The Legion Project is a test bed for ideas I have developed about the nature of  subjectivity and agency in cyberspace.  In particular, I maintain that in the cyberspace domain our subject stance becomes fractured and distributed in a way that can be modeled by Dissociative Identity Disorder, and that this is a natural and adaptive mode of being in that domain.  The main feature of this project is a 3D avatar virtual world, LegionWorld, which embodies these ideas in its architecture in a way that makes them functionally explicit and testable.  My goal was to create this virtual world and then user test it to see if it validates my ideas.  What follows is a presentation of the theory that informs this project, plus a description of my implementation of the project and results gathered from testing.


The creation of LegionWorld required a great deal of 3D modeling and VRML conversion, and I worked with a team  of two talented undergraduate students, Bart Landry and Tom Ivey.  Their technical abilities, feedback and input helped to make LegionWorld a reality.


Part 1: Dissociative Identity Disorder and cyberspace


Pathology or paradigm?


The very nature of cyberspace is acting as a sort of “universal solvent” to break up our long cherished notions of monolithic self and subjectivity.  The ease with which we can take on new identities and new connectivities in the virtual domain implies a self which is multivalent, fluid and de-centered-- a self which often operates in parallel, in different modalities and different capacities.  I propose that a model for this new, postmodern, postrationalist, cyberspatial self can be found in the cluster of psychological manifestations currently known as Dissociative Identity Disorder (DID), and formerly and more popularly as Multiple Personality Disorder (MPD).


There is precedence for my proposal to use what is commonly seen as a form of mental illness as a ground for more fully realized being.  Deleuze and Guattari put forth the ideal of the happy schizophrenic as a paradigm for dealing with our postmodern condition of existential free-fall (Deleuze and Guattari, 1977).  Through a process of “schizo-analysis” they transform the concept of madness to one of primal creativity analogous to artistic creativity.  This schizo-analysis is an essentially destructive process designed to break down the sorts of Modernist, totalizing  institutions that Deleuze and Guattari see as oppressing and confining our true selves. The result is a nomadic, rootless self, free of family, belief and structure, and dedicated to relentless change.


The schizophrenic lives in a more or less completely self-constituted world, where fragments of personality are imbedded in a matrix of delusional projections.  It is a world where inner-driven states are substituted for any kind of consensual reality, and a world where it is almost meaningless to talk about having “personality” at all.  The task of Deleuze and Guattari’s happy schizophrenic is to harness this condition of non-referentiality to the use of forging a new kind of self, in an essentially linguistic transformation, free of the moral and psychological despotism of Modernism.


Schizophrenia is very different from DID, however, and while much of what I am proposing may seem at first glance similar to Deleuze and Guattari’s project, there are in fact vast differences.  The challenge for both myself and Deleuze and Guattari, is to take a condition held by most to be maladaptive and pathological, and turn it into a paradigm for better and more effective living.  I believe that I have the easier task.


The concept of dissociation


The concept of Multiple Personalities is itself a fluid one, and it has evolved over not just centuries, but millennia, and throughout cultures as varied as nomadic hunter gatherers of the sub-arctic to contemporary industrial societies.  At its core, this concept involves the inherent capacity of the human psyche to dissociate, to spread itself out over psychic space and time.  Colin Ross defines dissociation quite simply: “Dissociation is the opposite of association...For definitional purposes the psyche may be reduced to a collection of elements in complex relationships with each other.  Psychic elements include thoughts, memories, feelings, motor commands, impulses, sensations, and all the other constituents of psychic life.  Any two psychic elements may be in a dynamic relationship with each other, which is to be associated, or relatively isolated and separate, which is to be dissociated.” (Ross, 1997, p 116).


Dissociation is an important factor in normal psychological functioning, allowing for a degree of mutability and adaptability which would be impossible without it.  It is probably best to think of dissociation on a continuum, from ordinary daydreaming, through such phenomena as forgetting where one is going on the freeway, to the particularly florid manifestations of dissociation which are commonly labeled as Multiple Personality.   Although the phenomenon is usually thought of as purely psychological in nature, it can actually be either biologically or psychosociologically driven, as given by Ross (Ross, 1997, p 116):


1.      Normal psychosocial dissociation--  Daydreaming during a boring lecture.

2.      Normal biological dissociation--  Forgetting that you got up in the night to go to the bathroom.

3.      Abnormal psychosocial dissociation--  Amnesia for incest.

4.      Abnormal biological dissociation--  Amnesia following a concussion.


These biological aspects of dissociation reinforce the notion that it is an entirely natural ability, and serve to lessen emphasis on pathological manifestations.  This sort of grounding in the ordinary is necessary, to blunt criticism that I am building some kind of fool’s paradise on a foundation of mental illness.  Nothing could be further from the truth, as I am claiming that dissociation is not only an adaptive, but a highly desirable strategy for dealing with the complex, multi-dimensional social relationships enabled by cyberspace technologies.  It is more “sane”, if you will, than strategies which try to preserve a monolithic unit of self and agency in the face of almost exponentially increasing demand for parallel processing and multiplicity.


The earliest known references to dissociative phenomena are to be found in Neolithic cave paintings, in which shamans can be seen transforming into animals and spirits, presumably for purposes of enabling the clan in its survival efforts.  In fact, throughout the long march of what we call pre-history, such dissociative abilities had great survival value, by creating a closer connection with the worlds of spirits and animals, with which humans coexisted.  Persons adept at such dissociations were given positions of considerable power and respect in “primitive” societies.


The shamanistic tradition has survived almost to the present day in the circumpolar regions of Asia and North America, and study of these cultures provides our best view of how shamanism operated over tens of millennia.  Some of the arguments made about shamanism are similar to arguments which might be given to my proposed schema, so it is illustrative to examine this incredibly ancient tradition.


Most shamans, as near as can be determined, seemed to be healthy and not suffering from mental disorders.  They were integral members of their societies and often went through long periods of training to become adept, all of which argue against any sort of pathological component to their craft.  The shamans of the far north rarely, if ever, used hallucinogenic plants, relying instead exclusively on self-hypnosis.  They were, in effect, rigorously trained professional dissociators who functioned variously as priests, weather forecasters, doctors and conveyors of the oral tradition.


Colin Ross (Ross, 1997, p 153) lists 11 dissociative features of the shamans work, and relates each to DID.  These parallelisms are worth mentioning here, as they further ground the experiences of DID patients in a milieu that is not pathological:


Structured, Meaningful Hallucinations--  Shamans deliberately induced special states of being, in which they could communicate with other realms of being, such as the spirit and animal worlds.  Like the “hallucinations” of DID patients, however, these events were not indicative of psychosis, but rather clearly dissociative in nature.


