Categorical versus Dimensional Models of Affect (Consciousness & Emotion Book Series) 9027241570, 9789027241573

One of the most important theoretical and empirical issues in the scholarly study of emotion is whether there is a corre

135 22 4MB

English Pages 356 [357] Year 2012

Report DMCA / Copyright

DOWNLOAD FILE

Polecaj historie

Categorical versus Dimensional Models of Affect (Consciousness & Emotion Book Series)
 9027241570, 9789027241573

Citation preview

Categorical versus Dimensional Models of Affect

Consciousness & Emotion Book Series Consciousness & Emotion Book Series publishes original works on this topic, in philosophy, psychology and the neurosciences. The series emphasizes thoughtful analysis of the implications of both empirical and experiential (e.g., clinical psychological) approaches to emotion. It will include topical works by scientists who are interested in the implications of their empirical findings for an understanding of emotion and consciousness and their interrelations. For an overview of all books published in this series, please see http://benjamins.com/catalog/ceb

Editors Ralph D. Ellis

Clark Atlanta University

Peter Zachar

Auburn University Montgomery

Editorial Board Carl M. Anderson

McLean Hospital, Harvard University School of Medicine, Cambridge, MA

Bill Faw

Brewton Parker College, Mt. Vernon, GA

Eugene T. Gendlin

Maxim I. Stamenov

Jaak Panksepp

Douglas F. Watt

Valerie Gray Hardcastle

Martin Peper

University of Chicago

Bowling Green State University, OH

Bulgarian Academy of Sciences Quincy Hospital, Boston, MA

Advisory Editors Bernard J. Baars

Wright Institute, Berkeley, CA

Thomas C. Dalton

Virginia Polytechnic Institute, Blacksburg, VA

California Polytechnic Institute, San Luis Obispo, CA

Alfred W. Kaszniak

Nicholas Georgalis

Alfred R. Mele

George Graham

Natika Newton

East Carolina University, Greenville, NC Wake Forest University, Wake Forest, North Carolina

University of Arizona, Tucson, AZ Florida State University, Talahassee, FL Nassau County Community College, New York

University of Marburg, Germany

Edward Ragsdale New York, NY

Howard Shevrin

University of Michigan, Ann Arbor, MI

Lynn Stephens

University of Alabama, Birmingham, AL

Kathleen Wider

University of Michigan, Dearborn, MI

Volume 7 Categorical versus Dimensional Models of Affect. A seminar on the theories of Panksepp and Russell Edited by Peter Zachar and Ralph D. Ellis

Categorical versus Dimensional Models of Affect A seminar on the theories of Panksepp and Russell Edited by

Peter Zachar Auburn University Montgomery

Ralph D. Ellis Clark Atlanta University

John Benjamins Publishing Company Amsterdam / Philadelphia

8

TM

The paper used in this publication meets the minimum requirements of the American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984.

Library of Congress Cataloging-in-Publication Data Categorical versus dimensional models of affect : a seminar on the theories of Panksepp and Russell / edited by Peter Zachar and Ralph D. Ellis. p. cm. (Consciousness & Emotion Book Series, issn 1566-5836 ; v. 7) Includes bibliographical references and index. 1. Affect (Psychology) 2. Emotions. I. Ellis, Ralph D. II. Zachar, Peter. BF175.5.A35C38   2012 152.4--dc23 2011051987 isbn 978 90 272 4157 3 (Hb ; alk. paper) isbn 978 90 272 7475 5 (Eb)

© 2012 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa

Table of contents

chapter 1 Introduction: Categories, dimensions, and the problem of progress in affective science Peter Zachar

1

chapter 2 In defense of multiple Core Affects Jaak Panksepp

31

chapter 3 From a psychological constructionist perspective James A. Russell

79

chapter 4 “Nature proposes…and science disposes” tertiary vs primary process approaches to emotions and affects: Commentary on Jim Russell’s position Jaak Panksepp

119

chapter 5 Preliminary comments on Panksepp James A. Russell

129

chapter 6 Discrete emotions: From folk psychology to causal mechanisms Andrea Scarantino

135

chapter 7 Nothing in mammalian psychology makes sense except in light of primary-process affective capacities Paul Sheldon Davies chapter 8 Lessons for affective science from a metascience of ‘molecular and cellular cognition’ John Bickle

155

175

vi

Categorical versus Dimensional Models of Affect

chapter 9 Affect as appraisal Meaghan A. Leddy, Gail C. Robertson and Jay Schulkin

189

chapter 10 What should theories of emotion be about? James R. Averill

203

chapter 11 Valence, reductionism, and the ineffable: Philosophical reflections on the Panksepp–Russell debate Ronald de Sousa chapter 12 Functional and empirical presuppositions in Russell and Panksepp: Neural predispositions of affect Georg Northoff chapter 13 Comparison of affect program theories, appraisal theories, and psychological construction theories Agnes Moors

225

243

257

chapter 14 Final remarks James A. Russell

279

chapter 15 My reflections on commentaries and concluding perspectives Jaak Panksepp

301

chapter 16 Concluding observations: Comparisons, contrasts, and some important convergences  Ralph D. Ellis and Bill Faw Index

321

347

chapter 1

Introduction Categories, dimensions, and the problem of progress in affective science Peter Zachar Department of Psychology, Auburn University Montgomery, Alabama

1. Introduction The romantics of the 18th century were concerned that scientific explanation had drained natural philosophy of color, feeling, and meaning. One expression of this concern was the organicist perspective of Schelling and Goethe which held that we should use insights gained from being a part of nature to understand natural phenomena (Gregory, 2008). In psychology this tradition evolved into the question of whether and to what extent scientific psychology should adopt the methods of a natural science like physics or whether some topics of central concern to psychology also require a historical and interpretive analysis. For the early psychologists, the methodenstreit was a disagreement about the merits of methodological pluralism. Always lurking in the background of this debate, however, was a (Kantian) assumption that some topics lie beyond the boundaries of science. In addition to the German founders of scientific psychology, one can also detect this perspective in both religious thinkers who wanted to protect theological notions such as the soul from being completely colonized by the sciences and scientific thinkers who were suspicious of the theological roots of certain psychological concepts (such as the self). Both the anti-modern Romantics and the most committed empiricists in the tradition of Hume saw some traditional topics in philosophical psychology as being beyond scientific reach. In the last four decades of the 20th century, in both psychology and philosophy, some empiricists backed away from being so sure about the boundaries of science and currently hold that we should not prejudge where and to what extent some clever scientist can find a way to make a phenomenon empirically tractable (Dennett, 1991; Quine, 1969). In part, this is one reason why topics such as self, consciousness, and emotion are scientifically respectable again.



Peter Zachar

Both Jaak Panksepp and James Russell advocate for the scientific study of what was once considered an empirically intractable concept, namely, affect – although they have very different views on the matter. The most obvious difference between them is their contrasting positions on the nature of affect. According to Panksepp affect is quantized in the physicist’s sense of that term. In physics “quantized” means that at the subatomic level properties such as energy level, momentum, and charge are not continuous; rather, like whole numbers, they come only in discrete quantities. In Panksepp’s psychology primary affect refers to a set of discrete natural kinds that are qualitatively distinct from each other. These primary affects, of which there at least seven, are the products of distinct neurophysiological systems that are rooted in subcortical structures and that predate human evolution. Panksepp is a committed scientific realist about theses affects. Russell holds that all affective phenomena share the same fundamental structure – being composed of two independent dimensions. The first dimension is pleasure versus displeasure and the second is low versus high arousal. Rather than being quantized in discrete packets, Russell views affect as continuous, analogous to the energy waves of the electromagnetic spectrum. Core affect in particular represents a dynamically shifting position on Russell’s two continuous dimensions of valence and arousal. Russell implicitly adopts a realist interpretation of core affect and his colleague Lisa Feldman Barrett (2006) opines that it may be a natural kind.

2. An intellectual biography of Jaak Panksepp To a limited extent, one can trace the development of these different perspectives by examining the background each thinker brings to their study. Jaak’s bête noire is the radical behaviorism of B. F. Skinner, the midcentury empiricist who believed that some constructs such as affective feelings (construed as subjective entities) are not credible topics of scientific study and need to be analyzed in terms of observable behavior, or eliminated from psychology. Panksepp vigorously rejects such eliminativism by proposing that subjective feelings such as joy and pain are neuroscientifically tractable aspects of animal minds. Panksepp began his doctoral studies in clinical psychology in 1965 at the University of Massachusetts in Amherst. He was expecting to learn about emotional disorders and the nature of affect, but the largely behaviorist faculty did not share his interests, nor did he share their exclusive focus on behavior modification. His first-year traineeship at the Northampton Veterans Administration Hospital allowed him the flexibility to pick his own research problems and he chose



1.  Introduction

self-stimulation reward in rats (Bingman, in press). This work hooked him and he switched to physiological psychology, studying with Jay Trowill (1939–1983). Trowill had recently completed a Ph.D. under Neal Miller, who is well known in psychology for joining John Dollard’s attempt to reconcile psychoanalysis with learning theory (Dollard & Miller, 1950). One advantage to Panksepp of being in Trowill’s lab was that he was not forced to adhere to strict behaviorist doctrines. A second advantage was that he could pursue his own research topics, which ended up being on the incentive properties of “brain reward” (Trowill, Panksepp, & Gandelman, 1969), which he later conceptualized as the seeking/expectancy­ system (Panksepp, 1981, 1982, 1986, 1991). After earning his Ph.D., Panksepp completed two post-doctoral fellowships (in nutrition and sleep physiology) before being hired at Bowling Green State University (BGSU) in 1972. As an assistant professor he worked on topics such as sleep and feeding (Panksepp, 1974), but after gaining tenure in the mid 1970s he began to emphasize emotion with studies on attachment and social bonding. His group mapped out the brain systems for separation distress (his panic/grief System) and identified its major neurochemistries, yielding testable ideas about the neural sources of social attachments. Crucially in terms of academic politics, his interests were supported by Bob Conner (1939–2000) and John Paul Scott (1909–2000), two senior members in his department (AJP, 2010). After Panksepp began working on the brain mechanisms of social attachment, he became more familiar with the work of John Bowlby (1969) – an ethologically influenced psychodynamic theorist. Panksepp had no ideological opposition to psychodynamic theory, and reported to me that he was attracted to the possibility of providing a neuroscientific foundation to a grand clinical theory. As the cognitive revolution in psychology advanced, it included an increasing emphasis on emotion, but not on the neurobiology of emotion (until the advent of modern brain imaging in the 1990s). Furthermore, federal research funding policies helped motivate interests in behavioral neuroscience among psychologists trained in behaviorism, and they brought with them their opposition to any hint of anthropomorphism. Panksepp told me that the discussion of and physiological study of emotional subjectivity in animals quickly became taboo in the newly minted discipline of behavioral neuroscience, so he decided that the best option for him was to introduce a subdiscipline called “affective neuroscience” (Panksepp, 1991). Having left the behaviorists behind only to have them conquer his chosen territory and, in a sense, make him a scientific outsider deepened his antipathy toward the unwillingness to allow the study of affect in animals or to explore the implications of animal research for understanding human feelings. In Panksepp’s view all mammals are endowed with conscious affective feelings. One of the findings that he is best known for is his discovery in the mid 1990s that





Peter Zachar

rats “laugh” during play and when tickled (see his Affective Neuroscience, p. 287). That “laugh,” says Panksepp, takes the form a 50 kHz chirp. He believes that this research serves as another bridge toward ethological measures that can be directly employed to monitor affective changes in animals. Although he advocates for a bottom-up approach based on studying subcortical structures, Panksepp does not believe that interpretation must be constrained in a bottom-up way; rather, information about experience and behavior as they appear across species (including humans) is needed to fully understand sub-cortical activity. If we use all the evidence available to us, says Panksepp, we can see that rats do, in fact, have a response that is homologous to our laughter, along with many other emotional reactions. This argument for seeing laughter in rats is similar to Galileo’s argument for seeing four moons around Jupiter. If you can show that your method of observation provides information that can be readily confirmed (the subjective feeling of human neural-behavioral-affective patterns; what is on the deck of the far distant ship that will eventually be in port), and convincingly argue that what it reveals about more distant objects is reliable information (chimpanzee affect, craters on the moon), then we should consider accepting what it suggests about even more distant objects (rat chirps and never before seen points of light in the night sky in the vicinity of Jupiter). Panksepp retired from BGSU in 1998. The previous decade had been very challenging, beginning with the death of his 16 year old daughter Tiina on Good Friday of 1991 – which led to depression – and was soon followed by two separate year-long disruptions in his work due to asbestos abatements in the Psychology Building. All this factored into his decision to write Affective Neuroscience, the book in which he summarized the scientific evidence for at least seven affective systems and also decided to forcefully defend his perspective on the reality of experienced affect in animals, whatever the cost to his scientific reputation (­Panksepp, 1998). At this time his university was discontinuing its early retirement program and, because of drop-outs by more senior faculty, Panksepp learned at the last minute that he would be eligible to retire, but with an imminent deadline; he had to make an immediate decision on whether he would accept. With Affective Neu­ roscience soon to be published, and a deep sense of fatigue, he decided to see what the world had to offer him. One month after officially retiring, it was discovered that he had a massive lung lymphoma which could be treated. Once his fatigue abated, he returned to his scholarship with gusto. (Jaak underwent even more radical treatments for another occurrence just as this current project commenced). In the first year after his retirement and recovery, he taught at the University of Michigan at the invitation of Warren Holmes. The following year he moved to



1.  Introduction

Chicago to work with biochemist Joe Moskal, who had spent time in Panksepp’s lab while he was an undergraduate student at Notre Dame. Panksepp eventually joined Moskal at the Falk Center for Molecular Therapeutics at Northwestern University. In Moskal’s lab they have been using the affective neuroscience framework to search for better pharmacological treatments for psychiatric disorders, especially major depressive disorder. As of this writing, a new medication derived from this animal research program (GLYX-13) has made it to a Phase 2 FDA approved human clinical trial. In 1999, the psychoanalyst Mark Solms invited Panksepp to join the editorial board of Neuropsychoanalysis and to work on a neuroscience counterpart to the psychoanalytic theory of emotions. Although one can see some shared terminology between Freud and Panksepp (e.g., primary process), psychoanalytic theory had limited influence on Panksepp’s developing views. Both Freud and Panksepp, however, began their respective projects wanting to develop a biology of the mind rooted in ancient, subcortical regions of the brain. The classical Freudians believed that all psychological phenomena are fueled by organic needs which are represented in consciousness as drives. Freud’s biology of the mind, based on 19th century findings, had no explicit role for emotion. Panksepp was more influenced by the work of Walter Hess (1957) and Paul MacLean (1990), which showed that not only motivations but also emotions are rooted in limbic structures. By the late 1990s object relations theory, with its emphasis on positive and negative affect, had become a dominant perspective in the U. S. – which only increased the parallels between psychoanalytic theory and Panksepp’s research. Panksepp reports that Solms’ invitation was the first time that a major clinical group had shown interest in his work (aside from the small Autism community that had become intrigued by his new ideas about the genesis and treatment of autistic aloofness in the late 1980s) – the world indeed had something attractive to offer. Joining the editorial board led to his giving a keynote presentation at almost every subsequent meeting of the International Neuropsychoanalysis Society and also to a series of what Panksepp describes as “free-wheeling workshops” with many prominent members of the psychoanalytic community participating. Since 2005 he has been the Bailey Endowed Chair of Animal Well-Being ­Science at the College of Veterinary Medicine at Washington State University. Invited to apply for this position, Panksepp was selected from a list of five finalists. Just prior to occupying the endowed chair, he taught a course on affective neuroscience, which he had offered to do at Boston College for Jim Russell’s and Lisa Feldman Barrett’s groups. In addition to continuing his empirical research at both Northwestern and Washington State, he has actively explored the connections between his work and neuropsychoanalysis and also explored the connections





Peter Zachar

with an area of interest that first made him consider studying psychology: the philosophy of mind – e.g., see Archaeology of Mind: Neuroevolutionary Origins of Human Emotions (Panksepp & Biven, in press).

3. An intellectual biography of James Russell Jim Russell entered graduate school at UCLA planning to study personality processes with Albert Mehrabian, who researched non-verbal communication and had written a book on personality theory. At this time Mehrabian was becoming interested in environmental psychology – i.e., the scientific study of person and environment interactions – and tasked Russell with reading the literature and developing research topics. They chose to study the extent to which environments influenced people by conveying moods – what they eventually named an emotion theory of places (Mehrabian & Russell, 1974). In seeking a parsimonious model of how emotions might mediate between the place and the person, they proposed a three dimensional structure – pleasure versus displeasure, low versus high arousal, and dominance versus submissiveness (Russell & Mehrabian, 1977). Russell was later hired to develop the environmental psychology program at The University of British Columbia. It was challenging for a trained scientist to work with applied people such as architects. For example, unlike scientists (in that era) the applied disciplines tended to see knowledge as being proprietary. EP was a complicated mix of different disciplinary perspectives, and the program did not take off as was hoped. During this time Russell also became more interested in the nature of the emotional mechanisms that he had studied in graduate school. In the late 1970s Russell returned to the affect dimensions, but decided to work with the circumplex model he had originally preferred (dropping the third dimension). Jerry Wiggins’ (1979) work on circumplex models in personality theory was an important influence on this decision. The way he approached the nature of affect was to ask whether there were any parallels between the representation of affective qualities in the environment and the cognitive representation of affect. He found that there were (Russell, 1980). Both kinds of representations could be modeled using the circumplex. He also used principal components analysis to uncover the latent variables underlying affective self-reports, and discovered that the resulting dimensions matched the patterns seen in the studies of the cognitive representation of affect. Shifting to a topic that was more interesting to him, the circumplex was re-interpreted as referring not only to the representation of affect, but about affect itself. Around this same time Eleanor Rosch (Rosch, 1976) had published her work on natural categorization and prototype theory. Russell wrote a paper attempting



1.  Introduction

to apply the prototype model to the emotions and shared it with a class of graduate students. Beverly Fehr, one of the social psychology students in the class, used the paper as the inspiration for her master’s thesis. In these studies she discovered that emotion was conceptualized as a superordinate category that had good instances (anger), less good instances (pride) and instances that seem to be on the boundary of the emotion category (boredom). Furthermore, the basic categories such as anger were themselves structured so that there were good and less good attributes of anger. In the article reporting these results, Fehr and Russell (1984) made it clear that they were talking about concepts – not actual emotion events. Drawing on Russell’s earlier paper, they offered the first public presentation of what philosophers would refer to as Russell’s nominalism. As stated in the article, any particular psychological event can be considered an emotion only to the extent that it can be brought under an emotion concept. Interestingly, they also referred to a discussion among researchers about whether Panksepp’s neural systems (like FEAR) were really emotions. Pointing out that scientists need some concept of emotion to guide their study of events, they observed that many researchers act as if folk concepts of emotion should play a normative role in science. In sympathy with Panksepp, they stated that scientists are not bound by folk concepts. New concepts can pick out a new class of events that they unite. When novel scientific concepts pick out unfamiliar classes of events, researchers may see them as not answering the kinds of questions that the entrenched folk concepts suggested were important, and mistakenly reject them. By the 1980s Russell had successfully made a shift into emotions – working on both affect and on emotion concepts. Russell reported to me that, like most psychologists, he readily accepted that Ekman’s research on the universality of facial expression constituted solid evidence in favor of basic emotion theory. He even spent some time in Ekman’s San Francisco lab when he was on a sabbatical at Berkeley. Later, while teaching a class in the psychology of emotion, he and his students discovered that a neutral face can be seen as sad when placed next to a happy face, or happy when placed next to a sad face. They also began playing around with the forced choice format used by Ekman and learned that they could manipulate the choices to obtain different judgments about what emotion a face was expressing. For example, a putative angry expression could be seen as contempt or disgust if those were the choices instead. The historians tell us that Darwin’s training with Henslow at Cambridge oriented him to seeing variability within a species, which was very helpful once he arrived on the Galapagos (Desmond & Moore, 1991). Likewise, one can speculate that Russell’s background in environmental psychology attuned him to potential flaws with Ekman’s model of culturally invariant basic emotions. While at





Peter Zachar

Berkeley on his sabbatical he also began a systematic study of anthropology and believes that this exposure was an important influence on his subsequent thinking about Ekman’s and Izard’s research. The result of his discoveries and his ponderings were several important critiques of the universality thesis that included a listing of alternative explanations of the findings (Russell, 1991, 1994; Russell & Fehr, 1988). These articles led to a spirited dispute with Ekman, in which Russell defended the alternative explanations as more plausible than Ekman’s explanations (Russell, 1995). A social psychologist might hypothesize that the act of defending these alternatives made Russell more likely to believe in them as well, especially the alternatives that were consistent with his dimensional model. Although the role of ecology and context was an important part of Russell’s critique, he was not opposed to invariance or the even universality, in principle. Earlier Russell had shown that his circumplex model not only described the structure of the representation of affect for college students; it worked just as well with children as young as three and also across cultures (Russell, 1983a, 1983b). In critiquing the facial expression research, he observed that there was decent agreement on recognizing happy faces, but less agreement on expressions of anger, fear, sadness, and disgust. However, although anger might be called disgust or frustration, it was never labeled happy. One interesting alternative explanation was that what is invariant is recognition of pleasurable and displeasurable affect. By the mid 1990s Russell had proposed the circumplex, explored the relevance of classical versus prototype categories for emotion concepts, and participated in a high impact critique of the research on the cross-cultural invariance of emotional expressions. In offering the alternative explanations of the research findings, Russell had not articulated a coherent theory and he was beginning to feel that he was, in his words, “working around the edges of emotion.” Foucault once argued that a critic should not be required to offer her or his own positive model. Russell did not hold to this view. He began working on a more comprehensive theoretical structure in 1996, but his first opportunity to publish this work came when Ed Diener invited him and Lisa Feldman Barrett to contribute an article to Journal of Personality and So­ cial Psychology for a special section on the structure of emotion. Feldman Barrett had sent Russell some of her work while she was in graduate school and it resulted in their collaborating on a project together. In this special issue, the term core af­ fect was proposed to account for what is modeled by the circumplex, and it was distinguished from prototypical emotional episodes, which they claimed cannot be modeled by a circumplex. Russell and Feldman Barrett (1999) also argued that their particular circumplex was empirically better than the alternatives, implicitly suggesting that it was a literally correct carving of nature at the joints. Since that



1.  Introduction

time, Feldman Barrett has become the most vocal proponent of the reality of the dimensions of core affect in contrast to discrete affect theories such as Panksepp’s (Barrett, 2006; Barrett et al., 2007). The more comprehensive theory, which integrated the work of many leading scholars into a single psychological model, was published in 2003. Very importantly, Russell (2003) disputed Ekman’s essentialist affect program model (and any other similar model) and suggested that his own non-essentialist, nominalist alternative was scientifically preferable. He argued that emotions are families of related states that are composed of a shifting menu of subcomponents. There are many different states and only a very few of them have been lumped together under common names. Neither the biological essences of Ekman and the neuroscientists nor the historical-cultural essences of the social constructionists can be consulted to specify a scientific taxonomy of emotional episodes. Instead of these traditional essences, what scientists should seek and study are the various components that feed into the observable patterns (for those episodes we name and those we don’t). An important aspect of this theory is that those particular episodes that are not named are rarely categorized as “episodes” because the patterning of the components does not match a conventional script, but they occur all the same. At this time Russell also more explicitly adopted some of the eliminativist claims that had been de rigueur in the philosophy of mind during the previous decade. Always enjoying the role of the student, he read widely and his views on this matter were gleaned from neither Skinner nor the Churchlands. A minimalist eliminativism is part of the story of scientific progress and modernism. His argument was that emotion concepts can be taken from both folk and scientific psychology, but that the two should not be confused, and that uncritical adherence to folk concepts is potentially retarding scientific progress. He concluded that one of these folk concepts is “emotion” itself. There are emotions and emotional phenomena, and scientists can specialize in studying these things, but there is no coherent natural kind called emotion. For this reason, there can no more be a science of emotion, than there can be a science of white things.

4. What is at stake in this scientific debate about affect? Both Panksepp and Russell agree that “affect” refers to subjective feelings and that such feelings can be studied scientifically. They differ on what they take to be the fundamental nature of affect, but agree that the resolution of these differences is potentially an empirical matter. Panksepp’s view is that he and Russell are studying different levels of analysis. His primary process affects emerge from the experimental study and manipulation of subcortically-based neural circuits,



10

Peter Zachar

whereas, according to Panksepp, Russell’s core affect is a structural model derived from correlational analyses of self-report data – which reflects cortical activity. Panksepp uses the terms primary and tertiary to differentiate these two research programs, but one should not conclude that he uses tertiary to mean third rate. All affect is important to Panksepp. Russell remains attracted to the hypothesis that core affect is fundamental (or simple), psychologically. He asserts that any further decomposition of affect can only be made by recourse to underlying neural mechanisms, not to a more finegrained psychological analysis. One might suppose that Panksepp’s research has uncovered these underlying neural mechanisms, but as we have seen, Panksepp’s understanding of these mechanisms does not occur in isolation from the study of experience and behavior. With his particular take on integrating information from multiple levels of analysis, Panksepp believes that lumping care and play together into an abstraction called “pleasure” would entail a loss of information (i.e., “pleasure” is too course-grained for Panksepp). In his dispute with Ekman, Russell’s general view was that the affective feelings associated with basic emotion categories are quite variable even within categories and can be more parsimoniously modeled in term of valence and arousal. He makes a similar argument with respect to the affective features of Panksepp’s primary process categories as well. Underlying neural mechanisms can support a variety of affective feelings, and in coordination with other mechanisms produce many kinds of pleasure, but the variation among different types of pleasure can be empirically accounted for by a single, underlying latent variable. There is an episode in Alice’s Adventures in Wonderland where Alice and a group of animals had just climbed out of a pool of tears onto the shore and needed to dry off. The Mouse said the driest thing he knew was the story of the Norman invasion of England. Unfortunately the Mouse’s recounting of the Norman Conquest didn’t dry anyone off. Once it was apparent that they were still wet, the Dodo bird suggested that they set up a caucus race. He outlined a roughly circular race course and everyone in the group was placed at different points on the course and began to run around (starting and stopping as they wished). After a half an hour they were all dry and the Dodo called an end to the race. When asked to declare a winner, after considerable thought the Dodo declared – everybody has won, and all must have prizes. Hopefully, recent work in the science of affect has amounted to more than running around in circles and we have gotten somewhere. Panksepp and Russell both believe so, as seen in their mutual attraction for a kind of scientific realism about their respective theories. A commitment to scientific realism seems inescapable in understanding progress and the growth of knowledge. As an exercise in considering­ alternatives, though, I invite readers to consider the possibility that a



1.  Introduction

commitment to scientific realism may not be the only way to account for progress. A more minimalist, empiricist analysis suggests that one scientific model can outcompete another without its advocates having to assert its literal truth or without explaining why it is warranted by making reference to “the approximation of reality.” It is unrealistic to believe that ontological commitments can be dispensed with, but when disputes evolve into spirited arguments about philosophical concepts like reality and natural kinds, there is a tendency for the disputants to become emotionally invested in their ontological commitments. One problem with such an investment is that evidence that than be interpreted to confirm one’s view of reality is noticed and emphasized, while disconfirming evidence is explained way. One could argue that none of the substantive empirical claims of Panksepp or Russell requires advocating scientific realism. A tentative instrumentalism (or provisional scientific anti-realism) might seem to be more of an option for ­Russell than for Panksepp, as it would be for any latent variable theorist. Core affect could be seen as a mathematical abstraction similar to general intelligence, average income, and a center of gravity. Panksepp’s neural systems are concrete physical things, not mathematical abstractions; so it might seem that he has to be a scientific realist. However, complicated neural systems are themselves models, and the description of how the model works is an idealization rather than a literal description of every particular instantiation. Kenneth Schaffner (1993, in press), who is neither a philosophical lightweight nor a scientific amateur, understands biological mechanisms along these lines. Panksepp could just as easily say his model outcompetes that of LeDoux or Damasio without having to assert that he is carving nature at it joints. Panksepp tends to be an antirealist about core affect theory and Russell tends toward anti-realism about Panksepp’s categories. Many of our commentators will not only be scientific realists about one or the other theory on the nature of affect, they will also be scientific realists about basic emotion categories. Someone who adopts a consistent scientific anti-realism will be tempted to agree with Panksepp on core affect and Russell on primary process affects and basic emotions. Such an anti-realist will, however, join with Panksepp and Russell on asserting one important point: for there to be progress, it cannot be the case that everybody wins and all must have prizes. Progress involves showing that two models are in competition, and that one is better than the other according to agreed upon standards of comparison (Zachar, in press). As Jaak’s final comments will note, despite some initial hopes for compatibility, it appears as if their two models of affect are in competition. In the following pages both Panksepp and Russell will have multiple opportunities to convince the readers of how they believe that scientific progress in the study of affect is best achieved.

11

12

Peter Zachar

5. The plan of the book and chapter summaries This book was inspired by a series of articles in the Journal of Theoretical and Philo­ sophical Psychology (Panksepp, 2008; Russell, 2008; Sundararajan, 2008; Zachar, 2006, 2008). After those articles were written and published, it became clear that there were outstanding issues between Panksepp and Russell and a more thorough exploration involving different scholars with a variety of perspectives might be helpful. This book was planned by Panksepp, Russell, Ellis, and Zachar with that goal in mind. The list of commentators was proposed and mutually agreed on by the four of us. The list included both established and up-and-coming scholars. We were pleased when almost every commentator from the original list agreed to participate. In order to provide some coherence to the volume, it was decided that all the commentators would be asked to respond to opening chapters by Pankseep and Russell. To further structure the book, the four of us developed a list of eleven questions to be answered in these two opening chapters. These questions are as follows: 1. What is affect? What is its structure? Is affect fundamental, and if so, what makes it so? 2. How does affect relate to prototypical emotional episodes? 3. What kind of affect might be had by other mammals and/or human infants? How does such affect relate to the affect of adult humans? 4. How can we infer affect from observation of behavior? 5. Is affect pre-linguistic? How can we study what is pre-linguistic? How does language alter affect? 6. Is affect always conscious? 7. What is the relationship between affect and cognition (conceptualization, intentionality, etc.)? 8. Is affect altered or filtered through culture? In what ways? Are there universal aspects of affect? 9. What roles do the biological sciences and psychological sciences have in studying affect? 10. What roles might Panksepp’s primary affects play in Russell’s model of psychological construction? What implications might Russell’s model of psychological construction have for Panksepp’s construct of primary process affects? 11. Considering the diversity of theoretical viewpoints in the scientific study of affective and emotional phenomena, might one hope for a coherent synthesis in this splintered field?



1.  Introduction

In addition to their opening chapters, Panksepp and Russell respond to each other with a second chapter. These are followed by the commentaries. All the commentators were given both of Panksepp’s and both of Russell’s chapters. They were asked to comment on the chapters and also to offer their own position on whichever aspects of the general topic they believed was most important. Following the commentaries, Panksepp and Russell each have a third chapter in which they respond to the commentaries. The book closes with a summary chapter coauthored by Ralph Ellis and the fair-minded and academically rigorous biological psychologist Bill Faw. These summaries were written over a period of six months as the commentaries came in. I have no good justification for why some are longer than others – although the longer sections were written when I had less time to tighten them up. Each description of a commentary is more of a précis, whereas the chapters of Panksepp, Russell, and Ellis-Faw are briefly summarized.

The opening chapters by Panksepp and Russell The first two chapters contain Panksepp’s and Russell’s answers to the eleven questions. These chapters are followed by their responses to each other. In his response, Panksepp contends that core affect theory should be revised to accommodate information about underlying neural mechanisms. He believes that the neural evidence favors categories of affect at the most basic level, whereas the question of arousal (categorical or dimensional) is not yet decidable. He also responds to four questions proposed by Russell in his opening chapter. In his response, Russell questions the specificity of Panksepp’ s attribution of feelings to animals, noting that the underlying neural mechanisms are complex and multifaceted. He does not dispute that mammals possess conscious feelings, but is not convinced that they can be directly read off from the underlying neural activity, because variations in pattern (even in one of Panksepp’s circuits) might produce variations in affect. The lumping of all these variations into one experience called hunger or fear is done by us and relies on additional psychological processes such as attention, memory, evaluation of context, etc. He would likely allow that we can lump animal affects together as well – but the variation that is actually there is more parsimoniously modeled in terms of his dimensions. Andrea Scarantino Early in his chapter, Andrea Scarantino makes an astute observation. Both Panksepp and Russell, are skeptical about any possibility of grounding a progressive research program in the study of folk emotion concepts. With this shared starting point, however, they end up in radically different places. Panksepp claims that his primary process affective systems should not be referred to with vernacular­

13

14

Peter Zachar

(folk psychological) terms like fear and anger. Likewise, Russell claims that a folk emotion concept such as fear does not refer to a single scientific kind, but to a family of particular states with varying degrees of overlap. Yet, observes ­Scarantino, Panksepp advocates for the existence of an internal infrastructure that causes and/or coordinates discrete emotional packets, whereas Russell rejects this same causal model. Scarantino proceeds to explore four criticisms that Russell, Panksepp or both give for the inadequacy of folk emotion concepts. Scarantino considers only one of these four criticisms to be valid. The first criticism is that discrete emotions are not structured like classical categories with (a) necessary and sufficient defining properties and (b) clear boundaries between members and non-members of the category. The presence of such a structure is one way of identifying a natural kind (Zachar, 2000). ­Russell suggests that, since folk emotion concepts do not have a classical structure, they are not scientifically legitimate kinds along the lines envisioned by the affect program model. In response Scarantino notes that there are alternative notions of natural (i.e., scientifically legitimate) kinds that do not adhere to the classical category model. The most prominent of such notions is Richard Boyd’s homeostatic property cluster model. It is well known that Darwin’s model showed that, in contrast to the Aristotelian view, species do not have fixed necessary and sufficient properties that all members share, nor is the boundary between members and non-members always distinct. Boyd famously argued that members of a species such as “tiger,” with all the variability noted by Darwin, are still unified by a set of shared causal mechanisms. Scarantino suggests that if species can be scientifically legitimate kinds, so can prototypically structured emotional constructs. The next criticism, shared by Panksepp and Russell, focuses on how emotion concepts vary across cultures. For Russell the extensive variation that exists is a telling criticism against Ekman and Izard’s proposal of a universal set of basic emotions that are grounded in biology. Opposing Russell on this point, ­Scarantino argues that cultural variability has little to say about the potential ­scientific legitimacy of a concept. Science, he suggests, can refine and improve upon cultural assumptions, and some cultures may better approximate scientific reality than others. The third criticism is more specific to Panksepp’s justification for naming his systems with capitalized titles such as fear, seeking, etc. Scarantino notes that Panksepp believes that fear is a subset of a family of states that in the vernacular are called “fear.” Never should a part (fear) of a conceptual whole (fear), be confused with that whole, says Panksepp. As Scarantino notes, the problem with this justification is that on Panksepp’s own model, the relation between the scientific part and the vernacular whole does not hold. For example, the vernacular whole



1.  Introduction

“fear-panic-anxiety” needs to be analyzed with respect to two distinct scientific parts – fear and separation distress. Rather than using vernacular-sounding terms, alternative naming conventions would have been possible. For example, Panksepp could have defined these systems in terms of their evolutionary function. If one word was wanted, he could have introduced a technical term like WIS4. Scarantino, however, is reluctant to recommend using neologisms in the style of Jerome Kagan or Raymond Catell. Instead he offers two suggestions for when it would be appropriate to use terms drawn from the vernacular to refer to affective systems. First, vernacular-sounding terms would be appropriate if fear is not a part of “fear” but rather a subset of “fear.” In other words, not every state that the folk call “fear” has to instantiate fear, but if something is fear, it is also “fear” in the same way that if something is a dog, it is also a mammal. Second, vernacular-sounding terms would be appropriate if fear had a high degree of overlap with “fear,” but without being a subset proper. In the latter case, there may be some cases of fear that are not members of “fear.” He believes that such fuzziness could be accommodated if Panksepp’s affective systems were themselves conceptualized as non-essentialist property clusters in Boyd’s sense. Scarantino considers the fourth criticism (favored by Russell) to be legitimate. To understand this criticism we need to return to the notion of natural kinds. One of the ideas underlying the conventional construal of a natural kind such as “gold” is that the necessary and sufficient properties are also causal properties; they are analogous to hidden essences that make gold be what it is. Consider the 79 protons in the nucleus of gold. The 79 protons bear a causal relationship with all other properties of gold, and if you want to reliably determine whether something is really gold, all you have to do is to verify the number of protons in the nucleus. In affective science, the hypothesized causal essences are usually referred to as affect programs. Scarantino agrees with Russell that the scientific search for single homogeneous essences that are causally responsible for discrete emotions has been a failure. The extension of the concept “fear” is too heterogeneous to be unified under a single causal essence. Scarantino does, however, take issue with Russell’s nominalist view that there are many particular emotions that are called anger, but no such thing as a unique anger mechanism that produces them and thereby makes them be the same kind of thing. Like Russell, he advocates for a non-essentialist perspective and does not believe that there is a single mechanism that produces a basic emotion called fear. In contrast to Russell, however, he believes that the unity of an emotion is not something that emerges on the fly when the proper conditions occur. According to Scarantino, whole emotional episodes are “designed” in an evolutionary sense and are therefore the outcome of selected, albeit complex mechanisms.

15

16

Peter Zachar

He concludes by offering a path of reconciliation between Panksepp’s and Russell’s views. On the issue of whether emotions as objects of scientific study can be considered to be the result of coordinating infrastructures, he sides with Panksepp and suggests that Russell could do so as well without giving up on his theoretical and philosophical commitments. In order for this reconciliation to proceed, however, Panksepp would have to abandon his view that primary process systems are fixed structures and start thinking of them along the lines of non-essentialist property clusters in Boyd’s sense. Scarantino is (at least) as tough on Panksepp. Although he questions Russell’s committed opposition to explaining emotional episodes with respect to coordinating affect programs in the head, his suggestion that Panksepp’s own understanding of mechanistic models may be scientifically inadequate might be a more damaging critique.

Paul Davies Our heart has reasons that reason knows nothing of is a phrase often attributed to Pascal. It is regularly used as a pithy introduction to psychodynamic views about our ignorance of the true causes of our behavior. Panksepp’s hostility to radical behaviorism has long been paired with an implicit affection for psychodynamic notions. In his tightly crafted chapter, Paul Davies defends this aspect of Panksepp’s thinking, albeit by trading the psychodynamic unconscious for the cognitive unconscious of recent scientific psychology. Davies notes that Panksepp’s experimental research provides compelling evidence for the existence of distinct affective systems reaching down to the subcortical level. When considered in an evolutionary context, these homologous affective systems can be said to constitute the foundation of our emotional lives. Also important is the fact that mammals can have coherent emotional lives without cortical structures, but not without subcortical structures. Furthermore, claims Davies, any causal role played by non-primary emotions supervenes on the casual efficacy of subcortically-grounded systems. For these reasons, he holds that an account of our emotional lives that is based primarily on reports of experiences and reflections on those experiences does not contain enough information to make valid inferences about the nature of emotion For example, knowledge about the underlying mechanisms that produce emotions is not amenable to discovery via introspection. It is unlikely that we will be able to develop plausible theories about the functions of emotions without understanding these mechanisms. Davies notes that this viewpoint is also consistent with those theories in social psychology that hold that the reasons for and causes of our own behaviors are opaque to us, and that our own ideas about those reasons and causes are often mistaken.



