The Spandrel of Virtue: Radical Behaviorism and the Science of Optimism

Marr, A. J. (2001) Why behaviorism, to survive and triumph, must abandon its very name. An Open Letter Behavior and Social Issues, 11(1), 92-99

A credo of humanistic psychology, as stated by the JHP, is that humanistic psychologists encompass a "lifelong learning community that values authenticity, choice, self-determination, self and social responsibility, empathy, mutual respect, and the integration of mind, body and spirit".

But the question arises, why must this be so, and should it be representative of the core values of psychology as a science? Could this be because of a confluence of genetic tendencies selected fortuitously by evolution? Could it be because of an inferred computational mechanics that reconciles such sentiments with the inputs of experience? Is it mere prejudice, or at worst, wishful thinking? Or could it be because it arises like the epiphenomenon of consciousness from the elementary workings of the human mind? Can epigenesis, or the unpredicted side effects of the concerted workings of genetically determined structures, provide us not only with consciousness, but with art, literature, empathy, and the unexplainable and 'transcendent' qualities of human nature?

The answer may be found in the neuro-biological bases of incentive motivation, the mechanisms that provide learning and the impetus to decide what to learn. In the last twenty years, bio-behavioral models of learning that integrate molar (overt behavior) and molecular (neural behavior) behavior have been constructed that integrate embodied or affective experience with the heretofore disembodied computational notions of behavioristic and cognitive science. This 'affective neuroscience' (Panksepp, 1998) or 2^nd generation cognitive science (Lakoff and Johnson, 1999) adds back into the equation of learning the non-computational conscious and nonconscious pleasures and pains that are the ultimate measure of our experience. It by its very nature reconciles the subjective with the objective, the metaphorical with the literal, and the behavioral with the neural. It is, in other words, a holistic or integrated psychology that is nonetheless firmly rooted to empirical or inductive principles. But what metaphors may we entertain that can simply explain this momentous shift in psychological thought that embodies all our intentions?

The description of incentive motivation or intentionality as derived from the traditions of behavioristic and cognitive science was elegantly phrased by the philosopher Daniel Dennett (1996) using the metaphor of a learning machine. If, as the metaphor holds, we are information processors, then the information we process and attain must be first selected. It its most rudimentary form, information is selected from a simple process of trial and error. The behavior of a Skinnerian machine is selected or shaped through the reinforcement of successive approximations of behavior. The individual learns successful behaviors through an experience in real terms with the natural contingencies of reinforcement. Thus a biological organism, as an information processor, learns through actual manipulation of the environment and the positive consequences or reinforcement that follows behavior.

In contrast to the Skinnerian machine, the Popperian machine (named after the British philosopher Karl Popper) learns successful behaviors through an experience in virtual terms with the natural contingencies of reinforcement. In a Skinnerian sense, a mouse learns to avoid a mousetrap because of its vicarious experience with mousetraps. In a Popperian sense, a mouse avoids a mousetrap altogether because it can model in its mind mousetraps.

We are essentially Skinnerian and Popperian creatures. We learn from experience and from experiences that we model in our minds. The behaviors we select in reality and in virtual reality are thus chosen because of associational or computational principles. In Skinnerian and Popperian machines, it is inferred that approach behaviors whether real or virtual prepare for real culminations that ultimately provide reinforcement. In other words, the real or virtual doing is not where the reinforcers are, but in the actual accomplishment of the goal. Like walking down a path to a goal, whether the journey occurs virtually or in reality, reinforcement only occurs when the goal actually occurs. However, as humanistic and social psychologists are wont to point out, reinforcement is not that simple. Indeed, we take pleasure and pain, or are intrinsically reinforced, by events both virtual and real that are far removed from the behavior's substantive products, or 'reinforcers'. Thus we take pleasure in accomplishment, shame in our misdeeds, pride in our private virtue, and all independently of the ultimate results of acts both imaginary and real. By its very nature, reinforcement 'teaches' us which way to go, but if reinforcement is something more than the computational mechanics culminating in the acquisition of an object, then something else and more elementary must be added to the equation of learning.

Learning involves teaching, or the feedback that enables our brains to select or modulate the images that are important for problem solving, and in the large, survival. With Skinnerian and Popperian machines, it is commonly inferred that an organism is 'taught' from informative or discriminative feedback from the action itself. In other words, the activity of the computational organelle of our brain, or the neocortex suffices entire for learning. But the expansive neocortex of Homo sapiens is a literal late bloomer, and the product of only the last fifty million or so years of evolution. For our mammalian cousins as well as ancestors, the lack of a substantive neocortex referred choice to more primitive neural systems and the simple decision or teaching signals that guided the choices those systems mediated. This secondary teaching signal does not however reflect computational processes, but a non-computational hedonic sensitivity to abstract qualities of information.

To understand this elementary or ‘affect’ logic, it is important to understand the elementary decision rules that our ancestors had to make to secure survival. It all had to do with surprise. In the ancient environments our ancestors had to face, survival depended upon their ability to respond to the unpredicted changes in its environment. The smell of a new source of food, the sound of a new predator, the sight of a receptive female, or even the exploration of a new territory that presaged such events would require precedence over other aspects of the environment that were more predictable. Hence, our ancestors would be sensitized toward unexpected changes in their environment that was represented neurally by the creation of neuro-modulators (neurochemicals that modulate global areas of the brain) that fix attention, improve synaptic or thinking efficiency, and had hedonic value.

From the perspective of moment-to-moment or molecular behavior, this sensitivity may be termed Pavlovian incentive salience (Berridge, 2001) or behavioral discrepancy (Donahoe and Palmer, 1993). In turn, the same behavior over larger or molar time scales may be termed a seeking response, or a universal foraging instinct (Panksepp, 1998). Discrepancy theories of reward (Hollerman and Schultz, 1998), represent a second teaching signal or source of reinforcement that is different psychologically and physiologically from neo-cortically situated computational or associational processes. Moreover, this secondary teaching signal may cohere or be positively or negatively incoherent with goal states as rationally conceived.

