Brain Organization and Psychodynamics

Abstract

Any attempt to link brain neural activity and psychodynamic concepts requires a tremendous conceptual leap. Such a leap may be facilitated if a common language between brain and mind can be devised. System theory proposes formulations that may aid in reconceptualizing psychodynamic descriptions in terms of neural organizations in the brain. Once adopted, these formulations can help to generate testable predictions about brain–psychodynamic relations and thus significantly affect the future of psychotherapy. (The Journal of Psychotherapy Practice and Research 1999; 8:24–39)

It is generally assumed that mental functions are correlated with neural activity in the brain. In 1895 Freud tried to explain the physical (neuronal) roots of mental phenomena in an ambitious manuscript titled “The Project.”¹ Today also, with modern knowledge about the brain, Freud's quest is still probably a highly relevant enterprise. Any attempt to link brain neural activity and psychodynamic concepts requires a tremendous conceptual leap. Such a leap may be facilitated if a common language between brain and mind can be devised. New concepts from system theory are beginning to suggest various ways in which the leap from basic, simple neurological processes to complex psychological phenomena might be possible.^2–4 By introducing concepts from system theories, we attempt to propose a language that links psychodynamic concepts to neural organization in the brain.

It is just as important to understand the effects of psychotherapy on the brain as it is to study the effects of medications on the brain. Initially the formulations from this work may serve only as a theoretical paradigm, but already at this stage they enable us to ask questions about brain–psychotherapy relationships not possible otherwise. Subsequently this paradigm may provide testable predictions related to reciprocal alterations between psychodynamic therapy and brain functions. Results of studies designed on the basis of these predictions will have great impact on the future of psychoanalysis as a science of the brain.

In a very simplistic way, the brain can be described as a system composed of many interconnected units. These units can be taken to be single neurons, groups of neurons, neural systems, or brain regions. Similarly, interconnections may be synapses or neural pathways. However simplified, this formulation of the brain system is sufficiently powerful to conceptualize a theoretical framework within which psychodynamic concepts can be related to the brain.

The state-space formulation of systems describes the dynamic activity of the brain in physical terms. The formulation of brain organization in terms of hierarchically and multiply constrained systems proposes a possible mechanism that may underlie psychodynamic phenomena. Attractor systems in the space-state of the brain represent information (i.e., the complex internal representations of psychic experience). The dynamic interplay between the internal representations and the neural organization provides the language proposed to bridge the relations between psychodynamic phenomena and neurophysiological events in the brain.

SYSTEMS: BASIC CONCEPTS

State-Space Configuration of Systems and the Concept of Attractors

Imagine a system formed from many elements. The arrangement of the elements in the system represents the states of the system. Each distinct arrangement in the system forms a different state for the system. If the elements are arranged randomly, all of the states in the system are similar to each other. If the elements of the system can form many distinct patterns of arrangements, then the system has many possible states. If the system can form only one type of arrangement, then the system is represented by one state only. The space of a system is represented by all of the possible states a system can assume. If the system changes over time, it is called a dynamic system. In this case, the system changes its arrangement from one point in time to the next.

To visualize systems and their dynamics, the physicist William Hamilton and the mathematician Karl Jacob devised the concept of state-space necessary for describing dynamics in physical systems.⁵ A dynamic system is generally defined by a configuration space consisting of a topological manifold (Figure 1). A point on the configuration space represents the state of the system at a given instant. Each point is a combination pattern in the activity of the elements (i.e., their arrangement). The configuration space of the system represents all of the possible states that the system is capable of assuming (i.e., all of the possible combinations in the activity of the elements). This configuration space is sometimes termed landscape (Figure 1). As the dynamic state of the system changes over time, the combinations in the activity of the elements change (that is, the points on the space change). The dynamics of the system are described in terms of state-space as “movement” from one point to the next point on the landscape, defining a trajectory, or curve, on the configuration space (Figure 1, states 1, 2, and 3, over times t₁, t₂, and t₃).

Figure 1

If the system “prefers” certain states (i.e., arrangements) over other states it will tend to be “drawn” or “attracted” to form these states. Once certain states are preferred by the system, they form attractors (basins) in the topological surface.³ If a metaphorical ball were rolling on the surface (space) in Figure 1, it would be easy to see that peaks represent repellors (those states the system tends to avoid) and basins represent attractors (those states the system tends to assume).

It has been shown that learning in the brain involves strengthening certain patterns of neuronal activity.^6,7 These patterns embed information in the form of memories. The brain has a tendency to be “drawn to” and to activate these patterns when memories are remembered. Thus, one can view information (i.e., memories) in the brain as embedded into attractor configurations of neuronal systems.

Hierarchy in Systems

Systems tend to have hierarchies. For example, if one considers a social system (e.g., a commercial company or a military force), one finds that various levels of organization characterize these hierarchies. Commercial companies have boards of directors at the management level. Typically these companies have subordinated departments, each with its own director responsible for an even lower level, and so on down to the single employee. Similarly, military forces are hierarchically organized in a command chain, from a central command down to single units.

Biological systems show increasing hierarchical organization as they ascend the evolutionary scale. In the human body, the leading system is the nervous system, which controls and coordinates subordinated systems such as muscle activity (through motor pathways) or internal organs (through autonomic nervous system pathways). Hierarchy is also evident within the nervous system itself. High-level information processing is computed by the associative areas, which occupy the highest levels of the hierarchy. These areas integrate widespread information from all parts of the brain system. Primary cortical regions deal with specialized lower level processes and are thus subordinate to the associative processes. Similarly, departments of a commercial company or units of a military force are subordinated to the controlling and coordinating functions of the leading management or command.

Hierarchical order can be extended both in time and space. To describe evolution is to describe the development of hierarchy in time. Over time, biological organisms have developed increasingly hierarchical organizations by increasing subordinating processes. Hierarchy in space is described in terms of microscopic versus macroscopic dimensions. Molecules organize to form membranes, membranes organize to form cells, cells organize to form tissue, tissues form organs, and the hierarchy continues from microscopic units to macroscopic whole entities.