1.  Introduction

Having established his ground rules, Davies turns a critical eye toward ­ ussell’s psychological construction model. According to Davies, Russell’s claims R about the necessity of emotional prototypes and emotion scripts for both having emotions and recognizing ourselves as being in a particular state of emotion means that the existence of any emotion is relative to and contingent upon the bearer’s culture (i.e., scripts are cultural products). He correctly notes that if Russell’s theory necessitates accepting such radical social constructionism, then it cannot be reconciled with Panksepp’s view that emotions are natural kinds that “actually exist within the brains of newborn animals before they become modified by individual experiences” (p. 138). Davies’ first substantive criticism of Russell goes to the heart of much recent work in scientific psychology – not only on emotion, but also on personality and psychiatric disorders in general. His criticism is that psychometric studies using factor analysis and related techniques depend on self-report (reflections on experience), a category of evidence that was earlier argued in the chapter to be scientifically inadequate. The patterns revealed by these techniques do not tell us what things are really like, but only what we perceive them to be like according to the traditions of our cultures. In a second substantive criticism, he questions Russell’s canvassing of findings from neuroscience that are consistent with core affect theory. According to Davies, one could plausibly claim that Russell’s circumplex is a useful abstract model for thinking about the structure of the affects across Panksepp’s subcortical systems, and therefore the supporting neuroscientific evidence does not discriminate in favor of Russell over Panksepp. In contrast, Davies claims that the evidence adduced by Panksepp does discriminate in favor of his model. He then concludes that Panksepp’s categorical model of affect should be accepted and ­Russell’s dimensional model rejected. Davies closes with a general criticism of the social constructionist position which he attributes to Russell. The problem is that without empirically-based knowledge about the nature of primary emotional systems, any theory about non-primary systems or about the relationship between the primary and the non-primary can be little more than speculation. This critique applies to both social constructionist and cognitive theories that distinguish between different kinds of emotions (anger, guilt, shame, etc). Furthermore, the farther away we get from a more secure biological grounding, notes Davies, the more speculative (and “imagined”) our theories must be. In this category he places speculative ideas about emotional scripts and what counts as a genuine blue ribbon emotion.

17

18

Peter Zachar

Meaghan Leddy, Gail Robertson, and Jay Schulkin Leddy, Robertson and Schulkin offer a functionalist account of affect that has strong echoes of John Dewey’s famous The Reflex Arc Concept in Psychology article. Instead of construing affect as a fundamental particle in the emotional universe, they argue that its adaptive nature requires that it be fluidly interconnected, even continuous with other so-called psychological elements – especially cognition. Although their perspective could be considered to conflict with Russell’s description of core affect as a psychological element, on the whole Leddy, ­Robertson and Schulkin favor Russell’s dimensional model of affect more than they do Panksepp’s discrete model. Like Russell, they construe affect as a constant background against which all other emotional phenomena occur, but they are not so atomistic as to reduce affect to being a single state of the body at any one moment. Given its close connection with cognitive appraisal, their view of affect is more psychological (where “primary” and secondary “levels” act in concert – so cannot be neatly separated). One consequence of this stream of affect-cognition, they hint, is that someone can occupy different affective states simultaneously (like different programs running on the same computer). Interestingly, neither do they construe cognition monolithically as one kind of thing that can be decontextualized. In this spirit, the chapter concludes by going outside the head, and examining the role of the cultural context in better understanding the stream of affect-cognition. John Bickle John Bickle has long been among philosophy’s most enthusiastic proponents of the reductionistic and eliminativist arguments popularized by Paul and Patricia Churchland in the 1980s. His excitement about the possibilities of a ruthless re­ ductionism has always been so genuine that even those who do not agree with him can appreciate the emotional power of his commitments. What is interesting about his current chapter is that it betokens some reflection about possibilities not turning out as expected, resulting in a still palpable enthusiasm, but one that is tempered by a weariness of promises, proclamations, and traditional philosophical projects. He begins his chapter agreeing with the eliminativist tone to both Panksepp and Russell’s perspectives – specifically their mistrust of using vernacular language and folk psychological concepts as technical scientific terms. He criticizes the original program of eliminative materialism for being transformed into a question about the ontological status of folk psychological concepts such as beliefs and desires (i.e., are they real or are they fictions?). Rather than getting caught up on perennial issues in the philosophy of mind or the philosophy of science, what eliminative materialism should have attempted, says Bickle, was an



1.  Introduction

analysis of why folk psychological concepts have no role to play in the discovery of causal mechanistic explanations. He names this more deflationary view of the eliminativist project as eliminative materialism with a little ‘e.’ It is also a work in metascience, not philosophical analysis. Bickle proceeds to describe his work with neurobiologist Alcino Silva and Bickle’ former student Anthony Landreth in specifying the kinds of information that is sufficient for justifying inferences about causal mechanisms in the field of molecular and cellular cognition (MCC). Four categories of information, which they call the Convergent Four, are needed. The first two types of information (positive manipulation and negative manipulation) are interventionist. One intervenes to increase or decrease the probability of the cause (such as adding or deleting genes) and then examines what effect the manipulation has on the dependent variable (in this case behavioral measures of cognition). The third type of information is measures of the correlational relationships between genes (or gene products such as proteins and enzymes) and cognitive measures (which are necessary but not sufficient for causality). The fourth type of information, called integration, examines a plurality of data from the first three categories to develop models of the causal structures of the phenomena being studied. From this perspective, Bickle comments on Panksepp’s and Russell’s own answers to Questions 4, 7, 9 and 11. With respect to Question 4 (How can we infer affect from observations of behavior), he defends the use of behavioral measures of cognitive functions (such as spatial memory) in animal research. He does not claim that an operationalization is identical with the cognitive construct as did the early behaviorists, but insists that using behavioral measures of cognitive and affective phenomena are the only way scientific work on causal mechanistic explanations can proceed. In Paul Meehl’s terms, these measures are “fallible” indicators of the constructs in question – but indicators all the same. In this respect Bickle is in agreement with Panksepp’s insistence on the importance of animal research – but, he emphasizes, for the purposes of developing causal mechanistic explanations. With respect to Russell’s use of factor analysis to uncover the phenomenological structure of affect, Bickle does not dismiss it, but contends that it is a different kind of scientific project. Question 7 asks: What is the relationship between affect and cognition (con­ ceptualization, intentionality, appraisal, etc); Bickle expresses some skepticism regarding Russell’s claims about core affect being encapsulated from cognition along the lines suggested by the modularity model. His reason for being skeptical is the lack of adequate scientific work of the causal mechanistic sort to address this question. With respect to the primary versus tertiary distinction that is key to Panksepp’s chapters, Bickle suggests that if there were a qualitative difference between kinds of human affect and the affect of other animals (such as rodents),

19

20 Peter Zachar

then some very clever research designs will be needed to study these phenomena from a causal mechanistic perspective – and no one has yet figured how to do that. Tertiary, verbal phenomena are not the kind of thing that reductionistic science can yet address. On Question 9 (What role do the biological sciences and psychological sciences have in studying affect), Bickle answers from his ruthless reductionistic standpoint. What he says is that experimental psychology will always be needed to operationalize cognitive and affective functions into reliable and valid behavioral measures. He disagrees with Russell’s claim that psychology and biology are “complementary” primarily because he states that the information needed to obtain a more scientifically valid understanding of emotion concepts will come from the biological and molecular data. He is in general agreement with Panksepp on this issue, but finds Panksepp’s use of philosophy of mind terms such as “emergence” to be scientifically irrelevant. Finally, the ruthless reductionist believes that molecular biology’s role in discovering causal-mechanistic explanations could potentially be supplanted by nanotechnology and biophysical engineering – but those two methodologies will still depend on experimental psychology. Bickle closes with some thoughts on Question 11 (Considering the diversity of theoretical viewpoints in the scientific study of affective and emotional phenom­ ena, might one hope for a coherent synthesis in this splintered field?). According to Bickle, it is a myth that interdisciplinary work characterizes cognitive science. The problem is that the participating research traditions are Balkanized and tend to dismiss those who use methods or levels of analysis different from their own. Usually the call for being interdisciplinary means “The other guy should start seeing things from my perspective.” There is, he implies, a little bit of that attitude in the interaction between Panksepp and Russell. Bickle believes that too much work on levels and interaction between levels gets sidetracked into metaphysical speculation, and that scientists are too willing to write about the interdisciplinary relevance of discoveries about things such as mirror neurons, even if the relevant evidence of the causal mechanistic sort that would form a foundation for the integration of levels is totally absent. Bickle, like Panksepp, refuses to accept that we can develop adequate causal mechanistic understanding without taking the lower levels into account (including levels lower than Panksepp prefers himself). The need to consider upper levels in understanding cognitive and affective functions will become evident once we consider information from the lower levels, but Bickle believes that an impediment to this realization is a Wimsatt-influenced view about explanatory pluralism (i.e., the view that explanations at different levels of analysis may not be reducible to one another, but still do good explanatory work).



1.  Introduction

James Averill Jim Averill takes both Panksepp and Russell to task for what he considers their dismissal of higher levels of analysis (and everyday experience) in favor of their own preferred research programs. Averill does not dispute the value of these research programs, but he is unwilling to assign either one of them ontological priority. They are scientifically important, but not scientifically privileged. The study of the “emotions” to which folk concepts refer, writes Averill, is also scientifically legitimate (and rigorous). Such a research program is more than “hermeneutic.” It is a proper scientific psychology, not a study in arts and humanities or cultural anthropology. The core of his chapter is spent explicitly justifying this claim. Averill questions both Pankseep’s and Russell’s acceptances of traditional empiricist/positivist dogmas seen in each, implying that their respective approaches have unmediated connections to observation and evidence. Averill’s argument amounts to a claim that everyday emotional concepts are necessary for framing what we study, for organizing the evidence, and for thinking about its relevance. Consider, he says, the concept of species. Darwin came to believe that “species” was not a real category in nature, but the species category inspired Darwin’s study of evolution and remains crucial for understanding its relevance. Likewise, Averill believes that the folk concept of “emotion” can also be used to inspire a scientific research program and understand its relevance. Even if the category of species does not refer to a coherent natural kind, ­Darwin never doubted that lions and tigers and bears are real. Averill makes a similar argument about the category of emotion; states such as happiness, guilt, and romantic love are real, although, like species, not necessarily universal across time or place. One key issue for both Panksepp and Russell is whether everyday emotional concepts can be used to explain behavior. Although he does not put it this way, Averill adopts a pragmatist’s view that a good explanation is a satisfactory answer to a “why” question. A satisfactory answer can, but does not need to specify casual mechanisms, at least not in the overly restrictive sense of “efficient cause.” In a broader sense, a causal explanation simply indicates why X happened rather than Y happened. In this broader sense, personal dispositions and social roles can be considered “causes.” As a social constructionist, Averill emphasizes how emotional behavior and experience vary across cultures. These complex social roles/dispositions are not fictions, they are subject to scientific study, and they have real effects on behavior. Anger is a real social role for people in the U.S and liget is a real social role for the Ilingot, a head-hunting people in the Philippines, but neither is a real social role for the Utku, an arctic tribe of Eskimos. Depending on what you want an

21

22

Peter Zachar

e­ xplanation to do, it may serve practical needs to say that the American male stomped out of the room when his wife yelled at him rather than kissing her because he was angry, but that answer is not available for explaining Eskimo behavior. Similar considerations apply to theoretical concerns, which cannot be divorced altogether from practical needs. Concluding with some pragmatist musings about realism and anti-realism, Averill makes the Peirceian point that some of our “instincts” (and related folk concepts) have to at least approximate reality, or we would not have survived our long evolution. This approximated reality includes the social reality that differs from culture to culture and epoch to epoch. Congruent with both James’ and Rorty’s distrust of absolutes, he suggests that any retreat to universal and nomothetic truths has to be so abstracted away from sociocultural reality, that to some extent the supposed truths are themselves fictions. Lab experiments and questionnaires are important tools for research but, as the old saw goes, when all you have is a hammer, the whole world looks like a nail. The dismissal of the scientific study of real, lived experience in favor of lab experiments and questionnaires, Averill suggests, is also a dismissal of what, ultimately, emotion theories should be about.

Ronald de Sousa In a philosophically rich chapter, Ronnie de Sousa examines the role of valence, reductionism, and the ineffable in light of the debate between Panksepp and Russell. On the issue of valence, he expresses some suspicions about core affect theory. He notes that the core affect model constitutes an uneasy mix of direct experiences and of theoretical constructs. In terms of its theoretical formulation, de Sousa is more concerned about the dimension of valence than of arousal. A key issue is whether valence (pleasantness and unpleasantness) is intrinsic to the subjective quality of an experience such as the taste of caviar or whether it is distinct and separable. In other words, can two people have the same experience of the taste of caviar, but one find that experience pleasurable and the other unpleasurable – or is valence intrinsic to the experience of taste itself so that a pleasurable experience of caviar must taste different from an unpleasurable experience of caviar? De Sousa notes that this is a question on which philosophers should defer to the empirical evidence. He is willing to go along with Russell that pleasure/displeasure is a psychological atom out of which other experiences are constructed, but is not convinced that these affective experiences have fixed behavioral correlates such as approach for pleasure and avoidance for displeasure. On the issue of reductionism, de Sousa sides with the reductionist perspective because he sees it as integral to the scientific enterprise. He has little sympathy for anti-reductionist claims which fundamentally state that “to learn MORE



1.  Introduction

about a phenomenon by discovering how it arises from underlying components results in LESS understanding.” This view, he suggests, is simply illogical. Echoing James’ distinction between the tender-minded and tough-minded, he notes that a preference for reductionism may be a matter of philosophical temperament. In this respect, he explores the different ideas Panksepp and Russell have about what counts as bottom-up. By bottom-up Panksepp means beginning with homologous biological structures, whereas Russell means beginning with simple psychological primitives. De Sousa is sympathetic to Panksepp’s notion of primary process affect programs, but hints that Russell’s dimensions of valence and arousal could be considered to be more primitive/simple with respect to affective subjectivity. Even if Russell’s core affect is the affective primitive, de Sousa believes that there exist biologically-influenced emotional kinds and that these “emotion categories/lumps” have more ontological weight than would be recognized by Russell’s splitting strategy (which emphasizes the existence of thousands and thousands of particular states of emotion as opposed to types/kinds). At the very least, it is informative to use a reductionist program to study these categories. In addition to Pankseep’s primary, secondary, and tertiary distinction, de Sousa notes the potential importance of Marr’s distinction between computational, algorithmic and implementation levels, and Dennett’s intentional stance, design stance, and physical stance. De Sousa also spends some time exploring the paradoxical issue of affect being prelinguistic and primitive even though we can only talk about such affect using higher brain functions. This raises the question of whether the experience of emotion is ineffable. It is not a simple question. Clearly language can refer to any emotional state. Although trying to filter affect through language might have a blunting effect, language can also help us better attend to certain aspects of our experience and contribute to its enhancement. This issue is related to the question of the difference between affect we attend to versus that we do not attend to. Is the latter an instance of “unconscious affect?” De Sousa believes that the notion of unconscious affect has some validity, but also that it raises problems for core affect theory. The basic problem is that if there can be some aspects of our current background affect that we attend to and others that we do not, this suggests that affect is not just one thing/state of the body as core affect theory suggests.

Georg Northoff Georg Northoff takes on the challenging task of attempting to find common ground between Panksepp and Russell by exploring the neural predisposition to affect, or what he calls a basic affect unit. By “predisposition” Northoff means not substrates that are sufficient for producing affect, but functional transformations that are necessary for the possibility of affect (in a Kantian sense). He is

23

24

Peter Zachar

­ articularly concerned with the private, subjective, and conscious nature of affect. p The transformations that are necessary for privacy, subjectivity and consciousness include the double transformation of non-affective into affective stimuli, and of objective into subjective stimuli. He presents an interesting theory of how an individual brain’s resting state activity interacts with different stimuli to achieve such transformations. His speculations in this regard provide some possible neural mechanisms for explaining Russell’s characterization of Core Affect as a continuous background state of the body in contrast to Panksepp’s model of discrete affects. In the latter half of the chapter, Northoff addresses the issue of why it is that the subcortical regions are those responsible for affect, and what it is that enables them to transform non-conscious into conscious states. In the first case, he notes that subcortical regions are more attuned to internal stimuli than are cortical regions. In the second case his argument is fairly complex. That is, once internal and external stimuli are neurally processed, the processing of the interactions between those stimuli leave us unable to process them as neural states – we have information only about relations. When these relations are processed relative to the various resting states of the brain (and this is a private, individual, and subjective event), the possibility for consciousness arises.

Agnes Moors William James contended that our initial experience of the world is a blooming, buzzing confusion. The confusion is reduced by (a) distinguishing between different aspects of experience, (b) naming things, and (c) freezing what is a flux into relatively fixed categories. Scientific categorization can be seen as an elaboration of this same process. In her closely argued chapter, Agnes Moors analyzes the different ways in which affect program theory, appraisal theory, and the psychological construction approach categorize the blooming buzzing confusion of emotional phenomena. Her analysis begins by offering a model for comparing and contrasting the three theories. The first part of the model looks at the basic components of emotional phenomena, specifically the somatic, cognitive, motivational, behavioral, and subjective components. The three theories generally agree on the menu of components, but they weigh them differently. For example, emphasizing the causal importance of central processes such as affect programs (largely a somatic process) versus emphasizing the effects of a family of components none of which are essential for emotion leads to radically different taxonomies of possible emotions. A second aspect of the model examines how the theories differentiate emotional from non-emotional states. To a large extent this follows from how the components are weighted. One reason the psychological construction approach



1.  Introduction

does not advocate a firm distinction between emotional and non-emotional phenomena is that it refuses to weigh any component as more essential to emotion than another. A third aspect of the model examines how the theories split the broad category of emotion into subtypes (what Moors called extensional definitions). As noted, how the components are weighted influences how emotions are divided up. For instance, affect program theory’s emphasis on the somatic components leads to a basic versus non-basic distinction – which is viewed with some suspicion by appraisal theories and the psychological construction approach. Moors then proceeds with a detailed analysis of the differences between the three theories with respect to their views of causation. All three theories agree that emotional reactions are elicited by some kind of a stimulus and processed in the organism, but beyond that differences emerge. For example, prototypical affect program theories emphasize the somatic component, particularly the role of neural circuits in explaining the occurrence of the other components. These theories tend to locate the causal force at Marr’s implementation level (i.e., neurological processes). Panksepp’s complex cognitive-affective circuits, she notes, are also described at Marr’s algorithmic level (i.e., a specification of underlying mechanisms). In contrast, appraisal theories use the cognitive component to explain the occurrence of the other components and emphasize Marr’s functional level (which takes into account events outside the head). Appraisal is not a unidimensional construct; rather, it can be divided into subtypes such as assessment of goal relevance, goal congruence, and coping potential. In one form of appraisal theory (Lazarus), these subtypes combine to form a global pattern of appraisal that subsequently influences the occurrence of the other components of emotion (via an emergent affect program). In a second “decompositional” form of appraisal theory (Scherer) each appraisal subtype is causally related to the components of emotion in different ways. Like Scherer’s version of appraisal theory, the psychological construction approach is inconsistent with affect program theory. In psychological construction, neither the neural/somatic nor the cognitive component is used to account for the other components. Instead, each component is explained on its own terms (with reference to underlying mechanisms). Moors expresses some skepticism about what she sees as Russell’s failure to specify causal relations among components. She suggests that the substrates of any component may be independent of the substrates of the other components, but the components themselves could still be causally related to each other. Moors also compares the three theories with respect to what kinds of affective experiences and associated behaviors they account for. There are three types. Type 1 consists of the affective and behavioral dimensions of ­pleasantness/

25

26 Peter Zachar

approach­ versus unpleasantness/avoidance. Type 2 is the affect and behavior associated with basic emotions such as anger and sadness. Type 3 is the affect and behavior associated with higher order emotions such as jealousy. Considering this scheme, she notes that the affect program theories are the most limited, conceptually. They tend to account for Type 2 affects in a hard-wired fashion, and account for Type 3 affects as derivative from Type 2 affects. The theory of psychological construction can potentially account for all three types of affectbehavior, as can some appraisal theories (those that do not involve central affect programs). To the extent that they are less likely to construe affect and behavior as hard-wired, these latter two theories have the conceptual resources to explain a wider variety of emotional states. The chapter concludes with a careful look at the kinds of experimental evidence (direct and indirect) that would be needed to validate one theory relative to the others. This is made more complicated by the fact that the theories overlap in a variety of ways. According to Moors, all of the following are experimentally testable: hypotheses that components are caused by a single mechanism, central process; whether particular causal processes are specific to subsets of emotions; the relation between causal processes and outputs; and whether these processes are hardwired or learned. Some problems with even these tests, she suggests, derive from the fact that they cannot occur without some background assumptions about what counts as an emotion, or what counts as a single mechanism. These background assumptions are, in part, what differentiates the three theories discussed in this chapter. Although progress can be made in distinguishing between the merits of the different theories, no set of experimental tests can ever definitively, on their own, validate one theory type rather than another.

Russell’s and Panksepp’s responses to the commentaries In his closing remarks, Russell clarifies that he has proposed two separate theories, specifically, core affect theory and the theory of psychological constructions, and he addresses some mischaracterizations of them by some commentators. He notes areas of agreement with each commentator, but on some key issues he dissents. Russell particularly questions Davies’ literalist reading of Panksepp’s categories and also Davies’ dismissal of the validity of evidence based on self-report. He maintains his eliminativism about basic emotion categories despite Averill’s defense of the scientific relevance of folk concepts and Scarantino’s Boyd-inspired rethinking of natural kinds. A final key issue is Russell’s re-assertion of the simplicity of pleasure against several claims – especially de Sousa’s – that pleasure can be further decomposed, psychologically. Russell agrees that there are different kinds of pleasurable experiences, but believes that those kinds are more, not less complex than the valence dimension of core affect theory.



1.  Introduction

Pankseep’s closing remarks are largely conciliatory, highlighting areas of agreement with each commentator. He also summarizes his own position and emphasizes the importance of developing, testable, evidence-based theories of affect. He also addresses some key disagreements. With respect to Scarantino, Panksepp acknowledges the problem of using vernacular terms to describe events at the subcortical levels of analysis, but asserts the practical reason for doing so. Specifically, using language with vernacular roots draws attention to the connection between lower and higher levels of analysis, which is especially important in understanding the biological basis of psychiatric disorders. He offers another version of this claim in addressing John Bickle’s reductionism. Also in his commentary on Scarantino, Panksepp states that one can posit coherent neural systems without advocating essentialism – and he claims that he is not an essentialist. In his response to Davies he again notes the existence of variation within each of his subcortical circuits, thereby also addressing this aspect of Russell’s critique by suggesting that such a critique mischaracterizes Panksepp’s own position.

Ellis and Faw’s concluding chapter The final chapter by Ellis and Faw serves as an additional commentary that takes all six chapters by Panksepp and Russell into account. No other chapter in this book provides a more detailed analysis and side by side comparison of Panksepp and Russell’s perspectives. It also includes an assessment of their respective strengths and weaknesses from the perspective of current thinking in biological psychology and affective science. Adopting scientific realism as a regulative ideal, Ellis and Faw ask whether Panksepp’s and Russell’s different theories end up in the same place (triangulate on the same reality). The claim in the first part of the chapter is that they do not, nor will they unless they aim for each other in a co-evolutionary sense. Bill Faw’s background in biological psychology makes Panksepp’s theory more preferable to him in that regard, but nowhere is it suggested that Panksepp’s model is literally true. It is comparatively better given the standards of biological psychology. Of course Russell is not claiming to be doing biological psychology, so those standards of comparison are not the only standards one might select. Another standard of comparison is introduced in the second part of the chapter. As a bit of background, let me note that Ralph Ellis is a phenomenologist who has long been developing a theory of consciousness based on the integration of phenomenology with scientific research, especially neuroscientific research. From Ellis’s standpoint, an essential psychological reality is to be found at this higher level of analysis and revealed by systematic introspection guided by the discipline of careful phenomenological self-questioning, which can guide us away from common presuppositions about the aims, objects, and intentional meanings

27

28

Peter Zachar

of emotional experiences. Like Panksepp, Ellis rejects the behaviorist reflex model and also construes consciousness as more than an appendage to neurophysiological activity. Panksepp’s inferences about the fundamental role played by consciousness throughout the mammalian line are very phenomenology-friendly. Interestingly, Ellis argues that theory of psychological construction could be seen as also consistent with phenomenological analyses, especially the phenomenologist’s emphasis on intentionality as a constraint on the reduction of consciousness to something simpler and automatic. Another way he might put it is that someone who is skilled at constructing emotional meta-experiences would be very psychologically-minded along phenomenological lines. Given that Russell believes that a good theory about emotional phenomena would be a theory of psychology in general, its ability to accommodate a phenomenological analysis is not surprising. How Russell’s nominalistic and eliminativist predilections could be integrated with a more essentialist version of phenomenology remains unanswered. At the end of the day, with scientific realism as our regulative ideal, all cannot have prizes. If the real, as Peirce says, is the referent of those opinions that are fated to be ultimately agreed on by those who investigate, then some parties to a disagreement must be mistaken. A nominalist would contend that all theories are partial and perspectival, but even there, scientific values ask us to try to classify the world as it is, not how we want it to be. In being guided by this ideal, we must also guard against premature foreclosure. If scientific realism it must be, then let us be partial and pluralistic realists and allow that a select number of different theories can have prizes for today.

Acknowledgement Thanks to Jaak Pankseep and Jim Russell for patiently answering my questions about their intellectual biographies.

References AJP. (2010). Science of the brain as a gateway to understanding play. American Journal of Play, 2(2), 245–277. Barrett, L. F. (2006). Are emotions natural kinds? Perspectives on Psychological Science, 1(1), 28–58. Barrett, L. F., Lindquist, K. A., Bliss-Moreau, E., Duncan, S., Gendron, M., Mize, J., et al. (2007). Of mice and men: Natural kinds of emotions in the mamalian brain? A response to Panksepp and Izard. Perspectives in Psychological Science, 2, 297–312.



1.  Introduction

Bingman, V. P. In press. The maturation of affective neuroscience: Celebrating the research career of Jaak Panksepp. Neuroscience and Biobehavioral Reviews. Bowlby, J. (1969). Attachment and loss (Vol. 1). New York: Basic Books. Dennett, D. C. (1991). Consciousness explained. Boston: Little Brown. Desmond, A. J., & Moore, J. R. (1991). Darwin: The life of a tormented evolutionist. New York: Warner Books. Dollard, J., & Miller, N. E. (1950). Personality and psychotherapy: An analysis in terms of learn­ ing, thinking and culture. New York: McGraw-Hill. Fehr, B., & Russell, J. A. (1984). Concept of emotion viewed from a prototype perspective. Jour­ nal of Experimental Psychology: General, 113, 464–468. Gregory, F. (2008). Natural science in western history. Boston, MA: Houghton Mifflin Company. Hess, W. R. (1957). The functional organization of the diencephalon. New York: Grune and Stratton. MacLean, P. D. (1990). The triune brain in evolution. New York: Plenum Press. Mehrabian, A., & Russell, J. A. (1974). The basic emotional impact of environments. Percpetual and Motor Skills, 38, 283–301. Panksepp, J. (1974). Hypothalamic regulation of energy balance and feeding behavior. Federa­ tion Proceedings, 33, 1150–1165. Panksepp, J. (1981). Hypothalamic integration of behavior: Rewards, punishments, and related psychobiological processes. In P. J. Morgane & J. Panksepp (Eds.), Handbook of the hypo­ thalamus: vol 3, part A: Behavioral studies of the hypothalamus (pp. 289–487). New York: Marcel Dekker. Panksepp, J. (1982). Toward a general psychobiological theory of emotions. Behavioral and Brain Sciences, 5, 407–467. Panksepp, J. (1986). The anatomy of emotions. In R. Plutchik (Ed.), Emotion: Theory, research, and experience: vol III: Biological foundations of emotion (pp. 91–124). Orlando: Academic Press. Panksepp, J. (1991). Affective neuroscience: A conceptual framework for the neurobiological study of emotions. In K. Strongman (Ed.), International reviews of emotion research (pp. 59–99). Chichester, England: Wiley. Panksepp, J. (1998). Affective neuroscience: The foundations of human and animal emotions. New York: Oxford University Press. Panksepp, J. (2008). Carving ‘natural’ emotions: ‘Kindly’ from bottom-up but not top-down. Journal of Theoretical and Philosophical Psychology, 28(2), 395–422. Panksepp, J., & Biven, L. In press. Archaeology of mind: Neuroevolutionary origins of human emotions. New York: Norton. Quine, W. V. (1969). Ontological relativity and other essays. New York: Columbia University Press. Rosch, E. (1976). Basic objects in natural categories. Cognitive Psychology, 8(3), 382–439. Russell, J. A. (1980). A circumplex model of affect. Journal of Personality and Social Psychology, 39, 1161–1178. Russell, J. A. (1983a). Dimensions underlying children’s emotions concepts. Developmental Psy­ chology, 19, 795–804. Russell, J. A. (1983b). Pancultural aspects of the human conceptual organization of emotions. Journal of Personality and Social Psychology, 45, 1281–1288. Russell, J. A. (1991). Culture and the Categorization of emotions. Psychological Bulletin, 110, 426–450.

29

30

Peter Zachar

Russell, J. A. (1994). Is there universal recognition of emotion from facial expression?: A review of the cross cultural studies. Psychological Bulletin, 115, 102–141. Russell, J. A. (1995). Facial expressions of emotion: What lies beyond minimal universality. Psychological Bulletin, 118, 379–391. Russell, J. A. (2003). Core affect and the psychological construction of emotion. Psychological Review, 110(1), 145–172. Russell, J. A. (2008). In defense of a psychological constructionist account of emotion: Reply to Zachar. Journal of Theoretical and Philosophical Psychology, 28(2), 423–429. Russell, J. A., & Barrett, L. F. (1999). Core affect, prototypical emotional episodes, and other things call emotion: Dissecting the elephant. Journal of Personality and Social Psychology, 76, 805–819. Russell, J. A., & Fehr, B. (1988). Reply to Ekman and O’Sullivan. Journal of Experimental Psy­ chology: General, 17, 89–90. Russell, J. A., & Mehrabian, A. (1977). Evidence for a three-factor theory of emotions. Journal of Research in Personality, 11, 273–294. Schaffner, K. F. (1993). Discovery and explanation in biology and medicine. Chicago: University of Chicago Press. Schaffner, K. F. In press. A philosophical overview of the problems of validity for psychiatric disorders. In K. S. Kendler & J. Parnas (Eds.), Philosophical issues in psychiatry II: Nosologydefinition of illness, history, validity, and prospects. New York: Oxford University Press. Sundararajan, L. (2008). It’s turtles all the way down: A semiotic perspective on the basic emotions debate. Journal of Theoretical and Philosophical Psychology, 28, 430–443. Trowill, J. A., Panksepp, J., & Gandelman, R. (1969). An incentive model of rewarding brain stimulation. Psychological Review, 76, 264–281. Wiggins, J. S. (1979). A psychological taxonomy of trait-descriptive terms: the interpersonal domain. Journal of Personality and Social Psychology, 37, 395–402. Zachar, P. (2000). Psychiatric disorders are not natural kinds. Philosophy, Psychiatry, & Psychol­ ogy, 7(3), 167–182. Zachar, P. (2006). The classification of emotion and scientific realism. Journal of Theoretical and Philosophical Psychology, 26(1–2), 120–138. Zachar, P. (2008). A triptych on affective science: Response to the commentary. Journal of Theo­ retical and Philosophical Psychology, 28(2), 444–453. Zachar, P. In press. Progress and the calibration of scientific constructs: the role of comparative validity. In K. S. Kendler & J. Parnas (Eds.), Philosophical issues in psychiatry II: Nosologydefinition of illness, history, validity, and prospects. New York: Oxford University Press.

chapter 2

In defense of multiple Core Affects Jaak Panksepp

Emotion studies would not be as important as they are if affective experience was not one of the defining attributes of emotionality. Since felt experience is the essence of consciousness, the nature of affects lies at the very core of the mind-matter dilemma that has haunted experimental psychology since its inception – leading on the one hand to the pseudo-solution of “behaviorism,” and the modern approximations of “cognitivism” on the other. However, no aspect of consciousness can be scientifically observed directly through any external measure yet imagined. This remains an embarrassment to science, a highly successful human endeavor where objective facts have priority over subjective claims. Affect studies now lie at the very fulcrum of the long-standing debate between objective-scientific­ and introspectivehumanistic understandings of our nature as living, feeling organisms. Modern physics has helped clarify how scientific approaches can penetrate under external appearances to reveal deeper levels of reality. Some of us feel that our understanding of nature has progressed to a point where certain experiences of organisms are finally amenable to objective scrutiny. My research, since tenure, has been devoted toward the possibility that the foundations of raw emotional experience in humans are best understood by studying the corresponding brain processes in our fellow animals. Such work may eventually provide the first causal decoding of what it means, in brain terms, to have internal experiences. My goal in implementing animal models for understanding the nature of emotional affects was clarification of the brain-mind sources of psychiatric disorders in humans. Psychiatry, right or wrong, has been so far premised on the likelihood that human minds are subject to many distinct types of affective imbalances. I believe the animal work is affirming the fertility of that intuition. As I reflect on the questions posed by the editors, my overall conclusion is that at the foundation of mammalian mental life, there exist neural potentials for ­diverse primary-process­ affective feelings – a variety of Core Affects – which are profoundly important tools for living as well as for the emergence of many higher mental abilities. I hope to understand how Jim Russell’s Core Affect view may further illuminate this ­evidence-based view of mental life.

32

Jaak Panksepp

1. What is affect? What is its structure? Is affect fundamental, and if so, what makes it so? Affects are feelings that guide our thoughts and our actions. Primary-process affects are intrinsic (unconditioned) neuropsychological functions of the brain, responsive initially to only limited sets of environmental events (i.e., unconditional stimuli). Psychologically, they are characterized by distinctly valenced feelings accompanied by generalized arousal and varying degrees of power (surgency) in the mind. In other words, diverse feelings – some “good” and some “bad” and perhaps some that can become ambiguous (e.g., anger) because of variable outcomes – are psychological compasses for our lives. For systematic progress in understanding the hierarchical organization of the affective BrainMind, it is important and useful, at minimum, to subdivide affective-emotional life into at least three levels of organization: i. Primary-process emotional networks are those that generate a variety of evolutionarily provided raw emotional-affects (for which, I believe, the term “basic” should be reserved). ii. Secondary-processes reflect how primary-process affects provide the fuel for additional complexities arising from basic learning mechanisms such as classical and instrumental conditioning. iii. Tertiary-process affects, so important for cultural issues, reflect the highly nuanced feelings that emerge, at least in part, from the previous levels of control interacting with our highest cognitive processes – perceiving, thinking, ruminating, fantasizing, etc. in the context of each person’s storehouse of autobiographical-cultural memories. At lower MindBrain levels, we are bound to be closer to the nomothetic (universal) processes laid down early in mammalian evolution. As we move further up in complexity, there is bound to be ever more idiographic (individualistic) differentiation of affects, especially since language is well positioned to make fine and metaphorical, often arbitrary, discriminations. So here is a synopsis of my primary-process perspective. At the beginning of life, affects are fundamental (primary-process, basic, raw “natural-kinds”). As far as we know, their initial mental forms and functions were heavily dictated by evolutionary selection yielding BrainMind functions that appeared earliest in neuromental evolution. Admittedly, they are poorly understood and rarely expressed by adult humans in full fidelity (except perhaps in the arts). They are ancient in mammalian brain evolution since the decisively important neural networks are densely represented in relatively caudal and medial subcortical areas of the brain, from regions that are clearly among the most ancient in the brain. Brain evolution



2.  In defense of multiple Core Affects

seems to have retained those ancient solutions in homologous forms, and they remain major operating systems for mental existence. Primary-process views, going back to Charles Darwin (1872/1965), are well supported by cross-species neuroscience. However, because affects penetrate so many aspects of our mental apparatus there is now a semantic-conceptual jungle of complexity to be negotiated before neuroscientific views can become an optimal frame of reference for understanding primary mind processes in psychology and philosophy. This might be an easier process if levels of analysis were not so conflated in emotion studies. In my estimation, the common failure of investigators to be clear about their own level of analysis, as contrasted with other levels, is the source of the abundant theoretical disagreements and chaos in this important field of inquiry. My own research is almost exclusively devoted to the primary-process nature of emotions, and I welcome a critique of discoveries and interpretations made at that level of analysis. For me, all implications for tertiary-process views are more inferential, but useful if they lead to concrete theoretical predictions. I have pointedly chosen not to use vernacular terms for primary-process emotional systems and their affects since all such words have variable meanings because of our socio-cultural diversity. By using a new semantic convention – capitalization of emotional/affective primes – I have tried to avoid those ambiguities and have made a strong case for the existence of at least seven fundamental emotional urges, each with distinct affects, within mammalian brain (the “Big 7” as noted below). In seeking to avoid confusions inherent in our vernacular words and also to avoid part-whole confusions (i.e., higher order psychological wholes in humans are surely constituted by more than ancestral emotional parts), I have used capitalizations as designation for these 7 systems (seeking, rage, fear, lust, care, grief/panic, and play). With these terms, I am referring to specifiable brain networks that are important parts of intrinsically ambiguous conceptual wholes. Obviously scientists can only study parts of complex wholes, so the capitalizations and the specific terms selected highlight not only which specific brain systems are critically important for various core affects, but also how they may relate to the psychological emotional wholes labeled by everyday human languages. This allows one to make many specific predictions in future higher-order psychological studies of how these systems influence human minds (Panksepp, 1999; Panksepp & Harro, 2004). There is a menagerie of affectively experienced primary-process BrainMind functions that we share with other animals because of our shared evolutionary heritage. To the best of our knowledge, the “Big 7” described in the next section are fundamental emotional substrates of mammalian brains, self-similar across mammalian species, in underlying principles, albeit not specific details.