For example, consider a Skinnerian fixed-ratio or piecework schedule of reinforcement in a factory environment. A worker may have to repeatedly pull a lever in a button-making machine a fixed number of times in order to receive a weekly paycheck. But if payment in some varying size occurs after each pull rather than on a set weekly basis, payments would therefore come in surprising or unpredicted regularity. By morphing from a button machine into a slot machine, the worker would be enthused, excited, and likely unmindful as to whether his average weekly winnings would ever match his former paycheck. The reinforcement value of the continuous discrepancy would thus be negatively incoherent with his rational appraisal of behavior that would otherwise maximize reward. In other words the affective value of gambling would be incoherent with the rational value of a predictable routine at work.

As another example, consider a playwright with a rather Popperian task of estimating the fruits of an otherwise simple commission. If the playwright was commissioned to write a play filled with sex and violence that meets the undiscerning needs of a popular audience, he may look past this relatively easy and predictable task to the virtual implications of the judgment of posterity, securing a girlfriend’s favor, surpassing a competitors talent, or impressing the Queen. These additional uncertain interdependencies are stimulating and exciting, but not ultimately necessary to pay the bills. However, they may be enough to raise an original idea for a banal and popular play about Romeo and Ethyl, the Pirate’s Daughter to an inspired masterpiece called Romeo and Juliet that meets all demands, both real and virtual. And as the movie ‘Shakespeare in Love’ so demonstrated, this was indeed the case.

In both examples, the uncertainty of virtual outcomes ‘energize’ behavior, but deflects its aims from what is logically required. Sometimes this is beneficial (as with our playwright), and sometimes its not (as with our gambler). But the illogic of discrepancy is also ubiquitous, and is a near constant aspect of our behavior. Thus our behavior is constantly influenced by daily distractions (e.g. checking email, idle conversations, daydreaming) that are valuable because of their affective rather than logical value. Similarly, when discrepancy is continuous and positive, as when we perceive a moment to moment string of positive uncertainties due to a near matching of demand and skill, the corresponding affective states may be continuous, and reported as ‘flow’ states (Csikszentmihalyi, 1990) that may nonetheless bestow value that is also incoherent with logical ends. Thus the successive uncertainties of a rock climber perched precariously on a ledge, a creative artist grasping for inspiration, or a football player driving his team down the field for a last second score achieve a greater affective value that overshadow the fact that these situations are not logically preferable to surer and safer alternatives.

As Popperian machines, the ability to plan ahead as bestowed by a substantive neo-cortex allows us to be taught by hypothetical events that are cognitively separable from real events, but more primitive teaching signals from deeper neural structures cannot make this distinction. Thus we can be as rewarded by constructing hypothetical castles in the sky as if they were in fact real. The fact that the discrepancies entailed by hypothetical action plans virtually reinforce necessarily causes behavior to diverge from its logical reinforcers. This divergence or incoherence is ultimately the linguistic root of all evil, and of all good. Thus, if our behavior is positively incoherent, we call such behavior virtuous, and if it is negatively incoherent, it is vice. And so a gambler has an addiction, a disease, or a moral defect, whereas a triumphant playwright has inspiration, a muse, or the spark of genius.

If it is assumed that the purpose of life, or ‘happiness’ is to maximize value, or reinforcers, the means-ends values that ensure survival must be combined with the discrepant values that allow us to plan for the surprising exigencies of survival. Thus by modeling discrepancy, we maximize reinforcement, and as a matter of course maximize the empathy that enables us to model the minds of other people. But doing that will alter behavior from its logical course, and propel us to acts both hideous and sublime. Thus to fully and virtually prepare for the contingencies of existence is to betray them or transcend them!

It is to the survival interest of humanity that our ‘affect’ logic and the means-ends logic of our worlds cohere, and multiply each other in a beautiful and infinite synergy. Hence the design and purpose of any culture is not to maximize economic value alone, but the boundless discrepant value that comes from the stimulating value of the thought of not just mechanical universes but of the empathic modeling of the minds of men and women.

Humanism is at root an epigenetic system of values that on one level transcends evolution but on another level perfectly coheres with a Darwinian universe. On one level of thinking, we may say like a Dostoyevskian anti-hero that two plus two equals five, but on a lower or neural level, the mechanics of his thinking is as determined as the orbit of a planet about a star. Thus on one level behavior is indeterminate, and on another level it is not. Thus freedom and mechanism is dependent upon the perspective you take, but neither one is any less ‘real’. The great irony of human existence is that the ‘logical’ aims of our genes entail ‘illogical’ behavior. On one level (of mechanism) it makes Darwinian sense, but on a higher emergent level (of embodied consciousness) it transcends Darwin. So humanistic instincts are right because we must as a species know the value of virtual things, of the beauty and the pleasure of desire. And because we also must know that human nature, to ensure its survival, must drive us slightly, remarkably, and transcendentally mad.

Consciousness and emotion are ancient topics as old as culture, still in their scientific infancy, and both slowly emerging into full respectability after decades of systematic neglect by science. Despite a modest resurgence in interest in the subject, emotion remains perhaps the least understood subject relative to its importance in human life in the whole of neuroscience. This is probably overdetermined. It may be in part a hangover from Lange-James perspectives in which the richness of experienced emotion was reduced almost to an epiphenomenon, a sensory feedback from autonomic and other efferents. This made emotion a compelling but ultimately almost irrelevant sensory after-image of the "real action" of autonomic and motor efferentation. Additionally, the explosion of cognitive neuroscience, in concert with the extensive discrediting of much of psychoanalytic thinking, has left emotion in a largely secondary role, despite dramatic lessening of the stranglehold that behaviorism had over thinking in psychology. Cognition is very much in ascendance these days, including in consciousness circles, with some even assuming its foundations are fundamentally independent from affect, a position for which there is little evolutionary or neurological evidence.

