Boden: Multiple Personality _M_U_L_T_I_P_L_E _P_E_R_S_O_N_A_L_I_T_Y _A_N_D _C_O_M_P_U_T_A_T_I_O_N_A_L _M_O_D_E_L_S Margaret A. Boden ÈÕº«ÎÞÂë Some of you may have seen the re-runs, on BBC-TV recently, of the "Face to Face" interviews done by John Freeman in the 1960's. One of these was with the singer Adam Faith, then a startlingly beautiful young man with the grace to be amazed at being chosen to be sandwiched between Martin Luther King and (if I remember aright) J. K. Galbraith. The re-runs were accompanied, where possible, with a further interview with the same person. What I found almost as startling as his lost beauty was Faith's referring to himself-when-young in the third person. After watching the re-run interview, the now middle-aged man commented to Freeman, on several occasions, that "He said such-and-such", "He told you so-and- so", and the like. This curious self-distancing was presumably based partly in a lack of memory. It would be amazing if one remembered much of what one had said in an interview twenty-five years ago, and Faith clearly did not. But cognitive-emotional distancing may have been involved, too. On a couple of occasions, Faith remarked either that he disagreed with what "he" had said then, or that he could not fathom why "he" had said it. The marked physical differences between the younger and older _p_e_r_s_o_n_a_e, and their very different life-styles, may have contributed also. But the effect was occasional, rather than sustained. Only once in a while did the third-person pronouns displace their first-person equivalents. There was no indication that the older Faith wishes to repudiate the younger, even if he disagrees with some of the views he (or even "he") held then. In short, his attitude to his younger self seemed to be much the same as ours, when we look at a faded photo-album, or read a wince-evincing diary left over from our younger days. His surprised forgetfulness and occasional embarrassment have been shared by us all. Only his fleeting use of a few unexpected pronouns was unusual. In some clinically-described cases, the self-distancing is much more pronounced. It affects current behaviour and thinking, not just past experience. There may be two (or more) memory-streams associated with one and the same physical body, two (or more) internally consistent sets of motivations and beliefs, and two (or more) distinct streams of consciousness. Some conscious experiences may be "shared", but if so the co-consciousness is often non-reciprocal: one motivational stream has access to all the experiences associated with the other, but not vice versa. There may even be two (or more) distinctive sets of physical attitudes -- involving not just the gait and body-language, but even the lines and planes of the face. Having seen the clinicians' film of the actual patient described in _T_h_e _T_h_r_e_e _F_a_c_e_s _o_f _E_v_e [Thigpen & Cleckley, 1957], I can confirm that these bodily changes are both instantaneous and compelling. Eve White's way of holding herself, in walking and sitting, was recognizably different from Eve Black's. So, too, were her facial expressions, and the lines of her face in repose. And the one physiognomy was replaced by the other in (as they say) the twinkling of an eye. Page 1 Boden: Multiple Personality What's more, the repudiation of "one" personality by the "other" is not only apparently sincere (sometimes expressed as total ignorance of the "other's" existence), but often passionate. "One" personality will refer to "another" using third-person pronouns, and perhaps a different name, and deny any responsibility for "her" actions or opinions. "She" may say that she hates "her", or is hated by her. "She" may even set cruel traps for "her", and gleefully report on the trouble and embarrassment experienced by "her" [Boden, 1972, ch. 7; Prince, 1905/1957; Wilkes, 1988]. Indeed, "she" will insist that the embarrassment (to which "she" claims to have quasi-telepathic access) is actually experienced _o_n_l_y by "her". As though all this were not enough, the "she's" and "her's" associated with one and the same body may be of very different ages, and may even be "she's" and "him's." Some clear differences in observable bodily attitudes have already been mentioned. To cap it all, there may be striking physiological differences between the various "personalities" -- such as the body's reaction to drugs, or the galvanic skin response associated with emotional arousal [Humphrey & Dennett, 1989, pp. 88-9]. Clinicians call