The Computational Self
                by Robert M. Galatzer-Levy, M.D.

This is a paper original delivered at the First Annual Mathematics
and Psychoanalysis Meeting in New York, N.Y. on June 6, 1988. Any
comments are very welcome.

	What I have to say today is more by way of posing a problem
and indicating an area where I suspect the solution to lie than a
coherent presentation of a new theory. I am going to talk about
some elementary ideas from a branch of psychoanalysis called self
psychology and some elementary ideas from computer science that
seem to me to provide a framework for thinking about the self of
self psychology and then invite you all to let me know whether what
I have said has made sense and whether you can see directions for
the development of these notions.
	Freud's effort to explain mental life on the basis of drives
that are the psychological representations of biological
disequilibria fell on hard times as he tried to work out the theory
in detail. He introduced a new entity, the ego, dangerously close
to a homunculus within the mind, that performed certain functions
and vigorously protected itself from being "overwhelmed" or
traumatized. The ego's functions included managing the persons
relation to reality, regulations of drives, object relations,
thought processing, defensive functions, perceptions and motor
activity, and integration of all other psychological functions -
its so called synthetic function. 
	The concept of the ego became the center of American
psychoanalytic theory in the forties ,fifties and sixties. Despite
heroic, rigorous efforts to sharpen the terms's meaning, the
confusion Freud left between the ego and the subjective experience
of the self continued. This persistent confusions was not merely
the result of intellectual sloppiness. Nor was it, as Bruno
Bettleheim proposes, the result of Freud's English translators'
discomfort the soul-like implications of the Freud's original idea.
The difficult is more fundamental. The terminologic and theoretic
confusion reflected a clinical reality.
 	It often happens that people who functioned badly in the areas
called "ego functions" also have major disturbances in their
experience of the self and that the two types of difficulty are
exacerbated or diminished in concert. The idea of the ego as a
unitary entity is not just as a convenient, if confusing, name for
the set of functions described earlier, a sort of waste basket for
what is neither id nor superego. The term reflect the commonly
observed covariation in these functions. 
	The systematic exploration of the self experience began in
psychoanalysis in the years following the second world war, though,
of course the concept of self has been the object of study since
the dawn of civilization. Although he had significant
psychoanalytic precursors, notably in the work of Paul Federn, Erik
Erikson was the first to propose that the core of much
psychopathology lies in disorders of self experience. Erikson's
concept of identity, which amalgamated the many sources of beliefs
about who one is is both evocative of common experience and proved
clinically useful. Many kinds of difficultly, as well a normal, and
supernormal psychological development can be usefully explored as
experiences of loss or diffusion of identity or attempts to
establish a satisfactory identity where one was lacking. 
	Erikson's work is problematic from a psychoanalytic point of
view for two reasons. First, reading Erikson carefully one
discovers that his wonderful portrayal of emotional states through
imagery, metaphor and clinical detail is not matched by explicit,
clear theoretical formulations. Second, his writings often focus
on external environmental effects rather than people's
psychological worlds and the manner of their construction.
	Erikson never systematically described his therapeutic
approach to his patients. However, it is clear that he consistently
placed a positive connotation on his patients' struggles. He
demonstrated how manifest psychopathology could be understood as
potentially successful attempts to achieve valuable identities,
that while there might be difficulties in the way the patient's
basic project and his ways of attempting to accomplish it were
closer to healthy development than the patient or the society might
recognize. Erikson's psychobiographical studies of Luther, Gandhi,
Hitler and Shaw are messages to readers, many of them young, about
the value of their struggles to form workable identities. Erikson's
implicit view is that an appreciative stance toward the patients'
struggles which include or dominated by external realities is
	In the years following the second world war Harry Stack
Sullivan, observed that the experience of the self of many of his
schizophrenic patients was grossly disturbed. Borrowing from the
Chicago School of Sociology, most notably George Herbert Mead,
Sullivan conceptualized the self as a summation of social roles,
some of them retained without full awareness from archaic periods
of development. In this "interpersonal theory" of psychology
pathology resulted from a self system that was internal incongruent
or problematic in terms of the environment. Therapeutic
intervention consisted in understanding and appropriately revising
the self system in the light of more mature and current
understanding. What is central to our discussion is Sullivan's view
that the self system was both the product of the external
environment and made no sense whatever outside of a social system.
