A Model of the Psyche (fifth draft )

Roy D'Andrade

UCSD May 1 1999

INTRODUCTION

Over the past thirty or so years no general model of the way the human psyche is organized and operates has had widespread currency. A general model has great scientific value; it helps put together findings from different areas of research, it forms a target for testing and advance, and it gives a sense of unity and common enterprise to a field. In the forties and fifties there was general model that incorporated psychoanalytic ideas and learning theory which fulfilled this role. This model centered on the notion of 'drive.' However, in both academic psychology and psychoanalysis the drive concept became increasingly problematic. The cognitive revolution, the rise of object relations and self theories in psychoanalysis, the great increase in knowledge about the brain, and the work in artificial intelligence all relegated drive related phenomena to a peripheral position. Systems, feedback and control, cognitive representations, neural models, and distributed processing became the central organizing concepts. A new framework has been a developing, starting since at least since the work of Bowlby in the early seventies. This model has been able to incorporate new findings about the brain and human evolution, and incorporating new ways of modeling systems. This paper is an attempt to continue this line of development and indicate something of its applicability to the study of culture and society.

SYSTEMS

One of the basic problems involved in trying to formulate anything about the organization of the psyche is to be clear about what is meant by the term 'organization.' In cognitive science theories, organization is usually described in terms of 'systems.'

There are several senses in which the term 'system' can be used:

1. A simple system. A set of interacting units connected together by causal links. Not every unit needs to be connected to every other unit, but all units in the set must have some connection to some other unit. The causal links may be bidirectional (a affects b and b affects a) or unidirectional. They may or may not have feedback loops. Good examples of simple systems are the solar system or the flow of traffic on connected roads.

2. A simple adaptive system. A set of units which maintain themselves in some environment. The units may have specialized functions, and the system may be self-reproducing and self-repairing. There is no central executive. Good examples are a single cell, a plant, or the human antibody system. An ant colony or perhaps an acephalous tribal society are multi-organism examples.

3. A complex adaptive system. Specialized units are organized into sub-systems with a central executive. The executive may be conscious. A simple insect that crawls toward the light and tries to find food is an example. The US postal system is another.

The difference between a system and a structure is that, in normal parlance, structures are generally thought of as something having 'parts' - mutually exclusive divisions of some kind. For example, houses are structures that have rooms, halls, stairways, etc. A room is a part of the house, and whatever is part of one room is not part of some other room. In classic psychoanalytic theory, if something is part of the superego, it is not part of the ego. Unlike structures, in systems, elements or units can participate in more than one system.

THE ARCHITECTURE AND OPERATION OF A COMPLEX ADAPTIVE SYSTEM

Cognitive scientists have generally taken as their main task the modeling of intelligence - as complex adaptive systems. In the early models of intelligence everything ran under the control of a central executive. The central executive had a series of goals, and given the executive's understanding of the situation, various sub-programs were activated to accomplish these goals. The executive tests to see whether goals have been achieved and then adjusts its future decisions about what should be done on the basis of the new situation.

Such a model of the psyche requires some kind of memory for what has happened, representations of its own capacities, and representations of the environment. Normally one can communicate with the program, adding the further requirement of a language - a lexicon and grammar by which things in the world and states of the program can be represented and expressed. A good example of this kind of program is Terry Winograd's block building program. One can type instructions to the program which it will then attempt to follow. One can, for example, ask the program to stack three yellow blocks to make a tower and the program, attached to an 'eye' and an 'arm,' would locate three yellow blocks and position them on top of each other.

The early triumphs of the field of artificial intelligence were followed by a series of set-backs. The programs often crashed, could only do a very small number of tasks, and lacked flexibility. When trying to act in the real world with a mechanical body these programs failed to be able to do simple things like climb obstacles. These problems resulted in a restructuring of the basic architecture of the models. An evocative overview of the new vision of intelligence is presented in Marvin Minsky's book, The Society of Mind (1985). Part of Minsky's vision is that the mind consists of a variety of 'agents' which operate with considerable autonomy, so that control of what is happening is not always or even usually managed by the top executive. In this vision there is a complex heterarchy of agents, with some agents activating other agents, who in turn activate other agents, etc., often outside of the supervision of the top executive. (A heterarchy is like a hierarchy in having levels except that a lower level node can be dominated by more than one upper level node.)

In this type of model each system is constructed primarily out of agents. For those unfamiliar with the idea of mental agents, some background material may be useful. The human brain is a vast complex electrical circuit made up of many billions of neurons and trillions of synaptic connections. It is currently said that the brain is the most complex mechanism in the universe. Each neuron is connected to other neurons - sometimes to only one, sometimes to thousands. Basically, the neuron is a cell which can transmit electricity in millisecond bursts. The electrical charges passes along the axon of the cell to a number of branches. These branches connect to the dendrites of other cells. The connection is called a 'synaptic juncture.' Neurotransmitters transmit by chemical means the electric charge across the synaptic juncture from the branch of an axon to the dendrite of another cell. Whether the charge will cross the synaptic juncture depends on the chemical concentration of the neurotransmitters in the area of the synaptic juncture and the degree to which the dendrite is responsive to particular neurotransmitters. There is still debate about how a dendrite learns to become responsive; a variety of chemical processes appear to act on a dendrite once it is is activated by a branch from another cell to make it even more responsive to neurotransmitters from this branch in the future. While the popular literature is full of accounts of the various effects of different neurotransmitters, neurotransmitters have no direct effects on mental processes; what they do is increase the ease or difficulty of activation across synaptic junctures. Different neurotransmitters affect the synaptic junctures of different kinds of neurons in different regions of the brain. It is interesting that more is known about the behavioral and emotional effects of a number of neurotransmitters than about which neural circuits they facilitate or inhibit.

The brain has a complex and as yet incompletely understood architecture. The general account follows very roughly MacLean's idea that the brain has three major parts. Although MacLean was incorrect in many of his ideas concerning localization of functions in particular brain structures, most authorities generally treat the brain as organized into three regions; the cortex (a thin densely interconnected enfolded sheet about the size of a large handkerchief covering the outside of the brain) where sensory information is analyzed; the limbic system (a loosely defined complex of structures below the cortex, including the amygdala, hypothalamus, thalamus and hippocampus), in which a variety of emotional and motivational centers are found as well as centers for the integration of various kinds of sensory information sent from the cortex; and the cerebellum, located below the limbic system at the top of the spinal cord, which organizes and sends on messages to the various motor structures of the body.

This much of the story is well known. What is less well understood by many social scientists is how such a complex of neurons can form something one can reasonably call 'agents.' Over the past fifteen years large advances have been made in constructing neural models of how the brain processes information. Psychologists and others interested in constructing neural models have been able to build computer simulations of neural networks which can do a variety of information processing functions. A neural net is simply a set of neurons which are interconnected and which respond differentially to different kinds of inputs from other neurons or from sensory excitation with different kinds of outputs. Such networks, called 'schemas' in the some of the psychological literature, are active processors, they can function as templates with default values and correlated variables, they can embed or be embedded in other networks, they are able to fill in missing information, they can discover organization or structure present in input information, they are context sensitive, they display 'graceful degradation' when parts of the network are destroyed, they can form 'content addressable' memories, and they naturally produce such psychological processes as prototyping, generalization and priming (see D'Andrade 1995). It is not possible to explain here how a smallish network of neurons can do such a wonderful variety of things here; readers unfamiliar with connectionism and neural modeling and interested in the mind owe themselves a survey of the the growing connectionist literature.The non-specialist reader who wishes to find out more about connections may find the accounts by Stinson and Palmer 1991 and Strauss and Quinn 1997 accessible; the classic presentation can be found in Rumelhart et al 1986.

Every agent consists of a number of neural nets or cell assemblies. One of these nets is able to identify from inputs the degree to which some dimension of a situation falls within a given range. Once the network determines that the situation falls outside the range , the net activates other nets which eventuate in actions which will affect the situation and bring it back within the proper range. For example, the human brain contains a complex balancing system which keeps us, in most circumstances, from falling down. The system contains a number of nets that receive information from parts of the sensory system about whether one is off-balance to the left or right, front or back. Once the degree of being off-balance reaches a certain threshold, these nets activate other nets which affect the body movements in ways which will bring one's body back into balance. Normally, this happens automatically, without consciousness. Miller, Galanter, and Pribram (1960) labelled such a complex a 'TOTE' unit; assemblies of cells that can test to see if some condition falls within certain limits, then operate to change things so that the world is changed in the direction of bringing things back within these limits, then test again to see if things have really come back to where they should be, and if so, then exit the unit. What most agents or TOTE units do is select from among a range of other more specific agents the best agent for the situation, who in turns tests the situation and selects even more specific agents. At the end of the chain of command are simple operations which don't test; when activated they simply do something.

The automatic and out-of-consciousness balancing system is usually sufficient to keep us from falling while sitting or walking or running. However, sometimes we become vividly conscious of balance problems. When this happens we simultaneously become aware of a sensation of being off balance and a strong wish to regain our balance. The automatic system has failed and emergency measures are needed. All other conscious agents are suspended, and a flood of body sensations and sense impressions rush to consciousness - how far one is from the ground, how one's body is positioned, what objects are nearby to grab or avoid, etc. - which aid in selecting the right agents to avoid disaster.

Note that two systems are at work here; one an automatic system with its own agents and the other a conscious system which can operate very generally on almost any problem.