Trance States--  Trance states were essential prerequisites for spirit communication, out of body experiences, possession, etc.  DID patients frequently enter trance.  One of the major diagnostic criteria for DID is ease of hypnotizability.


Hypnotic Anesthesia--  Such activities as walking naked in the arctic winter, holding hot objects, and self-piercing indicate hypnotic anesthesia.  DID patients often report anesthesia for the pain of abuse, or self-inflicted pain.


Symbolic Dreams--  Shamans’ dreams were often lucid and directed at conveying information about the spirit world or the real world.  Likewise, DID patients’ dreams often involve conveying information about past abuse, organization of the personality system or other buried information to the waking personality.


Ritual Dismemberment--  Often initiation as a shaman involved ritual dismemberment analogous to death, associated with  spiritual fragmentation.  DID patients have, by definition, undergone dismemberment of the self as a strategy for survival.


Possession by the Souls of Ancestors--  In trance states, shamans would often deliberately become posed by the souls of dead ancestors, who would impart advice and wisdom. It is fairly common for DID patients to have at least one alter who claims to be a dead ancestor.


Possession by Helping Spirits--  These “familiars” were often invoked by shamans during trance states as helpers and guides.  DID patients almost always have helper personalities, some of whom claim to be from other dimensions or to be spirits.


Exhaustion Following Strenuous Trance Work--  Intense trance events often left shamans tired for a period of days.  Intense DID treatment sessions have a similar effect.


Stimulating Dissociation through Intoxication-- As mentioned, it was rare for shamans of old to use intoxicants, except, rarely, hallucinogenic mushrooms.  DID patients, on the other hand, frequently resort to chemical paths to dissociation, and often manifest substance-abusing alters.


Out-of-Body Experiences--  A common technique for shamans was astral projection, or out-of -body travel, where they would project themselves into the spirit or real world to accomplish tasks.  Many DID patients have had out-of-body experiences, often at the time of childhood abuse, as a means of psychic escape.


Transformation of Identity--  During the possession state of trance, the shaman became reciprocally identified with the possessing spirit in a manner analogous to alter switching in DID patients.


As can be seen from this list of correspondences, there is much commonality between the “professionally” achieved and culturally approved dissociative states of the sub-arctic shaman, and the dissociative states of the DID patient.


A present day culture in which dissociative phenomena are active in daily life is described by  Suryani and Jensen in their study of trance and possession phenomena in Bali (Suryani and Jensen, 1993).  They found that highly dissociative trance states were an accepted part of the culture and no negative feelings attached to persons so dissociating.  In fact, there are certain persons who are easy to put into trance states by startling and other techniques, and there is a great deal of good-humored fun made at the expense of these persons, without it becoming at all mean-spirited.


Multiple personalities


The above discussion of DID as related to non-pathological functioning is important, as it will serve to ground our perceptions of this phenomenon while we tread the increasingly exotic terrain covered by the manifestation of DID usually called Multiple Personalities.  The category called DID in the DSM-IV (Diagnostic and Statistical Manual, Fourth Edition, of the American Psychiatric Association) was in fact called Multiple Personality Disorder (MPD) in the DSM-III, and so quotes from some materials will refer to the syndrome by that name.


According to Putnam, “The core feature of MPD is the existence of alter personalities who exchange control over an individual’s behavior.  It is important to state from the outset that whatever an alter personality is, it is not  a separate person.” (Putnam, 1986, p 103, italics the author’s).  Putnam goes on to quote the definition of an alter personality given by Braun and Kluft as “an entity with a firm, persistent, and well-founded sense of self and a characteristic and consistent pattern of behavior and feelings in response to given stimuli.  It must have a range of functions, a range of emotional responses, and a significant life history (of its own existence).” (Kluft, 1984, p 23).


Stephen Braude has constructed an interesting interpretation of multiple alters in terms of what he calls apperceptive centers  (Braude 1991).  First he postulates that mental and behavioral states are indexical  if a person believes such states to be his/her own.  This is a purely epistemological property of such a state.  In contrast, if a person experiences  a state as his/her own, then that state can be said to be autobiographical,  a phenomenological property of that state.  Apperceptive centers are defined as “an individual most of whose autobiographical states are indexical” (Braude, 1991, p 78), and he proposes that distinct alters tend to be distinct apperceptive centers.  He distinguishes multiples from other strong forms of dissociation by claiming that multiples tend to have more than one apperceptive center at one time, whereas other forms of dissociation, and non-dissociated states, do not.  I find this interpretation to be the most appealing of any I have read.


The sense of self maintained by alter personalities is solid enough that they often appear to each other as being very different, physically, emotionally and psychologically.  They often claim vociferously to be independent persons, in the face of all evidence to the contrary.  For instance, one alter might insist that another’s clothes are not just in bad taste, but are of the wrong size, or even appropriate for the wrong sex.  Sometimes alters even see other alters as being differently located physically, as sitting in a different chair or being in a different room.  In really extreme cases, one alter might wish to kill another, adamantly denying any connection whatsoever with that alter.


In addition to such well delineated alter personalities, it is typical to find a number of personality fragments-- smaller units of personality of limited scope and ability.  Typically a personality fragment will be limited to one affect, such as anger or joy, or one functionality, such as driving a car or absorbing pain.  This sort of fracturing can yield fragments so specific that it is ultimately useless to consider them as personalities.


Alter personalities can be thought of as units of functionality which enable the DID patient to function in the external world, or cope with the complex demands of the inner world such people inhabit.  External duties include holding a job, managing relationships with other people, taking care of bodily needs, etc.  Internal duties include such activities as managing which alter is “out” (in control of the physical body) at any time, holding and managing traumatic memories, and settling internal squabbles between alters.  A complicated set of personalities often arises to accomplish these tasks, frequently numbering a few tens or scores of distinct alters.


Although alter personalities tend to be distinct entities which resist stereotyping, there are nevertheless patterns of types which appear to be common across DID patients.  Putnam has generalized a set of such types (Putnam, 1986, p 230), varying along lines of functionality and affect.  It is worth noting  them here, as they give us some insight into the organization of multiple personalities:


The Host Personality--  Usually defined as the alter who has executive control of the body most of the time.  This personality is typically depressed, compulsively good, conscience-stricken, and feels overwhelmed by circumstances.  Quite often, the host personality is completely unaware of the existence of other alters until presented with them during therapy.  Hosts may not be single alters, but are sometimes elaborate facades put up by a group of cooperating alters.


Child Personalities--  These seem to be found in virtually every DID patient.  They are frozen in time and do not age.  Such personalities often serve the function of holding traumatic memories of childhood abuse; they are frequently counterbalanced by other child alters who embody pure childhood innocence.