	Several analysts, notably Klein, Winnicott, Khan, Fairburn,
Bion, Spitz, and Modell emphasized the role of holding environment,
environmental container or the "mother" in the development of the
self. From their very different perspectives each emphasized how
the self's growth required an external situation of being "held"
as the emerging and vulnerable self gained strength and autonomy.
People whose psychopathology centered in problematically self
development, a condition that all these authors equated with
difficulties in the first two years of life, could work out their
problems if provided with an analytic situation that allowed them
to reengage those phases with the analyst experienced as the
archaic maternal environment of that era. Some of these analysts
believed, like Melaine Klein, that these very early situations
involved inherent conflicts that now be resolved through
interpretation in analysis. Others like Donald Winnicott held that
new experiences, "beyond interpretation," with a good-enough object
were needed so that the developmental failure could be righted
through new development. The theoretical formulations of many of
these authors was either so inherently fantastic, or so abstruse,
or so unsystematic that their work has had relatively little
influence on psychoanalytic theory beyond the range of their
immediate followers. It is only now being integrated into the
mainstream of psychoanalytic thought.
	Margaret Mahler and her coworkers also concerned themselves
with he early development of the self. They centered their
attention on the era of late toddlerhood that involved the
difficulties of the child emerging from a state they called
"symbiosis" in which the experience of the self includes the care
taking environment into a state of being an individual in one's own
right. Based on treatment experiences with youngsters and adults
who seemed to have difficulties in the area of the self experience
and observations of toddlers, which unfortunately were dominated
by their preexisting theory, Mahler and her group concluded that
much of the difficulty in self experience arose from a failure to
adequately separate from the mother of infancy. Although there are
many well informed analysts who would disagree with me, I will
assert that the overwhelming data of infant and later developmental
studies demonstrate that Mahler's symbiotic phase is not part of
normal development nor is separateness, in the sense she meant it,
characteristic of ordinary or healthy more mature psychological
	However the clinical observations that lead to Mahler's
thinking and that have been explained in terms of her theories are
certainly common. That is, there are many people who seem to have
shaky experiences of themselves and function with a conflicting
notions that on the one hand they desperately need other people if
they are to function at all reasonably and that some core aspect
of themselves is in danger precisely in these urgently needed
	Starting in the sixties in Chicago Heinz Kohut initiated a
psychoanalytic study of disorders of the self. His approach to
these researches was methodologically distinct and is worth a
moment's pause. First he took a radical position, that he claimed,
incorrectly at the time, to be a standard one, that psychoanalytic
data was collected in a manner different from that of the natural
sciences. He asserted that it is possible and usually to
immediately comprehend complex psychological configurations in
others and that such understanding ordinary mode of operation of
the working analyst. Empathy for other's internal states, as a mode
of comprehension, was, for Kohut, similar to the way we perceive
faces - as a complete and immediate gestalt. Analytic training and
technique are designed to maximize the analyst's ability to use
this investigative tool and to overcoming its pitfalls, just a
training in microscopy enables us to vastly extend ordinary visual
capacities. While Kohut claimed to be making explicit what everyone
did anyway, his position, right or wrong, was deeply antithetical
to Freud's view of psychoanalysis as a natural science-like
investigation and also Hartman's explicit statements that empathy
in the sense that Kohut meant it had no appropriate role in
psychoanalytic investigation. 
	A second, and less problematically, position about
psychoanalytic investigative method was Kohut's position on
transference. In his early writings on self psychology Kohut
assumed that the only data to be taken seriously in psychoanalysis
were the data of the transference. The various stories the patient
told, the analyst's conceptual framework and responses and all the
other stuff the analyst commonly use to frame a picture of the
patient's psychology was of minimal importance compared to the job
of describing and understanding the interaction between patient
and analyst. Kohut also believed that premature interpretations to
the effect that the patient was avoiding knowing something about
himself often interfered with the full blossoming of the
transference. According to Kohut, premature interpretations,
particularly premature interpretations of defense often resulted
in the analyst discovering evidence that confirmed their
preexisting notions because they misunderstood possibly contrary
clinical facts as representative of the patients' avoidance of
already known realities.