33

34

Jaak Panksepp

The basic emotional-affective circuits of mammalian brains Empirical work using localized electrical and chemical stimulation of mammalian brains supports the existence of at least seven distinct types of emotional ­arousal – seven basic types of highly interactive emotional systems – in all mammalian brains (see Panksepp, 1998a and 2005a for fuller anatomical, neurochemical and behavioral descriptions of these systems). All are concentrated subcortically, consisting of extensive longitudinal neural networks interconnecting critically important midbrain regions such as the Periaqueductal Gray (PAG) and Ventral Tegmental Area (VTA), with various basal ganglia nuclei, such as nucleus accumbens and amygdala as well as higher limbic cortical regions such as the orbitofrontal, anterior cingulate, insular and medial frontal cortex regions, via pathways that course through the periventricular brain stem and lateral hypothalamus (e.g., most notably the Medial Forebrain Bundle (MFB)) and dorsomedial thalamus. Although much is known about these systems (Panksepp, 1982, 1998a 2005a), there is an enormous amount left to be learned. Each emotional system has abundant descending and ascending components, not yet well detailed (especially in humans), that presumably work together to coordinate various instinctual emotional behaviors as well as certain affective feelings (i.e., operationally defined in animals by the existence of various distinct types of “rewards” and “punishments”) normally associated with those behaviors. The primary-process experiential (i.e., affective) concomitants are much harder to study than the instinctual behaviors, but they appear to arise from the same neural networks. These systems are of critical importance for higher decisionmaking abilities. In mature mammalian brains, these systems can be aroused by higher brain mechanisms but they also regulate how higher brain functions process information. Presumably the accompanying affective states veridically signal specific types of survival issues. The following seven systems are supported by abundant evidence, and all emotion theories need to consider how they fit into their infrastructures: 1.  The SEEKING/Desire System. This general-purpose appetitive motivational system is essential for the many other emotional systems to operate effectively. It may be a major source of the life-energy that has at times been called “libido” – it generally enhances a broadly based positive affect that resembles what might be commonly called, from low to high arousal respectively, interest, enthusiasm, exhilaration and euphoria. It induces animals to be intensely engaged in exploring their world and with the automatic anticipatory learning that results, it gradually leads them to become excited when they are about to get what they desire. It eventually allows animals to find and eagerly anticipate all kinds of resources



2.  In defense of multiple Core Affects

they need for survival, including water, food, warmth or coolness, and the ultimate evolutionarily dictated mammalian social-emotional survival needs: sexual urgency, maternal care and social bonding. When fully aroused, seeking fills the mind with interest and motivates organisms to move their bodies seemingly effortlessly in search of the things they need, crave, and desire. In humans, this system generates and sustains curiosity from the mundane to our highest intellectual pursuits. When this system becomes under-active for various reasons, such as drug withdrawal, chronic helplessness/stress or neural deficits of old age, a form of depression results. When the system becomes spontaneously overactive, which can happen as a result of various drugs, thoughts and behaviors can become excessive and stereotyped, guided often by psychotic delusions and manic confabulations. Neuroanatomically, the seeking system corresponds to the major self-stimulation system that runs from the midbrain VTA up to the nucleus accumbens and medial frontal cortex via the MFB. Animals will activate this system readily, whether with drugs of abuse such as cocaine or direct electrical or chemical stimulation of the brain (after they have been surgically prepared with the necessary delivery devices, of course). This system has long been problematically conceptualized as “The” Brain Reward or Pleasure or Reinforcement system. Instead, it appears to be a general-purpose system that coaxes animals and humans to move energetically from where they are presently situated to the places where they can find and consume resources needed for survival. It permits learning by readily engendering predictive reward relationships in the world. A critically important chemical in this system is dopamine, but seeking regulation is also facilitated by neuropeptides such as endogenous opioids, neurotensin, oxytocin and orexin. Many of these peptides participate in several other basic-emotional and motivational processes, and hence could be envisioned as part of a generalized positive affect system. However, diverse forms of affective specificity may also be engendered by the specific brain sites of action of these brain chemistries. This system has more extensive cortical innervation in hominid brains than more simple-minded mammals. In addition to brain opioids, this is the closest we have to a generalized positive affect system in the brain. 2.  The RAGE/Anger System. Working in tandem, often in opposition, to seeking is a system that probably mediates what is commonly called “anger” in the vernacular. RAGE is aroused by frustration arising from any attempts to curtail an animal’s freedom of action. It has long been known that one can enrage both animals and humans by stimulating very specific circuits of the brain, ones that lie rather close to the trajectory of the fear system discussed next. The rage system invigorates aggressive behaviors when animals are irritated or restrained and also

35

36

Jaak Panksepp

helps animals defend themselves by arousing fear in their opponents. We predict that human anger gets much of its psychic energy from the arousal of this brain system; there are a number of well documented cases where humans stimulated in these brain regions have exhibited sudden, intense rage attacks, with no external provocation. Key chemistries in this system are Substance P and glutamate, which activate the system, and endogenous opioids, which inhibit the system, but these chemistries (especially glutamate and opioids) also participate in many other emotional responses, with specificity determined by the specific emotional systems in which they operate. 3.  The FEAR/Anxiety System. A coherently operating fear circuit was designed during brain evolution to help animals reduce pain and the possibility of destruction. When stimulated intensely, the circuit leads animals to flee as if they are extremely scared. With much weaker stimulation animals exhibit freezing, which is also common when animals are placed back into an environment in which they have been previously hurt or frightened. Humans stimulated in these same brain regions report being engulfed by an intense free-floating anxiety that appears to have no environmental cause. Key chemistries that regulate this system are Neuropeptide Y and Corticotrophin Releasing Factor (CRF); specific anti-anxiety agents such as the benzodiazepines inhibit this system. Modest anxiety in rats can be monitored through their emission of 22 kHz ultrasonic vocalizations (USVs), but intense fear (i.e., impending foot-shock) terminates these alarm calls. Animals “sigh” (exhibit a double-inspiration) when given safety signals in the midst of the signaled punishments commonly used in fear learning experiments (Soltysik & Jelen, 2005). Most investigators now agree that this system participates in various anxiety disorders. 4.  The LUST/Sexual System. Sexual urges are mediated by specific brain circuits and chemistries; they can be distinguished in males and females (for recent overview, see Wu, et al., 2009). They are aroused by male and female sex hormones, which control many brain chemistries including two neuropeptides whose synthesis is strongly controlled by sex hormones: Oxytocin transmission is promoted by estrogen in females and vasopressin transmission by testosterone in males. These brain chemistries help create gender specific sexual tendencies. Oxytocin promotes sexual readiness and acceptance postures in females and vasopressin promotes assertiveness, and perhaps jealous behaviors, in males. Distinct male and female sexual circuits are constructed very early in life, and are activated by maturation of gonadal hormones at puberty. Because brain and bodily sex characteristics are independently organized, it is possible for animals that are externally male to have female specific sexual urges and, likewise, for some to be female in external appearance but to have male sexual urges. Some of the chemistries of



2.  In defense of multiple Core Affects

sexuality, for instance oxytocin, have been re-used to mediate maternal care – nurturance and social bonding – suggesting there is an intimate relationship between female sexual rewards and maternal motivations. 5.  The CARE/Maternal Nurturance system. Brain evolution has provided safeguards to assure that parents (usually the mother) take care of offspring. The massive hormonal changes at the end of pregnancy (declining progesterone and increasing estrogen, prolactin, and oxytocin) set the stage for the activation of maternal urges a few days before the young are born. This symphony of hormonal and neurochemical changes, especially the heightened secretions of oxytocin and prolactin, facilitate maternal moods which assure strong social bonding with the offspring. Similar neurochemicals, especially oxytocin and endogenous opioids, promote infant bonding to the mother. These changes are foundational for one variant of love. Indeed, virgin female rats can be induced to become maternal under very specific testing conditions (see Chapter 13, Panksepp, 1998a). 6.  The PANIC/GRIEF/Separation Distress System. All young mammals are dependent on parental care, especially maternal care, for survival. Young animals have a powerful emotional system to indicate they are in need of care, as reflected in their intense crying when left in strange places by themselves. These separation calls alert caretakers to seek out, retrieve, and attend to the needs of the offspring. The separation distress system has now been mapped in several species; it is powerfully inhibited by endogenous opioids, oxytocin, and prolactin – the major social-attachment, social-bonding chemistries of the mammalian brain. These basic separation-distress circuits are also aroused during human sadness, which is accompanied by low brain opioid activity. Sudden arousal of this system in humans may contribute to the psychiatric disorder known as “panic attacks.” 7.  The PLAY/Rough-and-Rumble, Physical Social-Engagement System. Young animals have strong urges for physical play. This takes the form of pouncing on each other, chasing and wrestling. These actions can seem outwardly aggressive but they are accompanied by positive affect – an intense social joy. During these activities, rats make abundant high frequency, 50 kHz chirping sounds that have many features resembling human laughter (Panksepp & Burgdorf, 2003). It is interesting to note that there seem to be similarities between the subcortical brain circuits that mediate human laughter and play-induced chirping in rats. The most powerful evidence for an evolutionary relationship between positive affect and chirping is the fact that, if humans tickle rats, these vocalizations often go up to maximal levels, and the young animals rapidly return to the individual human and attempt to solicit more tickling. In contrast to positive affect, when negative feelings are aroused, animals begin to exhibit 22-kHz “complaint” type ­ vocalizations and

37

38

Jaak Panksepp

play ­temporarily ­ceases. A key function of the social play system is to facilitate the natural emergence of social dominance. Play helps young animals to acquire more subtle social interactions that are not genetically coded into the brain but must be learned. Thus, the play urge may be one of the major emotional forces that promotes the epigenetic construction of higher social brains – promoting humor and teasing in humans. This system, like all the other emotional systems, is concentrated in specific subcortical regions of the brain. To reiterate, the primary evidence for the existence of such executive brain systems for diverse primary-process emotions is our ability to artificially activate various kinds of emotional patterns by applying the appropriate kinds of chemical or electrical stimulation to specific subcortical regions of the brain. Emotional theories that fail to deal with the evidence for their existence may be deemed to be profoundly incomplete – perhaps premature, shortsighted or outdated from a neuroscience perspective. To consider these as “mere behavioral systems” is to neglect the mass of evidence that their arousal can serve as “rewards” and “punishments” in traditional learning paradigms. They can clearly influence human higher order mentation, since stimulation of these brain regions activates more intense emotional experiences than stimulation of any other regions of the brain. It is easy to speculate that affects reflect higher brain “readout” of unconscious types of arousal that these systems subserve, but that is to neglect abundant available evidence to the contrary. For instance, radical decortication of animals in infancy (surgical removal of the dorsal cerebral mantle, exposing many of the aforementioned brain structures) leaves all of the above emotional-instinctual urges seemingly intact, even though the capacity to learn new behavior patterns is severely impaired. Most importantly for the present discussion, that affective feelings arise from these subcortical neurodynamics is indicated by the fact that animals are rarely neutral about such artificially induced emotional arousals. Animals are attracted to environmental location and manipulanda where the positive emotions are aroused (seeking, lust, care and play); in contrast, animals escape or avoid arousal of the negative emotional circuits (rage, fear and panic). For instance, all brain sites that activate 50 kHz ultrasonic chirps, an apparent unconditional indicator of arousal of the seeking system in rodents, also support self-stimulation behavior (Burgdorf, et al., 2007). Sites that arouse fearful behaviors can be used to provoke escape and to punish animals (Panksepp, 1990). In lieu of contradictory evidence, the simplest explanation is that raw affective feelings are constituted substantially from the neurodynamics that engender the ­visually evident instinctual emotional responses. Brain-behavior studies have long indicated that if extensive damage occurs to brain regions through which these systems course,



2.  In defense of multiple Core Affects

the survival of animals is severely compromised. They can no longer look after their own welfare (i.e., starting with the work of Bailey & Davis, 1942, 1943). These neural networks are clearly major ancestral tools for living. Presumably, all the primary-process emotional systems are also intrinsic value (i.e., affect) systems that code how animals are faring in their quest to survive. Such affective feelings may be critical for simple forms of learning (classical and instrumental conditioning) as well as for structuring higher mental abilities. The various positive affects index animals’ return to “comfort zones” that support survival. The negative affects constitute “discomfort zones” indicating animals are pursuing trajectories that may impair survival. It is possible that such intrinsic affective potentials constituted the first kinds of internally driven experiences (i.e., primal affective consciousness) that existed in living creatures. There are reasons to suppose that without them, reflective consciousness would not exist. It is possible that in early BrainMind evolution, raw affects and emotional instinctual behaviors arose from the same underlying neurodynamics, providing critical substrates for more subtle affective elaborations as forebrain encephalization expanded.

Cognitive diversification of affects does not negate basic affects My research has almost exclusively been devoted to primary-process issues – those highlighted by cross-species behavioral and neural homologies (Darwin, 1872/1965; Panksepp, 1998a, 2005a). I have long promoted the view that basic affects constitute the foundation of consciousness, with the corollary that higher mind functions simply could not exist without these lower brain functions. Because of the general absence of evolutionarily layered levels of analysis outside of cross-species neuroscientific approaches, I regret that my work has been brought into long-standing disputes in psychology between those advocating other reasonable “basic emotion” views and reasonable dimensional theories of higher affective life, neither of which have been based on neuroscientific inquiries. I think these views could easily subsist peacefully with each other (Panksepp, 2006). In my estimation, the only way we can make substantive scientific progress understanding the raw nature of affects is through brain research. To be on the right track, it would also have to be concordant with the best psychological and cross-species behavioral analyses. Since the brain is an evolutionarily layered organ, unlike any other in the body, affects need to be discussed at several distinct conceptual levels of analysis (e.g., at least at primary, secondary, and tertiary process levels, as envisioned above). The failure to fully appreciate such strategic issues is, I believe, a major source of the scientific dilemma we are discussing, and

39

40 Jaak Panksepp

the ultimate source of many perennial, seemingly unresolvable, debates in the psychology and philosophy of emotions. Because of tertiary process complexities at the human level, we find thousands of highly nuanced affective terms, some of them culturally unique. At the primary-process level of analysis one does not have the luxury of such conceptual flexibility. Very few investigators work at the primary-process affective level – I may be the only one left in the USA after the passing of Paul MacLean (whose work and thinking I admire – see Panksepp, 2002). The work of many others, now amounting to thousands of investigators, has proceeded largely at the secondary and/or tertiary process level. These tertiary process psychology-only analyses naturally tend to take us toward a greater appreciation (and hopefully eventual deeper scientific understanding) of the endless manifestations of affective experiences in lived human lives, for which we sometimes have good verbal descriptors, sometimes not (e.g., as in wine-tasting). That is an important project, but it should not be confused with the important goal of understanding the primary-process causal infrastructure of affective life shared by all mammals. We cannot understand primary-process affects/emotions by staying at higher semantic-psychological or even facial levels of analysis. A psychological analysis is important for characterizing the “objects” of study, along with their mentally expressed dynamics, but an understanding of their causal infrastructure must include a robust program of neuroscientific research that can manipulate the brain with causal variables. Many current dilemmas arise because psychologists and philosophers have not dwelled sufficiently on the experimental constraints and strategies needed in order to clarify the causal mechanisms that create raw affects within the MindBrain. As readers will have noticed, I have used this conflated term, double capitalized in either order (e.g., BrainMind), to highlight that basic aspects of mind – the various evolutionarily “given” experiences of existence – are thoroughly neurobiological, while also accepting the obvious fact that these systems are designed to be exquisitely responsive to diverse events in the body and the world. From this levels-of-control perspective, it is no surprise that the study of neural events does not begin to capture the large-scale intricacies of subjectively experienced cognitions. In contrast to basic affects, there appears to be no discrete neurochemical coding of cognitions, and many higher regions of the brain initially are pluripotent (e.g., so-called somatosensory and auditory cortices can develop visual competence if the occipital cortex is damaged early in life: see Sur & Rubenstein, 2005). All are dependent critically on balanced glutamateric and GABAergic transmissions in cortical and linked basal-ganglia thalamic loops of the brain (i.e., ancient subcortical systems, along with cerebellar “smoothing” functions, are essential for cognitions: see Koziol & Buddings, 2009)!



2.  In defense of multiple Core Affects

Although affects also arise from the arousal of those highly generalized neurochemical systems, there has been no major problem in also identifying more distinct neurochemical linkages for basic affects: There is abundant neural-network­ and neuro-chemical (i.e., neuropeptidergic) coding of basic emotions and their affects in animal models (Panksepp, 1998a, 2005a; Panksepp & Harro, 2004). From an evolutionary perspective it is unlikely to be different in humans. To the best of our knowledge, we mammals all have homologous subcortical networks, remarkably similar, from which primary-process affects emerge (please note, the behaviorists called these entities “rewards” and “punishments” which could be defined as external objects and events, because they had no empirical way to deal with internal BrainMind processes). Let me briefly address models of affect that posit two fundamental dimensions in terms of primordial positive and negative feelings and accompanying arousal states. I have often made the point that affective neuroscience has to consider not only i. the emotional affects, some of which surely arise from inborn neural systems that engender distinct pre-linguistic feeling tones, but also ii. the sensorially triggered affects (e.g., pleasures & displeasures of sound, taste, touch, etc.), as well as iii. the fact that body-state detectors can engender homeostatic affects (e.g., feel­ ings of hunger and thirst) that directly promote and modulate various consummatory satisfactions. It is a challenge for me to imagine all this evolutionary complexity, well represented in subcortical neural circuits, arises from a bivalent affective infrastructure. I have contemplated that parsimonious view from many angles, but the brain evidence does not fall neatly onto such a conceptual simplification. A bivalent affect structure may be a “natural” way for our higher mind to cognitively conceptualize and categorize feelings, as basically “good” or “bad” – reflecting our taste for dualities, like “reward” and “punishment.” However, I do not see how we can begin to understand the diversity of primary-process affects if we remain at a bivalent level. Perhaps the slow pace at which mainstream psychology is being realistically neurologized (finally spurred on by the brain-imaging revolution!), and the general failure to conceptually and empirically deal with primary-process neuropsychological issues, helps explain the root cause of our present debate and persisting scientific dilemmas surrounding the “What is affect?” question. Very few individuals are trying to unravel affects through causal neuroscience work, even in behavioral neuroscience (where there is abundant work on secondary-process emotions such as the learning of fear) and human affect science (where most of the neural work is currently on brain-imaging and peripheral psychophysiological correlates).

41

42

Jaak Panksepp

On defining emotional affects Primary process emotional-affective systems cannot be defined just psychologically and peripherally, for that would only outline the inter-relationships among human concepts (mentalistic words with the usual host of problems: see ­Machery, 2009) and surface descriptions of behaviors, along with autonomic and some neural correlates (typically from the larger higher brain regions where neural firingrate changes are much easier to monitor than from within highly interdigitating subcortical emotional systems where neural firing rates are much lower). The intrinsic processes of the mind also have to be defined by their neural and bodily biochemical intermediaries. A general neurally-based definition of primaryprocess­ emotional networks is provided below: Neural definition of an emotional system (Panksepp, 1982, 1992) – Cognitions instigate emotions 5

– Emotions control cognitions 6 – Positive feedback 4 1 – Unconditional sensory inputs

– Gating 3 of inputs 2 – Coordinated physiological and behavioral outputs

7. Affect arises from activity of the whole system

The key to defining “the structure” of all primary-process aspects of the BrainMind, including basic affects, is to include intelligible neuroscientific descriptions of what we are talking about. Emotional feelings cannot be defined only in semantic terms, no more than the color “red” can be so defined. Neural network characteristics must be a substantial part of the definition. In my estimation, all emotion generating neural systems share six intrinsic characteristics: 1. Primary-process brain networks are triggered by a limited number of sensory/ perceptual inputs that can be expanded enormously by learning and thought. The number of somatic unconditioned stimuli that can initially arouse emotional-affective unconditioned responses are limited. In other words, emotional systems are largely born “objectless” and most object-relations are



2.  In defense of multiple Core Affects

2.

3.

4.

5.

6.

7.

established through learning. The emotional primes initially only have a few connections to the world, as with pain evoking fear in all species and the smell of predators evoking fear in some species (e.g., rats become fearful in the presence of the smell of cats), etc. The number of internal visceral and hormonal processes that may modulate emotional networks unconditionally are bound to be much larger. Through the latter, brain emotional systems are unified with bodily states resulting in various two-way communications, most of which remain to be empirically characterized. The ability to coordinate sets of behavioral and autonomic outputs, including hormonal and immunological parameters. In other words, emotional systems have a coherent, behavioral-affective coherence promoting infrastructure – i.e., there are intrinsic brain networks that coordinate emotional expressions via poorly-understood self-sustaining activities. It is through the somatic outputs that we see emotions in external actions; it is through the autonomic-visceral outputs that we see the many internal actions of emotional systems. The capacity to gate and modulate incoming sensory inputs directly as well as through general arousal systems of the brain (e.g. corticopedal – heading toward cortex – systems, such as via acetylcholine, norepinephrine and serotonin state-control systems). Emotional networks reverberate longer than the duration of precipitating circumstances, at least in part by the role of neuropeptides that can establish “pressures” for action for extended periods of time. Emotional networks have inputs from cortical systems to emotional primes (corticofugal – heading away from cortex – controls allowing cognitions to influence emotions). These last two attributes lie at the heart of well-recognized­ cognition-emotion interactions that are leading many to conflate the two highly distinct types of processes in the brain (Panksepp, 2003). Emotional networks have the reciprocal capacity to modulate cortical information processing (corticopedal controls allowing emotional primes to influence cognitions). Various emotional affects – the psychologically experienced, valenced-feeling aspects of emotional action systems – are created by these complex network properties, with subcortical parts being necessary for the higher brain components.

All these aspects combine ultimately to generate affects, but the primary-proces­s variants can occur without the last two attributes. I should add another attribute that adds much complexity to the primary-process level: (8) All affective networks interact with each other in complex antagonistic, synergistic and opponentprocess­ ways, albeit most such interactions remain to be studied ­empirically – e.g.,

43

44 Jaak Panksepp

perhaps all emotions are modulated to some extent by levels of hunger, etc. Certainly the sensory affect of taste is modulated by hunger and satiation (Cabanac­, 1992). In other words, primary-process emotional networks should never be conceptualized in terms of isolated modules, but as distinguishable networks of brain-body control in complex living systems. Remember, the functions of the liver depend heavily on heart, kidneys, and gastrointestinal functions, etc. To not consider them scientifically as distinct organs, just because they interact would be foolhardy. In sum, primary-process emotional affects arise from various core emotional operating systems in action. The executive infrastructure of each emotional operating system (i.e., the first three of the above attributes) may be more critically important than the others for understanding basic affects. There must be such “commanding” neural processes for emotions since we can turn on emotionalbehavioral episodes in all mammals by stimulating specific regions of their brains. The same is true for birds and, to some extent, ectothermic vertebrates too. The most critical points for our discussion are (i) to achieve such effects, the stimulations need to be applied to very specific regions of the brain, and (ii) such stimulations, in order to be constitutive of affects, must routinely provoke internal states that can serve as “rewards” and “punishments” in various learning and evaluative choice tasks. This attribute has been repeatedly affirmed using various operant tasks (escape and avoidance paradigms) as well as instrumental choices (conditioned place preference (CPP) and aversion (CPA) procedures). This gives us an entry point into understanding such subtle MindBrain processes in some causal detail. The question before us now is whether the primaryprocess affects of the human-mammalian brain come in two primordial types or many varieties. The brain evidence so far suggests “many,” but one could attempt to reinterpret the data to support a simpler duality (but that reinterpretation would, in my estimation, have to be very convoluted in order to be consistent with the evidence).

The function of affect: Evolutionary considerations Might evolutionary “Why would there be many affects?” reasoning shed any more light on the issue? Our work (concordant with Cabanac’s (1992) studies of sensory affects), has been premised on the assumption that the evolutionary “why” for the existence of affects, is that they were simple-minded but useful (i.e., survival supporting and informative) heuristics to anticipate survival issues. Namely, they are the unconditioned experienced “voices” of certain genetically-constructed brain networks.



2.  In defense of multiple Core Affects

Just consider the fact that “raw hunger” (as distinguished from appetite) is not telling you that your body energy stores are dangerously low, but rather, that it is time to “top up your tank” when the opportunity arises. Indeed, that is the main function of the nervous system at all its hierarchical levels of control: to anticipate the future in ways that promote survival and successful reproduction. Emotions and their associated affects do this by allowing animals to make rapid, snap-judgments­ about key survival issues that have repeatedly and persistently confronted their ancestors in their implicit evolutionary aspirations to live and thrive. At a ­primary-process level such snap-judgments are surely simple-minded­, and do not suffice to solve most of the real-life challenges and contingencies of mammalian existence. For all that vast complexity of “everything else” we need learning and higher cognitive processes. However, my (our?) only concern here is with the intrinsic nature of the affective life. As many have recognized, the various positive affects coax animals to keep doing more of the same; the various negative affects coax animals to stop doing things that are followed by aversive feelings. Indeed, this is how the original “Law of Effect” enunciated by Edward Thorndike (1874–1949) was generally phrased (however, rather than “positive and negative affect” it used the terms “satisfactions” and “discomforts”; see Thorndike, 1911). But there are many things that need to be done in this world for survival, and raw affects, in order to have optimal veridical “signaling” value for learning, would need to come in many varieties. Otherwise, “everything else” in affective life would have to arise from simple reward and punishment linked learning. As already highlighted, a dilemma that confuses debates in this field of inquiry is that different scholars are not clearly specifying at what level of MindBrain organization they are focusing their inquiries. The perspective one takes and the “understanding” one achieves is very different if one is working to reveal the dark basement of the mind, or the lofty attic of the mind, or the ordinary living space in between. Also, affect scientists often do not clearly distinguish what aspects of affect they are talking about – sensory, homeostatic or emotional. With this amount of complexity, I believe dimensional models of affect would run promptly into trouble (if they only considered all the facts). Let us just consider one example – the classic Minnesota Starvation Experiments at the end of WWII (for modern summary of the heroic 1944–45 studies by Keys, et al., 1950, see Kalm & Semba, 2005). When food was withheld for a few days, humans rapidly developed extreme levels of “hunger” along with nearly psychotic cravings for food. These feelings did not spill into cravings for water, sex or social companionship. Those occur only with extremes of thirst, lust, and loneliness. We already know of dozens of neuropeptides that specifically control

45

46 Jaak Panksepp

aspects of energy-balance regulation (Panksepp, 2010) as well as other affective feelings (Panksepp & Harro, 2004). No human in dire straits would experientially confuse hunger and thirst, and this is not just a conceptual act unique to humans. All mammals would be equally clear and forthright through their prompt behavioral choices. Indeed, the best brain imaging of such homeostatic affects (as summarized by Denton, 2006), indicates that thirst and air-hunger produce different arousal patterns, largely concentrated in subcortical regions that are known to regulate such bodily states in animals. These, and many other bodily states, involve affective issues. Perhaps at some early point of MindBrain evolution, some kind of primordial negative and positive affect may have existed before the evolution of more discriminating affects, but little has been discovered in presently existing brain organization to suggest that such primitive bivalent states still captivate our minds. In sum, one can easily speculate that from a cogent evolutionary-historical perspective some kind of raw “good” and “bad” feelings may have been primordial substrates from which a diversity of primary affects emerged in brain evolution. How many distinct raw affects exist within the mammalian brain may also inform our understanding of the evolutionary foundations of consciousness. Affects, at their most primal level, reflect intrinsic valuative processes of the BrainMind. They directly inform animals of the importance of certain sensations from the world (sensory affects) as well as certain actions and interoceptive processes that emerge from the inside of the body (homeostatic affects) as well as more directly from within the BrainMind (e.g., emotional affects). Affects indicate that ‘something’ inside or outside of organisms is of clear importance for the survival of animals. If raw affects initially had no veridical signaling value beyond “go” and “stop” (well developed in all invertebrates), then everything else in our affective lives would have to be learned by reward (positive affect) and punishment (negative affect) learning processes. Still, I can easily envision that in conjunction with the more specific affects, that the brain may also have more generalized positive and negative affective processes, perhaps through the global brain-spritzers such as acetylcholine, dopamine, norepinephrine and serotonin networks as well as many neuropepties such as endorphins and dynorphins. For instance, sudden alerting-arousal functions are surely subserved by brain acetylcholine and norepinephrine systems, usually accompanied by negative valence. In contrast, dopamine is essential for many positive emotional arousal functions while serotonin can tone down and “relax” many specific emotional systems, and hence may engender a general positive feeling of being at ease. Wouldn’t it be neat if the specific primary-process affects and dimensional affects can subsist harmoniously within the vast complexities of the higher reaches of the MindBrain as well as within the lower emotional systems of the brain?



2.  In defense of multiple Core Affects

2. How does affect relate to prototypical emotional episodes? The evidence suggests raw emotional feelings are closely enmeshed perhaps isomorphic with the neural circuits that engender intrinsic instinctual-emotional action tendencies (i.e., the so-called “prototypical emotional episodes”). Thus, distinct emotional affects may relate to instinctive emotional actions simply through their massively shared executive neural circuitries. It may be somewhat different for sensory and homeostatic affects, also subcortically concentrated, but I will not dwell on those issues here since the available database is smaller. How do we know the above? Animal studies clearly and repeatedly have demonstrated that one can stimulate specific regions in ancient parts of the brain and provoke distinct emotional behaviors, and these arousals can serve as “rewards” and “punishments” in a large variety of learning tasks. To the best of our knowledge, the neocortical mantle is not necessary for these effects. The discussion here is also relevant for Question 5 below, better covered now than later. Half a dozen compelling and reliable aspects of such seminal discoveries, starting initially with stimulation-bound rage (Hess, 1954, 1957), became evident during the last half of the 20th century (Panksepp, 1998a): i. Coherent emotional behaviors were evoked by applying unpatterned electrical stimulation into the brain, causing the animal to shift from a neutral emotional state to one of apparently being enraged, fearful, full of curiosity and so forth. What this tells us is that no higher cognitive activity – attributions and other evaluations of the world – are required to arouse behaviorally distinct and coherent emotional flare-ups. Walter Hess who first discovered stimulusbound RAGE (see his two cited books summarizing his vast compendium of work) was hesitant to suggest such brain arousals were experienced by animals and a major source of the feelings of human “anger,” but late in life he acknowledged that the only reason he failed to be true to those personal convictions was that he did not want to have his work marginalized by the American behaviorists. ii. In any event, these “stimulus-bound behaviors” are essentially never affectively neutral. Every time one evoked the visible instinctual emotional states, it was typically found that positive or negative affective characteristics were also present, as monitored with variety of procedures – i.e., as “rewards” and “punishments” in a variety of operant tasks, as well as CPP and CPA tasks). Still, it is important to emphasize that one can even evoke key behavioral indicators of such circuit functions in anesthetized animals, for instance sniffing for the SEEKING circuit. Remarkably, the thresholds for evocation of such simple exploratory responses in unconscious rats correspond strictly to thresholds

47

48 Jaak Panksepp

iii.

iv.

v.

vi.

for self-stimulation reward monitored when the animals are awake, with correlations typically above .9 (Rossi & Panksepp, 1992). Comparable approaches can be pursued for the FEAR and RAGE systems where hind leg thrusting and teeth chattering can be observed in anesthetized animals (Panksepp, et al., 1991). When the psychological experiences of humans were observed during similar brain stimulations (typically as secondary findings in ongoing neurologicalclinical studies), the affective changes evoked corresponded reasonably well to the emotional activations seen in animals (Heath, 1964; Panksepp, 1985). Admittedly, some very large scale studies of this sort (e.g., Sem-Jacobsen, 1968) have yielded ambiguous results, perhaps because of the enormous range of stimulation sites used since understanding emotions was not the major goal of such studies. Practically all of the above are obtained only from stimulating subcortical regions of the brain, homologously organized in all mammals, and never from stimulating higher neocortical areas that enrich our lives with tertiary-process­ cognitive mentality (this should give us pause to worry about how the study of human words relates to primary-process affects). When these subcortical brain regions are severely damaged in animals and humans, consciousness at all levels of mental processing is globally compromised (Baily & Davis, 1942, 1943; Parvizi & Damasio, 2001; Watt & Pincus, 2004). In contrast, total elimination of higher neocortical regions of the brain at birth, leaves the affective vitality of animals intact, even enhanced (Panksepp, et al., 1994). Also, human children with no higher brain functions still appear to be affectively coherent organisms, and exhibit preferences for diverse forms of affective stimulation (Shewmon, et al., 1999).

Regrettably, human investigators of emotional words and cognitions have largely ignored the cross-species brain stimulation and neuropharmacological evidence. In addition, they have yet to pursue illuminating neuro-evolutionary inquiries of their own. At the very least, the above lines of evidence strongly indicate the existence of ancient, evolved affect generating circuits in subcortical regions running through homologous brain regions in all mammalian brains. The data strongly indicate that core emotional affects are intimately related to the subcortical instinctual-action apparatus of the brain, situated rather far from the brain regions that generate the words dimensional theorists study to glean insights about similar matters. The discovery of abundant homologies among the underlying neurochemistries, especially among the many neuropeptides, in all mammals studied so far,



2.  In defense of multiple Core Affects

provide additional cogent ways to translate animal affective/behavioral studies to human work, with abundant clinical implications and predictions (Panksepp & Harro, 2004). Our understanding of human affects could be enriched already through studies with at least a dozen neuropeptides that could falsify the affective views I have espoused.

3. What kind of affect might be had by other mammals and/or human infants? How does such affect relate to the affect of adult humans? Shall we consider animal infancy to extend through the period when babies are dependent on mother’s milk for survival? And human infancy? Shall we consider the corresponding period, as estimated from our Environments of Evolutionary Adaptation (EEA) from studies of tribes like the !Kung San, as ranging up to 2–3 years of age, since kids of those ages are being provisioned with human milk (i.e., under conditions simulating the EEA that existed before the advent of artificial modern manufacture of infant formulas, with preservatives along with many other baby food products)? If so, the likely best answer to the question would be very many kinds of affect, and about the same for all mammals at a primary-process lev­ el. However, cognitive enrichment proceeds very rapidly after birth, with human babies rapidly developing a great deal of emotional sophistication (Reddy, 2008), presumably because of their maturing higher mental abilities (Tomasello, 2009). Most of the basic emotional networks (aside from lust and care) are maturing during infancy – giving young children robust seeking, fear, rage, panic and play urges early on. Ditto for the non-emotional affects – e.g., sensory-taste and homeostatic-hunger: Cross-species research has found evidence for four distinct affective responses to the four primary taste qualities quite early in life (Sewards, 2004; Steiner, et al., 2001). Surely hunger and thirst and other bodily imbalances affectively impact infant lives. How might such primary-process affects influence how adult humans experience their emotions? I expect the homeostatic affects remain qualitatively the same but perhaps get weaker than in infancy because of higher regulatory functions. In contrast, the sensory affects surely get more complex in adulthood, as one learns to navigate and combine the details of one’s culinary, somatosensory, and other sensory terrains. I would imagine that the primary-process emotional affects are foundational for the emergence of the more complex emotional-affective experiences throughout human life, where they can be combined into the seeming infinity of social-emotional nuances from admiration to envy and Schadenfreude, for example (here one must note that the supposed “love molecule” oxytocin can promote both of those latter feelings: Shamay-Tsoory, et al., 2009). It might be

49

50

Jaak Panksepp

noted that one’s capacity to process such “higher” emotions is dependent on medial frontal cortical regions (Shamay-Tsoory, et al., 2007) that have been implicated in the emotionally guided construction of a core self (Northoff­ & Panksepp, 2008; Panksepp & Northoff, 2009). With cognitive development, one can even conceptually cross seeming impenetrable categorical boundaries like that between sensory and emotional affects. For instance, even basic homeostatic/sensory affective states such as disgust, may be used to mold higher order emotional responses such as social disgust and disdain through social learning. Whether other animals also have more subtle emotional feelings such as jealousy, shame, guilt or a sense of humor – feelings that are created by the interrelations of basic emotions with higher cognitive processes – remains an open issue. Subtle analyses of animal behavior in natural environments certainly suggest that other animals do have more subtle emotions built upon the Big 7 described above. But, how would one study “worry” in animals? Compassion? Embarrassment? Envy? Teasing? Sublime feelings? We have no compelling experimental methodologies to study those, albeit PET scanning has been promising in monitoring jealousies in other primates (Rilling, et al., 2004). At present, we also have no unambiguous evidence that other animals cognitively reflect on their own feelings or the mental states of other animals (except perhaps in other great apes: de Waal, 2006). Thus, we do not know if they can exhibit cognitive forms of suffering – prolonged and intensified psychological pain because of their ongoing thoughts about their feelings. Those aspects of mind have less distinct behavioral indicators than the basic emotions, and most of the compelling data on the complex socially-constructed emotions is bound to come from neuropsychological studies on humans. In short, in emotion-affect studies, chaos will prevail if we do not adequately recognize levels of control within the brain. Rather than focusing on the many differences among individuals and species that are bound to arise from analyses of higher MindBrain activities, let me re-focus briefly on the affective attributes of primary-process emotional networks that have received comparatively little attention among psychologists and philosophers. In other words, what might be the relation between the “Big 7” and the common sense emotions of adults. My perspectives on this were highlighted in Question 1 where these systems were introduced. However, let’s briefly consider fear-anxiety, the “traditional” emotional system that has been studied most by neuroscientists. Does the basic emotion of fear, help constitute a homogenous family of “anxieties”? Certainly there is a mountain of human brain imaging evidence indicating that a large variety of very modest anxiety-related stimuli arouse the higher reaches of the primary-process fear system in the amygdala (for reviews, see Delgado, et al., 2006; Sehlmeyer, et al., 2009). Indeed, it seems that



2.  In defense of multiple Core Affects

practically every anxiety ever conceptualized and evaluated with brain imaging has yielded arousal of the amygdala. Further, ethologically correct paradigms of fear applied to humans can envision arousal all the way down to the midbrain PAG (Mobbs, et al., 2009). In addition there is now abundant evidence that benzodiazepine anti-anxiety agents (receptors for which are enriched along fear networks – see Panksepp­, 1990), can reduce many anxieties – from highly specific phobias to vague neurotic fears. However, typically there is not enough data to answer certain questions, such as are some members of the anxiety spectrum of disorders characterized by a fuller activation of the fear system? But we can say that separation anxiety and panic attacks have typically been quite unresponsive to traditional benzodiazepines, while they respond well to drugs that are marginally effective for anticipatory and neurotic anxieties – e.g., imipramine and opiates (Panksepp, 1998a). Thus, we might provisionally conclude that if an “anxiety” problem does not powerfully impact the fear system, running from amygdala to PAG, then we are probably dealing with a different type of negative affect. This again highlights why we must consider multiple negative affects at the primary-process level. It also leads me to wonder whether there may exist some type of unified negative-positive­ affective integrator at some higher-order psychological level (Russell’s Core Affect?). But to my knowledge there is no hard neuroscience evidence for that yet. Accordingly, I wonder whether most higher emotional permutations arise more from conceptual acts than evolved neuropsychological realities.

4. How can we infer affect from observation of behavior? This has already been substantially addressed above for the primary-process level. To reiterate, when one arouses a coherent emotional-instinctual behavioral sequence with brain stimulation, that stimulation will prove to be rewarding or punishing in typical animal behavior learning tasks. This suggests that at a neuroethological level, one can probably use instinctual emotional responses as proxies for monitoring distinct affective states. In contrast, at the tertiary-process level, psychologically experienced cognitive aspects are not yet capable of being monitored empirically other than by verbal subjective self-reports (a cognitive read-out measure). And there is now abundant human data that when one expresses emotional facial behaviors, corresponding feeling states are also aroused (Schnall & Laird, 2003; Strack, et al., 1988). However, with just facial actions, the affective changes are rather modest. When one renders full-body dynamic expressions of emotional actions, such as laughter and crying, one can voluntarily evoke quite intense

51

52

Jaak Panksepp

a­ ffective feelings (Panksepp & Gordon, 2003). Also, we should not forget the complexities of emotional facial action and bodily patterns in infants are impressive in both humans (Reddy, 2008) and other higher primates (Ross, et al., 2009). Spontaneous human affective sounds – e.g., “Duchenne laughter” – may be especially direct measures. Likewise, because of recent developments in the study of emotional vocalizations of animals (Brudzynski, 2009), it remains possible that we can also get “self-report” types of information through the careful study of emotional vocalizations of other animals (Burgdorf & Panksepp, 2006; Panksepp, 2009), but such lines of research are not as well developed as they could be. For those who doubt that animals have feelings, let me elaborate on a remarkable finding from rats noted earlier (Soltysik & Jellen, 2005): When rats are put into standard (and rather awful from a humane perspective) fear-conditioning­ tasks (tone-shock, classical conditioning), they show abundant complaints (22 kHz squeals) between trials, which cease as soon as the scary tone comes, leading one to wonder whether two slightly different anxiety-type affects are involved – worry and full-blown fear? – or whether the shift in 22 kHz USVs is just a matter of differences in fearful arousal intensity. In any event, if one also inserts “safety signals” into the ongoing distress, indicating that no shocks will be forthcoming for a while, all animals promptly sigh – a double inspiration that humans often exhibit during moments of relief. Thus, perhaps we can develop models for understanding some higher-order (cognition­-triggered) emotions such as “relief ” even in rats! Also, the recent discovery of empathy behaviors in animals as lowly as mice (e.g., most recently Chen, et al., 2009), suggests that we must keep an open mind to the subtlety of higher emotional experiences in other animals, although they are typically much harder to study than the primary-process emotions. If we accept verbal reports to be good indicators of emotional-affective states in humans, then let us accept natural emotional behaviors to be equally veridical read-outs of animal feelings. Further, as we triangulate such behavioral evidence with our growing understanding of the underlying brain control mechanisms, with concrete predictions concerning human feelings, the existence of affective experiences in animal BrainMinds gains further support. It is long past time for neuroscientists to be realistic about animal minds (Edelman & Seth, 2009). It is good that investigators like Jim Russell do not see this to be a problematic issue. If the above analysis is correct, then in animals and human infants who do not censor their emotional actions, we can probably read their affective/emotional states from their behaviors more readily than we can in human adults. In contrast to our enormous capacity for higher brain inhibition of lower brain processes, the other animals probably do not have as much reason to try to deceive others. Their cognitive self-esteem is not tied up in what others think of them (at least



2.  In defense of multiple Core Affects

not to the degree that is common in our species). Further, our urge and capacity for “self-understanding” and “self-esteem” probably did not develop to be highly reliant on our ability to express emotions– they had different selection pressures. The overall point being that we may scientifically understand primary-process emotionality better through animal than human research.