Persecutor Personalities--  These personalities will try to sabotage the patient’s life and even inflict bodily harm.  They are sometimes responsible for “suicide” attempts which are really cases of “internal homicide”, where a persecutor personality attempts to kill the host.


Suicidal Personalities--  As distinct from above, these are personalities driven to kill themselves, as opposed to other alters.  They may be unaware of the existence of the host, or other alters.


Protector and Helper Personalities--  These personalities serve as a counterbalance to the destructive and suicidal alters.  They come in a number of forms, depending on what the multiple requires protection from.  In female multiples, they are often male alters of great physical strength.


The Internal Self-Helper--  Relatively emotionless personalities, who provide information bout the internal workings of the personality system.  Very useful allies for therapy.


Memory Trace Personality--  A passive personality who usually has a more or less complete memory of the DID patient’s life.  Most alters have access to only a severely limited range of memories.


Cross-Gender Personalities-- It is very common for multiples to have alters of opposite sex.  In females, male alters tend to serve in such masculine roles as physical protection and operation of machinery.  In male DID patients, female personalities are older ”good-mother” figures who provide counsel, and tend to be active in the patient’s internal dynamics more than manifesting outwardly. 


Promiscuous Personalities--  These alters exist to express forbidden impulses, often sexual in nature. 


Administrators and Obsessive-Compulsive Personalities--  These personalities frequently emerge in the workplace, where they may be quite competent professionally, although seen by fellow workers as distant and authoritarian.  They may also function to manage the internal organization of the DID patient’s fragmented personality.


Substance Abusers--  Drug abuse in DID patients is frequently limited to one alter, and this alter may be the only one to suffer withdrawal symptoms.


Autistic and Handicapped Personalities--  Autistic personalities are often sent “out” during periods when no other alter is interested in executive control, or especially during situations of confinement or control.


Personalities with Special Talents or Skills--  Alters displaying great skill in work-related, artistic or athletic fields are common, and often tend to be more like fragments than complete alter personalities.


Anesthetic or Analgesic Personalities--  These alters seem to be formed during initial episodes of abuse, and deny feelings of pain.   They are activated when the body is injured by self or others.


Imitators and Impostors--  Some multiples have alters who imitate the functions of other alters.  In some instances, the impostor will handle situations the personality they are mimicking cannot.


Demons and Spirits--  These alters sometimes manifest, especially in persons from rural areas or with fundamentalist religious beliefs.  Spirit alters are similar to Internal Self-Helpers and demon alters are similar to Persecutors.


The Original Personality--  Many multiples have a “core” personality from which the others are derived.  Typically this alter is not active, and is described by others in the system of alters as having been “put to sleep” to protect it from trauma.


As can be seen from the above list, alter personalities tend to assume some sort of functional role, either in the external life of the multiple, or in the complex internal dynamics of the system of alters.  Such diversification according to functionality will be very useful in developing a model of distributed agency in the cyberspace domain.


Causes of Dissociative Identity Disorder


The fact remains, and it must be addressed, that DID is presented as a pathological state, not as a condition of elaborated opportunities for interaction with the world.  To understand the pathological nature of DID, it is necessary to consider its causes.


Colin Ross elaborates four pathways to DID (Ross, 1997).  These are:


1.      Childhood Abuse Pathway

2.      Childhood Neglect Pathway

3.      Factitious Pathway

4.      Iatrogenic Pathway


The Childhood Abuse Pathway is by far the most common pathway to DID, a sad commentary on the ability of our species to terrorize its young.  Ross has an interesting gloss on this pathway to DID: “It seems to me that the fundamental problem in DID is the problem of attachment to the perpetrator.”  (Ross, 1997, p 65, italics the author’s).  The abused child is put in a classic double bind by conflicting biological and psychological imperatives.  On the one hand, the child must attach to its parents in order to thrive, or even to physically survive.  On the other hand, the parent to whom she (by far, most DID patients are women) must attach is abusing the child, typically sexually, as well as physically.  The child is powerless to predict, stop, or avoid the abuse and so falls back on the only avenue of escape available, extreme dissociation.


Dissociation is a skill all children are adept at.  It manifests as deep role-playing behavior in play, as total immersion in entertainment, and as “imaginary playmates.”  These dissociative abilities are drawn upon by the abused child to help keep the attachment systems up and running, so the biological organism does not die.  These attachment systems are personified as separate identities, who idealize the abuser and are amnesiac for most or all the abuse. An additional drive for creating alter personalities is the need  to create stable internal persons who will always be available for comfort and attachment.  This strategy is typically effective to the extent that it allows the child to survive, but becomes maladaptive when the abuse is no longer a factor.


Seen in this light, the cause of DID is not the abuse per se, but rather the bind that such abuse puts the abused child in.  I have been thinking about DID as a model for cyberspace subjectivity for some time, and the heavy involvement of child abuse in the etiology of DID has always left a bad taste in my mouth, like I was basing my model on something inherently evil.  I feel better proposing this model, as I can frame DID as a creative response to a horrible condition, rather than a direct result of that condition.  A subtle, but meaningful distinction.


The Childhood Neglect Pathway typically results in DID less severe in nature and more responsive to treatment than that caused by sexual and physical abuse.  In this pathway, the abuse suffered by the child is extreme neglect, often to the point of being locked in a closet or room for prolonged periods, resulting in the loss of a secure attachment figure.  Children thus abused use their dissociative abilities to retreat into an internal world, and they populate this world with internal figures with which to form attachments.


Factitious Pathway DID is the result of elaborate faking on the part of the patient.


Iatrogenic Pathway DID likewise results from factors other than (at least primarily) childhood abuse.  Poor therapy techniques are at the root of this pathway, creating DID where there was none, or exacerbating mild cases into much worse ones.


It is worthwhile to note that the factitious and iatrogenic pathways to DID do not necessarily involve child abuse, or indeed any abuse at all.  This leaves open the possibility of a form of intense dissociation on the order of DID, but not identical with it, based on a postulated “corrosive” effect of information technologies on the subject stance.


A new kind of self


From the foregoing, we can see that dissociation can be an adaptive response to trauma, one that enables an individual to maintain some sort of integrity in situations where basic survival systems might break down without such intervention.  I believe that the enabling technologies lumped under the heading of “cyberspace” have so altered our relationships with the world that it amounts to severe trauma.  A dissociative strategy will serve to cope with this information trauma, and allow us to thrive and even flourish.


Humans have evolved over the eons as communicating creatures of a certain bandwidth, as dictated by our hard-wired sensory and nervous systems, mediated through the software of language.  Until very recently, we have been interacting with our physical and social worlds through channels which were essentially serial and hierarchical, a mode whose final realization resulted in the totalizing intrusions of the Modernist project.


