The Twin-Earth Thought Experiments
Suppose that you had always had a physical twin, Chris, who on a different planet went through life having physical characteristics, sensory experiences, utterances, and brain processes exactly the same as yours in every physical and sensory respect. Chris’s perceptually accessible environment also was physically identical to yours. So throughout your and Chris’s lives you experienced or had identical visual fields, uttered identical words with identical intonations, saw qualitatively identical objects at the exact same times from the very same perspectives, etc. Surely everything in your and Chris’s respective and physically identical local environments seems precisely the same to you and Chris from the inside, from your own cognitive perspectives. It seems that you must have had the same thoughts too. When at age six Chris thought "I would like a cookie" you also had that thought. Of course, when Chris says that sentence the occurrence of ‘I’ picks out Chris, not you. But even so, complexities like that aside, it seems that you and Chris live the same cognitive life. Whenever you thought that walleye are freshwater fish, Chris thought that walleye are freshwater fish. Thus, it seems that one’s physical makeup determines one’s thoughts. Call this view individualism.
As reasonable as this is, it is easy to offer reflections that are comparably intuitive and yet suggest the very opposite: your physical makeup does not determine your thoughts, at least not exhaustively. As soon as you realize that thoughts have intentionality or content, individualism becomes suspect. For how can my thought that Florence is spectacular be determined solely by my physical makeup? My thought is about Florence, the distant city; how could the purely local, internal, or intrinsic physical characteristics of my body fix the contents of my thoughts when those contents but not those physical properties seem to reach out into the world, toward Florence? Perhaps we can make sense of a brain process, for instance, having "aboutness" (e.g., being about Florence) and counting as a thought that Florence is spectacular, but could we make sense of the idea that just the local, intrinsic properties of that brain process give it its aboutness? Perhaps the brain process has some extrinsic, historical or causal, physical properties that indirectly connect it to Florence and account for how it can be a thought about Florence. But it is difficult to see how the intrinsic physical properties could make such a connection—that seems contrary to the very meaning of ‘intrinsic’. When considered in this light individualism may look ridiculous, even magical, although in my opinion such an argument is not an adequate refutation. Still, this is enough to show that even when we limit ourselves to intuition, before proceeding to argument, we can see that individualism is by no means obvious.
In the 1970s Hilary Putnam introduced us to Twin-Earth (1975). As it was originally conceived, Twin-Earth looks, feels, sounds, smells, tastes, and in every practical way seems just like Earth, with the exception of a few details that are utterly irrelevant—irrelevant to everyday life, that is, but crucial to those nose-to-the-grindstone philosophers who try to construct excruciatingly exact characterizations of the thoughts of Twin-Earth’s inhabitants. According to one camp of these philosophers—the externalists or anti-individualists—Twin-Earthians have thoughts, beliefs, suspicions, and realizations that are different from those we on Earth have, even though this fact makes no practical or more importantly even physical difference in their lives. The same central point was urged by the closely parallel thought experiments of Tyler Burge (1979, 1982a, 1982b, 1986b). If these theorists are right, then we have an adequate refutation of individualism: since your Twin-Earth twin does not have the same thoughts as you, we have a case of two people with the same physical makeup but different thoughts. It is the task of this essay to describe several kinds of Twin-Earth thought experiments (section 1), the central argument generated from these thought experiments (section 1), the primary lessons drawn from the thought experiments (section 2), the most influential objections raised against them (section 3), and their most important alleged implications (section 4). In addition to going over the basics, in each section I will be presenting original arguments and theses; so there is plenty for both the novice and the expert. Both Putnam-style and Burge-style thought experiments are addressed.
1. The Thought Experiments
Here is an example of a Putnam-style Twin-Earth thought experiment. Words have their meanings contingently. The word ‘walleye’, for instance, could have referred to trout or sauger or maybe even nothing at all. As things actually stand, ‘walleye’ refers to walleye and my father Ron knows that walleye are freshwater fish. Ron has actually told me ‘Walleye are freshwater fish’. And I believe him based on his word. So as things actually stand I believe that walleye are freshwater fish.
Now suppose that in an imaginary world ‘walleye’ referred to swalleye and not walleye, where swalleye is some nonactual but possible kind of fish that is not a kind of walleye but is very similar to walleye. Suppose further that there were no walleye. So if in this imaginary scenario someone wrote in a fishing article ‘Walleye are freshwater fish’ they would be saying that swalleye are freshwater fish. Well, suppose there was that very remark in a fishing article. And suppose further that my father Ron read that article and believed what it said. (Since he did not do this in the actual world, he is physically different across worlds.) Then he would believe that swalleye are freshwater fish—although of course he would utter ‘Walleye are freshwater fish’. Let us suppose that in this imaginary situation I get my opinion from Ron telling me ‘Walleye are freshwater fish’, just like in the actual world. So I end up with the same belief as Ron; I believe that swalleye are freshwater fish.
None of this is anything but innocent. The principle behind our evaluation of the imaginary scenario is the following. Suppose
Let A be the conjunction of 1-4. The operative principle of the thought experiment is the claim that if A is true, then S’s assents to ‘Fa’ would have expressed her belief that Fb (Swalleye are freshwater fish). Thus, in the actual world S believes that Fa whereas in the counterfactual world she believes that Fb.