CONSCIOUSNESS

It is a general finding of psychology that for most learning or memory to take place, the content that is learned or remembered must have been conscious. While subliminal learning does occur, it's effects are difficult to uncover and tend to involve either emotional conditioning or the recovery of previous memory (see LeDoux's summary of this research) . However, a distinction must be made between being consciousness of the contents of the specific elements involved in the learning process and being aware of what has been learned. For example, people can be taught to categorize certain characters of Chinese writing into semantic categories, and they can learn the task so well that they can correctly categorize characters they have never seen before into the correct semantic category, but they cannot say how they do this - they are not aware of the basic patterns they have abstracted (Hull did such an experiment in the thirties). They must be, however, conscious of the shapes of each of the characters they learn to categorize. Similarly native English speakers know they should not say "I am wishing you were here" but they do not know the semantic features they use to make this discrimination. They are only aware of the sequence of individual words and that they sound wrong. To remember, learn, or in any way reorganize one's mind, generally there must be consciousness of the material processed, but not necessarily consciousness of what it is that has been remembered or learned. Thus someone can learn that world is a secure and supportive place without being aware that they have learned this; what they must to be conscious of are the incidents through which they learned this. Such learning is called 'implicit learning.'

The account presented above places a heavy functional load on consciousness. It needs to be stressed it is not being asserted that people are aware of everything they do. A huge number of goals and procedures never enter consciousness (how one stays balanced, how one knows the meaning a sentence, etc.), and a great number of procedures are repressed or kept out of awareness. For example, a young man may know quite consciously that after he has had a number of dates with a new girl that he begins to lose interest in her. He may even observe that this is a pattern - he does it over and over. But the procedures by which this happens may be totally out of awareness - it's just that this funny thing happens. The procedures that result in a loss of interest happen in an automatized way. However, that said, in the normal operation of the psyche, consciousness plays the crucial role in handling life's complexities or which automated mental action is ineffective or maladaptive.

In this account, consciousness itself is not a motivational system, but rather a complex system of another type. It is a system that operates in an automatic, inbuilt way. How it does what it does - how neurons can give rise to conscious sensation and thought - is the great mystery with which we begin the second millennium, bereft of even a single plausible mechanical model. However, work on the neural associates of consciousness has continued to progress. Most recently, Tononi and Edelman presented in Science (4 Dec 1999 p 1846) the hypothesis that consciousness is the result of the neural action of a neural dynamic core within the thalamocortical system. These authors have suggested that a key element of conscious experience involves reentrant interactions between posterior thalamocortical areas related to perceptual categorization and anterior areas related to memory, value, and planning for action. They argue that consciousness involves the activation of widely distributed brain areas which shift over time. The authors go on to argue that fast, strong, and distributed neural interactions which can be maintained for 500 milliseconds or more are necessary for the production of conscious experience. Further, they suggest that a group of neurons can contribute directly to conscious experience only if the group achieves a high degree of synchronous integration. They say:

This vision of consciousness as a moving swarm of simultaneously activated neural nets, always moving, incorporating new nets and dropping others, is a great improvement on a model of specially privileged neurons (a place in the brain), or a model of simple activation. It raises many questions, especially the question of how the dynamic core selects the nets to be incorporated into the dynamic core. However, in this model at least such questions are amenable to empirical investigation.

Another point which needs to be stressed is that consciousness is very limited. The number of items one can be aware of and hold in working memory is very small - approximately five or six things. Despite the great number of things people know and the great number of operations taking place in the brain at any moment, only a small amount of all this can be conscious at any one time. It may be that the serial nature of consciousness is the result of this limitation, which chokes off parallel processing into a small trickle of items. In turn, the size limitation of consciousness may be due to the 'binding problem;' that is, the problem of how different features of an object, each analyzed by separate neural networks, are combined into a single percept. For example, given a blue square and a red circle, how does the brain, after analyzing by means of separate neural networks 'blueness' and 'redness' and 'squareness' and 'circularity,' know how to put these these feature back together - to construct a 'blue square,' not a 'red square?' One hypothesis, discussed by Crick in his book The Astonishing Hypothesis, is that for material that does not have an already well built neural net for just these connections (a special neural net dedicated just to 'blue squares'), but has to built connections rapidly, that the neurons of the blue detecting network and the square detecting network fire in a cycle of synchronized bursts, while the red detecting network and the circle detecting network also fire in a cycle of synchronized bursts, but a few hundreds of a second out of phase with blue and square network bursts . The time difference in the firing contains the information about which color goes with which shape. My speculation is that given that all the networks in the dynamic core must fire within the same brain wave cycle operating in tenths of second, there may only be room for five or six synchronized groups of networks within the same cycle, thus limiting working memory and consciousness to five or six complex conceptual items.

PROBLEMS OF COORDINATION

Once this kind of society of agents is envisioned, problems of coordination arise. What if two agents get into a battle concerning which one gets to do something next? Or if two agents activate mutually incompatible agents? How should such conflicts be settled? A strict chain of command in which every agent has only one boss (a strict hierarchy), which would obviate such problems, will not work because often the same low level agent is need to be used to accomplish different goals. For example, if a child is building a tower, the 'build a new tower' agent and the 'take down an old tower' agent both need to be able to activate the 'grab a block' agent, but if the task to is take down one tower and build another, something has to be done to coordinate control of the 'grab a block' agent. Minsky proposed that such conflicts 'migrate up' the system to be dealt with by higher level agents, and in the usual course of events the upper level the activation of one agent (the 'take down an old tower') will suppress the other agent (the 'build a new tower' agent) because it has the support other agents (the 'get proper materials' agent gives the 'take down an old tower' agent support because there are no free blocks for the 'build a tower' agent). Control by some agents - and hence the yielding of control by other agents - passes in complex pathways through the entire system. Problems in control migrate up and often are resolved without ever reaching the top level executive. However, sometimes they reach the top level executive, and if they can not be resolved there, the system may crash, freeze, or behave erratically.

In general, Minsky equates consciousness with the top level executive. He says

Minsky, like Mandler, see consciousness as the 'place' where upper level trouble shooting occurs. Most mind-like programs have a structural division between a short term memory - a 'work-sheet' where a small number of immediate problems are represented and acted upon - and long term memory, where much larger stores of information are kept and brought when needed to short term memory. It is an easy but controversial jump to equate short term memory with consciousness.

It has been, for me, surprising to see the models of machine intelligence evolve towards a perspective in which complex internal dynamics and consciousness play crucial roles. The new perspective reminds one strongly of Freud's scientific project which attempted to construct a neural model of the psyche.The emerging computational view of the human psyche as balanced between distributed decision processes and a central executive is a powerful perspective, permitting the integration of findings in cognitive psychology, neuroscience, cognitive anthropology, psychoanalytic theories and computational models of mind.

THE OPERATION OF THE SYSTEM

Dynamics - the operation of the agents of a system - are constrained by the system's organization. Let us start with how the adaptive system of the human psyche might operate before major problems in operations set in. The description here follows the outline presented by George Mandler (1975, 1985). The world is perceived and objects and events are categorized by a series of neural networks that operate outside awareness. The results of these categorizations can result in a flow of activated goals and procedures also without awareness. I am sitting at my desk typing. A heterarchy of procedures accomplish the typing of words. These typing procedures flow from perception to behavior without conscious direction. In some cases, one can become aware of the parts of some normally automated process -one has trouble typing a word and experiences consciously one's finger reaching for the letter 'z.' In some cases one can never become aware of the parts of the process - people have no consciousness of the process whereby a forefinger is lifted.

Using Minsky's metaphor, people are aware of steering but not how the steering is accomplished. Humans are born with a number of inbuilt automatic operations. And through learning a second very large number of complex operations comes to be automated. However, at any point, if there are problems and the operation does not run off properly, the operations that are interrupted attempt to 'post messages' to consciousness. The executive responds to these conscious contents with the decision procedures that have worked successfully in the past, over-riding learned automated sequencing with a new sequence of procedures which attempt to adapt to the local circumstances.

There are many purposive agents which do not generate conscious registrations in the human psyche, as in the example given above of the human balancing system. However, the human psyche also consists of many purposive agents that don't have in-built test-operate-test procedures. Suppose an agent give rise to the conscious goal of filling a glass with water. A set of operations are activated to turn on the tap and get water and put a glass under the tap and fill the glass. But none of these procedures have in-built testing for their success or failure. Instead, the 'turn on the tap procedure' must be coordinated with the visual system to generate in consciousness information about whether water is coming out of the tap. The executive system then can trouble shoot the situation if there are problems with the water flow. Most people need to consciously watch and see if the glass is correctly positioned under the tap so that water will go into the glass. Finally, visual and kinesthetic information about the amount of water in the glass is registered in consciousness so that the executive system can decide when to stop the procedure. However, a bartender, standing at the same tap and filling the same kind of glass many times a day, can almost completely automate the process so that very little consciousness need be involved. Unlike computer programs, in which programmers build in specific test-operate-test procedures at every decision point in the program, for many procedures in the human psyche the testing is be done by the executive system through consciousness. This gives great flexibility and plasticity of behavior, since the same procedures can be used for very different goals and in different circumstances. It also gives rise to very complex monitoring problems. The human psyche is a complex blend of two models; one in which a global level executive determines what will be done next, and the other a number of distributed systems in which many decisions are made locally.

At almost all times the human adaptive system is goal directed. Goals are selected by the executive from the contents of consciousness, or goals activate sub-goals directly without conscious intent. The goals may not be of any great moment, as when we idly watch a bird fly across the sky. Sometimes we can put a brake on having any goals, turn off the executive, and just day dream or start to go to sleep and let whatever comes to consciousness float in and float out. The executive can sometimes turn the executive off. Perhaps hypnosis can turn part of the executive system over to someone else, and perhaps disassociation occurs when other systems are unable, for various reasons, to reach consciousness and the executive shuts down.