this phenomenon multiple personality disorder, or MPD. Cases of MPD were described well over a century ago, but most cases have been reported in recent years. The numbers have grown since the 1960's, and some clinicians report "dozens" of fragmentary selves, having only "a few minutes of exclusive biography per week" [Dennett, 1991, p. 425]. This sudden surge of MPD has been taken by some as reason for scepticism. For people differ about whether MPD is "real", and about what causes it. Some critics claim that MPD is an artefact of hypnosis and other forms of suggestion. (Similar "age-shifts", for example, are routinely elicited by stage-hypnotists.) They argue that its recent explosion in the USA (in 1990, the median number of "alternates" was eleven) is due more to trendiness and publicity-seeking, on the part of therapist and patient, than to disinterested diagnosis. An unbiassed reader of the literature must conclude that there is some truth in this. However, one should note also that MPD-symptoms may be evident prior to visiting the therapist (and, in the past at least, prior to reading the newspapers). As for aetiology, the received view (among those who regard MPD as a genuine condition) is that it is grounded in childhood abuse, either sexual or physical. One woman (most MPD-patients are female) had a number of "alternates", with names such as Rachel, Ruth, Esther, and Rebecca. It turned out that -- barring total fabrication by the patient -- the woman's father, when she was a young girl, had continually abused her sexually, saying "Let's play Nazis and Jews again. You be the little Jewish girl. This time, you can be Rebecca ... " [Humphrey & Dennett, 1989]. It would not be surprising if the child at the time, and the woman in later years, wanted to distance herself as far as possible from this horror. Other case-histories of abuse are less extreme. According to sceptics, many of these may be moulded, or even initiated, by suggestion (including publicity). For our purposes, however, the aetiology is largely irrelevant. Even if many cases of MPD arise as a result of the patient's suggestibility and the therapist's desire for publicity, the phenomenon Page 2 Boden: Multiple Personality does exist. We need to understand why it is described in terms of multiple "personalities", and what it is about the mind which makes it possible. And, if I am to justify the title of this paper, we need to understand how a computational psychology may be relevant. In the clinical cases, there always seems to be some psycho-dynamic split, some dissociation between varying motivations, beliefs, and behaviour-patterns. None of the alternating personalities is emotionally well-rounded (although they often appear to be psychological complements of one another -- Eve White and Eve Black, for example). This is no accident, for it is largely what leads us to speak of two personalities in the first place. Identifying the various alternates (and encouraging their "outing" by suggestion, if that is what is going on) involves being able to distinguish separate, and perhaps psychologically incompatible, personal narratives in interpreting the observed behaviour. In short, distinct motivational and cognitive patterns are criterial of the various "personalities", and characterize their differing "biographies". To describe these cases in terms of several alternate personalities is the natural way to express the existence of two or more distinct, and continuing, cognitive-teleological patterns. For "personality" is a concept that picks out distinct types of narrative unifying (describing, explaining, and predicting) a person's life. It is rooted in what Dennett [1987] calls the intentional stance, being primarily an "outsider's" word: one used by third parties to mark someone's enduring personal style and commitments. However, it can also be used by the person concerned, in self-examining mode. We all allow that a person may sometimes act "out of character". And we allow that someone's personality may change, perhaps because of material hardship or a spouse's betrayal, and that later she may (temporarily or permanently) become "her old self" again. These ways of speaking are naturally applied, and extended, in describing MPD- patients. That's not to say that we really can make sense of there being "dozens" of personalities attached to a single body. If a putative personality makes its appearance for only a few minutes a week, it's unclear what reason we could have for calling it a