Our evolution has reached a new phase now, one dominated by cultural and technological forces much more fluid and mutagenic than any previously known.  We have created an electronic skin, a bioapparatus, that shapes our reality and determines the flow of information through our synapses.  The crucial and defining feature of this new bioapparatus is that it operates in parallel mode, presenting us with fundamental incompatibilities with our existing processing structures.


This incompatibility, as I see it, is the source of the information trauma being inflicted by cyberspace technologies, and the way to deal with this trauma is to enable ourselves for parallel processing.  DID provides a paradigm for just such multivalent operation.


As given by Braude above, the defining characteristic of DID is the presence of more than one apperceptive center (alter) at one time (Braude, 1991).  This is all well and good, but most interpretations of DID treat these alters as essentially functionally distinct units.  Indeed, therein lies the pathology of DID: alters do not know what other alters are up to.  Such a completely fragmented world view would model the parallel input streams of our postmodern information realm, but it would be unable to interpret or synthesize any of it.


Braude counters this commonly held view about the distinctness of alters thus: “Nevertheless, certain features of experience seem to presuppose a type of psychological unity that can only be explained adequately in terms of a single underlying subject, even in cases of MPD.”  (Braude, 1991, p 166, italics mine), and “We have seen, not only that an alter’s characteristic functions inevitably overlap with those of other alters, but also that they cannot literally be isolated from a common pool of dispositions attributable to the multiple as a whole.  In that case, however, we need not deny either the reality and functional distinctness of the alters or the reality and functional complexity of the underlying subject to whom the entire repertoire of abilities belongs.” (Braude, 1991, p188)


A practical interpretation of this model of DID can be found in those clinical cases where treatment has stopped short of the usual goal of complete integration.  In these cases, what is left is a group of alters who are now cognizant of each other and capable of cooperation to preserve the viability of the aggregate (The Troops for Truddi Chase, 1990).  Such multiples walk a fine line between unity and multiplicity, and this is the very arena where I place the evolutionary advantage that Deleuze and Guattari attempted to find in schizophrenia.


This, then is the model of DID that I am proposing to deal with the psychic shock of living in cyberspace.  It may even be thought of as an elaborate ruse, a pretending to be multiple, and that is fine as long as the depth of the pretense is sufficient to enable the required level of dissociation, as is certainly the case in DID.


The new, cyberspatial self will employ those same enabling technologies, which the linear self found so threatening, to extend into the parallel domain.  In particular, we can conceive that the various alters in a DID multiple’s repertory might be represented by semi-autonomous agents.  These agents are able to parse multiple input streams for relevance to the “core” personality, and perform some form of buffered parallel-to-serial conversion on any data stream selected.  This conversion will be necessary, because of the limitations imposed by our biological apparatus.  Possibly advances in genetics will enable us to address these limitations, as well, at some future time.


Our dissociated, cyberspatial self is modeled thus:  There is a core personality, a body in space and psyche in information space.  This body is extended, through the bioapparatus of cyberspace technologies, into a realm of enhanced sensory and communication modalities and this enhanced information load, in turn, places greater processing and integrating strain on the psyche.  Parts of the psyche are now off-loaded, however, onto agents capable of acting in behalf of the core personality, and capable of coordination with each other.  Like the alters of a DID personality, these agents will come and go, as required by circumstances of the information environment.


These dispersed agents will be interacting with agents dispersed from other core personalities, and so much of the success of this model will depend on the ability to faithfully encode into them the desires and capabilities relevant to core needs.   This, of course, requires faith that we will be able to solve the problems inherent in creating such “smart” agents.  I believe that current research in “bottom-up” evolution of intelligence has the promise to provide such breakthroughs, however, so I am optimistic that our cyberspatially dissociated self will become a reality.


Part 2: Implementing the LegionWorld test bed


Initial design decisions


The LegionWorld 3D avatar virtual world is meant to test the notion of distributed and fractured subjectivity in the cyberspace domain.  I planned to embody these ideas into the actual structure of the virtual world, much as our assumptions about monolithic identity are built into the structure of normal consensual reality.


For the initial design of LegionWorld, I wanted to create special composite avatars which, rather than being singular and specific entities, were more like a small “gang” or “swarm” of sub-avatars.  Each of these sub-avatars would, in turn, be emblematic of a state of affect and/or engagement, such as Angry/defensive or Excited/engaged.  These states were chosen for initial testing, because they are relatively easy to recognize and represent.  The sub-avatar most closely representing the user’s current state would be at the head of the gang, and the others would recede to a stand-by position in the rear.  Most importantly, however, they would not disappear,  but be always visible as a reminder that many personalities are in constant contention for control.  As the affective state of the user changed, the makeup of the composite avatar would likewise change to reflect that state, by swapping different sub-avatars to the head position.  Persons using such avatars in LegionWorld would be constantly reminded of the fractured and distributed nature of the subject position, and it was my hope that this would rapidly become a natural and intuitive mode of self-representation.


Initial avatar design called for switching sub-avatars according to key words and phrases extracted from the text stream that the user was typing.  In this way, it was hoped that the makeup of the composite avatar would at least roughly track the affective state of the user, without the requirement of deliberately choosing specific avatar representations.  For instance, parsing phrases associated with an active, engaged state would automatically put the sub-avatar representing that state at the head of the gang; if the user’s text output turned morose, then the composite avatar would change to reflect that state.  I decided early on to concentrate on the user’s text, as opposed to text streams from other avatars, as a way of reducing complexity.  Also, it seemed to me that such a strategy would more closely model real life personality switching, which often happens for purely internal reasons, as opposed to being driven by external factors.


The next design decision concerned the arrangement of the sub-avatars to create the required effect.  I decided that a group of three or four sub-avatars would be sufficient for an initial proof of concept, providing enough complexity to be effective, but with few enough variables to be analyzable.  Given that number of components, the composite avatar could assume a variety of forms, including such geometric shapes as squares, triangles and circles, as well as more irregular clumpings.  It seemed that the effect I was striving for would be best achieved by having the dominant sub-avatar up front, with the rest of the sub-avatars in an irregular group in the background.  It also seemed that a certain amount of random jostling among the sub-avatars might add to the verisimilitude of the model.


The issue of avatar movement through space was the next to be addressed.  I decided that the ideal for my model would be to employ some sort of flocking algorithm, so that the head sub-avatar would be free to navigate at will, and the rest of the sub-avatars would follow along the way birds follow a lead bird.  I realized, also, that this might prove too complex to achieve, and I decided as a backup that the group would move as a single unit.


For the first phase of testing, I decided to use three avatars representing the following states: Angry/defensive, Excited/engaged and Shy/disengaged.  I deferred the issue of designing individual sub-avatars until later.