However, most of the time the executive is operating and the contents of consciousness correspond to noticing sub-system glitches whose problems have migrated up through various higher level processes until they come to threaten the achievement of the currently regnant goal, at which time they become conscious. For example, one may have the conscious activated goal of writing a paper on something. Much conscious processing is required because many unautomated sequences of ideas have to be constructed and rearranged at each point in the process to reach the various sub-goals involved in writing a paper - having a general introduction, selecting which topic to begin with, introducing this topic, presenting thoughts about this topic, discovering that some idea is not clear or is vague or even contradictory, reformulating the idea, and so on. As Mandler stresses (1975), at each point glitches and blockages result in new conscious contents which the executive attempts to resolve with respect to priorities relevant to the upper-level goal.

The process described above can be programed in a standard computer programming language like FORTRAN. However, the operation of the human psyche involves complicated sensations like love and fear which are not easy to simulate. There is no FORTRAN of the sensations. Sensations, defined here as including emotions, feelings, and felt body states are a constant part of consciousness. Some of them have external reference - the hot feeling on the tongue comes from the hot coffee one is drinking. Some are locatable in one's body - the pain comes from one's foot, the itch is just behind the shoulder blade. Some have no specific body location, although various body sensations may be involved, and are called emotions and desires - one feels sad, or wants a cigarette, or is amused.

Sensations have neural sources. How it is that some kinds of neural activation give rise to the sensations of anger and things being green while others give rise to ideas and thoughts is part of that great mystery of consciousness. The account here is that certain inputs to consciousness are somehow experienced as feelings, emotions, desires, etc., while others are experienced as ideas. Human consciousness is a complex combination of sensations and ideas.

A basic hypothesis about sensations is that they effect the operation of the executive system just like other representations. The executive system uses decision procedures which say 'doing this feels bad, so don't do it.' Or 'this situation feels frightening, so get out of this situation.' The same kind of information can also be represented by consciousness ideas and thoughts without sensations; for example, one might think without any emotion 'Oh, this is a dangerous situation' or 'Oh, this is not a good way for me to talk to this guy.' But sensations have special properties that thoughts do not. Provided the activators of the sensation continue to operate, sensations can have persistence in consciousness no matter what immediate goals are involved; they can stay around in consciousness while thoughts shift. Also it is hard to have two thoughts at once but one can feel hungry or sad while thinking about almost anything. Another property of sensations is that they are naturally scalar; sensations can be strong or weak. A strong sensation seems to take over a greater extent of consciousness, and very strong sensations can it difficult for the executive to continue with the old goals because there is not much else for the executive system to respond to. A third property of sensations is that the underlying sensation activators often have direct automated connections to other procedures, and if the neural activation is strong enough, the executive gets over-ridden. For example, in extreme fear the bladder empties and the body may become immobile, not because of conscious intent, but because an automatic over-ride has taken place. There is often consciousness awareness of impending over-rides, adding even more factors for the executive system to deal with. Strong anger often does this, and also may thereby produce a tertiary effect in which anticipation of the disaster which might be caused by the over-rides caused by anger all of which results in conscious anxiety, resulting in a complex blend of sensations.

Given the complexity of all this ; the complexity of the flow of control down the heterarchy; the interlacing of goals; the problems of disequilibrium between and within systems, the gating concerning which appraisals get to consciousness and which do not, it is amazing that more humans are not severely dysfunctional. Walking across the street is an incredibly complex feat. Taking care of a child is even more so.

This is an account of the normal dynamics of the human psyche. It is close to the thousands of year old folk model of the mind, which although implicit, can be discerned through text analysis of the ancient stories found in the Bible and the Iliad. However, in many ways this account differs from the standard folk model of the mind. Here the extent of conscious control is greatly limited and the role of consciousness in trouble-shooting is emphasized. Also here the affective sensations - emotions, feelings, desires, etc. - are treated as signals as well as prompts to action. In the folk model, wishes, impulses, and feelings are treated as if they did things directly. In the model presented here the actual 'movers' are the agents or activated goals. The energy is not from some great energy source, it is simply the energy of neural nets. External sense organs and internal sources activate nets. Nets activate other nets. Agents are complex networks of neural nets.

However, this is not how humans experience themselves. This is because agents, when activated, give rise to the consciousness content we call 'wishes' - a representation of the goal and a felt sensation. To consciousness it seems that wishes do things. But there is no direct causal path from conscious wish to action. The executive system may carry out decision making procedures with respect to the wish, or because of other considerations it may not. The pathway to action is from activation of agents to the awareness of a conscious wish to the executive system to activation of sub-agents or procedures which can lead to action. Or the pathway is from an agent or neural source of sensation directly, without consciousness, to sub-agents or procedures which may lead to action. Because we are not conscious of the activation of agents, only the conscious representation of the goal and associated sensations, we experience conscious wishes as prime movers.

The account given above of the operation of the agents of the human balancing system, and its occasional emergence into consciousness, is relatively straightforward. However, a perspective which treats all motivational systems as composed of agents, often operating outside consciousness, is more problematic. For example, consider the case of two young new employees (A and B) of a firm. At office meetings the normal discussion involves exchanges of opinion and coalitions which shift by topic. Let us supposed that it becomes noticeable that one of the new employees (Ms. A), often disagrees with Ms. B, sometimes interrupting, sometimes expressing non-verbal disapproval, sometimes simply silent, but rarely agreeing or being supportive. An observer would conclude that rivalrous feelings and motives are involved on the part of A, and feel that this observation is supported by the fact that Ms. A's disagreements with Ms. B's are sometimes inconsistent with previous opinions given by Ms. A. However, Ms. A, when somebody mentions her rivalry with Ms. B, disclaims having such feelings; "it's just," she says, "that M gets things wrong and tries to throw her weight around." Consciously, it seems that Ms. A is not experiencing rivalrous feelings, but simply the reactions she believes that anyone would have to Ms. B's lapses in judgment (and it surprises her that the rest of the office does not agree with her about Ms. B's bad judgment).

The situation described is not unusual. Most people recognize what A is doing. But most people find it hard to believe that A really doesn't know she has rivalrous feelings and motives. She so obviously has goals of competing with and derogating B that she must be aware of them - otherwise how could she be doing it? Part of the problem is that A is fully conscious of the goals of the sub-agents, goals which include pointing out people's faults and lapses of judgment. What A is unconscious of is the upper level agent's goals - defeat of B so that A alone will have the esteem and affection of the group.

This example of a mixture of conscious and unconscious agents raises a variety of theoretical issues. Most immediately, it raises the issue of how human motivational systems are organized.

THE HUMAN ADAPTIVE SYSTEM

The approach I have taken here to identity human motivational systems is to follow the lead of psychoanalytic theorists who have already worked out integrations between cognitive science models and psychoanalytic models; primarily the work of Bowlby (see his three volume work: Attachment, Separation and Loss, 1971) and Rosenblatt and Thickstun (Modern Psychoanalytic Concepts in a General Psychology, 1977). Bowlby, who was an active researcher at the Tavistock Institute and served as the Training Secretary and the Deputy President of the British Psycho-Analytical Society, is well know in the field of academic psychology for his work on attachment, but less well known for his role in the development of the British object relations school of psychoanalysis. Rosenblatt and Thickstun wrote as practicing psychoanalysts interested in the integration of psychoanalytic theory within a general theory of human behavior. I have also been influenced by the work of Kenneth Colby, a psychoanalyst who has developed complex computer simulations of neurotic and paranoid processes (see, for example "Simulations of Belief Systems" in Computer Models of Thought and Language, Shank and Colby, eds., 1973), the work of Mardi Horowitz, a psychoanalyst at UCSF interested in the integration of clinical psychodynamic knowledge and cognitive science (see, for example, Person Schemas and Maladaptive Interpersonal Patterns 1991), and Drew Westen, a Harvard Medical School psychiatrist, interested in the integration of psychoanalytic theory with cognitive science and the study of culture and society (Self and Society: Narcissism, Collectivism, and the Development of Morals, Cambridge 1985, The scientific legacy of Sigmund Freud: Towards a psychodynamically informed psychological science, Psychological Bulletin 124:333-371, Beyond the binary opposition in psychological anthropology: integrating contemporary psychoanalysis and cognitive science, in The Psychology of Cultural Experience, Moore and Mathews, eds, Cambridge, in press)

A basic question confronted by all these theorists concerns the issue of adaptation. If the human psyche is a complex adaptive system, what does it adapt to? Obviously, humans need to adapt to their environments in ways that will maintain the functioning of their own bodies and reproduce them. Standard concepts of the role of physiological drives such as hunger, sex, and avoidance of physical pain, in directing the individual towards body maintenance and reproduction, need to be included in the model. But a series of independent drives does not by itself make a system.

A major conceptual problem in the study of a system is to determine its architecture. Systems are composed of sub-systems, and the ways in which sub-systems are interrelated constitutes the major architecture or organization of the total system. It is exactly on the this issue that one finds disagreements among the different psychoanalytic theories. The classic Freudian model of ego, id, and superego no longer holds pride of place. The idea that there is a particular amount of energy within the system provided primarily by the id has almost disappeared from psychoanalytic theorizing. There is considerable terminological confusion about the difference between the ego and the self, although some aspects of something like the ego are found in most theories. The superego still retains its theoretical utility but its role is often less central than in the classic formulation. Finally, other structures, such as internal objects and self-objects, now play major causative roles in psychoanalytic theories. A large number of commentators have remarked on the splintering and diversity of models in current psychoanalytic thought (for example, see Psychoanalytic Versions of the Human Condition, edited by Marcus and Rosenberg 1998); from a history of science perspective this splintering and diversity is expectable when the main presuppositional framework of a theory is discredited. Patricia Kitcher, a philosopher of science, has written a detailed analysis of the changes in neuroscience and related fields which undermined Freud's metatheoretical assumptions and his interdisciplinary agenda (Freud's Dream, 1995). Once the basic notion of drive as a store of energy became both outmoded in neurology and of questionable importance in the practice of therapy, there was no common framework to guide the construction of a model of the psyche.