Once a system is hierarchically organized, two types of processes are formed in the hierarchy: a top-down and a bottom-up. The top-down processes reflect the coordinating and commanding influences traveling from higher levels down to the lower levels. These are the instructions of the management to the departments of the commercial company or the instructions of the central command to its units. The bottom-up processes reflect the influences exerted by the lower level units on the higher level units. This is best demonstrated in social organizations where leaders are elected. The lower level components of the system determine the higher level leadership. However, even in organizations that are not democratic, such as a military force, the command has to consider the situation of its subordinated units in order to reach optimal function. Thus, the influences exerted upward toward the command level by the conditions of the units can be viewed as a bottom-up process.

One of the first descriptions of hierarchical organizations of the nervous system was introduced by John Hughlings Jackson in 1869⁸ under the concept of “neuro-axis.” The neuro-axis is a rostral-caudal dimension of the nervous system. Jackson argued that the more “atomic” aspects of sensory and motor processing take place further down the neuro-axis, leaving the cortical areas to deal with increasingly generalized issues. Lower level components of the sensory system process specific incoming stimuli (e.g., visual, auditory) and higher levels integrate these processes to form the complex multimodal human experience. Higher level motor systems plan actions and decisions; lower level motor subsystems execute the numerous specific motions that carry out behavioral responses. Jackson's perspective of evolutionary biology led him to conclude that during development of the brain, lower level organizations interact, evolving into more complex, higher level formations. This is the bottom-up process.

Once formed, higher levels of neuronal organization regulate and dominate the activity at the lower levels. A good example of this top-down process is demonstrated by the primitive reflexes of newborns (sucking reflex, grasp reflex, and others). These reflexes result from lower level neuronal organizations of reflex arches in the spine and brainstem. As the cortex develops, these primitive reflexes are inhibited, controlled, and subdued by higher level organizations of more complex behavior. These reflexes are not lost, but are integrated in the more complex behavioral patterns of the mature nervous system. When neuronal damage destroys the cortical higher levels of organization, these reflexes reappear and are typically used in the clinical diagnosis of cerebral damage. One can conceptualize the primitive reflexes of the lower levels of the brain as the building blocks of more complex behaviors. These more elaborate behaviors result from the multiple interactions among the simple reflexes. Such interactions are also those that repress the expression of the reflexes in their original form. Damage to cortical structures can be viewed as accompanied by a breakdown in multiple interactions of certain brain functions. Without any interactions, the primordial reflexes are left to function independently and reappear in the behavioral patterns of the individual.

Multiple Constraint Satisfaction and Functional Connectivity

Imagine a system in which each element can assume a large number of different states and dynamically change its state over time. If each element is left to itself and can change its state freely without “concern” for the states of the other elements, this element can indeed assume each of all of its possible states. Once connectivity is established between the elements of this system, the connections create interdependencies between the elements of the system. Because of influences exerted via the connections, each element in the system can no longer act independently of the other elements in the assemblage. Some of the states that were possible for the elements before connectivity was established are now unrealizable; they may “contradict” the constraints of the connections. As a result, the number of states that elements in an interconnected system may assume is reduced. The activity of the element is constrained, via the connections, by the activity of coupled elements. Constraint satisfaction is achieved if the states of the interacting elements comply with the connection between them (i.e., assume values that comply with connection strength).

An example of multiple constraint activity can be taken from social interactions: people tend to behave and plan their actions and responses on the basis of a large range of considerations. In other words, decision-making (the “state” that the individual assumes) is often guided by multiple considerations. For most people, deciding on a vacation involves considering parents' and children's needs, occupational (work) constraints, and other circumstances. This is because we are part of a social network where the connections made are also constraints to consider.

Information theory states that information in a system increases with the decrease of probability (chance). The more multiple constraints in a system, the less chance it allows, and thus the more information it can embed.³ The process of deciding on a vacation is a good example of how increasing constraints increases information. The information about the type, timing, and duration of a vacation is primarily dependent on the constraints. If you do not know when the children get a vacation from school, what your financial constraints are, and what the preference of your wife is, you have many uncertainties regarding your vacation. If a colleague at work asks you where, when, and for how long you plan to go on a vacation, the amount of information you can provide him about your vacation will be reduced compared with a situation where all the constraints mentioned above are known.

Once multiple constraint satisfaction is achieved by a system, elements that do not satisfy the constraints do not have a place in the system. If a new element is to be added to the system, it has to comply with the multiple constraint pattern of the system or else it is excluded from the system. If many elements that do not satisfy the multiple constraints are introduced to the system, they may disrupt the pattern of existing multiple constraints.

Each neuron in the brain receives input of excitatory and inhibitory stimuli from many other neurons via its dendrites, and it sends output to many other neurons via its axon. The activity (i.e., firing action potentials) of the neuron is determined mostly by the inputs it receives. Thus, the activity of each neuron is multiply constrained by the neurons that are connected to it. Because neurons are organized in networks where high connectivity is typical, a multiple constraint system is created. In such a system, complex interdependencies among neurons do not allow “independent” (chance) activity in the neurons; rather, the neurons have to “consider” the mutual dependencies due to the connections of the network.

Functional connectivity is the influence that the activity in one particular group of neurons has on the activity of another distinct group of neurons.⁹ The importance of functional connectivity to brain function was recognized by pioneering neuroanatomists as early as 1911.¹⁰ Ramcn y Cajal¹⁰ was one of the first to realize that information can be stored by modifying the connections between communicating nerve cells in order to form associations. Later this idea was formalized by Hebb,⁶ who proposed that the connection from one neuron to another, rather than being a fixed, passive conductor like a piece of wire, could be increased in strength when both neurons were simultaneously active, so that a neuron could subsequently be made to excite the next neuron more easily than before.⁶ Hebbian modification of influences among neurons alters the multiple constraint satisfaction formed by the brain. If one neural system strongly influences another system, it also heavily constrains its activity. The stronger the connectivity, the stronger the constraints. As mentioned above, information increases in systems of multiple constraint satisfaction. This explains the relationship between increase of constraints in the brain and the ability to store information (i.e., memories and representations).