personality, as opposed to a fleeting mood, an uncontrolled emotion, or even playfulness. If each of these few-minute periods appeared to be part of an integrated life, with identifiable projects being followed from week to week (as in a weekly soap-opera with exceptionally short episodes), such a description might be appropriate. Failing that, the claim that there are "dozens of personalities" seems extravagant and unhelpful. If two (or more) coherent personal narratives can be supported by the patient's behaviour, then the clinician speaks of two (or more) personalities. The identification (sceptics would say, the hallucination) of multiple personalities is influenced by the proper names and pronouns that the patient uses in talking about the various actions and experiences associated with the body in question. But other criteria are important too, including the dominant motivational "flavour" of behaviour at different times, and the degree of means-end consistency. Page 3 Boden: Multiple Personality Inconsistencies in means-end patterns can be part of the evidence leading the clinician to posit distinct biographies. Sometimes, the same bodily hand tears up on Tuesday a letter it had written on Monday, yet wipes the tears from the face on Wednesday when the destruction of the letter is "discovered". We can make narrative sense of this apparently irrational behaviour by attributing Tuesday's handiwork to one "personality", and Monday's and Wednesday's to another. But means-end inconsistency is not enough. After all, everyone has changed their mind and then changed it back again. What distinguishes this case from that of mere vacillation is the background existence of two differentially appropriate streams of activity. The relevant activity includes verbal statements. These may involve "self-distancing" uses of third-person pronouns, denials of memory and/or responsibility regarding the hand's work on Monday, and explicit acceptance or denial of the motives and beliefs distinguishing the two teleologically coherent streams. This ferreting-out of separate streams of motivational consistency is not an exact science. In one familiar philosophical terminology, it is a hermeneutic exercise, involving the ascription of intelligible interests to our fellow humans. In Dennett's [1987] terminology, it is a matter of taking the intentional stance, and using judgments of rationality in attributing one motive (or belief) rather than another to one "personality" rather than another. But rationality is a slippery concept [Stich 1983], and intelligibility runs us around the hermeneutic circle. Even if we discount cultural differences in aspiration and opinion, it is not always easy to be persuaded -- still less, to be confident -- that a given action is or is not "consistent" with (intelligible in terms of) a given narrative. And since none of us is a paragon of consistency in any case -- a point to which I shall return, below -- judgments that the inconsistency is so great as to require attribution to a distinct narrative, or "personality", will often be less than convincing. Hermeneutic, or intentionalist, interpretation therefore has to face two challenges, concerning plausibility and validity. The one asks "Does it make sense?", the other "Is it actually true?". A plausibility-challenge is one which questions an interpretation on the grounds that it does not make intentional sense. Thus one might say to a detective: "But no-one who planned to murder Mrs. Jones would dive off the pier to save her from drowning, or give her the kiss of life on the beach. That's absurd!". Similarly, someone might challenge a Freudian dream-interpretation by arguing that there is no semantic connection between the manifest and (putative) latent content of the dream, or (more likely) a connection so distant and convoluted that it strains our interpretative capacity too far. In other words, it is implausible because it is unintelligible. In a validity-challenge, the argument is that the story in question, although perfectly intelligible, simply could not be true. This time, it might be the detective herself who offers the challenge: "You were given the missing diamonds by a visiting Martian? They fell off a lorry? ... Come off it!". There are, she assumes, no Martians in the district, and no diamond-dispensing lorries either. Produce a Page 4 Boden: Multiple Personality Martian, or a suitably leaking lorry, and she would be prepared to think again. In the context of MPD, the validity-challenge would