In trying to think about systems, one important issue is the degree to which one sub-system is subordinate to another. Considering humans abstractly, one might imagine that the physiologically based motivational systems involved in obtaining food, sex, and escape from physical discomforts would dominate the other systems because they serve such vital functions. Some simple animals - fish, eels, snakes, lizards, etc. - seem, on causal observation, to be organized this way. However, animals with more complex social relationships seem to be organized psychologically in a different way. With all the apes and monkeys, as well as many other mammals, the vital physical systems seem to be enmeshed and even subordinate to motivational systems that involve seeking and maintaining social relatedness. A speculation about this in evolutionary terms is that because social relationships affect so strongly the outcomes of physically vital activities for some species, whatever systems organize the activities that maintain and enhance social relatedness have evolved in these species to organized and be integrated with the physiologically adaptive systems. Obviously, the physiologically adaptive systems must have enough autonomy so that at life threatening degrees of hunger or pain or physical danger these systems will over-ride the system that maintains and enhances social relatedness. Physical attack, the threat of death, extreme deprivation, etc., result in the activation of strong escape or aggression goals which take priority over goals of social relatedness. However, as an empirical assumption, the model here assumes that humans from birth seek to form social relationships and that the systems which organize such activities are persistent, pervasive, and coordinated with other systems. In Attachment, Separation and Loss, Bowlby presents a wealth of observational data concerning the numerous inbuilt behavioral patterns which maintain and enhance relatedness beginning at birth and continuing throughout life.

In classical Freudian theory the infant is thought to be dominated primarily by the motivational systems that serve vital physical needs with maladaptation occurring when the goals of these systems are frustrated or result in trauma. But even without early frustration or trauma, according to Freud, complexities arise because, impelled by these motivational systems, the child encounters problems with parents and others in the oedipal period. The vicissitudes and rewards of social relatedness, involving emotions such as love, envy, hate, fear, etc., create a world of internalized objects which form a platform for further psychological development. Thus, in classical psychoanalytic theory, social relationships per se do become a motivational center of personality, but not until after four or five years of age. The results of a large body of infant and childhood observational research contradict Freud on this issue.

It should be stressed that the other motivational systems are generally integrated into the social relationship system. In the normal case, social relationship systems not only guide other systems, but the operation of the other motivational systems also act to increase the effectiveness of the social relationship systems. In the canonical account, the baby seeks the breast, is nourished, and at in doing so, discovers the mother as an object and relates itself to its mother.

The next problem is to distinguish the different motivational systems. A basic question concerns the selection of the kinds of phenomena that count as indicating that something is an element of a system. An ideal method would be to be guided by knowledge of the neural organization of the brain. There have been great advances in mapping the cortex, but the neural and biochemical interactions between the cortex, thalamus, hypothalamus, hippocampus, amygdala, and other structures of the limbic system are at present incompletely known (For an accessible account of what is known about mapping networks involved in emotion, see LeDoux's The Emotional Brain, Simon and Schuster 1996). At this point the best one can do is try to present an account which has neural plausibility - an account that does not contradict what is known.

Back then to the question: what kinds of elements should we be concerned with? There are some heuristics for isolating and defining systems. Adaptive systems do something - they consist of agents that have characteristic goals which have conscious representations along with sensations of wanting or desiring something. A system of agents which can have conscious goal representations and characteristic sensations constitutes a motivational system. Accomplishment of particular goals involves particular actions, some of which are unique or nearly unique to the accomplishment of that particular goal. Since humans experience feelings and emotions that function as information about what is happening inside and outside themselves, each motivational system should generate characteristic feelings which serve as appraisals of how well the system is doing with respect to its relevant goals. Setting up the goals requires cognitive discriminations about the world; if one wants to find food, one must have a complex of discriminations that distinguishes the eatable from the inedible.

It might be informative if someone had completed a comprehensive factor analysis of a large data set of coded observations across many time slices of the goals, actions, feelings, and discriminations of many people. The closest available proxy is the result of the synthesizing and differentiating operations of the minds of clinicians. Over the hundred or so years of psychoanalytic therapeutic practice many thousands of patients have been seen for many hundreds of hours. This collective data base contains millions of observations. These observations are not in a single mind or computer file, but the distribution of these observations among a large number of clinicians has the advantage of providing many independent judgments about the nature of the systems which make up the human psyche. As a heuristic we can take the general systems that clinicians have found and attempt to align these with characteristic patterns of goals, actions, feelings, and discriminations. The historical place to begin is with the tripartite structure of ego, id, and superego.

THE EGO OR EXECUTIVE SYSTEM

The psychoanalytic ego corresponds relatively closely to the cognitive science notion of an executive, discussed above. An executive system is like the operating system of a computer; it consists of a complex series of decision procedures. In this account, the executive system will be treated as operating only on the contents of consciousness. The operations of executive itself may or may not be conscious. Normally we are not aware of the actual machinery involved in decision making unless we interrupt the flow of decision procedures and are able to retrieve conscious inputs from where the process stopped. The executive system responds to the contents of consciousness, comparing and evaluating items, attempting to bring new information to consciousness by searching memory when it is needed, and focusing attention on particular features of a problem. The items in consciousness can come from anywhere in the system; from perceptions of the external world, from subsystems that suddenly confront an operational problem, and from activation of various sources of conscious sensation and affect. One of the central tasks of the executive is to decide among competing goals, in relation to information about the world, the goal that will be pursued. Related to this, the executive makes decisions concerning the selection of the best sub-goals to accomplish the regnant goal.

The top-level goal of the executive system is to accomplish goals. The most elusive elements to describe in the executive system are the characteristic feelings of the ego; to the degree we experience feelings to inform us about how the executive system is doing, these feelings seem to be senses of effectiveness or accomplishment, fatigue of a special mental kind involved in loss of concentration, etc. The characteristic actions can be introspectively reported (for example, in problem solving think-aloud protocols) and primarily involve setting up priorities, formulating problems and outcomes, finding out about the world, and holding firm against other material entering consciousness. The characteristic discriminations are that some action is doable or not doable, risky or safe, that some situation is known or unknown, and that some situation can be managed or is unmanageable.

THE IDENTITY SYSTEM

There appears to be consensus among current psychoanalytic theorists that a distinction needs to be made between 'ego' and 'self.' Reading the self theorists, discussions seem most concerned with what is often called 'a sense of self' rather than the self as an executive. For example, discussions of narcissism deal primarily with the 'grandiose sense of self,' the 'sense of entitlement,' etc. What is being referred to seems to be identity rather than executive functioning. While this term is associated with Eric Erickson's developmental sequence, it will be used here to refer to a distinctive motivational system. The characteristic goals of this motivational system are affirmation and enhancement of the self based on appraisals generated by match to self-representations, as well as goals of obtaining responses of acceptance and admiration from others.

According to the model presented here, there need be no one top-level goal for a motivational system. On this issue Leininger comments:

The cognitive core of this system is a linked network of self-representations with varying degrees of fit and conflict among them. Following Rosenblatt and Thickstun, the total "self-representation ... can be considered to be the integrate of those appraisals which, if felt, constitute the experience of self. When the integration is not complete ... there may be multiple self-representations" (p. 300). The characteristic feelings sent to consciousness are pride, self-esteem, shame and embarrassment. Self-representations are typically gendered. Representations in this system include ideals for the self. Characteristic discriminations involve judgments concerning what is acceptable, respected, prestigious, honored, etc., along with judgments about social rank. Within this model, identification refers to a process by which self-representations are modified to be similar to the cognitive representations of some person. Characteristic disorders of this system are pathological narcissism and acute inferiority problems.

THE AUTONOMOUS COGNITIVE SYSTEM

Cognitive operations occur throughout all the systems of the psyche - formation of cognitive representations, prototyping, etc. However, typically these cognitive operations are tied to the particular procedures which invoke them. In Piagetian terms, these cognitive operations are primarily sensory-motor or based on concrete operations. Such reasoning is used to reach a particular goal, not to understand things in general, and might be called 'embedded cognition' in contrast to the more general cognition of the autonomous cognitive system. The achievement of abstract thinking is, in many areas of life, quite difficult. The general moral principles of Buddha or Jesus are very abstract formulations and without cultural tuition such formulations are difficult for most people to achieve. Much if not most of the material related to interpersonal matters in the autonomous cognitive system is culturally learned.

In psychoanalytic theory this ability to develop and reason with more abstract and general propositions is usually considered part of the ego. Within the model presented here, the autonomous cognitive system is treated as a resource of the executive system, but it has a certain degree of motivational autonomy, sometimes sending it own messages to consciousness. The voice of reason can come to us without being called. The pressure to be rational seems distinctively human. The cognitive core of this system is a series of abstract schemas about the world and people which can generally be put into propositional form. The characteristic goals are trying to be consistent, generalizing, and formulating new representations of problems. Characteristic feelings are what Zeno Vendler called the 'cognitive emotions' of wonder, surprise, bewilderment, the feeling of comprehension, the "ah-ha" feeling, etc. Characteristic actions are testing of ideas and speculating about things. The characteristic discriminations concern truth and consistency.

THE ID

The id is present in current theoretical debates mainly through its absence. However, there are id phenomena that Freud noted which are important. For Freud, the id was a source of energy. Current models of the psyche and brain do not need a source of energy per se; each neuron has substantial energy. The modern problem is still one of energy economics, but rather than involving total amount, the problem is shifted to the distribution of energies. Neural nets need to be organized in complex ways so that the patterns of firing neurons has a systematic relation to whatever the neural net is supposed to convey. It is this distribution of energies, not total energy discharge, that is important.