	Using empathy and the exploration of transference as their
primary tools, Kohut and his students treated a group of patients
whose distress took three overlapping forms. One group of patients
suffered from feelings of depletion, emptiness, triviality and/or
fragmentation. These experiences often took symbolic expression in
the form of hypochondriasis. Another set of patients were engaged
in activities that seemed enormously driven or addictive such as
sexual promiscuity and perversion, shop lifting, desperately
clinging relations to other people and substance abuse. Finally,
some of the patients had chronic and acute states of tantrum like
	In analysis, at least as conducted by Kohut and his followers,
these patients developed characteristic attitudes to the analyst
that Kohut labeled selfobject transferences. Characteristically,
often against considerable internal resistance, these patients came
to experience the analyst as essential to their well being. His
physical or psychological absence variously precipitated great
distress and/or the reemergence of symptoms that had been
previously remitted. For example, a young man who had entered
analysis much distressed by his promiscuous homosexual behavior
reported what for him was a major business success during a
session. The analyst, noting that the patient's anxiety had
interfered with an even greater accomplishment, made the plausible
interpretation that the patient had inhibited himself from doing
even better because he experienced his business competitor as like
the analyst and feared the analyst's reprisal if the patient beat
him in competition. The interpretation was bolstered by several
significant details that made it plausible and the patient thought
it was "right on the mark" and promised "to try to do better next
time." Retrospectively he said he had felt irritable and "headachy"
during and immediately after the interpretation was given. That
evening he returned to a gay pornographic movie theatre were there
was much sexual activity among the patrons. Before the analysis
this was one of his regular haunts but he had stopped patronizing
the theatre many months before. The patient allowed several men to
perform fellatio on him. He felt angry and painfully excited as he
thought the fallators really appreciated what he had. In response
to what he felt was the analysts inadequately appreciative response
the patient had desperately turned to a more concrete indication
that someone could appreciate his accomplishments. Another patient
experienced every weekend as "like being sent away to live in the
Sahara in a desert" and the return to the analysis as "like coming
back to the oasis." 
	When their feelings are not interrupted these patients like
these experience the analyst in characteristic ways that Kohut
described with oversimplifying systemticity. Some patients
idealized the analyst seeing in him the embodiment of strength and
good and feeling alive and whole in his presence. Others find
relief in the sense of being in a unity with their analyst, or
being like him or being appreciated by him. Interruptions in these
states of mind commonly bring with them inordinate distress or
symptoms which could be reasonably understood as experiences of a
fragmented or devitalized self or attempts to avoid those
	From these clinical experiences Kohut posited that there were
a group of people for whom the maintenance of a satisfactory self
experience was centrally important because it was so problematic.
The analyses of these patients was characterized by the use of the
analyst to maintain the a cohesive and vital self by using the
image of the analyst as part of the self or as a support for the
self. Any interruption in the capacity to use the analyst in this
manner lead to the reemergence of problems in this area. The
situation within the analysis was equated with postulated normal
developmental states in which the caretaker ordinarily performs the
functions for the self. These functions Kohut called selfobject
functions and he believed his patients to be suffering from
disorders of the self resultant on traumatic failures of early
selfobject functions. As in normal development small, empathically
supported, failures in the selfobject function allow patients to
identify with the image of the way the analyst should have
functioned and to make those functions more their own. However
mental health does not consist in giving up self objects. Kohut
asserted that selfobject functions normally continue across the
course of life and that it is their qualities, not their existence,
that is altered with maturity. (Having made this assertion Kohut
never elaborated or demonstrated it. Recently Bertram Cohler and
myself have undertaken the task of exploring the empirical evidence
for Kohut's position.)  
	Kohut's findings, and the findings of many of those who have
examined the psychology of the self from other viewpoints, have
been questioned in too apparently distinct ways, whose
interconnection I will show you in a moment.
	The first objection is that Kohut's theories serve to avoid
painful psychological truths. Many of the phenomena Kohut observed
had been observed previously and classified as defensive
operations. For example, idealizations of the analyst were commonly
understood as ways both to avoid knowing of the unconscious
demeaning of the analyst and to arrange for disappointments when
the analyst fails to live up to the idealization as he inevitably
must. The idea that the patient "needs" the analyst to function in
some certain fashion lest his core being be seriously damaged could
be understood as a fantasied misunderstanding designed to
rationalize wishes whose non-fulfillment may be extremely
frustrating but not inherently, must less psychologically fatally,
	The second set of objections has to do with the theory of the
self. Kohut never clearly defines his central concept of the self.
