Some epistemological concerns about dissociative identity disorder and diagnostic practices in psychology

Abstract

In this paper we argue that dissociative identity disorder (DID) is best interpreted as a causal model of a (possible) post-traumatic psychological process, as a mechanical model of an abnormal psychological condition. From this perspective we examine and criticize the evidential status of DID, and we demonstrate that there is really no good reason to believe that anyone has ever suffered from DID so understood. This is so because the proponents of DID violate basic methodological principles of good causal modeling.

When every ounce of your concentration is fixed upon blasting a winged pig out of the sky, you do not question its species’ ontological status.

James Morrow, City of Truth (1990)

Acknowledgments

Thanks to Hal Brown, Edward Erwin, William Bechtel, Scott Lilienfield, and an anonymous referee for helpful comments on earlier versions of this paper.

Notes

 The dissociative disorders include dissociative amnesia, dissociative fugue, dissociative identity disorder, depersonalization disorder, and dissociative disorder not otherwise specified. Our criticisms concerning the evidential status of DID should be extendable to any disorder defined in terms of dissociation or repression.

 For more or less recent professional consideration of clinical treatments for DID/MPD see Braun (1993), Fine (1993), Kluft (1983, 1993a, b), Loewenstein (1993), Brassfield (1983), and Horevitz (1983).

 It is interesting to note that DID is the new name for multiple personality disorder (MPD) introduced in DSM-IV. The controversy over the status of MPD as specified in DSM-III led to some minor emendations in DSM-IV, including renaming. But, for the purposes of this paper, they are indistinguishable.

 See Grünbaum (1984, 1993), Erwin (1996) and Crews (1995).

 See Ross (1997) (especially p. 31) and Hacking (1995) for historical perspectives on Freud's attitude toward dissociation and how it both differs from and is similar to repression. We do not intend to suggest here that dissociation is Freudian in origin, only that it bears some resemblance to Freudian repression. Given this resemblance, the current, highly critical, attitude towards Freudian psychology also suggests that the practices and theoretical assumptions of the proponents of dissociation should be scrutinized along similar lines.

 Elizabeth Loftus is perhaps the most well known critic of the mechanisms of memory repression and memory recovery. Loftus has labored to show that people do not store complete records of events in their minds to which they later have access. Loftus' work on memory and suggestion tends to support the view that many of the ‘repressed’ or ‘dissociated’ memories are iatrogenic in origin. See, for example, Loftus (1980a, 1992, 1993, 1994), Loftus and Ketcham (1994), Loftus and Banji (1989), Spanos (1994), and Lilienfeld, Lynn, Kirsch, Chaves, Sarbin, Ganaway and Powell (1999) for discussion of iatrogenesis, suggestion and recovered memories. The methodological implications of these studies will be given a more satisfactory treatment in §7.

 The argument presented here will most closely resemble those presented in Grünbaum (1984, 1993).

 Disorders characterized in terms of dissociation are not, however, limited to the dissociative disorders proper. Acute stress disorder, posttraumatic stress disorder, and somatization disorder are also all characterized by dissociation.

 Of course, the converse of the NMP is not true.

 See, for example, Machamer, Darden and Craver (2000), Craver (2000), Craver and Darden (2001), and also Bechtel (2001) and Glennan (1996, 2000) for related approaches.

 See Salmon (1984) and Machamer et al. (2000), p. 7.

 See Glennan (2000), p. S347.

 See Shaffer (2000) for a more or less complete account of partial explanation.

 The point is noted in Machamer et al. (2000), p. 21, and is implicit in the discussion of the adequacy of mechanical models in Darden and Craver (2001).

 The DID model so sketched assumes that DID is post-traumatic in origin, and this is the typical assumption that has been made by proponents of DID as a legitimate psychological disorder (Boon & Draijer, 1993, ch.s 1, 2). As we shall see, however, this is not the only possible causal model of DID. See, e.g., Spanos (1994, 1996) and Lilienfeld, et al. 1999 for elaboration of the so-called sociocognitive model of DID. On this model, DID is not the result of a real trauma induced splitting of personalities, but rather is only the iatrogenic, socially conditioned, play-acting of the patient as if he or she had multiple personalities. This need not be intentional, but it is iatrogenic. In what follows, when we refer to the theoretical model of DID, we are explicitly referring to the post-traumatic model as characterized by D1-D5. In our opinion, it seems to be something of a mistake to even treat the sociocognitive model of DID as a bona fide model of DID, for the sociocognitive model denies that symptom A. of the DSM-IV characterization really ever occurs. Moreover, insofar as the sociocognitive model treats DID as largely (if not wholly) iatrogenic in origin, to treat such cases as real DID would be rather like categorizing false pregnancies as real pregnancies simply because these two phenomena are characterized by very similar symptoms. We hope that, for obvious reasons, this is methodologically inappropriate. If it turns out that all the symptoms exhibited by supposed DID patients are iatrogenic in this manner, then the patients in question are not really suffering from DID as it is typically understood. So, pace Lilienfeld et al. (1999) we do not accept that the issue of the existence of DID is a pseudo-issue that is tangential to the debate about the etiology of DID. The question of the etiology of a disorder is, in an important sense, just the question of the existence of a disorder caused in such-and-such a manner.