Optimization

Optimization is defined as the condition in which elements in a multiple constraint system “satisfy” their constraints. A system may achieve different levels of optimization depending on the extent to which multiple constraint satisfaction is achieved by its elements.¹¹ The nature of the multiple constraints depends on the nature of the interactions between the units (strong or weak, excitatory or inhibitory). Multiple constraint organization in the brain is set according to the Hebbian algorithm.⁷ Once a set of constraints is established, the interconnected units act to satisfy these constraints. When multiple constraint satisfaction is achieved, the system reaches a specific pattern of activity. In other words, the system reaches a state in the space of the system (see Figure 1). It is as if the set of constraints pushed the system to assume a certain state. As described above, the system is drawn to a certain state (an attractor). Thus, one can see how setting the values of the multiple constraints creates attractors in the space-state of the system. As mentioned above, information (memories or representations) can be embedded in attractors,¹² in concert with the notion that the Hebbian algorithm embeds information (learning) in the system.⁶ It does so by setting the constraints to form new attractors. The information embedded as a whole can then be viewed as attractor systems in the state-space of the system (the brain).

To “activate” a memory embedded in the system, the pattern representing this memory is activated. This pattern is a point in the space-state of the system. This point is “located” in an attractor (i.e., a basin) because of the setting of constraints by the Hebbian learning. The activation of memory patterns in multiple constraint systems was described as a gradient descent,¹² as if a ball were rolling into a basin of a topological space. Another way to describe this activity is as a convergence into an attractor. The system tends to converge and assume the state of the attractor because of the strengthening of connections between the units activating the attractor state. An additional term used to describe memory retrieval is optimization. The units optimize their values to satisfy the set of constraints, gradually reaching the best pattern that satisfies the constraints (i.e., the gradient descent) and thus fits the memory activation pattern.

The brain is dynamic.¹³ Constraints change over time, and memories in the attractors can be activated and then deactivated. Being dynamic, the system can activate sequences of attractors moving from one state in the space of the system to the next state (see Figure 1). As mentioned above, sequences of convergences into attractors form trajectories in the space-state of the system. Presumably the physical system of the brain provides a highly complex multidimensional attractor system that embeds representations of the world. Such mental activities as thinking (i.e., sequences of representations) presumably delineate a complex and continuous trajectory in the attractor-space of the brain.

THE BRAIN: A DYNAMIC SYSTEM

Neuronal Plasticity

The past 20 years have witnessed something of a revolution in our understanding of synaptic transmutation and its regulation. Synaptic transmission is no longer viewed as being mediated by static processes of fixed strength, but rather by dynamic ones, continuously regulated by many factors. In 1973 a group of researchers published the first detailed reports on artificially induced modification of synaptic strength.¹⁴ They found that stimulation of certain neuronal fibers with high-frequency electrical pulses caused the synapses of these fibers to become measurably stronger (i.e., to have increased capability to stimulate postsynaptic potentials) and stay so for many weeks. Their observation, which they called long-term potentiation (LTP) was probably the first report of synaptic plasticity.

One critical component of the induction of synaptic plasticity in virtually all experimental models is a change in postsynaptic (sometimes presynaptic) membrane potential, usually a depolarization. There are two other common features of synaptic plasticity. First, Ca²⁺ typically plays an indispensable role in triggering synaptic change. The elevation of Ca²⁺ may arise via flux through membrane channels, release from intracellular stores, or both. Second, plasticity usually comes in two general forms: short-term, which is dependent on post-translation modifications of existing proteins, and long-term, which is dependent on gene expression and de novo protein synthesis. It is increasingly apparent that for many experimental models, a vital bridge between initial induction of plasticity and its maintenance across time is the activation of adenylate cyclases and protein kinases A. One of the more studied mechanisms of regulating Ca²⁺ flux in synaptic transmission relates to the N-methyl-d-aspartate (NMDA) excitatory amino acid receptor. Over the years, it has become apparent that many subcellular systems combine in a complicated way to regulate Ca²⁺ flux and levels—for example, the phosphoinositide system, the G protein systems, and the neuronal membrane currents. (For detailed explanation of the relevance of these systems to synaptic plasticity, see Wickliffe and Warren.¹⁵)

Experience-Dependent Neuronal Processes

In a series of experiments with the marine snail Aplysia, Kandel has demonstrated how synaptic connections can be permanently altered and strengthened through the regulation of learning from the environment.¹⁶ Kandel found structural changes in neuronal pathways and changes in the number of synapses related to learning processes in the Aplysia. Essentially, LTP is the mechanism by which Aplysia learns from experience at the synaptic level, and the experience-dependent process then translates into structural, “hard-wired” alterations.¹⁷

In another series of experiments, with monkeys, the map of the hand in the somatosensory cortex was determined by multiple electrode penetrations before and after one of the three nerves that enervate the hand was sectioned.¹⁸ Immediately following nerve section, most of the cortical territory that previously could be activated by the region of the hand enervated by the afferent nerves became unresponsive to somatic stimulation. In most monkeys, small islands within the unresponsive cortex slowly became responsive to somatic stimulation from neighboring regions. Over several weeks following the operation, the previously silent regions became responsive and topographically reorganized.

Studies of the primary visual cortex in mammals typically show experience-dependent changes.^17,19 The blockade of spontaneous retinal discharge prevents the segregation of the afferents from the two eyes into ocular dominance columns; this finding suggests that spontaneous activity may promote axon sorting. Ganglion cells in the developing retina engage in coherent oscillatory activity,¹⁷ which enables the use of synchronous activity as a means of identifying the origin and neighborhood relations of afferents. However, a substantial fraction of neurons in the primary visual cortex, especially those in layers remote from thalamic input, develop feature-specific responses only if visual experience is available. Visual cortical “maps” in these layers can be modified by manipulating visual experience during a critical period of early development.¹⁷

On the more general and psychological level, experience-dependent alterations have been reported in humans²⁰ and other mammals²¹ suffering from maternal and social deprivation in critical periods of infancy and childhood. A complex set of social and intellectual skills proved to depend on experiential stimulations at an early age.^20–22

Assembly-Forming Connections and High Mental Functions

As mentioned above, Ram¢n y Cajal was one of the first to realize that information can be stored by modifying the connections between communicating nerve cells in order to form associations.¹⁰ Acquisition and representation of information basically entail the modulation of synaptic contacts between nerve cells.¹⁹ Information is stored by facilitation and selective elimination of synaptic links between neuronal aggregates that represent discrete aspects of the environment. Thus, memories are essentially associative; the information they contain is defined by neuronal relationships.