Essentially he says that everyone knows from experience what the
self is and leaves it at that. After studying the many discussions
of the meaning of the "self" in the psychoanalytic literature one
is reminded of the Buddha's comments on the self. He said that
those who believe in the self are like  "a man who says that he is
in love with the most beautiful woman in the land, but is unable
to specify her name, her family or her appearance" (Digha Nikaya
I 193, quoted in Carrithers (1983).) The essential theoretical
difficulty was clarified by Meissner who pointed out that the term
self as habitually used by Kohut and most other writers whose work
places the self at the center of psychological life, is
consistently used to refer to both a psychological representation
and also a psychological agent. Although more systematic
researchers, for example Hartman, limit the concept of self to a
psychological representation of the person, they also give the self
a markedly subsidiary role in psychology. Meissner's argument is
quite similar to Schafer's later discussions of internalization in
which Schafer observed that the elaborate analytic theories of
internalization were in fact nothing more then the translation into
psychoanalytic jargon of unconscious fantasies and did not, in his
view represent, represent actual psychological mechanism and in
fact obscured, what actually happens when we have experience that
had been described as the taking in of another person or aspects
of that person. 
	The two problems with self psychology, its use as a defense
against painful insight and its confusion of agent and image, are
related. Notice that if the self is "only" a psychological
representation it would follow that the patient's idea that had
will be dysfunctional as a direct result of some impairment in this
representation seems mistaken - or at least so it seemed to many
thoughtful psychoanalysts. Only the impairment of some mental
agency could really result in dysfunction. It was if the patient
complained that his car did not function because part of a picture
of the vehicle had been obliterated. The idea that the patient is
in error in this regard supports the clinical stance that the
patient's fears in these matters are not an accurate assessment of
the situation but rather fantasies motivated by their unconscious
desires to hide deeper psychological realities.
	Now of course we all know that there are "mere"
representations that are very good for actually doing things and
whose faultiness causes no end of problems. These representations
are called programs.
	Now, I suspect that once stated the notion that the self is
a program which like other programs is capable of change by
altering its representation and at the same time is an active agent
is neither a surprising or remarkable idea. However, when one
notices that fifty years or so of both clinical and theoretic
psychoanalytic thinking about the self has been profoundly
influenced by the idea that the existence of such an object is a
logical impossibility the point seems more worth making. The other
advantage of making this point is that it invites us to use what
we know about programs to think about the self and suggests the
systematic characterization of the self as a program.
	Let us begin the selfobject function whose enemies are want
to equate it with some form of mysticism. We know, of course, that
programs have meaning and function only within computational
environments. An inappropriate computational environment can alter
the meaning and operation of the program or render it altogether
meaningless. For example a routine that calls a global variable
gives a different value depending on the value of that variable;
a program written in C for which one has no compiler is totally
useless. The use of the term "computation environment" in computer
science is relative to the process being discussed and only has
meaning once one specifies what program is being referred to. An
expression only has meaning within an environment. Having bound a
global variable that value then becomes part of the computational
environment of the programs running within that context. Of course
from a different viewpoint the program that sets up the environment
for our first program itself has an environment. Thus ordinarily
we expect that program will "need" appropriate environments in the
same way that self psychology predicts that people need
	What one chooses to call program and what environment
obviously effects the picture of the situation that emerges and is
a function of the interest of the investigator. Similarly the
boundaries of the self depend on the point of view we adopt based
on the focus of our interests. It is only important to notice that
the choice is ours, not intrinsic to the system under study and
that it is important not to become confused about the principles
governing the entities we have defined. A few decades ago von
Bertalanfy made a minor industry of pointing out the inappropriate
application of conservation principles to "open" systems that were
mistakenly treated as having no energy flux across their
	The mechanics of the selfobject or the environment is
naturally important but by no means definitive in terms of its
function. In one since it is obviously of considerable importance
whether a subroutine that is called is available in RAM, is
currently located on a easily accessed storage device or is located
on a tape that the machines operator must fetch and mount before
it can be used. In another sense these mechanical considerations
are of minor importance in our understanding of the program.
Likewise whether the capacity to be soothed is a readily available
group of psychological functions represented within the cranium,
the activity of a caretaker who is but a cry away or requires some
elaborate undertaking - say a few years of psychoanalysis - can be
regarded as involving no essential difference in this function.
Although he never would have put it in this way this is an
essential aspect of what Kohut was trying to point to in the idea
of the selfobject - something that functions as an essential aspect
of the self or of the support of the self but which because of the
mechanics of its availability is at times less efficiently
accessible than other aspects of the self that we are more
accustomed to including in our idea of the self. This computational
relative inaccessablity commonly is associated with the need for
particular perceptual inputs and computational assistance.