 This distinction is closely related to that discussed in Bechtel (2001).

 We are here borrowing the term ‘candidate phenomenon’ from Humphrey and Dennett (1989).

 See Ross (1997), ch. 2, and Lilienfeld and Lynn (2003).

 See Fine and Forbes (1984) for the suggestion that psychoanalytic theories ought to be judged more like theories in somatic medicine, and see Erwin (1996), ch. 2 for a forceful response.

 This point is closely related to Nancy Cartwright's conception of inference to the best cause. See Cartwright (1983), p. 92. To clarify, by ‘eliminate’ we mean only that competitor hypotheses must be shown to be significantly less explanatory, or less likely.

 The sort of reasoning going on in such cases is similar to inference to the best explanation as described in Harman (1965). It might be best understood as inference to the best causal model. In such reasoning we appear to be attempting to select the best causal model with respect to the data. As such, it is not really strictly required that we falsify all alternative existential claims, only that we make it clear that one of the set of plausible existential claims associated with the competing models is more likely than its competitors. The main point is that one must at least take into account all alternative models that have existential claims with non-marginal likelihoods.

 This condition is designed to rule out so-called ‘only game in town’ cases. As such, epistemic support for existential claims is differential. See Erwin and Siegel (1989).

 Again, as suggested in fn. 21, this principle will need to be weakened somewhat. We need not require that all plausible alternative existential claims be ruled out in the sense of being falsified. But one might be tempted to insist on the strong formulation of CMP qua strict falsification of such alternatives where the issue is a matter of the bare existence of some phenomenon.

 We take ‘plausible existential claim’ here to mean one that has a significant likelihood. How high we fix this requirement at is a partially contextual issue. See Shaffer (2000) for discussion and elaboration on this sort of theme.

 See Mayo (1996) for extensive discussion of such practices.

 Of course, all such reasoning is nonmonotonic, and so any conclusion arrived by such a procedure is revisable.

 As in the case of the simpler formulation of the CMP, we must interpret unlikely here to mean of sufficiently low likelihood, and, again, we believe that this is partially a contextual matter.

 This, of course, is a common problem in science in general.

 As we have suggested above, there is little in the way of other evidence for the existential claim associated with the mechanical model of DID, and, more, specifically with the existential claim asserting the instantiation of the mechanism of dissociation. First, as a survey of the literature shows (see the references in fn. 50), there is little or no physiological evidence for DID, and there do not seem to be any tests for DID other than these sort of test batteries. All of the evidence for the existence of DID is the result either of (frankly specious) physiological tests or from the components of test batteries. If the former evidence is, in point of fact, as weak as it appears to be and conflicts with our background knowledge concerning the nature and functioning of human memory (see the references in fn. 53), then unless the test batteries provide good evidence for the existence of DID, we have no good evidence for the existence of DID.

 There are others as well, e.g., the Dissociative Disorders Interview Schedule (DDIS).

 For a detailed explanation and evaluation of the DES and SCID-D see North, Ryall, Ricci and Wetzel (1993) and Boon and Draijer (1993).

 See, for example, Armstrong (1996), North et al. (1993), Boon and Draijer (1993), and Putnam (1989).

 See, for example, Butcher (2002a), Exner and Erdberg (2002), and Acklin (2002).

 See the various articles in Butcher (2002b).

 See Graham (1990), ch. 1.

 The criticisms raised here are similar to those raised in, e.g., Lilienfeld et al. (1999), Lilienfeld & Lynn (2003), Merskey (1992), and Spanos (1994). However, we believe that our criticisms are, nevertheless, novel in focus.

 For details of the MMPI and MMPI-2, see North et al. (1993), and, especially, Greene (1991).

 We employ the terms ‘reliable’ in the sense of epistemological reliabilism, whereby a method is reliable, if and only if, it tends to produce true beliefs rather than false beliefs.

 The criticisms we raise in this section are not entirely new as, for example, reading of Graham (1990), Levitt (1989), and Levitt and Gotts (1995) illustrates. However, despite general criticism of the MMPI and MMPI-2 as truly reliable diagnostic instruments in the sense intended by its authors, Starke Hathaway and J. C. McKinley, it has nevertheless continued to be employed with startling frequency in psychodiagnostic test batteries. This is made especially clear in the case of DID, for example, in North et al. (1993), ch. 4, Putnam (1989), ch. 4, Ross (1997), ch. 7, and Armstrong (1996).