So in both worlds I hear the same sounds from Ron: ‘Walleye are freshwater fish’. But my actual and counterfactual beliefs are different (they are the same as Ron’s): in the imaginary world I believe that swalleye are freshwater fish whereas in the actual world I have the walleye belief. Since walleye are not swalleye, the propositional kind of content of the belief that walleye are freshwater fish is different from the propositional kind of content of the belief that swalleye are freshwater fish. And since those contents are distinct, the beliefs (belief types) themselves are distinct. Here it is assumed that in cases similar to the walleye one differences in propositional content mark differences in belief type.
The challenge to individualism comes in when we observe that the similarity of my intrinsic physical makeup in the two worlds has no bearing on the above reasoning. Obviously I need not ever eaten, seen, or touched either walleye or swalleye in either world. In both worlds my knowledge and other attitudes regarding the fish are second-hand. In particular, my beliefs ascribed with ‘Walleye are freshwater fish’ are acquired in physically identical situations of Ron uttering ‘Walleye are freshwater fish’. Furthermore, I acquired my concept corresponding to ‘walleye’ in the same scenario in each world: as a boy I listened to my father, mother, and uncle discuss their fishing trips on several occasions. In the actual world when they uttered ‘walleye’ they were talking about walleye; in the imaginary world they were talking about swalleye. The similarity of walleye and swalleye guarantees that all the same descriptions could have been used in each world. The concept I acquire corresponding to ‘walleye’ is their concept. One might want to protest that in each world I cannot without help distinguish walleye from swalleye. Under a natural interpretation that is true, but it is irrelevant. Although my relatives did not give me any means to differentiate the two kinds of fish I nevertheless acquired different beliefs in the two worlds—I acquired their beliefs. One might not like this practice, but that’s just the way belief attribution works.
So in the two worlds I am as physically unchanged as you please even though my beliefs have different propositional contents (i.e., truth conditions). If a difference in propositional content entails a difference in belief type, which is reasonable but can be questioned, then even though my physical makeup is the same in each world I have different beliefs (belief types). Thus, what happens in the mind isn’t exhaustively determined by what happens in the body; individualism is false.
There have been many objections to the above reasoning; I will address some of them in section 3. However, one of the objections is best dealt with here since it leads to alternative Twin-Earth thought experiments. Roughly put, the objection is that swalleye might be a kind of walleye (twater might be a kind of water, twalum might be a kind of aluminum, etc.). If so, then it is not clear that my beliefs or belief contents are distinct across worlds. I think that the objection fails, but even if it were sound it is not difficult to construct Twin-Earth stories that avoid the objection. Instead of using a hypothetical natural kind (swalleye) in the imaginary world, we can use an actual natural kind that we know is distinct from the natural kind used in the first, non-imaginary world. Walleye are Stizostedium vitreum and sauger are Stizostedion canadense; I take it that these are distinct but quite similar kinds of North American freshwater and game fish. Now suppose that in world W1 ‘walleye’ refers to walleye but there have never been any sauger; in world W2 ‘walleye’ refers to sauger but there have never been any walleye. In W1 ‘Walleye are freshwater fish’ means that walleye are freshwater fish; in W2 it means that sauger are freshwater fish. Clearly these are distinct beliefs and contents: up until last year I, for one, did not believe that sauger are freshwater fish (I had never heard of them) even though I did believe that walleye are freshwater fish. The rest of the thought experiment is not importantly different from the walleye-swalleye one.
Alternatively, and this point is not found in the Twin-Earth literature, the protagonist of the thought experiment can employ words for both natural kinds present in both worlds. In my judgment, these are the most convincing thought experiments. As things actually stand, ‘walleye’ refers to walleye, ‘sauger’ refers to sauger, and my sister Leslie insists that walleye get bigger than sauger. A fishing expert told her so during the summer of 1998. Later that year Leslie actually told me ‘Walleye get bigger than sauger’. And I believe her. So as things actually stand I believe that walleye get bigger than sauger. I do not believe that sauger get bigger than walleye. Now consider an imaginary world in which ‘walleye’ refers to sauger and ‘sauger’ refers to walleye. So ‘Walleye get bigger than sauger’ would mean that sauger get bigger than walleye. So if in this imaginary scenario someone wrote in a fishing article ‘Walleye get bigger than sauger’ they would be saying tha t sauger get bigger than walleye. Suppose that in that scenario my sister Leslie read that article and believed what it said. Then she would (incorrectly) believe that sauger get bigger than walleye—although of course she would utter ‘Walleye get bigger than sauger’. Finally, suppose that in this imaginary situation I get my opinion from Leslie telling me ‘Walleye get bigger than sauger’, just like in the actual world. So I end up with the same belief as Leslie; I believe that sauger get bigger than walleye. Obviously the content of this belief is distinct from that of my actual world belief that walleye get bigger than sauger. And of course the belief types are distinct as well: no one is going to argue that the belief that walleye get bigger than sauger is the same as the belief that sauger get bigger than walleye.
We can call the first thought experiment, in which we have swalleye, the science fiction story; the one in which we have walleye beliefs in one world and sauger beliefs in the other, the realistic story; and the one in which we have both walleye and sauger beliefs in both worlds, the dual-concept story.