5. Is affect pre-linguistic? How can we study what is pre-linguistic? How does language alter affect? Primary-process affects have to be pre-linguistic for them to make any neuroevolutionary sense – namely that very similar “rewards” and “punishments” are critically important tools for living and learning in all mammals. This, of course, causes great mischief in emotions studies since we have no option but to talk about such BrainMind functions with our most sophisticated higher brain functions, using concepts that can only be approximations of that in which we are interested. Language obviously can be used in highly idiographic ways, which is one reason scientists always need to operationalize their concepts. For me primaryprocess emotional systems are operationalized most clearly and profoundly by the types of coherent emotional behavior patterns that can be induced by stimulating specific regions of the brain, along with measures of the affective associates of such arousals. Since, at birth, emotional-affective potentials of the nervous system are minimally connected to world events they must also be initially pre-intentional in the cognitive sense (they occur before we can exhibit “intentions to act”) even though they are surely “intentional” in the evolutionary sense – they spontaneously engender “intentions in action” (Panksepp, 2003). The fact that affects are pre-linguistic in animals (as well as human babies), supposedly engenders great difficulties for experiential emotion research in animals. Hopefully the preceding discussion has put that misconception to rest. Still, the critical empirical issue, barely addressed experimentally, is the extent to which we can claim that various affective states evoked by brain stimulation are as diverse as the emotional action patterns we can evoke. Although this research is in its infancy, as already noted, it can be broached using discrimination studies, namely evaluating the extent to which animals can distinguish stimulation of two brain sites that yield distinct emotional behaviors (Stutz, et al., 1974). I would predict that such studies will yield at least 5 distinguishable types of emotional affects in young mammalian brains (seeking, rage, fear, grief/panic and play/ Joy), without even considering the many sensory and homeostatic affects. I am not as confident that the brains of young animals could distinguish the ­diverse ­

53

54

Jaak Panksepp

pro-social affects, for instance lust and care feelings, which, because of their adult hormonal controls, are not fully developed in childhood. Indeed, those may be more subtle feelings, dependent heavily on seeking affects, so that perhaps even adult animals could not distinguish them from each other when evoked in primary-process forms. I hope such critically important and difficult “pre-linguistic” studies are eventually conducted in animal models. Again, one reason I have confidence that such experiments will support the existence of multiple positive and negative affects is because in drug-discrimination studies, artificial arousal of practically all of the distinct neurochemical systems of the brain, many of which have affective properties, are discriminated by animals (for overview, see Panksepp, 2010). That is a short answer to a difficult question that has not received the neuro-empirical attention it deserves either in academic psychology, or by the behavioral neuroscience community. In addition, since all laboratory animals are pre-linguistic, it is very hard to relate their feelings to the ones humans talk about. Hence, the validity of provisional conclusions derived from animal work needs to be validated through brain-stimulation and psychopharmacological predictions that can be made to humans (Panksepp, 1999; Panksepp & Harro, 2004). However, as already noted, we do have wonderful objective measures of animal likes and dislikes, preferences and aversions. Also, all mammals have diverse patterns of emotional vocalizations, so that we can even envision studying their feelings through vocal affective “self-reports­” (Panksepp, et al., 2002). The evidence is overwhelming that ~50 kHz chirps can index positive affect in rats, while ~22 kHz ultrasonics can index negative affects across a diversity of situations (Burgdorf & Panksepp, 2006; Knutson, et al., 2002). These vocalizations can be used as general indicators of positive and negative affect, and animal welfare states, but this should not be taken as strong evidence that those are the only affective categories generated by the brain. However, it may indicate that at some level of neural organization, even animals relate their feelings into positive and negative categories. As previously noted, the ~50 kHz chirps, which can be especially robustly activated by play and tickling, may have a more than passing resemblance to human laughter, both of which are known to be subcortical processes (Panksepp, 2007a). When we mapped the circuitry for this response, every animal that exhibited brain stimulation induced ~50 kHz chirps also exhibited self-stimulation reward for activation of those brain sites that run along the dopamine facilitated brain SEEKING system which, in our estimation, participates in every emotional process (Burgdorf, et al., 2007). This kind of causal work in “lower” animals can lead inquiries, with concrete neuroscientific predictions, in humans as well as in the analysis of laughter patterns in great apes (Ross, et al., 2009). For instance, we



2.  In defense of multiple Core Affects

would predict that dopamine receptor antagonists that reduce rat chirping will also reduce human laughter. Obviously for sounds to be pre-linguistic does not mean that they are precommmunicative. The more subtle qualities of emotional vocalizations clearly need to be studied more intensively. Many pet owners, especially of dogs and cats, usually claim they can distinguish all kinds of emotional nuances in their simple barks, whimpers and meows. Much more work needs to be done on such highly nuanced communicative issues in both animals and humans, though, we can all easily recognize the squeal or moan of pain across species. Emotional states are also obviously communicated and provoked by the prosodic quality of speech. If someone speaks angrily at us, it can provoke strong feelings. This signaling effect of emotional speech quality deserves much research attention (Zinken, et al., 2008). Human laughter, easily instigated by being tickled by words (i.e., humor), highlights a fine example of how language can provoke emotions. Although we do not know much about how all this occurs within the brain, we do know that the emotional language of poetry and theatre often moves us emotionally in aesthetically profound but often idiographic ways. Songs certainly move us affectively more potently than propositional sounds, presumably because songs stylize our natural emotional prosodic vocalizations, accompanied by a cognitive overlay of appropriately selected words. One challenge for dimensional emotion theories is why happy and sad music selectively modifies those self-reported affective feelings, without any effects on measures of anger and anxiety (Panksepp & Bernatzky, 2002, see Figure 3, p. 146). Why is it that various string instruments can tear at our heartstrings better than many other instruments? My own meager hypothesis is that certain instruments and certain pieces of music render the dynamics of our ancient separation-distress calls that remind us of the psychic pain of social loss melded with the possibility of redemption (Panksepp, 1995). However, language works both ways, it can trigger feelings and it can inhibit feelings. When I have been intensely sad, and talked about what has happened with friends, the feelings fade for a while, returning soon after I am alone again. When I have been angry and let my anger out in the way I say things, it used to intensify my anger, but now, it makes me sad. Has anyone studied such issues formally in a scientifically rigorous way? When I am in love and I express that love deeply and sincerely, and it is received openly, my spirits are elevated for quite a while. For me, the greatest mystery is how the brain generates those raw affective experiences. My main hypothesis would be that various cognitive acts play upon our emotional primes like skilled pianists on the ivories. But how does mere propositional speech modify our affects? We simply do not know scientifically, but we can be sure it has to be a top-down effect within our MindBrain.

55

56

Jaak Panksepp

An important consideration here is that there is commonly a see-saw effect between higher and lower brain regions. With abundant higher cognitive activities, our affects are diminished (Liotti & Panksepp, 2004). When our feelings are intense, the top of the brain tends to shut down (Damasio, et al., 2000; for a synopsis of that effect see Figure 2 in Panksepp, 2003). Thus, in brain imaging of emotional processes, we must remain sensitive to the fact that if one is requesting affective self-reports while scanning the brain for affective change, one can easily be polluting the relevance of the brain-scans for understanding affect. Clearly, such see-saw interactions between higher and lower brain regions need to be handled with finesse when trying to image human affects. Indeed, those who have evaluated affective changes that occur in fMRI environments to emotional pictures while minimizing pollution from cognitive decision making (i.e., the affective changes are harvested offline so as to minimize the effects of cognitive decision making on fMRI signals), the most positive correlations to affective changes, as already noted, are typically found in subcortical sites, while negative correlations are typically found in higher brain regions, perhaps reflecting how higher regions are dampening affective intensity (Northoff, et al., 2009). In considering how language can modify feelings, it needs emphasis that there is essentially no evidence that the top of the brain, so essential for our language, can engender emotional feelings on its own. Damasio (1994, 1999), has suggested that the neocortex can engender “as if ” feelings in humans, which depend on past bodily enactments, but more research needs to be done on those issues before we know where or how such feelings are felt in the brain. Perhaps primary-process affects can be re-represented in higher regions of the brain, just as our old motor functions which are so dependent on subcortical cerebellar and basal-ganglia circuits are converted into more subtle cognitively guided actions when those systems interact with the neocortex (Koziol & Budding, 2009).

6. Is affect always conscious? It depends on how one defines consciousness. Consciousness is just like any other evolved aspect of BrainMind, with multiple levels of processing. To my way of thinking, three levels need to be considered at minimum – anoetic, noetic, and autonoetic consciousness in Endel Tulving’s terminology (for overview, see Vandekerckhove & Panksepp, 2009). Anoetic is raw experience without knowledge – primary-process­ in my terminology. Noetic is experience associated with knowledge about the world – secondary-processes that have gone through the “mill” of simple learning. Autonoetic, is awareness of our lives and our unique experiences on the vast neocortical cognitive stage of mental existence – tertiary-process in my terms.



2.  In defense of multiple Core Affects

Thus, if one defines consciousness as the ability to be cognitively aware that one is experiencing something, then by definition, all the other animals may be unconscious zombies. That view can lead to intellectual tragedies of the highest order. Accordingly, just from a philosophical-ethical point of view, I think it is wiser to envision primary-process anoetic consciousness to constitute internally felt experiences. This would include all primary-process affects, which is not to claim that such experiential processes do not have unexperienced neural components. Freud was rather consistent in his opinions. By definition, affect was always experienced. I agree. However, many of the cognitive antecedents of affective change may be unconscious, especially if you define “cognitive” as incoming information through our exteroceptive portals, even though admittedly it could be defined in other ways (Moors, 2009). Clearly, affect relevant information inputs into the brain (e.g., emotional words), verified to have measurable brain effects through cerebral evoked potentials, can be deeply unexperienced subjectively; for instance, unpleasant words presented subliminally induce significantly greater evoked potential amplitudes in the left rather than right hemisphere (Bernat, et al., 2001). Likewise, such unconscious stimuli (e.g., 1 msec word exposures, where signal-detection procedures indicate that subjects did not detect (experience) anything to have occurred in their visual fields), can evoke distinct muscular activities in human subjects. For instance, electromyographically monitored corrugator muscle tension in the brow, as well as skin conductance responses, to happy and angry schematic faces presented below the objective identification threshold are greater for angry faces (Bunce, et al., 1999; Wong, et al., 1994). Even some choice behaviors have been reported to be demonstrably modified by the totally unconscious presentation of emotional words (Shevrin, 2001), suggesting the existence of unconscious affects. When I asked Howie Shevrin whether they had ever measured affective change in such experiments, he indicated they had not. We agreed to collaborate on such an experiment, using emotional terms, presented at 1 millisecond (i.e., below the absolute detection threshold) that had previously been validated to influence brain activity as monitored by evoked potentials (Snodgrass & Shevrin­, 2006). Since a diverse set of emotional stimuli were used, for efficiency we ­decided to evaluate affective changes with the more global estimates allowed by dimensional approaches. We used Peter Lang’s (see Bradley & Lang, 1994, 2000) self-assessment Manikin scales, with a continuous visual analog measure under each of the three pictorial scales (each scale represented in 5 pictorial Manikins) – for arousal, valence and potency (the last being the most subtle concept in dimensional theories of emotions but basically corresponds to the “power” of an emotion in the mental apparatus, which can be described more vividly perhaps by how much it fills the mind, as compared to “arousal” which is the felt internal

57

58

Jaak Panksepp

intensity of an emotional feeling). In two successive experiments, even though “valence” and “arousal” were not modified, the third measure was modified in the same direction each time; this measure of “power” was increased more by pleasant than unpleasant emotional words, totally unconsciously presented, both times the experiment was run (Panksepp, et al., 2004). Although we could not agree on how to interpret this finding, I interpreted the result to mean that totally unconscious cognitive information could marginally modify background affective experience, the power of ongoing mood, so to speak. Of course, the fact that such unconscious cognitively-induced changes in affect were not valence congruent, but only evident in the rather abstruse “power” measure, remains perplexing. It is very difficult to use these dimensional concepts of “arousal” (intensity of feelings) and “power” (perhaps corresponding to the concept of internally experienced salience) in animal research. In contrast, it is conceptually easy (methodologically hard: again, see Stutz, et al., 1974) to do valence-type discrimination studies in animals using different types of brain stimulation reward. This is as close as we can get to empirically distinguishing primary-process states in animal brains. Hopefully readers will appreciate the incredible difficulties in contrasting dimensional and basic-emotion approaches in animal models. They are comparable to the near impossibility of studying the primary-process emotional affects in human beings. To what extent are we thereby caught between a rock and a hard place in contrasting tertiary-process human dimensional views with primaryprocess emotion views derived from animal brain research?

7. What is the relationship between affect and cognition (conceptualization, intentionality, appraisal etc?) Neocortical “limbs” and “branches” that constitute unique cognitive canopies of the BrainMind are bound to exhibit abundant diversity across species, allowing more distinct mental lives in different species than prescribed by the affective evolutionary “trunks” and “roots” that are more closely shared. From my perspective, a distinction needs to be made between cognitions and the primary-process emotions. This is not difficult. For instance, higher cognitive brain systems were designed to respond to the incoming “information” from our externally directed senses; primary-process affects are intrinsic value-codes of the brain. Both neuroscientific evidence and evolutionary reasoning strongly suggest that there are many basic affects. However, additional diversity is surely added initially by “simple” learning mechanisms shared by all mammals, and subsequently by capacities for complex thoughts in well encephalized species, many of which are bound to be unique in the human species (e.g., language-based ones). Human minds,



2.  In defense of multiple Core Affects

because of cortical complexities, are able to conceptualize practically anything in words, even total fantasies. This is a big dilemma for those who study emotional words – tertiary-process aspects of mind. In the intact adult human brain, the relations between affect and cognition are enormous and almost impossible to disentangle. In the current era where emotion research is fashionable, psychologists have extensively conceptualized and discussed how higher cognitive processes interpenetrate with more ancient primary affective processes. Indeed, the claim that cognitions and emotions are totally interpenetrant (see whole issue of Cognition & Emotions, 2007, vol. 21, issue 6) and that meaningful distinctions between the two types of neuro-mental processes cannot be made are quite misleading (or shortsighted – namely they are not really considering primary-process issues). With cross-species brain research (e.g., brain stimulation and decortication studies), primary-process affects can be disentangled from the higher cognitive processes with which they dramatically interact in the intact human MindBrain (Panksepp, 2003). The empirical work of most emotion researchers arises largely from the “bubbling” top of the cauldron of emotional complexity, while only a few study the rich broth at the bottom of the evolutionary crucible of emotional life. For constructivists (perhaps most dimensionalists also), cognitive appraisals are very important for understanding emotions; for me they are comparatively recent evolutionary issues that are of secondary importance for understanding the nature of Core Affects. And those higher issues simply cannot be approached easily through crossspecies affective neuroscience research. Maybe if we had more people working on the foundational issues, we could begin to envision productive human-animal comparative experiments that can bring the diverse levels of knowledge into a more coherent harmony (Panksepp, 1999). In any event, from the primary-process view, the notion that cognitions and emotions are totally interpenetrant is untenable. Here are a few simple lessons from cross-species neuroscience: 1. If animals are decorticated when they are very young – whether an infant rat through brain surgery or a human child through developmental misfortunes – their primary-process emotional-affective responsivity remains largely intact in “simple” animals (Merker, 2007; Panksepp, et al., 1994) and substantially intact in the humans that have been reared humanely (Shewmon, et al., 1999). However, both will be severely impaired in cognitive abilities. 2. It is next to impossible to evoke strong affective experiences by stimulating neocortical regions of the brain that are essential for our cognitive lives; it is very easy to evoke affect by stimulating specific subcortical regions of the brain that are necessary for us to remain cognitively aware.

59

60 Jaak Panksepp

3. If one considers a mass of brain imaging data, there seems to be a reciprocal see-saw relationship between cognitive activities and affective states (Liotti & Panksepp, 2004; Northoff, et al., 2004). When emotional arousal is high, cognitions are impaired. When one is indulging vigorously in cognitive activities, most emotional feelings tend to be inhibited. 4. Indeed, if affects were cognitively mediated, one would generally expect positive correlations between our experienced emotion arousals and activations of higher brain regions; in fact those correlations tend to be negative while affective correlations with arousals of lower brain regions are generally positive (Northoff, et al., 2009). 5. There are many other reasons to believe there are meaningful distinctions between cognitions and affects, not the least of which is that cognitions transform perceptual inputs into ideas while subcortical affects motivate decision making, as highlighted dramatically by recent studies of shopping (Haber & Knutson, 2009). The cortex is needed for all conceptualizations and appraisals but not for primary process affects. The primary-process emotions are brain systems that mediate “intentions in action” but they are not systems that create our “intentions to act” (Panksepp, 2003). In other words, the coherence of our emotional-instinctual responses, reflecting evolutionary memories, have a directedness that is intrinsic, rather than cognitively mediated, leading many humans to delude themselves (Wegner, 2002). Here is where the interesting new philosophical concept of “alief ” – an affectively salient action that counters our cognitively understood beliefs – raises havoc with our rationality (Gendler, 2008). As every psychotherapist knows, our old emotional-affective “intentions in action” and habitual emotional ways of behaving can easily override our cognitive belief systems. In any event, an answer to how affects interpenetrate our appraisals, conceptualizations and intentions (now with the participation of abundant mirror neurons in Broca’s Area – the expressive-language and bodily gesture regions of the brain: Iacaboni, 2008), is a more difficult neuropsychological problem, at least at a causal level of analysis, than the neuroscientific clarification of primary process affects. The latter can be clarified in animal models where detailed neurochemical work on underlying BrainMind mechanisms can be done; in contrast, relatively little insight about human cognitions (except for the basic mechanisms of memory) can be achieved with animal brain research. Indeed, at the cognitive level, emotions and personality dimensions become highly idiographic. Accordingly, I personally remain more gratified by the diverse emotional understanding that novelists, film-makers and other artists of the human affective experiences have provided than experimental psychologists.



2.  In defense of multiple Core Affects

However, some objective progress is being made, largely because of the advent of modern brain imaging. I am especially intrigued by the fact that in borderline personality disorders, where affective instability looms large, patients are more likely to recruit phylogenetically older “instinctual” networks more than healthy subjects who generally tend to recruit higher “cognitive-reflective” brain regions (Koenigsberg, et al., 2009a, 2009b). In the field of depression, brain-imaging is highlighting the fact that the medial frontal sectors of the brain that elaborate active “resting states” become hyperactive, as if the depressed brain is ruminating more on life problems than more typical individuals (Grimm, et al., 2008, 2009). Without doubt, all these higher cognitive issues that impact emotional processing can be better studied in humans right now.

8. Is affect altered or filtered through culture? In what ways? Are there universal aspects of affect? For basic neuroscientists working on animal models of affect, these questions can only lead to the sharing of opinions. These issues are largely inaccessible except through human subjective reports and perhaps through some forms of associated brain imaging. To my knowledge, these issues have rarely been brought into the laboratory and evaluated in causal ways. I look to anthropologists, culturologists, psychologists and sociologists to debate and illuminate such questions. In addition, brain imaging is also now capable of providing novel cross-cultural images that can sway our arguments one way or the other. Of course the key issue will be whether provisional conclusions reached at any of these levels are sound as opposed to just-so stories. I don’t think one has comparable doubts about the primary-process issues that are best illuminated through cross-species animal brain research. In lieu of dipping deeply into that emerging literature, I would only share a few generalities: No doubt affects are filtered through culture in highly idiographic ways. Let’s begin with the proposition that the systems that generate primary process affects are genetically encoded. A key issue for neuroscientists would be how much affective differences among cultures may be due to genetic factors and how much to how extensively shared experiences within a culture differentially modify its members’ brains. The human genome is full of single-nucleotide polymorphisms (SNPs) many of which are neutral (have no consequences for human phenotypes), but others clearly influence emotional tendencies in humans and other primates (Suomi, 2006). Some progress on the differential experience issue at a neuroscience level is currently being made, for example, through the recognition that cortical “mirror neurons” can be programmed in different ways

61

62

Jaak Panksepp

in different­ cultures, leading to culturally-distinct gestures of various kinds (for examples, see Iacaboni, 2008). An equally interesting question will be how culture is filtered through the genetic control of affects. Different breeding populations can concentrate different alleles of the same gene leading to genetically controlled temperamental differences as has been seen in different subspecies of rhesus monkeys (Suomi, 2006). Such SNPs (i.e., single nucleotide changes in DNA that often lead to amino acid substitutions in the encoded proteins) also predict emotional temperamental differences in humans. Such genetic factors have already been implicated in aggressive temperaments, susceptibility to post-traumatic stress disorders, as well as many other neuropsychological traits (Kilpatrick, et al., 2007; van der Wee, et al., 2008). Some of these SNPs are likely to segregate in human populations across traditional cultural lines, which may influence the emergence of various cultural differences. For instance, differential expression of short and long alleles of the serotonin transporter have been evident in different cultures, and Steve Suomi and colleagues have found that such differences are also present in various rhesus monkey sub-species (i.e., the majority of rhesus subspecies are laid-back and carry two copies of the long alleles while members of the most temperamental subspecies, Rhesus Macaques (Macaca mulatta), typically have one or two copies of the short alleles. Thus the similarities between genetically promoted temperaments in monkeys and humans are quite similar across different breeding populations and across cultures (for extensive reviews of this research see Kalueff, 2009). Thus, not only are affects filtered through culture but cultural differences may arise from genes that regulate affective temperaments. When one adds to this the dramatic changes in gene expression patterns in different environments (e.g., epigenetic, methylation and acetylation changes), the culturally induced sources of variability mount enormously (e.g., Roth, et al., 2009; Weaver, et al., 2006). For instance Steve Suomi’s group, with assistance from the ­Michael Meaney lab, has recently documented genome-wide methylation changes as a function of early experiences – mother rearing vs. nursery rearing in rhesus monkeys, and almost half the genome is affected (Unpublished data). Of course, the extensive work from Michael Meaney’s lab on the methylation genes that calibrate the reactivity of the pituitary-adrenal stress axis are of great relevance for both primary-process as well as tertiary-process approaches to understanding the neurobiology of emotionality. In sum, I suspect that cross-cultural differences in affective temperaments are influenced as much by genetic differences in affective tendencies molding cultures as they are by the traditional accepted direction of culture molding human affects. Environments where individuals are confronted by more challenging hardships,



2.  In defense of multiple Core Affects

can easily lead to the selective reproduction and survival of certain genotypes. The comparative ease of living in the south-sea islands, as opposed to the polarregions, are bound to promote the flourishing of different genotypes. Thus, if we consider all the evidence about the ways of nature and nurture, then we must remain open to two-way causation: genetically promoted affective strengths and weaknesses can promote cultural trends, and cultural differences and individual vicissitudes can magnify the importance of certain affects. With regard to the question of whether there are universal aspects of affect, neuroscience will perhaps offer clearer (more empirically-based) answers than psychology-only dimensional, componential and constructivist approaches. Affective neuroscience supports a view that an enormous amount of complexity is built into the spectrum of affective feelings – hunger, thirst, delight, disgust, the “Big 7”, etc. – that are our birthright.

9. What roles do the biological sciences and psychological sciences have in studying affect? Without mind sciences, it would be very difficult to identify nomothetic psychological processes that need to be understood through biological approaches. Conversely, it is hard to imagine that we will ever have an adequate understanding of such psychological processes without neurobiological approaches. It would be good if the two approaches could work in harmony, but historical-cultural differences among existing scientific traditions are hard to overcome. The foresight to consider both types of findings concurrently and to evaluate their utilities at different levels of analysis, with a clear focus on areas of overlap, can facilitate a lasting synthesis. In contrast, parochialism along with little cooperative thinking, can sustain polarization of views for a long time to come. In pursuing the dialectical approach, as we are here, perhaps a new synthesis can emerge. So let me re-iterate the affective neuroscience thesis: All pre-linguistic universals of the mind can only be defined coherently through an understanding of the underlying neurological processes. Words only suffice, with luck, to point us in the right direction. For instance, no matter what we say about the primal perception of “red” that all humans (with normally functioning color vision) experience, we will never capture the experience semantically. Even brain research may not close the explanatory gap completely, but it certainly can narrow it substantially. The dual-aspect monism view (for discussion see Panksepp, 2005b) encouraged by the discovery that emotional instinctual behavioral responses and their affects arise from the same ancient neural systems could be a substantial step in that direction.

63

64 Jaak Panksepp

My view is that cross-species neuroscience has finally narrowed the gap enormously between our intuitions about basic emotional feelings in humans, and key brain substrates (Panksepp, 1998b, 2005b, 2007b). The data strongly suggests a multiplicity of built-in feeling potentials in the brain. In contrast, Jim Russell (2003) has now developed the view that the psychological primitives in the field of affect studies are simply dimensions of positive and negative valence, and all agree that arousal is a big aspect of emotionality. Let me briefly focus on the latter where conceptual agreement might easily be found but solid evidence hard to obtain. We can perhaps agree that general arousal is a fundamental process mediated by a large number of neurochemical systems of the brain, and fortunately, we do not yet know the major known regulators (e.g., acetylcholine, dopamine, histamine, norepinephrine, and serotonin), while a host of neuropeptides from corticotropin releasing factor (CRF) to orexin may regulate certain more specific types of affective-emotional arousals. One could make a case that acetylcholine and CRF promote negative affective arousals more than positive ones, and dopamine and orexin promote positive arousals more than negative ones. In contrast norepinephrine probably modulates both about evenly. If such issues could be nailed down, it would constitute lasting cross-disciplinary scientific progress. Hence, I call upon all emotion-psychologists to consider evidence related to such neurochemical issues. It gets more complex for the dimension of valence, especially if there are various distinct types of primal positive and negative affects at the primary process level, with distinct neurochemical systems in the brain (e.g., neuropeptides) contributing heavily to which specific affects are aroused. The behavioral neuroscience database for specific modulation of distinct behavioral tendencies by neuropeptides is enormous, and many of these peptides also mediate conditioned place preferences and aversions. These are hard affective facts in the puzzle that needs to be solved. I call on psychologists interested in affects to consider these key issues.

10. What role might Panksepp’s primary affects play in Russell’s model of psychological construction? What implications might Russell’s model of psychological construction have for Panksepp’s construct of primary process affects? The apparent disagreements between Jim Russell and myself may amount to little more than the fact that the two of us have been approaching a difficult problem that interests us from two seemingly incommensurate perspectives – with Jim arriving at core issues from a top-down perspective, where the study of human



2.  In defense of multiple Core Affects

words and concepts has led the way, and with me proceeding from a bottom-up view, where one probes ancient brain territories that are critical for the emotionalaffective life of all mammals. Thus, in his earlier work, Jim promoted the reasonable possibility that certain widely used emotional categories emerge partly from learning, while I have shared abundant data indicating that complex subcortical networks are substrates for many distinct aspects of our emotional lives (in the vernacular, processes such as anger, fear, loneliness, sexual desire, maternal nurturance, social joy, and a generalized curiosity-enthusiasm-exploration-foraginginterest-expectancy system) along with many generalized arousal functions. At those distinct levels of analysis there need not have been any substantive differences between our views. We were simply plowing different territories in the vastness of the MindBrain. However, now that Jim has put forward the concept of intrinsic Core Affect, it is a new ball game. However, if there are core differences in our ideas, then they can only be cashed out in terms of differential predictions at the most relevant levels of analysis. In that spirit, let me consider how we can positively influence each other’s agendas.

Bottom-up issues: From Panksepp to Russell Right now, the postulation of Core Affect is a working hypothesis, and one immediately wants to know what kind of research programs might allow us to test the idea. Indeed, what kind of brain research would strongly support Russell’s hypothesis and what kind of evidence would constitute negations? If his view is going to advance substantive cross-disciplinary scientific inquiries that can lead to causal understanding, I suggest that Core Affect, not just the components “underneath,” has to be tested using neurobiological approaches. Psychology-only approaches cannot address the overall adequacy of the Core Affect proposal. The same would have applied to me had I remained in clinical psychology, where I started my graduate career. However, my decision to shift from human clinical training to animal psychobiology was premised on the likelihood that deep progress on these issues required relevant animal models. I saw no other realistic way to understand the affective foundations of human mind. Accordingly, I sought to cultivate a cross-species, neuroscientifically-defensible view that might one day illuminate primary-process emotional feelings in humans. Regrettably, the discourse between Russell and myself, till now, has officially been indirect – between two independent and non-interactive cultures. I thank the editors for facilitating a direct cross-cultural conversation. Successful dialog should eventually yield true reciprocity where we will, hopefully together, craft synergistic views that assimilate the relevant evidence we already have on both sides of the cultural divide.

65

66 Jaak Panksepp

To facilitate a pointed dialog and synthesis let me offer at least one potentially critical experiment: The most powerful positive affective systems discovered in the brain so far are endogenous opioids and dopaminergics, with some evidence for serotonin from the dorsal raphe also promoting a form of generalized positive affect. If their reinforcing properties operate through a homogenous shared higher mechanism for positive affect, then by using animal models, one could place sufficient amounts of the respective agents into the various brain regions where they have already been shown to generate “reward” (e.g., have “reinforced” operant choices, and mediated CPPs) and evaluate whether there are confluent “nodal networks” in evolutionarily more recent brain regions (as can be done directly with cfos immunocytochemistry in animals and perhaps estimated indirectly with diffusion tensor imaging in humans) and then determine if damage to those brain regions in animals attenuates, even eliminates, the rewarding effects of both the opioid and dopamine stimulants. If such higher brain areas exist, then one begins to be empirically justified in talking about the shared function of the two systems as operating through a possible unitary higher “positive affect” network. The same can be done for several distinct negative-affect manipulations such as acetylcholine, CRF, CCK and glutamate. If damage to those higher brain regions reduces general “reward” and “punishment” functions in the brain, Russell’s viewpoint will have been supported. If such research does not yield coherent evidence for global positive and negative affects, the concept of a shared supramodal network would be weakened, albeit perhaps not terminally disconfirmed. Conversely, mine will be negated if those kinds of damage markedly reduce all “reward” and “punishment” effects across different “reinforcing” objects throughout the brain. Of course, one could also imagine various human psychopharmacological/ psychoethological experiments, for instance, the careful study of drug-induced hedonic responses, as decoded through language about the affective changes caused by each agent. If one found an abundance of shared descriptors, one has some positive evidence for a generalized positive-affect mechanism in the human brain. If however, humans insist that those drugs are engendering highly distinct, even incommensurable sets of positive and negative feelings, one would have to rethink the Core Affect idea. Such experiments could also be done with various other human experiences, including sensory and homeostatic affects. In short, discriminating neuroscience types of experimental proposals would be most welcome. In my estimation, both human experience and neuroscientific analysis currently are consistent with the existence of multiple affective processes in human MindBrains. For instance, the pleasures of sensation, like the pleasing taste of sweetness, appear to be controlled by opioid reward mechanisms in specific parts of the brain (Peciña et al., 2006), and these are substantially different than the



2.  In defense of multiple Core Affects

rewarding properties of opioids overall, where one finds other brain regions (always quite low in the brain) that are rewarding by conditioned place-preference measures (Olmstead & Franklin, 1997). The positive affective (i.e., rewarding) properties of dopaminergics, are instigated from yet other brain regions such as medial frontal cortex (Alcaro, et al., 2007; Ikemoto, 2007). Can such findings be incorporated into a Core Affect conceptualization? I have much respect for psychology-only data, and my reading of evidence harvested for “basic emotion” views is as substantial as for traditional dimensional views. Indeed, there is good data to argue that point from some of our own data: I would note that our Affective Neuroscience Personality Scales (Davis­, et al., 2003; Davis & Panksepp, 2011), using language based self-reports, easily parses emotional-affective space into at least six distinct entities – 3 positive (seeking­, care, & play) and 3 negative feelings (fear, rage, panic/grief), each with their own statistically distinct factor structures (loadings). However, overall factor analysis also highlights that the above two groupings constitute overall positive and negative affect superfactors. Perhaps this can be integrated with Jim’s views, but I am not clear how.

Implications of Russell’s views for Panksepp to think about It is important to try to empirically distinguish Russell’s Core Affect view from the multiple affect view of Panksepp. One experiment I never thought of doing (item 9, vide supra) could offer the beginning of a resolution to the critical question of whether Russell’s proposal has neuropsychological merit. I don’t think any of Russell’s work is inconsistent with my long espoused view that neocortical regions of the brain are relatively free of affective circuitry; the neocortex is largely devoted to the mediation of higher cognitive processes, perfectly designed for dimensionalist as well as constructivist enterprises. It is within these cognitiveconceptual regions of the brain where one can construct emotional strategies useful for maximizing our affective experiences in the world. For my research, it might be useful for dimensionalists to promote better scientific languages for the higher mental entities that they typically study. For instance, I have already noted how I sought to generate a useful language for the study and discussion of primary-process emotions. At the very least, I hope the convention of capitalizing emotional primes has minimized “part-whole” confusions regarding the mind processes I am talking about. I suspect the majority of vernacular affective and emotional labels may remain scientifically ambiguous for a long time to come. However, here is where language-based theories of human emotions can make enormous contributions by developing some bridging rules,

67

68 Jaak Panksepp

which may require psychologically sophisticated investigators to analytically study human mental self-reports during the many kinds of deep-brain stimulation that modify affective feelings and are presently increasingly being conducted in humans because of the utility of deep-brain stimulation for treating various neurological and psychiatric disorders (Coenen, et al., 2011; Heath, 1996). Also, I hope Russell can suggest some additional ways that my views might be negated. For instance, if he is correct that emotions such as fear and anger are merely family resemblance terms, with no underlying functional “integrity”, might he be able to indicate why some of “my” primary-process systems are better members of the postulated families of basic emotions than others? I trust that the more we entertain such possibilities, the closer our views will become.

11. Considering the diversity of theoretical viewpoints in the scientific study of affective and emotional phenomena, might one hope for a coherent synthesis in this splintered field? The easy answer is: “One can only hope.” However, I do not think it can happen unless all become better immersed in the relevant data sets, and all recognize that there are distinct levels of emotional-affective processing in the MindBrain. I believe that any lasting synthesis requires both a behavioral and psychobiological understanding of instinctual aspects of emotions, but that behavior-alone, psychology-alone, neuroscience-alone and human-alone approaches cannot provide coherent primary-process neuroevolutionary levels of understanding upon which this field needs to be grounded. I have been repeatedly surprised that this has not become a more common view in recent emotion studies, especially in my own field of behavioral neuroscience. The newly-minted neuroscientists that started working on emotions in the 1980s were largely interested in secondary emotional-behavioral learning processes (e.g., fear conditioning). The increasing numbers of investigators that are studying the brain mechanisms of animal emotional learning (the secondaryprocess level), have made spectacular progress on the acquisition, extinction and reconsolidation of fear learning (e.g., Davis, et al., 2009; LeDoux, 2000; Maren, 2005). However, I remain puzzled by why they have spent so little time in trying to understand the unconditional-instinctual-affective responses of fear – studying the neural circuits where the affects “reside”. In my experience they even object to the discussion of affective issues in the animal models they use (hopefully that is changing now that clinicians have recognized the potential utility of their knowledge). The failure of primary-process neuroscience-based emotion views to have more prominence in psychological thinking is clearly due to the ­continuing



2.  In defense of multiple Core Affects

hegemony of behavioristic biases among those who study animal models of neurobehavioral controls. In contrast to the study of basic learning mechanisms, most “traditional” psychologists that entered the field seemed most fascinated by the psychophysiological correlates of emotions (perhaps because of the prevailing James-Lange perspective) as well as the nature of higher BrainMind regulatory mechanisms. All these levels of analysis can surely be blended conceptually and empirically. Perhaps a synthesis of relevant views can only be achieved by broadly trained, empirically-immersed neurophilosophers. Besides the critically important cross-species animal data, new human findings supplement the above conclusions. There are a growing number of examples where the primary-process animal work and the human brain imaging results are yielding concordant conclusions. The brain arousals (monitoring brain blood flow in PET studies) observed when humans feel sadness, fear, anger and joy, are highly concentrated subcortically (Damasio, et al., 2000) in brain regions where emotional operating systems are concentrated (Panksepp, 1998a). In studies of mild anxiety, the amygdala routinely lights up, but when clever investigators have generated more intense fear responses, the periaqueductal gray (PAG) becomes aroused as expected from primary-process emotional analyses (Mobbs, et al., 2007). Especially neat concordance has been found during human imaging of seeking system activities following animal models (Haber & Knutson, 2009), but the findings are still, all too often, conceptualized more in terms of non-affective behavioristic jargon – e.g., “reward-prediction error hypotheses” – rather than more naturalistic (ethological) emotional terms (Panksepp & Moskal, 2008). I think the field would become more coherent if investigators explicitly situated themselves more clearly at the level at which they are really studying such phenomena. Russell’s work started from the top and mine from the bottom, and I think we could easily achieve a coherent synthesis if we accept the limits of each approach. I find it a bit tragic for our field of science that by far the fewest investigators are currently working at the neuroscientifically focused primary process level. The intellectual territory carved out early on by basic emotion theorists (e.g., Tomkins, Ekman, Izard, Plutchik and Buck) was compelling from a primary-process vision of emotions but not from a tertiary process level. And that has been the rub. Even as certain investigators, usually working at the primaryprocess level, have pointed out ways to synthesize levels of understanding (Izard, 1994, 2007), others working usually at tertiary-process levels, accepting the seemingly endless human mental complexities (e.g., Barrett, 2006; Ortony & Turner, 1990), have found the primary-process level of analysis hard to swallow. But I do not think their critiques validly extend to our ancient emotional nature.

69

70 Jaak Panksepp

Still, many psychologically oriented investigators are currently not comfortable with animal research; this is understandable considering the long-term hegemony of behaviorism that diverted psychology from human complexities. They do not dispute the validity of animal findings, just the generalizability. There is no easy solution to this. Most investigators of the human condition recognize how impoverished a view animal research provides of human mental life, especially of our unique cognitive abilities (however, see Tomasello, 2009 who delightfully studies both young humans and chimpanzees). Hopefully they will increasingly see the value of the primary-process analysis that animal models provide. It is very important for most of their enterprises to be linked to detailed neuroscientific inquiries. This is not to deny the importance of a human oriented science of psychology. Thus, in closing I would like to briefly reflect on the vast cognitive territory of tertiary process emotions. The last few years have brought us a cornucopia of books on the cognitive and emotional subtleties of human minds and our brethren great apes (e.g., Cacioppo & Patrick, 2008; Decety & Ickes, 2009; deWaal, 2006; Hrdy, 2009; Iacoboni, 2008; Keltner, 2009; Rizzolati, 2008; Tomasello, 2009). With this level of enthusiasm, an affective neuroscientist like myself is coaxed to consider to what extent the massive social and emotional complexities of our family of anthropoid apes, especially humans, has developed unique emotional capacities that are encoded in our genome and the genetically dictated organizations of our brains (and bodies) as opposed to being emergents of accumulated general purpose brain expansions and culture. When we discuss the evolution of the human mind, to what extent are we indulging in the hubris of new creation stories, versus fact-based conclusions derived from solid neuroscience? How much do we really know about the evolution of the social MindBrain? Do the unique neocortical expansions of human brains contain new and intrinsic circuits for affective experiences that may be unique for our species? These and other momentous questions remain, to my knowledge, without much credible evidence one way or the other. Just to take one poignant and important example – the “mirror neurons” that have captured the imagination of all of us (see Iacoboni, 2008 and Rizzolati, 2008). At present, the most likely conclusion is that their functions are learned, but are there unique paths to that type of learning (promoting imitation, cooperation and empathy) within the human brain? We simply do not know. What we do know is that the human brain contains much more randomaccess­ type of memory (RAM) space among the “columns” and “minicolumns” – the banks of chip-like computational units of the mammalian neocortex. How are emotions represented in these circuits? We do not know. Are there “natural” paths to global positive and negative affects there? We don’t know. Are there ­fine-



2.  In defense of multiple Core Affects

grained evolutionary engravings that promote various ways of thinking? I don’t know. Nor am I aware of anyone else that does. Nor even a good strategy to get at these issues. With such doubts, the minimalist position advocated by Jim Russell is refreshing. I can imagine various intriguing hypotheses with the neuroscience facts that we already have at hand, but without the necessary functional data it is not worth sharing them at this time. Let it suffice to say that with a large enough RAM space and association fibers between different brain regions, the neocortex can become the home for deliberative thought, reason and the birth of language. The programming of the higher brain is probably achieved with the primary-process sub-neocortical emotional, motivational, perceptual and attentional circuits. In concert with all of the exteroceptively driven senses and perceptions, brains construct a variety of types of memories – procedural, declarative and the autonoeticautobiographical aspects of our higher selves. All this helps highlight how important dynamic brain views are for an understanding of how experience programs our higher mental apparatus. As higher brain systems assume their functional specializations, many will develop the ability to modulate, even repress, the operations of the lower affective circuits that I have studied. Others who have been confronted by developmental crises may not be so lucky. For them, most of the affective “action” may remain subcortically concentrated. There is an abundance of fine scientific literature to already consider. Let me simply note that early affective vicissitudes percolate throughout the maturing brain, and even across generations (Roth, et al., 2009). Throughout our discussion of how higher brain functions may elaborate emotionality, we should keep in mind one of the more amazing discoveries of modern neuroscience: Despite the best wishes of Evolutionary Psychologists, there is no evidence for genetically dictated evolutionary modules of any kind up within the neocortical expansions of our brain except perhaps for small genetically dictated differences in posterior sensory and frontal motor cortices. As I have emphasized in many writings, even the visual cortex derives it functional specialization developmentally and epigenetically (Sur & Rubenstein, 2005). This highlights the importance of individual learning and culture in the construction of our higher minds, and the need to make sure our kids get the best early start our culture can provide (Dunbar & Barrett, 2007; Narvaez, et al., 2012; Panksepp, 2007c; Worthman­, et al., 2010). From these developmental complexities, along with the ancestral complexities of primary-process affects, the higher cognitive mind is born to yield intentions, complex emotions and the deliberative forms of self-awareness, social imitation, cooperation, competition, and community – all the higher psychological and cultural complexities of the human MindBrain that psychologists continue to

71

72

Jaak Panksepp

describe from seemingly endless vantages. Whether any of that required the evolution of functionally dedicated higher cognitive-affective networks, before the onset of learning, remains a debatable issue. If new forms of Core Affect reside in those higher mental spaces, as dictated by evolutionary selection, we have some exciting surprises in store for us in affective neuroscience.