 See Greene (1991), ch. 1, Graham (1990), and Colligan et al. (1983) for historical accounts of the MMPI, the MMPI-2 and the rationale behind their construction.

 This is especially troubling in the case of psychological phenomena that may not, in principle, be publicly observable.

 It is interesting to note that this seems to be one of the reasons behind the construction of the DES and the SCID-D, and these instruments were supposed to provide for more reliable differential diagnosis of DID and other dissociative disorders. But, we shall see that these instruments suffer from related, but somewhat different methodological flaws.

 See Bernstein and Putnam (1986), p. 729.

 See Carlson et al. (1993).

 See Boon and Draijer (1993).

 In spite of the caveats issued concerning the inability of the DES to ground differential diagnoses, Ross, for example, still claims that, ‘… it is the best self-report instrument for measuring dissociation available (1990), p. 173)’.

 This is the method of validation adopted by Bernstein and Putnam (1986).

 See Boon and Draijer (1993) for extensive discussion of this attempt to validate the DES.

 For examples of SCID-D questions and structures, see Steinberg et al. (1990).

 See Merskey (1992), Piper (1997), Spanos (1994, 1996), Lilienfeld et al. (1999), and Lilienfeld and Lynn (2003) for defense of the view that DID is not post-traumatic in origin, but rather is likely iatrogenic in origin.

 An examination of the relevant literature on the physiological aspects of dissociation shows that there is some anecdotal evidence concerning the physiological basis of dissociation, as well as a strained effort to mount a kind of parity argument between the phenomenon of hypnosis and the phenomenon of dissociation. But, as demonstrated in §4 there is no real deep mechanical understanding of how dissociation is supposed to work. Consideration of the physiological basis of dissociation is addressed in Ludwig (1983), Miller and Triggiano (1992), Braun (1983a, b). For extensive discussion of the attempt to ground the existence of dissociation by parity with hypnosis see Ludwig (1983), Bliss (1983), Bartis and Zemansky (1986), Strass (1986), Speigel (1986), Yapko (1994), Bloom (1994), Ewin (1994), Loftus et al. (1994), Lynn et al. (1994), and ch. 6 of Aldridge-Morris (1989).

 One alternative possibility for improving the situation might be to appeal to the bootstrapping methods as understood in Cronbach and Meehl (1955) in order to improve the reliability of the instruments we have discussed. Via bootstrapping we are supposed to be able to develop tests with greater validity than the criterion on the basis of which it was originally developed. For example, our primitive test for heat by the use of our tactile senses is not very reliable, and has been replaced by far more sophisticated and reliable operational methods for determining temperature, i.e. thermometers, that correlate (roughly) with human temperature sense. So, initially fallible criterion can be replaced by conceptual enrichment/change and lead to the development of much more reliable instruments that test for the property in question. In effect, one pulls one's self up by the bootstraps. While this suggestion is perhaps useful here, it, of course, depends on the development of those more highly enriched tests and our showing that they correlate with our prior, less reliable, tests for that property. However, bootstrapping can only occur if the original property that one is trying to develop an improved criterion for actually exists. Absent any independent reasons to suspect that the original test for the criterion is at least partially reliable, there is no purchase for such bootstrapping. Such bootstrapping in no way provides the kind of evidence needed to make existential claims of the sort with which we are concerned.

 Of course, the inclusion of memory repression in this theory is a Freudian influence.

 In other words, memory is taken to be a reliable source of beliefs about the past.

 Representative literature on the topic includes Loftus (1980a, 1992, 1993, 1994), Loftus et al. (1994), Loftus and Banji (1989), Loftus and Hoffman (1989), Loftus and Ketchum (1994), and Thomas and Loftus (2002).

 See Loftus (1979).

 See Loftus and Hoffman (1989), p. 103.

 They are, or course referring to the behaviorism of Watson (1939).

 See Loftus (1993).

 These factors are detailed in Toland et al. (1991), pp. 237-239.

 As the technique of Maltz illustrates, ‘ [s]pend time imagining that you were sexually abused, without worrying about accuracy, proving anything, or having your ideas make sense’ (quoted in Garry et al., 1996, p. 209).

 Garry et al. (1996), p. 213. This includes even pseudo-memories that are extremely bizarre. See Thomas and Loftus (2002).

 Additional positive evidence in favor of this view is collected in Lilienfeld and Lynn (2003).

 Here we differ somewhat in our assessment of the importance of the existential question concerning DID from, for example, Lilienfeld et al. (1999) and McHugh (1993). We take the existential question to be inextricably intertwined with the etiological question for the reasons just given.

 See Gifford (1994) and Marquis (1983) for further discussion of the therapeutic obligation.