Hebb⁶ proposed that “two cells or systems that are repeatedly active at the same time will tend to become associated, so that activity in one facilitates activity in the other.” This is called the “principle of synchronous convergence.”²³ Through summation of temporally coincident inputs, neurons become associated with one another, such that they can substitute for one another in making other cells fire. Further, connections between input and output neurons are strengthened by recurrent fibers and feedback. By these associative processes, cells become interconnected into functional units of memory, or Hebbian “cell assemblies.” The functional importance of synchronous convergence in the mammalian cortex is well documented.¹⁷

Simple sensory memories, or images, are probably formed in cell assemblies or modules of the sensory or parasensory areas of the cortex.²³ There is a hierarchy of perceptual memories that ranges from the sensorially concrete to the conceptually general.²⁴ At the bottom resides the information on elementary sensations; at the top, the abstract concepts that, although originally acquired by sensory experience, have become independent from it in cognitive operations.²³ This information process is most likely to develop, at least partially, by self-organization^25,26 from the bottom up; that is, from sensory cortical areas toward areas of association. Thus, memory networks appear to be formed in the cortex by such processes as synchronous convergence and self-organization.²³ In the higher levels, the topography of information storage becomes obscure because of the wider distribution of memory networks, which link scattered domains of the association cortex, representing separate qualities that, however disparate, have been associated by experience. Because these higher memories are more diffuse than simple sensory memories, they are in some respects more robust. Only massive cortical damage leads to the inability to retrieve and use conceptual knowledge—the “loss of abstract attitude” described by Ariety and Goldstein.²⁷

Like sensory information, motor information on planning and deciding has also been hierarchically described.⁸ As first suggested by Jackson,⁸ the cortex of the frontal lobe computes the highest levels of motor information. At the lowest cortical level is the primary motor cortex, representing and mediating elementary motor acts. The prefrontal cortex, conventionally considered the association cortex of the frontal lobe, represents the highest level of the motor hierarchy.²⁸ This position connotes a role not only in the representation of complex actions (concepts of action, plans and programs) but also in their enactment, including those such as working memory.²⁹ The prefrontal cortex develops late, both phylogenetically and ontogenetically, and receives fiber connections from numerous subcortical structures, as well as from other areas of the neocortex.³⁰ This extensive connectivity links reciprocally the perceptual and conceptual information networks of the posterior cortex with prefrontal motor networks, thus forming perceptual-motor associations at the highest level.^28,31

Usually information, although stored in the connections of neuronal networks, is “dormant.” A network is said to be reactivated when the information it represents is retrieved by the associative process of recall or recognition.^2,23 High mental functions such as working memory and abstraction involve activating more complex associations of multiple networks. For example, in delayed-response tasks such as those involving working memory, a multitude of networks spread in the brain are activated. Typically, working memory tasks involve choosing between stimuli presented in a sequence.²⁹ Developing a certain pattern of correct selection requires the formulation of a plan for choosing the stimuli. Delays between consecutive stimuli are necessary to hold this plan in mind over time. Thus, the performance of working memory tasks demands the simultaneous and parallel association between short-term memory (the past stimulus), instantaneous perceptual memory (recognition of the current stimulus), and long-term abstract conceptual memories (the rules for choosing the next stimulus). In short, the higher the mental function performed by the brain, the more intricate are the associations among distributed neuronal networks. The more complex the mental function, the more the brain draws upon associative resources of functional connectivity. Similarly, the more “connectional power” achieved by the brain, the higher the potential for better mental function in the individual.

An important aspect of the multiple connections among neural networks is consistency,³² which results from the multiple constraints that are formed once different neuronal systems become connected.³³ If prior to establishing a connection two neural systems could act independently of one another, once their activity is interdependent, the activity of one neural system or network will influence the activity of the other. This might explain the internal consistency we experience in our mental functions and why reality is perceived as being coordinated auditorily, visually, and tactually. Planning, thinking, and acting also have consistency; thoughts and reactions are goal-directed to the stimuli at hand, and they match situational events. Finally, our entire experience seems united in one complete logical and meaningful continuity.

Integration Versus Segregation in Cortical Organization and Neural Complexity

The important interplay between integration (functional connectivity) and segregation (functional specialization of distinct neural subsystems) is captured by introducing the mathematical concept of neural complexity (C_N).³⁴ C_N is low for systems whose components are characterized either by total independence or total dependence. C_N is high for systems whose components show simultaneous evidence of independence in small subsets and increasing dependence in subsets of increasing size. Different neural groups are functionally segregated if their activities tend to be statistically independent. Conversely, groups are functionally integrated if they show a high degree of statistical dependence. Functional segregation within a neural system is expressed in terms of the relative statistical independence of small subsets of the system, and functional integration is expressed in terms of significant deviations from this statistical independence.³⁴

A more complete characterization of the functional connectivity of the brain must relate it to the statistical structure of the signals sampled from the environment. Such signals activate specific neural populations, with the result that synaptic connections between them are strengthened or weakened. In the course of development and experience, the fit or match between the functional connectivity of the brain and the statistical structure of signals sampled from the environment tends to increase progressively through processes of variation and selection mediated at the level of the synapses.²⁶ These processes are particularly well demonstrated by the organization of primary visual areas. Within a visual area, the connectivity is initially organized in a uniform way. During development and experience, it undergoes a selection process such that the groups of neurons responding to similar orientations become preferentially connected.³⁵ The resulting functional connectivity, which constitutes a basis for various gestalt criteria, matches the prevalence of extended colinear edges in the retinal image.