It is often said that in Putnam-style thought experiments the difference in the actual and counterfactual beliefs gets traced to differences in the kinds (e.g., walleye and sauger) referred to by the relevant term (e.g., ‘walleye’). However, it appears that this is an accidental feature of the thought experiments given in the literature. The thought experiments may apply to empty, mythical, and fictional kinds just as well as they apply to natural kinds. For example, suppose that in world A it is not uncommonly thought that bigfoot exists. There also is, in A, a separate footbig story about a creature quite similar to bigfoot. In world A Alf’s friend who has concluded that bigfoot is real but footbig isn’t passes his opinion on to Alf so that in world A Alf comes to believe that bigfoot is real but footbig isn’t, and he expresses that belief with the sentence ‘Bigfoot is real but footbig isn’t’. In counterfactual situation B both bigfoot and footbig stories are p resent, but bigfoot is called ‘footbig’ and footbig is called ‘bigfoot’: the terms have switched. It is easy to imagine that in the situations in which in world A the bigfoot story started, the term ‘bigfoot’ could, in world B, have been replaced with ‘footbig’ (similarly for the beginning in world A of the footbig story). In world B Alf’s friend has come to the reverse conclusion that footbig is real but bigfoot isn’t—though of course he expresses that opinion with ‘Bigfoot is real but footbig isn’t’. In world B Alf gets his opinion from this footbig believer. So in B Alf ends up hearing and uttering ‘Bigfoot is real but footbig isn’t’—just as he did in world A. The details of the rest of the story (e.g., the bit about how the concepts are acquired) are similar to those given in the thought experiments given above. Thus, he actually believes that bigfoot is real but footbig isn’t whereas he counterfactually believes the reverse. It seems pretty solid that some people actually believe that bigfoot exists but the Loch Ness monster does not—even if neither creature exists. The bigfoot-footbig story is not importantly different.
Here is a Burge-style thought experiment. Bert is like many of us who have no first-hand or expert knowledge of arthritis. As a child he acquired his concept corresponding to ‘arthritis’ by being told something like the following over several occasions.
Arthritis is a crippling disease that usually shows up in elderly people’s hands, knees, and hips. It makes it painful to do things with one’s hands, etc. It prevents one from doing all sorts of activities that one usually takes for granted. It can be difficult to type, to write, to dress, to fish, and to climb stairs. It’s worse in high humidity. Aunt Jean had it in her left hand; Grandpa had it in both hands. Elizabeth’s cat has it in her hips.
Like many people Bert has no idea whether arthritis is a joint disease that must by definition be caused by calcium deposits. The answer is ‘no’, but Bert has a normal amount of ignorance about medical matters and so doesn’t know what to think about arthritis and calcium deposits. Even so, this lack of expert knowledge hardly prevents him from knowing that Aunt Jean had arthritis in her left hand. Now imagine a counterfactual situation in which Bert’s entire physical life is the same as it actually is. However, in the counterfactual world the medical community has explicitly defined ‘arthritis’ to apply to a smaller class of rheumatoid ailments, including all and only those joint inflammations caused by calcium deposits. Call this smaller class tharthritis. He hears all the same sentences containing ‘arthritis’ in both worlds, most of them true. He acquires his concept corresponding to ‘arthritis’ in the same situations in both worlds—the concept his teachers had and were passing on to him. The details about concept acquisition are those of the Putnam-style stories. In the counterfactual history Bert doesn’t believe that his father had arthritis—just as we, in the actual world, fail to have tharthritis beliefs. In the latter world he has no idea whether tharthritis is a joint disease that must by definition be caused by calcium deposits. He is under no misconception in either world, but like virtually everyone his knowledge concerning the definition corresponding to ‘arthritis’ is incomplete.
Burge’s kind of thought experiment differs from Putnam’s in that the difference in belief gets traced to a difference in the explicit definition of a word—a difference that the protagonist of the thought experiment is unaware of. Thus, it is often said that Burge-style thought experiments show that our beliefs depend not only on the makeup of the natural world (e.g., walleye or swalleye) but on the social or linguistic world (arthritis definition or tharthritis definition). (Although I will not broach the subject, Keith Donnellan has offered some thoughts on the relations between the two kinds of thought experiments (1993).) Burge’s thought experiments admit of some of the same variations as Putnam’s do. For instance, as I understand it a heart attack is defined as the death of heart-muscle cells from reduced or obstructed blood flow through the coronary arteries. These are usually caused by blockage of coronary arteries by a blood clot, spasm or contracture; or a severe disruption in the heart’s rhythm. A stroke is defined as a sudden decrease in the blood supply to part of the brain, damaging the area so it cannot function normally. Hardening of the arteries or high blood pressure are the usual causes of strokes. Most people, at least adolescents who have thus far been removed from people suffering strokes and heart attacks, are not aware of these definitions, although they know that both problems have to do with the heart and are quite serious. Jazmine knows that her Uncle Fred had a heart attack many years ago, that her great-grandmother Ester died of a stroke, etc. And on the basis of the testimony of her trusted friend Reggie, she believes that strokes are more serious than heart attacks. Now consider a world in which the terms ‘stroke’ and ‘heart attack’ have switched definitions. Suppose further that in that world Fred had not a heart attack but a stroke and Ester died of not a stroke but a heart attack. Finally, suppose that Reggie had come to the reverse opinion, that heart attacks are more serious than strokes—although of course she expressed her opinion with ‘Strokes are more serious than heart attacks’. So Jazmine hears all the same sentences involving ‘stroke’ and ‘heart attack’ in each world. But she obtains different beliefs in the two worlds (just like Reggie): in the actual world she believes that strokes are more serious than heart attacks whereas in the counterfactual world she believes that heart attacks are more serious than strokes.
It must be remarked that Burge has given yet another kind of thought experiment, one that does not depend on the protagonist’s lack of expert knowledge of a kind (e.g., of walleye) or (Burge claims) lack of expert knowledge of a meaning (e.g., of ‘arthritis’) but on her having a nonstandard theory regarding an item referred to by her belief (e.g., regarding sofas, as in his 1986b). I suspect that the primary point of this third kind of thought experiment is that if sound it expands the class of terms and beliefs subject to Twin-Earth thought experiments. I will not address this further kind of thought experiment here.