Additionally, the relative disregard for emotion (until recently) in neuroscience may have major contributions coming directly from the intrinsic scientific and methodological difficulty of the subject itself. To borrow from Chalmers (1996), affect is no easy problem. It is elusively multi-dimensional, a complex composite of patterned autonomic, endocrine, motor-executive, pain/pleasure (valence), social/signaling, and cognitive (other/self appraisal) activations. Additionally, there are formidable terminological and nosological issues. Emotion can be defined quite broadly (as emotional meaning, or emotional learning, which is vast and virtually interpenetrant with almost every higher activity in the CNS) or narrowly (the prototype emotional states of fear, rage, sadness, lust, etc.) Adding further to the complexity, emotional processes can be conscious (in which case one has "feelings") or unconscious (i.e., unconscious valence assignments or unconscious affective behaviors like avoidance). This has further divided the focus within the emotions research community: should we focus on feelings, or are they just a scientific distraction, while the real "action" of emotion is largely unconscious. And in terms of neural substrates, emotion broadly defined in humans is spread out through many neocortical, paleocortical, subcortical, diencephalic, midbrain and pontine-medullary brainstem systems, eluding neat localization in any "limbic system" unless such a system is exceedingly diffuse. This has led many reviewers to suggest that the concept of a limbic system no longer is meaningful at all.

Finally, although emotion’s complexity in humans reflects "re-entry" between many subcortical and cortical systems, emotional prototypes or "primitives" (emotion narrowly defined) are organized in diencephalic and midbrain structures that have been largely ignored in most work on emotion, except by Panksepp, Damasio, and a few others before them such as Nauta. Much work instead has focused on the cognitive processing of emotion, its cortico-cognitive elaborations, and the complex "valence tagging" systems (studied largely so far in the amygdala and not as much in other basal forebrain systems involved in affect). These mostly basal forebrain systems (best studied in the amygdala) allow correlations (via various LTP mechanisms) between activations encoding primitive biological values and higher cognitive encodings, forming the basis for adaptive emotional learning. But the more primitive base of biological and social value (in diencephalic – midbrain – brainstem networks?) gets relatively short shrift, as these regions tend to be seen as mostly passive outputs targets for higher systems, like amygdala, and not particularly integrative themselves.

The empirical evidence actually favors the more complex view that the amygdala is quite dependent on those lower diencephalic and midbrain systems to instantiate and organize a coherent emotional response. Conversely, the lower systems by themselves (without the connectivities to basal forebrain and thalamocortical systems) cannot develop the crucial correlations with higher encodings essential for the emotional meanings (our emotional associations to complex stimuli) that inform adaptive functioning. Lesion evidence (summarized by Panksepp, 1998) suggests that disruption of the connections from amygdala to the hypothalamic-midbrain systems affects emotional learning, and top-down adaptive activations of emotion (largely for fear and rage but not so much for the positive affects). But disruption of these connections from amygdala to lower systems does not fundamentally destroy the potential for an emotional response, while destruction of those same hypothalamic and midbrain systems or their outputs devastates any potential for an emotional response. In other words, the primitive core of emotion may be deeply ventral and subcortical, while the amygdala and the thalamocortical systems are necessary to determine when it is "appropriate" to generate such a response. Thus the amygdala is best conceptualized as a high level correlator system central to affective learning, particularly for negative emotional meanings, linking primitive defensive responses with higher cognitive encodings. But this is still somewhat different from the dominant "party line" in the neuroscience literature: that emotion (conscious or unconscious) may just require cortex and amygdala, and a few "brainstem systems" (meaning many structures in hypothalamus, midbrain, medulla and pons) that are "passively activated" by these higher systems.

Given all the "poles" of these several debates (the wide vs. narrow definitions of emotion, the focus on conscious vs. unconscious aspects of emotion, and the focus on telencephalic (paralimbic and subcortical) vs. more primitive (diencephalic-midbrain-brainstem) systems, it is safe to say that the whole field of emotion studies is still fragmented, with little theoretical coherence. Overall though, in both the "harder" neuroscience investigations (and in softer cognitive science approaches as well) most emotion research has been largely focused on the "top" of the processing hierarchy (and analogously the cognitive literature has focused largely on appraisal). There has been much less appreciation of the whole, scientifically formidable and intimidating, neural integration of primitive and higher systems. I would argue that this fundamental integration of higher cognition with basic social-biological value is the core scientific challenge for mapping emotion in humans. Such an approach implies the need for more distributed hierarchical models, an approach recently taken by Lane (1998), and Damasio (1998).

From the other side of the fence, within the burgeoning literature about the neural basis of consciousness, affect has been largely relegated to the back of the bus as an interesting "coloration" to the "hard problem" of consciousness, which is often assumed to have a more cognitive-sensory foundation. Most current theories of consciousness are deeply cogno-centric, and neglect evidence that emotion is a central organizing process for consciousness, even though there is increasing acknowledgement that cognition and emotion are not "oppositional" (Damasio, 1994) and that affective meaning might even drive and vitally inform most cognitive activity. There has been a quiet tendency within consciousness studies to view emotion as just an interesting type of "qualia" among many other types of qualia. LeDoux (1996), the best known researcher in affective neuroscience who has done much elegant work on the amygdala, has also advanced this very same conceptualization. He suggests that the main point of intersection of affective neuroscience and a science of consciousness is only the limited domain in which "emotion enters experience through representation in consciousness mechanisms." This is seen as a small part of the big picture of emotional processing, much of which LeDoux sees (accurately) as going on unconsciously. Thus the point of intersection of consciousness studies and emotion studies is "reduced" to the problem of the neural substrates of conscious vs. unconscious emotional processes. Not only is affect seen as just another qualia from both sides of the fence (the emotion literature and the consciousness literature), but a disadvantaged poor sister qualia at that, competing with better mapped visual awareness, which several adherents (notably Koch and Crick, 1995) offer as a best available neural network model for consciousness itself. Thus, overall, little mention is made in "affective neuroscience" or in consciousness theory of any role that emotion might have in organizing or underpinning consciousness, as opposed to their being two basically "orthogonal" processes Work by Panksepp and Damasio provide virtually the only major exceptions to this (along with some theoretical overview work of my own).