References Alcaro, A., Huber, R., & Panksepp, J. (2007). Behavioral functions of the mesolimbic dopaminergic system: an affective neuroethological perspective. Brain Research Reviews, 56, 283–321. Bailey, P., & Davis, E. W. (1942). Effects of lesions of the periaqueductal gray matter in the cat. Proceedings of the Society for Experimental Biology and Medicine, 351, 305–306. Bailey, P., & Davis, E. W. (1943). Effects of lesions of the periaqueductal gray matter on the Macacam mulatta. Journal of Neuropathology and Experimental Neurology, 3, 69–72. Barrett, L. F. (2006). Are emotions natural kinds? Perspectives on Psychological Science, 1, 28–58. Bernat, E., Bunce, S., & Shevrin, H. (2001). Event-related brain potentials differentiate positive and negative mood adjectives during both supraliminal and subliminal visual processing. International Journal of Psychophysiology, 42, 11–34. Bradley, M. M., & Lang, P. J. (1994). Measuring emotion: the Self-Assessment Manikin and the semantic differential. Journal of Behavior Therapy and Experimental Psychiatry, 25, 49–59. Bradley, M. M., & Lang, P. J. (2000). Measuring emotion: behavior, feeling, and physiology. In P. J. Lang & L. Nadel (Eds.). Cognitive neuroscience of emotion (pp. 242–276). NewYork: Oxford University Press. Brudzynski, S. M. (Ed.). (2009). Handbook of Mammalian Vocalization, Academic Press, Oxford, UK. Bunce, S. C., Bernat, E., Wong, P. S., & Shevrin, H. (1999). Further evidence for unconscious learning: preliminary support for the conditioning of facial EMG to subliminal stimuli. Journal of Psychiatric Research, 33, 341–347. Burgdorf, J. & Panksepp, J (2006). The neurobiology of positive emotions. Neuroscience and Biobehavioral Reviews, 30, 173–187. Burgdorf, J., Wood, P. L., Kroes, R. A., Moskal, J. R., & Panksepp, J. (2007). Neurobiology of 50-kHz ultrasonic vocalizations in rats: electrode mapping, lesion, and pharmacology studies. Behavioral Brain Research, 182, 274–283. Cabanac, M. (1992). Pleasure: the common currency. Journal of Theoretical Biology, 155, 173– 200. Cacioppo, J., & Patrick, W. (2008). Loneliness: human Nature and the Need for Social Connec­ tion. Norton. Chen, Q., Panksepp, J. B., & Lahvis, G. P. (2009). Empathy is moderated by genetic background in mice. PlosOne, 4(2), e4387. Coenen, V. A., Schlaepfer, T. E., Maedler, B., & Panksepp, J. (2011). Cross-species affective functions of the medial forebrain bundle-Implications for the treatment of affective pain and depression in humans. Neuroscience & Biobehavioral Reviews, 35(9), 1971–1981.



2.  In defense of multiple Core Affects

Damasio, A. R. (1994). Descartes’ error: Emotion, reason, and the human Brain. New York: Avon Books. Damasio, A. R. (1999). The feeling of what happens: Body and emotion in the making of con­ sciousness. New York: Harcourt Brace. Damasio, A. R., Grabowski, T. J., Bechara, A., Damasio, H., Ponto, L. L., Parvizi, J., Hichwa, R. D. (2000). Subcortical and cortical brain activity during the feeling of self-generated emotions. Nature Neuroscience, 3, 1049–1056. Darwin, C. (1872/1965). The expression of the emotions in man and animals. Chicago: University of Chicago Press. Davis, K. L., & Panksepp, J. (2011). The brain’s emotional foundations of human personality and the Affective Neuroscience Personality Scales. Neuroscience & Biobehavioral Reviews, 35(9), 1946–1958. Davis, K. L., Panksepp, J., & Normansell, L. (2003). The affective neuroscience personality scales: normative data and implications. NeuroPsychoanalysis, 5, 21–29. Davis, M., Walker, D. L., Miles, L., & Grillon, C. (2009). Phasic vs sustained fear in rats and humans: role of the extended amygdala in fear and anxiety. Neuropsychopharmacology Re­ views, 1–31. Decety, J., & Ickes, W. (2009). The Social Neuroscience of Empathy, MIT Press, Cambridge, MA. Delgado, M. R., Olsson, A., Phelps, E. A. (2006). Extending animal models of fear conditioning to humans. Biological Psychology, 73(1), 39–48. Denton, D. (2006). The primordial emotions: the dawning of consciousness. New York: Oxford University Press. de Waal, F. (2006). Primates and Philosophers: how Morality Evolved, Princeton University Press, Princeton, NJ. Dunbar, R. I. M., & Barrett, L. (eds.). (2007). The Oxford Handbook of Evolutionary Psychology, Oxford, UK: Oxford University Press. Edelman, D. B., & Seth, A. K. (2009). Animal consciousness: A synthetic approach. Trends in Neurosciences, 32(9), 476–484. Gendler, T. S. (2008). Alief and belief. The Journal of Philosophy, 634–663. Grimm, S., Beck, J., Schuepbach, D., Hell, D., Boesiger P., Bermpohl, F., Niehaus, L., Boeker, H., & Northoff G. (2008). Imbalance between left and right dorsolateral prefrontal cortex in major depression is linked to negative emotional judgment: An fMRI study in severe major depressive disorder. Biological Psychiatry, 63, 369–376. Grimm, S., Boesiger P., Beck, J., Schuepbach, D., Bermpohl, F., Walter, M., Ernst, J., Hell, D., Boeker, H., & Northoff., G. (2009). Altered negative BOLD responses in the default-mode network during emotion processing in depressed subjects. Neuropsychopharmacology, 34, 932–943. Haber, S. N., & Knutson, B. (2009). The reward circuit: linking primate anatomy and human imaging, Neuropsychopharmacology Reviews, 1–23. Heath, R. G., (1964). Pleasure response of human subjects to direct stimulation of the brain: Physiologic and psychodynamic considerations. In R. G. Heath (ed.). The Role of Pleasure in Human Behavior (pp. 219–243). Hoeber, New York. Heath, R. (1996). Exploring the mind-body relationship. Baton Rouge, LA: moran Printing. Hess, W. R. (1954). Diencephalon: autonomic and extrapyramidal functions. Grune and Stratton, New York.

73

74

Jaak Panksepp

Hess, W. R. (1957). The functional organization of the diencephalons. Grune & Stratton, New York. Hrdy, S. (2009). Mothers and Others: the Evolutionary Origins of Mutual Understanding Harvard University Press, Cambridge, MA. Iacoboni, M. (2008). Mirroring people: the science of empathy and how we connect with others. Farrar, Strauss and Giroux, New York. Ikemoto, S. (2007). Dopamine reward circuitry: two projection systems from the ventral midbrain to the nucleus accumbens–olfactory tubercle complex. Brain Research Review, 56, 27–78. Izard, C. E. (1994). Innate and universal facial expressions: evidence from developmental and cross-cultural research. Psychological Bulletin, 115, 288–299. Izard, C. E. (2007). Basic emotions, natural kinds, emotion schemas, and a new paradigm. Per­ spectives on Psychological Science, 2, 260–280. Kalm, L. M., & Semba, R. D. (2005). They starved so that others be better fed: remembering Aancel Keys and the Minnesota Experiment. Journal of Nutrition, 135, 1347–1352. Kalueff, J. F. (2009). Experimental models in serotonin transporter research. Cambridge University Press, Cambridge, UK. Keltner, D. (2009). Born to Be Good: the Science of a Meaningful Life, Norton, New York. Keys, A., Brozek, J., Henschel, A., Mickelsen, O., & Taylor, H. L. (1950) The Biology of Human Starvation, Vols. I–II. University of Minnesota Press, Minneapolis, MN. Kilpatrick, D. G,, Koenen, K. C., Ruggiero, K. J., Acierno, R., Galea, S., Resnick, H. S., Roitzsch, J., Boyle, J., & Gelernter, J. (2007). “The serotonin transporter genotype and social support and moderation of posttraumatic stress disorder and depression in hurricane-exposed adults”. The American Journal of Psychiatry, 164, 1693–1699. Knutson, B., Burgdorf, J., & Panksepp, J. (2002). Ultrasonic vocalizations as indices of affective states in rat. Psychological Bulletin, 128, 961–977. Koenigsberg, H. W., Siever, L. J., Lee, H., Pizzarello, S., New, A. S., Goodman, M., Cheng, H., Flory, J., & Prohovnik, I. (2009a). Neural Correlates of Emotion Processing in Borderline Personality Disorder. Psychiatry Research: Neuroimaging, 172, 192–199. Koenigsberg, H. W., Fan, J., Ochsner, K., Liu, X., Guise, K. G., Pizzarello, S., Dorantes, C., ­Guerreri, S., Tecuta, L., Goodman, M., New, A., & Siever, L. J. (2009b). Neural Correlates of the Use of Psychological Distancing to Regulate Responses to Negative Social Cues: a Study of Patients with Borderline Personality Disorder. Biological Psychiatry, 66(9), 854–863. Koziol, L., & Buddings, D. (2009). Subcortical structures and cognition: Implications for neuro­ psychological assessment, Springer, New York. LeDoux, J. E. (2000). Emotion circuits in the brain. Annual Review of Neuroscience, 23, 155–184. Liotti, M., & Panksepp, J. (2004). On the neural nature of human emotions and implications for biological psychiatry. In Panksepp J. (ed.). Textbook of Biological Psychiatry (pp. 33–74). Wiley, New York. Machery, E. (2009). Doing without concepts. Oxford University Press, New York. Maren, S. (2005). Building and burying fear memories in the brain. Neuroscientist, 11, 89–99. Merker, B. (2007). Consciousness without a cerebral cortex: a challenge for neuroscience and medicine. Behavioral and Brain Sciences, 30, 63–134. Mobbs, D., Petrovic, P., Marchant, J. L., Hassabis, D., Weiskoft, N., Seymour, B., et al. (2007). When fear is near: threat imminence elicits prefrontal-periaqueductal gray shifts in humans. Science, 317, 1079–1083.



2.  In defense of multiple Core Affects

Mobbs, D., Marchant, J. L., Hassabis, D., Seymour, B., Tan, G., Gray, M., Petrovic, P., Dolan, R. J, & Frith, C. D. (2009). From threat to fear: the neural organization of defensive fear systems in humans. Journal of Neuroscience, 29, 12236–43. Moors, A. (2009). Theories of emotion causation: a review. Cognition & emotion, 23, 625–662. Narvaez, D., Panksepp, J., Schore, A., & Gleason, T. (Eds.). (2012), in press. Evolution, Early Experience and Human Development: from Research to Practice and Policy. New York: Oxford University Press. Northoff, G., Heinzel, A., Bermpohl, F., Niese, R., Pfennig, A., Pascual-Leone, A., et al. (2004). Reciprocal modulation and attenuation in the prefrontal cortex: an fMRI study on emotional–cognitive interaction. Human Brain Mapping, 21, 202–212. Northoff, G., Schneider, F., Rotte, M., Matthiae, C. Tempelmann, C., Wiebking, C., Bermpohl, F., Heinzel, A., Danos, P., Heinze, H. J., Bogerts, B., Walter, M., & Panksepp, J. (2009). Differential parametric modulation of self-relatedness and emotions in different brain regions. Human Brain Mapping, 30, 369–382. Northoff, G., & Panksepp, J. (2008). The trans-species concept of self and the subcortical-cortical midline system. Trends in Cognitive Sciences, 12, 259–264. Olmstead, M. C., & Franklin, K. B. J. (1997). The development of a conditioned place preference to morphine: effects of microinjections into various CNS sites. Behavioral Neuroscience, 111, 1324–1334. Ortony, A., & Turner, T. J. (1990). What’s basic about basic emotions? Psychological Review, 97, 315–331. Panksepp, J. (1985). Mood changes: In P. J. Vinken, G. W. Bruyn, & H. L. Klawans. (Eds ). Handbook of Clinical Neurology (Revised Series). Vol. 1. (45): Clinical Neuropsychology. (pp. 271–285). Amsterdam: Elsevier Science Publishers. Panksepp, J. (1990). Can “mind” and behavior be understood without understanding the brain?: a response to Bunge. New Ideas in Psychology, 8, 139–149. Panksepp, J. (1992). A critical role for “affective neuroscience” in resolving what is basic about basic emotions. Psychological Review, 99, 554–560. Panksepp, J. (1995). The emotional sources of “chills” induced by music. Music Perception, 15, 171–207. Panksepp, J. (1998a). Affective Neuroscience. The foundation of human and animal emotions. Oxford University Press, New York. Panksepp, J. (1998b). The periconscious substrates of consciousness: affective states and the evolutionary origins of the SELF. Journal of Consciousness Studies, 5, 566–582. Panksepp, J. (1999). Emotions as viewed by psychoanalysis and neuroscience: an exercise in consilience, and accompanying commentaries. NeuroPsychoanalysis. 1: 15–89. Panksepp, J. (2002). Foreword: the MacLean legacy and some modern trends in emotion research. In G. A. Cory, Jr. & R. Gardner, Jr. (Eds.). The Evolutionary Neuroethology of Paul MacLean, (pp. ix–xxvii). Praeger, Westport, CT. Panksepp, J. (2003). At the interface between the affective, behavioral and cognitive neurosciences: decoding the emotional feelings of the brain. Brain and Cognition, 52, 4–14. Panksepp, J., (2005a). Affective consciousness: core emotional feelings in animals and humans. Consciousness & Cognition, 14, 30–80. Panksepp, J. (2005b). Affective consciousness: Core emotional feelings in animals and humans. Consciousness & Cognition, 14, 19–69. Panksepp J. (2007a). Neuroevolutionary sources of laughter and social joy: modeling primal human laughter in laboratory rats. Behavioral Brain Research. 182, 231–44.

75

76

Jaak Panksepp

Panksepp J. (2007b). Affective Consciousness. In M. Velmans & S. Schneider. (eds.). The Black­ well Companion to Consciousness (pp. 114–129). Malden, MA: Blackwell Publishing, Ltd. Panksepp, J. (2007c). Can PLAY diminish ADHD and facilitate the construction of the social brain. Journal of the Canadian Academy of Child and Adolescent Psychiatry, 10, 57–66. Panksepp, J. (2009). Emotional causes and consequences of social-affective vocalization. In Brudzynski, S. M. (Ed.). Handbook of Mammalian Vocalization (pp. 201–208). Oxford, UK: Academic Press. Panksepp, J. (2010). Energy is delight: the affective pleasures and pains of brain system for eating and energy regulation In L. Dube, A. Bechara, A. Drewnowski, J. LeBel, P. James, & R. Y. Yada. (Eds.). Obesity Prevention Handbook, (pp. 5–14). New York: Elsevier. Panksepp, J. & Bernatzky, G. (2002). Emotional sounds and the brain: the neuro-affective foundations of musical appreciation. Behavioural Processes, 60, 133–155. Panksepp, J. & Burgdorf, J. (2003). “Laughing” rats and the evolutionary antecedents of human joy? Physiology & Behavior, 79, 533–547. Panksepp, J., & Gordon, N. (2003). The instinctual basis of human affect: affective imaging of laughter and crying. Consciousness & Emotion, 4, 197–206. Panksepp, J., & Harro, J. (2004). The future of neuropeptides in biological psychiatry and emotional psychopharmacology: goals and strategies. In Panksepp J. (ed.). Textbook of Biologi­ cal Psychiatry, pp. 627–660. New York, Wiley. Panksepp, J., Knutson, B., & Burgdorf, J. (2002). The role of emotional brain systems in addictions: a neuro-evolutionary perspective. Addiction, 97, 459–469. Panksepp, J. & Moskal, J. (2008). Dopamine and SEEKING: subcortical “reward” systems and appetitive urges. In A. Elliot. (ed.). Handbook of approach and avoidance motivation, (pp. 67–87). New York: Taylor & Francis Group, LLC. Panksepp, J., Normansell, L. A., Cox, J. F., & Siviy, S. (1994). Effects of neonatal decortication on the social play of juvenile rats. Physiology & Behavior, 56, 429–443. Panksepp, J., & Northoff, G. (2009). The trans–species core self: the emergence of active cultural and neuro-ecological agents through self related processing within subcortical-cortical midline networks. Consciousness & Cognition, 18, 193–215. Panksepp, J., Sacks, D. S., Crepeau, L., & Abbott, B. B. (1991). The psycho- and neuro-biology of fear systems in the brain. In M. R. Denny (Ed.). Aversive Events and Behavior (pp. 7–59). New York: Lawrence Erlbaum Assocs. Inc. Publisher. Panksepp, J., Shevrin, H., Brakel, L., & Snodgrass, M. (2004). On Differences Between Conscious and Unconscious Affect and Cognition. Toward a Science of Consciousness, April 2004, Tucson, AZ. Parvizi, J., & Damasio, A. (2001). Consciousness and the brainstem. Cognition, 79, 135–160. Peciña, S., Smith, K., & Berridge, K. C. 2006. Hedonic hot spots in the brain. The Neuroscientist, 12, 500–511. Reddy, V. (2008). How infants know minds. Cambridge, MA: Harvard University Press Rilling, J. K., Winslow, J. T., & Kilts, C. D. (2004). The neural correlates of mate competition in dominant male rhesus macaques. Biological Psychiatry, 56, 364–375. Rizzolati, I. (2008). Mirrors in the Brain: How Our Minds Share Actions, Emotions, and Experi­ ence, Oxford University Press, Oxford UK. Ross, M. D., Owren, M. J., & Zimmermann, E. (2009). Reconstructing the Evolution of Laughter in Great Apes and Humans. Current Biology, 19, 1106–1111. Rossi, J., III & Panksepp, J. 1992. Analysis of the relationships between self-stimulation sniffing and brain-stimulation sniffing. Physiology and Behavior, 51, 805–813.



2.  In defense of multiple Core Affects

Roth, T. L., Lubin, F. D., Funk, A. J., & Sweatt, J. D. (2009). Lasting epigenetic influence of earlylife adversity on the BDNF Gene. Biological Psychiatry, 65, 760–769. Russell, J. A. (2003). Core Affect and the psychological construction of emotion. Psychological Review, 110, 145–173. Schnall, S., & Laird, J. D. (2003). Keep smiling: enduring effects of facial expressions and postures on emotional experience and memory. Cognition & Emotion, 17, 787–797. Sehlmeyer, C., Schöning, S., Zwitserlood, P., Pfleiderer, B., Kircher, T., Arolt, V., & Konrad, C. (2009). Human fear conditioning and extinction in neuroimaging: a systematic review. PLoS One. 4(6): e5865. Sem-Jacobson, C. W. (1968). Depth-electroencephalographic stimulation of the human brain and behavior. Springfield, IL: Charles C. Thomas Publishing. Sewards, T. V. (2004). Dual separate pathways for sensory and hedonic aspects of taste.Brain Research Bulletin, 62, 271–83. Shamay-Tsoory, S. G., Tibi-Elhanany, Y., & Aharon-Peretz, J. (2007). The green-eyed monster and malicious joy: the neuroanatomical bases of envy and gloating (schadenfreude). Brain, 130, 663–78. Shamay-Tsoory, S. G., Fischer, M., Dvash J., Harari, H., Perach-Bloom, N., & Levkovitz, Y. (2009). Intranasal administration of oxytocin increases envy and schadenfreude (gloating). Biological Psychiatry, 66, 864–870. Shevrin, H. (2001). Event-related markers of unconscious processes. International Journal of Psychophysiology, 42, 209–18. Shewmon, D. A., Holmse, D. A., & Byrne, P. A. (1999). Consciousness in congenitally decorticate children: developmental vegetative state as self-fulfilling prophecy. Developmental Medicine and Child Neurology, 41, 364–374. Snodgrass, M. & Shevrin, H. (2006). Unconscious inhibition and facilitation at the objective detection threshold: replicable and qualitatively different unconscious perceptual effects. Cognition, 101, 43–79. Soltysik, S., & Jellen, P. (2005). In rats, sighs correlate with relief. Physiology & Behavior, 85, 598–602. Sur, M., & Rubenstein, J. L. (2005). Patterning and plasticity of the cerebral cortex. Science, 310, 805–810. Steiner, J. E., Glaser, D., Hawilo, M. E., & Berridge, K. C. (2001). Comparative expression of hedonic impact: affective reactions to taste by human infants and other primates. Neurosci­ ence and Biobehavioral Reviews, 25, 53–74. Strack, F., Leonard, M., & Stepper, S. (1988). Inhibiting and facilitating conditions of the human smile: a nonobtrusive test of the facial feedback Hypothesis. Journal of Personality and Social Psychology, 54, 768–777. Stutz, R. M., Rossi, R. R., Hastings, L., & Brunner, R. L. (1974). Discriminability of intracranial stimuli: the role of anatomical connectedness. Physiology & Behavior, 12, 69–73. Suomi, S. J. (2006). Risk, resilience, and gene x environment interactions in rhesus monkeys. Annals of the New York Academy of Sciences, 1094, 52–62. Thorndike, E. (1911). Animal Intelligence. New York: MacMillan. Tomasello, M. (2009). Why we cooperate. MIT Press, Cambridge, Mass. Vandekerckhove, M. & Panksepp, J., (2009). The flow of anoetic to noetic and autonoetic consciousness: a vision of unknowing (anoetic) and knowing (noetic) consciousness in the remembrance of things past and imagined futures. Consciousness & Cognition, doi:10.1016/ j.concog.2009.08.002

77

78

Jaak Panksepp

van der Wee, N. J., van Veen, J. F., Stevens, H., van Vliet, I. M., van Rijk, P. P., & Westenberg, H. G. (2008). Increased Serotonin and Dopamine Transporter Binding in Psychotropic Medication–Naïve Patients with Generalized Social Anxiety Disorder Shown by 123I-β(4-Iodophenyl)-Tropane SPECT. The Journal of Nuclear Medicine, 49, 757–763. Watt, D. F., & Pincus, D. I. (2004). Neural substrates of consciousness: implications for clinical psychiatry. In J. Panksepp. (Ed.). Textbook of biological psychiatry (pp. 627–660). Hoboken, NJ: Wiley. Weaver, I. C. G., Meaney, M. J. & Szyf, M. (2006). Maternal care effects on the hippocampal transcriptome and anxiety-mediated behaviors in the offspring that are reversible in adulthood. Proceedings of the National Academy of Sciences, 103, 3480–3485. Wegner, D. M. (2002). The Illusion of Conscious Will. MIT Press: Cambridge, MA. Wong, P., Shevrin, H., & William, W. J. (1994). Conscious and nonconscious processes: an ERP index of anticipatory response in a conditioning paradigm using visually masked stimuli. Psychophysiology, 31, 87–101. Worthman, C. M. , Plotsky, P. M., Schechter, D. S., & Cummings, C. A. (eds.). (2010). Formative experiences: the interaction of caregiving, culture, and developmental psychobiology. New York: Cambridge University Press. Wu, M. V., Manoli, D. S., Fraser, E. J., Coats, J. K., Tolikuhn, J., Honda, S.-I., Harada, N., & Shah, N. M. (2009). Estrogen masculinizes neural pathways and sex-specific behaviors. Cell, 139, 61–72. Zinken, J., Knoll, M., & Panksepp, J. (2008). Universality and diversity in the vocalization of emotions. In K. Izdebski. (ed.). Emotions in the human voice, Vol. 1. Foundations. (pp. 185– 202). San Diego, CA: Plural Publishing Inc.

chapter 3

From a psychological constructionist perspective James A. Russell

We are not going to develop a successful science of emotion by sticking closely to our traditional ways of thinking. Here I briefly summarize a non-traditional alternative conceptual framework that I am developing and that is described elsewhere in greater detail along with its empirical support (Russell, 2003a, 2005, 2006, 2008, 2009). That framework is still under construction; for example, biological and social-cultural levels of analysis remain to be added. For some questions posed by Peter Zachar and Ralph Ellis, my answers are aimed at clarifying, refining, and developing that framework. I am disheartened by how often my previous attempts at articulating my ideas have been misunderstood, and I am grateful for the opportunity to clarify. My initial reaction to other questions, however, was “I don’t know!” Some questions raise issues that are not addressed by my framework. I decided to respond to the questions anyway, in the belief that doing so would help reveal my underlying assumptions and in the hope that my speculations would be taken as steps in a continuing dialogue. I have not inserted phrases such as “I suppose” or “I guess” in enough places, and perhaps the reader will kindly insert them where needed. My answers to the questions posed will be clearer if I make explicit several background assumptions and establish some terminology, before outlining my overall perspective on human emotion, a perspective I call Psychological Construction. The framework that I am proposing is an alternative to the standard, age-old approach to the topic of emotion in which the focus is on a set of discrete emotions, each named by a common everyday word such as fear, anger, happiness, and so on. I have no philosophical argument against the notion of discrete emotions. Instead, my argument against that approach is empirical. Over the last fifty years, that approach was articulated into a highly productive research program in psychology known as Basic Emotion Theory. Much evidence has now accumulated, and that evidence has uncovered problems in the program’s concepts

80 James A. Russell

and major propositions. No one or two pieces of negative evidence are fatal, but cumulatively the prognosis for that type of theory is poor. I am deeply concerned about the words that we use in the science of emotion. Affect, emotion, mood, feeling, anger, happiness, fear, and so on are all everyday words, some quite ancient, some surprisingly recent, pressed into scientific service, as happens in the early stages of scientific investigation of most domains. These words, while a starting place, have become a source of difficulty in furthering a scientific analysis of emotion. These words bring with them the assumptions of our religious, philosophical, and intellectual traditions, including such dubious distinctions as that between body and mind or that between reason and emotion. As culture-specific, historically shifting, family resemblance cluster concepts, they lack defining properties, leaving no principled way to resolve questions about whether certain cases are or are not instances of emotion. We have two tasks, not completely unrelated but not always sufficiently distinguished (Russell, 1991b). Analysis of what such words mean in everyday discourse (their intension) is one task; analysis of the events to which those words refer (their extension) is another. Consider the first task. As everyday words, emo­ tion, anger, fear and the like are labels in the English language for mental concepts that play a role in everyday life. Like other folk terms found on any page of the dictionary, such as wallet, water, winter, and wizard, these terms are parts of psychological and cultural processes and practices. Each has an interesting history of changing meanings, sometimes under the influence of scientific discoveries. As labels for folk concepts, they must be understood, for these concepts influence thought and action. The second task – the scientific study of the events to which words such as emotion, affect, and jealousy refer – is altogether different. This task is the focus of the questions asked in this project. It is for this task that I have become skeptical about the ability of everyday words to play the technical role scientists have asked them to play. When treated as scientific terms, our everyday folk words bring fuzziness, hidden assumptions, and cultural specificity. At the least, they have to be used in science with great caution and with eyes open to the problems they import to the job. Panksepp and I share this suspicion of folk words. He wrote (2008: 402), “folk-psychological words are not good scientific labels.” He then used capital letters to name his basic neural circuits (fear, panic, seeking, etc.), although I worry that fear is commonly treated as fear. I have coined phrases such as Core Affect so that I could define them as precisely as I could; to mark the phrase’s newly minted quality, I capitalized the first letters. Even so, I worry that readers often treat Core Affect as a vernacular term. Scientific work on emotion often begins with this question: What is the definition of emotion? This endlessly discussed question is ambiguous. When emotion



3.  From a psychological constructionist perspective

is taken as a folk term, the question calls for a descriptive definition, and I have worked to uncover empirically its meaning and properties (e.g., Fehr & Russell, 1984). But that descriptive definition does not determine how emotion should be defined as a scientific term. When emotion is taken as a scientific term, the question calls for a prescriptive definition. My prescription is to use the term in an honorary role (such as a chapter title), but not to give it real scientific work to do. I use emotion as a way station, a sign pointing imprecisely toward many real and important events. My approach has many precedents. Science often approaches a difficult problem by breaking it into smaller parts each easier to solve. Like William James (1884), I use the tactic of searching for parts that can be delegated to another research specialty. James analyzed what he called emotion (by which he meant the subjective conscious experience of having an emotion) as a form of perception, thereby handing this part of the problem over to perceptionists. Perception of what? According to James, one’s emotion is a perception of one’s own bodily reaction, and he therefore handed this part of the problem over to those who study bodily reactions. And of course emotion begins with a perception of the “exciting fact” – another problem for perceptionists. In short, James analyzed emotion as a process consisting of parts and found no part that could not be handed over to another already established research specialty. By Occam’s razor, James thought that we should be unwilling to postulate a new entity (emotion) when already established ones (perception, bodily reaction) can explain the data.

1. Psychological construction In folk psychology and much of scientific psychology, emotion is assumed to arise from (or to be) a process qualitatively different from other psychological processes such as memory, cognition, perception, or behavior. This assumption underlies the search for a “theory of emotion” – one theory that explains all and only emotions and that is therefore separate from theories of other major psychological processes. Such a theory might look like this: All emotions are produced by (or are) a single type of entity (such as a set of affect programs in the limbic system). That entity produces (or is) an organized pattern of component responses. Consider a specific example of an emotional episode made famous by James: Sally’s response to the sudden appearance of a wild bear in her path through the woods last Saturday morning. In the standard account, fear (or the fear program) made Sally’s heart pound, her palms sweat, her face broadcast danger; fear focused her thoughts on the bear; it made her freeze and then flee. The evolutionary function of emotion is to produce just such an organized effect quickly and without much pondering.

81

82

James A. Russell

According to Psychological Construction, in contrast, emotion cannot be equated with any one process or component. Emotional episodes are not all of one kind, and they do not all stem from one mechanism dedicated just to emotion (unless perhaps we take the entire MindBrain, to use Panksepp’s term, to be that mechanism, but then the MindBrain produces everything psychological). Thus, to study emotional episodes is to study all of psychology with certain interests and questions in mind. Just as cognitive psychologists have abandoned the assumption that memory is a single entity, I suggest that we abandon the assumption that emotion is a single kind of entity or process. Psychological Construction thus does not offer one process as the explanation for emotion. It does not point to an affect program, perception of bodily reaction, Core Affect, arousal, appraisal, or attribution. It does not point to arousal + labeling or to Core Affect + an act of categorization. Perhaps more controversially, Psychological Construction makes the same claim about each discrete emotion such as fear and anger. Nevertheless, each token emotional episode (such as Sally’s encounter with the bear last Saturday) requires an explanation (cf. Moors, 2009). That explanation begins with a focus on the “components” – facial movement, vocal tone, peripheral nervous system change, appraisal, attribution, behavior, subjective experience, emotion regulation, etc. etc. – that occur in a particular instance of emotion. So, we need a theory of facial movement, a theory of vocal tone, and so on. Psychological Construction is an umbrella term for three sets of processes, those that produce: (a) the components, (b) the associations among these components, and (c) the categorization of the pattern of components as a specific emotion. Once these three sets of processes are understood, we have what we need. We need no extra mechanism. We need no emotion. Fear episodes, for instance, do not all stem from a single mechanism dedicated just to fear. Nonetheless, on Psychological Construction, each token occurrent event called fear requires an explanation. In the case of Sally’s encounter with the bear last Saturday, that explanation begins with each of the components (her heart pounding, palms sweating, etc). Each component has its own causal chain, which research has gone a fair way in describing. Thus, we need an explanation of how the encounter with the bear resulted in changes in the autonomic nervous system, but that explanation would not include the assumption that fear did it. And so on, through each component that actually occurred during the emotional episode. In the spirit of William James, Psychological Construction thus tries to divide emotion into parts and where appropriate pass a part to another research specialty. For example, we need an explanation for Sally’s overt behavior, and I have suggested Strack and Deutch’s (2004) dual-process model of behavior, a model that does not include the assumption that fear did it. Most parts can be passed on to another research specialty, but I suggest that Core Affect remains. Our emotional



3.  From a psychological constructionist perspective

lives are ones of agony or ecstasy, of excitement or lethargy, but you wouldn’t know that from reading much modern research on emotion. The patterning among the components too must be accounted for: Most of the components are on-going processes: we are continually appraising our surroundings; our autonomic nervous system is continuously functioning, our head and face move frequently, and so on. Therefore, the various components are always in some pattern. The components are more weakly linked than commonly assumed, but they are not completely independent, for three reasons: (a) features in the environment have a correlational structure, which then creates correlations among components, (b) one component can influence another, and (c) two components are correlated when they are both influenced by the same central mechanism such as attention. Finally, we need an explanation for how this pattern is seen (by a witness, a scientist, or by Sally herself) as a specific emotion. Occasionally, the pattern resembles the mental prototype we have for that emotion. Resemblance is a matter of degree, and therefore there are borderline, mediocre, good, and prototypical cases. When the pattern resembles a prototype/script sufficiently, the token may (or may not) be categorized as an instance of the emotion category. Psychological Construction thus provides directions for a specific research agenda. One critique of that agenda has been that, even if each component has its own causal chain, all the same, we must posit some additional mechanism – the emotion – to account for the coherence among the components for each discrete emotion. Coherence is a matter of degree, and this critique, like discrete emotion theories and folk psychology, presupposes a high degree of coherence within each emotion category. By coherence, I mean a high correlation in time and intensity among the components. In the science of emotional episodes, the degree of coherence is an empirical question, and sufficient evidence has now accumulated to have a tentative answer (Russell, 2003, 2006): correlations among components are surprisingly weak. In the mental prototype for fear, all the components cohere, but in reality, they rarely do. Rather than presupposing coherence, we must look to see what links actually exist. A similar response has been that even if the research program of Psychological Construction were successfully carried out, it would not explain the blueribbon­ cases of emotion. That is, Psychological Construction may be plausible for moods and mild non-prototypical instances of emotion, but not for the prototypical cases. My answer is that there is nothing about these prototypical cases that would require an additional explanation. They can be accounted for by the same general principles as the less prototypical cases. Consider an analogy from astronomy: How do we explain those special configurations we see in the sky as constellations? Beyond the laws of physics and statistics, astronomy does not need

83

84

James A. Russell

an additional explanation for the appearance of constellations. There is nothing special about the constellations. Another analogy comes from the game of poker: beyond the rules of the game and statistics, there is no need for an extra mechanism (except in cheating) to explain the very occasional occurrence of a royal flush. Still another analogy comes from the historically pervasive belief in vitalism, the belief that even after the chemistry is known, some additional force, some élan vital, is needed to explain the processes of life. Interestingly, young children’s naïve biology assumes vitalism (Inagaki & Hatano, 2004). The presupposition that emotion is a single specifiable process led to an interesting misinterpretation of Psychological Construction. In an overview of emotion theories, Scherer (2009) represented my account (Russell, 2003) with a flow chart consisting of four steps: (a) event, (b) production mechanism, (c) Core Affect, and (d) feeling. Of course, some emotional episodes fit the pattern so charted, but others do not. The claim of Psychological Construction is that there is no fixed order to events, no fixed number of components, and so on. Scherer’s representation of Psychological Construction also leaves out many events that are parts of some emotional episodes, including changes in peripheral physiology, perceptual and cognitive processes (especially appraisal and attribution), the mechanisms that lead to behavior, and emotion and affect regulation. Importantly, Psychological Construction also emphasizes an observer (a witness, the self, a scientist) who may (or may not) categorize the observed pattern of components as a type of emotion. In short, Psychological Construction is not another flow chart but rather a different perspective on the events in these flow charts. My research program may seem radical, but much actual research is already consistent with it. Many scientists have abandoned the search for the one process that IS the emotion and, instead, view emotion as a multi-component process, no one component of which defines the emotion; this perspective was first articulated by Averill (1982). Many scientists have abandoned the search for one mechanism to explain all emotions and turned instead to the search for one mechanism for each discrete emotion. Psychological Construction goes one step farther down this path. Indeed, much of the research seeking that mechanism for a discrete emotion has already taken that step in that it actually addresses only one or two individual components of the emotion. So interpreted, such research is part of my research program rather than at odds with it.



3.  From a psychological constructionist perspective

2. What is affect? What is its structure? Is affect fundamental, and if so, what makes it so? The word affect lacks an agreed upon definition. Affect is sometimes used in a broad sense capturing emotion, mood, feeling, or even any way in which a person is affected or moved. But affect is also used in a narrow sense limited to private subjective conscious feelings. I take the word affect here in this narrower sense. I distinguish three different types of affect, without claiming that these three exhaust the domain: Core Affect, Affective Quality, and Emotional Meta-experience. So defined, affect must be distinguished from what I call an emotional epi­ sode. Emotional episodes are all those particular (token) events referred to by the everyday English word emotion. An example of an emotional episode is Sally’s full response to the bear in the woods last Saturday. An emotional episode typically includes not only affect, but a sequence of other components as well, such as Sally’s perception of and appraisal of the bear, all the peripheral physiological changes that occurred, facial and vocal changes, her behavioral reaction of freezing and then fleeing, and all the neural processes that underlie these changes. In other words, an emotional episode includes both subjective and objective components of a particular person’s reaction in a particular place at a particular time. Common sense provides suggestions about affect. For example, we distinguish different types. People who speak English report feeling angry and afraid; those who speak Ifaluk report feeling fago and metagu. Affect is a larger class of events than just these experiences of specific emotions. That is, we subjectively experience not just anger, fear, fago, metagu, and so on, but also moods and simple affective reactions to events. We open our eyes and see an affectively charged world: the garden is calm, the music exhilarating, the person next to us pleasant. Affect has valence (sometimes you feel great, sometimes miserable). Affect varies in subjective sense of mobilization, from the lethargy of depression to the energy of mania. Affect seems to be largely involuntary, but we can find ways to alter it. All these common sense suggestions are, well, suggestions. We cannot ignore them, and we must account for why people believe them. We can formulate them as scientific hypotheses and seek evidence that would support or refute them. If the hypothesis is rejected, we then have the task of explaining why affect appears to us a certain way, even when reality is otherwise. Indeed, part of my proposal has just this feature of violating common sense and then trying to explain the discrepancy between appearance and reality. Core Affect. At any point in time, a person can answer the question, how do you feel? Studies of their answers have led to the concept of Core Affect. Core Affect

85

86 James A. Russell

is, at a psychological level, the most elementary simple primitive affective feeling. A map of Core Affect is seen in Figure 1, which shows a circumplex representation of self-reported moods and feelings. A person’s Core Affect at any one moment in time is represented by one point somewhere inside the space. The person thus has only one Core Affect at a time. The center can be thought of as an adaptation level (a neutral point midway between pleasure and displeasure and midway between low and high arousal), with distance from the center representing intensity or extremity of feeling. Core Affect can be extremely intense at times, milder at other times. Arousal Tension

Excitement

Displeasure

Pleasure

Depression

Serenity Sleep

Figure 1.  A schematic map of Core Affect, showing a circumplex model

Let me caution the reader against certain misinterpretations of Figure 1. Core Affect is defined by bipolar dimensions, but also can be broken into categories. For example, it includes feeling states that might be categorized as excitement (feels good + highly energized), serenity (feels good + low in energy), tension (feels bad + highly energized) and depression (feels bad + low in energy). In the previous sentence, the words excitement, serenity, tension, and depression are not used in their full senses, but only as short hand for the specified combinations of pleasure and arousal. In the same way, the terms shown on the periphery of many empirical versions of the circumplex are not used in their full sense. More generally, Core Affect is part of, but not the whole of, the states labeled in the circumplex. Put differently, Core Affect is not the set of all states listed in a figure, but rather a feature of all those states. Furthermore, all token events called fear, for example, do not fall at the same point in Core Affect space.



