1 Swampman has been cited/ sighted in several places but the treatment of him has been generally dismissive. Thus Donald Davidson writes: ‘[The Swampman] can’t mean what I do by the word “house,” for example, since the sound “house” it makes was not learned in a context that would give it the right meaning or any meaning at all. Indeed, I don’t see how my replica can be said to mean anything by the sounds it makes, nor to have any thoughts’ (‘Knowing One’s Own Mind,’ Proceedings and Addresses of the American Philosophical Association 60 [1987] 441-58, at 444).

Other philosophers, such as Dretske, Fred (Explaining Behavior [Cambridge, MA: MIT Press 1988], 98Google Scholar) and Millikan, Ruth (Language, Thought and Other Biological Categories [Cambridge, MA: MIT Press 1984], 93; 337-8Google Scholar) concur with the idea that only by reference to a creature’s history are we justified in saying that and what the creature means; hence a newly created double will just have to wait for his or her representational skills. Advocates of narrow psychological content have not challenged this assumption. Stich, Stephen (From Folk Psychology to Cognitive Science [Cambridge, MA: MIT Press 1983]Google Scholar), for example, holds that context and history are determining factors in the attribution of broad psychological states but denies that the latter are the true subject matter of a scientific psychology.

2 The Hubots and Rumans are hybrid robotic evolved creatures from Millikan (Language, Thought and Other Biological Categories, 127ff.).

3 H-particles are Fodor’s, Jerry invention (Psychosemantics [Cambridge, MA: MIT Press 1987], 33ff.)CrossRefGoogle Scholar. He asks us to imagine that he has stipulated that every particle in the universe is an H-particle or a T-particle, depending on whether he flips his gen-u-ine United States dime heads up or tails up. But of course, he concludes, physics would be crazy to take such relational properties of particles into account since they do not affect the causal properties of particles. Similarly, he claims, psychologists can safely ignore the relational properties of psychological states.

4 Twin briskets and contracts are examples from Burge, Tyler (‘Individualism and the Mental,’ in French, Peter Uehling, Theodore E. Jr. and Wettstein, Howard K. eds., Midwest Studies in Philosophy, Volume 4 [Minneapolis: University of Minnesota Press 1979] 73-121)Google Scholar.

5 The question is: Can we assume from the physical type identity of two creatures that they are syntactic twins whilst denying in the case of one the possibility of a semantics for its protolinguistic output? The answer depends on whether syntax is independent of semantics, and, in linguistic circles at least, that is an open question. A competent speaker must be capable of generating an utterance of ‘Visiting philosophers can be boring’ from either of two deep structures. But which one is actually realized depends on the speaker’s linguistic intentions—on how he or she intends the sentence to be interpreted.

There is a further question about whether one can consistently be an extemalist about semantics and an individualist about syntax. After all, if one is committed to the view that meaning is determined by socio-linguistic factors, why not suppose that syntactic structures are also partly conventional? For example, if the meaning of ‘the doctor gave the arthritis patient the water’ is not in my head, why suppose the correct parsing is? See also Seager, ‘Thought and Syntax’ (ms., University of Toronto 1989).

6 It also begs the question to assume intentionality does or would make a difference to the Swampman’s behavior. I take up the question of whether we can dispense with intentionality in explaining the Swampman’s behavior in section V.

7 This line is sponsored by Uncle Arthur (Ripstein’s) Supervenience Stores, Pty. Ltd.

8 Two-bitsers are available from Dennett, Daniel (The Intentional Stance [Cambridge, MA: MIT Press 1987])Google Scholar.

9 Perhaps I should emphasize that it is crucial to the story and to the philosophical issues I’m raising that we suppose that there is no causal connection between Swampman and Raoul. If there were, then Swampman might inherit Raoul’s intentionality and personal identity.

10 I thank Andre Gallois for this suggestion and related comments.

11 Experiences which are caused by external events and those which are internally caused could be said to involve different brain processes or different modules (see Fodor’s The Modularity of Mind [Cambridge, MA: MIT Press 1983]). But it is questionable whether identification of a perceptual state could proceed via self-identification of its modular function. No one argues that the latter would be introspectively accessible.

