity theory be internalists about the location of mental states?¹¹

The distinction between M and being in M, between, e.g., pain and being in pain, shows us how to answer this question. Although it is John Doe who is in pain, not his brain, his brain is where his pain is located. How is this possible? Here is how. For John Doe to be in pain is for a part of John Doe to play the pain role and the part that is that way is a part of his brain. The property instance that, so to speak, locates the pain, is the part of the brain playing the role. The underlying idea that location comes down to location of tokens is correct, but the token in question isn’t the subject but rather a part of the subject’s brain. That’s how internalism turns out to be true on the identity theory.

I now turn to the second objection to the type–type version of the identity theory, the objection that it cannot give a satisfactory account of the causal role of mental properties.¹²

8 The causal role of mental properties

Suppose John Doe believes that a certain figure, F, is an ellipse. I invite him to reproduce the shape of F on a whiteboard. He accepts my invitation and draws an ellipse on the board. It is very plausible that a cause of the appearance of an ellipse on the whiteboard is his belief’s being that F is an ellipse. His belief’s having the property of being that F is an ellipse causes an ellipse to appear on the whiteboard.

In saying it this way, I am taking it for granted that we can sensibly talk of the property-like, of types, standing in causal relations – as in, e.g., ‘The sharpness of the bend caused the car to slide out of control’ and ‘The extent of the damage surprised us.’ This is to some extent controversial.¹³ Some say that causation only ever relates items (events, say) thought of as tokens, not as kinds. To say otherwise is to confuse the question of what stands in causal relations with questions about causal laws and causal generalizations. Here I will simply go with the majority (and common sense) and take it for granted that we can sensibly ask after the causal impact something has qua kind or type of thing or event that it is.

What, from the viewpoint of the identity theory, is going on in a case like the one just described? The first thing to say is that a token mental state, namely, a certain belief of Doe’s, is causing an ellipse to appear on the whiteboard. This token state will be a token brain state. That bit’s easy. The hard bit is what to say on the subject of which property of the token brain state is responsible. The obvious position for the identity theorist to take is that the property (the key one, I mean; obviously many will be involved, along with the properties of surrounding happenings, etc.) is the very brain state, i.e., brain state type, that the type–type identity theorist holds to be identical, in Doe at the time in question, to the belief that F is elliptical.¹⁴ The cause will be, that is, a neurological state. This makes good sense neurologically, and it is surely good news for the type–type identity theory that, in giving this answer, it is holding to the compelling idea that mental states cause what they cause in part because of the kinds of mental states that they are.

All the same, many feel that there is a problem with this answer on behalf of the identity theory. The concern is that even if the belief that F is elliptical – in Doe at the time in question, and maybe in Doe-like creatures more generally – is neurological state N, it remains true that what makes it the case that N is that belief in Doe is its functional nature, or something along those lines. Analogy: although arsenic is Agatha Christie’s favourite poison, what makes arsenic her favourite poison isn’t its being arsenic. It is her use of it in her novels.

But why, exactly, is this a problem? The worry seems to be that when Doe draws the ellipse, the way the mental cause presents itself to Doe will be in terms of what makes it the case that he believes F is elliptical. (It certainly isn’t presented to him as N.) It follows, seems to be the thought, that the cause isn’t N but rather the property that makes N the belief that F is elliptical, that is, its functional role.¹⁵

It is, however, a familiar fact that often our response to the impact of some given property on us, and the way the property presents itself to us, is governed by the state of our system at the time of the impact. Consider the following well-known examples. Doe looks tall to me. The property I am responding to and registering is his being tall. But the property actually impacting on my senses at the time is his being a certain height, H say. However, my system has stored in it, somehow or other, the fact that H is at the upper end of heights of people. In consequence, I see him as tall. It isn’t like a case where someone sends me a letter saying that Doe’s height is H and I then infer that he is tall using my knowledge of the normal range of heights. Or think of what happens when someone looks angry to me. What’s impacting on me at the time in question will be a certain facial configuration (and way of holding the body, but we’ll keep it simple). But it is a certain facial configuration impacting on a system primed to register anger when acted on by that kind of facial configuration. This is why I see the person as angry, and indeed the causally operative facial configuration may be, to an extent, unknown to me. Finally, consider the way a written sentence in English looks grammatical or ungrammatical to native speakers. The look is a causal response at any given time to a shape property of the sentence at that time. We know this because we know that the look at that time supervenes on the shape at that time. However, the sentence’s looking grammatical, or the opposite, is the result of the shape’s acting on a system primed to register that shape as playing the role of being grammatical or of being ungrammatical.

I think type–type identity theorists should say the same about Doe’s experience when his belief that F is an ellipse causes him to draw an ellipse on the whiteboard. N is doing the causing in the sense that it is the crucial property initiating Doe’s movements at the time in question. But his system has coded into it that N plays the role that makes N the belief that F is an ellipse. In consequence, the cause presents to him as that which plays that role. All the same, N – that is, N qua N – is doing the causing in the sense of being the causal trigger at the time. It is just like the grammar case.