However, dismissal of the id as a unitary source of energy does not affect the observation that people cathect things. The conscious experience of the person is that they find something interesting, like it, want to be close to it, care about it, etc. Or that they don't like it, want to escape from it, etc. One can describe cathexis as the establishment of special patterns of activation of neural nets in response to certain objects. As a result of the establishment of these patterns of activation an object (or, more accurately, a particular relation to an object) becomes a goal. These special patterns of activation are generally thought to involve limbic as well as cortical areas. One can theorize that cathexis occurs within the brain when an object is recognized by various neural nets and, once recognized, neural projections descend from these nets into the limbic system producing various sensations and stirrings to action that humans experience as wanting, love, liking, etc. A probable hypothesis is that all the basic motivational systems depend on special linkages to the limbic system.

In research I have carried out with a variety of attitude rating scales, I have found that the factor which explains the greatest proportion of the correlational variance of these scales is how important something is judged to be. The scales which have high loadings on this factor are 'how important ... is to me,' 'how much I care about ...', 'how much I value ...', whether '... is one of my goals', whether '... is one of my interests,' whether '... motivates me', whether 'I often think about ...', whether 'I like ...,' whether 'I have strong feelings about ...', and whether 'I am involved with ...'. This cluster of responses corresponds closely to the expected effects of cathexis as a limbic system involvement which transforms a representational schema into a goal with emotional characteristics. Unfortunately, the history of the term 'cathexis' may activate in some readers negative affects based on the battles in the seventies concerning the libido as a store of energy, but whatever term is used, the phenomenon is distinct and important.

THE ATTACHMENT SYSTEM

The total collection of things a person cathects is too large and heterogeneous to form a single system. There is a tendency in current psychoanalytic thinking, discussed by Drew Westen (n.d.), for current psychoanalytic and other theorists to simply posit a variety of motives or goals with little organization. The position taken here is that while a person can become enamored of many things, there are in-built neural systems which make certain kinds of objects naturally engaging - more easily cathected. Perhaps given the right conditions, anything can be cathected - pepper, quadratic equations, cabbages and kings. But the usual experience of clinicians is that while the store of things cathected is indeed diverse, things are cathected because they are associationally connected with a small number of general motivational systems.

Among this class of general goals, one general set of goals has been found to be extremely active in creating a large class of cathected things. The system built around this set of goals will be called here the 'attachment system.' The core goals of this system are the establishment, maintenance, and enhancement of relationships to others (See Bowlby 1971, and Rosenblatt and Thickstun 1977, for a similar statement: p 123)

The characteristic feelings produced by this system are love, feelings of closeness, intimacy, and relatedness. The attachment system has its dark side, reflected by feelings of hate, jealousy and envy which occur when interference with attachment goals is experienced. There are also the affects that go with loss, such as feelings of insecurity, sadness, estrangement, and loneliness. Characteristic actions are seeking contact, exchange, support and rivalry. Characteristic discriminations involve the presence or absence of someone, the responsiveness of someone to oneself, the presence of rivals, etc. A large complex of behavioral manifestations of this system are described in Bowlby's three volume Attachment, Separation and Loss, and are the subject of a continuing line research in child development.

As argued above, among the motivational systems of the psyche, the attachment system is the most pervasive and, if not the most peremptory, the most general organizer of the psyche. In this account, a primary goal of the infant is appropriate relatedness. This goal is 'instinctive' - it is inborn, shaped by evolution, and a human universal. Disruption of it is catastrophic for psychological development (See Bowlby on separation and loss). The physiological system of sex, with its own characteristic feelings, goals, actions and discriminations, from infancy onwards is, to some degree, integrated into the attachment system. The sexual system may be diffusely integrated into the attachment system, involving love for the entire person, or may be limited to momentary sexual desire for kinds of body contact and physiological satisfactions. The two systems need to be distinguished because, as Bowlby argues:

To this I would add that the characteristic affects of the two systems are also different.

Conflicted integration of the sexual and attachment systems is a frequent occurrence, resulting in disturbances with respect to both. Problems in one system may be displaced onto compensatory activities in the other. However, as a senior analyst in San Diego, H. Serota, once mentioned to me with respect to the question of whether problems in the sexual system always meant that there are problems in the attachment system, "I've seen it every which way."

THE SUPEREGO

The cognitive core of the superego is a system of moral prohibitions and prescriptions. The operation of the superego system is distinguished by the generation of characteristic feelings of guilt and righteousness, a powerful sense that one ought to do or not do something, etc., characteristic goals to not do unnecessary harm, to help others, to be fair, to obtain justice, characteristic actions involving punishments and rewards, confessions, self-punishments, etc., and characteristic discriminations involving what is morally right or wrong, intended or unintended, consequential or inconsequential, etc. These elements form interlocking complexes which affect consciousness and the operation of other systems. It does not seem to be the case that the normal superego has a single principal, rather, the system seems to consist of a set of interrelated goals and affective appraisals which are not reducible to one principle.

Figure 1 presents a summary of these five systems. They all operate with some degree of autonomy, but the attachment system is the basic system from which the superego and identity system differentiate. Development of superego and identity systems is a continual life-time process. The autonomous cognitive system is a late maturational development and is always fragile. It may be the case some humans do not develop a superego. Whether or not conscienceless people exist, the development of the superego seems to vary greatly by individual and society with respect to the extensiveness and generality of its provenience.

INTERNALIZED OBJECTS

Beginning with Freud's first venture into building a structural model, close connections have been hypothesized between the formation of internal objects and the structures or systems of the psyche. In his paper 'The Ego and the Superego' Freud presented the hypothesis that the superego is formed in males primarily through identification with the father. This identification forms a precipitate within the ego (an internal object, or introject), which is then differentiated from the ego and forms the superego. Fairbairn, in opposition to Freud, argued that both the id and the superego were split off aspects of the ego's identification with split internal objects.

The first issue concerns the nature of internal objects. Part of what is typically said about internal objects in psychoanalytic theories fits neatly with formulations in cognitive psychology concerning neural nets or schemas (see, for example, M. Horowitz, Person Schema and Maladaptive Interpersonal Patterns). The infant and child forms strong attachments to parents. This results in the formation of powerful schemas for the parents and for the interaction of the self with them. Activation of these schemas results in the conscious representations of the parental figures and expectations about what they will do and how they will respond to oneself. Even in the absence of the parents, these schemas continue to assimilate experience to the kinds of persons the child experienced and the kinds of relationships experienced in interaction with the parents. The schemas formed by interaction with the parents are not carbon copies of the interaction of the parent and child as viewed by an adult observer. The child's immature understandings of the world results in what seems like distorted schemas, strongly affected by the child's needs and the traumas and frustrations experienced in maintaining relatedness. These various object schemas have complex relationships to each other, both with respect to similarity and with respect to various 'who does what to whom' relationships.

However, equating 'schema' and 'internal object' is not an entirely satisfactory formulation. Psychoanalytic observations indicate that internal objects have causal powers that standard schemas for squares and circles do not. For example, internal objects can split into separate objects. Further, the schemas for the relationships of self and parent can stay active in the absence of the external objects and are able to produce complex feelings and thoughts that are like the feelings and thoughts the child experienced in interaction with parents. More than this, the child's parental schemas may be shaped by fantasies of things that never happened. Even further, these internal objects not only may attack with a savagery never actually experienced, but the attack itself may be a transformation of the child's angers towards the parents. The academic writing on schemas does not have much to say about how to give schemas such causal powers. However, it would not be too difficult to develop neural nets models that would do such things.

One of the questions that has interested cognitive psychologists is whether the schemas or neural nets that are built up by experience are basically configurations of exemplars or configurations of associated features. My understanding is that there is no clear answer here; some neural nets seem to be modeled best as configurations of exemplars, others as configurations of associated features. This has relevance for the understanding of internal objects. Introspectively, one can usually find a set of images, events, thoughts and feelings centered around one's memories of another person. This would be a set of exemplars - things or events that one experienced - collected together. However, most people seem to have a net of associated features that is not conscious, or less so, derived from normal experience with another person. For a maternal figure, this net typically consists of attributes of caring, nurturance, love, and the like, which organize the way certain people are consciously and unconsciously experienced and responded to. Often there is another net, quite different, consisting of attributes of coldness, hostility, rejection, and the like, also derived from normal experiences with one's mother, which also affects the way certain people are experienced and responded to. These last two nets appear to consist primarily of associated features. Perhaps the problem of splitting involves the relation of three networks; if all three stay well connected, certain people will be experienced, based on the relevant features, as more or less like one's mother, and more or less likely to have both her good and bad features. But if the original connections between the three networks are split, perhaps because the set of exemplars - the conscious memories of one's mother - has been affected in various ways so that the memory of one's mother has been sanctified or demonized (perhaps caused by a preponderance of either satisfying or frustrating and traumatic experiences), then the linkage to one of the feature nets is inhibited and the person has 'split' the maternal object. As a result, real people may be assimilated to either the 'good' or 'bad' network of features, but not to both. As Westen points out, the splitting so characteristic of patients with borderline personality disorder maybe due primarily to histories of abuse or to genetic vulnerability to extreme emotional reactions which in turn draw malignant or unempathetic responses from caretakers. (Westen, n.d.).

The next issue concerns the relation of internal objects to the motivational systems of the psyche. The position taken here is that internal objects are incorporated or recruited into the various sub-systems, but, contra Fairbairn, are not themselves the basis for these systems. The same internal object may be recruited into the operation of more than one motivational system. The internal object(s) derived from experiences with the father, for example, may be recruited into both a boy's identity system and his superego, as well as having a place in his attachment system. Internal objects may be recruited into the identity system through the process of identification, in which various self-representations are given properties of internal objects. They may also play an important role in affecting self-representations based the way an internal object is represented as acting towards the self - e.g. 'I must be worthless because I am rejected.'