Since the main emphasis in LegionWorld would be on avatar design and behavior, the nature of the 3D environment itself was seen as important, but not critical.  One promising strategy was to set up different areas of the world with artifacts that might evoke affective states, so that avatars could be changing as a result of interactions with the virtual environment itself, as well as from interactions with other avatars. 


Initial implementation decisions


The first, and most critical, implementation decision was to decide whether to create a virtual world from scratch, or to use commercially available world building/browser software.  My first inclination was to create a world using VRML2 and Java, as this would give the greatest amount of control and flexibility.  Indeed, it seemed difficult to realize many of the unique features of LegionWorld, without being able to code them into the architecture of the browser itself.  I was definitely interested in using a browser that was VRML2 compliant, rather than a proprietary system, as I feel that open standards will evolve around VRML2 and I want my work to be a part of that movement.


Starting, as I was, from a position of almost total ignorance in VRML2 and Java, I decided to take a two tiered approach to this problem:  I would begin a crash effort to acquire enough capability to create a world from scratch, but at the same time, I would choose a commercially available system and begin proof of concept experiments immediately.  This turned out to be a wise decision.


I was already familiar with the Traveler software from Onlive! Inc., a VRML1 system with proprietary extensions to make it somewhat  similar to VRML2.  This would have been a natural for a proof of concept platform, except they had quit developing it before it was VRML2 compliant.  Also, making avatars with that system had proved to be remarkably difficult.  I looked at the SGI Cosmo VRML player, but it had no well developed method for dealing with avatars in shared spaces, a problem it had in common with most VRML browsers.


I finally discovered a browser I had not previously encountered, Community Place from Sony.   This was actually a suite of products which seemed to be almost tailored to my needs.  There was the fully VRML2 and Java compliant browser client, but also a freeware server designed to run on Windows95/NT machines, and a simple VRML2 authoring tool.  There were also a few shared, multiuser worlds to visit for seeing how others were developing for the platform.  After downloading the latest version of this package, I decided that this would be my proof of concept development system, and would become my primary development system, if creating a shared world from scratch proved to be an unattainable goal.


Experiences with initial LegionWorld build


While I was deep into studying VRML2 and Java, in preparation for world building from scratch, I had my team start immediately on prototyping possible avatars in Studio3D Max.  I had decided to start with a triad of avatars: Angry/defensive, Excited/engaged and Shy/disengaged.  It was now time to design the sub-avatars which would comprise these three types.  After some debate about various body types and attributes, the team agreed that the avatars did not necessarily have to resemble human forms, or to even adhere to any realistic standards of morphology.  Once achieved, this breakthrough was quite liberating in our discussions of viable body types.


The first sub-avatar to emerge was Shy/disengaged, known at first as Morose/disengaged.  In keeping with our decision to abandon realistic morphology, and since avatars in this world would not walk in any real sense, this first sub-avatar had no legs at all, and very long arms which reached almost as low as legs would have. It had a very appealing expression of resigned indifference, and was colored a medium green with red eyes.  Its rather thin torso was humanoid in the same sense that common alien stereotypes are humanoid, but certainly no more so.  Still, it seemed to perfectly characterize the affective state I was trying to evoke.


The next sub-avatar was Angry/defensive, a close analog to the classic protector personality common in DID patients.  Like Shy/disengaged, this sub-avatar was legless and had very long arms, but was otherwise quite different.  The dominant feature of this sub-avatar was its sheer bulk, which made it impossible to overlook.  Combined with this bulk was a visage suggesting unrelenting obstinacy and a predisposition for confrontation.  Its dull red color and horns led me to think of it as “raging bull”.  The entire team agreed that it was very evocative of its target state.


The final sub-avatar of this triad was Excited/engaged.  Unlike the previous two, this one lacked both legs and arms to connect it to the ground plane, and instead appeared to float free.  Its visage was somewhat reminiscent of an insect, with large dark eyes on a light blue body and a subtly shaded “frill” behind the head.  Its sprightly appearance was quite in keeping with a state of excited engagement, and I found myself calling it “perky”.


The first iteration of the 3D world itself was ready at about this time.  It was a large, flat plane containing an object in each corner designed to elicit affective states.  One corner held a large, jail-like structure, all in gray, containing scenes of death and torture.  I dubbed this the “house of pain”.  Opposite this was a bright blue structure resembling a temple, with a large set of steps in front and a spiral staircase inside, leading up to the roof.  The other two corners were furnished with a portico containing an animated insect, and a spinning 3D star.


The world file, created in Studio3D Max and exported to VRML, did not at first match the scale of the avatars, but was much larger.  After a few rounds of first scaling the world and then the avatars in Studio3D Max, and then exporting them to VRML, I instead began using VRML code directly to change the sizes.  This made the whole process less tedious, needing only a text editor to effect the changes.  Eventually, I decided  to match my avatars to the size of the Sony Community Place default avatars, and then scale the world to fit. 


The first composite avatars were created in Studio3D Max by grouping three subavatars in a rather loose delta pattern, with the head subavatar somewhat farther in front of the other two than they were from each other.  The first tests in the world were most interesting!  One thing I had not thought of during all this time was where the avatar viewpoint was, in relation to the virtual bulk of the avatar.  In my mind, due to overidentification of avatar experience with human experience, the viewpoint was going to be through the eyes of the head avatar.  Of course, that was all wrong, and the viewpoint was really at the virtual “center of mass” of the three subavatars.  I quickly discovered that navigating this avatar through the virtual world felt like navigating an aircraft carrier; the virtual “polar inertia” manifested by the widely spaced subunits was extremely disconcerting.  I also discovered that certain features in the world, in particular the spiral staircase in the Temple, were completely inaccessible, due to the sheer size of the avatar.


By now, I felt competent enough with VRML to dispense with using Studio3D Max for most modifications.  Instead, I began using VRML coding to arrange three subavatars in a group, and modified the size and shape of this group until navigation through the world was as effortless as I could make it.  It was no surprise that the grouping which turned out to be best in this regard was a tight equilateral triangle, so tight in fact, that the subavatars actually interpenetrated each other a bit.  Even at this reduced avatar size, the spiral staircase proved unusable, and was regretfully scrapped.  I decided to balance this optimum navigation against conveying the idea that one sub-avatar was “up front” and in control, by using a tight isosceles triangle configuration.  This had the head subavatar slightly farther from the common center, but not so much so that it made navigation too awkward.


One annoying feature that I discovered was that the viewpoint of my avatars seemed to be somewhat above their  heads, meaning that one avatar could look down at another, even though they were the same type avatar--something completely impossible in real life, and rather disconcerting.  There is a VRML2 node defined to handle this viewpoint, but I could not get it to work.