	For instance the phenomenon of "social referencing" has been
studied extensively from a social psychological point of view.
Starting at about age seven months given a novel situation or a
situation with elements that suggest danger babies look to
caregivers for cues about whether to proceed and base their actions
on the caretaker's response. Toddlers as they move away from mother
in a play ground frequently turn around, checking mother's
expression before proceeding further. In the toddlers experience
the decision does not take its basis in the issue of whether
mother, as a person approves or disapproves of the action, rather
the mother's approving response registers as an impersonal "It is
okay." The toddler has not called the person "mother" in this
situation but has rather expanding his computational resources
which happen at the moment to be located in the being we would
refer to as his mother. The child needs loves nor hates the mother
in this context but does need her to function. If she is
functioning well like any computational resource he remains unaware
of her presence. It is only her failure of availability that makes
her of interest, just as we are generally unaware of our memories
except when we have difficulty recollecting something we need to
continue our thinking.
	Those of you familiar with Marvin Minsky's work recently
summarized in The Society of Mind will recognize in these ideas a
particular application of the multi-hierarchy computational model
that can be used to explore processing within many levels of human
function from neurons to societal organizations. The issue of a
non-pejorative attitude to what we call mysticism comes to mind
here. Much of what is referred to as mystical might well be
considered as attempts to comprehend hierarchically higher
computational structures within the computational world of lower
order entities. 
	The self as a program does two important things that are the
subject of our constant attention in our analytic work. The program
monitors its own operation and ordinarily modifies itself in
response to such monitoring. The type of programs we are familiar
with in daily work with computers generally have facilities to
monitor and modify their own execution to a limited extent. Error
trapping of one type or other is virtually universally employed so
that unexpected or undesirable situations do not result in the
continuation of "business as usual" but instead lead to some kind
of branching in the process. In an "error" situation the new
execution often takes the form of enlarging the computational
environment to include the operator who is asked how to proceed or
to correct some situation that impedes the computation or to
authorize the use of additional computational resources.  For
example if the execution of a program requires more than a certain
amount of time the systems operator may be asked whether to
continue or abort the execution.
	Similarly, but much more extensively, the self is engaged in
a constant process of monitoring its own function and functional
needs, arranging for them to be met or attempting to compensate for
their not being met. We have already implicitly discussed the
ongoing monitoring of computational resources and the recognition
of the need to evoke devices such as the perception of other people
to serve as selfobjects. The detailed study of the nature,
functions and situations in which these additional computational
devices are called or where calls to such devices is avoided
constitutes a major area of psychoanalytic investigation that
encompasses much of object relations theory, including self
psychology, attachment theory, the concept of the transitional
object and the role of cultural experience.
	In the von Neumann architecture computer design was dominated
by the wish to avoid programming errors. This was accomplished by
carefully separating data, programs and processing functions and
forcing sequential processing so that except in terms of the
overall duration of computation the outcome of a computation was
unaffected by the time required for each computational step.
Furthermore building this basic architecture requires the
anticipation at least the basic architecture of the system from
its beginning. It cannot result of the evolutionary piecing
together of elements designed for other functions as the brain must
have evolved.
	The von Neuman architecture is so excellent an environment for
humans to design programs for that it dominated computer design for
almost four decades. However as von Neumann noted from early on
this architecture is a poor model for brain functioning. The
microsecond firing times of neurons are much to slow to allow
brains to do the things they do all the time with a von Neumann
machines. Furthermore brains are the result of a bioevolutionary
process, not a unitary design and its programmer is not an
individual who sets out to explicitly specify processes but an
environment with many other things on its mind than programming
brains. Of course we know from direct study of brains that they
operate through massively parallel processing. 
	Fortunately for those of us interested in brains and their
productions it has become clear that the technological limitations
inherent in the von Neumann architecture make it essential that
other architectures be explored in depth to make more capable
computers. The last five years has seen an explosion of
publications about parallel processing architecture and we will be
among the beneficiaries of the resultant intellectual advances.
	But, of course, the problems that von Neumann sought to avoid
in computer design are precisely the problems that emerge in
parallel processing. It is simply much more difficult to predict
what is going to happen when things do not go on sequentially, when
the distinction between memory and processing is abandoned and
simple hierarchies of bindings are abandoned. Now rather then
building the absence of these difficulties into the architecture
of the system it becomes necessary to discover ways to overcome
them. A much more elaborate system of error trapping and control
becomes essential.