Internal human objects are formed by the action of distributed neural networks. A big brain animal will develop neural nets for whatever it has consequential interaction. The importance of internal human objects for the human psyche is a result of the centrality of the attachment system; the fact that the human attachment system is instinctively organized to adapt cognitively, emotionally, and motivationally to other humans. A creature whose attachment system is instinctively organized to interact primarily with nuts and fruits would have a very different set of internal objects. Cats probably have a different set of internal objects than hamsters. While most animal species show interest in con-specifics, humans, apes and canines are remarkable for the strength of pervasive attachments to particular others and the complexity of the discriminations and feelings they experience about these others.

Another instinctual property of the human psyche which should also be discussed is the human experience of things being good or bad. The notion of goodness extends across the sense of utility (this is a good knife - it cuts well), the sense the something being a par excellence example of its kind (Peter blew a good fart - a good example of fart), the sense of liking something (this is good ice cream - I like it), the aesthetic sense (I heard a good recording of Schubert - it was beautiful) and the moral sense of treating people in the right way (Mother Theresa is a good woman - she has dedicated her life to helping the poor). The sense of goodness seems to be a kind of experience or sensation (a qualia?) which is unanalyzable - like the experience of pain or time or space. Unlike time and space, but like pain, goodness does not exist outside of our experience of it. It is hard to know whether other animals experience it, although if one examines sign language communication between humans and apes, one finds apes learn with great ease the signs for good and bad and use them appropriately and egregiously. (Koko the gorilla signs "Penny bad, bad Penny, devil Penny, good Koko, tickle Koko.")

This sense of goodness and badness is of importance in the functioning of the human as an adaptive complex system. The sense of badness is an appraisal - an internal signal - that something has gone wrong. Feelings of anxiety and pain have similar functions. Perhaps for the infant these three experiences are all one. However that may be, pain is a signal that something is presently wrong with the body and anxiety is a signal that something dangerous is about to happen within or without. What then is badness a signal of - especially the sense that oneself is bad? Being derogated or scolded by people one loves seems to be a major trigger for this experience. Agents of the superego and the identity system can also produce the conscious experience of oneself being bad. These operations of internal objects seem to be based on the real experiences of being punished or derogated. It is hard to shame a cat but easy to shame a dog or a child - or an adult if one is a relevant alter.

The relation between the sense of badness and anxiety is complex. People can become anxious because they want to do something which they know will result in feeling they are bad. They can also become anxious because they believe people in the external world will do something bad to them - sanction them, shame them, or hurt or deprive them. Anxiety seems to be a generalized signal of disasters which may occur inside or outside the person, while the sense of personal badness seems to be a more specific signal that a particular kind of disaster (derogation, blame) has happened with respect to attachments - attachments either to external or internal objects. Shame and guilt are specific varieties of this sense of badness.

Anxiety and the sense of goodness/badness discussed above are sub-type of the processes that characterize appraisal. Rosenblatt and Thickstun summarize of appraisal process as follows:

GLITCHES

So far only normal operations of the system has been discussed. But individual human psyches often operate at non-optimal levels. Consider the not unusual example, mentioned above, of a young man who dates various women, and after a relationship with one of them has started, finds himself becoming interested in other women. He is conscious of no reason for this - it just happens. He even feels a little bad about this continuing state of affairs but seems unable to find the right woman - a woman in whom he can stay interested. He interprets this as caused by his having too strong a sexual drive to confine himself to one woman.

System glitches that could account for the young man's problem might run as follows: As a result of various attachment system events and vicissitudes, he has developed a complex internal object. This object has partially split into a rejecting object and an exciting object. Various women interest him because they, in part, are assimilated to the exciting object. The agents of his attachment and sexual system generate wishes and emotions which are both arousing and satisfying. But once he gets into the relationship, bad associations connected to deep intimacy and closeness result in an assimilation of the external object - the real woman he is dating - to the rejecting object.

If this were the only glitch in the system, one would theoretically expect him to experience the woman he is dating as rejecting him in various ways and for him to experience feeling bad about himself and a wish to escape from her and the grief she causes. But while he does seem to be trying to escape from relationships, he does not recount actual situations with his partner in which he has been rejected and feels bad, nor does he experience an active wish to escape, only a loss of interest in his partner and an interest in other women.

Given the story as presented, there obviously must be other glitches at work. Perhaps when much younger he did experience bad feelings in interactions with his mother and others and somehow learned ways of escaping from these feeling by avoiding intimacy and closeness. In time these procedures became automated and are no longer evoked a conscious representation of wishes or feelings.

There are gaps in this account. Basically, what has been left out is the process of repression. One of the general points made by Minsky is that given a conflict between procedures, one way of dealing with the conflict so the system can go forward is for the activation of one procedure to inhibit the activation of the other. Neural nets can affect other neural nets either through inhibition or activation.

Simple inhibition is easily learned. Again, take typing as an example. A novice typist tries to reach the right forefinger up to hit the letter 't.' The other three fingers naturally move up with the forefinger. However, this interferes with typing the next letter if it must be struck by the other fingers of the same hand - like 's' for example, which is struck by the right ring finger. The novice is aware of trying to keep all non-striking fingers on the home keys. The executive system must work at it. However, in time when the procedure to strike 't' is activated, movement of the other fingers is inhibited. Somehow - through Hebbian strengthening of connections, both excitatory and inhibitory, let us say - the upper level goal of striking 't' both moves the forefinger and inhibits the movement of the other three fingers. The inhibition becomes automatized with enough effort and repetition.

Simple inhibition is close to the psychoanalytic concept of repression. But the process of inhibition described for the letter 't' requires much repetition and effort. Another way inhibition may come about is through the effect of a strong unpleasant conscious experience. Suppose a dog is presented with a very attractive bone, then badly shocked when he notices and moves close to the bone. Here, as soon at the goal is activated and the wish to get the bone is registered in consciousness, a very unpleasant experience occurs. After this happens several times, the dog will stay away from the bone, not look at it, etc. In time the dog ignores the bone as if it were not there.

In psychoanalytic terminology the wish for the bone has been repressed and is not conscious. After the first few shocks the sight or thought of the bone resulted in the dog's experiencing anxious anticipations. Avoidance of thought of getting the bone prevented this anxiety, and quickly became strongly learned. In the terminology used here, the agent originally activated by the sight of the bone is now inhibited by the sight of the bone. Since the agent is not activated, there is no conscious awareness of a wish. Obviously, one can say either that 'the wish has been repressed' or that 'the agent has been automatically inhibited and as a result does not generate the consciousness of wish' as an account of the same phenomena. The advantage of psychoanalytic terminology here is that it is closer to ordinary language. But while neural models of inhibition are well understood and easy to simulate, models for the mechanics of the repression of wishes are not. By what mechanism, neural or otherwise, is a conscious wish repressed? In the account given here, the dog's executive system did not operate on the wish to get the bone and push it into unconsciousness. Rather, the dog's experienced a painful state and learned to inhibit 'get food' agent and activate instead other procedures.

As Rosenblatt and Thickstun describe this:

To return to the young man and his problem, what may have happened is that he may have experienced such strong unpleasant conscious states when goals associated with maintaining intimacy were activated that these goals were inhibited and no conscious awareness of wishes for deep intimacy now takes place. From the perspective of his conscious experience, he loses interest. Then other women, not associated with the rejecting internal object, activate various attachment and sexual goals and the process starts all over.

This leaves one further question; what might have created the original unpleasant conscious state when goals associated with maintaining and enhancing intimacy were activated? Any one of a number of scenarios are possible. His mother may have left him at various times in his early life, or she may have been very difficult or abuse, rejecting his attempts at closeness. But it is also possible that the problem was internally generated. In the classic oedipal situation the boy would have been frightened of what his father might do to him for his attempts at intimacy and closeness with his mother. Activation of a particular goal, instead of resulting in pleasant conscious experience, results in an unpleasant conscious experience because of anticipations of danger.

Just to complicate things further, the father may never have actually been punitive with respect to the child behaving in this way. It could be that because of the child's anger at the father for his father's close relation to his mother the boy anticipates that his father will be angry at him for doing the same thing and this frightens him. At various points each piece of this complex layering of glitches would have been conscious, but given very strong unpleasant experiences, many pieces would be quickly replaced by automated avoidances and automated anticipations.

The foregoing is an incomplete account of either normal dynamics or glitches. For example, there are other defenses besides repression, each of which would need to be described within the 'society of agents' model. In this account emphasis has been placed on the importance of automated processes, both with respect to glitches and the ordinary operations of the psyche. The contrast between automated and conscious processes has been found useful in cognitive psychology in the understanding of a wide range of phenomena, including memory (see Mandler, Cognitive Psychology, An Essay in Cognitive Science, 1985 ) and language (Bates, Harris, Marchman, Wulfeck, and Kritchevsky (Production of complex syntax in normal ageing and Alzheimer's disease, in Language and Cognitive Processes, 1995:10:487-539). The figure below is adapted from Bates et al.

Criteria for distinguishing between automatic and controlled processes

Automatic Controlled

unconscious conscious

effortless effortful

no attention required attention required

parallel serial

high frequency low frequency

high predictability low predictability

no decisions required decisions required

fast slow

immune to strategic factors affected by strategic factors

difficult to interrupt of modify prone to interruption and modification

does not interfere with parallel tasks does interfere with parallel tasks

ASSOCIATIVE NETWORKS

Let us consider another example of a maladaptive systems problem. I once knew a middle aged woman who was badly afraid of birds. How can one account for this within the framework presented here? To explicate this, the notion of associative nets needs to be presented.

The human brain is probably composed of millions of neural nets. The ordinary process of association that occurs when one hears a word and various ideas come to mind is a process by which the net which recognized that word activates other nets. Some of these by means of the great mystery reach consciousness as the 'associated' ideas. Read the sentence "Clinton is in trouble again." What comes to mind? This is the associative network at work. And interesting example of the analysis of the associational linkages within a myth has been presented by Edwin Hutchins in his paper Myth and experience in the Trobriand Islands, in Cultural Models in Language and Thought, edited by Holland and Quinn, Cambridge 1987. Claudia Strauss has analyzed associational nets in the discourse of Rhode Island working class about work and achievement in a chapter in A Cognitive Theory of Cultural Meaning. I carried out an analysis similar to Claudia Strauss' on the content of three sessions of a young man in psychotherapy (The identification of schemas in naturalistic data, in Person Schemas and Maladaptive Interpersonal Patterns).