A change of plans


After spending about two months of intensive study of VRML2 and Java, I began to realize that the project to create a completely custom virtual world, which would implement all my ideas, was simply beyond my capabilities, primarily because of my limited command of Java.  This was a disappointing, but not entirely unforeseen realization, given the time limitations I was working within and the extreme complexity of creating such a client-server system.  I evaluated the capabilities of the Sony Community Place browser and server, and decided that the work I had been doing on that platform could be extended to implement most of the concepts I wanted to address.  This meant that all the work so far created was no longer mock-up work, but rather the real project.


Part 3: Refinement and testing


A project web site


At this time, I began building a web site to present the project to the public.  It has links to the Dissociative Identity Disorder theory informing the project, to Sony for downloading the browser, and a link to download a zipped file of LegionWorld itself.  It also has images of the avatars and screen shots of the world, plus a brief description of the project, to give visitors an idea of what the LegionWorld experience might be like.


Avatar model refinement


It was now time to refine the avatar design, so that it would more completely convey the notion that several personalities were in contention or had potential, but that only one was up front and in control.  So far, the only device I had employed to achieve this was the arrangement of the subavatars in space, and while this did achieve the desired result to some extent, I knew I needed more.  It seemed like I needed to employ some scheme to limit the legibility of the standby subavatars, and after discussing several options with the team, I decided on two such schemes.


The first method was to simply remove color from the standby subavatars, making them a solid medium gray.  The second method was to keep the color information, but to make them semitransparent.  The standby subavatars in the gray version seemed to be lifeless, like statues,  and to have a  relationship to the head subavatar more like a group of objects to an owner.  Those subavatars in the transparent version seemed to better evoke the notion of a potential presence.  Their relationship with the head subavatar felt a lot more dynamic, like they were still infused with life, only in a diminished amount.  It was more believable to me that they might take over executive control of the avatar, but I resolved to submit the matter to user testing, to determine which avatar model I would finally choose.


I was anxious at this point to try extracting words from the text stream, to use for keying avatar selections.  My conversations with Sony about such capabilities were not encouraging, however.  It seems there is no way for their browser to either extract such information, or, having acquired the text through other means, to allow it to effect avatar selection.  This was a great disappointment to me, but I had no choice but to work within the capabilities of the browser.  


As a backup plan, I would have users pick from a list the avatar they felt most closely reflected their affective/engagement states.  How to represent the avatars in this list now became a critical issue.  Community Place has a standard format for choosing avatars, a pull-down list where each avatar is represented by a bitmapped image and its filename.  As a working model for initial development, I had made simple bitmapped logos for each avatar, consisting of just a nickname, either in color for transparent type avatars, or in gray for the gray type.  These nicknames were “Perky” for the Excited/engaged avatar, “Shy” for the Shy/disengaged avatar and “Gruff” for the Angry/defensive avatar.  It seemed that perhaps using such highly loaded words as the only representation of the avatars was not a good strategy, so I substituted bitmapped images of the various avatars themselves, and gave them their old nicknames as filenames, with “T” or “G” suffixes to indicate transparent or gray type.


Initial testing of avatar models


At this point, I felt that LegionWorld was beginning to cohere as a viable project, but that I needed to settle on one form of avatar, either the transparent type or the gray type.  To accomplish this, I tested LegionWorld on two groups of seven subjects each.  After giving them some background about what I was trying to accomplish with representing multiple affective states by these avatars, I demonstrated first the gray type avatars and then the transparent type.  I then asked the subjects to rate each type as to how effective it was at conveying the idea of multiple affective states.  The results of this test were what I had expected, although not as decisively so as I had expected.  In each of the groups, four people thought the transparent type most effective, while three people thought the gray type to be most effective.


Further attempts at avatar refinement


I wanted the group of subavatars that made up the composite to be constantly shifting and jostling, as I thought this would make them seem more like contending entities, rather than passive fellow-travelers.  To achieve this effect I used VRML coding to make a small animation driver for each subavatar, which would constantly shift its place in the group by varying small increments.  I loaded these avatar files directly into the browser to test them, and the effect was quite close to what I had in mind.  The next step was to incorporate these “nervous” avatars into LegionWorld, but at this point I ran into another intractable impediment.  Whenever I chose an avatar of this type, the scene in the browser window would freeze, not allowing navigation and stopping any animation of objects.  Worse than that, after a few minutes, the system would crash, after giving me a “low memory” warning.  Clearly the Sony browser was having serious difficulties with handling these avatars, but I did not know why, and my help requests to Sony and the mail list were fruitless.


At about this time, through my discussions with Sony,  I discovered a proprietary Sony VRML2 node that enabled me to place my avatar viewpoint anywhere I chose to.  Using this proprietary node, Sony_eyepoint, solved two problems, but created two new ones.  It did, in fact, allow me to place my viewpoint properly, at the front of the head subavatar’s face.  Having the viewpoint thus located made navigation in the world seem much more lifelike, and the avatars were no longer able to look down at each other.  Instead viewpoint now corresponded to each avatar’s individual height.


Another feature that I wanted to incorporate into my avatars was a visual cue in the viewport, letting the user know which avatar model was in current use.  I was frustrated at first in my attempt to provide this functionality, however, as the avatars were unable to see any part of themselves. The Sony_eyepoint node, coupled with another proprietary node, Sony_rederingpoint, allowed avatars to see themselves.  I took advantage of this newfound capability to provide a small single-surface VRML rectangle in front of each avatar, color keyed to each individual avatar.  These rectangles were visible from the avatar’s viewpoint, but being single-sided, were invisible when facing an avatar.  They appeared in the lower left corner of the viewport, and functioned quite well for providing feedback on current embodiment.


Using these proprietary nodes caused some unforeseen problems, however.  For some reason, such avatars were unable to see the bitmapped background images I used for such things as walls, ceiling and floor.  The lighting model degraded as well when using these avatars.  Lighting became much darker, and the headlight would no longer shine on other avatars, even though it would shine on other objects in the world.  These side effects were very strange and unexpected, and Sony was unable to help me overcome them.


I was now forced into a decision whether to go with avatars which allowed for proper viewpoint and self-awareness, or with avatars which provided a proper lighting model and background images; the two feature sets seemed mutually exclusive.  I resolved to put that decision off until the last minute, hoping that someone at Sony would be able to come up with a fix.  Otherwise, I was pleased with the avatars, and informal testing on a few people seemed to indicate that they functioned as planned.


World model refinement


During this time, I had the team working on another space to add to LegionWorld.  The first space, which I now called the Plaza, was very much an artificial, “urban” kind of setting.  I was curious to see if users in a more “naturalistic” setting might choose different avatars, or if this mattered at all in what was, in fact, a totally artificial environment.


