From Aaron Sloman Thu Sep 3 05:55:45 BST 1998 To: PSYCHE-B at LISTSERV.UH.EDU Subject: Re: Theories of consciousness and MPD (The architectural basis) Cc: drwatt at classic.msn.com,dawso007 at MAROON.TC.UMN.EDU Doug Watt wrote Date: Mon, 31 Aug 1998 [DW] > 1) Strongly agree with Bernie Baars that the empirical evidence is > overwhelming that MPD is a real and not an iatrogenic disorder and that a > position of complete disbelief and dismissal should raise some questions > about the dismisser's assumptions .... with much else of interest discussed below. To which George Dawson responded Date: Tue, 01 Sep 1998 [GD] > I think that the lines of evidence to support MPD and associated > mechanisms of dissociation and repression are actually far less > compelling. > This has been a historically long and divisive debate. The > constructs of both repression and dissociation have been controversial. > .... > ... > I think a lot of the bias in the consideration of MPD from a clinical > perspective comes from the treatment of real people with the disorder. > There are a group of clinicians who believe that this is an > extraordinary condition that requires extraordinary treatment measures. > There is another group (where I include myself) who recognizes that > people with dissociation and MPD-DID need treatment, but that standard > methods properly applied are quite successful. > .... A multi-pronged attack on this problem (or set of problems) may be helpful. Most readers of this list will appreciate the need for both clinical investigations and for attempts to find neural underpinnings. Some will also appreciate the need for conceptual analysis to help clarify the questions that are being discussed and to ensure that arguments do not proceed at cross-purposes. There is a fourth approach that may be useful: explore information processing architectures that may be capable in principle of explaining various types of normal human mental functioning and then see whether they admit the possibility of specific sorts of abnormalities if stressed or damaged or developmentally disadvantaged in some way. A useful slogan is: don't expect to understand "abnormal" cases until you know how the "normal" system works, in considerable detail. Only then can you know what sorts of things can go "wrong", etc. This exploration of architectures can be more or less constrained by various kinds of evidence, and various kinds of requirements, e.g. evolutionary plausibility, requirements for learning, requirements for social interaction, requirements for neural implementability, etc. In recent postings to this group I've sketched ideas about a coarse-grained multi-layered architecture, within which I think some of the phenomena of MPD are to be expected. In this architecture there are various components which evolved at different times in response to a variety of evolutionary pressures. In particular the sketch includes (at least) the following concurrently active mechanisms: (1) a layer containing a collection of very old "reactive" mechanisms, e.g. in the cerebellum, brain stem, spinal cord, etc. ("reactive" needs careful definition, but I'll not do that now.) (2) a layer containing a collection of somewhat newer and less biologically widespread "deliberative" mechanisms -- requiring a variety of supporting mechanisms, including motive generators, associative memory, inference mechanisms. ("deliberative" also needs careful definition.) (3) a meta-management layer which can (to some extent) monitor, evaluate, and redirect processes in the other layers (and itself?) (4) one or more (possibly trainable) global "alarm" systems which can very rapidly detect patterns in sensory data or internal processing and deterine whether those patterns indicate problems or opportunities which require rapid redirection of activities in many parts of the architecture. (5) multi-layered sensory (perceptual) mechanisms, providing different sorts of information about the environment, at different levels of abstraction, to the previous layers and alarm systems. (Likewise proprioceptive sensors). (6) multi-layered motor mechanisms controlled by the previous layers. Actions may be initiated at different levels of abstraction. Sometimes the global alarm systems can grab control of the effectors. The deliberative and meta-management mechanisms in humans seem to make use of frontal lobe sub-systems. The global alarm mechanisms apparently use the brain stem, the limbic system, and maybe more specialised state-switching mechanisms in the frontal lobes. (I am only summarising things I've read, possibly inaccurately: I am no expert on brains.) NOTE: I've pointed out previously that three different classes of emotions can be associated with reactive, deliberative and meta-management mechanisms, and shown how the presence of a meta-management system with access to intermediate sensory databases can (in intelligent, reflective, animals), lead to recognition of and theories about qualia, etc. (This is loosely related to the global workspace theory.) I think phenomena related to some types of MPD are to be expected in the normal functioning of the meta-management system. Other phenomena may require serious metamanagement malfunction. I'll try to explain. 1. It seems that humans are not born with all these architectural sub-mechanisms fully fledged. Rather they develop during childhood. E.g. it takes a while before a child develops the ability to attend to its own motives or learns that they can be evaluated (e.g. as selfish or not, imprudent or not, and in some cultures as sinful or not, etc.) Likewise it takes a while before a child discovers it can remember and use the contents of its thought processes and motives a short while previously. (Before that happens wise parents can defuse some kinds of nagging by distracting a child. When the architecture is more developed and the child has learnt how to use it, e.g. by reviving temporarily suspended concerns, that ruse fails.) 