12 I thank an anonymous referee for this journal for pointing out this response.

13 Kaplan’s distinction between character and content is helpful in relation to this point. Suppose that two men in green sweaters are sitting on a sofa when I make reference to ‘the man in the green sweater’ and suppose that, due to some malfunction in my visual system, I cannot discriminate between them. An interpreter may not be able to fix any determinate content to my utterance. But in a perfectly good sense we both know my utterance is meaningful. We know what, for example, the expression could refer to. In Kaplan’s terms, my utterance would have character or type-meaning but no determinate content. See Kaplan, David ‘Demonstratives,’ in Almog, Joseph Perry, John and Wettstein, Howard eds., Themes from Kaplan (New York: Oxford University Press 1989)Google Scholar.

14 See section II and n. 5.

15 See Sachs, Jacqueline Strunk ‘Recognition Memory for Syntactic and Semantic Aspects of Connected Discourse,’ Perception and Psychophysics 2 (1967) 437-42CrossRefGoogle Scholar; Johnson-Laird, Philip N. and Stevenson, Rosemary ‘Memory for Syntax,’ Nature 227 (1970) 412CrossRefGoogle ScholarPubMed; and, for a short summary of this literature, Garnham, Alan Psycholinguistics: Central Topics (London: Routledge 1985), 138-41Google Scholar. Experiments carried out suggest that subjects only remember syntactic details by ‘rehearsing’ them, when, for example, they anticipate a memory test.

16 See Haugeland, John ‘Weak Supervenience,’ American Philosophical Quarterly 19 (1982) 93-103Google Scholar.

17 I thank Ronald de Sousa for pointing this out to me.

18 Thanks to Kim Lycos for drawing my attention here to dispositions.

19 This is close to Davidson’s general picture of the preconditions for interpretation and, hence, meaning. There must be correlations between words and external causes for meaning to be possible since it is such correlations which interpreters rely on. See Davidson’s ‘Knowing One’s Own Mind’ and ‘Epistemology Externalized’ (ms., University of California, Berkeley 1990).

20 See, for example, Fodor’s (Psychosemantics, ch. 4; A Theory of Content [Cambridge, MA: MIT Press 1990], ch. 3 and 4).

21 Davidson, in ‘A Nice Derangement of Epitaphs,’ in Lepore, Ernie ed., Truth and Interpretation: Perspectives on the Philosophy of Donald Davidson (Oxford: Basil Black-well 1986) 433-46Google Scholar, argues in relation to the case of Mrs. Malaprop that if non-standard uses of words conform to regularities of their own then they are not misapplications or errors. Rather, they just have a non-standard meaning in the person’s idiolect. This is not to say (as pointed out to me by a referee for this journal) that a view like Davidson’s cannot accommodate regular misapplications. We might have holistic reasons for thinking a person regularly misapplies a word. For example, we might decide that a person who uses the word ‘truth’ according to standard English, understands how the suffix ‘fully’ works in English, and yet uses ‘truthfully’ non-standardly, is better understood as misapplying the word ‘truthfully,’ since to interpret this word as having a different meaning would jeopardize, beyond reason, our interpretations of the speaker’s other words, such as ‘truth.’ But it remains true that, for Davidson, provided there are no overriding holistic considerations, regularities in the words-to-world relationship is constitutive of meaning and in that domain, no regular misapplications are possible. I regard that as a questionable tenet of externalism for the reasons given above.

22 Millikan (in Language, Thought and Other Biological Categories; ‘Biosemantics,’ Journal of Philosophy 86 [1989]281-97) is often keen to stress the problems of trying to build a semantics on the basis of statistical correlations between representations and their external causes. Her focus, however, is much more specific. She wants to accommodate the fact that structures internal to organisms have been selected by evolution for their representational functions (e.g., prey, predator, food, rival- or mate-detecting functions) even where there is no strong correlation between the tokening of those structures and the state of affairs the structure is supposed to represent.

23 What makes the externalist position so attractive is the intuition that until we can say what a Swampman’s psychological states are about in the world we have no reason to suppose they have any content and what, given the absence of causal connections, could they be about? I don’t want to discount that intuition — it is compelling and, believe me, any sequel to Swampman (Designator II: The Swampman Returneth?), is not going to get away with such a low budget. But I don’t feel that I must address that issue in this context. Here I am only concerned to point out some difficulties facing externalists when they take on the Swampman. I make no claims, yet, to have developed a semantics for a Swampman, only to be showing that there must be one.

24 This is the ‘crude’ version of the causal theory Fodor (in Psychosemantics) entertains but moves beyond. I take up a less crude version below.