3.  From a psychological constructionist perspective

Searle (1992) suggested thinking of consciousness as a single field that is modified. If so, Core Affect is one property of the field, a property that changes from time to time. Like sensations and perceptions, Core Affect is experienced as involuntary: it is given rather than as chosen, but can be changed indirectly by creating conditions that alter it. A person is always in some state of Core Affect, just as a particular color falls at some point in the color space. That is, although its structural description requires two dimensions, Core Affect is a single feeling. This feature is parallel to the way that a description of a particular color requires three dimensions (hue, saturation, and brightness), but the sensation of a particular color is a single sensation. In both cases, dimensions combine in an integral fashion to form the experience. The conscious side of Core Affect is primary consciousness (Farthing, 1992). Here is my attempt at a definition of Core Affect: Core Affect is “a neurophysiological state that is consciously accessible as a simple, non-reflective feeling that is an integral blend of hedonic (pleasure–displeasure) and arousal (sleepy– activated­) values” (Russell, 2003: 147). This definition is a promissory note; it points to a consciously accessible experience and to whatever neurophysiological mechanism underlies that experience. Clearly, an essential step is specifying the mechanism so far left unspecified. I can mention some research moving toward filling in this piece of the puzzle. The neural basis of Core Affect in an active research concern (Gerber et al., 2008; Posner, Russell, & Peterson, 2005). Smith, Mahler, Pecuña, and Berridge (2010) suggested a subcortical pleasure circuit involving hot spots in the nucleus accumbens, ventral pallidum, and parabrachial nucleus of the pons. This circuit is involved in pleasures of all sorts, including sex, music, drugs, humor, and money. Leknes & Tracey (2008) point to evidence that mechanisms of pleasure and pain are closely related. Pfaff (2006) outlined a “generalized arousal system” that is “elementary and fundamental, primitive and undifferentiated and not derived from higher CNS functions” (p. 5). This arousal system is fundamental to alertness, attention, orientation, behavior, and “variations in moods and feelings that make up our emotional lives” (p. 2). The neurophysiological system Pfaff described is complex, but a key mechanism is centered in the brainstem. Smith et al. (2010) theorized that their subcortical pleasure circuit relays hedonic signals to the prefrontal cortex “perhaps for translation … into conscious feelings of pleasure and cognitive representations” (p. 36). Similarly, Kringelbach (2010) suggested that the orbital prefrontal cortex and the anterior cingulate cortex provide a global workspace that integrates information from other brain regions for access to consciousness. In this way, output from the Smith et al. pleasure circuit and the Pfaff arousal circuit may come to be integrated and made available to consciousness.

87

88

James A. Russell

If these various neurobiological hypotheses pan out, my definition of Core Affect would evolve into something like this: Core Affect is activity in the Smith et al. (2010) hedonic circuit and the Pfaff (2006) arousal circuit, activity which, by way of the prefrontal and cingulate cortices, is consciously accessible as a simple, non-reflective feeling that is an integral blend of hedonic (pleasure–displeasure) and arousal (sleepy–activated) values. Core Affect, again defined through verbal report of conscious experience, has also been shown to be related to changes in the autonomic nervous system, facial and vocal behavior, instrumental behavior, cognitive processes, reflexes, and a host of other things (Russell, 2003). Core Affect is clearly related to stimulants, depressants, and euphoric and dysphoric drugs. These relations are all empirical discoveries, rather than part of the definition of Core Affect. Core Affect is fundamental in the sense that it cannot be decomposed at a psychological level, although, of course, it can at a neurological level. Rather, Core Affect is an elemental building block of other psychological events. For example, it is a building block of an emotional episode: It is the hedonic tone and subjective sense of energy or mobilization in the episode. Speaking loosely, prototypes of various emotions can be characterized in terms of pleasure-displeasure, but not vice versa. Fear includes the anticipation of displeasure. Hope includes the anticipation of pleasure. Anger includes displeasure attributed to someone’s blameworthy action. And so on. (For more elaborate characterizations of individual emotion concepts, see Ortony, Clore & Collins, 1988; Wierzbicka, 2009.) Core Affect is not a substitute term for emotion, nor is it the essence of emotion, nor an additional discrete emotion. Emotional episodes are said to begin and then, after a short time, end. In contrast, one is always in some state of Core Affect, which simply varies over time (sometimes slowly, sometimes rapidly) without beginning or end. Emotional episodes are typically directed at something (one is angry with, afraid of, or sad about something). In contrast, Core Affect is not necessarily directed at anything. (In this regard, Core Affect is closer to the English word mood than to emotion, although the everyday concept of mood typically implies a long lasting and mild state, which Core Affect does not.) Core Affect per se can be free-floating (as in feeling down but not about anything and not knowing why), but it can come to be directed at something. The full experience that includes Core Affect and what it is directed at is thus intentional in the philosophical sense, in much the same way that the full experience of a pain and its bodily location is intentional (Searle, 1992: 251). Core Affect is, at a psychological level, a pre-conceptual primitive process. Indeed, I tentatively hypothesized that Core Affect has many of the features of modularity (Faucher & Tappolet, 2006): fast, mandatory, unique output, an



3.  From a psychological constructionist perspective

e­ volutionary explanation, dedicated brain circuitry, and encapsulation (Russell, 2008). This unitary view of Core Affect contrasts with some alternative views. For example, for the horizontal dimension of Figure 1, some have theorized many qualitatively different types of pleasure (multiplicity view) (Russell 2003b). Similarly, for the vertical dimension of arousal, others have theorized qualitatively different types of arousal. That Core Affect is encapsulated from general beliefs and desires is consistent with the commonsense idea that affect is involuntary. Core Affect does not change simply because we desire it to change or because we believe it will change. Instead, we must create conditions that then influence Core Affect. I return later to the topic of encapsulation. Core Affect is caused by both bottom-up and top-down processes (Ochner et al., 2009). Most (but not all) changes in Core Affect result from a single salient event. The pleasure and arousal elicited by the event are not intrinsic properties of the event but are provided by the observer. I also believe that a mechanism is needed to bind the change in Core Affect to the event that elicited it. Although we often have a good idea of why we feel as we do, we sometimes do not, as shown by free-floating emotions and moods and everyday feelings. Because there are usually too many influences on Core Affect for a person to track and because some of those influences are beyond our ability to detect (infrasound, for example), we can make mistakes about the cause of our Core Affect. Noting that people do not have conscious access to the origin of hedonic processes, Kringelbach (2010) suggested that people are “happy to confabulate about the causes” (p. 203). That is, in philosophers’ terms, our knowledge about our Core Affect is not incorrigible. Instead, we attribute our Core Affect to a specific cause (Weiner, 1985), but we can get the wrong cause, with interesting consequences on just what emotion we perceive in ourselves (Neumann, 2000). It is tempting to become poetic about the importance of Core Affect, but let me illustrate that importance by listing its functions. Core Affect is a continuous assessment of one’s current state, and it influences other psychological processes accordingly. A change in Core Affect evokes a search for its cause and therefore facilitates attention to and accessibility of like-valenced material. Core Affect thus guides cognitive processing according to the principle of mood congruency. The more positive is Core Affect, the more positive seem events encountered, remembered, or envisioned – provided that the Core Affect is not attributed elsewhere (Schwarz & Clore, 1983). Core Affect is part of the information used to estimate affective quality and thus is implicated in incidental acquisition of preferences and attitudes. Core Affect influences behavior from reflexes to complex decision making. One can seek to alter or maintain Core Affect directly – affect regula­ tion – from the morning coffee to the evening brandy. People generally (but not

89

90 James A. Russell

always) seek behavioral options that maximize pleasure and minimize displeasure. Decisions thus involve predictions of future Core Affect (March, 1978). Core Affect is involved in motivation, reward, and reinforcement. Affective Qualities. When we think about objects and events and their features, we think of them as beautiful, ugly, awful, soothing, and so on. We know not just their objective properties but their affective properties as well: how pleasant or unpleasant or how energizing or soporific something promises to be. Of course, beauty is actually in the eye of the beholder, but, subjectively, beauty seems to reside in the beautiful object. Typically, experiencing the pleasure of a pleasant event leads to coding its Affective Quality and this code is consolidated in long-term memory along with the episodic features of the event. To remember, to anticipate, or to perceive something as pleasant typically increases pleasure. Nevertheless, I want to allow the possibility that the awareness of an affective quality is a cold cognitive process that can occur even when Core Affect is not moved by the object or event. For example, someone may take an objectivist or intellectual stance toward the beauty of a piece of art without having Core Affect altered by the piece. So, I distinguish Affective Quality from Core Affect and treat as an empirical hypothesis the proposition that the two are related. (In their neurobiological account, Smith et al., 2010, made a similar distinction between neural mechanisms that cause pleasure and those that code pleasure in support of some other function such as decision making.) A structural description of Affective Qualities looks like the same circumplex as Core Affect (Russell & Pratt, 1980). Perceptions of affective quality are also fundamental. They cannot be decomposed at a psychological level, although they can at a neurological level. They are building blocks in our representations of objects and events. Emotional Meta-experience. From time to time, a person consciously experiences having a specific emotion, such as fear, anger, jealousy, enthusiasm, awe, love, and so on. Indeed, at one time, the conscious experience of emotion was equated with the emotion. The phrase Emotional Meta-experience is my way of characterizing this type of affect. Like James (1884), I contend that Emotional Meta-experience is a perception: to feel afraid is to perceive oneself as afraid. My account thus resonates with Gazzaniga’s (1985) interpreter theory of consciousness, in which conscious experience includes an interpretation, a making sense of and sometimes even a confabulation about, the raw data of experience. Perceiving oneself as having an emotion is as complex as other perceptions. Percepts are often compelling, but they are not simple. First, they are complex because they are, in the philosopher’s term, intentional mental states; that is, they



3.  From a psychological constructionist perspective

are about something (Solomon, 1976). In our example, Sally’s fear is not freefloating. Rather, Sally is afraid of the bear, or, more precisely, that the bear will harm her. One is not just in love, but in love with someone or something. Second, they are complex because they are meta-experiences. That is, the experience includes other experiences, such as Core Affect, somatosensory feedback, appraisal of the eliciting event, attribution, beliefs, desires, plans, impulses, and behavior. The meta-experience includes these ingredients in the way that the perception of a family dinner includes the food, music, individuals, setting, and so on as parts of the whole. Similarly, Lambie and Marcel (2002) characterized the experience of having an emotion as a second-order experience – that is, an experience that depends on and emerges out of first-order experiences. In Farthing’s (1992) terms, Emotional Meta-experience is a secondary level of consciousness. Third, an Emotional Meta-experience is complex because, like any other perception, it is the end product of a complex process. It is not entirely driven by the raw data, but depends on prior expectations, attention, and categorization: the process is not just bottom-up but also top-down. To perceive is to form a representation and involves concepts, learning, attention, and memory (Smith and Neumann, 2005). One consequence is that, in an emergency, attentional resources may be focused on, for example, the wild bear in the path and on discovering an escape route, with none left to form the self-perception of being afraid. Thus, the experience of being afraid can occur late in or even after the emotional episode, or never. Finally, like other perceptions, an Emotional Meta-experience is an interpretation that can be elusive or mistaken. In some cases, persons try to understand their own emotional episode, but cannot: Is this really love, or just infatuation? Was I angry, or jealous? Is this shame I’m feeling, or fear? I should be feeling grief, but am I? The Emotional Meta-experience can also be “wrong” in the sense that the person later admits that he was indeed jealous, although he didn’t realize it at the time. Or, wrong in the sense that others think it obvious that he’s jealous no matter how vehemently he denies it. Put differently, emotional episodes and Emotional Meta-experience overlap, but not completely. LeDoux (2008) offered a detailed model of Emotional Meta-experience (his term was feeling) involving working memory. A single representation (such as a feeling of fear) is created by integrating sensory properties of the eliciting stimulus with cognitively processed information about it, long-term memories activated by it, and input of low-level emotional processes. That single representation must be attended to and maintained in working memory if it is to become a conscious feeling. On my account of Emotional Meta-experience, to perceive oneself as afraid is to categorize oneself by means of the concept of fear. It is to establish the meaning

91

92

James A. Russell

of one’s state via the concept of fear. In turn, the concept of fear can be unpacked as a mental script laying out a series of subevents (the components) in a temporal and causal order (Fehr & Russell, 1984). Some emotion theorists have now abandoned the search for the one feature that defines the concept of fear (or any other emotion) and write of each emotion instead as a “multicomponent process.” To perceive oneself as afraid is to see a resemblance between one’s current state and the mental script for that emotion. A separate hypothesis is that there are both similarities and differences in emotion concepts across cultures and languages. We return to this topic in a later question. Emotional Meta-experience does not lend itself to a universal descriptive structure, because it depends on an uncountable number of ingredients, some of which, especially concepts, are influenced by culture or even more personal experiences. English contains hundreds of terms for types of emotion, but other languages contain fewer, perhaps just a handful (Russell, 1991). The concepts for separate emotions can be ordered in a fuzzy hierarchy. That is, broad generic terms, such as, in English, emotion can be subdivided into basic-level concepts such as fear, anger, love, and so on. Fear, in turn, can be subdivided into panic, anxiety, apprehension, and so on. I call this hierarchy “fuzzy” because there is no rigid class inclusion system that it represents. Rather, the concepts vary in breadth and degree of overlap. Conclusion. To study affect is to study consciousness with certain interests and questions in mind. Thinking of consciousness as a collection of separate items is less helpful than thinking of consciousness as a single field. Consciousness at a given moment is typically a single Gestalt that has a single Core Affect, various objects and events each with a perceived affective quality, and other ingredients, all organized into one meaningful whole. When it occurs, an Emotional Metaexperience­ of fear, for example, is not another element in the collection; it is not just the isolated experience, ah fear! – instead, it is a Gestalt including all the ingredients on which that perception is based. Izard (1977) offered the insight that a person is always in a state of emotion. The analysis here comes to a similar conclusion. Affect is always present. Emotional Meta-experience occurs only sporadically, but is only one kind of affect. Core Affect is always present, always accessible, even if not attended to. Perceived objects and events and their features have affective qualities. Some states are more prototypical emotional episodes than are others, and a person only occasionally forms the percept of having an emotion. A person always has at least some of the ingredients of an emotional episode: the autonomic nervous system is always active; the face moves frequently; we constantly appraise our surroundings; we are constantly behaving, and so on. Whether the pattern of ingredients constitutes



3.  From a psychological constructionist perspective

an emotional episode is the question of whether the pattern crosses a subjective threshold of resemblance to a mental prototype. Cultures form different prototypes; different individuals within a culture have different thresholds.

3. How does affect relate to prototypical emotional episodes? Prototypical Emotional Episodes. Emotional episodes are real events, whether or not they are labeled or categorized. They exist out there in the real world as the referents of our concept of emotion. Because this definition of an emotional epi­ sode relies on the folk concept of emotion, the category of emotional episodes may not cut nature at its joints. Still, all its referents must be explained. Also because the definition relies on a folk concept, it is not surprising that some emotional episodes are better exemplars of emotion than are others – just as some shades of red are better examples of red than are others: fire-engine red is a more prototypical red than is the brownish red of a robin’s breast, which some people might not judge to be red at all. Emotional episodes can be arrayed along a continuum of their degree of prototypicality within the concept of emotion, the degree to which they resemble the prototypes of the concept. At one end are prototypical, blueribbon cases, which come readily to mind when the word emotion is mentioned. Mediocre cases fall in the middle, and borderline cases fall at the other end fading off into non-emotions with no clear boundary to be found. This continuum of prototypicality is a property of our folk system of categories and not of the emotional episodes themselves. Thus, as in the case of fire engines and robins, prototypical cases are not necessarily more natural than mediocre cases. The membership of a borderline case in the category of emotion is difficult to adjudicate not for ignorance about the case, but because the concept of emo­ tion lacks clear boundary conditions. Borderline cases are just as real, important, natural, and worthy of scientific explanation as are the more prototypical ones. Because more prototypical cases come more readily to mind and are remembered better, we must be cautious when stating the properties of emotional episodes not to overemphasize the prototypical cases over the equally real mediocre and borderline cases. The psychology of emotion has been bothered by questions about nonprototypical­ cases: Which of the following are really emotions: love, anxiety, trust, liking, respect, hunger, pain, preference, tranquility? Opinions differ. Which of the following cases imply real emotions: a charging bear, a hooked fish, a fleeing flatworm? From the point of view articulated here, questions about what really is an emotion are questions more about the meaning of folk concepts than about the events referred to. Thus, if you tell me all the facts about love, then adding that

93

94 James A. Russell

love is an emotion provides me with no additional scientific information. Similarly, if you tell me all the facts about the charging bear, then adding that the bear is having an emotion provides no additional scientific information. Emotional episodes can be divided into kinds. Every language provides words for kinds of emotion. In English, we have fear, anger, love, jealousy, and so on. A prototype analysis for each kind of emotion can be had parallel to that for the superordinate concept of emotion. Many emotion theorists suppose that the kinds specified by our English terms will serve as scientific kinds. For various reasons, this proposal is problematic. Other languages recognize emotions not recognized in English. Borders are fuzzy and definitions lacking. Again, we must distinguish properties of the category system from properties of the events referred to. For instance, consider the claim that anger is a basic emotion. Anger does fall at a basic level of categorization: that is, within a fuzzy hierarchy, anger is included (largely) within the superordinate emotion; and, in turn, anger can be subdivided into kinds of anger: outrage, annoyance, temper tantrum, cold anger, and so on. It is a short but unfortunate step to calling anger a basic-level emotion or basic emotion. In this context, “basic” is a property of the categorization system, and tells us nothing about the nature of the events referred to. Relation of Emotional Episode to Affect. The relation of emotional episodes to affect is not simple. Core Affect exists continuously and therefore exists throughout any emotional episode. Core Affect gives an emotional episode its subjective feelings of pleasure-displeasure and energy. For example, the prototypical case of sadness includes a feeling of displeasure and low energy. The prototypical case of elation includes an opposite feeling, that of high pleasure and high energy. But of course not all cases are prototypical, and cases of sadness and elation may exist without these specific values of Core Affect. Core Affect varies from one case to the next and varies over time during the emotional episode. A specific value of Core Affect is a central feature of some emotion concepts, but less central to others and missing in still others. That is, at one end of the continuum, Core Affect is so central to the concepts of sadness and elation that without the appropriate Core Affect the case would be so aberrant that it would be borderline or outside the category. At the other extreme, one can feel love, but the concurrent Core Affect depends on circumstances. Most emotion concepts fall between these extremes. For example, fear allows a range of values of Core Affect. The prototypical fear episode includes a Core Affect of displeasure and high energy. Nevertheless, the category of fear included less prototypical cases with unusual values of Core Affect. One can seek out and enjoy the scary roller coaster ride or the horror film. One can calmly read the horror novel, find it scary, and still nod off to sleep. One can honestly say, “I fear that the climate will warm over



3.  From a psychological constructionist perspective

the next century” without having a Core Affect of displeasure and high arousal. (Are these really cases of fear? That is a question of semantics. They are all cases that need to be explained.) The more interesting question is empirical: what role in fact does Core Affect play in emotional episodes? Core Affect would not have evolved unless it serves a function. It serves as a part of the behavioral planning and goal setting. It influences what memories are accessible. We must, however, resist the assumption that Core Affect plays the same role in all emotional episodes. In the prototype of fear, for example, Core Affect shifts in response to a dangerous event. If we consider all cases of fear, and not just the prototypical ones, then changes in Core Affect and the role those changes play in the episode may be different. For example, the change in Core Affect can come before the danger, as when a bad feeling creates vigilance for threats. When an event occurs that can be interpreted as a threat, then we have a case likely to be called fear. Emotional Meta-experience is also included in the mental script for prototypical emotional episodes. It is a central but not a defining feature of the script. Consider someone facing a clear danger, who experiences a strong shift in Core Affect to displeasure and high arousal, who initially freezes, who undergoes numerous bodily changes preparing to cope with the danger, and who then flees. In this prototypical case, in addition to the components just mentioned, this person also very likely perceives him- or herself as frightened, just as would an outside observer. Yet, people do sometimes report afterwards that during the episode they were too focused on the danger to realize that they were frightened. Once again, the meta-experience of being frightened is a prototypical feature of the concept of fear, but not a necessary feature. In some of the more mediocre cases of fear, the person may not perceive him- or her-self as afraid, even if outside observers do. Human adults sometimes report that only at the end of the day did they realize that they had been angry or jealous all day without knowing it. How can one not know one’s own emotion? In some cases, the representation may simply not be formed, even when the ingredients are present, or the representation might be formed but not attended to.

4. What kind of affect might be had by other mammals and/or human infants? How does such affect relate to the affect of adult humans? Given my deep ignorance on the topic, I am reluctant to discuss conscious experience in non-human animals and human infants. Nothing in my proposed framework depends on an answer to these vexing questions. In this section, I do not answer the questions, but, in the spirit of this exchange, and in the hope that

95

96 James A. Russell

the questions will be seen as empirical issues rather than invitations to a priori pronouncements, I speculate on a research agenda that might move us in the direction of answers. I assume that everyone agrees that non-human animals and human infants have emotional episodes. Clearly, early in animal evolution, it became useful to have a system for detecting dangers, opportunities, and so on; to engage in “emotional behavior” (fight, flight, approach, avoidance, and so on); to alter one’s peripheral physiology to support that behavior; to remember outcomes of one’s behaviors; and, among social species, to have signals to communicate to others about behavioral intentions, abilities, threats, and the like. Nevertheless, even if we agree to all these assumptions about emotional episodes, we will not have answered the question about affect. Non-human animals and human infants may show, for example, “fear behavior” before they are equipped to have an Emotional Meta-experience of being frightened. Ditto for other “signs of fear,” such as peripheral physiological changes. From these behaviors I would not automatically infer the conscious meta-experience of feeling specifically frightened. Similarly, Core Affect can operate unconsciously (Berridge & Winkielman, 2003), and so from behavioral evidence of Core Affect in nonhuman­ animals and human infants I would not automatically infer the conscious feeling of Core Affect. How might we infer consciousness? I am conscious, and the adult humans I ask tell me they are conscious. Because my consciousness is produced by my brain, and their brains are similar, I believe them when they say they are conscious. Preverbal human infants and some non-human animals have similar brains, and so I believe they have similar (although perhaps not identical) conscious experiences. As we consider animals with brains less and less similar to human brains, and we consider younger and younger human fetuses, the situation becomes less and less clear. How about affect specifically? An answer requires a comparison of the neural processes underlying affect in human adults with those in the creature in question. We need first to agree on how to measure affect in human adults. Then, we need to find in human adults the relevant neural correlates of affect, show which neural events cause affect, and develop an account of how those neural events explain the properties and functions of affect. Finally, we could compare these neural processes in human adults to those found in non-human animals and human infants. I cannot speak to the neural level, but we can begin this speculation by asking about the function of affect. I assume that Core Affect evolved as an adaptation and therefore has a function. Core Affect monitors one’s state along the dimensions of pleasure and arousal. Doing so is useful in setting current goals and plans



3.  From a psychological constructionist perspective

to achieve those goals. For example, an immune system response to infection moves Core Affect toward displeasure+low arousal, which, in turn, discourages ambitious goals and plans. Cabanac (1971) describes the pleasure-displeasure dimension as a common currency, a means of comparing different outcomes. Core Affect, which also includes the dimension of arousal, monitors one’s energy level and therefore serves as a consideration in planning action. If this characterization of the main function of Core Affect is correct, then Core Affect is useless to an organism whose behavior is totally reflexive. Core Affect has a use only when behavior is flexible. By “flexible,” I mean that the behavior is determined, not directly by a stimulus, but by the animal setting some outcome as a goal and having knowledge of the causal relation between a sequence of actions and achievement of that goal. Core Affect plays a role in selecting the goal and the plan to achieve it. Thus, Core Affect likely co-evolved with behavioral flexibility, that is, alongside the ability to choose among different goals to maximize pleasure and minimize displeasure and to choose among different plans to achieve those goals. If Cabanac (1992) is correct that pleasure-displeasure exists in amniotes – which includes all land dwelling vertebrates: mammals, reptiles, birds, and extinct mammal-like reptiles (theropsids) and dinosaurs – then I expect that amniotes show behavioral flexibility. Similarly, predicting how one’s Core Affect will change in response to an event is useful for decisions about that event. One possible mechanism relies on coding a current event for Affective Quality, with this code consolidated in longterm memory along with the episodic features of the event. For example, pleasure upon encountering a new person, place, or thing leads to coding it as pleasant. Remembering this coding is handy for anticipation of results of future encounters. This information is useful in planning, for example, whether or not to seek out that person, to travel to that place, or to approach the thing. Finding that a particular place is relaxing is useful information when one seeks a place to rest. These examples highlight acquisition during a lifetime based on initial changes in Core Affect, but acquiring such information can also be a phylogenetic change. Even if what I’ve said so far is correct, Core Affect could still operate unconsciously (Berridge & Winkielman, 2003). Why is Core Affect accessible to consciousness in human adults? Balleine and Dickinson (1998) hypothesized that the function of consciousness is to provide an interface between a basic motivational system and a cognitive, goal-setting system. (Indeed, Cabanac, 2010, suggested that consciousness first evolved as consciousness of pleasure-displeasure.) In present terms, one such interface would be between a change in the response of Core Affect and the cognitive system encoding the Affective Quality of the event that caused that change. Balleine and Dickinson used their account to provide experimental evidence consistent with the existence of consciousness in the

97

98 James A. Russell

l­aboratory rat. They relied on the Garcia effect to study just such an interface in rats. A rat was trained to bar press to obtain sugar water. At some point, it was given an emetic after ingesting the sugar water. Nevertheless, on the next trial, the rat continued to bar press to obtain sugar water, presumably because the cognitive goal-setting mechanism did not have access to the association of sugar water with nausea. When the rat received the sugar water, tasted it, and presumably became nauseous, it stopped bar pressing. In my terms, although the rat developed a conditioned aversion to the sugar water, that information was not transferred to the goal-setting cognitive system mediating bar pressing until the rat was conscious of the change in Core Affect, attributed the change to the sugar water, and then re-coded the affective quality of the sugar water. (Another possible use of consciousness was mentioned earlier: conscious simulation of a future behavior and its outcome and thereby experiencing the resulting Core Affect.) How about Emotional Meta-experience? Even if we conclude that Core Affect exists in conscious form in human infants and other species, the question remains unanswered for Emotional Meta-experience. We cannot directly infer Emotional Meta-experience from Core Affect, conscious or not. Core Affect is primary level of consciousness; Emotional Meta-experience is secondary. I also believe that we cannot directly infer Emotional Meta-experience from an emotional episode because Emotional Meta-experience is not required for an emotional episode. Again, we can begin speculation on an answer by asking when Emotional Metaexperience became useful. Emotional Meta-experience is the process of using information about one’s current emotional episode in order to categorize oneself as having a specific emotion. One also categorizes the emotions of others. I speculate that one use of Emotional Meta-experience is bringing social and personal norms to bear on oneself and others. If so, Emotional Meta-experience co-evolved with sociality, and humans are especially social animals (as are some other species such as canines and other primates). This speculation implies that categorization as to emotion arose first as a way of perceiving others rather than perceiving self. On the other hand, the existence of norms, emotion categories, and a secondary level of consciousness might be found to require the existence of language or a certain level of cognitive complexity or cultural transmission, in which case Emotional Meta-experience might be limited to humans.

5. How can we infer affect from observation of behavior? I take this question to express a worry about the inference of conscious private subjective experience – affect – from behavior. Of course, this is the worry that so exercised the behaviorists. I believe that the idea that other human adults, human



3.  From a psychological constructionist perspective

infants, and members of other species have conscious experiences began with the first-person experience of consciousness. When the early psychologists introspected their own conscious affective feelings, they found dimensions similar to those of Core Affect (see Reisenzein, 1992, and Reisenzein & Schonpflug, 1992 for details). Various sources of behavioral evidence more acceptable to modern scientists point to a similar conclusion (for details, see Russell, 2003): The dimensions of Core Affect (although sometimes only one or the other of its two dimensions) has been called on as part of the explanation of (a) semantic similarities found across all human languages; (b) various behaviors including sex, aggression, eating, drug use, and altruism; (c) cognitive effects, including decision making; (d) changes in the peripheral nervous system; and (e) hemispheric differences in the central nervous system. Within each of these domains, observable behavior can be explained through the operation of an underlying process like Core Affect. (Hence, similar principles appear in theories formulated from different perspectives: hedonism, utility, the law of effect, the pleasure principle.) For example, the varieties of sexual behavior can be best understood by hypothesizing that they bring pleasure. Although something like Core Affect appears in many research domains, the details of my hypothesis of Core Affect stem most directly from psychometric studies of self-reported mood, emotion, or affective reactions to various stimuli. Psychometric studies rely on such statistical models as factor analysis. In a study of mood, for example, participants might rate how intensely they are currently feeling various moods (such as anxiety, depression, enthusiasm, etc.), each on a 1–10 rating scale anchored by “not at all” and “extremely.” Different participants typically report different moods. The fact on which factor analysis relies is that an individual’s rating for one mood descriptor is not independent of his or her ratings on all other mood descriptors. Indeed, ratings for different moods are found to be highly related to one another. Factor analysis accounts for the correlations among observed ratings in terms of a small number of unobserved underlying dimensions (the factors). More specifically, a correlation coefficient is computed for each of all possible pairs of the mood ratings. This set of correlations is subjected to factor analysis in order to infer the factors that influence the ratings. When someone provides a high number on one rating scale, what else do they do? One might expect that if they provide a high numerical rating for fearful, they will simultaneously provide high ratings on words synonymous with fearful (ap­ prehensive, concerned, frightened, anxious), and that is indeed what they do. One might also expect that those who provide high ratings to the set of fear words will provide low ratings on antonyms of fear (calm, relaxed, reassured, serene), and that too is what they do. Now, here is the interesting finding. On the basis of the standard assumption of discrete emotions, one would expect that high ratings on the cluster of fear words would be independent of ratings on clusters of words for

99

100 James A. Russell

other discrete emotions (anger, sadness, and so on). Or, if one discrete emotion inhibits all others, one would expect negative correlations (in which high ratings on one cluster predict low ratings on another cluster). Neither independence nor negative correlations between discrete emotions has been found. Instead, high ratings on any negative word predict high ratings on any other negative word and low ratings on any positive word. High ratings on any positive word predict high ratings on any other positive word and low ratings on any negative word. This empirical finding appears in the factor analysis as a dimension of positive vs. negative. One such result is seen in the horizontal dimension of Figure 1. In factor analysis, the first dimension can then be set aside, and the search can start over to account for remaining correlations. The next dimension found is typically based on high ratings on one high-arousal word predicting high ratings on all other high-arousal words. This result can be seen in the vertical dimension of Figure 1. Factor analysis can be used to extract additional factors, but they have typically been difficult to interpret because the signal to noise ratio is low. Numerous such studies have been conducted, with cumulatively strong support for the underlying operation of pleasure-displeasure and degree of arousal. (See Russell, 2003, for a review of research up to that point, but research continues; Stanley & Myer, 2009.) Details vary of course from study to study, but there is a remarkable convergence on at least these two dimensions: Whether participants are responding to stimuli presented or simply reporting their current mood or emotion, factor analytic studies point to two large factors. The labels used for these factors vary, but are variants of valence and arousal.

6. Is affect pre-linguistic? How can we study what is pre-linguistic? How does language alter affect? Clearly, most of the ingredients of emotional episodes are pre-linguistic, and so the question is, again, affect. Whether affect is pre-linguistic may depend on which aspect of affect we consider. Core Affect is pre-linguistic in that it preceded language in phylogenesis and precedes language in ontogenesis and in that it continues to operate on a daily basis independently of language. Core Affect operates continuously whether or not it is spoken about, labeled, categorized, or even brought into awareness. Berridge and Winkielman (2003) offered experimental evidence for the operation of Core Affect (although their term was emotion) without conscious awareness. As an analogy, Core Affect is similar to the mechanism for monitoring body temperature: you can access your body temperature at will, but the body adjusts to temperature changes even without your being consciously aware of that temperature.



3.  From a psychological constructionist perspective 101

The more interesting case is Emotional Meta-experience. On my account, it is a perceptual process that relies on concepts. Are the concepts through which we categorize our emotional episodes, concepts such as fear and anger, acquired through language? The answer to this question is independent of my proposed framework for the study of emotion. Human and other animals form non-linguistic­ concepts, but language is also a vehicle for transmission of newly formed concepts, as occurs in education and other forms of cultural transmission. The role of language in acquiring concepts and the role such concepts play in perception and cognition are fascinating topics. Evidence is accumulating that language does indeed play a role (e.g., Roberson & Davidoff, 2000). As is discussed in a later question, I believe that language is responsible for the acquisition of at least some emotion concepts. Perhaps those concepts that are defined as simple combinations of pleasure and arousal are near universal, but other emotion concepts are inherited through language. If so, it would follow that speakers of English and Ifaluk have Emotional Meta-experiences that, although in some ways similar, are in some ways different, because of the different languages they speak.

7. Is affect always conscious? At one time, emotions were assumed to be conscious states, but nowadays some speak of unconscious emotions. For instance, some folks occasionally report that in an emergency situation they were so focused on the external world (the gun pointed at them, the earth trembling beneath them, the car spinning out of control) that they were unaware of their own intense emotion. People occasionally admit that at the end of the day, upon reflecting back, they realized that they were jealous or envious, but had not been aware of it at the time. Psychologists Berridge and Winkielman (2003) offered experimental evidence of unconscious emotion: subliminal presentation of pleasant stimuli increased drinking of orange juice – unconscious liking. The anthropologist Levy (1973) offered ethnographic evidence. Tahitians lack a word for sadness, even though on occasion they show clear signs of being sad. Levy concluded that they are sad but not consciously aware of it – unconscious sadness. Let me try to make sense of seeming cases of “unconscious emotion” in the terms of my framework. I believe we can speak of unconscious Core Affect but not of unconscious Emotional Meta-experience. Core Affect is always accessible to consciousness, but we do not always attend to it. Schooler and Mauss (2010) offered evidence that even an intensely pleasant state can occur without one attending to that state, although the state is readily available when attention is turned to it. I interpret Berridge and Winkielman’s (2003) finding as showing that Core Affect continues to operate

102 James A. Russell

even when we are not attending to it. Its role in certain psychological processes, however, does presumably vary with how conscious it is. That is, consciousness has a function, and so whether or not we attend to Core Affect matters on what role it can play. Emotional Meta-experience is conscious. It is a self-perception based on a pattern of ingredients. (Here I assume that perceptions are conscious events. However, the discovery of blind-sight, blind-touch, deaf-hearing, and numbsense shows that perceptual processes can occur in the absence of awareness, although such “perceptions” are very limited, Weiskrantz, 2008. I know of no cases in which something as detailed as an Emotional Meta-experience has been discovered to occur unconsciously, and so I set that possibility aside.) The percept is based on the pattern of ingredients as conceptualized in terms of a specific emotion such as fear for English speakers or fago for Ifaluk speakers. The ingredients can all occur, even when the self-perception of the emotion does not. When that self-perception does not occur, the person has not, so to speak, put two and two together. Perhaps this is what happens in the emergency situations when people report no awareness of their own emotion. How about unconscious emotion? We can interpret cases of “unconscious emotion” without using that way of speaking. Return to the case of Levy and a Tahitian showing signs of sadness but not experiencing sadness. On my account, the Tahitian may be consciously aware of all the ingredients of sadness (feeling miserable, lethargy of movement, downcast visage, and so on) in much the same way that Levy is aware of at least some of those ingredients in the Tahitian. But whereas Levy conceptualized that pattern of ingredients as sadness – as fitting together into one thing – the Tahitian did not. The ingredients occur in a pattern, and although we English speakers have learned to categorize this pattern, Tahitians have not. In short, in such cases, we can say that the Tahitian is aware of the ingredients but has no Emotional Meta-experience of the emotion we would judge to fit that pattern of ingredients. Is this a case of unconscious emotion? Calling it that would be misleading. In general, because most ingredients of an emotional episode are continuous processes, they are continuously in some pattern. We can name some of these patterns, but not all. There are an uncountable number of possible patterns, and patterns could be categorized in multiple ways, as is done across languages. Indeed, many different patterns could be defined – just as we could define many new constellations in the night sky, although finding good or useful patterns is a different matter. Thus, if we were to speak of unconscious sadness in the Tahitian, we would have to admit for the same Tahitian on the same occasion simultaneously many other unconscious emotions, one for each possible way to categorize the pattern. Of course, this issue is not unique to Tahitians or to sadness; the same



3.  From a psychological constructionist perspective 103

principle applies to everyone and all emotional episodes. Referring to each of the countless unnamed patterns as an unconscious emotion is useless. There is a further problem in the notion of an unconscious emotion. In English, the names for the patterns include sadness, but the name is not the emotion. The pattern, its ingredients, their underlying biology, and the act of categorizing the pattern in a certain way are all real events. There is no additional event – no event that would be the unconscious sadness. The pattern is not caused by some underlying entity called sadness. It is thus misleading to say we are conscious of or not conscious of the sadness, as if we were detecting or not detecting the entity that causes the pattern. To experience sadness is to interpret the pattern of ingredients as sadness. To not experience sadness given the same pattern of ingredients is to fail to interpret the pattern as sadness. You can perceive yourself as being sad, even when others might reject the label. Or you can later interpret yourself as having been sad after the fact. (The reasonableness of that interpretation is yet another matter. For example, you might incorrectly infer what you felt given the situation.) In short, sadness is not an entity that is conscious or unconscious. It not an entity at all. In summary, my way of describing these matters runs counter to psychologists’ current way of speaking. My way stems from my not using everyday emotion terms as scientific terms, but simply as folk concepts, and from not thinking of emotions as causal entities. Core Affect can operate unconsciously, but can be brought to consciousness at will. Emotional Meta-experience, in contrast, is conscious. There is currently no reason to believe that you can perceive yourself as sad and not be aware of that perception. Speaking of unconscious sadness is not helpful.

8. What is the relationship between affect and cognition (conceptualization, intentionality, appraisal etc.)? A person’s Core Affect is often influenced by cognitively processed information and is often embedded within an intentional state. Receiving a gift makes you happy. You can be happy anticipating it, even before actually receiving it. You can be happy later, after the gift is gone, when you remember it. With the gift in hand, Core Affect depends on whether the gift is interpreted as a thoughtful gesture, as a bribe, as inappropriate, as fulfillment of an obligation, or whatever. Through the cognitive process of attribution, Core Affect becomes part of an intentional state: happy about receiving the gift. Even in cases of a stimulus eliciting a strong change in Core Affect, as in tasting a delicious food or hearing a loud sound, some mechanism must bind the change in Core Affect to the stimulus.

104 James A. Russell

Nevertheless, Core Affect need not be categorized, conceptualized, attributed, labeled, or even brought to awareness to do its work, and it need not be based on cognitive processes of any kind. In this case, it is likely to be called a mood or background feeling or free-floating emotion. An example would be waking up in a good or bad mood for no known reason and not about anything. Another example would be when a drug is slipped into your coffee without your knowing it. Core Affect is influenced by events that are not cognitive and are difficult or impossible to perceive. Internal examples include immune system reactions and hormonal changes. External examples include drugs, ionization of the air, ultrasound, and infrasound. I suggested earlier that Core Affect is modular, implying that it is sometimes encapsulated from general knowledge. (An example of encapsulation occurs in the Muller-Lyer illusion in which the perception that the two lines are unequal in length is immune to knowledge that they are in fact equal.) At first thought, encapsulation seems to contradict my claim that Core Affect responds to (among other things) cognitively processed information, as in anticipating or remembering a gift. This and similar examples of cognitive influence, however, may all have another feature in common: conscious attention to the information. When the information fades from consciousness, its influence on Core Affect similarly fades. If the thoughts about the gift continue to invade consciousness, then the gift continues to shape Core Affect. But if one is distracted, such that thoughts about the gift are pushed out of consciousness, then the excitement fades. Thus, I offer the hypothesis that Core Affect is isolated from general information, with the massive exception of information currently seen through the window of consciousness. Encapsulation thus resonates with the Virtual Reality Hypothesis. One half of this hypothesis is that Core Affect can be impervious to general information when that information is not part of current consciousness. The second half of this hypothesis is that Core Affect responds to the drama playing in the theatre of consciousness – regardless of general knowledge about that drama such as whether it is fact or fiction. This is most clearly seen when Core Affect responds to virtual reality – daydreams, art, music, theatre, films, and novels – when we know that what we see is fiction. Similarly, Core Affect responds to the fortunes and misfortunes of others as if they were our own – even though we know they are not our own – in what is known as empathy. But empathy does require conscious attention. Core Affect responds to a fantasized future, even when we know it won’t come true. Core Affect responds to a vividly imagined event, even when we know it is physically impossible. By consciously simulating a possible behavioral path and experiencing the resulting changes in Core Affect, one can put this feature of Core Affect to good use.