If emotion and consciousness are not just orthogonal, how might one conceptualize their relationship? Conceptually, the generally segmented approaches in neuroscience to emotion, pain, attentional functions, and executive functions (volition) neglect that these are all "global state functions" speaking to different global integrative processes in the brain that are all probably foundational for consciousness in different ways. Put differently, consciousness may be the largest "umbrella concept" subsuming all the global state functions (selfhood, pain, attention, executive function, and emotion probably constituting the "big five.") These functions address a basic integrative and multi-component neural envelope for consciousness. I recently argued (Watt, 1998) that these putatively separate functions might be better conceptualized as just different "slices" of the consciousness pie, different aspects of the global integration that brains achieve intrinsically, but that neuroscience struggles to understand with much more effort. Deep interpenetrations of all global state functions (both functionally and in terms of their highly distributed neural architectures) are little appreciated in neuroscience. For example, both working memory and executive function are critically dependent upon an embedding of value within WM frames and behavioral goals. Without emotion generating and informing a central representation of value, executive and attentional functions are collapsed at their base, as are personal meaning and any viable image of agentic active self (e.g., akinetic mutism and confusional states in bilateral cingulate lesions). In marked contrast to these basic suppositions, much of current research and theory in neuroscience sees the brain in terms of fine grained modularity, and promotes a vision of the brain as a discrete set of processing modules. What holds these modular aspects together, what allows a serial train of consciousness to emerge from the workings of all these parallel processors, as Bernie Baars (1993) rightly asks us to focus on, is relatively neglected. However, and most critical to emphasize, modularity and mapping global state functions must be seen not as opposed points of view, but as two sides of a theoretical coin. But without a (generally neglected) understanding of these integrative global state architectures underwriting global state functions, modularity cannot explain anything like a serial and coherent stream of consciousness in which not only are the "lights on" but "somebody is home." Without integrative architectures we cannot explain how a highly distributed and very complex neural system generates a person who feels continuous with their experience and "qualia." This is Chalmers’ "hard problem," and clearly the core of the old mind-body dualist dilemma. Only these interdependent global state functions can provide the basic bridge concepts here for a neuroscience of consciousness, illuminating:

1. how a primitive yet superordinate and coherent neural image of the self is instantiated, such that qualia are "owned" by an agentic self in normal consciousness;
2. how executive processes arise contiguous with this primitive self-representation such that volition, our sense of being an active agent continuous with ourselves, becomes a basic property of consciousness;
3. what supports an "attentional matrix" enabling selection of stimuli for further processing so that we experience the smooth streaming of working memory and other conscious content consistent with that "agency";
4. what neural foundations for organismic value in pain and pleasure (primitive states of qualia) inform and drive these basic executive and attentional mechanisms, as biologic value cannot be separated from attention or volition without truly "gutting" these functions;
5. how basic emotional processes build on primitive foundations in pain and pleasure to further the complex neural representation of value and how do such processes support prototype feeling states.

It is essential to frame this larger theoretical context to understand the importance of Jaak Panksepp’s summary work, Affective Neuroscience, a compilation of much of his previous publication and research. In this larger context, Panksepp’s Affective Neuroscience is a vital work for those interested in the development of a global mind-brain theory (obviously a primary concern in consciousness studies), and especially for those interested in the uncertain place of emotion in that still developing global theory of mind-brain. Panksepp has been for almost thirty years one of the leading investigators of the diencephalic and midbrain roots of primary or prototypic affect. Panksepp’s theoretical work (work scattered through many journals over 25+ years) is as original and significant a contribution to emotion literature as there has been in the past thirty years by any researcher, but its wide spread has been poorly appreciated, and his potentially most salient theoretical contributions have been scattered throughout many journals, and as a whole not widely visible or appreciated. Instead having to hunt through all these sources, Affective Neuroscience performs the service to readers by summarizing Panksepp’s work on differential and specific structural and chemical architectures for each of the key emotional prototypes or emotional primitives. However, it may be those primitive foundations of emotion in the relatively neglected diencephalon and midbrain that shed the potentially brightest light on emotion and its possible intrinsic ties to consciousness. (For extended discussions on this subject involving a number of the leading figures in affective neuroscience and consciousness theory, see Watt, (1998a), and the associated commentaries in the ASSC Electronic Seminar on Emotion and Consciousness at http://server.phil.vt.edu/assc/esem.html.)

Affective Neuroscience argues for the primacy of emotional systems and affect generation in mental evolution in a comprehensive textbook form. This textbook is the first attempt at a genuinely comprehensive coverage of the basic neural systems that generate emotionality in the mammalian brain, an area that has not received the attention it deserves from either psychologists or neuroscientists. Panksepp is hoping that there will eventually be more formally organized academic courses and graduate programs in the area of affective neuroscience, but to my knowledge, they presently do not exist. Indeed, some (notably LeDoux) question whether a designation of such a topical area of "affective neuroscience" only further fragments and separates cognition and affect, when almost all agree that we are looking for deeper integration. Whether the designation of "affective neuroscience" further fragments cognition and affect is open to debate, but any neat division between the two is clearly not a concept Panksepp would support at all. In any case, Panksepp’s textbook summarizes the emerging area of affective neuroscience in three convenient modules:

1. Part I, contains six introductory chapters providing essential theoretical and conceptual materials as well as excellent integrative chapters on the relevant neuroanatomical, neurophysiological and neurochemical issues.
2. Part II consists of 5 chapters covering vigilance states, regulatory states, as well as exploratory/anticipatory, aggression and fear systems.
3. Part III consists of 5 chapters on social emotions, starting with sexuality, followed by maternal behavior, separation distress, playfulness and a concluding chapter on self and consciousness. Three appendices provide essays on human evolutionary origins, language, and dualism in the brain sciences.