2. Anti-Individualistic Theses
Anti-individualism is usually roughly characterized as the denial of psychophysical supervenience: it isn’t true that one’s propositional attitudes are fixed by or supervene on one’s intrinsic physical qualities. This result is important because, among other things, it rules out most forms of functionalism for propositional attitudes. Very roughly, on these theories someone has a certain thought property just in case she has a network of intrinsic physical properties that satisfy a certain pattern. It isn’t the intrinsic physical properties but the pattern of properties that is claimed to be common to everyone who thinks that P. The thought experiments show that this is incorrect: the protagonist has all the same patterns of intrinsic physical properties in both the actual and counterfactual worlds of the thought experiments, but she has different beliefs nonetheless.
However, there are many ways to make the anti-supervenience idea precise, depending on all sorts of factors. Furthermore, there are anti-individualistic theses of a very different form, not adequately characterized as negations of supervenience. In this section I will attempt to sort out only some of the simpler issues.
The thought experiments can be run with two people on two planets in the same possible world: you and your Twin-Earth duplicate. In that case the thought experiments support the corresponding one-world anti-supervenience content thesis that there could be physical twins (intrinsically physically type-identical throughout their lives in the same possible world) with attitudes with distinct contents. If one accepts that the difference in content requires a difference in belief type—you and your physical duplicate not only have different belief contents but different beliefs—then we have the one-world anti-supervenience belief-type thesis that there could be physical twins with distinct attitude types. The thought experiments can also be run with one person in two possible worlds: Alf in the actual world and counterfactual world. In that case we get the two-world anti-supervenience content/belief-type theses that a person could, counterfactually, have distinct attitude contents/belief-types in the absence of any intrinsic physical difference in her body throughout her life. This holds for both Burge-style and Putnam-style stories.
The anti-supervenience theses are usually and with justification taken to be the immediate consequences of the thought experiments. Even so, if one is interested merely in metaphysical theories of thought and belief, then the anti-supervenience theses may be stronger than what is needed. Individualistic theories of the mind have a common consequence: if there is a certain cross-world sameness in intrinsic physical or functional types in one’s body, then there is cross-world sameness in the contents of one’s propositional attitudes. Obviously no individualist thinks only complete intrinsic physical sameness is required for sameness of content (what happens in one’s toenails need not be of any moment), so only a relevant physical similarity is demanded by the individualist. Thus, in order to refute those theories the protagonist of the thought experiments does not have to be completely physically identical across worlds; all we need is enough physical similarity to engage individualistic theories. This may seem to be an irrelevant complication, but arguments suggested by some remarks made by John Wallace and Gene Mason (1990) show that it is not clear how to work out these details regarding intrinsic physical identity across worlds.
Philosophers have also considered the thought experiments adequate grounds for various externalist theses, all of which are variations of the rough idea that one’s having a certain kind of belief requires that one’s social or physical environment satisfies certain conditions. Here are some specific externalist theses.
If in a possible world one has a mental state, then in that world there are contingently existing objects other than the agent (Putnam 1975, 220).
If in a possible world one believes that K is F (e.g., that water is wet), where K is a natural kind, then there are physical objects in that world (cf. Fritz Warfield 1998).
If in a possible world one believes that K is F (e.g., that water is wet), where K is a natural kind, then in that world natural kind K exists, but perhaps without any instances (cf. Cynthia MacDonald 1990, 387 and 1995, 156; Colin McGinn 1989, 7-8).
If in a possible world one believes that K is F (e.g., that water is wet), where K is a natural kind, then in that world one has had some causal contact with K or instances of the kinds that make up K (cf. Stephen Schiffer 1987, 36; McGinn 1989, 7; Ned Block 1991, 215; and Brian McLaughlin and Michael Tye 1998, 301).
If in a possible world one believes that K is F, where K is an existent natural kind, then in that world either there are instances of the kind, one has cohorts who have beliefs involving that notion of K, or one is not relatively ignorant and indifferent about the nature of K (Burge 1982, 116).
The central idea behind all these externalist theses seems to be the principle that (a) in order to have the belief that water is clear, for instance, one must have the concept of water (i.e., the ability to have de dicto ascribed water thoughts), (b) in order to have that concept of water one must obtain it, and (c) in order to obtain it one must be connected to the physical world in some one of a variety of ways. Typically, (i) one obtains a kind concept in something like the theoretical way scientists come up with concepts of natural kinds that for all they know have no instances; (ii) one gets it "naturally", in something like the way each of us presumably acquires the concepts of up, before, etc.; (iii) one gets it via someone else by teaching with or without the presence of instances (e.g., the concept of an electron); (iv) one gets it oneself by observing supposed instances of the kind and making a linguistic baptism (e.g., ‘Let’s call this shade of re d ‘Christmas stocking red’’), or (v) one gets it via some ordinary, non-theoretical description such as ‘the plant that would result from the breeding of these two plants’.
It has yet to be made clear how the externalist theses relate to either the anti-supervenience theses or even the thought experiments themselves. In fact, I have not understood any publication to have treated the topic. (However, Brian McLaughlin tells me he appreciates the problem.) Although the anti-supervenience theses are reasonably thought to be the immediate upshot of the thought experiments, there is reason to think that some externalist thesis is the principle behind the negation of supervenience. However, new investigation is required to properly address the relations among externalism, the negation of supervenience, and the thought experiments.