3.  From a psychological constructionist perspective 105

Emotional Meta-experience is an intentional state; it is the perception of oneself as having an emotion about something. It is a Gestalt composed of many ingredients. Although the information available about oneself and about another person differ somewhat, perceiving the emotion of oneself and that of another are the same type of process. Although some of the features on which an Emotional Meta-experience is based occur without concepts, Emotional Meta-experiences themselves rely on concepts. The features can all occur, and just the same the person might not conceptualize the pattern of features as an emotion. Such is said to occur in alexithymics, in human infants, and in other species. In short, emotional episodes can occur unaccompanied by an Emotional Meta-experience.

9. Is affect altered or filtered through culture? In what ways? Are there universal aspects of affect? Core Affect occurs in all humans, of all cultures and historical periods, and, indeed, in many other species as well. However, the forces that change Core Affect, how we think about Core Affect, and how we try to alter Core Affect can all be influenced by culture. The forces that change Core Affect range from those purely determined by nature to those purely determined by nurture, but most are likely determined by an interaction between the two. As a paradigm case, consider how Core Affect responds to food (Rozin, 2007). The pleasantness of a sweet taste is likely hardwired, but the pleasantness or unpleasantness of an odor is likely learned (Engen, 1982). Cultures differ in their preferred Core Affect (Tsai, 2007). And, different cultures and individuals have developed different mechanisms for achieving these preferences, or, more generally, altering Core Affect. Consider nicotine, caffeine, ethanol, cannabis, and other drugs different people of different cultures and subcultures use to adjust Core Affect. Emotional episodes are built by both nature and nurture. Components of emotional episodes draw from a common pool, especially if described at an abstract enough level. A huge number of patterns of those components occur, with both commonality and some culture specificity. A controversial issue centers on the categories into which emotional episodes (patterns of components) are to be divided for scientific purposes. Tradition, common sense, and many emotion theorists have assumed that emotional episodes can be divided into a small number of universal categories labeled by such English words as happiness, surprise, fear, anger, disgust, and sadness. My account of emotional episodes offers an alternative approach. Of course, we can impose any categories we choose, but they might not turn out to be useful scientifically.

106 James A. Russell

Like emotion, the conscious experience of emotion has traditionally been thought of as divisible into a small number of universal discrete categories such as the seven basic emotions. On this view, subjective experience of emotion is either equated with the emotion or considered a detection of the emotion. The seeming variety of subjective emotional experiences is thought to be due to variations in intensity, co-occurrences, or blends of the few basic ones. Put differently, the standard view postulates a small number of universal discrete emotional qualia. My notion of Emotional Meta-experience offers an alternative approach. Because there is no entity or essence that underlies an emotional episode, there is no entity or essence to detect. Different individuals can interpret even the same emotional episode differently, and some of that variation is due to culture. The folk concepts of emotion familiar to speakers of English today – emo­ tion, fear, anger, terror, outrage and on and on – are not identical to those found in other languages or even to English speakers of other historical periods. Some languages lack a superordinate equivalent to emotion. Basic-level concepts do not have a one-to-one translation between languages. English has concepts with no exact equivalent in other languages, and other languages offer concepts with no equivalent in English. The Japanese concept of amae may be the most famous example. Fago in the language of the Ifaluk and liget in the language of the Ilongot are others. Even concepts that initially appear translatable between languages turn out not to be (Russell & Sato, 1995; Hurtado de Mendoza, Fernandez-Dols, Parrott, & Carrera, 2010; Wierzbicka, 2009). In addition to the well known linguistic evidence supporting the claim of cultural differences in emotion concepts (Wierzbicka, 2009), there is evidence from a surprising source. In their effort to find a universal signature in the autonomic nervous system (ANS) for each basic emotion, Levenson et al. (1992) studied the Minankabau of West Sumatra. Participants were instructed to contract facial muscles into the prototypical configurations hypothesized for each emotion. Doing so, in turn, alters ANS activity. For North Americans, this alteration of facial muscles and ANS activity resulted in some reports of the experience of specific emotions. For the Minankabau, however, the same procedure failed to produce the emotional experience, perhaps because their emotion concepts differed from those of North Americans.

10. What roles do the biological sciences and psychological sciences have in studying affect A psychological account must be complemented by biological and social-cultural accounts. Theories at these different levels have to be compatible and are mutually



3.  From a psychological constructionist perspective 107

beneficial. For example, the brain is doing the work and we need to know how it does it. Such biological knowledge would then put constraints on what psychological theories can propose. Conversely, clarity at the psychological level also aids the search for biological underpinnings. For example, if I am correct that the extension of the concept of emotion is too heterogeneous for scientific purposes, then the search for the biological basis of emotion will discover a heterogeneous set of different processes. And, indeed, the search for neural circuitry common to all emotions has been largely abandoned (LeDoux, 1996) in favor of the search for circuitry for individual classes of emotion such as fear. And of course I raise the same questions about those individual classes. My suggestion is that the concept of, for example, fear does not cut nature at the joints. If so, then biological theories that use that concept are headed for the same fate. To illustrate, consider neuroscientific theorizing on the “central state of fear” (e.g., Davis, 1992). The original hypothesis that the startle reflex is potentiated by this “central state of fear” has been replaced by the hypothesis the potentiation occurs with any unpleasant state (Lang, 1995). One recent study examined two different behaviors thought to be effects of this “central state of fear”: freezing and inhibition of eating (Petrovich et al., 2009). In a conditioning paradigm, these behaviors were elicited by the same conditioned stimulus, but shown to be supported by different brain circuitries. The two behavioral components of fear were dissociable – a finding consistent with the notion that a “central state of fear” is too heterogeneous for scientific purposes (See Killcross, Robbins & Everitt, 1997, for related evidence.)

11. What role might Panksepp’s primary affects play in Russell’s model of psychological construction. What implications might Russell’s model of psychological construction have for Panksepp’s construct of primary process affects? At first glance, Panksepp’s account may appear to be at odds with Psychological Construction, but I believe they are close and complementary. Of course, some healthy competition exists, but even that is not easy to pinpoint. I begin with some areas of agreement, then turn to several questions that may help sharpen the possible points of disagreement, and finally offer an illustration of how the two accounts are complementary. Areas of Agreement. First, Panksepp and I agree that emotion must be analyzed at complementary levels, including biological, psychological, and social-cultural.

108 James A. Russell

Second, Panksepp endorsed the valence and arousal dimensions. He wrote of a positive-negative dimension in his treatment of drugs and electrical brain stimulation (2005a). Speaking more broadly, he wrote that “human animals are utilitarian, because life-affirming events feel good, and life-detracting events feel bad. As utilitarian Jeremy Bentham said in 1789, ‘utility’ reflects ‘that property in any object, whereby it tends to produce benefit, advantage, pleasure, good, or happiness … or … to prevent the happening of mischief, pain, evil, or unhappiness’.” (2005b: 22– 23). Panksepp (1998) also included a dimension of arousal (Cannon, 1927; Duffy, 1941; Hebb, 1955). Panksepp (2008: 407) wrote about “emotional dimensions of valence-type … and overall arousal… as part and parcel of each and every primary-process emotional system.” He added that “each of the subcortical emotional systems can generate different degrees of valence and arousal” (p. 407). Third, we agree that the subcortical circuits he delineated are involved in many emotional episodes. I am not qualified to compare Panksepp’s theory to other neuroscientific theories of emotion, and, in any case, neuroscience is a work in progress. I assume that subcortical circuits, such as Panksepp’s (1998) fear circuit, are universal aspects of mammalian nature. Whatever else they may produce, these circuits change, downstream, the autonomic nervous system, endocrine system, and certain simple behaviors. Upstream, these circuits influence perceptual, cognitive, and complex behavioral processes. Fourth, we agree that the effects produced by his subcortical brain circuits are only part of what is seen at a psychological level in emotional episodes. Panksepp (2008: 402, emphasis added) wrote that his circuits are “critical parts of the emotional wholes described by the many, lower-case vernacular labels that dimensionalists typically study in their experimental work.” He added that the products of his circuits “dissolve into the rich broth of lived lives to such an extent that they cannot be easily retrieved” (p. 404). And, fifth, we agree that there is no one-to-one correspondence between, for example, the effects of his fear circuit and the events in the category picked out by the everyday word fear. That is, neither of us assumes that changes produced by the Panksepp fear circuit occur in all and only cases named fear. A telling example comes from Panksepp’s (1998) research on separation anxiety. Although the word anxiety is semantically considered a subcategory of fear, separation anxiety is not mediated by his fear circuit. Consistent with Panksepp’s model, a child could honestly report the experience of fear/anxiety when undergoing separation anxiety, and yet this experience is not mediated by the Panksepp fear circuit. Conversely, there may be occasions in which the fear circuit is active but that would not be labeled fear. Therefore, the subjective experience of fear is not invariably read from activity of the fear circuit. Similar statements can be made for the other circuits.



3.  From a psychological constructionist perspective 109

Four Questions. My account also raises questions for Panksepp. Here are four. 1. Given that what a Panksepp circuit produces is only a part of an emotional episode, what produces the other objective parts? Consider fear behavior. Behavior when afraid is not limited to freezing, avoidance, and flight. Consider Sally’s many fears: Afraid of the bear, she flees; afraid of a spider, she squishes it; afraid that she’ll miss the train, she runs toward it; afraid that her child is sick, she telephones her doctor; and afraid of an economic recession, she sells her riskier stocks. Even in encounters with wild bears, different individuals behave differently. Depending on circumstances, some people might gather together their children and restrain their pets. Those who heed the advice of the park ranger do not flee, but retreat slowly and speak softly. I have read about people who did not heed the ranger’s advice and who played dead, punched the bear, or threw their camera at it. I have read about people who try to feed it or pet it. And of course some people shoot bears. I do not mean to limit this question to humans. In laboratory situations in which a fear stimulus is presented, an animal’s options may be limited to freezing, but in situations with more options, animals show a range of behaviors to fearful stimuli, including exploration, which if anything might be considered a type of approach, (MacDonald & Pinel, 1991; Pinel & Mana, 1989). In the wild, animals sometimes react to frightening stimuli with defensive aggression, such as an entire flock mobbing a predator (Lorenz, 1966). Earlier, I mentioned a study by Petrovich et al. (2009) showing that different fear behaviors in the rat (freezing and inhibition of eating) were supported by different circuitries. So, my question about fear behavior is this: What is the precise contribution of the Panksepp fear circuit to the actual specific behavior that occurs in cases called fear? What other mechanisms are needed to explain the various specific behaviors that actually occur? A similar question can be raised about the changes in the autonomic nervous system during episodes called fear. For example, fear elicited by an escapable stimulus involves heart-rate acceleration, whereas fear elicited by an inescapable stimulus involves heart-rate deceleration (Hamm, Cuthbert, Globisch, & Vaitl, 1997). So, again, what is the precise contribution of the Panksepp fear circuit to autonomic nervous system activity, and what other mechanisms are needed to explain the various specific changes that actually occur? Finally, parallel questions can be raised for emotions other than fear. What does the Panksepp circuit alone produce and what other circuitry is involved in producing the actual changes seen? 2. My last three questions concern affect. Panksepp suggested that his circuits create subjective conscious affective feelings in mammals. My questions concern the precise nature of those feelings. Admittedly, this is a difficult topic, especially

110 James A. Russell

in non-human animals, but advancing the field requires that we propose testable hypotheses as specific as possible. Let’s begin with the question of how Core Affect fits into Panksepp’s theory of conscious experience. As mentioned earlier, Panksepp wrote of the dimensions of valence and arousal. Does he agree with me, which he may not, that Core Affect is a single, non-reflective, pre-conceptual, and pre-linguistic combination of valence and arousal that integrates across modalities and other systems and is accessible to consciousness? Does he agree with my tentative equation of the valence dimension with Smith et al.’s (2010) pleasure circuit and my tentative equation of the arousal dimension with Pfaff ”s (2006) arousal system? What is the relation of Core Affect to the various Panksepp circuits such as fear? On a psychological level in humans, pleasure-displeasure is a more primitive reaction than fear. After all, fear involves the anticipation of displeasure. Fear conditioning in rats begins with an aversive stimulus. Recall Sally’s encounter with the bear. Part of her reaction included a change in Core Affect: the pleasant calm she felt strolling through the forest shifted to intense displeasure and extreme arousal. Does the fear circuit produce this shift Sally’s Core Affect? More generally, what is the relation of Core Affect to the full set of ­Panksepp circuits? (For another example, aggressive behavior is increased not only by ­angereliciting events but by negative Core Affect more generally, Berkowitz, 2003.) Core Affect is involved in emotional episodes mediated by all of the Panksepp circuits. 3. My third question concerns the precise conscious experience of having a specific emotion. English-speaking people report experiencing fear, anger, shame, guilt, and jealousy, for example. However, in situations such as confronting a bear or a near auto accident, some persons report that they did not immediately experience fear, but were too busy attending to the bear or the car. Jealous and angry persons sometimes honestly deny feeling that way, although may later acknowledge they were in fact angry or jealous. I believe that an emotional episode can occur without the person perceiving him- or herself as having that emotion. So, one question is what mechanism determines whether the conscious experience of the specific emotion occurs or not. Anthropologists report that speakers of different languages do not divide their emotional experiences in the same way that English speakers do. A speaker of Gidjingali might describe an experience such as Sally’s as one of gurakadj (Hiatt, 1978), a concept that includes what in English would be distinguished as shame and fear, although gurakadj is not simply the union of the two sets shame and fear. A speaker of Ifaluk might describe the experience as one of metagu (Lutz, 1988), a concept that includes what in English would be distinguished as guilt and



3.  From a psychological constructionist perspective

fear, although again it is not the union of the two. More interestingly, a speaker of Pintupi would have no word for fear, but must distinguish among 15 different emotions that English groups together as fear (Morice, 1978). For example, ngulu is fear of another seeking revenge; nginyiwarrarringu is a sudden fear that causes the person to try to see what caused it. So, the next question is this: When Panksepp writes of a conscious affective experience produced by each circuit, what is the precise nature of that experience? Is it one quale per circuit? And how is that precise nature of that experience to be established? Is it exactly as specified by an English word such as fear, or does it include the Gidjangali gurakadj, the Ifaluk metagu, the Pintupi ngulu or the Pintupi nginyiwarrarringu? What are the properties of the subjective experience that Panksepp hypothesizes? For example, is it an intentional state, in the philosophical sense? Is it a first-order or second-order experience? At this point, our two accounts may differ, and so this question can be re-phrased this way: How would we empirically differentiate various claims about what conscious experience, for example, the Panksepp fear circuit produces, such as: (a) it produces all and only specific conscious experiences of what in English is called fear, (b) it produces the specific conscious experience of what in Ifaluk is called metagu, or (c) it produces the Core Affect of displeasure + high arousal (“distress”)? 4. Finally, my last question is closely related to the third. Let us assume that there is no one-to-one correspondence between activation of the fear circuit and the extension of the concept of fear. Nor is there a one-to-one correspondence between activation of the fear circuit and reports of fear. That is, let us assume that the fear circuit is neither necessary nor sufficient for fear as an emotional episode or as an Emotional Meta-experience of being afraid. This assumption raises several questions. First, what does the fear circuit alone produce? What mechanisms besides the fear circuit are needed to generate the subjective conscious experience of fear? For example, what mechanism is needed to account for Sally’s specific experience as being one of fear of the bear rather than a free-floating fear? On another occasion, when Sally fears she is losing her boyfriend to another woman, what generates her Emotional Meta-experience of being jealous? Complementary Levels of Analysis. Let me illustrate my claim that our two frameworks are complementary by combining them in an “explanation” of a single fictional case of fear, offered not of course as evidence of anything but as a heuristic story. Consider again Sally and the bear, a case very near the prototype of fear. Panksepp’s fear circuit (consisting of the lateral and central nuclei of the amygdala, the anterior and medial hypothalamus, and the periaqueductal gray) results in momentary freezing, inhibition of ongoing behavior, and changes in Sally’s autonomic nervous system and endocrine system.

111

112 James A. Russell

Other parts of the brain are also involved. Sally must perceive the bear, and doing so involves the thalamus, the sensory cortex, sensory association cortex, and motor association cortex. The thalamus does provide information directly to the amygdala, but that is unlikely to be sufficient to recognize the bear as a bear. The cortex sends impulses to the ventrolateral region of the inferior pons, which eventuates in facial changes. The cortex also provides information to the Panksepp circuit. The amygdala projects to the prefrontal cortex and other parts of the mesolimbic system, which shifts Sally’s Core Affect in a negative direction; this shift is sudden enough to become part of consciousness. The Panksepp fear circuit also connects to the monoaminergic arousal system, shifting Sally’s Core Affect toward high arousal; again the shift is sudden enough to become part of consciousness. The shift in Core Affect, when mediated through the orbital prefrontal and cingulate cortices, is felt as a single strong feeling we might call tension (but Sally does not have time to name it). The amygdala and the monoaminergic arousal system enhance cortical perceptual-cognitive processing of the bear: Sally attends to its every movement, searches her memory for information about bear attacks, and searches the environment to see what it affords in the way of escape or protection. The cortex links the danger presented by the bear to the shift in her Core Affect. The amygdala and monoaminergic arousal system also enhance the activity of the prefrontal cortex in formulating a plan for what to do about the bear. Remembering the park ranger’s advice, Sally slowly backs away and then runs to her car. All the while, this episode is being stored in memory. In the safety of the car, she feels her heart pounding and mentally rehearses what just happened. With her attention now on her bodily feedback, on the serious danger she just faced, and on her escape, she perceives for the first time just how scared she was. In this fictional story, I assume that large changes occur in the autonomic nervous system, but I do not assume that the changes form a pattern common to all and only cases of fear, for we know that different patterns occur in different cases (Larsen et al., 2008). For example, a different pattern occurs when escape is possible than when impossible (Hamm, Schupp, & Weike, 2003). I do not assume that the face shows a signal of fear, although I do assume that Sally’s face shows rapt attention and tension. I do not assume that the Panksepp fear circuit directly produces the conscious experience of fear; for that Emotional Meta-experience requires memory, attention, concepts and, in this particular story, somatosensory feedback; the perception occurs late in the episode. The Panksepp fear circuit directly produces some of the components of Sally’s entire fear episode, and indirectly contributes to some others. Yet, the fear circuit is not sufficient for the full episode. Perceptual-cognitive processing of the bear, attributing the shift in Core Affect to the bear, planning and executing her escape, and then recognizing that



3.  From a psychological constructionist perspective 113

she was scared – all these components require parts of her brain, both subcortical and cortical, in addition to the fear circuit. The story offered for this specific prototypical case cannot be generalized to other cases of fear, even other good examples of fear. And, some of the same processes occur in emotional episodes that would be categorized as emotions other than fear or in states not categorized as any emotion at all. For example, feedback from autonomic changes was a major contribution to Sally’s Emotional Metaexperience­ on this occassion, but more generally, it likely plays a minor role. Some important human fears are directed at objects and events (ill health, financial loss, future anger of a parent) that require more and different cognitive processing than the encounter with the bear required. Some fears are directed at objects and events (violation of religious rules, loss of status) that require a cultural background. Some intense fears arise gradually. (Sally notices that over the course of the day her daughter is showing increasing signs of illness. As the symptoms worsen, Sally’s fear grows. Other fears grow over weeks or months.) In these cases, the Panksepp fear circuit may play only a small role. In some cases, the Core Affect of distress comes before the fear object is perceived. (After a stressful day, Sally encounters an important decision. She fears the worst. Her friends assure her that she is just stressed and has nothing to worry about.) I also speculate that the structures of the fear circuit are active during episodes that would not be labeled fear: We might freeze and step backward in a situation we experience as disgust. Sympathetic activation occurs in strenuous exercise. In some cases, the fear circuit is very active, but Sally does not interpret it as fear. (She and her daughter ride the roller coaster. Especially at the highest peaks at the greatest speed, she is frozen in her seat and feels intense feedback from her autonomic nervous system. Her Core Affect is extremely aroused but on the pleasant side. They laugh and scream as they hurdle downward. Both mother and daughter feel the ride was not so scary after all but thrilling. They decide to ride it again.)

12. Considering the diversity of theoretical viewpoints in the scientific study of affective and emotional phenomena, might one hope for a coherent synthesis in this splintered field? The diversity of theoretical viewpoints stems in part from the heterogeneity in the extensions of the terms scientists have borrowed from everyday language: emo­ tion, fear, anger, jealousy, mood, and so on. Different theoretical statements based on these terms may each be limited to different subsets of the extension. Empirical studies typically include a small non-representative fraction of the extension. To illustrate the problems facing a science based on a heterogeneous term,

114 James A. Russell

Griffiths (1997) pointed to superlunary objects, which was a scientific category in Aristotle’s physics, but which, we now know, refers to a heterogeneous collection of objects. Griffiths wrote, “The search for a theory of superlunary objects and their characteristics was a mistake. Similarly, the idea that we need a theory of the emotions, or a theory of some specific emotion, may be a mistake” (p. 2). Once the term emotion is abandoned, then different theoretical viewpoints may be reconciled. Consider the classic debate between William James and ­Walter Cannon. If where James used the word emotion, we substitute “subjective experience of emotion” (or Emotional Meta-experience), and if where Cannon used the word emotion, we substitute “emotional behavior,” then their accounts are not in any clear conflict. Neither account is the final word empirically, but so phrased, they are complementary hypotheses. A similar re-phrasing would eliminate much of the appearance of diversity in today’s theoretical viewpoints. I see the potential to unite my account with Panksepp’s. In the end, we won’t have a single theory of “emotion,” unless all of psychology is taken to be that theory. All the same, we will have accounts of the diverse phenomena referred to. Admittedly, moving beyond our folk terminology is not easy. These terms appear in the questions asked and the answers scientists have provided. These questions motivate us as scientists and as human beings. These terms therefore frame the historic debates and current scientific background to any research project. These are the questions and debates that appeal to funding agencies that make research possible. On the other hand, other sciences have faced similar problems and solved them. Emotion, anger, fear and the like are puzzles today. Puzzles are, well, puzzling – until solved. When we think back on puzzles to which we now know the solution, it is hard to empathize with the sense of mystery once experienced. Around 1800, it was extremely puzzling how life could emerge from non-living­ chemicals; indeed, it seemed impossible. But now that the general solution is known, it does not seem mysterious at all. I am confident that the puzzle of emotion will be solved.

References Averill, J. R. (1982). Anger and aggression. New York: Springer-Verlag. Balleine, B. W., & Dickinson, A. (1998a). Consciousness – the interface between affect and cognition. In I. Cornwall (Ed.), Consciouness and human identity (pp. 37–83). Oxford: Oxford University Press. Balleine, B. W., & Dickinson, A. (1998b). Goal-directed instrumental action: Contingency and incentive learning and their cortical substrates. Neuropharmacology, 37, 407–419.



3.  From a psychological constructionist perspective 115

Berkowitz, L. (2003). Affect, aggression, and antisocial behavior. In R. J. Davidson, K. R. Scherer­, & H. H. Goldsmith (Eds.), Handbook of Affective Sciences (pp. 804–823). Oxford: Oxford University Press. Berridge, K. C., & Winkielman, P. (2003). What is an unconscious emotion? (The case for uncouscious “liking”). Cognition & Emotion, 17, 181–211. Cabanac, M. (1971). Physiological role of pleasure. Science, 173, 1103–1107. Cabanac, M. (1992). Pleasure: the common currency. Journal of Theoretical Biology, 115, 173–200. Cannon, W. B. (1927). The James–Lange theory of emotion: A critical examination and an alternative theory. American Journal of Psychology, 39, 106–124. Davis, M. (1992). The role of the amygdala in fear and anxiety. Annual Review of Neuroscience, 15, 353–75. Duffy, E. (1941). An explanation of “emotional” phenomena without the use of the concept “emotion.” Journal of General Psychology, 25, 283–293. Engen, T. (1982). The perception of odors. New York: Academic Press. Farthing, G. W. (1992). The psychology of consciousness. Englewood Cliffs, NJ: Prentice-Hall. Faucher, L., & Tappolet, C. (Eds.) (2006). The modularity of emotions. Canadian Journal of Philosophy, Supplementary volume. Fehr, B., & Russell, J. A. (1984). Concept of emotion viewed from a prototype perspective. Jour­ nal of Experimental Psychology: General, 113, 464–486. Gazzaniga, M. S. (1985). The social brain. New York: Basic Books. Gerber, A. J., Posner, J., Gorman, D., Colibazzi, T., Yu, S., Wang, Z., Kangarlu, A., Zhu, H., Russell, J. A., & Peterson, B. S. (2008). An affective circumplex model of neural systems subserving valence, arousal, and cognitive overlay during the appraisal of emotional faces. Neuropsychologia, 46, 2129–2139. Griffiths, P. E. (1997). What emotions really are. Chicago: University of Chicago Press. Hamm, A. O., Schupp, H. T., & Weike, A. I. (2003). Motivational organization of emotions: Autonomic changes, cortical responses, and reflex modulation. In R. J. Davidson, K. R. Scherer, & H. H. Goldsmith (Eds.), Handbook of Affective Sciences, 187–211. Oxford: Oxford University Press. Hebb, D. O. (1955). Drives and the CNS (conceptual nervous system). Psychological Review, 62, 243–254. Hiatt, L. R. (1978). Classification of the emotions. In L. R. Hiatt (Ed.), Australian aboriginal concepts (pp. 182–187). Princeton, NJ: Humanities Press. Hurtado de Mendoza, A., Fernandez Dols, J. M., Parrott, W. G., & Carrera, P. (2010). Emotion terms, category structure, and the problem of translation: The case of shame and verguenza. Cognition & Emotion, 24, 661–680. Inagaki, K., & Hatano, G. (2004). Vitalistic causality in young children’s native biology. Trends in Cognitive Science, 8, 356–362. Izard, C. E. (1977). Human emotions. New York: Plenum Press. James, W. (1884). What is an emotion? Mind, 19, 188–205. Kringelbach, M. L. (2010). The hedonic brain: A functional neuroanatomy of human pleasure. In M. L. Kringelbach & K. C. Berridge (Eds.), Pleasures of the brain. Oxford: Oxford University Press. Lambie, J. A., & Marcel, A. J. (2002). Consciousness and the varieties of emotion experience: A theoretical framework. Psychological Review, 109, 219–259.

116 James A. Russell

Lang, P. J. (1995). The emotion probe: Studies of motivation and attention. American Psycholo­ gist, 50, 372–385. Larsen, J. T., Bernstson, G. G., Poehlmann, K. M., Ito, T. A., & Cacioppo, J. T. (2008). The psychophysiology of emotion. In M. Lewis, J. M. Haviland-Jones, & L. F. Barrett (Eds.), Handbook of emotions, (3rd ed.) (pp. 180–195). New York: Guilford. LeDoux, J. (1996). The emotional brain. New York: Touchstone. LeDoux, J. (2008). Emotional colouration of consciousness: How feelings come about. In L. Weiskrantz & M. Davies (Eds.), Frontiers of consciousness: Chichele lectures (pp. 69– 130). Oxford: Oxford University Press. Leknes, S., & Tracey, I. (2008). A common neurobiology for pain and pleasure. Nature Reviews Neuroscience, 9, 313–320. Levenson, R. W., Ekman, P., Heider, K., & Friesen, W. V. (1992). Emotion and autonomic nervous system activity in the Minangkabau of West Sumatra. Journal of Personality and So­ cial Psychology, 62, 972–988. Levy, R. I. (1973). Tahitians. Chicago: University of Chicago Press. Lorenz, K. (1966). On aggression. London: Methuen. Lutz, C. (1988). Unnatural emotions: Everyday sentiments on a Micronesian atoll and their chal­ lenge to western theory. Chicago, Ill: University of Chicago Press. MacDonald, S. E., & Pinel, J. P. J. (1991). Information gathering: A component of the defensive behavior of rats and Old-World monkeys. Psychological Record, 41, 207–215. Moors, A. (2009). Theories of emotion causation: A review. Cognition and Emotion, 23, 625–662. Morice, R. (1978). Psychiatric diagnosis in a transcultural setting: The importance of lexical categories. British Journal of Psychiatry, 132, 87–95. Neumann, R. (2000). The causal influences of attributions on emotions: A procedural priming approach. Psychological Science, 11, 179–182. Ochsner, K. N., Ray, R. R., Hughes, B., McRae, K., Cooper, J. C., Weber, J., Gabieli, J. D. E., & Gross, J. J. (2009). Bottom-up and top-down processes in emotion generation. Psychologi­ cal Science, 20, 1322–1331. Ortony, A., Clore, G. L., & Collins, A. (1988). The cognitive structure of emotions. Cambridge, UK: Cambridge University Press. Panksepp, J. (2005a). Affective consciousness: core emotional feelings in animals and humans. Consciousness and Cognition, 14, 30–80. Panksepp, J. (2005b). Toward a science of ultimate concern. Consciousness and Cognition, 14, 22–29. Panksepp, J. (2008). Carving natural emotions: Kindly from bottom-up but not top-down. Jour­ nal of Theoretical and Philosophical Psychology, 28, 395–422. Petrovich, G. D., Ross, C. A., Mody, P., Holland, P. C., & Gallagher, M. (2009). Central, but not basolateral, amygdala is critical for control of feeding by aversive learned cues. Journal of Neuroscience, 29, 15205–15212. Pfaff, D. (2006). Brain arousal and information theory: Neural and genetic mechanisms. Cambridge, MA: Harvard University Press. Pinel, J. P. J., & Mana, M. J. (1989). Adaptive interaction of rats with dangerous inanimate objects: Support for a cognitive theory of defensive behavior. In R. J. Blanchard, P. F. Brain, D. C. Blanchard, & S. Parmigiani (Eds.), Ethoexperimental approaches to the study of be­ havior (pp. 137–150). Norwell, MA: Kluwer Academic.



3.  From a psychological constructionist perspective 117

Posner, J., Russell, J. A., & Peterson, B. S. (2005). A circumplex model of affect: An integrative approach to affective neuroscience, cognitive development, and psychopathology. Devel­ opment and Psychopathology, 17, 715–734. Reisenzein, R. (1992). A structuralist reconstruction of Wundt’s three dimensional theory of emotion. In H. Westmeyer (Ed.), The structuralist program in psychology: Foundations and applications (pp. 141–189). Toronto, Ontario, Canada: Hopgrefe & Huber. Reisenzein, R., & Schonpflug, W. (1992). Stumpf ’s cognitive–evaluative theory of emotion. American Psychologist, 47, 34–45. Roberson, D., & Davidoff, J. (2000). The categorical perception of colors and facial expressions: The effect of verbal interference. Memory and Cognition, 28, 977–986. Russell, J. A. (1991a). Culture and the categorization of emotion. Psychological Bulletin, 110, 426–450. Russell, J. A. (1991b). Natural language concepts of emotion. In D. J. Ozer, J. M. Healy, Jr., & A. J. Stewart (Eds.), Perspectives in personality: Self and emotion (pp. 119–137). London: Jessica Kingsley. Russell, J. A. (2003a). Core Affect and the psychological construction of emotion. Psychological Review, 110, 145–172. Russell, J. A. (2003b). Introduction: The return of pleasure. Cognition and Emotion, 17, 161–165. Russell, J. A. (2005). Emotion in human consciousness is built on Core Affect. Journal of Con­ sciousness Studies, 12, 26–42. Russell, J. A. (2006). Emotions are not modules. Canadian Journal of Philosophy, Supplement Volume, 32, 53–71. Russell, J. A. (2008). In defense of a psychological constructionist account of emotion: Reply to Zachar. Journal of Theoretical and Philosophical Psychology, 28, 423–429. Russell, J. A. (2009). Emotion, Core Affect, and Psychological Construction. Cognition and Emotion, 23, 1259–1283. Russell, J. A., & Pratt, G. (1980). A description of the affective quality of environments. Journal of Personality and Social Psychology, 38, 311–322. Russell, J. A., & Sato, K. (1995). Comparing emotion words between languages. Journal of CrossCultural Psychology, 26, 384–391. Rozin, P. (2007). Food and eating. In S. Kitayama & D. Cohen (Eds.). Handbook of Cultural Psychology (pp. 391–416). New York: Guilford. Scherer, K. R. (2009). Emotion theories and concepts (psychological perspectives). In D. Sander­ & K. R. Scherer (Eds.), The Oxford companion to emotion and the affective sciences (pp. 145– 151). Oxford: Oxford University Press. Schwarz, N., & Clore, G. L. (1983). Mood, misattribution, and judgments of well-being: Information and directive functions of affective states. Journal of Personality and Social Psychol­ ogy, 45, 513–523. Searle, J. R. (1992). The rediscovery of the mind. Cambridge, MA: MIT Press. Smith, E. R., & Neumann, R. (2005). Emotion processes considered from the perspective of dual-process models. In L. F. Barrett, P. M. Niedenthal, & P Winkielman (Eds.). Emotion and consciousness (pp. 287–311). New York, NY: Guilford. Smith, K. S., Mahler, S. V., Peciña, S., & Berridge, K. (2010). Hedonic hotspots: Generating Sensory pleasure in the brain. In M. L. Kringelbach & K. C. Berridge (Eds.), Pleasures in the brain. Oxford: Oxford University Press. Solomon, R. C. (1976). The passions. Garden City, NY: Anchor Press.

118 James A. Russell

Stanley, D. J., & Meyer, J. P. (2009). Two-dimensional affective space: A new approach to orienting the axes. Emotion, 9, 214–237. Strack, F., & Deutsch, R. (2004). Reflective and impulsive determinants of social behavior. Per­ sonality and Social Psychology Review, 8, 220–247. Tsai, J. L. (2007). Ideal affect: Cultural causes and behavioral consequences. Perspectives on Psychological Science, 2, 242–259. Weiner, B. (1985). An attributional theory of achievement, motivation, and emotion. Psycho­ logical Review, 92, 548–573. Weiskrantz, L. (2008). On the ubiquity of conscious-unconscious dissociations in neuropsychology. In L. Weiskrantz and M. Davies (Eds.), Frontiers of consciousness: Chichele lectures (pp. 323–334). Oxford: Oxford University Press. Wierzbicka, A. (2009). Language and metalanguage: Key issues in emotion research. Emotion Review, 1, 3–14. Yik, M. (2009). Studying affect among the Chinese: The circular way. Journal of Personality As­ sessment, 91, 416–428.

chapter 4

“Nature proposes…and science disposes” tertiary vs primary process approaches to emotions and affects Commentary on Jim Russell’s position Jaak Panksepp

Emotionality, with all its diverse manifestations, from affective feelings to behavioral and autonomic consequences, is a natural process of the BrainMindBody. Affective issues also percolate to the highest reaches of the MindBrain. My research and theoretical focus has been on how the raw affective side of mental life arises from homologous neural systems in mammals. This provides neuroevolutionary­ approaches that are essential for understanding what affects really are. Jim ­Russell studies the fuller complexities of affective life, but recently proposed a primary-process type concept with Core Affect. It subsumes traditional dimensional analyses of affect (with orthogonal arousal and valence variables) but I question whether the valenced infrastructure of Core Affect is just dimensional, or also has more discrete aspects. Long-standing “battles” between Constructivist and Basic Emotion views of affect are of little relevance from my primary-process cross-species neuroscientific­ perspective. Generally, thorough analysis of primary-processes requires ­ animal models. Russell’s concept of Core Affect has yet to be neurologized, which presumably would require research models like mine. Obviously, neuroscience and psychology need to work together to understand how our affective dynamics were constructed by evolution. The cultural aspects of affect need to be explicated by anthropologists, psychologists, sociologists, and scholars in the humanities. Those sciences, at higher levels of analysis, typically look toward more basic sciences for causal insights. Russell hopes the job of untangling how Core Affect emerges from brain activities will be elucidated by neuroscientists. My work may provide some help.

120 Jaak Panksepp

Of course, the hoped-for partnership would require psychologists to frame their questions in ways that can be linked to causal brain research. Correlative ­issues can already be addressed with modern functional brain imaging (vide ­infra), but the causal issues cannot. Arousal controls can be substantively addressed neuroscientifically (Pfaff, 2006), but how they relate to psychological issues is not well understood. Causal analysis of valence has been partly addressed by animal work, through studies of subcortical networks that not only control distinct emotional actions but mediate rewarding and punishing effects. To surmount problems for achieving productive cross-disciplinary fertilization, those working at higher psychological levels must consider the utility of existing neuroscientific evidence, much of which has been collected using basic-emotion models quite different from those working at psychology-only levels of analysis. Evidence at primary-process levels already supports a multiplicity of “basic affects” – a larger category than “basic emotions” since there are distinct sensory and homeostatic affects. This does not mean we must discard dimensional theories of Core Affect, but at least consider whether they may work best at higher tertiatry-process (e.g., verbal, conceptual) levels of analysis. The primary-process level analyses cannot currently eliminate “basic emotion” theory, because so many distinct emotional coherences are evolutionarily built into the brain.

1. The affective neuroscience of raw emotional feelings The most compelling primary-process evidence for distinct affects comes from the imposition of totally un-natural forms of experimental “stimuli” into animal brains – electrical and chemical stimulation to specific subcortical regions. Not only can we evoke diverse unconditioned emotional responses in this way, indicating emotional coherence was constructed, in part, by evolution, but we can evaluate the affective properties of these networks. Empirical demonstrations of affective experiences consist of using such artificial brain stimulations as “rewards” and “punishments.” The correspondence of such reward and punishment sites with strong evocation of emotional feelings in humans is compelling (for early work, see Heath, 1996; Panksepp, 1985). It supports the thesis that distinct affects were built into subcortical brain regions by evolution. It would be a valuable heuristic maneuver if Core Affect could be situated in the database I have spearheaded.



4.  “Nature proposes…and science disposes” 121

2. The dimensionality of affective experience If we constrain our debate to primary-process issues, we need to ask whether “dimensional” or “categorical” views of affect currently have more evidence. In my estimation it is the latter for valence. Arousal is not troublesome. Pfaff (2006) has well summarized data, accepted by most psychobiologists for the past three decades (Panksepp, 1986, also see Figure 4, in Panksepp, 1982: 419), that diverse brain systems contribute to emotional (and cognitive) arousal. Many distinct neurochemical systems have been implicated (acetylcholine, dopamine, epinephrine, histamine, norepinephrine, and serotonin) as well as a host of neuropeptides (corticotrophin releasing factor to orexin, so to speak), all of which interact in multiple ways. Working together, they may simply evoke one type of psychological arousal, but we must remain alert to distinct patterns of arousal under different emotional states. Pfaff may prefer a unitary arousal concept at the psychological level (his book, p. 7), but it is empirically undecided (p. 142). Good neuropsychological studies are needed to resolve ambiguities. Animal research can address differential emotional valences more easily. Indeed, we may wish to broaden discussions to include non-emotional affects like hunger, thirst, etc. Regrettably, I phrased my invitation to do this poorly in my opening essay (i.e., “if they only considered all the facts” p. 45). Evidence suggests they are subcortically elaborated with striking correspondences in animal and human research (Denton, 2006; Panksepp, 1974). I mistakenly thought that Russell’s Core Affect was also attempting to fully account for such states. That seems not to be the case. My current understanding is that there do exist a variety of distinct valence dynamics in all mammalian brains. In any event, primary-process homeostatic and sensory affects are surely important for understanding the foundational affective infrastructures of the mind and perhaps Core Affect.

3. Kantian critique: Concepts without evidence are empty. Facts without concepts are blind Jim Russell says Core Affect is psychologically primitive and cannot be analyzed further without venturing into biology. If we did venture into biology, how would that analysis proceed to fertilize psychological understanding? How could neurobiological analyses provide valuable new perspectives for conceptualizing Core Affect (as well as everyday feelings that humans and animals have)? We all agree that testable hypotheses are essential, and hopefully together we can craft differential predictions that can resolve theoretical options. I claim that further biological

122 Jaak Panksepp

analysis would support discrete affects, but the critical evidence would be at the primary-process level, which is not open to analysis through the tertiary-process information on which most psychological analysis relies. It may be that dimensionality of affect provides one reasonable conceptual framework to translate my findings at subcortical levels to higher mental functions, but it is also possible primary-process events cannot be rigorously resolved in human psychological analyses. In any event, I would love scholars working at psychological levels to conduct comprehensive open-ended surveys inquiring about emotional experiences of tens of thousands of individuals at different ages within many different cultures (e.g., emulating Buss’s 1989 work). Surely we need to know which kinds of emotional feelings different people actually claim they experience and thereby acquire fuller descriptions of our affective lives. It would be good to catalog (i) those feelings that are perceived aspects of everyone’s makeup, (ii) those perceived as learned aspects while growing up in different cultures, and (iii) those perceived to arise from combinations or mixtures of first and second list replies. Have such large-scale psychological surveys ever been done? They could serve as universal databases for all psychological levels of theorizing and might reduce needless squabbling. Russell sees problems, as do I, in scientifically analyzing emotions with traditional “folk-psychological” terms, but that admirable goal is not inconsistent with the need for more comprehensive phenomenological surveys in the field. This has been done with those who rear animals in the UK, and the results seem to support distinct categories of affect (Morris, et al., 2008), but it is easy to see why (that is the way we communicate). Just because something is “folk science” does not automatically make it “bad science”. My own research has arisen from different intellectual traditions than practically all other psychological investigations of emotions, including the classical “basic emotion theories” that are often contested by constructionists. My view of emotionality is ruthlessly empirical but not ruthlessly reductionistic (i.e., unlike most neuroscientists, I have no urge to discard experience as if it were an epiphenomenon in the brain, and have advanced ways affective consciousness might arise in the brain and what causal functions it may serve). We all agree that the brain and consciousness, understood by no one yet, is infinitely complex; any linkages to specific psychological issues will need various simplification stages for any systematic progress at all.