The writing style is quite richly metaphoric at times, and this is one textbook that does not suffer from dryness-to-the-point-of-desiccation that typically makes reading neuroscience such a deadly process for most students and all but the most passionate and most interested. As a broad-based textbook on emotion and neuroscience, the work earns high marks. But hidden in the textbook format are some seminal insights and several original contributions. Panksepp’s own empirical work on the neural underpinnings of affect is seminal on several accounts:

1. its emphasis on the real roots of emotion in primitive "value generators" in basal midbrain-diencephalic regions, in contrast to the reigning (and reining?) assumptions that emotion is mostly a telencephalic construction (subcortical gray structures in the hemispheres, and paleocortical paralimbic regions). There is much empirical material illuminating this basic confusion between affect qua primitive affect, vs. the affective meanings and "valence taggings" that require higher telencephalic systems;
2. the "unpacking" of "valence" (a confusing and poorly sorted-out basic property of emotion) into a core group of emotional primes or prototypes, revealing the "modularity" and evolutionary specificity of basic emotion, with associated chemoarchitectures for each prototype;
3. outlining that all these prototype emotional systems (rage, fear, separation distress/attachment/ bonding, sexuality, play) project to a midbrain-diencephalic structure known as "periaquaductal gray," thus generating a primitive global value mapping in PAG. PAG may represent the ventral core of the "limbic system," a structure almost totally neglected in most accounts of the neural substrates of emotion. (Although several theorists have suggested that this concept is virtually meaningless due to the system’s lack of real integration, PAG may supply crucial integration of emotional processing but at a deep ventral position in the brain, and PAG’s essential role in emotion is more neglected largely by the very theorists that claim that the limbic system concept is dead);
4. that these various midbrain-diencephalic systems projecting to PAG are much more highly reactive to modulation by neuropeptides than by monoamines. Neuropeptides have much more behavioral specificity than amines, with PAG having significant efferents to all of the core monoamine "spritzers" in the brainstem, influencing global state in part by tuning them;
5. in combination with other recent work, Panksepp’s work helps to outline direct pathways for the neural interdigitating of this midbrain-diencephalic architecture for prototype emotion with the ERTAS architecture (extended reticular thalamic activating system) thought to provide the key substrates for consciousness, facilitating a global workspace of basic neural systems and connectivities (Baars and Newman, 1993; Taylor, 1998; Newman, 1997). These pathways include particularly connections to ILN, MRF and other non-specific thalamic and reticular systems from PAG, as well as much limbic and paralimbic modulation of nRt and the other non-specific thalamic systems.

These prototype emotional systems (what Panksepp calls the "grade A blue ribbon emotions") deservedly get the lion’s share of attention in Affective Neuroscience; they are same ones we appear to share strongly with all mammals. These are (in our crude and still evolving typology): fear, rage, separation distress (grief/sadness forerunners), play (affection), lust, nurturance, and perhaps a few others that are not charted much at all (a system for social dominance?), and an important and often misunderstood system for non-specific motivational-affective arousal that Panksepp calls a "seeking" system. This is the dopaminergic VTA-mesolimbic-mesocortical system which has been generally misrepresented as the brain’s "reward system" (it is only one of several "reward" and "punishment" systems). Additionally, this complex set of architectures extends well beyond what are probably phylogenetically earlier systems for fear and rage centered in distributed networks with the amygdala at the dorsal end of the subcortical processing chain. The more appetitive-attachment systems include other crucial systems in the basal forebrain, particularly the bed nucleus of stria terminalis and ventral septum. This makes for a much broader picture of emotionality than what is typically accessible in most neuroscience texts, underlining our deeply social neurobiology: play (affection), lust, attachment, and separation distress get extended attention by Panksepp. Many of these prototypes (particularly the bonding/separation distress, play, and non-specific seeking) are subjects to which he has made major empirical contributions. Separation distress has been perhaps as deeply investigated by Panksepp as fear has been by LeDoux. Separation distress is inherent in the state space of mammalian attachment systems in which the object of attachment appears lost. (Distinctions between fear and separation distress are consistently obscured in both the clinical and experimental neuroscience literature.) Panksepp’s wide angle focus on the full range of defensive and social emotional systems is a refreshing tonic for the fear-centric focus in most emotion research. There is still some tendency to assume that if we understand fear (as one crucially important prototype defensive state) we can understand emotion broadly defined, an assumption that Joe LeDoux himself probably would not want promoted.

Although initially the relationship of a non-specific VTA-LH seeking system to primary or prototypical affect may seem unclear, this "seeking system" probably underpins a most basic emotional capacity without which others make little or no sense – the capacity to experience hope. We are driven to go out into the world and interact by this VTA - lateral hypothalamic system, searching for food, drink, and other rewards and pleasures. But this system is also just as essential for searching for more abstract treasures such as purpose and meaning that are essential to humans with their more advanced cortical evolution, and this VTA system has a huge role in prefrontal system synaptogenesis, myelination and development (see Schore, 1994). Re-entrant connectivities between VTA – LH and higher cortical systems are essential for normal curiosity and cognitive explorations (see Schore, 1994 for an encyclopedic review). Thus, in effect, Nature says, "for God’s sake, at least do something !#&!" This "seeking" happens despite most of us having a clear sense that we certainly could get clobbered out there in the world, putting our curious noses into various things to check them out (quite literally in the case of animals with heavy dependence on olfaction). This largely-DA and neuropeptide modulated VTA – lateral hypothalamic system with widespread limbic and cortical projections appears to be a basic underpinning for the most basic and primitive aspects of motivation. It is, however, quite motivationally and goal non-specific, facilitating only the relative activation of other potentially affectively rich interactions with others and the environment. Thus the VTA-LH system may "broker" activations of the other primes or prototypes, and relies directly on hypothalamic mediation of homeostasis in the seeking of food, drink, and other biological requirements. The following table summarizes a basic outline for a preliminary chemoarchitecture for emotion, extracted from Panksepp (1998).