be the charge that it is impossible for distinct personal narratives to be sustained by one and the same human body. (It would follow that the patient must be fabricating, and the doctor colluding.) An adequate answer would have to explain what it is about human beings -- specifically, about human minds -- which, to the contrary, makes it entirely possible for this sort of thing to happen. Notice that to say that there are two (or more) personalities is not the same as saying that there are two (or more) _m_i_n_d_s. Indeed, the clinical facts suggest that there is only one mind. The varying aspects of this mind, instead of being functionally integrated within one system, have been partially dissociated to form two (or more) sub-systems. Each sub-system shows a degree of psychological integration over time (if it did not, we would not be tempted to identify it as an alternate "self" in the first place). But each seems less than a whole person, since it excludes certain types of behaviour and experience typical of normal minds. Moreover, as noted above, the different selves typically seem to be complementary. What one possesses, the other lacks: think of Eve White and Eve Black, or of Dr. Jekyll and Mr. Hyde (a fiction based on nineteenth-century clinical case-histories). Imagine putting the parts together, and you would have a more complete human mind -- though still an abnormal one, swinging between two motivational extremes. We need to understand, then, how it is possible for a mind -- which we usually think of as a unitary system -- to be "split" in this way. It might seem that computational psychology cannot help us here. Given the account of MPD outlined above, how can computer models be relevant? Normal personality involves many different, partly incompatible, goals and many potentially conflicting motives. The clinical evidence, and the theory about self-distancing being prompted by abuse, both suggest that it is when these motives cannot be reconciled that they are separated into two or more functionally independent streams. But computer models, it may be said, are boringly single-minded. They are concerned only with single goals, or at best with a variety of sub-goals all chosen and followed in light of the one overall goal. Hence they are thoroughly unsuitable as theories of personality, whether normal or abnormal. This is certainly true of most "goal-directed" AI-models. In some of these, such as the "General Problem Solver" and traditional planning-programs in general, the goal-setting is explicit [Newell & Simon, 1961; Boden, 1987, ch. 12]. In the class of programs known as production systems, it is implicit [Newell & Simon, 1972; Boden, 1988, ch. 6]. But in either case, we are dealing with essentially single- minded systems. (To be fair, one of the earliest AI-programs did try to model multiple goals, in a simulation of personality; but the attempt was premature [Reitman, 1962, 1965].) What of connectionist AI? PDP-systems (using parallel distributed processing) are tolerant of conflict. Because their representations are embodied as best-fit equilibrium patterns defined over a network of Page 5 Boden: Multiple Personality units, they can function adequately in the face of conflicting evidence [Hinton, McClelland, & Rumelhart, 1990]. They might, therefore, seem even less likely than classical AI to support a functional "split" between conflicting goals. However, most current connectionism does not -- and cannot -- model goal-seeking at all. Only a few networks are designed to deal with hierarchical structure or sequential decison- making, and most of those are reminiscent of traditional AI [e.g. Hinton, 1991]. This is no accident: even leading connectionists argue that a neural network must simulate a digital computer to model means- end activity [Hinton _e_t _a_l., 1990; Clark, 1989]. Recent work in "situated robotics" shows that, up to a point, apparently purposeful and integrated behaviour can be produced without any central supervision or top-down planning [Maes, 1991]. The behaviour of situated robots results from a number of functionally separate mini- agents. Each of these simply does its own (simple) thing, with no guidance from above -- other than the "unthinking" inhibition of certain actions in certain situations. The relevant actions and inhibitions are all utterly automatic. The apparent integration "emerges", rather than being explicitly imposed. Barring physical injury, the integration cannot break down. Nor can any apparent