Tononi and co-workers³⁴ introduced a statistical measure called matching complexity (C_M) that reflects the change in C_N observed when a neural system is receiving sensory input. Through computer simulations, they showed that when the synaptic connectivity of a simplified cortical area is randomly organized, C_M is low and the functional connectivity does not fit the statistical structure of the sensory input. If, however, the synaptic connectivity is modified and the functional connectivity is altered so that many intrinsic correlations are strongly activated by the input, C_M increases. They also demonstrated that once a repertoire of intrinsic correlations has been selected that adaptively matches the statistical structure of the sensory input, that repertoire becomes critical to the way in which the brain categorizes (i.e., perceives) individual stimuli.

SYSTEM THEORY AND CONSCIOUSNESS THEORY

Information Processing and Consciousness

Building on a “contrastive analysis” that compares conscious versus unconscious processes across numerous experimental domains, Baars³² presents an integrative theory of consciousness called global workspace theory. Baars's theory is founded on the view that the brain is composed of many different parallel processors (or modules), each capable of performing some task on the symbolic representations that it receives as input. The modules are flexible in that they can combine to form new processors capable of performing novel tasks and can also decompose into smaller component processors. Baars treats the brain as a large group of separable partial processors, very specialized systems that function at the unconscious level much of the time. At least some of these partial processes can take place at the conscious level when they organize to form global processes. Global processes carry the conscious information and are formed from competing and cooperating partial processors.³²

According to Baars,³² conscious awareness is subject to internal consistency. This implies that multiple constraint satisfaction characterizes the interacting partial processors when they participate in the global process. This model of the brain is fairly well supported by evidence from brain studies and studies of patients with brain damage.²⁸ The model also complies with the notion that systems are composed of interacting elements (i.e., information processors) and are multiply constrained.

To explain the differences between conscious and unconscious processes, Baars turns to the popular models of distributed processing systems (i.e., neural network models³). These models rely on a globally accessible block of working memory to mediate communication and novel interaction between the various individual processors. Baars proposes that a similar structure exists in the human brain, and that it supports conscious experience. The structure, which he terms the global workspace, is accessible to most processors, meaning that most processors potentially can have their contents occupy the working memory. The global workspace can also “broadcast” its contents globally in such a way that every processor receives or has access to the conscious content. Significant, though, is the idea that only one global process can be conscious at one instant of time. In other words, consciousness is a serial phenomenon even though its unconscious predeterminants are parallel processes.

Baars's important claim about consciousness is that it has internal consistency, a property not shared by the collection of unconscious processes in the brain. Baars cites as an example of this property the experience of viewing a Necker cube, an optical illusion that we can consciously see in one of two different orientations. The two views of the cube can flip back and forth, but we cannot entertain both of them simultaneously. In other words, our conscious experience of the cube is consistent. A similar situation is found with ambiguous words. People seem to be capable of having but one meaning of a given word in mind at one time. There is evidence, though, that the alternative meanings are represented unconsciously in the brain at the same time as the conscious meaning, in that the other meanings of such words often show priming effects on sentence comprehension.^36,37 This indicates that, while conscious processes are consistent, the collection of unconscious processes is not.

To summarize: Baars postulated a theoretical workspace where global processes are formed from the interactions of many partial processes. He postulated that the global formations in the workspace carry the global dominant message of conscious awareness.³² Partial processes are specialized processes, each processing its information in an independent fashion; they function in parallel, and if not involved in any global organization, they proceed disconnected from other processes. Partial processes compete, cooperate, and interact to gain access to and participate in global organizations. The global formation may be viewed as a complex network of partial processes.

In global formations there are internal consistencies, and thus multiple constraints are formed between the partial processes. When partial processes participate in the organization of a global process, they are constrained by the activity patterns of the global formations. Thus, partial processes can no longer function (i.e., process information) without regard for the global message. Partial processes are fast, highly specialized, and aimed at handling certain specific types of information. They are, however, limited in the extent of the information they can process, and they lack the flexibility and adaptability acquired when many partial processes combine and cooperate to act together. Global formations have the advantage of both complexity and flexibility needed for efficient and elaborated information processing.³²

Combining Baars's theory with the notions we described earlier about hierarchical organization of information (memories) in the brain, it is reasonable to consider that lower level partial processes in the nervous system interact to form higher level neural global organizations. In addition, the idea of internal consistency in global formations captures the basic notion of multiple constraint organizations. It is assumed that the dynamic activity of partial processes in their organization demonstrates both hierarchical and multiple constraint organizations. For example, once the partial process makes up part of the global organization, it is interconnected with all the other processes and thus is broadcast globally. Thus, it contributes to, or influences, the global organization by virtue of its connections—by exerting its output through the connections to the rest of the system. On the other hand, because it is a multiple constraint system, many other processes will constrain (through the connections) its activity. One may conclude that from the information processing point of view, the information delivered by partial processes influences and at the same time is influenced by the global message.

Because of internal consistency, if the information structure (i.e., activation pattern) of the partial process “contradicts” (i.e., markedly differs from) the information being represented in the global formation, the partial process will have “difficulty” gaining access to (or fitting with) the global process. This difficulty is due to the multiple constraints between the partial process and the global formation that are not satisfied in this case. Since global formations are higher levels of organization (from the hierarchical point of view), constraining partial processes, which are probably of lower levels, permits formation of top-down controls. Partial processes compete for access to global formation, creating the bottom-up procedure.

By unifying the theoretical considerations detailed so far, we assume that 1) consciousness arises as a property of global organizations; 2) different levels of consciousness and awareness correlate with different levels of organization in global formations; 3) since partial processes in their segregated form do not support conscious phenomena, they remain in the “unconscious domain”; thus, the unconscious is simply a lower level of information processing that has no access to global formations; and 4) the subconscious is the level of information on the border of gaining access to, or dropping out from, the global formations that are part of higher level organizations.

SYSTEM APPROACH TO PSYCHODYNAMIC FORMULATIONS

System Approach to Freudian Concepts in Psychoanalysis

The first concepts introduced by Freud in his topographic model¹ were related to the levels of consciousness. We now have the tools to define his description of conscious, unconscious, and subconscious as levels of integration that partial processes achieve to form global organizations. Conscious awareness is the property of global formations. Unconscious information is presented by partial processes that do not make part of the global organizations. The subconscious is characterized by those processes that are about to make part of, or drop out of, the global formations.