Distributed Midbrain - Diencephalic - Basal Forebrain Chemoarchitectures for Prototype Emotions (Extracted from Panksepp, 1998)

Keys [(–) inhibits prototype, (+) activates prototype] [CCK = choleocystokinin, CRF = corticotrophin releasing factor, ACTH = adrenocorticotropic hormone, DBI = diazepam binding inhibitor, ACh = acetylcholine, DA = dopamine, MSH = melanocyte stimulating hormone, NPY = neuropeptide Y]

As noted, the neural architectures that Panksepp outlines are much lower in the neuroaxis than most emotion theorists have tended to look (where the attention is typically focused on the telencephalon – particularly various paralimbic regions and subcortical gray matter structures (principally the amygdala). Mostly interestingly, these prototype affect-specific neural architectures in various midbrain-diencephalic circuits all project to either dorsal and/or ventral periaquaductal gray (PAG), with the positive systems projecting much more to the ventral regions, while the negative systems project much more to the dorsal regions.

The reticular regions in ILN and MRF (which PAG projects to), in concert with highly distributed regions with which they are highly interactive, form the essential neural architectural foundations for consciousness in various iterations of ERTAS theory (Baars, Newman, Taylor, Llinas et al, Engels et al, being among the major contributors to this point of view). Additionally, there are neural connectivities between the global value mapping supported in PAG, ambient spatial mapping in superior colliculus, and primitive motor mapping in the deep tectal and tegmental motor systems. These midbrain and pontine regions are in register with each other, and their connectivities may form some of the most primitive and essential substrates for a neural representation of the self. Indeed there may be several other primitive systems in the base of the brain that contribute to the creation of such a primitive self map. (Note that this map probably can not come from one structure acting as a convergence zone.) This is an argument Panksepp’s last chapter advances cogently. Panksepp emphasizes the predominance of motor maps, a position that I don’t fully agree with, as I believe that essential foundations for consciousness cannot be unimodal by their very nature (granting of course that the body motor mapping is essential) and can only be in the cross-talk between different types of systems that establish correlations between basic forms of neural "information." It would be very hard to know that one existed if one could not correlate on-going sensory changes with activated action schematas, and both of these with value schematas that generated and predicted inherent internally valenced rewards and "punishments." It is these correlations that may enable the most basic and primitive feeling that we exist, and that what we do matters, and has effects, good and bad. Without primitive value correlates (possibly largely contributed from PAG) interacting with the primitive sensory and motor mappings, sensory-motor correlations might not mean anything by themselves. This poorly appreciated midbrain-brainstem integration of sense, value and action may form foundations for a primitive yet superordinate "self-model" that Metzinger sees as an essential foundation for consciousness (Metzinger, 1998), underpinning the normal "ownership" of qualia, a basic property of consciousness as yet unexplained.

Panksepp’s work has personally aimed me in the direction of an active interest in further defining PAG function from its scattered and fragmented literature in behavioral neuroscience. PAG may be one of the more important convergence systems in the brain. I am increasingly convinced (see Watt, 1998b for an extended treatment) of PAG’s foundational role in affect, and for some, albeit poorly understood, crucial role in consciousness, in concert with its many connectivities to reticular tissues in thalamus and brainstem. A key issue here is the meaning of its receiving projections from all of the affective prototype systems in the basal forebrain and diencephalon, which raises the question (not addressed at all in the PAG literature curiously) of whether it is computing both competition and agonism between these prototype states. If PAG is a clearinghouse for projections from the various distributed affective systems in the diencephalon and limbic basal forebrain, it may function as a primitive ventral nRt – "computing" some version of an ongoing competition between the relative activation states of the prototypes. The various positive and negative (defensive/appetitive) affective systems project quite differentially to four adjacent dorsomedial, lateral, ventral (more positive) vs. dorsal (more negative) columnar systems in PAG. Although it has been understood for years that there is radial organization in these columnar systems, their function has not been mapped, and the radial organization may be a substrate that allows the columns to interact. In any case, the results of PAG "computations" are highly distributed both dorsally and ventrally in the brain, probably tuning the response of various reticular (including MRF) and monoamine systems, and projecting upwards to several key non-specific thalamic systems that appear to critically underpin gating and binding (Newman and Baars, 1993; Llinas et al, 1994; Newman, 1997). Regarding the reticular connectivities, the four columnar systems in PAG project with overlapping as well as quite differential efferents to different portions of the ILN (intralaminar nuclei), some parts of MRF (midbrain reticular formation, particularly the cuneiform nucleus) and other midline thalamic systems. PAG may even be essential for the arousal or maintenance of a conscious state, or at least for virtually all behavioral intentionality, as PAG lesions seem to have very serious effects on the ability of the organism to maintain a conscious state. Unfortunately, the lesion work on rats has included lesions of superior colliculus in the ablations of PAG, so a clean and complete PAG lesion in rats has not yet been assessed to my knowledge, but full PAG lesions in primates seem to badly impair consciousness, resulting in a dim, profoundly abulic, twilight state. (See Watt, (1998a), and the associated commentaries in the ASSC Electronic Seminar on Emotion and Consciousness at http://server.phil.vt.edu/assc/esem.html for an extended discussion of the complexities of PAG functional correlates.)

Overall, I would argue that PAG is playing a central role in integration of the most primitive (and essential) features of affect that make affective experience intrinsically "valenced" (the essential "plus or minus" nature of affect) but without the essential higher cognitive correlates. By higher cognitive correlates, I mean the crucial adaptive ability to know when to activate these responses, which depends on appraisals, correlations and personal meanings that could only be formed in the more dorsal regions of the brain over the lifetime of the organism’s affective learning (mediated by various synaptic plasticity mechanisms). This is the crucial distinction made earlier between substrates for primitive affective states of the organism vs. substrates for affective meanings. Thus, PAG may be a complex convergence zone that underpins the organization of affective behavior at the most primitive and basic level in the brain, but one quite dependent on various higher systems to adaptively trigger the integrated emotional responses that PAG organizes. PAG then distributes those affective operators to core reticular systems at both the brainstem and thalamic levels that critically underpin conscious states and also feeds back to hypothalamus and core monoamine systems. PAG’s importance for all of the basic global state functions has yet to be fully appreciated (although its role in pain has been understood for years), but Panksepp’s work brings it much more into the neuroscientific spotlight.