dissociation be "cured", except by engineering. There is no top-down program, some change in which might alter the nature or degree of dissociation. On the contrary, behaviour is attributed to the creature's physical embodiment, not to its mental processes. It follows that this newly fashionable form of (decentralized) robotics does not offer us a helpful model of MPD. However, some computational theorists who (like the situated roboticists) see the mind as a set of mini-agents come rather closer to classical AI's position on the control of action. Both Marvin Minsky [1985] and Daniel Dennett [1991], for instance, depict the mind as a complex "society" of agents. In their theories, the societal mind has no single executive control, or master program. But it supposedly develops an integrative sub-system, which constructs a "self-model", or biographical "narrative", unifying the person's actions. This teleological unification consists partly in anticipatory direction, and partly in _p_o_s_t _h_o_c rationalization. Evidence from social psychology, never mind Freud, shows that such rationalization is very common, even where simple actions are concerned. People often don't really know why they did what they did, giving sincere (but false) explanations of their actions that make sense within a (confabulated) personal story [Nisbett & Ross, 1980]. According to both these "societal" computational theorists, the internal narrative coherence is not perfect. Personal integration involves functional negotiations of many types, and at many levels. Accordingly, the integration -- or dissociation -- of the mind varies on many different dimensions. The self-system constructs the best reflexive interpretation it can, and then (to some degree) imposes it in guiding future actions. Many years before computer models were dreamt of, some clinicians were describing MPD in terms of hierarchical integration and dissociation. Indeed, William McDougall [1926; Boden, 1972, ch. 7] even combined this language with a societal analogy, and used that analogy to explain MPD. He saw the normal human mind as a "colony" of competing and Page 6 Boden: Multiple Personality cooperating agents, or "monads". These agents were said to be organized into various cognitive-motivational-affective structures: idiosyncratic "sentiments" focussed on wealth, learning, ice-skating ... and so on. And the sentiments, he argued, are unified in turn by a special agent: the self, or "chief monad", associated with "the master-sentiment of self-regard". Lower-level monads may sometimes "escape" from the control of the chief monad, the master-sentiment being unable to integrate them all as well as usual. If these escaped lower-level monads are themselves mini-masters of others, functioning in a reasonably consistent and integrated fashion, we may -- McDougall said -- have a case of "multiple personality". The idea of a societal mind, and of its relevance to MPD, is therefore not new. But by grounding it in computational theory, we may hope to get a better purchase on just what may be involved, and just how the striking clinical phenomena are possible. Admittedly, there are no large-scale computer models exemplifying the mental processes underlying personality. We still know very little about the mental architecture required to support complex systems of motivation and emotion. But some progress is being made on this front, in identifying the sorts of task which such a system must be able to perform. A human mind includes a number of motives, often competing for the person's attention and for the use of their hands and time. Certain general questions therefore arise, about how the limited mental attention and bodily resources can be allocated between the various motives, and how these motives can be scheduled relative to each other. We must distinguish motives of differing urgency, insistence, and importance: speaking of "strong" and "weak" motivation is not enough. Compare, for instance, the urgency of a motive to buy bread just as the bakery is about to shut, the insistence of a motive to locate a shop in the neighbourhood selling one's favourite type of bread, and the importance of a motive to eat some food every day. These distinctions, and many others, are made in a recent computational account of multiple motivation [Beaudoin & Sloman, 1993; Sloman, 1990, 1991]. Aaron Sloman discusses a control system for scheduling motives in a teleologically consistent way, one which is sensitive also to mere preferences, to changes