In the structural model,¹ psychic “compartments” such as the ego and id were conceived. The ego is described as developing from where initially all was id in the infant. The id is described as a disorganized system where concepts are disconnected or dissociated in every “strange” possible way. Freud named this form of inconsistency primary thought process. From the system point of view described so far, primary thinking can be conceptualized as a feature of a system without internal consistency, or, in other words, where multiple constraints are not satisfied. This enables conflicting ideations to coexist and concept formations that do not make any sense to predominate. Biological evidence shows that in infants, synaptic connectivity is just beginning to develop.²⁸ Thus the biological neural correlate at this phase of development cannot support the needed multiple constraints organization that forms the basis of ordered mental activity. Ego development involves the formation of secondary thought process.¹ This process is described by Freud as the normal thinking that characterizes each one of us. In other words, secondary thinking emerges from multiple constraint satisfaction organization of the neural system; and in fact, synaptic connectivity fully matures from infancy to adulthood. By introducing the concept of superego, Freud suggested what were later to be developed as internal representations of social and interpersonal norms. It gave the ego (i.e., its superego portion) not only the scope of organizing the disordered id processes, but also the entire responsibility of representing, and adapting to, psychosocial reality.

Introduction of the dynamic model¹ added the interplay among the psychic compartments of Freud's model. Defense mechanisms are probably the most described dynamic factors in this model. According to Freud, the ego makes use of an unconscious domain of mental activity (also referred to as id) into which undesirable drives and ideas are repressed. Repression has been described as the mental mechanism that “guards” the conscious awareness from the intrusion of inadequate and intolerable ideas or drives. Repression keeps them unconscious. Freud indicated that the intruding ideas and drives from the unconscious actually threaten ego integrity.¹ Based on the formulation described so far, repression can be reconceptualized as the dynamics of participating, as well as nonparticipating, processes in the global formations that support conscious phenomena. Partial processes that do not gain access to the global process remain unconscious (i.e., repressed).

Because of the multiple constraints that characterize global organizations, certain partial processes may encounter difficulty in accessing the global formations. This is especially true if the partial processes carry information that is entirely removed from, or contradictory to, global messages. On the basis of these assumptions it is possible to conceive what type of information will be denied access to the global organization; it will be the contradictory and unfitting messages (i.e., partial patterns that do not satisfy the constraints of global patterns). In fact, Freud described the repressed contents as conflicting topics or unbearable ideas. Here, “unbearable” stands for information (of the partial process) that is removed from (i.e., unfitting with) the information presented by the global formation.

The unbearable partial process cannot be incorporated in the general message without damaging its internal consistency (i.e., its multiple constraint satisfaction organization) and therefore is bound to be excluded. For example, to a mother of a newborn baby, the idea of killing her baby is extremely contradictory to the regular loving and caring state of mind typical of a new mother. If inadequate partial processes somehow gain access to the global organization, they are inclined to destabilize or even disrupt it. If many conflicting and disrupting processes gain access to the global formation, the whole global message may be destroyed and the neural system representing it is bound to destabilize. Indeed, the types of thoughts that involve killing one's newborn baby often emerge in mentally disturbed patients. It is thus conceivable that in fact certain partial processes actually do threaten the integrity of global organizations and the actual stability of neural systems. This description conforms with Freud's notion of ego integrity being threatened by repressed mental processes of conflicting ideas or drives.

Occasionally, inadequate partial processes may gain access to the global organizations and be “transformed” in order to accommodate the global pattern. For example, immoral ideation is contradictory to the dominating content of a moralistic conscious awareness. Transforming the wish to behave in an immoral way into moralistic ideation may accommodate the dominating global organization of a puritanical message. This type of transformation is known in the psychoanalytic literature as reaction formation. Another transformation of unbearable ideation is known as isolation. Here the ideation is not excluded from awareness, but certain relevant parts of it are “neutralized.” These are the parts that are incompatible with the rest of the conscious message. The partial process is included in the conscious awareness only to the extent that it is removed from certain contents of the conscious awareness (i.e., isolated). If isolation is not enough to satisfy the constraints of global formations, then dissociation might occur, and certain contents of awareness will thus be ignored or experienced as independent and unrelated. The transformations described above are needed in order to protect the global formation from being disrupted by contradicting partial processes. Therefore it is conceivable that these transformations justify the term defense mechanism. They protect the global formations and prevent destabilization of the multiple constraint activity in the neural system. From the biological point of reference, this may translate into destabilization of the interrelations between groups of neurons, which presumably has direct neuropathological outcomes on transmitter-receptor activity.

The Dynamics of Internal Representations and Personality

The psychologist Carl Rogers³⁸ suggested that the best vantage point for understanding behavior is from an “internal frame of reference” of the individual himself. He called this frame of reference the experiential field, and it encompasses the private world of the individual. Neuroscience research demonstrates that the brain uses internal “maps” to represent information. One of the more famous examples is the homunculus of sensory and motor representations spread over the cortex.²⁸ Just as the homunculus is probably formed from the strengthening of synaptic pathways, the experiential field probably results from experience-dependent plasticity in the brain.⁹ In terms of space-state formulation, the experiential field can be conceptualized as a configuration of attractor systems in the brain.

According to Rogers, organismic evaluation is the mechanism by which a “map” (i.e., an internal configuration) of the experiential field assesses the psychological events of everyday life.³⁸ Evoking the formulation of state-space for internal representations, organismic evaluation can be reconceptualized as convergence into, or activation of, relevant experience-dependent attractor configurations of the internal map. If the incoming experience is identical to the previous internal representation of that experience, no change will occur and the map of internal representation will activate familiar past experiences. On the other hand, if the new experience is slightly different from the past experience, this will be enough to “reshape” the topological map and add attractor systems to the internal configuration.