25 I am not suggesting that even the crude causal theory would be so crude as to suppose that being a murder was a primitive physical property or ‘murder’ a simple concept. Presumably the causal dependency of the Mentalese symbol type MURDER on murders is mediated by the tokening of simpler symbols and inferences. See Fodor’s ‘demure foundationalism’ in Psychosemantics (118ff.). The bottom line, nevertheless, is that concepts are causally dependent on the properties they denote.

26 Appealing to the functional role of indicators within the system as a second factor which determines what MURDER should indicate is not an option, for that is something that, on this model, the indicator acquires during the learning period. The first MURDER has no history of serving the system and hence no function. See Dretske’s Explaining Behavior, 98.

27 Think of the innumerable ways in which one may commit a murder. What physical properties do all murders share which explains why they have the same effect on the psychological states of observers?

28 I thank Calvin Normore for helpful discussion on these issues.

29 One may feel that the case I’m defending here is too strong. After all, what if we just built a machine which could make these kinds of discriminations? Wouldn’t we then have a purely physicalist-reliabilist description of the machine’s actions? Of course, we could, in principle, build a machine like us by cloning or arranging cells in a human-like pattern without having the foggiest idea how to non-intentionally describe how such ‘machines’ succeed in complex categorization. But the real issue is whether or not we can build a Turing machine which is capable of simulating all of our behavior. Andre Kukla pointed out to me that if I deny this I start to run afoul of Church’s thesis and that that would be an unpopular move on my part. Church’s thesis is that any function that can be computed is primitive recursive— i.e., can be computed by a Turing machine. So if our discriminatory capacities are rule-governed or computational then we ought to be able, in principle, to build a Turing machine which computes the same functions. And it would follow that there would be an appropriate non-intentional description of the machine’s human-like behavior.

The question of our machinehood is a fascinating topic, too large for this context. My intuitions tell me that we will never really answer this question until we answer the question: How much of our behavior is computational? Some of our behavior seems quite rule-governed (e.g., playing chess) but some of it seems much more ‘open-textured,’ to use one of Wittgenstein’s terms (e.g., discriminating between first and second degree murder). If my intuitions are correct, Church’s thesis is left standing but may turn out to be somewhat restricted in its domain of application.

30 One view I don’t discuss here, and I’m grateful to Christopher Gauker and Bernie Kobes, among others, for pointing out that I should, is instrumentalism. Part of the reason why I haven’t discussed it is that I don’t see it as a threat to the view I want to defend and that is that we should have a uniform explanation of our behavior and that of the Swampman. Swampman passes all the verbal and non-verbal Turing tests for having a mind. On instrumentalist criteria, his interpretability is on a par with our own.

31 Jack Macintosh suggested to me that all of Swampman’s problems could be solved by regarding him as Raoul’s closest continuant. In that case Raoul didn’t die; he just, as Macintosh put it, ‘got a close shave.’ Tempting, but I’ll have to say no. Had Raoul lived, Swamp man wouldn’t have been Raoul and the problem of his (Swampman’s) intentionality would still need to be solved. The same would be true if Swampman had no ‘ancestor.’

32 If there is anything deep about this paper it’s likely to reside in this cheap plot device. One of the problems with arguing that we should treat Swampman as a reliable, non-intentional Turing machine is that reliable, non-intentional Turing machines can’t lie. They can simulate lying behavior by computing a function to withhold the appropriate output given certain input under certain circumstances, or malfunction, but neither of these is lying. It’s a good question whether a non-intentional Turing machine could have anything like free will.

Moreover, Swampman’s lying means that he had the concept of murder prior to any perceptual contact with actual or staged murders and that should undermine the claim that we can build a foundationalist account of Swampman’s conceptual stock in a way compatible with a causal theory of mental content.

33 During the Swamptour ‘91-’92 I received many comments on this paper from which I benefited greatly. Thanks especially to Ronald de Sousa, Andre Gallais, Christopher Gauker, Laurence Goldstein, Bernie Kobes, Andre Kukla, Kim Lycos, Jack Macintosh, Victoria McGeer, Calvin Normore, Paul Pietroski, Arthur Ripstein, Don Ross, William Seager, Paul Teller and the Davis mob, the philosophers at the University of Melbourne, the University of Queensland, and the Traditional and Modem department of philosophy at the University of Sydney, the philosophy graduate students at Syracuse University (1991), and The Algonquin Circle. I would also like to thank the referees for the Canadian Journal of Philosophy for their comments, and the editors, in particular Bernie Linsky and Mohan Matthen, for their daring.