It is usually granted that in the realistic story the actual walleye-are-freshwater-fish and counterfactual sauger-are-freshwater-fish beliefs differ in one kind of content, call it linguistic content. Linguistic content may be thought of as propositional content or truth-conditional content. But it is often held that the beliefs share some other important kind of content, call it psychological content, that supervenes on intrinsic physical makeup and could be had by a solipsistic mind—so the anti-supervenience and externalist theses would all fail for psychological content. In this way theorists accept the central result of the anti-individualist’s reasoning but attempt to undercut its scope and importance. We can call this the dual content view since it posits two kinds of belief content. Although I will briefly describe this objection and its weaknesses, I will not examine it in any detail since it is consistent with anti-individualism. Afterwards I will discuss objections to anti-individualism itself.
There are two primary reasons one might find the dual-content objection attractive. First, one might think that any kind of scientific study of thought must, in order to be scientific, consider the thinker in isolation from her environment. One idea here (there are others) is that the kinds used in science have to be "individuated by causal powers", entailing that if two people’s thought tokens have the same causal powers, then they cannot fall under different scientific kinds (Jerry Fodor 1987, 1991; cf. Joseph Owens 1993, Robert Wilson 1995, and Burge 1989, 1995). Since you and your Twin-Earth twin are physically identical, it seems to follow that your thought tokens have the same causal powers: roughly, when put in the same contexts they will produce the same effects. So if those thought tokens are to fall under scientific psychological kinds, then they must fall under the same scientific kinds. Since they have different linguistic contents, these linguistic contents are scientifically defective and any serious science of thought will have to use psychological kinds other than linguistic contents. On this view the anti-individualist is correct in saying that ordinary belief contents and types are not determined by intrinsic physical qualities. But she is incorrect if she goes on to claim that theoretical psychology will appeal to non-individualistic psychological kinds.
These arguments have been found wanting, eventually by their most prominent advocate, Fodor (1994, xii). Furthermore, there seem to be counterexamples to the principle that physically identical objects must fall under the same scientific kinds: a moon might have not fallen under ‘moon’ even without any physical change if it counterfactually failed to orbit a planet. Finally, it appears to many that cognitive science is doing just fine appealing to belief contents that are non-individualistic (the ones treated in the thought experiments). But this is a complex issue belonging to the philosophy of science and cannot be resolved in a short compass.
Alternatively, one might think that reflections on not science but ordinary common psychological explanation show that there has to be a kind of individualistic content (Brian Loar 1988a, 1988b; cf. Sarah Patterson 1990, Bryan Frances 1999, Robert Stalnaker 1990). It is thought that even if some linguistic contents of our attitudes are not determined by our physical makeups, nevertheless people who are physical duplicates inside the skin must be psychologically the same when it comes to evaluating their rationality and explaining their actions and attitudes. If the only difference between me and my Twin-Earth twin is that absolutely everything I believe about and desire of walleye, he believes about and desires of sauger, and neither of us could distinguish the fish even if our lives depended on it, then in some interesting and philosophically relevant sense our beliefs about fish pack the same explanatory punch despite being of distinct belief types. Even though I but not my duplicate believe that walleye are freshwater fish—so we differ in linguistic content—this difference between us is somehow explanatorily irrelevant. After all, the line of reasoning continues, it is a datum that my duplicate and I conceive of the fish in precisely the same way, and in ordinary psychological explanation we are interested in how people conceive of what is around them. Since our beliefs are equivalent for the purposes of ordinary psychological explanation, even if anti-individualists are correct in thinking that our linguistic contents are distinct, our beliefs must be of the same explanatory type—the property somehow appealed to in or most relevant for ordinary psychological explanation. It isn’t enough, according to these theorists, to notice that there is some deep isomorphism between our thoughts; that is true enough but it seems to leave out the crucial fact that our conceptions are literally the same despite involving distinct linguistic contents.
There appear to be two problems with this approach. First, it has been very hard to articulate anything regarding this additional kind of content, often called narrow content. No theory of narrow content has ever come close to getting off the ground (Owens 1987 and 1990, Stalnaker 1990, Burge 1986a). Second, there is some reason to think that many of the intuitions that suggest the existence of narrow content can be accounted for with just the familiar non-individualistic linguistic contents (Frances 1999).
Other objections, ones that do not grant the truth of anti-individualism, are less convincing. I will very briefly treat ten of them here. Most target Burge’s thought experiments.
First, even if, due to Bert’s ignorance of arthritis, we cannot properly describe in some contexts his attitudes without the addition of descriptive psychological glosses (such as ‘but he doesn’t know its definition’), the argument does not seem to be thereby weakened—as Donald Davidson (1987, 449) appears to think. Although I think Davidson is right to imply that there are some insights to be gained from examining the use of glosses, the mere addition of explanatory glosses doesn’t on the face of it effect the conclusion that Bert lacks some beliefs in the counterfactual world that he has in the actual world.
Second, one might argue that the whole apparatus of belief types, tokens, contents, and propositions is irremediably flawed—so the debate over individualism is an ultimately incorrect though temporarily insightful product of wrong-headed philosophy. This response, well motivated as it sometimes is, seems insufficient since the thought experiments and theses can be formulated with no obvious reference to either belief types, tokens, concepts, contents, meanings, Fregean senses, functional roles, propositions, referents, extensions, intensions, or for that matter possible worlds. I have a belief, one that many have, that walleye are freshwater fish; I could have lacked this belief independently of how physically unchanged I was. This thesis does not appear philosophically theory laden. And it can be plausibly defended with the realistic and dual-concept stories given earlier—which need not make any reference to any exclusively philosophical concepts.