	Parallel systems are highly vulnerable to internal conflicts
and instabilities. Attempts to remove these features from the
system usually entail the loss of precisely what has been gained
through parallelism. To give an very elementary but quite everyday
example, when a database can be updated through several different
inputs there is considerable danger that attempting simultaneous
updating of a record will result in loss of data or undesirable
results. Suppose I am making a deposit in my savings account at the
same time that interest is being calculated and recorded in the
same record. In many database systems the entire record is
retrieved updated and stored again. So in this instance the
original record is retrieved by both the deposit and the interest
function. Each, independently updates the record and then writes
it to the storage device. Either the deposit or the interest
payment, whichever is stored last, will be recorded but not both.
A simple solution that is used in many database systems is to make
the record available to only one potential input at a time by
locking it to other users while it is in the hands of a potential
inputter. In essence one suspends parallel processing and goes to
sequential processing in the face of such potential errors. This
is an awful solution for simple database management, although as
anyone who has worked with such a system knows it can be thoroughly
annoying. But such a general solution for a massively parallel
system would slow the whole thing to a snails pace. Thus special
mechanism for recognizing, protecting against and resolving
conflicts are expected to be a central aspect of massively parallel
	But notice how close we have gotten to the ordinary stuff of
psychoanalytic clinical work. A lot of what we do in analysis has
to do with successes and failures to resolve conflicts between
computational results achieved through parallel processing of
situations. To give a much oversimplified instance, a young man
who might displace a supervisor by putting forward his own ideas
expresses them but muddles their presentation. Analysis reveals
that his actions result from two parallel, conflicting computations
and an attempt to resolve that conflict. On the one hand are a
variety of factors including his wish for greater prestige and
material wealth that in turn reflect a long sequence of
developmental processes and on the other his assumption (which is
outside of awareness) that he will be harmed in various ways if he
pursues these wishes results in a state of conflict. This conflict
and potential conflicts are dealt with variously by some higher
order resolutions or through the isolation of the processes from
one another by a variety of means. The resulting action,
unfortunately called a "compromise formation" in psychoanalytic
jargon is an attempt to synthesize the results of these two groups
of computations.
	An even greater danger to the system than partially
contradictory computational results is its own instability.
Computational process may become chaotic, disorganized or pass
through a catastrophe as we recognize in depth when we study them
in terms of dynamical systems. It is reasonable to expect that a
computational system can only function in anything like a
satisfactory manner if such situations is rigorously limited to
lower levels of function and if the system has extensive safeguards
against higher level catastrophes or chaos. 
	Again this is precisely what we find clinically. The most
central concerns in disorders of the self frequently are concerns
about discontinuous and disorderly change. A typical error trapping
procedure in the area where catastrophic change seems a danger is
to avoid all change whatsoever and to attempt to isolate the
computational processes from outside influences that might result
in change. Recently I described how the process of working through
in psychoanalysis, the repeated reexamination if slightly different
versions of paradigmatic situations within an analysis, could
usefully be regarded as the reestablishment of a Boltzman
algorithm-like psychological function by which existing "solutions"
are repeatedly and automatically reexamined both to achieve greater
optimality and to integrate data that may have been unavailable at
the time they were formed. I said that much psychopathology could
be usefully characterized as the interruption of this ordinary
process in the face of a perceived threat of disruption or
disorganization and that what we often think of as the curative
factor of working through is just the resumption of normal
psychological function.
	This brings us to the third way in which the self differs from
the programs we are most familiar with from the study of computers.
The self is self developing. Here my opinions are somewhat
different from many of my psychoanalytic colleagues, so let me
spell them out briefly. As she attempted to explore the concepts
of normality and pathology in childhood, Anna Freud discovered that
the presence or absence of symptoms per se was not an adequate
guide in assessing children. She concluded that childhood was
normatively a period of change and development and these were its
primary tasks. The failure of such for such development to be
ongoing was the essence of psychological disturbance in childhood.
For Anna Freud, who had a clear picture of what psychological
health was like in adulthood, the task of childhood was move toward
such mature functioning and she posited a drive to "the completion
of development." 