The woman in our example has one or more neural nets that recognize the presence of birds. When this net is activated, it activates neural systems within limbic system - according to LeDoux, in the amygdala - in a way that results in the sensation of fear becoming conscious along with the sight of the bird. But why should this be? By what means could the neural net for recognizing birds have become connected to the amygdala in this fashion?

There are several possibilities. One is that the woman was once frightened by some event while in the presence of birds, and that this experience connected the bird recognizing neural net with the amygdala. This could be the case, but one wonders why the woman does not inhibit this connection. After all, the woman knows that the birds can't really hurt her, that most birds are small timid creatures, not equipped with much in the way of weaponry. If she were afraid only of eagles or large hawks it might not seem so irrational, although even eagles and hawks are not known for attacking humans except under great provocation. Why don't these nets, normally associated with the recognition of birds, inhibit the fear response?

Furthermore, in this case, there is no evidence that the woman was ever frightened by something in the presence of birds. The phobia grew over several decades, beginning with only mild apprehensions about birds, and gradually became more and more severe. So direct conditioning as an explanation seems unlikely. Over the course of a number of conversations she provided certain hints or clues to a different explanation. Several times I heard her say that birds reminded her of her mother. The quick way they move, something about their beaks, the way they hold their heads, gave a peculiar sense that there was something about them like her mother. Not surprisingly, the woman had had a difficult relationship with her mother, who she reported to be a harsh person who had frightened her through much of her childhood.

This woman was timid and had many fears about not doing her job well, losing her position, and so on. My speculative account is that the bird recognizing neural net had a strong set of associative connections to both her mother and a variety of nets encoding experiences and anticipations about the world ('you can't do anything right') taught to her by her mother which were connected to amygdala based sources of fright sensations. As life became more difficult due to a divorce and ensuring job problems, the bird phobia increased because these nets developed stronger and stronger amygdala connections.

There are still problems with this account. In the normal course of the operation of associative nets, one would still expect that other features of birds - their timidity, small size, and lack of impressive means of violence - would act to reassure the woman. For example, one walks into a zoo and sees a very large and powerful bear. One recognizes that the bear is behind a deep trench, and that it cannot possibly get across the trench. One does not feel much fear. Why doesn't something like this happen for this woman?

In the case of the bear in the zoo, the neural networks that recognize the trench are strongly activated by the visual scene so that the conscious gestalt combines information which conveys the fact that the situation is safe. This woman does not have such a conscious gestalt because, one speculates, the neural nets associated normally with birds concerning their timidity and small size do not get sufficiently activated to become part of the conscious gestalt. But why might this be?

It may be that the connections between the bird recognition neural net and the normal associations to birds atrophied because they were so infrequently used. It is not that the connections were not there, but the strength of the connections weakened because they are not used. With each experience of seeing a bird, the fright resulted in no further processing of the associative network; instead the woman avoided thinking about any features of birds. Certainly when the topic came up it was difficult to get her to talk in general about birds. Instead, she would quickly move to associations about how difficult her fright was for her to manage, how it had incapacitated her on her last vacation, etc. Perhaps the more the normal associative connections atrophied, the less the more reassuring associated nets reached consciousness, and so she became less and less 'realistic' about birds. It isn't that her capacity to test reality is impaired, it is just that she doesn't test reality. I call this 'reality blotting.'

The story is probably more complex than this. More than simple connectionist atrophy due to infrequent use may be involved. It may be that something about the associative connection to her mother activated other nets that in turn inhibited the activation of the networks for the needed reassuring features of birds. Perhaps the timidity of birds was associated with her mother's own timidity. On a few occasions she mentioned her mother's timidity. Not that her mother was afraid of birds - the link was not a direct one. Thought of her mother's timidity seemed to cause her some distress, expressed primarily by vague gestures of waving something away. Since some of these nets concerning her mother's timidity may be the cause of distress, an avoidance of these particular nets might have been established. Seeing birds activates (usually unconsciously) a network connected with her mother. This mother network both activates the amygdala, and, at the same time, inhibits other networks which contain information about the harmlessness of birds.

It is interesting that after she went into therapy her bird phobia greatly improved. The occasionally conscious associative links between birds and her mother may have been useful in the therapeutic process, making possible the recovery of normal associations to birds and weakening of the associative connections between birds and her mother. Or, it may be that the woman simply became less frightened of losing her job and became more effective at handling life in general and the old associative links, still there, simply aroused less fear.

An notable part of this process is the loss of reality testing. The woman seems crazy because she is afraid of birds and birds are harmless. A theoretical explanation for her lack of reality testing is that the parts of the normal associative network which would reassure her about birds never reach consciousness. Generalizing this process, we might expect that a borderline who finds someone they liked last week now to be a demon from hell has similar problems of reality testing. The borderline is unable to become consciousness of the good things the new demon did for them last week, just as this woman seemed unable to consciously process material about the harmless of birds. In the language of the object relations theorists, the bad object has taken over, and reality blotting has been automatized.

This kind of complexity between consciousness and unpleasant experiences receives support from experimental work with animals. In the Emotional Brain LeDoux presents a series of experiments carried out with rats. If rats are shocked just after hearing a tone, they become frightened at hearing the tone alone. To trace out the brain pathways involved, LeDoux lesioned various brains areas. He found that damaging the auditory cortex did not affect the strength of the conditioned response of fear. However lesions between the thalamus and the amygdala, or lesions in central nucleus of the amygdala, completely prevented the fear response. Since auditory stimuli enter the thalamus and are go along separate pathways to the auditory cortex and to the amygdala, lesions in the amygdala result in the rat hearing the tone but feeling no fear. LeDoux also found out that if the auditory cortex was lesioned, the rat could not unlearn the fear response once established at it normally wold, no matter how many times the tone was presented without shock.

Interestingly, if the conditioning procedure is more complex, and the shock occurs only when the rat hears a tone at a particular frequency (tones at other frequencies are not followed by shock), a normal rat after training does not respond with fear unless the tone has that particular frequency. In this case, if the auditory cortex is lesioned, the rat responds with fear to any tone because the auditory cortex is needed to carry out a more complex analysis (is this the 'bad' tone). Since stimuli that enter the amygdala directly from the thalamus are not subjected to complex processing, distinctions between tones is not made. The auditory cortex can turn off the fear at the amygdala as soon as it recognizes that what it heard was not the 'bad' tone. (Since the processing in auditory cortex takes time, the fear producing signal which goes directly from the thalamus to the amygdala arrives before the fear inhibiting signal from the auditory cortex, and this should produce at least a small jolt of fear even in a rat that has been well trained to discriminate tones. Perhaps that is why seeing a real large bear up close is almost always a little scary, trench notwithstanding.)

LeDoux speculates that some kinds of abnormal fear are due to failures of this type. A complex situation, which should not give rise to fear, does so because the cortex fails to send down a message that there is nothing to fear. The stimuli from the thalamus that are sent to the amygdala are so broad-band and non-specific that they carry a message that something terrible is about to happen. This 'initial impression' can be corrected by more complex processing. LeDoux's speculation here is similar to the speculation presented above in pointing to the importance of complex processing in the production and inhibition of fear. As a result of associations between nets in the cortex a formerly neutral object (birds) came to be connected to a fear arousing object. More complex processing concerning the harmless of birds should have had the power to inhibit this fear, but repression or inhibition kept this more complex processing from happening.

The complexity of these processes makes explanation of behavior difficult. Why is this woman afraid of birds? The answer cannot be "They are associated with her mother and she is still frighted by her mother," because she knows that birds are harmless. There is a complex of cognitive processes that keep most people from being afraid of all the things that are associated with something frightening. What is different about this woman is that a particular kind of cognitive processing - the processing that informs the amygdala that there is nothing to fear - has been lost.

A psychologist who read this account of the bird phobia in an earlier draft of this paper remarked that it seemed complex and Rube Goldberg like. This is true. The point of all the specifications of 'what goes where' is to show that the model outlined here has plausible neural explanatory power. In the (hopefully near) future it will be possible to record the processing of the brain of a person with a bird phobia, or a person with a remarkable lack of conscience, and observe 'what goes where.' This will result in an enormous increase in knowledge about all kinds of mental processes, ranging from normal processes of perceiving and puzzle solving to neurotic and psychotic processes, that now are matters of speculation with few constraining facts. If bird phobias don't work as outlined above, they do work somehow.

It might be mentioned that from the perspective of this model, the process of psychotherapy is a process of bringing into consciousness materials which, because of various automated glitches, have not been conscious. Bringing such materials into consciousness makes it possible for the executive system to reorganize its operations and for re-learning to take place. According to the model presented here any psychotherapeutic focus that accomplishes this can have significant therapeutic effects, whether the therapist believes that libidinal desires, body parts fears, self-problems, traumatic memories, internal objects, social relationships, or whatever are the 'true' and 'deep' causal motors. In fact, the effects of consciousness on previously automatized operations can be quite surprising. V. Ramachandran, for example, has found that if he can trick people who have lost a limb and have phantom limb pain into experiencing that they are moving their phantom limb (he uses real mirrors), the pain will often disappear (Ramachandran and Blakeslee, Phantoms in the Brain, 1998). Of course, a psychotherapeutic focus which is causally relevant to the patient's glitches will aid the therapist to guide the patient's processing in fruitful channels. However, given good human insight and empathy, a therapist may be able to accomplish this despite mismatches between the therapist's theory and the actual operation of the patient's psyche.