2. There are good engineering design reasons why the behaviour of the metamanagement layer (e.g. the criteria used for self evaluation, the strategies for redirecting attention, the strategies for invoking deliberative processes, etc.) should not be rigidly fixed, but should be capable of (a) varying from one context to another, and (b) developing in sophistication over time (e.g. using more and more abstract characterisations of mental processes.) For instance, different types of self awareness and self control are required when relaxing with friends, when looking after one's children, when working with colleagues in the office, when playing a team game like soccer. This suggests a need for a mechanism which can detect changes of context and switch clusters of categories, rules, strategies, motivations into or out of the meta-management sub-system. Perhaps we could describe this as a "personality switching mechanism". (Examples of its operation are very familiar: the person who is a sensitive, loving and caring individual with his/her family, and a cold, calculating, ruthless ambitious individual at work. (Or vice versa.) This switching between control structures in the meta-management system is very crudely analogous to "context" switching in a multiprocessing computer which time-shares different sorts of processes on a single CPU. In a typical computer the architecture provides a collection of "hardware" central control mechanisms, such as registers, instruction counters, cache memories, etc. which can be occupied by different "software" according to what the current task is. Evolution may have got there first, though with a much more complex and subtle mechanism. In humans these changes of style or strategy according to context, can happen without the individual noticing that they do, even when it is evident to others (e.g. switching to an obsequious style in the presence of the boss). There are individual variations. A "rigid" person will be too inflexible about changing and may have too few "personalities" available for dealing with different situations. At another extreme a person may be too unpredictable, too volatile, switching between control regimes on the slightest provocation and not necessarily selecting useful subsets in all (or most) situations. Other possibilities include people who have a collection of effective control regimes, but have not developed good ones for dealing with all the commonly occurring contexts. In mild cases the person may simply be awkward or embarrassing to deal with (in some contexts), and in extreme cases totally unable to relate to others and maybe even dangerous. So far, all I've described is a collection of more or less normal and familiar phenomena with a hint at an architecture which could account for them (though there's much I have not explained). However, in more extreme cases this very same architecture could go wrong in various ways, some of which would support and explain ideas about multiple personalities. For example, normally switching contexts does not disallow access to (episodic) memories of events and actions in other contexts. However, the change of context will involve a change in criteria for directing attention, and as a result things that are important in one context may be forgotten or go unnoticed in another. I.e. memory access mechanisms and perceptual mechanisms are context sensitive. (I once failed to recognize my niece from London when she turned up unexpectedly in my office in Brighton. I thought she was a student with a vaguely familiar face, whose name I could not remember.) In an extreme case the accessibility of cross-context memories might be severely impaired, either because of a neural abnormality of some kind or because the "software" of the context mechanism has developed in a way that prevents information from different contexts being accessible at the same time. (There are many mechanisms which could achieve this within a content-addressable memory mechanism, e.g. using context-sensitive indexing mechanisms.) In that case there would be a collection of relatively independent, mutually uninformed, subsystems which take it in turns to be in the meta-management role. (For a very loose analogy think of loading different files into a word processor, or into the same spread-sheet package. Sometimes communication between file contents is possible, sometimes not, depending on the system). In some cases the accessibility barriers could be asymmetric: in context A, information about context B is accessible, but not vice versa, possibly because of the type of indexing used or because of how attention is controlled in the different contexts, or .... A more extreme pathology could arise if the early developmental process gets damaged in such a way that instead of one meta-management system developing which can at different times be "occupied" by different "software packages", the system grows several meta-management mechanisms in parallel, which thereafter vie for control, possibly unequally. Speculating wildly (partly inspired by discussions I have had with Doug) I wonder whether this could happen if severe maltreatment during the early stages of the development of the meta-management layer (the first few years of childhood) might lead to a partially grown sub-system being left unfinished because it doesn't produce good results, Then another one might be "sprouted", which is also abandoned when partly developed etc. If this sort of thing happens there could be a number of concurrently active incipient meta-management structures with only a subset ever managing to gain complete control from time to time. This would require a neurodevelopmental mechanism which is partly controlled by innate mechanisms and partly by the degree of success and failure of interaction with the environment (including other people). (Is that an accurate characterisation of early brain development in altricial species?) (Compare mechanisms involved in regeneration of capabilities after damage in adult life?) If all the above is correct we could expect to be able to distinguish (a) varieties of MPD which have to do with a not very extreme extension of normal mechanisms for switching "personalities" according to context (b) varieties