Third, finding answers to the question ‘In the counterfactual or actual situations does Bryan have such-and-such walleye beliefs?’ does not seem to rely on the implausible premise that there must always be an answer to the question whether or not someone believes that P: yes or no. Contrary to Wallace and Mason (1990, 194-195), the anti-individualist can advocate all the reasonable local indeterminacy she wants. The argument can deal with perfectly ordinary situations and beliefs (as in the walleye-sauger stories) for which there is no ground to doubt the ascription (unless, of course, one thinks all such belief ascriptions are false). Relatedly, with many others Dennett remains under the impression that like many other philosophical thought experiments the Burge-Putnam ones must be in the realm of philosophical fantasy (1996). He is correct in that some have been (some of the science fiction ones), but his criticism fails since the generalization fails.
Fourth, the thought experiments do not run afoul of any current theories of belief ascription. For instance, on any plausible theory of belief ascription the beliefs that walleye get bigger than sauger and that sauger get bigger than walleye have distinct propositional contents. The same holds for the Burge-style story involving heart attacks and strokes. One can try to argue for the existence of belief contents that are untouched by the thought experiments, but that threatens the scope, not truth, of anti-individualism.
Fifth, some theorists have complained that Burge-style thought experiments fail because the protagonist is confused about the notion in question (e.g., arthritis). However, no confusion or misconception whatsoever is required; all that is necessary is some lack of specialized knowledge. In both worlds Bert is unsure about the definition of ‘arthritis’; he is not making any mistakes, conceptual, linguistic, or otherwise. I am willing to bet that many thousands of people have arthritis beliefs and yet do not know whether it is exclusively a joint disease caused by calcium deposits. Even so, many of them know perfectly well that arthritis is painful. Once again, that just seems to be a fact about how belief attribution works.
Sixth, it has been said that ‘tharthritis’ is an empty term, so won’t that cause problems? But it is not empty: it includes all and only those joint inflammations caused by calcium deposits. Many of us who have arthritis also have tharthritis.
Seventh, philosophers have given compelling reasons for thinking that Bert really has the same concept corresponding to ‘arthritis’ across worlds. Talk about concepts is exceedingly slippery; so perhaps the claim is true. However, in order to count against Burge’s conclusion that there is a difference in belief across worlds one must hold that either Bert doesn’t have arthritis beliefs in either world or he has arthritis beliefs in both worlds. Thus, one has to show, contrary to Burge, that Bert either doesn’t have arthritis beliefs in the actual world or he has arthritis beliefs in the counterfactual world. The first option looks to be a nonstarter; it seems that people are actually in Bert’s situation. The second option looks implausible as well. If it were sound, then we would have to claim, by symmetry, that most of us (who do not have any opinion on the correct definition of ‘arthritis’) actually believe that tharthritis is painful. And of course one can multiply belief types as one pleases here, with as many belief types as thought experiments. Thus, poor Bert ends up with who knows how many hundreds of distinct beliefs each ascribed with his use of ‘Arthritis is painful’. In addition, we have already seen a plausible dual-concepts Burge-style story—the Jazmine case. In the actual world she believes that strokes are more serious than heart attacks whereas in the counterfactual world she believes that heart attacks are more serious than strokes. One can theorize about concepts until the cows come home and it will not change the fact that these are different beliefs with different contents (in one important sense of ‘content’). Furthermore, it seems that since she expresses these beliefs with the same sentence, ‘Strokes are more serious than heart attacks’, her concept corresponding to ‘stroke’ differs across worlds. A similar point holds for her concept corresponding to ‘heart attack’. If her concept corresponding to ‘stroke’ (or ‘heart attack’) were the same across worlds, as this objection has it, then it would seem that she would have the same belief across worlds. But this seems false. This suggests that the objection that Bert has the same concept in both worlds corresponding to ‘arthritis’ is false as well. This is not to say that there is no sense of ‘concept’ for which Bert’s use of ‘arthritis’ expresses the same concept in each world. Instead, it is to say that there is a sense of ‘concept’ for which Bert’s use of ‘arthritis’ expresses different concepts in each world.
Eighth, one might hold that even though Bert actually believes something (that arthritis is painful) that he counterfactually fails to believe, his beliefs really do not differ in content. It is important to note that this stronger than saying that his actual and counterfactual beliefs have a content in common. An anti-individualist can admit that Bert’s actual and counterfactual beliefs share a kind of content—narrow content—while insisting that they differ in the ordinary kind of content ascribed by ‘that’-clauses. In order for the objection to be against anti-individualism, it must conclude that Bert’s beliefs do not differ in ‘that’-clause content. Thus, his actual arthritis-is-painful and counterfactual tharthritis-is-painful beliefs do not differ in ‘that’-clause content. I think this objection fails for the arthritis case, but it appears we can simply sidestep it with the stroke-heart attack story.
Ninth, one might hold that even though Bert actually believes something that he counterfactually fails to believe, his beliefs really do not differ in content since they are de re, not de dicto. In the actual world Bert believes of arthritis that it is painful; in the counterfactual world he believes of tharthritis that it is painful. These de re beliefs do not differ in content. In a similar vein, it is said that Burge is incorrect in thinking that ‘arthritis’ has oblique occurrence in ‘Bert believes that arthritis is painful’. I don’t know of any compelling reason for believing either of these objections, but here are two considerations that count against them. First, pretend that ‘arthritis’ has some medical term as a synonym, say ‘calarthritis’. On the basis of the structure of the term itself, Bert believes that calarthritis is always caused by calcium deposits (he thinks the ‘cal’ part refers to calcium deposits). But he insists that he does not believe that arthritis is always caused by calcium deposits (he says that he remains agnostic). Aside from the counterintuitive Millian considerations advanced by Nathan Salmon and others (myself included) there is little reason to think that Bert believes arthritis is always caused by calcium deposits. So he believes that calarthritis is always caused by calcium deposits, but he does not believe that arthritis is always caused by calcium deposits. If this is right, then ‘arthritis’ has oblique occurrence in ‘Bert believes that arthritis is painful’. And if it has oblique occurrence, then it is not a traditional de re ascription.