Neuroscience is now so overwhelming and vast a frontier that even the most gifted and driven cannot keep up with events in one large subfield, while staying current in many or most is impossible. In this context, a stitching together of many ragged fabrics into a more smoothly seamed theory of global state functions -- that consciousness, self, emotion, executive functions, attentional functions, pain, etc., will all have to be understood together and not in isolation as differential manifestations of highly distributed integrative neural systems -- serves a vital and badly needed function for all of the many neuroscience disciplines. Such a theory of global state functions is in its infancy and is still being (at best ambivalently and erratically) stitched together from barely compatible fabrics. Its potential importance is rarely appreciated by those funding neuroscience research, which looks much more myopic and a place where the forest is frequently missed for the trees. This focus would bring us closer to being able to do real justice to the old notion, echoed in older neurological literature for many decades, but largely neglected by much of cognitive neuroscience, that we must understand the dynamical operations of the brain as a whole system, or else we really don't understand it at all, even in terms of the functions of its parts.

Unfortunately, there is a distressingly large gap between the potential emergence of concerted study of global state functions and the kind of thinking that controls most of the funding and the major purse strings of empirical neuroscience work. I am puzzled about how one might get pharmaceutical companies (which clearly have far and away the most cash of anyone to devote to primary empirical neuroscience) up to speed on the huge importance of a neuroscience of global state functions for a more informed, less myopic, more interdisciplinary, less theoretically isolated (and ultimately more effective and humanistic) psychopharmacology. There is mostly myopic vision in the researchers and administrators that have control over the labs in these big companies, with a total dominance of intracellular and receptor dynamics and other "microstructure" or "bottom-up" foci in these arenas. Psychopharmacology now completely dominates the managed care-behavioral health care landscape. This is due to the huge momentum of the clinical neurosciences, and how far the pendulum has swung from one extreme to the other from the 60's to the 90's around the value of long term psychotherapy. But we have not found anything resembling a balance point between "top-down" regulatory functions (such as those in high level cognitive-behavioral frameworks, or in the affective predispositions that form much of personality, or the long term effects of strong primary emotions experienced in primary relationships), with the bottom-up tuning of multiple receptor and intracellular dynamics. These are mutually constraining and deeply interactive domains of brain function. Panksepp's work argues that the current darlings of pharmacological manipulation (the monoamines) are not the great frontier in psychiatry, and that neuropeptides will potentially offer much greater therapeutic and behavioral specificity if their delivery problems can be addressed.

At a still broader philosophical level, this book addresses, at least implicitly, some of the most disturbing implications of biological and scientific perspectives on consciousness. The really hard problem is not consciousness (it’s only cognitively and scientifically a very hard problem), but our mortality (this one is hard in every possible way). The daily partial denial of this keeps us sane and free from a fundamental terror, and even milder dilutions of this primal awareness of death can be dysphoric and disruptive. The developing neuroscience of consciousness underlines that we are a thin film on a neural bubble that surely will burst someday, echoing again into our consciousness this much harder "hard problem" (our mortality). This intimation of mortality is enough of an emptying out of human existence for some that science itself either has to be rejected (as in all magical - fundamentalist points of view) or seen as just one point of view that answers some questions but not others (like basic affective questions about purpose and meaning). This splitting off of scientific perspectives from those deep questions can only yield another version of dualism.

In this context, Panksepp’s vision of our neural systems addresses at least indirectly the deep problems implicit in the widely perceived (almost de facto for some) intrinsic separation of value and science. This is a deeply divisive, polarizing notion that has in varying degrees split virtually all the major conferences on consciousness, often with little explicit identification of this as such during the fray, a curious form of unconsciousness at consciousness conferences. It has been particularly in evidence in its fractionating of perspectives at the benchmark Tucson series. From this polarization we have seen endless fights between those upset about soulless brains (a more disturbing problem for some than mindless brains) and neuroscience "reductionism" vs. those equally uneasy that souls appear to have lost their biological grounding and have headed off into the ionosphere like Icarus, quite convinced that their wings will not melt. I think that this polarization is not just unfortunate, (and unpleasant for those caught up in it) but quite destructive of better scientific inquiry. I personally see no conflict at all between the best in science and the best in religion (and also much similarity between the rigidly worst in each also), although this is not a widely shared belief. I also see no conflict at all between appreciating deep neural mechanisms and crediting that consciousness is an emergent property. Science is chock-full of examples in which more primitive levels of organization bootstrap emergent properties at virtually every level of structure. In fact, I would even argue that such bootstrapping of emergent properties from organizing processes at lower levels of structure may even be the most stunning and generally least appreciated feature of the rich landscape of hierarchically related disciplines that make up science as a whole.