The new space, called the Island, was a large open meadow, with a grove of trees on one side and a group of rather spiky hills on the other.  In contrast with the Plaza, I did not use any bitmaps for the sky, so the Island appeared to be floating in infinite blackness.  Of course, the Plaza would appear to be floating in blackness as well, if I chose to use the avatars which could not perceive the bitmap backgrounds.


One interesting aspect of the Island turned out to be its size.  It was significantly larger, in terms of virtual real estate, than the Plaza, being at least twice as far across.  Upon testing this space, I noticed for the first time that in the Community Place browser it was possible for an avatar to disappear from view at a certain distance.  When that happened, even though that avatar was still in the world, there was no indication of it, either visually or on the built in “radar screen” locating device.  I postulated that “hide and seek” would evolve as a natural activity in that space.


I now had two separate spaces in LegionWorld and I needed a way to travel back and forth between them.  I created a glowing red sphere to use as a portal object, and I inserted it into the Island by inlining the VRML code, using an Anchor node to attach the address of the Plaza.  I then performed the same operation on the Plaza, while attaching the address of the Island.  It was now possible for avatars to navigate from one space to the other by clicking on the portal object.  As LegionWorld grows, the collection of portal objects will expand to allow for maximum ease of navigation.


In testing the new portals, I noted one feature that tends to break  the model of retained identity across spaces.  The first time a user teleports through a portal to the other space, a default avatar is assigned, the one on top of the pull-down list.  Revisiting any space then results in the reassignment of the last avatar the user employed before leaving that space, even if another avatar was used in the space returning from.  I had much rather see the avatars tied to users and stay constant across spaces, instead of having them tied to the spaces.  Once again, I was forced to bend to the constraints of the browser design.


An amusing incident occurred while implementing the portals.  The Plaza and the Island were built in Studio3D Max using very different scales, and I used VRML scaling nodes equalize them.  The portal object was created for the Island and was about an avatar height in diameter.  When I inlined that same object into the Plaza, however, it turned out to be almost as large as the entire Plaza!  In fact after I placed it, I could not find it, even though I knew the coordinates where it was located.  The problem was, that it was in fact located there, but my viewpoint and everything else was inside  it.  Since solid VRML objects are only visible from the outside, I could not see it until I turned off wall collision and backed out through it.


One final detail was lacking: to provide some kind of ambient sounds in the spaces.  Bart Landry is an accomplished musician, so the task of composing such music fell to him.  The music for the Plaza was a medium tempo piece, fairly neutral as far as any kind of emotions it might evoke.  It had a loop time of about 7 seconds, but the loop was so seamless that the repeat was not noticeable.  The background sound for the Island was a looped wind sound, which went very well with the “wide open” feel of the space.


LegionWorld was now as fully realized as I could make it in the time available to me.  It encompassed two very different spaces, and was populated with avatars which I hoped would convey to its users the idea of multiple subjectivity.  It was now time for the first round of testing to see if my ideas were valid.


Testing the avatar model


I tested LegionWorld on seven subjects, one of whom was familiar with the project and six who were not.  Of those six, LegionWorld was the very first VR experience for three of them, so they had no preconceptions about what to expect.  I tested by installing the testers at one computer running LegionWorld, and myself at another in the same room, but not close.  All communication was through the browser chat window, as if we were in separate locales.  The sessions were logged for later analysis.


I wanted to establish some sort of initial base line by using a tester who was famliar with the ideas behind my project, although not with the details of its implementation.  I encountered this first tester in the world and asked questions about the effectiveness of the avatars in conveying the concept of multiple agency, all the while cycling through the range of available avatars.  I was gratified when that person replied, “I definitely get the sense that I am speaking to a ‘different’ person.  Definitely.”  And also, “ terms of of how it gives one a sense of communicating differently, with a different person/space, I think it works well.”  This was very much in line with the informal testing I had done, and made me feel that my model was valid.


At this point I began testing volunteers who had no experience with my concepts, to see if they would be able to derive those concepts from their experiences in LegionWorld.  As before, I met the subjects in the world, but this time I did not ask about whether my model evoked my ideas.  Instead, I presented the testers with various aspects of the model, such as the different avatars, the objects in the world and the two different spaces, and I asked them how these things affected them.  While the responses varied from tester to tester, there was a large amout of consistency


The first issue I wanted to address was whether the architecture of my avatar model did, in fact, convey the idea of multiple personalities contending for executive control.  I wanted to do that without asking leading questions.  If the tester did not bring up that subject voluntarily, I asked something about his/her reaction to encountering these avatars which were basically small gangs, with some members seemingly fading away.  Most of the testers seemed to get the core idea of multiple representation, which was very heartening.  Here are some typical responses on that subject:  “...kind of like alternate personas not in use, and I know I’m representing as one of them, which seems strange...”, “...are you a clump or what?  Do you have three parts/identities?”, “Puts off the image that there are actually other people there.”, “...but why do you have a group with you?  Who are they, your alternate avatars?...So in a way you have multiple personalities in this world?  One person controls all three cyberbeings?”  This was exactly what I had hoped that my model would elicit from people, so  I was quite pleased with these reactions.


The next phase of testing was to see if the different avatar models were successful in conveying the desired states of affect/engagement.  To accomplish this, I faced off in the world with each tester, and asked them to respond to my presence as I switched from one avatar to another.  If they had no response, I would try to elicit one by asking questions like “How does this make you feel?”, or “How would you feel embodied by this avatar?”.  I started this test phase each time with the Excited/engaged avatar, and then proceeded to Shy/disengaged and finally Angry/defensive.  Here are the results:


The Excited/engaged composite avatar reliably evoked the desired responses.  In informal testing, a few people had questioned my choice of blue as the color for its dominant subavatar, saying that blue reminded them of the blues, and seemed inappropriate.  I even made another version in a lilac color, but the original just looked so perfect that I kept it.  Not one person in the formal test trials objected to the blue color.  The testers, for the most part, read this avatar as being expansive, friendly, silly, and likeable.  Here are some typical reactions:  “...I like the blue self, not the red or green.” “ = the likeable guy caught up in the bad crowd.”, “The little blue guy is very likeable.  Kind of like an alien duck.”, “Looks friendly.  The six protruding things around the neck make it seem clowny.  The big circular eyes and the small mouth are carricatured.”, “You look wide open and friendly.  Big eyes, classic cartoon features.”, “...neutral, like I am talking to a magui.”


The comments above were typical and bolstered my hypothesis that this avatar would convey such a state of  affect/engagement.  One tester responded very differently to this avatar, however:  “...least humanoid, most alarming...but I think I’d feel the most freedom with him...the pupil-less eyes are very effective--soulless!...but perhaps easier to inhabit.”  I found this response very intriguing.  It seemed to indicate that humanoid form was more desirable than I was giving credit for, and that big open eyes were not a universal sign of benign intent.