in belief, to teleologically appropriate emotions, and to shifting moods. His theory, and similar ones [_e._g. Johnson-Laird, 1988], cannot yet be fully implemented in a computer model. But the computational concepts used in it help us to think clearly about what is involved in choosing between our many aims and preferences. Sloman outlines mechanisms for selecting between competing motives, mechanisms which take seriously distinctions such as those listed above. That's not to say that detailed problem-solving goes on whenever there is a motivational clash. On the contrary, an urgent motive is one which has to be satisfied quickly, so that careful consideration of evidence and alternatives would be out of place. If the bakery is about to shut, you have no time to luxuriate in thinking just what sort of bread you want to buy: simply, you must get in there. If the motive is both urgent and important (escaping from an approaching tiger), the time available for thought is even shorter. Page 7 Boden: Multiple Personality On the other hand, an important but non-urgent motive may continually be placed at the head of a priority-queue, for action or for deliberation. (The more insistent the motive is, the more often this will happen.) Accordingly, when the system has no more urgent need, it will consider how the relevant goals might be achieved. This process may require complex problem-solving, involving evaluations of various kinds (such as personal preferences and moral codes), inference from stored beliefs, means-end analysis, and contingency-planning. Means-end planning allows for one and the same action to be done, at different times, under the control of different motives. So buying bread, for instance, may be done to appease one's own hunger, to seduce a hungry girl, to make money, to compete for a prize, to glorify God, or even to court a princess [Boden, 1981]. To discover which motive is underlying the observed action, we need to put it into an intentional context that provides a narrative covering the person's behaviour. Does the bread-buyer immediately eat the bread himself? Does he carry it round the corner with a lascivious smirk, depositing it in the lap of a beautiful starving girl? Does he try to sell the loaf (and others) at a profit to people who were unable to go to the bakery? Does he eat fifty loaves at one sitting, to get into _T_h_e _G_u_i_n_n_e_s_s _B_o_o_k _o_f _R_e_c_o_r_d_s? Does he bless the bread, and use it as the Eucharist? Or does he recount the tasks set to him by the king, whose daughter's hand is promised to the first suitor who can complete them? Such intentional patterns must be identified in someone's behaviour if we are to understand what moves them, what they are really doing, in behaving in "one and the same" way at various times. The more often these patterns are repeated, the more we see them as stable dispositions, aspects of character or personality. Moreover, in normal minds (and in well-designed computational architectures), two or more motives may be approached, or even achieved, by the same activity. A man who wishes to seduce a starving girl may be peckish himself, and share some of the bread as she wolfs it down. A system which wanted many different things, but could pursue only one goal at a time, would be unable to achieve this sort of _r_a_p_p_r_o_c_h_e_m_e_n_t between different motives. Control of its behaviour would flit from one motive to another, appearing excessively single-minded over brief periods of time. There would be much wasted effort: not only unnecessary repetitions (due to its not being able to kill two birds with one stone), but self-defeating actions too, wherein a goal or sub- goal that has already been achieved is later undone in pursuit of some quite different end. The similarity of such behaviour to certain aspects of MPD is evident. In general, one can think of mental dissociation as resulting from various types of compromise or breakdown in the complex control system that informs and "unifies" the normal mind. (The word _u_n_i_f_i_e_s needs scare-quotes because the motivational and cognitive unity, or consistency, of normal minds is by no means perfect.) Independent, alternating, and perhaps even competing motivational structures will very likely arise if the usual control-mechanisms for integrating motives break down. Differential access to memories, which plays an important criterial role in individuating "personalities" or "alternates", may also arise in this way. If we think of the mind