	Three groups of observation impressed me into extending her
notion. First the past quarter century has yielded a massive
demonstration that human development normal continues across the
entire life course - that the idea of a definite mature
developmental state whether occurring with the resolution of the
Oedipus complex or the end of late adolescence or whenever else is
mistaken. Second there seem to be quite diverse ways to be
psychologically healthy which becomes readily apparent if we avoid
employing a priori notions of the meaning of health. Finally the
work begun by Marsh to the effect that programs can be written not
with specific goals in mind but rather that proceed to explore and
develop in area that are vaguely defined by such criteria as
"interestingness" corresponded so well to the observations of
workers like Piaget who found that exploration and development were
self motivating that it seemed likely that the human mind is such
a system. It thus seems reasonable to posit that an ongoing
function of the self is its own reorganization and development.
	Indeed it was this point that first led to my interest in a
computation model of the self because the question of how the self
could be both agent and representation and in particular how it
could be an agent acting on itself as a representation has a long
standing concrete instaniation in Lisp. Lisp, one of the two oldest
high level programming languages in common use, was specifically
designed to manipulate list of symbols. Of course lisp programs are
themselves list of symbols so that lisp programs can be operated
on my lisp programs including the program itself. The species that
seemed so internally contradictory that analysts denied there
existence have in fact been around for a long time. 
	Now, of course such programs are not without very serious
problems - in particular they too can be much less stable and far
less predictable than those programs were program and data are kept
strictly separate. As with parallel processing one way to protect
from the dangers inherent in such a structure is to carefully limit
in advance the changes the program can make in itself. Another
possibility is to monitor the development of the program and
introduce error trapping and correction as untoward consequences
of the rewriting occur. A combination of the two approaches would
seem to be necessary. In a sequential system for example a fatal
error occurs if a real interminable loop is introduced into a
program. Here, however, parallelism and conflict can be of
considerable help. Freud's idea of a tripartite model of mind
essentially involves the parallel processing of data, the
consequent development and resolution of conflict so that a variety
of needs can be met through these various modes of processing. In
particular aspects of the mind can monitor the ongoing process of
the development of the self - interrupting and altering it when it
comes parlously close to instability or stagnation. 
	The hierarchical level at which these process can proceed are
various and new levels in the hierarchy seem to develop with
greater maturity. In particular greater capacities for abstraction
both from data and process appear to be a normal part of human
development. With these capacity comes increased abilities for
metacognition. Piaget's observation of the progressive decentering
of cognition with the related capacity, for example to think about
thinking, represents such an elaboration of abstraction
	Among the many objections that could be raised to my
discussion is the importance I lay on introspection and
subjectivity as a source of information about psychological
processes. From a computational viewpoint consciousness is an odd,
unnecessary, or at least peculiar phenomenon, while from the point
of view of classical psychoanalysis precisely what is most
interesting about people is barred from the conscious awareness.
Thus subjective reports about experience should be relatively
uninteresting to both groups. However, following Vygotsky and
Basch, I take a different point of view about consciousness.
Consciousness is a state that we employ when automatic functioning
becomes problematic. For example we only become aware of walking
when we stumble or when we are learning how to do it and only
attend to it in detail if something impedes are ability to walk.
It is thus precisely in areas of difficulty that we expect
awareness to appear. So it is the areas of difficulty that we
should find well represented in consciousness. Freud's idea of
bringing the unconscious into awareness then is nothing more then
the extension of this normal process into areas in which it is not
employed. In particular the mechanism of repression reflects a
special procedure to keep ideas separate from each other by not
bringing them into awareness. But more generally we can use
subjective experience as at least a preliminary guide to the
computational difficulty.
	I am well aware of having painted the picture of the
computational self with extremely broad strokes and having done
violence to many subtle and important issues in the process. At the
same time I am impressed that psychoanalysts having discovered that
the Freudian and ego-psychological paradigms are inadequate have
largely abandoned the attempt to develop broad theories that
encompass the particular data of the psychoanalytic field, choosing
instead to focus on smaller more tractable problems and maintaining
an unavowed theoretical agnosticism. 
	An exception to this abandonment of theory lies in the work
of the self psychologists. However their conceptualizations,
especially those of Kohut, while evocative remain vague. I think
it is clear that the computational properties of the mind must find
representation in personal psychology. I have suggested one
possibility for how this may occur using the computational self as
the central organizer for my thinking and attempting to show how
ideas from computer science may yield models that are congruent
with our clinical experience. Just as I believe development is the
central activity of the self so to I believe development should be
the central goal of our intellectual activities. Thus if this
paper, despite its flaws does nothing more then stimulate some of
you to think along these lines and to help me do so more cogently
I will be satisfied.