Activation of the internal map organizes the incoming stimuli into a meaningful perception. The newly perceived experience is meaningful when it relates to the previous experience already embedded in this map. This is a circular process in which the map of internal representation is at the same time influencing and being influenced by the incoming stimuli. In other words, the brain sustains a map of internal representations that is continuously updated through interactions with the environment. Recently this type of interaction between internal representations and perception of environmental stimuli has been referred to as context-sensitive processes.³⁹

Because of this interaction, internal representations can be viewed as approximated models of reality. It is reasonable to assume that a “good match” between internal representations (of the psychosocial world) and external psychosocial situations will enable efficient adaptive interpersonal relationships. On the other hand, a “mismatch” between the psychosocial events of the real world and their internal representation may “deform” both the perception and the behavioral responses of the individual. The concept of matching complexity (C_M), mentioned above, further indicates that mismatch will be related to reduced neural complexity in the relevant neural systems and thus will be responsible for more adaptive problems on the neurocomputational level.

Internal representations of interactions with psychosocial reality are predominantly significant when evaluating personality traits or disorders. Personality traits are enduring patterns of perceiving, relating to, and thinking about the environment and oneself. They are exhibited in a wide range of social and personal contexts.⁴⁰ Thus, specific contexts of internal representations will have firsthand impact on personality traits. For example, internal representations regarding hygiene, punctuality, and precision are more pronounced for some individuals; for other individuals, other representations such as vanity and pride are prominent. The first example is typical of individuals who give special importance to order and strive to achieve perfectionism. These individuals are often referred to as having obsessive personality traits. The second example is more typical of individuals who regard themselves as special and important. They are often referred to as having narcissistic personality traits. Individuals who attribute importance to hygiene (i.e., who optimize these internal representations of context) will perceive a stimulus involving information about dirt and filth differently than individuals who do not have this type of attribute.

Once decoded, the map of internal representation can both explain and predict the reaction of the individual to certain stimuli. In the case of personality disorders, the optimization of particular internal representations of context may be enhanced to the extent where certain stimuli may be perceived with incredible distortion. For example, even a little dust on the table may be perceived as extreme filth by someone with an obsessive personality disorder. Even a small disapproval may be interpreted as an extreme insult by an individual with a narcissistic personality disorder.

The process of creating the specific maps of attractor configuration in different individuals depends heavily on the background of the individual. Individuals reared in families that give emphasis to sanitation will probably incorporate this emphasis through experience-dependent plasticity into internal representations of context (obsessive traits). Individuals reared in environments where they themselves were considered of prime importance and were the center of attention will probably incorporate these attitudes by optimizing the need to receive affection and attention (narcissistic traits). The developmental experience-dependent processes responsible for the formation of internal representations of context may involve deviations from the “normal itinerary” of internal representations needed for “regular” psychosocial function. These deviations may form internal representations that are greatly removed from psychosocial reality. A large mismatch between the internal representations and the environmental reality is likely to provoke distortions that lead to disturbances in perceiving and reacting to the environment (such as personality disorders).

To a certain extent, incoming information from environmental stimuli may be conceptualized as partial processes competing to gain access to global organizations of conscious awareness. A large mismatch between the internal map of representation and the pattern of environmental stimuli is likely to create the same difficulties that conflicting partial processes may encounter when trying to gain access to global organizations of conscious awareness. This mismatch may distort the incoming information in a manner similar to the way unfitting partial processes that attempt to access the global workspace are distorted; they have to be transformed before they can participate in the dominant message of conscious awareness.

A good example of this distortion is seen in the phenomenon of transference. Transference is regarded as an attitude toward an event or individual that is based on previous experience with similar events or people that is not congruent with the current situation.⁴¹ Thus, the incoming stimuli from the psychosocial event are distorted to “fit” the internal representation of similar events already dominating the global processes in conscious awareness. Since incoming information is “evaluated” by the internal representations, and since these are formed by experience, it is only natural that many of the perceptions we have are related to past experience. When a set of stimuli of a new psychosocial event enters the system and causes it to converge to a set of attractors that represents similar past experience, that set of attractors activates the past experience in the global organization, bringing it to conscious level. The conscious awareness regarding the individual or event that provoked this process will be perceived in many connotations as being the past experience. If there is a substantial mismatch between the internal representations and the actual psychological event, the transference (i.e., the perception as past experience) may distort the perception of that psychological event. Matching complexity may be the future mathematical tool that will predict to what extent transference is likely to determine one's behavior.

Sometimes the set of environmental stimuli is so removed from any context of internal representation that it is totally unperceived by the individual. This is defined in psychodynamic terms as denial. An individual with narcissistic personality traits may not perceive signs suggesting that he is not desired. This is because in his map of internal representations there is no context or attractor system (basins) that can portray rejection. Furthermore, in this individual the part of the space-state topology involved in representing rejection probably lies on a peak, repelling any possible perception of this stimulus. Since the representation of rejection will not be activated at all, it will not manifest in the global organization and will remain entirely out of any conscious awareness (denial).

SYSTEM APPROACH TO PSYCHOTHERAPY

Psychotherapy: A Physical, Experience-Dependent Process in the Brain

The system approach to psychodynamic concepts offers new insights into possible physical processes taking place in the brain during psychotherapy. Roughly speaking, individuals seek psychotherapeutic treatment out of distress that originates from interpersonal relationships in psychosocial situations. In the process of psychotherapy, changes occur in the client that enable him or her to adapt to, act upon, and perceive these situations without accompanying discomfort.

The interpersonal relationship between therapist and client is the tool for creating the needed change. Initially the relations with the therapist will repeat the same patterns of interpersonal relations that caused the distress. The skilled therapist identifies these patterns and reacts in a way that gradually changes the attitudes of the client toward similar future situations. Successively, this change continues both in and outside of the therapeutic setting. Better coping in psychosocial situations reduces the former suffering and enables the relief from symptoms. This relief is the outcome of treatment. Psychotherapeutic procedures have been described as overcoming resistance, offering appropriate interpretations, and increasing insight into relevant aspects of interpersonal relations.^1,41

In a system approach, the psychotherapeutic process can be described as a physical change that takes place in the brain of the client. Initially, the relationships between the internal map of reference of the individual (internal representations) and some aspects of the psychosocial situations he or she encounters are incongruous (a mismatch). This incompatibility reaches the extent where perception and reaction to those psychosocial situations are distorted and interfere with the psychosocial functioning of the individual. The psychosocial dysfunction is generally accompanied by distress, which is typically expressed as symptoms of anxiety and depression. The goal of the therapy is to reshape the internal representations to include the appropriate representations to cope with the psychosocial situations at hand.