This appreciation (that there is really no conflict between neural "reductionism" and the values expressed in more holistic and spiritual perspectives) seems to still mostly elude the hotly contested infant science of consciousness studies. However, a mature neuroscience of emotion will potentially illuminate the direct connections between deep neural mechanisms and our most cherished values, including those embedded in culture’s mythic-iconic religious belief systems. Affective Neuroscience offers a neurology and a neuroscience in which value and value generation lie much closer to the center of the consciousness "machinery," and one in which emotion and cognition are placed in a proper and much closer evolutionary relationship, with cognition evolved not independent from emotion, but as an outgrowth of it. Cognition is an evolutionary enhancement of our ability to meet prototypic adaptive challenges that are intrinsically affective. Whether the infant field of consciousness studies eventually substantiates Jaak Panksepp’s hypothesis that affect is a central organizing process for sentience (and that therefore mammals have a primitive sentience) remains to be seen, but he is not alone in these suppositions. Panksepp’s vision of the brain ends up being much more resonant with basic wishes to feel grounded, and to believe that value is neither flimsy, nor dependent on doctrine that one accepts on the authority of a church or other cultural group, nor just a fleeting idiosyncratic "computation" that can be neither true or false, instantiated in a transient neural bubble. It suggests instead that nature does see to it (through evolutionary processes profoundly affecting the development of CNS structures) that value is built into us at the very deepest levels of our brains, that we have not been abandoned and left in an emptied out, mechanical universe, although we may do a poor job of interpreting those basic signals about value. This is not to suggest some happy or easy vision of emotion and the powerful and potentially destructive evolutionary legacies that emotion embodies. On the contrary, we struggle to varying degrees all our lives with fundamental emotional ambivalence towards others, potentials deeply bred in the uneasy co-existence and interpenetration of powerful defensive and attachment-appetitive systems that have their specific origins lost in the antiquity of mammalian and even reptilian evolution. Appreciation for the destructive power and the many subtle textures of human ambivalence might be the beginning of emotional wisdom. We must respect this neurobiological ground underneath us or we cannot make any progress in understanding human emotion. In this regard, fundamental needs for connection and attachment to con-specifics from our mammalian heritage were strengthened in the highly social lines of primate and hominid evolution, and serious early derailments and denials around those basic neurobiological needs damage us fundamentally (see Schore, 1994 for a review). The vicissitudes of gratification and frustration of these fundamental emotional needs determine epigenetic sequences propelling us towards the people we become.

But this more affectively resonate and hopeful view of human and animal brain function in which fundamental biopsychosocial value is deeply woven into the neural fabric would mean little without the scientific integrity of a solid empirical foundation. The truest measure of the worth of this work by Jaak Panksepp is that is has both. This book is a must-read for anyone seriously interested in emotion. Although there remains very much work to do on the problem of emotion, this work addresses some of the areas of our greatest ignorance and aims researchers in crucial directions generally not getting the attention they deserve. Panksepp’s Affective Neuroscience gives us a much needed shove towards a future neuroscience of the whole person, what will no doubt someday be an integrated cognitive-affective neuroscience having much to say about the nature and origins of consciousness.

References

1. Baars, B. (1988) A Cognitive Theory of Consciousness, Cambridge University Press.
2. Baars, B. J. (1993), 'How does a serial, integrated and very limited stream of consciousness emerge from a nervous system that is mostly unconscious, distributed, and of enormous capacity?' In CIBA Symposium on Experimental and Theoretical Studies of Consciousness, ed. G. Bock & J. Marsh (London: John Wiley and Sons), pp. 282-90.
3. Baars, B.J. & Newman, J. 1994. A neurobiological interpretation of the Global Workspace theory of consciousness. In (A. Revonsuo & M. Kamppinen, eds) Consciousness in Philosophy and Cognitive Neuroscience. Lawrence Erlbaum.
4. Baars, B.J. (1996), In the Theater of Consciousness: The Workspace of the Mind (NY: Oxford Press).
5. Baars, B.J., Newman, J. & Taylor, J.G. (1998) Neuronal mechanisms of consciousness: A relational global workspace framework. In S. Hammeroff et al. (Eds.) Towards a Science of Consciousness. Cambridge, MA: MIT Press.
6. Baddeley, A (1986) Working Memory. Oxford Univ Press, Oxford.
7. Chalmers, D. (1996) Facing up to the problem of consciousness, Journal of Consciousness Studies, 2 (3), pp. 200-19.
8. Crick, F. & Koch, C. 1995. Are we aware of neural activity in primary visual cortex? Nature 375: 121-23.
9. Damasio, A. (1994) Descartes’ Error: Emotion, Reason, and the Human Brain. New York: Avon Press.
10. Damasio, A. (1998) Emotion in the perspective of an integrated nervous system. Brain Research Review. May; 26 (2-3): 83-86.
11. Engel, A.K., Fries, P., Roelfsema, P.R., König, P. & Singer, W. (1997) Temporal Binding, Binocular Rivalry, and Consciousness Association for the Scientific Study of Consciousness. On Line Seminar, http://server.phil.vt.edu/assc/ esem.html
12. Freeman, Walter (1995) Societies of Brains. New York: Lawrence Erlbaum Associates
13. Lane, R.D. (1998) Subregions within the anterior cingulate cortex may differentially participate in phenomenal and reflective consciousness awareness of emotion. http://www.zynet.co.uk/ imprint/Tucson/2.htm
14. LeDoux, J. (1996). The Emotional Brain. The Mysterious Underpinnings Of Emotional Life. New York: Simon and Schuster.
15. Llinas, Ribary, Joliot and Wang (1994) Content and context in temporal thalamocortical binding. In G. Buzsaki et al. (Eds.) Temporal Coding in the Brain. Berlin: Springer Verlag.
16. Metzinger, T. (1998) "Being No One" Plenary Address, ASSC Conference on Neural Correlates of Consciousness, June 21, 1998. Bremen, Germany.
17. Newman, J. and Baars, B.J. (1993) A neural attentional model for access to consciousness: A Global Workspace perspective. Concepts in Neuroscience, 4:2, 255-290.
18. Newman, J. (1997) Putting the puzzle together: Towards a general theory of the neural correlates of consciousness. Journal of Consciousness Studies, 4:1&2, 47-66, 101-121.
19. Schore, A. (1994) Affect Regulation and the Origins of the Self. The Neurobiology of Affective Development. Hillsdale, NJ: Lawrence Erlbaum Associates.
20. Taylor, J.G. (1998) The Emergent Mind. MIT Press (in press)
21. Watt, D.F. (1998a) Implications of Affective Neuroscience for Extended Reticular Thalamic Activating System Theories of Consciousness. Target Paper - ASSC Electronic Seminar on Emotion and Consciousness at http://server.phil.vt.edu/assc/esem.html.)
22. Watt, D.F. (1998b) Affective Neuroscience and ERTAS (Extended Reticular Thalamic Activating System) Theories of Consciousness. Proceedings of the Ischia Conference on Emotion and Consciousness. In press.