as a Page 8 Boden: Multiple Personality computational system, we can see how it is possible for some memories to be shared between several (or all) alternates, and for others to be accessible only to one (or two ...). Reciprocal and non-reciprocal co- consciousness, too, can be understood in these terms. Whereas McDougall had to resort to positing "telepathic" (non-energetic) communication between the monads forming the society of the mind, we can distinguish computational (non-energetic) access and control of various specific kinds. As noted above, we can't yet do this with sufficient completeness and precision to explain just how normal minds function, or just how they can malfunction. Nor can we use computational ideas to explain the sudden changes in body-language, or the different sensitivities to drugs. But the general possibility of one (malintegrated) computational system supporting distinct "narratives" is clear. Insofar as our puzzlement over MPD concerned _h_o_w _i_t _i_s _p_o_s_s_i_b_l_e _a_t _a_l_l, it should by now have been allayed. Insofar as our puzzlement concerned _w_h_y the motivational splits happened, an aetiology of childhood abuse could be understood in computational terms. A complex system involving a reflexive representation, or "self-image", involving evaluative norms might well adapt to negative self-judgments by "splitting" the self-image so as to avoid them. Either: "This is not happening to me, but to someone else (Rachel)", or "This is not being done by me, but by someone else (Eve Black)." This sort of narrative split, with the associated displacement of responsibility, could also serve to avoid negative judgements of other people: "My father is not doing this to me, his daughter, but to someone to whom he owes no special care". Very early in the history of computer modelling, a psychoanalyst modelled the neurotic shifts in belief which he had observed in one of his patients [Colby & Gilbert, 1964]. To be sure, this program was simplistic in many ways -- not least, in its primitive definition of analogy and its quantitative measures of different types and levels of anxiety [Boden, 1987, chs. 2 & 3]. But, crude though it undoubtedly was, it suggested how a computational approach might model the denials, projections, introjections, and displacements described by Freud's theory of defence mechanisms. These cognitive shifts, with their anxiety-reducing effects, are an adaptation to otherwise insupportable experiences (in particular, negative self-judgments). One does not have to be a Freudian to allow that such semantic shifts can and do happen in human minds, with significant affective consequences. If a woman can transform "I hate father" into "I hate the boss", or "I hate you" into "You hate me", is it really so surprising that she may sometimes transform "I did it" into "She did it?" -- or even into "It's the sort of thing _s_h_e does"? And is it really so surprising that she can construct a serial narrative (or two, or three ...) in which actions done by her body are consistently attributed by her to someone else? If a serene middle-aged man can occasionally refer to his younger self in the third person, an anguished mind can surely seek (unconsciously) to escape the anguish by systematic denial and role-playing. The young Adam Faith had a hit-song asking "What do you want if you don't want money? Wish you wanted my love, baby." An MPD-patient would have given him two, or more, radically conflicting answers. Even if Page 9 Boden: Multiple Personality these had included the one he wanted, he'd have found it impossible to satisfy them all. But the older Adam Faith, if he ever reads the computational literature, could have a better understanding of what would have been going on. His understanding, like ours, would still be incomplete. But that just goes to show what a complex system the human mind is, and how much we still have to learn about its integration and control. Page 10 Boden: Multiple Personality _R_E_F_E_R_E_N_C_E_S Beaudoin, L. P., & A. Sloman. [1993] "A Study of Motive Processing and Attention". In A. Sloman, D. Hogg, G. Humphreys, D. Partridge, & A. Ramsay (eds.), _P_r_o_s_p_e_c_t_s _f_o_r _A_r_t_i_f_i_c_i_a_l _I_n_t_e_l_l_i_g_e_n_c_e, pp. 229-238. Amsterdam: IOS Press. Boden, M. A. [1972] _P_u_r_p_o_s_i_v_e _E_x_p_l_a_n_a_t_i_o_n _i_n _P_s_y_c_h_o_l_o_g_y. Cambridge, Mass.: Harvard University Press. Boden, M. A. [1981] "The Structure of Intentions". In M. A. Boden, _M_i_n_d_s _a_n_d _M_e_c_h_a_n_i_s_m_s: _P_h_i_l_o_s_o_p_h_i_c_a_l _P_s_y_c_h_o_l_o_g_y _a_n_d _C_o_m_p_u_t_a_t_i_o_n_a_l _M_o_d_e_l_s, pp. 133-156. Ithaca, NY: Cornell University Press. Boden, M. A. [1987] _A_r_t_i_f_i_c_i_a_l _I_n_t_e_l_l_i_g_e_n_c_e _a_n_d _N_a_t_u_r_a_l _M_a_n. 2nd edn., expanded. London: MIT Press. Boden, M. A. [1988] _C_o_m_p_u_t_e_r _M_o_d_e_l_s _o_f _M_i_n_d: _C_o_m_p_u_t_a_t_i_o_n_a_l _A_p_p_r_o_a_c_h_e_s _i_n _T_h_e_o_r_e_t_i_c_a_l _P_s_y_c_h_o_l_o_g_y. Cambridge: Cambridge University Press. Clark, A. [1989] _M_i_c_r_o_c_o_g_n_i_t_i_o_n: _P_h_i_l_o_s_o_p_h_y, _C_o_g_n_i_t_i_v_e _S_c_i_e_n_c_e, _a_n_d _P_a_r_a_l_l_e_l _D_i_s_t_r_i_b_u_t_e_d _P_r_o_c_e_s_s_i_n_g. Cambridge, Mass.: MIT Press. Colby, K. M., & J. P. Gilbert. [1964] "Programming a Computer Model of Neurosis," _J_o_u_r_n_a_l _o_f _M_a_t_h_e_m_a_t_i_c_a_l _P_s_y_c_h_o_l_o_g_y, 1, 405-417. Dennett, D. C. [1987] _T_h_e _I_n_t_e_n_t_i_o_n_a_l _S_t_a_n_c_e. Cambridge, Mass.: MIT Press. Dennett, D. C. [1991] _C_o_n_s_c_i_o_u_s_n_e_s_s _E_x_p_l_a_i_n_e_d. Boston, Mass.: Little, Brown. Hinton, G. E., J. L. McClelland, & D. E. Rumelhart. [1990] "Distributed Representations." In M. A. Boden (ed.), _T_h_e _P_h_i_l_o_s_o_p_h_y _o_f _A_r_t_i_f_i_c_i_a_l _I_n_t_e_l_l_i_g_e_n_c_e. Oxford: Oxford University Press. [First published as Chapter 3 of D. E. McClelland & J. E. Rumelhart (eds.), _P_a_r_a_l_l_e_l _D_i_s_t_r_i_b_u_t_e_d _P_r_o_c_e_s_s_i_n_g: _E_x_p_l_o_r_a_t_i_o_n_s _i_n _t_h_e _M_i_c_r_o_s_t_r_u_c_t_u_r_e _o_f Cognition, Vol. I: Foundations, pp. 77-109. Cambridge, Mass.: MIT Press.] Hinton, G. E. [1991] _C_o_n_n_e_c_t_i_o_n_i_s_t _S_y_m_b_o_l _P_r_o_c_e_s_s_i_n_g. Cambridge, Mass.: MIT Press. Humphrey, N., & D. C. Dennett. [1989] "Speaking for Ourselves: An Page 11 Boden: Multiple Personality Assessment of Multiple Personality Disorder," _R_a_r_i_t_a_n, 9, 68-98. Johnson-Laird, P. N. [1988] _T_h_e _C_o_m_p_u_t_e_r _a_n_d _t_h_e _M_i_n_d: _A_n _I_n_t_r_o_d_u_c_t_i_o_n _t_o _C_o_g_n_i_t_i_v_e _S_c_i_e_n_c_e. London: Fontana. McDougall, W. [1926] _O_u_t_l_i_n_e _o_f _A_b_n_o_r_m_a_l _P_s_y_c_h_o_l_o_g_y. New York: Scribner. Maes, P., ed. [1991] _D_e_s_i_g_n_i_n_g _A_u_t_o_n_o_m_o_u_s _A_g_e_n_t_s. Cambridge, Mass.: MIT Press. Minsky, M. L. [1985] _T_h_e _S_o_c_i_e_t_y _o_f _M_i_n_d. New York: Simon & Schuster. Newell, A., & H. A. Simon. [1961] GPS -- A Program that Simulates Human Thought. In H. Billing, ed., _L_e_r_n_e_n_d_e _A_u_t_o_m_a_t_e_n, pp. 109-124. Munich: Oldenbourg. Reprinted in E. A. Feigenbaum & J. Feldman (eds.) [1963], _C_o_m_p_u_t_e_r_s _a_n_d _T_h_o_u_g_h_t, pp. 279-96. New York: McGraw-Hill. Newell, A., & H. A. Simon. [1972] Human Problem Solving. Englewood Cliffs, NJ: Prentice-Hall. Nisbett, R. E., & L. Ross [1980] _H_u_m_a_n _I_n_f_e_r_e_n_c_e: _S_t_r_a_t_e_g_i_e_s _a_n_d _S_h_o_r_t_c_o_m_i_n_g_s _a_n_d _S_o_c_i_a_l _J_u_d_g_m_e_n_t. Englewood-Cliffs, NJ: Prentice- Hall. Prince, M. [1905/1957] _T_h_e _D_i_s_s_o_c_i_a_t_i_o_n _o_f _a _P_e_r_s_o_n_a_l_i_t_y: _A _B_i_o_g_r_a_p_h_i_c_a_l _S_t_u_d_y _i_n _A_b_n_o_r_m_a_l _P_s_y_c_h_o_l_o_g_y. New York: Meridian Books, 1957. (First published, 1905.) Reitman, W. R. [1963] "Personality as a Problem-Solving Coalition." In S. S. Tomkins & S. Messick (eds.), _C_o_m_p_u_t_e_r _S_i_m_u_l_a_t_i_o_n _o_f _P_e_r_s_o_n_a_l_i_t_y: _F_r_o_n_t_i_e_r _o_f _P_s_y_c_h_o_l_o_g_i_c_a_l _T_h_e_o_r_y, pp. 69-100. New York: Wiley. Reitman, W. R. [1965] _C_o_g_n_i_t_i_o_n _a_n_d _T_h_o_u_g_h_t: _A_n _I_n_f_o_r_m_a_t_i_o_n-_P_r_o_c_e_s_s_i_n_g _A_p_p_r_o_a_c_h. New York: Wiley. Reitman, W. R., R. B. Grove, & R. G. Shoup. [1964] "Argus: An Information-Processing Model of Thinking," _B_e_h_a_v_i_o_r_a_l _S_c_i_e_n_c_e, 9, 270-81. Sloman, A. [1990] "Motives, Mechanisms, and Emotions." In M. A. Boden (ed.), _T_h_e _P_h_i_l_o_s_o_p_h_y _o_f _A_r_t_i_f_i_c_i_a_l _I_n_t_e_l_l_i_g_e_n_c_e, pp. 231-247. Oxford: Oxford University Press. Page 12 Boden: Multiple Personality Sloman, A. [1991] _P_r_o_l_e_g_o_m_e_n_a _t_o _a _T_h_e_o_r_y _o_f _C_o_m_m_u_n_i_c_a_t_i_o_n _a_n_d _A_f_f_e_c_t. CSRP-194, School of Cognitive and Computing Sciences, University of Sussex. Stich, S. C. [1983] >_F_r_o_m _F_o_l_k _P_s_y_c_h_o_l_o_g_y _t_o _C_o_g_n_i_t_i_v_e _S_c_i_e_n_c_e: _T_h_e _C_a_s_e _A_g_a_i_n_s_t _B_e_l_i_e_f. Cambridge, Mass.: MIT Press. Thigpen, C. H., & H. M. Cleckley. [1957] _T_h_e _T_h_r_e_e _F_a_c_e_s _o_f _E_v_e. London: Secker & Warburg. Wilkes, K. V. [1988] _R_e_a_l _P_e_o_p_l_e: _P_e_r_s_o_n_a_l _I_d_e_n_t_i_t_y _W_i_t_h_o_u_t _T_h_o_u_g_h_t _E_x_p_e_r_i_m_e_n_t_s. Oxford: Clarendon Press. Page 13