At the initial phases of psychotherapy, the therapist is perceived by the client as being more like one representation of similar people encountered in the past (transference). This is because the client activates the attractor systems, described above, that represent these people. Since the therapist is not the same as the activated representation, a distorted perception of the therapist emerges. Because of this distortion, an inappropriate behavioral reaction to the therapist (transference) occurs. Most probably this distortion is there for other interpersonal situations outside the therapeutic sessions as well. This indicates that there is substantial mismatch between internal representation and psychosocial reality. The therapist has to enlarge the repertoire of representations of the individual to match many more different psychosocial situations. In other words, the psychotherapeutic process increases the neural complexity (C_N) in the brain of the client. An enlarged repertoire of internal representations will enable a better match between internal representation and psychosocial reality.

Increasing the repertoire of internal representations involves creating additional and more complex attractor systems to match the events of the psychosocial settings in reality (i.e., increasing matching complexity C_M). When the therapist reacts to the client in new ways that were never perceived by the client before, Hebbian mechanisms of plasticity will gradually create the new attractor systems needed for the additional internal representations. In this manner, the therapist shapes the space-state topology of the brain to form new internal representations. This involves actual changes in the functional connectivity of the neural systems involved, and thus it is a physical process in the brain.

The process is actually much more complex than the above description suggests. For example, due to a lack of representational systems, many times the interpretations offered by the therapist are denied and do not gain access to the global formations. These interpretations will never reach conscious levels. This is resistance in psychoanalytic terminology. In the system approach, the set of inputs from the interpretation simply conflicts with (i.e., does not satisfy the constraints of) the global formations concerning the internal representations. It has been wisely indicated that for an interpretation to succeed it must be delivered at the right time, when the individual is ready for it.⁴¹ There must be a certain constellation of the global organization that is favorable to including the new patterns of information proposed by the interpretation. The therapist first prepares the patient by repeated clarifications, confrontations, and other interpretations. This process changes the global formation, “moving” it slightly toward the pattern that will be favorable to accepting the critical interpretation (the one that will induce the change).

Freud indicated the importance of overcoming resistance in psychotherapy.¹ By gradually changing the global formation to a favorable pattern that will enable an interpretation to be incorporated, the therapist overcomes the resistance to that interpretation. Repeating this process over and over again will eventually reshape the state-space of the brain to increase the complexity of internal representations and thus the psychological repertoire of the individual. These changes occur, and are maintained, by means of the experience-dependent plastic processes of the brain. It is probably the increase in neural complexity that enables adaptability and reduction of suffering as the outcome of psychotherapy.

DISCUSSION

The Problem of Complexity Assessment in the Brain

The most challenging problems that stand before researchers who want to apply the system approach to psychotherapy are the problems of sampling, identifying, and decoding the multiple constraint organization. Sampling typically uses electrophysiological assessments such as electroencephalography, magnetoencephalography, or, more recently, the magnetic resonance signals that make up functional magnetic resonance imaging. The source of the biological signal and its sampling techniques pose many problems that are beyond the scope of this text.⁴² Regardless of the source of the signal, its interpretation regarding neural organization is a tremendous problem.⁴³

One measurement widely used in signal analysis is that of coherence in terms of correlation coefficients. This measurement proposes that regions in the brain influence one another if the signals emitted from these regions have the same waveform (i.e., frequency and amplitude).^44,45 If two regions have the same waveform in their signals but the signals are shifted in time (an expected situation, considering time delays in activity transfer), the correlation coefficient will be reduced and the time shifts must be taken into consideration. Both time shifts and the consideration of the three-dimensional special conductance in the brain medium greatly complicate coherency calculations. Principal component analysis can give a rough estimation of the dimensionality of a biological signal.⁴⁶ Dimensionality can be interpreted in terms of how many components contribute to the signal. Increased dimensionality indicates that more sources are contributing to the signal. Independent component analysis, a relatively new method, is designed to identify how many components contribute to the signal. For example, in schizophrenia, if one expects a disconnection syndrome,⁹ then one would expect an increase in the number of independent components in the signals from schizophrenic patients compared with those of normal control subjects.⁹ Most recently, mutual information measurements used for assessing neural complexity, C_N, are beginning to be applied to brain signals.^34,47 They hold the promise of detecting the degree of neural integration (described above) in the brain.

The degree of cortical integration in the brain will probably provide some insight into the degree to which neural systems are multiply constrained by each other via functional connections. However, this is only a rough indication of the actual pattern of constraints occurring in the brain. Future research will have to rely on more sophisticated combinations of signal analysis with mathematical tools in order to decode the precise patterns of interrelations among neural systems in the brain. Until such tools are devised, we are left with general, simplistic notions of multiple constraint satisfaction and optimization.

Questions Made Possible by the System Approach to Psychotherapy

Although acting only as a theoretical paradigm, or at best as a reconciling language between psychodynamic concepts and brain concepts, the formulations presented here permit asking questions that are not possible otherwise. For example, how do experience-dependent plastic processes alter the brain in relation to psychotherapy? Is there a relationship between neural complexity (assessed by electrophysiological or other signal analysis of the brain) and progress in psychotherapy (measured by symptom relief)? Can signal analysis (using measurements of neural complexity) help monitor progress and assess the outcome of psychotherapy? What are the interactions between medication effects and psychotherapy at the brain level? It is sometimes suggested in clinical settings that medications might interfere with psychotherapy. If medications curtail experience-dependent changes at the synaptic level, such an assumption may be plausible. On the other hand, substances that facilitate neural plasticity might prove to boost psychotherapeutic consequences. Conversely, medications that facilitate neural plasticity might be useful only in conjunction with psychotherapy. In this sense, these medications would “become activated” only in conjunction with psychotherapy, and thus psychotherapy could be viewed as an activator of neural plasticity–facilitating drugs. Results of studies designed on the basis of these questions will have substantial impact on the future of psychoanalysis as a science of the brain and on psychotherapy as a therapy of brain-related mental disorders.