of MPD which actually have abnormal, possibly stunted, physiological structures implementing various partly developed meta-management mechanisms which compete with one another for control (not necessarily on equal terms!). Cases of type (a) might be treatable by various kinds of therapy which train the patient's switching mechanisms to get better at classifying situations and turning on appropriate sets of perceptual strategies, evaluation criteria and responsive behaviours. Cases of type (b) may be impossible to restore to full normality, though it's possible that some sort of treatment might enable one of the meta-management mechanisms to develop furhter while the others atrophy and lose their ability to interfere with the main one. I have no idea what sort of therapy could do this. It has occasionally struck me that the above ideas might be closely related to phenomena like hypnotism, mass hysteria, religious ecstasy, "possession" by spirits, etc. I wonder whether readers who actually meet and treat people they classify as having multiple personalities find any of the above ideas help them understand actual cases they have encountered. I think these ideas resonate strongly with many of Doug's claims, e.g. > That MPD-DID is occasionally faked, often hyped, etc., is by no means > the same as proving that it doesn't exist. I'd add: and it may exist in several different forms, depending on whether the basic information processing architecture has been "deformed" or whether a normal architecture is being "used badly". > The real reason for the skepticism is the profoundly disturbing > implications for a unitary self/identity and a unitary stream of > consciousness, both of which we take for granted ... Which se should not do if the sort of information processing architecture I've described is at work. > There can be more hypnotic ability and less trauma, or less > hypnotic ability and more trauma in my experience. Differences which might be explained in part by the difference between context switching in one meta-management system and concurrent competition between different meta-management systems. > That we are willing and able to divide the global workspace to avoid > terror and horror at a young age seems to me anyway to be an > empirically established fact. That suggests frequently switching the occupancy of the single meta-management system with little communication between occupants. However, an alternative in some cases might be abandoning the partly grown neural architecture and starting another one. > Why we see that as having little direct implication for the possible > intrinsic relationships between emotion and consciousness is harder to > understand .... The use of the word "emotion" to label the types of implicit evaluations of success or disaster (or awfulness) which influence very early developmental trajectories may be misleading insofar as these are very different from the kinds of things we call "emotion" in normal adult life, which presuppose an already developed successfully functioning architecture, e.g. being apprehensive in the face of anticipated danger, grieving for a lost friend, being proud of an achievement.... What they all have in common is the powerful role of evaluation in control mechanisms, and large scale redirection of processing. > the outline of important connections between HC, nRt, nucleus ..... > etc. .... > might eventually provide some way of understanding the genuine > islands of awareness, the local workspaces that are partially > segregated within the global workspace of a divided personality system And also the varieties of context switching in a normal personality system? > we are talking about a fairly long time span, a "working identity" > analogue to working memory that has a much longer time frame than any > WM. Yes. the "working memory" would be only a portion of the whole meta-management mechanism. (Likewise individual registers in a computer can change their contents rapidly between less frequent global context switches.) > But there are no good neural network models. However, requiring the models to work at the level of the neural networks may be a mistake. E.g. some of them might involve higher level virtual machines which do not map onto neural architectures in any simple way. And chemical processes may play important roles (e.g. changes of mood, and modulating various forms of decision making, etc.) > What is not clear is how differential affective states prime > the "retrieval" of different alters, with the alters having such a > degree of non-access to episodic memories formed by other alters. If I am right, some of this may be just an extreme application of a useful normal mechanism for context switching, driven not only by affective states but also by detection of changes in the environment. Other cases may require suppression of competing concurrent subsystems. > These later (and what I consider) truer diseases of consciousness > reflect much more than the stripping out of modular processors and > their contributions to qualia, but instead show more fundamental > alterations and disturbances in self, qualia, and working memory (esp. > relevant to the nature of delirium ..... There are different ways in which the very complex mechanism outlined above can go wrong. Some have to do with losing or gaining accessibility to contents of various sensory buffers (qualia), some to do with various malfunctions or abnormalities in the processing of the contents. More dramatic cases involve malfunctions in the higher order control mechanisms as opposed to just the contents of what is controlled. I.e. I agree with Doug that there must be many different sorts of "diseases of consciousness" all needing to be explained, pointing to different aspects of the architecture and ways they can go wrong. cheers. Aaron ===== Aaron Sloman, ( http://www.cs.bham.ac.uk/~axs/ ) School of Computer Science, The University of Birmingham, B15 2TT, UK EMAIL A.Sloman@cs.bham.ac.uk Phone: +44-121-414-4775 (Sec 3711) Fax: +44-121-414-4281 See also ftp://ftp.cs.bham.ac.uk/pub/groups/cog_affect/0-INDEX.html