Next, I tested the Shy/disengaged avatar the same way.  Again, the responses and comments from the testers were generally favorable to my hypothesis about this avatar’s evocative potential.  It was perceived generally as standoffish, remote, cool, even strange and spooky.  Here are some typical responses:  “[it makes me feel] stoned:) tired and it maks me feel sort of ingnored, like s/he is looking right through me.”, “It’s a little spooky, especially the expression it has.  I like it not being connected to the ground.”, “In this one you look like the strange guy in high school behind the bleachers.”(!), “...non threatening, sullen, anorexic...the eyes look vacuous, general feeling of emptiness.”, “I don’t think I’d be sidling up next to this guy for a late night conversation...Just seems very non-friendly.”


These reactions to the Shy/disengaged avatar were a bit more extreme than I had anticipated, but they were absolutely in character with what I had planned, so this was again quite encouraging.


I was confident that the final avatar, Angry/defensive would be correctly percieived, and the testers proved my confidence well-founded.  With one exception, all of them read that avatar essentially the same; as being gruff, aggressive, confrontational.  Testers reacted like this:  “...reminds me of a bull.”, “This one looks very mythical.”, “Yeah, well I think I should be snarly or something...although it’s pleasantly absurd to just be and be repped by a fanged bull!”, “...angry.  I wouldn’t go up to this avatar and start talking...Looks rude and ruthless.”, “...looks like the av wants to start a brawl or a flame war.”, “You look scary.”


Such comments made it clear that this avatar evoked the desired affective/engagement states for most people.  However, there was one notable exception who gave the following response:  “I actually like this one lots.  Seems friendly to me in the same way that a :( face is...the way that a big tough guy is still very lovable / huggable.”  I liked this response, because I had not intended the avatar to be relentlessly menacing, and it was good that someone saw a kind heart behind that gruff facade.


Testing the world model


My testing of the LegionWorld world model was not as thorough as that of the avatar models, as befitting the relative importance I placed on those two aspects of the LegionWorld experience.  I had designed the world to elicit different affective states, however, so I took testers through a tour of both the Plaza and the Island, while noting their comments. 


The House of Pain produced the strongest reactions, which came as no surprise: “..first tell me about that poor bastard tortured.  That is f up.”, “..and the prison?...I feel weird about it being right here.  I feel like if I do something wrong, I’ll end up in there.”, “Looks pretty scary...gray, hanging...are they depressed or what?”  Such reactions were entirely consistent with the nature of that object, and I put it in the world deliberately to provide a “dark side”, which would make the world more complete.  I was not out to create a utopia, at least not for these initial tests.


Also equally predictable were testers’ reactions to the Island.  We always started in the Plaza, and only after a session there would we travel by portal to the Island.  Upon arriving, the general reaction was almost like being on holiday:  “It’s beautiful--I love the flickering trees!”, “The colors are nice, makes me feel refreshed.  Green always does that.”, “...the open space makes me want to fly...”  So it seems that the types of objects, and the nature of the space itself have the desired effects on affective states.  More testing will be required to determine if that translates into users’ choosing “kinder, gentler” avatars.


There were a few unexpected reactions, however.  Two testers just never felt at home in the virtual space, and gave reports like this: “I’m a little lost...I feel lonely--I think I’m the only person in here...”  And: “I have no body///it feels weird not to know my size or shape...Sorry I don’t like this.  I feel uncomfortable.”  Since this was the first time for either of these people to experience any virtual world, I was not overly concerned about how that reflected on LegionWorld in particular.  I think their concerns were about the basic experience of inhabiting a digitaly-mediated space.




These test results were very encouraging.  They indicated to me that the reasoning and design criteria that informed the avatar model were valid, and could serve as the basis for further work in this direction.  Even more exciting, I felt like LegionWorld, even in its current embryonic state, was capable of supporting valid research into the nature of embodiment and agency in the cyberspace domain.


There are aspects of this project which need work.  Testers had the following comments: “It’s very disembodied and how I’d feel if I were a ghost.”, “...again I need to see myself.  I should at least be able to see my own shadow(s).”  This is not really what I had in mind; I was after a feeling of enhanced embodiment.  There needs to be a robust method for giving users feedback about their current avatar model.   Possibly, replacing the color-coded rectangle with an actual avatar image would solve this problem.


Given the opportunity to further develop this project, I would try to build it without recourse to a proprietary browser, so that I could enable the parsed-text avatar switching mode.  That lack is a very serious flaw in LegionWorld in its current state, but in many other ways the project has lived up to, or even exceeded my expectations.  The efficacy of the composite avatars in evoking their discrete affect/engagement states is gratifying.  The way in which these composite avatars convey the notion of multiple, parallel idenities is even more gratifying.   The effect seems to be quite real, reproducible and intuitive.


 I believe I have made a small, but significant, step toward enabling humans to survive, and even flourish, in an age characterized by massive informational truama.





Braude, S.E. (1991), First Person Plural,  Routledge, London


Deleuze, G. and Guattari, F. (1977), Anti-Oedipus: Capitalism and Schizophrenia, Viking, New York


Kluft, R. P. (1984), An Introduction to Multiple Personality Disorder, Psychiatric Annals, 14:51-55


Putnam, F. W. (1986), Diagnosis and Treatment of Multiple Personality Disorder, The Guilford Press, New York


Ross, C. A. (1997), Dissociative Identity Disorder,  John Wiley and Sons, New York


Suryani, L. K., and Jensen, G.D. (1993), Trance and Possession in Bali: a Window on Western Multiple Personality, Possession Disorder and Suicide,  Oxford University Press, Oxford


The Troops for Truddi Chase (1990),When Rabbit Howls,  Jove Books





Thomas Vernon Reed was born in Pine Bluff, Arkansas on June 26, 1948, the son of Harleen Roberta Reed and Thomas Russell Reed.  After completing work at Pine Bluff High School, Pine Bluff, Arkansas, in 1966, he entered the University of Texas at Austin.  He received the degree of Bachelor of Arts from the University of Texas at Austin in December, 1971.  From 1972 until 1993 he pursued a career as an artist, having work hung in such prestigious venues as the Louvre in Paris, the Smithsonian Institution in Washington, D.C., and the Museum of Modern Art in Tokyo, Japan.  In September, 1993 he returned to the University of Texas at Austin to pursue further undergraduate studies.  In September, 1995, he entered the Graduate School at the University of Texas at Austin.  He is a member of the Phi Kappa Phi honor society, and of the International Interactive Communications Society.


Permanent Address:            5902 Haydens Cove

                                    Austin, Texas 78730


This report was typed by the author.