This point is similar to Freud's stress on the importance of consciousness in therapeutic change. Psychotherapy involves reorganization and learning new things. But because automated glitches are the cause of much distress and ineffectiveness, therapy requires bringing to consciousness these unconscious automatized operations. Consciousness is not all, for either patient or therapist, but it is a primary fulcrum of change.

CULTURAL AND SOCIAL IMPLICATIONS

Anthropology and the other social sciences - sociology, political science, history, and even economics - need a comprehensive model of the human psyche. I have been struck lately while reading the greats - Marx, Durkheim, Weber, Parsons - with how strongly their general theories of society and culture are influenced by their models of the psyche. Marx, for example, does not have an autonomous cognitive system nor a superego in his theoretical model of the psyche. He has many ideas about material self-interest and uses this motivational explanation extensively to account for various aspects of society and culture. The lop-sidedness of this view of the psyche results in a theory, which, for all of its insights, distorts social and cultural realities. It is interesting that all the varieties of present day Marxism, while having moved some way towards integrating the 'role of ideas' into their theories, still consistently ignore the causal importance of moral, religious, democratic and other values (Eric Fromm and others of the Frankfurt school are an exception).

INTERNALIZATION

Within cultural anthropology, having a comprehensive model of psyche is of importance because one of the primary goals of cultural anthropology is to produce ethnographic accounts. In trying to describe the patterned ways of life of a human community a sense of the potentials and possibilities of the human psyche is strongly needed to avoid narrow, truncated and distorted perspectives. But there are also more direct benefits. With respect to the understanding of culture, one the most important psychological processes which needs to be understood in any ethnographic study is the degree to which various cultural idea complexes are internalized. Spiro has made the crucial point that the people in a society may continue over generations to reproduce certain ideas, such as the Buddhist doctrines concerning the non-existence of self, but not internalize these ideas - that is, these ideas are held only as cliches and not-to-be-openly-questioned-pronouncements, not as salient explanations of events, nor as bases for important feelings, nor as significant sources of motivation. To do good ethnography it is not enough to simply report that the Burmese believe such and such. Any good ethnographic account should obviously describe both the ways ideas are institutionalized in various roles and the degree to which these ideas are internalized.

Sherry Ortner has said, "Unfortunately, anthropologists have generally found that actors with too much psychological plumbing are hard to handle methodologically..." (Theory in anthropology since the sixties, p 151). Actually, the problem isn't methodological at all, it's conceptual. Without an adequate model of the psyche it is impossible to understand anything psychological whatever methods one might use. How could one understand the loss of the sense of community worth, for example, described by Luhrmann for the Indian Parses in her book, The Good Parsee, without a psychodynamic perspective?

With respect to internalization, a schematic like the five systems model can be of help because cultural idea complexes are incorporated into different systems in different ways. The same idea complex - for example, the idea of a supreme deity - can be incorporated differently into the different systems. The moral prescriptions and proscriptions of the supreme deity can be internalized into the superego and become the source of strong feelings of guilt and goals relating to self-punishment, asceticism, etc. Or a supreme deity can be internalized into the identity system and become the source of a strong identification with a community of like believers along with contempt for unbelievers. Or a supreme deity can become part of the attachment system as an internal object with whom the believer has an important relationship, providing support and love. Or the theological ideas concerning a supreme deity can internalized in the autonomous cognitive system and become part of a highly general theological perspective, consistent and comprehensive. These are not mutually exclusive incorporations; people have a cultural idea complexes incorporated into a number of systems at different levels of internalization (Durkheim made a similar point in The Division of Labor.)

As another example, consider a modern American's ideas about politics. For some Americans, political ideas - general, Republican, Democrat, or Libertarian - are not much more than a joke. For others some of these ideas are incorporated into a highly moral sense of how American society should be lead and run. Other Americans seem to treat political ideas as public badges for distinguishing political parties which characterize one socially and serve as sources of personal status and identity. For other Americans political ideas are experienced through their attachments to presidents and other notable political figures. "The thing about Clinton isn't his politics, it's that some people really hate him," one news commentator said. For a few Americans the ideologies concerning democracy, or free markets, or the abuses of the state, are internalized strongly as part of their autonomous cognitive systems. Understanding the American culture of politics is impossible without understanding the ways different segments of American society have internalized political ideas and persons into different systems of the psyche. Even more complex are the dynamic processes by which such internalizations take place, often interacting with psychological defenses in the way Spiro describes for the Burmese ideology of the polluting vagina in Gender Ideology and Psychological Reality. At the bottom of Figure 1 some idea complexes which are likely to have important cultural instantiations are listed.

ROLE MOTIVATION

The most central idea in the social sciences about the organization of society is that humans relate to each other as they do and do the things they do because they are in roles - the boss and the employee, the husband and wife, the teacher and student, the friend with the friend. Spiro (1987) pointed out that people perform their roles because they are motivated to do so - a culturally system of activities has to be linked to a system of activated goals. Spiro distinguished two classes of goals: First, goals which are extrinsic to the activities of the role, such as the money earned, or the prestige involved, or the fear that one will be punished if one does not follow the proper norms of the role, or a sense of duty ; and second, goals which are intrinsic to the role, such as the destruction of enemies involved in the role of a soldier, or the control over others involved in the role of a leader, or the esteem generated by winning an election.

With respect to the five systems, each contains a variety of general and specific goals. In understanding the motivations involved in role performance, it is important to know not only the particular activities and outcomes involved intrinsically and extrinsically in performing the role, but also the heterarchy of more general goals into which these specific activities and outcomes have been incorporated. Using the systems discussed here, a social analyst can trace for any given role the likely linkages or incorporations to the five systems; the role of a judge or preacher, for example, has a rich set of potential incorporations into the superego, while being a movie star or nobel prize winner has a rich (perhaps too rich) set of potential incorporations into the identity system. Caste roles in India or Nepal, on the other hand, present a series of great potential difficulties with respect to the identity systems of individual untouchables, as Steven Parish has described. Becoming a mother or a friend is obviously strongly related to the goals of the attachment system, and so on. (See the social role and system linkages in Figure 1 for some further examples.)

Tracing out the actual linkages or incorporations for any given individual can only be done through interviewing and observation, but often some linkages are relatively easy to discern. Of course some linkages are difficult to discover because they involve goals that have been blocked from awareness, about which people cannot introspect accurately, and where what is said is likely to be defensive camouflage. But a great deal can be discovered from informants that does not require specialized listening skills.

CULTURAL REALITY MODELS

Like internal objects, the person's models of the world, culturally or idiosyncratically learned, provide a vast interpretive system by which events are organized and given meaning. The Development of Cognitive Anthropology contains a summary of the growth of the work of anthropologists and others oriented to the study of cultural models.The theoretical understanding of such schemas within a connectionist framework has been outlined by Strauss and Quinn in their book, A Cognitive Theory of Cultural Meaning. One thing missing from these accounts is a general theory of the relations between cultural models and motivational systems. Central to such an enterprise is an account of how cultural models become agents or motives. That something like this happens is clearly the case (see Human Motives and Cultural Motives, edited by D'Andrade and Strauss). There are a number of potential routes by which cultural and idiosyncratic models or connectionist schemas become agents; for example, the representational content of some schema can be subsumed by a more general schema (e.g, for someone helping others is a strong agent, children in the Sudan are assimilated to the category of relevant others, and helping these children become an agent), or a 'means-end' linkage can established (e.g., helping others is a strong agent and giving money to the Red Cross is thought to be an effective means to helping others). A variety of processes by which cultural models become active goals are discussed in Human Motives and Cultural Models, and an interesting theoretical account is given in Spiro's Gender Ideology and Psychological Reality, 1997, Yale University Press.

ANTICIPATED OBJECTIONS

I can anticipate that some social anthropologists might say - "Well, I don't really need to know all this stuff. I don't need to know what they have cathected, whatever that means, or how activated goals fit into five or twenty systems, or whether they are automated or not. All I need to know is the simple fact that they perform the role." There are a number of answers to this objection; for this paper suffice it to say that behind this objection is a misguided view of society. Roles are an important part of any society, but an account of just roles is not adequate. People in roles are constantly responding to various exigencies, trying out new strategies, re-negotiating role norms and systematically violating others, as Bailey in his book Strategisms and Spoils describes in detail. If these facts of life have to be excluded from ethnography, the result is a static and one-dimensional description. "Structures don't march in the street" - a critique Adam Kuper noted among anthropologists dissatisfied with structuralism. To get into the analysis of matters of strategy and negotiation and norm violation requires an understanding of which activities and outcomes have been cathected and which have not, along with understanding how these activities and outcomes fit into more general goal systems so that the individual's economy of choices can be understood and related to other social and cultural conditions. Agency requires agents. And human agents have in their psyches a whole society of agents.

TERMINUS

Because the human psyche is complex and not directly perceptible, any account of its organization and dynamics is likely to be wrong or even misleading. Accounts of the connections between the human psyche and culture and society are even less certain. The account given here is provisional. It may turn out that it is not the attachment system which has global top priority, but the identity system, as Kohutians argue. Or the sexual system, as Freud proposed. Or that there are other systems, not mentioned here, which need to included in the account of the psyche. Similarly, the account given here of the relation of goals to the experience of desire may be wrong, or it may turn out that repression is not central in producing maladaptive automated behavior, or that repression does not occur in the way described here. None of these issues affect the importance of working towards a general theory of the psyche. Without even an outline of a general theory, psychological anthropology, social psychology, cultural psychiatry, as well as psychoanalysis and related fields which attempt to relate the psyche to the real world of human life will continue to have the fragmented, patchwork quality currently characteristic of much current work. While there are problems in the endeavor, advances in the cognitive sciences over the past forty years have made it possible to see more clearly the outlines of such a useful theory.

***