Second, just as Putnam-style thought experiments can be run on empty kind terms (‘bigfoot’), with some creativity perhaps the same can be done with Burge-style stories. Since the de re belief that K is F entails the existence of K (at least on the standard views regarding de re belief), the protagonist’s beliefs cannot be construed as de re.
Tenth, some have been tempted with the idea that ‘walleye’ or ‘arthritis’ is indexical in some way that prevents the protagonist of the story from having different beliefs, at least in one sense. Even though I actually believe that walleye are freshwater fish and counterfactually believe that sauger are freshwater fish, despite having different truth conditions my beliefs share a common character in the same way that our beliefs corresponding to our utterances of ‘I am a philosopher’ have a common character even though they have different truth conditions (mine being true if I am a philosopher, yours being true if you are a philosopher). We do have an understanding of what this character is for words like ‘I’, ‘here’, and ‘now’; that is, we have rules for knowing how to generate truth conditions for utterances involving these indexicals. The rule for ‘I’ is something like this: the occurrence of ‘I’ refers to the linguistic agent who used it. But what is the character for ‘walleye’ or ‘stroke’? Is the rule for ‘walleye’ something like ‘the occurrence of ‘walleye’ refers to whatever satisfies description D relative to the context of the occurrence’? Then we can do Twin-Earth story on D to show that it does not supervene. In any case, this objection has yet to have been developed sufficiently to challenge anti-individualism (Almog 1981, Owens 1999).
4. Putative Implications of Anti-Individualism
As noted above, the immediate practical consequence of anti-individualism is the falsehood of functionalist theories that entail that physically identical individuals must share the propositional contents of their thoughts (i.e., functionalist theories formulated and endorsed in the ‘60s, ‘70s, and most of the ‘80s). Many leading theorists now accept this consequence and have tried to formulate theories that are consistent with anti-individualism.
However, externalist and anti-individualist theses have been used to generate even more ambitious refutations, ones that are absolutely stunning. They are said to refute physicalism, scepticism, authoritative self-knowledge, and the causal construal of psychological explanation. In this section I will indicate how these arguments go and what some of their weaknesses may be.
Owens has argued that under reasonable assumptions the two-world anti-supervenience content thesis refutes the causal construal of psychological explanation. Although there are several ways to make Owens’ argument precise, when applied to the water-twater thought experiment (in the actual world I believe that water is clear; in the counterfactual world I believe that twater—water’s Twin-Earth double—is clear) it is perhaps best formulated as follows (1993, 249-250).
Since the argument is valid and 7 and 9 are contradictory, at least one of the premises is false. The premises are 1 (two-world anti-supervenience content thesis), 2 (content individuation), 8 (causal supervenience), 6 (which is supposed to follow from the causal construal of psychological explanation), and 3 and 5 (which are innocent). Owens argues, in part, that since anti-individualism (1) and causal supervenience (8) are true, we have to reject either the causal construal of psychological explanation (6) or the principle that propositional attitude tokens are content individuated (2). He concludes that since these latter two principles are the backbone of our model of folk psychology, that model is mistaken. And if that model is mistaken, then cognitive science is built on a mistaken model.
However, even bigger consequences await us. As Burge points out the conjunction of the two-world anti-supervenience content thesis and content individuation—without 6, the causal construal of folk psychology—rules out some forms of physicalism. It follows from premises 1 & 2 that the water and twater tokens are distinct. But what are held constant across worlds in the thought experiments are not merely Alf’s intrinsic physical qualities, but the very stuff he is made of: Alf is physically token identical across worlds (Burge 1979, 111). Obviously, if the water token were numerically identical with some physical token P of Alf’s body, then since Alf is physically token identical across worlds the water token would have to exist in his body in the counterfactual world (since P exists there in that world). However, it is easy to see that the water token does not exist in the counterfactual world. By content individuation in order for it to exist in the counterfactual world it would have to have the water content in that world. That is, Alf would have to believe, in the counterfactual world, that water is clear. But he does not. Thus, the water token does not exist in the counterfactual world even though P does. So the water token is distinct from P. So the token-identity theory is false provided realism about attitude tokens is true. Additional argument inconclusively suggests that even weaker forms of physicalism would have to be false as well.
I will offer just one comment on these two arguments. It is not clear that we should accept content individuation, and if we reject it both refutations fail. One reason Burge accepts it is that the system of belief contents provides the only way to generate systematic understanding, description, and explanation of intentional mental tokens (1993, 110). For instance, we have no network of neurophysiological descriptions that performs these tasks. Burge’s comment is true if we confine our task to understanding the cognitive relations among mental tokens: only the system of content properties buys us any understanding. However, the system of economic properties provides the only way to generate systematic understanding, description, and explanation of the monetary properties of economic tokens, and we do not think that economic tokens (e.g., dollar bills) have their economic properties necessarily. Furthermore, if individualistic psychological content exists, then perhaps thought tokens have those contents necessarily while having their linguistic contents contingently. If so, then premise 1 is true for linguistic content but false for psychological content; and premise 2 is false for linguistic content but true for psychological content. Either way, we fail to generate a sound refutation of token identity (or the causal construal of psychological explanation or causal supervenience). In any case, it has yet to be determined what we should conclude about these two provocative arguments.
The alleged epistemological consequences of externalism are just as significant. Consider the externalist principle that necessarily, if someone believes that water is clear, then there is a physical world satisfying certain empirical conditions—conditions that allowed one to acquire the concept of water. Some philosophers think that if the thought experiments prove anything philosophically interesting, then they prove this modal principle a priori. After all, the arguments use mere philosophical thought experiments and the modal principle makes a claim about necessity. Of course if one can know the modal principle a priori, then one can know a priori the corresponding nonmodal principle that just leaves off the bit about necessity: if as a matter of actual fact someone believes that water is clear, then there is a physical world satisfying certain empirical conditions. Now if one knows a priori that if one thinks that water is clear then there is a physical world satisfying certain empirical conditions, then if one knows a priori that one thinks that water is clear, then one should be able to deduce a priori by modus ponens that there is a physical world and even that one’s physical world satisfies certain empirical conditions. Thus, if one had a priori knowledge of both the modal dependency principle and one’s thoughts, then one would be able to deduce from a priori premises nontrivial empirical facts thereby generating a stunning refutation of certain forms of scepticism. Alternatively, if one thinks that scepticism is tenacious, so it cannot be refuted in this way (while allowing that it might be refutable in another way), then one might conclude that the thought experiments show that we do not have privileged knowledge of the contents of our thoughts. Thus, it appears that if externalist theses’ truth requires their a priority, then a firmly established philosophical view will have to be rejected: either externalism, the tenacity of scepticism, or the existence of first-person authority (Michael McKinsey 1991, Warfield 1998, Paul Boghossian 1990).
A problem here is that it is not clear that externalist theses are knowable a priori—at least through the Twin-Earth thought experiments, which is all that matters here. One reason for this lack of clarity is that for most externalist theses it is highly controversial whether they are even true (and of course if they aren’t true, then they aren’t knowable). For instance, the thesis that necessarily if one thinks that natural kind K is F, then there must be K around, seems false: laypeople read popular science books that describe exotic elements or compounds that may never have any instances. It isn’t easy protecting externalist theses from clever counterexamples, even theses that are more sophisticated. Another reason for thinking externalist theses are not knowable a priori via the thought experiments is that even though these theses are the product of armchair speculation, that does not mean that they are a priori. Once again, the issue remains open and actively pursued.
Another set of arguments concerning the alleged epistemological implications of the thought experiments starts as follows (Boghossian 1990, Burge 1988 and 1998, Owens and Kevin Falvey 1994, Tony Brueckner 1997). Suppose that on his fifteenth birthday Oscar is unknowingly taken away from Earth with its concept of aluminum and shuttled to Twin-Earth with its concept of twalum, a distinct but superficially similar alloy. Before the move Oscar believed that pots are made of aluminum, but does he have aluminum or twalum beliefs at age thirty after spending the last fifteen years on Twin-Earth? At age thirty Oscar will utter in all sincerity ‘Fifteen years ago I believed that pots are made of aluminum; this is the same belief I have today, that pots are made of aluminum’. According to one possible view, thirty-year old Oscar no longer has an aluminum concept and thus has said that fifteen years ago he believed that pots are made of twalum; that that old belief is the same as the on e he has today, that pots are made of twalum. Further, at age thirty Oscar cannot (without more information) even think the thoughts he had when he was fifteen—even though it seems odd to say that he has forgotten anything. And on this interpretation Oscar is wrong: he says, at age thirty, that fifteen years ago he believed that pots are made of twalum. But he didn’t have that belief at age fifteen; at that age he had aluminum beliefs. Some find it highly dubious to charge Oscar with such a mistake. Alternatively, one might hold that when Oscar utters at age thirty ‘Fifteen years ago I believed that pots are made of aluminum; this is the same belief I have today, that pots are made of aluminum’ he says that fifteen years ago he believed that pots are made of aluminum and that this is the same belief he has today, that pots are made of twalum. In this case we allow him to retain his aluminum concept and express it with ‘aluminum’ on some occasions depending on th e context of utterance. But now we have the oddity that he thinks that his past and present beliefs are the same when in fact they are not by assumption.
The issues here are even more complex and have yet to be settled. However, we can make a worthwhile and original comment on the switching arguments as a class.
The truth of nearly universally held intuitions regarding these switching cases entails the two-world anti-supervenience content thesis. Here’s why. In world A Oscar is whisked from Earth with water to Twin-EarthA, which has twater. In world B Oscar is whisked off to Twin-EarthB, which is exactly like Earth in every way (e.g., it has water (not twater), our concept of water, ‘water’ refers to water, etc.). Since Oscar can be intrinsically physically identical in A and B, supervenience entails that there is no cross-world difference in belief. Since he obviously retains his water beliefs in world B (this being in principle no different from him moving from a community in Indiana to a physically identical one in Illinois, the languages the same in every respect), he must do so in world A as well. So according to supervenience (individualism) Oscar never acquires a twalum concept. Even philosophers such as Boghossian, who has had the most influence using switching cases to cast doubt on anti-individualism, think that eventually Oscar will acquire the twater concept on Twin-EarthA whether or not he retains his water concept (1994, p. 38). If these intuitions are correct, then of course anti-supervenience is true. So the argument does not seems to target anti-individualism. Instead, it poses a problem for all theories of authoritative self-knowledge.