4.  “Nature proposes…and science disposes” 123

4. Russell’s explicit questions Jim Russell explicitly asked me to address four issues. These are challenging questions and none can be answer precisely. Such questions have not been studied with sufficient precision either psychologically or neuroscientifically. However, in the spirit of perspective sharing (I would also appreciate Jim’s views, especially on #2 and #3): (1) “Given that what a Panksepp circuit produces is only a part of an emotional episode, what produces the other objective parts?” As our unconditioned emotional-affective responses, which intrinsically energize and guide behavior, are blended with learning, we develop more precise and thereby more discrete emotional responses to the world. As they proceed through the sieve of higher tertiary-mental processes, especially verbal concepts, we integrate these feelings in vast idiographic autobiographical landscapes of mind that allow autonoetic consciousness, with the capacity for mental time-travel, mindmirroring, and the capacity to speak, listen and understand. Artists of all kinds make many related “objective parts”. (2) “My questions concern the precise nature of those feelings. Admittedly, this is a difficult topic, especially in non-human animals, but advancing the field requires that we propose testable hypotheses as specific as possible.” Psychological precision is not possible in discussing the minds of other animals, nor of humans. However, affective qualities of emotional circuits in animals can certainly be studied empirically even to the point of discriminating between different forms of negative and positive affects through discrimination studies. As Darwin noted, studies of emotional vocalizations are especially important (Brudzynski, 2010), and wherever in the brain we evoke happy play chirps, rats self-stimulate the circuitry (Burgdorf, et al., 2007). I outlined other predictions in my opening piece, and hopefully I will have space for more in our closing remarks. Human words are better suited for describing the complexities of distinct feelings, but even here, Wittgensteinian “word-games” intrude. (3) “My third question concerns the precise conscious experience of having a specific emotion.” Again, the uncertainty principle applies. For the time being, I think it might be useful to envision primary-process emotional feelings as resembling the dynamics (perhaps chaotic attractor landscapes) of the instinctual emotional actions that

124 Jaak Panksepp

animals and human children readily exhibit (Panksepp, 2000), and can be evoked by localized brain stimulation of various “sweet” and “sour” spots in the brain. (4) “First, what does the fear circuit alone produce?” I would suggest that the feeling is “fear itself ” with no object relations. Second, “what mechanisms besides the fear circuit are needed to generate the subjective conscious experience of fear?” We do not yet have an accepted evolutionary view of consciousness, and my only empirical goal has been to provide an empirically based vision of the lowest anoetic forms of consciousness, in this case affective consciousness, based on a coreself concept (Panksepp, 1998, 2007) although I remain interested in the noetic and autonoetic forms of awareness (Vandekerckhove & Panksepp, 2009). I expect that higher subjective experiences of fear – various anxieties, worries and trepidations (Panksepp, et al., 2011) – may require better understanding of higher-order forms of consciousness, which we do not yet have, but which is essential for “access consciousness” – namely our capacity to be aware of the fact that we are having affective experiences. In saying this, I agree with Russsell’s “possibility that the awareness of an affective quality is a cold cognitive process” serving many “other psychological events.” Perhaps space will allow us to flesh out some of these ideas in our final statements.

5. Concluding challenges and reflections There are many ideas with which Jim Russell and I agree. We both envision “emo­ tional meta-experience… . as a perceptual” (or cognitive) “process that relies on concepts” and “that language is responsible for the acquisition of at least some emo­ tion concepts” albeit I believe that at primary-process levels it is worth “thinking of emotions as causal entities” and most especially that “advancing the field requires that we propose testable hypotheses as specific as possible.” We both probably agree that ultimately basic psychological constructs like Core Affect need to be cashed out in neuroscientific terms, and I hope Russell will aspire, with the aid of neuroscientific colleagues, to provide more relevant neuroscientific visions. I have shared (and evaluated) enough “testable hypotheses” in my own conceptually conservative (but hopefully not impoverished) scientific domain of crossspecies behavioral neuroscience (Panksepp, 2010) – deploying animal models where the evolutionary sources of our emotions can be systematically studied and partly understood. Regrettably, I remain unsure about Russell’s position on the utility of concepts such as primary-process emotionality and various types of ­affects (e.g., emotional, homeostatic and sensory), and how his concept of Core Affect relates to the data and theoretical visions I have shared.



4.  “Nature proposes…and science disposes” 125

I would ask Russell to explore the following two major issues – the viability of dimensional theories of valence at the primary-process level (I see no problem with its utility at the tertiary-process conceptual level), and the utility of Core Affect in sustaining progressive lines of neuroscientific research. Dimensional theories of valence are well situated for dealing with the constructed emotional complexities of our uniquely thinking higher minds, but perhaps not our lower affective-evolutionary mind. Might Russell agree that dimensional theories have progressed almost exclusively through studies of linguistic (neocortical) decoding of affective states? I am not aware of any robust data suggesting that Core Affect evolved in the brain above and beyond the kinds of brain functions I have been studying; so far it seems to be a man-made construct with no independent neuroscientific content. At the same time it is quite clear Russell’s Core Affect and my primary-process affects would “not have evolved unless” they served “a function.” But what is Russell’s evidence that Core Affect evolved? I can point to many neurally based affective homologies. Does this differentiate our approaches/theories? We remain at the beginning of scientific clarifications of how emotional feelings and other affects emerged in BrainMind evolution. Many approaches will contribute to this understanding. Animal models will be more important than most emotion researchers acknowledge or perhaps realize. And I would forewarn aficionados of brain imaging, ever increasingly enthused by affect research, to recognize various difficulties they must confront: (i) Many studies evaluating affects confound cognitive demands with affective changes – troublesomely, brain signals from higher cognitive activities are reciprocally related to affective feelings (Liotti & Panksepp, 2004). Thus, great care must be taken to disentangle ongoing cognitive evaluations from measurements of experienced affective changes. (ii) Proper correlation of brain changes with affective changes requires sensitivity to how much correlations change simply from data juggling – e.g., raw-data variability is commonly reduced through prior data-averaging into clusters, thereby artifactually inflating correlations (Panksepp, 1973: Vul, et al., 2009). (iii) Metaanalyses are lovely, but investigators interested in the “boundary conditions” of experienced primary-process affects would be wise to prioritize findings from PET studies, especially relevant neurochemical ones (e.g., Zubieta, et al., 2003), over most existing fMRI studies; fMRI methodologies, because of strict temporal constraints, simply don’t allow subjects to get into intense affective states as readily as PET procedures (e.g., Damasio, et al., 2000). It should be recognized that the best neuro-causal evidence comes from brain stimulation and psychopharmacological studies, done most systematically in animal models I ask: Will not our understanding of Core Affect remain incomplete without considering deep (cross-species) evolutionary issues? Arbitrary fencing off of

126 Jaak Panksepp

spheres of knowledge, as behaviorists did through their disdain for the “Black-Box”, is not to be commended in our era, where evolutionary epistemology, through the study of diverse animals, is so well developed. Accordingly, I encourage Russell to more fully clarify the role of cross-species “neuro-psycho-epistemology” for solving the Core Affect problem/solution he has posed for himself. This is not an easy task. In my estimation, psycho-evolutionary clarity is currently more easily achieved with subcortical studies of our animalian affective BrainMinds than through the study of the tertiary-process stratosphere of our uniquely human, deeply thoughtful (Premack, 2010), but developmentally constructed, conceptual minds. These varieties of mind need to be seamlessly integrated in theory and life. I respect Russell’s major contributions to the field and the perspectives of the many dedicated scholars who participated in this conversation, as we learn to deeply appreciate each other’s (i) theoretical perspectives, (ii) epistemological limitations and (iii) vastly different, but all highly important, research. Arguing that some approaches are more important than others, as I sometimes do, is less productive than seeing how all relevant approaches can be integrated.

References Brudzynski, S. M. (Ed.) (2010). Handbook of mammalian vocalizations: an integrative neuroscience approach. London & San Diego: Academic Press/Elsevier. Burgdorf, J., Wood, P. L., Kroes, R. A., Moskal, J. R., & Panksepp, J. (2007). Neurobiology of 50-kHz ultrasonic vocalizations in rats: electrode mapping, lesion, and pharmacology studies. Behavioral Brain Research, 182, 274–283. Buss, D. M. (1989). Sex differences in human mate preferences: evolutionary hypotheses tested in 37 cultures. Behavioral and Brain Sciences, 12, 1–14. Damasio, A. R., Grabowski, T. J., Bechara, A., Damasio, H., Ponto, L. L. B., Parvizi, J., & Hichwa, R. D. (2000). Subcortical and cortical brain activity during the feeling of self-generated emotions. Nature Neuroscience, 3, 1049–1056. Denton, D. (2006). The primordial emotions: the dawning of consciousness. New York: Oxford University Press. Liotti, M., & Panksepp, J. (2004). Imaging human emotions and affective feelings: implications for biological psychiatry. In J. Panksepp (Ed.), Textbook of Biological Psychiatry (pp. 33– 74). Hoboken, NJ: Wiley. Morris, P. H., Doe, C., & Godsell, E. (2008). Secondary emotions in non-primate species? Behavioural reports and subjective claims by animal owners. Cognition and Emotion, 22, 3–20. Panksepp, J. (1973). A reanalysis of feeding patterns in the rat. Journal of Comparative and Physiological Psychology, 82, 78–94. Panksepp, J. (1974). Hypothalamic regulation on energy balance and feeding behavior. Federa­ tion Proceedings, 33, 1150–1165.



4.  “Nature proposes…and science disposes” 127

Panksepp, J. (1982). Toward a general psychobiological theory of emotions. The Behavioral and Brain Sciences, 5, 407–467. Panksepp, J. (1986). The neurochemistry of behavior. Annual Review of Psychology, 37, 77–107. Panksepp, J. (1998). The periconscious substrates of consciousness: affective states and the evolutionary origins of the SELF. Journal of Consciousness Studies, 5, 566–582. Panksepp, J. (2000). The neurodynamics of emotions: an evolutionary-neurodevelopmental view, In M. D. Lewis & I. Granic (eds.), Emotion, Self-Organization, and Development (pp. 236–264), New York: Cambridge Univ. Press. Panksepp J. (2007). Affective Consciousness. In M. Velmans & S. Schneider (Eds.), The Black­ well Companion to Consciousness (pp. 114–129). Malden, MA: Blackwell Publishing, Ltd. Panksepp, J. (2010). Foreword: perspectives on passages toward and affective neurobiology of mind? In G. Koob, R. Thompson, & M. LeMoal (Eds). Encyclopedia of Behavioral Neuroscience. Elsevier, in press. Panksepp, J., Fuchs, T., & Iacobucci, P. (2011). The basic neuroscience of emotional experiences in mammals: The case of subcortical fear and implications for clinical anxiety. Applied Animal Ethology, 129, 1–17. Pfaff, D. (2006). Brain arousal and information theory: neural and genetic mechanisms. Cambridge, MA: Harvard University Press. Premack, D. (2010). Why humans are unique: three theories. Perspectives on Psychological Science, 5, 22–32. Vandekerckhove, M., & Panksepp, J. (2009). The flow of anoetic to noetic and autonoetic consciousness: a vision of unknowing (anoetic) and knowing (noetic) consciousness in the remembrance of things past and imagined futures. Consciousness & Cognition, 18, 1018–1028. Vul, E., Harris, C., Winkelman, P., & Pashler, H. (2009). Puzzlingly high correlations in fMRI studies of emotion, personality, and social cognition. Perspectives on Psychological Science, 3, 274–290. Zubieta, J. K., Ketter, T. A., Bueller, J. A., Xu, Y., Kilbourn, M. R, Young, E. A., & Koeppe, R. A. (2003). Regulation of human affective responses by anterior cingulate and limbic mu-opioid­ neurotransmission. Archives of General Psychiatry, 60, 1145–1153.

chapter 5

Preliminary comments on Panksepp James A. Russell

In the end, analyses of emotion at biological, psychological, and social-cultural levels must complement one another. We are not going to develop a successful science of emotion without understanding how the brain creates emotion. Nor without understanding the psychological and social-cultural processes involved. Still, we are far from the end, and a preliminary discussion between two researchers working at different levels – what Panksepp (p. 65) called a cross-cultural dialog – cannot be expected to sail along without some mishaps. Before we turn the discussion over to the commentators, it might be helpful if I refine, clarify, or sharpen some points. Here are six. First, throughout his chapter, Panksepp inferred specific conscious affective feelings (feeling afraid, feeling enraged, and so on) from specific behaviors or neural activities. What is the justification for this inference? A justification cannot be found in the general assumption (which Panksepp and I share) that human and non-human animals are conscious. The justification must be specific to the particular behavior, particular neural activity, and particular conscious experience in question, because behavior and neural activity can occur without consciousness even in fully conscious adult humans. For example, the brain constantly monitors and regulates the blood – blood volume, pressure, salinity, acidity, oxygen concentration, glucose concentration, etc. – all nonconsciously. From the existence of this ancient, vital neural activity, we cannot directly infer conscious access to that activity. My chapter mentioned cases in which a fully conscious human adult underwent the autonomic and behavioral changes characteristic of fear, for example, without simultaneously having the conscious feeling (the Emotional Meta-experience­) of being afraid. If so, might not the same occur in non-human animals? But let us suppose that in Panksepp’s experiments the animals are having some conscious experience. The challenging question is its precise form. For example, the form of the conscious experience of hunger cannot be inferred from its underlying neurophysiology. Hunger is moderately complicated at the conscious level, mindbogglingly complicated at the neurophysiological level. In the account I am

130 James A. Russell

defending, “feeling afraid” is not a simple sensation but rather a complex perception. We need a detailed empirical justification to allow a convincing description of the precise form of the animal’s conscious experience. We must examine the relation of consciousness to each specific type of neural activity separately. We must ask why some neural activities lead to consciousness whereas others do not, what mechanism brings a certain neural activity into consciousness, what function is served by doing so, what is the nature of the corresponding conscious experience, and why consciousness takes the form it does. We need an empirical approach to consciousness to replace, on the one side, a priori dismissal of consciousness and, on the other, a priori equation of consciousness with behavior or neural activity. Second, in support of his theory, Panksepp cited human brain imaging studies. He wrote, “it seems that practically every anxiety ever conceptualized and evaluated with brain imaging has yielded arousal of the amygdala” (p. 50–51). Indeed, but one problem is that the fear-amygdala association is the only association between a specific emotion and a specific brain location to emerge consistently across the meta-analyses of brain imaging data of which I am aware (Lindquist et al., in press; Murphy et al., 2003; Phan et al., 2002; Vytal & Hamann, 2010). ­Another problem is that the relation of the amygdala to fear is not simple. Bilateral destruction of the amygdalas does not eliminate an unconditioned fear response (Bechara et al., 1995). And the amygdala modulates neural systems underlying various cognitive and social functions (Phelps, 2006). Third, Panksepp and I used the phrase Core Affect differently. To understand my concept of Core Affect, certain distinctions and definitions should be kept in mind. In my hypothesis of Core Affect, intensity and arousal are not equated. The hypothesis of Core Affect is independent of the hypothesis of Psychological Construction. Core Affect is not on a par with the terms emotions, moods, or af­ fects in that such terms are superordinate categories, whereas Core Affect refers to a primitive feature of conscious experience. I coined the phrase Core Affect and made it a proper noun by capitalizing the first letters to indicate that Core Affect names a technical term in my account. I chose the phrase Core Affect as shorthand for a specific hypothesis because, to my knowledge, no one had used the phrase before, and I could therefore define it in a specific way. Panksepp used the phrase, sometimes with capital letters, something without, sometimes in the singular, sometimes in the plural, as if it were vague in the way most common everyday phrases are vague. I could not pinpoint the meaning of the phrase as he used it. Fourth, Panksepp (p. 48) wrote of “investigators of emotional words.” Such talk may inadvertently sound dismissive, as if my colleagues and I simply read dictionaries. Our topic is not words but the events to which those words refer. Our research uses various methods, many – although not all – of which do indeed



5.  Preliminary comments on Panksepp 131

involve words. When the topic is human consciousness, asking people to tell us of their experiences is an essential method. Of course, we cannot simply take their answers at face value. We need to understand the nature of the words they use. In the attempt to do so, researchers were inspired by Wittgenstein’s philosophical analysis to overturn long-held assumptions about the nature of words (Rosch, 1978), including emotion words (Averill, 1982; Fehr & Russell, 1984). We must also consider linguistic evidence of similarities and differences in the emotion words of different languages (Russell, 1991; Wierzbicka, 1999). Because words also play a role in scientists’ thinking and writing about emotion, a greater understanding of the nature of emotion words helps free us from unexamined naïve assumptions. Fifth, when Panksepp addressed “models of affect that posit two fundamental dimensions” (p. 41), he oversimplified. He seemed to suppose that such models claim there are only two dimensions in all human experience. This false supposition led him to find that “it is a challenge for me to imagine all this evolutionary complexity” – various emotions such as fear, anger, lust, maternal love, etc; pleasures and displeasures of sounds, tastes, pains, etc; and homeostatic mechanisms such as hunger and thirst – all accounted for with only two dimensions. He wrote, “dimensional models of affect would run promptly into trouble (if they only considered all the facts)” (p. 45). What facts? The fact, for instance, that deprived of food, people get hungry rather than thirsty. The fact that when hungry, they desire food rather than water: “No human would … confuse hunger and thirst” (p. 46). Do such facts cause trouble for dimensional models of affect? These models do not specifically address hunger, thirst, lust, pain, and the rest. They are focused on what are called emotions and moods. I cannot speak for all dimensional theorists, but I do not interpret any dimensional model of affect as implying that there are only two dimensions to all human experience. None implies that people confuse hunger with thirst or any other experience. To speak for myself, I do not propose that Core Affect is the only human experience. Even for emotions and moods, Core Affect alone cannot describe all that goes on. Core Affect is but one component of an emotional episode. Core Affect is the main ingredient of a mood, but other dimensions, such as time, need to be taken into account. Ditto for hunger, thirst, lust, and so on. Where Panksepp and I differ is not in the existence of many distinct experiences, but in how to characterize and account for them. I know little about such experiences as hunger, thirst, pain, and so on. In the spirit of this dialog, however, let me speculate by offering an analogy with my account of experiences of emotion (“emotional meta-experiences”). At a psychological level, hunger, thirst, pain and the rest are perceptions. As with other percepts, they depend on sensory input, attention, context, memory, beliefs, social

132 James A. Russell

situation, and so on. Various hunger experiences bear a family resemblance to one another. Rather than a simple feeling, each hunger percept is a Gestalt consisting of components. The components of a typical hunger experience include various sensations (sight, smell, and taste of food; bodily signals) all with affective qualities, a desire for food, time since last meal, memories, one’s current goals and plans, and a change in Core Affect. A conscious experience of thirst could have a similar Core Affect but different sensations and the unique component of a desire for water. Pain too is a perception with information components and a Core Affect component. Thus, some drug treatments for pain abolish the Core Affect component but not the information component. Patients treated with these drugs report that they still feel the pain, but are not bothered by it. Tomkins (1962, 1963) offered a similar analysis from his basic-emotion perspective. The experience of oxygen deprivation typically includes panic (in my terms, a Core Affect of displeasure + high arousal), but slow oxygen deprivation at high altitudes does not. At high altitudes, pilots slowly deprived of oxygen go happily to their death. Sixth, in responding to the question of how his concepts might play a role in my account, Panksepp proposed critical experiments. One set of these critical experiments would test the hypothesis that damage to or elimination of the higher brain regions involved in reward and punishment would damage or eliminate the reward and punishment functions of lower brain regions. I do not see these experiments as critical because, first, I do not equate Core Affect with reward and punishment. Learning through reward and punishment occurs in many species, including invertebrate ones, that are unlikely candidates for Core Affect. Second, I do not assume that Core Affect is a function of higher rather than lower brain regions. My definition of Core Affect invokes neurophysiological activity, but does not specify what or where that activity is. In my chapter, I mentioned some candidates, including pleasure (Smith, Mahler, Pecuna, & Berridge, 2010) and arousal (Pfaff, 2006) circuits in lower brain regions. Panksepp’s other set of critical experiments would explore whether humans can subjectively distinguish between various affectively charged drug experiences. This proposal echoes his comments elsewhere that Core Affect implies an inability to make distinctions, such as between hunger and thirst. I do not suppose that all affectively charged experiences are nothing but Core Affect. Core Affect is part of rather than the whole of conscious experiences, including drug experiences. Therefore, Core Affect does not imply an inability to distinguish among different drug experiences. Perhaps Panksepp based his proposal on my hypothesis that when, as is typically the case, there are multiple influences on how good or bad, how energized or enervated, one feels, one lacks sure knowledge of how much of the feeling is



5.  Preliminary comments on Panksepp 133

due to each separate source. People attribute their feelings to a source, but can be mistaken. So, if you take a drug that induces displeasure while hungry and thirsty and after a stressful day and while foreseeing danger, you have no mechanism to measure precisely how much each source contributes to the overall feeling. Indeed, some causes operate outside conscious awareness. You can estimate, but you can be wrong. This hypothesis does not, however, imply an inability to distinguish among various experiences.

References Averill, J. R. (1982). Anger and aggression. New York: Springer-Verlag. Bechara, A., Tranel, D., Damasio, H., Adolphs, R., Rockland, C., & Damasio, A. R. (1995). Double dissociation of conditioning and declarative knowledge relative to the amygdala and hippocampus in humans. Science, 269, 1115–1118. Fehr, B., & Russell, J. A. (1984). Concept of emotion viewed from a prototype perspective. Jour­ nal of Experimental Psychology: General, 113, 464–486. Lindquist, K. A., Wager, T. D., Kober, H., Bliss-Moreau, E., & Barrett, L. F. In press. The brain basis of emotion: A meta-analytic review. Behavioral and Brain Sciences. Murphy, F. C., Nimmo-Smith, I., & Lawrence, A. D. (2003). Functional neuroanatomy of emotions: a meta-analysis. Cognitive, Affective & Behavioral Neuroscience, 3, 207–233. Pfaff, D. (2006). Brain arousal and information theory: Neural and genetic mechanisms. Cambridge, MA: Harvard University Press. Phan, K. L., Wager, T., Taylor, S. F., & Liberzon, I. (2002). Functional neuroanatomy of emotion: A meta-analysis of emotion activation studies in PET and fMRI. NeuroImage, 16, 331–348. Phelps, E. A. (2006). Emotion and cognition: Insights from studies of the human amygdala. Annual Review of Psychology. 57, 27–53. Rosch, E. (1978). Principles of categorization. In E. Rosch & B. B. Lloyd (Eds.), Cognition and categorization (pp. 27–71). Hillsdale: NJ: Erlbaum. Russell, J. A. (1991). Culture and the categorization of emotion. Psychological Bulletin, 110, 426–450. Smith, K. S., Mahler, S. V., Peciña, S, & Berridge, K. (2010). Hedonic hotspots: generating sensory pleasure in the brain. In M. L. Kringelbach & K. C. Berridge (Eds.), Pleasures in the brain. Oxford: Oxford University Press. Tomkins, S. S. (1962). Affect, imagery, consciousness: Vol. 1. The positive affects. New York: Springer. Tomkins, S. S. (1963). Affect, imagery, consciousness: Vol. 2. The negative affects. New York: Springer. Vytal, K., & Hamann, S. (2010). Neuroimaging support for discrete neural correlates of basic emotions: A voxel-based meta-analysis. Journal of Cognitive Neuroscience, 22, 2864–2885. Wierzbicka. A. (1999). Emotions across languages and cultures: diversity and universals. Cambridge, UK: Cambridge University Press.

chapter 6

Discrete emotions From folk psychology to causal mechanisms Andrea Scarantino Department of Philosophy, Neuroscience Institute, Georgia State University

1. Introduction Panksepp and Russell are eager to find common ground in their fascinating debate. To combine the theoretical insights each approach has to offer, one central issue of contention must be resolved: Are discrete emotions genuine causal mechanisms with respect to the components commonly associated with them? By “discrete emotion”, I will mean any emotion episode designated by an individually separate and distinct category (e.g. fear, anger, sadness). With respect to fear, for instance, we can ask: Does it bring about the heart rate changes, action tendencies, facial expressions and feelings commonly associated with it? The way we settle the issue of emotional causation has key methodological implications. If discrete emotions are not genuine causal mechanisms, as posited by Russell, they should not be allowed into affective science’s basic explanatory toolbox. If at least some discrete emotions are genuine causal mechanisms, as suggested by Panksepp instead, providing emotion-based explanations of physiological, behavioral, expressive, and phenomenological events should be one of affective science’s primary tasks. In my commentary, I side with Panksepp on the issue of emotional causation. But I will argue that there are key lessons to be learned from Russell’s opposition to standard accounts of discrete emotions. In the spirit of fostering discussion, I will conclude by proposing a revision of Panksepp’s discrete model that takes into account Russell’s critique. To avoid mixing substantive disagreements with disagreements that are merely terminological, I begin with a quick summary of the two accounts, presented in as theory-neutral a fashion as possible.

136 Andrea Scarantino

2. Panksepp on primary, secondary and tertiary affects According to Panksepp, “affects are feelings that guide our thoughts and our actions” (32). On this view, affects are by definition associated with a subjective experience, i.e. a feeling. In other places, however, Panksepp uses the term “affects” more broadly to designate emotional operating systems, on the assumption that such systems produce feelings once activated. To increase clarity, I distinguish between the functional notion of an “affective system”, defined by its ability to coordinate organismic resources towards an end (e.g. a danger-avoiding system), and the phenomenological notion of an “affective feeling”, defined by the subjective experience attached to it (e.g. a feeling of fear). Whether affective systems always produce affective feelings once activated is a question I will leave open. Panksepp distinguishes between three major classes of affective feelings: emo­ tional affective feelings (e.g. the feeling of fear), sensorially triggered affective feel­ ings (e.g. the feeling of pleasure produced by tasting truffles), and homeostatic affective feelings (e.g. the feeling of hunger). Within the class of emotional affective feelings, he distinguishes between pri­ mary-process, secondary-process and tertiary process affective feelings. Primaryprocess affective feelings, a.k.a. “core affects”, are produced by “primary-process emotional systems” (33). There are “at least seven” of those, labeled by Panksepp seeking, rage, fear, lust, care, grief/panic, and play. Each primary system is associated with a genetically based subcortical brain network unconditionally triggered by a limited number of sensory/perceptual inputs connected to major life-challenging circumstances. But it can also be conditionally triggered by neutral stimuli. For instance, primary-process systems can combine with general-purpose learning processes such as “classical and instrumental conditioning” (32) to give rise to secondary-process affective feelings. They can also combine with higher cognitive processes such as “perceiving, thinking, ruminating, fantasizing, etc.” (32) to give rise to tertiary-process affective feelings. Panksepp’s belief in the existence of primary affective systems is largely due to “our ability to artificially activate various kinds of emotional patterns by applying the appropriate kinds of chemical or electrical stimulation to specific subcortical regions of the brain” (38). For instance, Panskepp argues that the electrical stimulation of the fear network can lead animals to manifest fearlike behaviors, from freezing responses to flight responses depending on the intensity of the electric current. . Unless otherwise stated, all page numbers refer associated with Panskepp’s quotes refer to Chapter 2 in this volume.



6.  Discrete emotions: From folk psychology to causal mechanisms 137

Panksepp declares that his research “is almost exclusively devoted to the pri­ mary-process nature of emotions” (33), in particular to the understanding of their “causal infrastructure” at the neural level.

3. Russell on core affect, affective quality and meta-experience Russell begins by stating that he will use the term affect “in a narrow sense limited to private subjective conscious feelings” (85). Later on, however, he allows for the possibility that some varieties of affect (e.g. Core Affect) may occur without being attended to, a circumstance that gives rise to unconscious affect. Russell emphasizes that “affect must be distinguished from what I call an emotional epi­ sode” (85), namely a particular event referred to by everyday English words such as “emotion”, or “fear”, or “anger”. Consider Sally’s fear of a bear in the woods last Saturday at 10 am. For Russell, this episode includes an “affect”, i.e. an affective feeling, jointly with “a sequence of other components as well, such as Sally’s perception of and appraisal of the bear, all the peripheral physiological changes that occurred, facial and vocal changes, her behavioral reaction of freezing and then fleeing, and all the neural processes that underlie these changes” (85). As we shall see, Russell does not think that there is an internal emotion mechanism responsible for the co-instantiation of such components. Russell distinguishes “three different types of affect, without claiming that these three exhaust the domain: Core Affect, Affective Quality, and Emotional Meta-experience” (85). Core affect is “a neurophysiological state that is consciously accessible as a simple, non-reflective feeling that is an integral blend of hedonic (pleasure–displeasure) and arousal (sleepy–activated) values” (Russell, 2003: 147). Like Panksepp’s primary-process emotional systems, Core Affect is assumed to be associated with specific brain networks, and Russell makes some suggestions as to what such networks may be (cf. Pfaff, 2006; Smith et al., 2010). Core Affect need not be directed at anything, but it is “an elemental building block of other psychological events” (88) directed at something, including discrete emotions. Sally’s fear episode, for instance, will include as its proper part a feeling of high arousal and high displeasure directed at the bear. The primary, but by no means exclusive, source of evidence for the existence of Core Affect comes from psychometric studies of self-reported moods and . Unless otherwise stated, all page numbers associated with Russell’s quotes refer to Chapter 3 in this volume..

138 Andrea Scarantino

emotions. These studies have shown that self-reports of how one feels or felt at a particular moment or expects to feel in the future correlate in ways that can be parsimoniously explained along the two dimensions of valence and arousal. In a nutshell, subjects tend to report experiencing moods and emotions with the same valence together and moods and emotions with the same arousal together. If they report feeling afraid, for instance, they also tend to report feeling angry. What underlies this surprising correlation, it is proposed, is that feelings of fear and anger typically have in common high displeasure and high arousal. The second type of affect Russell considers is the “affective quality” of a stimulus, namely “its capacity to change core affect” (Russell, 2003: 149). As with every capacity, the capacity to change Core Affect need not be manifested in order to be possessed. A given sunset on Daytona beach may have had the affective quality of being soothing even though no one observed it, or even though the only person observing it was too depressed to be soothed. The sunset is still soothing because it would lead an observer in suitable circumstances to be soothed by it. The third type of affect distinguished by Russell is “emotional metaexperience­”, which amounts to the categorization of one’s discrete emotion under a concept. “On my account of emotional meta-experience”, Russell suggests, “to perceive oneself as afraid is to categorize oneself by means of the concept of fear” (91). Since concepts have correctness conditions for their application, one’s metaexperience may be mistaken. Sally may categorize herself as being afraid when she is angry instead. Furthermore, anyone who lacks the concept of fear can’t have an emotional meta-experience of fear.

4. Are discrete emotions causal mechanisms? With these terminological preliminaries out of the way, let us consider what I take to be the central issue of contention between Russell and Panksepp. Whereas Panksepp posits seven “core” affective systems, i.e. seeking, rage, fear, lust, care, grief/panic, and play, each associated with a discrete affective feeling, Russell argues that there is only one “core” affective system, i.e. Core Affect, associated with affective feelings resulting from blends of pleasure and arousal values. At first blush, the two accounts are compatible: Panksepp’s seven affective systems could in principle co-exist with Russell’s Core Affect system. Which of such systems should be called “core” strikes me as being more a matter of terminological predilection than theoretical substance. But the devil is in the details. Panksepp believes that there are discrete emotions that cause the suite of physiological, behavioral, expressive, and phenomenological components associated with them. His seven “core” affective systems



6.  Discrete emotions: From folk psychology to causal mechanisms 139

provide an example of causally powerful discrete emotions. Such systems have the “ability to coordinate sets of behavioral and autonomic outputs [and] have a coherent, behavioral-affective coherence promoting infrastructure” (43). Russell’s position is instead that the suite of behavioral, physiological, expressive, and phenomenological components characteristic of episodes of discrete emotion is simply the evidence on the basis of which emotion concepts are ascribed. His view is that “[w]e need no extra mechanism” that is the emotion to explain why physiological, behavioral, expressive, and phenomenological components co-occur to the (limited) extent that they do. An important corollary is that emotion, fear, anger, happiness, sadness and so on are of scientific interest to Russell merely as folk concepts. We can usefully explore which patterns of components are categorized under one folk emotion concept rather than another. This amounts to engaging in a descriptive project, whose objective is the “analysis of what [emotion] words mean in everyday discourse” (80). But when we shift to the prescriptive project, i.e. the “scientific study of the events to which words such as emotion, affect, and jealousy refer” (80), folk emotion concepts are no longer suitable. This is why affective science must abandon “traditional ways of thinking” (79) about emotion, and shift to the study of Core Affect, affective quality and emotional meta-experience. At this juncture we encounter a puzzle. Russell’s rejection of the causal import of discrete emotions is grounded in skepticism about the scientific suitability of folk emotion concepts. Everyday emotion concepts, he writes, “have become a source of difficulty in furthering a scientific analysis of emotion” (80). Yet, ­Panksepp seems to share the same skepticism: “I have pointedly chosen not to use vernacular terms for primary-process emotional systems and their affects” (33). How did skepticism about folk emotion concepts lead Russell and Panksepp in such different directions? Can their differences be reconciled?

5. What’s wrong with folk emotion concepts? Russell and Panksepp are both convinced that folk emotion concepts are unsuitable for scientific purposes. This puts them in contrast with the great majority of emotion theorists, who characterize their objects of investigation in folk psychological terms. Competing scientific theories of emotion/fear/anger/etc. disagree on content, but they share the assumption that there is nothing wrong with emotion, or fear, or anger as scientific concepts. Russell begs to disagree. On his view, folk emotion concepts have five fundamental flaws: they “lack defining properties”, they admit “borderline cases”, they are “culture-specific”, they carry “hidden assumptions” from our religious,

140 Andrea Scarantino

­ hilosophical, and intellectual traditions, and they are “too heterogeneous”. Acp cording to Panksepp, folk emotion concepts have two fundamental flaws: they have “variable meanings due to our socio-cultural diversity” and they generate “part-whole confusions”. I will consider these alleged problems one at a time, and conclude that heterogeneity is the only fatal flaw of folk emotion concepts. The rest of my paper considers what methodological consequences follow from the heterogeneity of folk emotion concepts, specifically with respect to the investigation of discrete causal mechanisms. 5.1

No defining conditions, borderline cases and homeostatic property cluster kinds

Russell rightly emphasizes that it is very difficult to formulate classical definitions of folk emotion concepts such as emotion or fear, namely definitions consisting of individually necessary and jointly sufficient properties for being an episode of emotion or an episode of fear. Despite centuries of attempts, no theory has been able to unveil agreed upon defining properties. Furthermore, there seem to be borderline cases of folk emotion concepts, understood as cases that produce widespread disagreement about membership in the language community. For instance, when asked whether calmness is an emotion, 48% of subjects state that it is, and 52% of subjects state that it is not (Fehr and Russell, 1984). This is the sort of response distribution found when we ask whether a 39-year old person is middle-aged. As there does not seem to be a fact of the matter as to whether a 39-year old person is middle aged, there does not seem to be a fact of the matter as to whether or not calmness is an emotion in the folk sense. The difficulty to find classical definitions and the existence of borderline cases must be explained by a satisfactory theory of folk emotion concepts. Russell’s central proposal is that folk emotion concepts are mental representations of scripts, which stand to events as prototypes stand to objects (Fehr and Russell, 1984). An emotion script “contains prototypical causes, beliefs, physiological reactions, feelings, facial expressions, actions, and consequences” (Fehr and Russell 1984: 482) associated with specific folk emotion categories. For example, the fear script may contain a danger of some kind as a prototypical cause, heightened heartbeats and muscle tension, an unpleasant feeling, a facial expression that includes raised upper eyelids and dropped open jaw, an action tendency of avoidance, and a general physiological preparation for escape. This mental script can be



6.  Discrete emotions: From folk psychology to causal mechanisms 141

the basis for the following categorization rule: something is “fear” in the folk sense just in case it achieves enough similarity with the fear script. This proposal has several virtues. First, it explains why classical definitions are so hard to come by for folk emotion concepts. It would be purely coincidental if all items that achieve enough similarity with some folk emotion E’s script happened to share properties that are individually necessary and jointly sufficient for being E. Second, it explains why there are borderline cases of folk emotion categories, which emerge when the similarity threshold to the relevant script is approached but not exceeded. My point is that lacking a classical definition and having vague boundaries are not obstacles to the scientific suitability of a concept. Biological species concepts are a case in point. There is a great deal of variation among members of the same species. Certain characters are marked as ‘highly variable’; some species are divided into several geographic ‘races’; some larger groups of birds vary in such complex and overlapping ways that they are denoted a ‘species complex’. All this was well known to Darwin and it is what evolutionary theory should lead us to expect. Yet, despite not being classically definable and having blurred edges, species concepts are scientifically suitable. A good way to make sense of this fact is to consider species concepts to be natural homeostatic property cluster (HPC) kinds (Boyd, 1999). Boyd introduced this influential theory of scientific kinds to account for the fact that many of the central theoretical constructs in the special sciences have internal variability and lack sharp boundaries. According to Boyd, “[t]he natural definition of…homeostatic property clusters kinds is determined by the members of a cluster of often co-occurring properties and by the (“homeostatic”) mechanisms that bring about their co-occurrence” (Boyd, 1999: 141). On the HPC view, something is an instance of a biological species concept like Canis Lupus by virtue of sharing a cluster of “often co-occurring” properties, but there are no properties individually necessary and jointly sufficient for membership. Furthermore, some individuals may have just enough of the co-occurring­ properties to place them on the borderline for membership. These co-occurring properties tend to cluster together by virtue of causal mechanisms such as interbreeding, shared ancestors and exposure to common selection pressures. On account of such “homeostatic mechanisms”, the members of the Canis Lupus species tend to imperfectly share properties. Finally, the often co-occurring­ properties of members of this species are of central interest to biology, in the sense that biologists can formulate lots of reliable inductions and explanations about Canis Lupus.

142 Andrea Scarantino

This example demonstrates that we can find scientifically suitable categories that are not classically definable and admit borderline cases. The converse holds as well. We can find scientifically unsuitable categories that are classically definable and lack borderline cases. The ancient astronomical concept of a superlunary object has classical defining conditions – an object is superlunary if and only if it is located beyond the orbit of the moon – and it lacks blurred edges, but it is clearly unsuitable for contemporary astronomy. A critic may retort that neither biological species concepts nor the concept of a superlunary object are vernacular concepts, unlike folk emotion concepts. This is true, but irrelevant to the point being made, which is that lacking defining conditions and having borderline cases are not per se obstacles to the scientific suitability of folk emotion categories. If they were, we could not find scientifically suitable concepts, whether grounded in the vernacular or not, with such properties. Since we can find them, the case against folk emotion concepts must find support elsewhere. 5.2

Cultural specificity and hidden assumptions

Russell and Panksepp have both suggested that there is another source of trouble with folk emotion concepts, namely that they are “culture-specific” (Russell, 80) or have “variable meanings due to socio-cultural diversity” (Panksepp, 33). It is certainly true that emotion concepts differ to some extent across cultures. As Russell reminds us by way of example, Gidjingali speakers do not have any concept that directly translates with “fear”, combining instead shame and fear into the culture-specific concept of “gurakadj” (110). Languages also differ in the overall richness of their emotion vocabulary, and in the extent to which specific emotions are hypercognized or hypocognized, namely associated with a lexically rich or poor conceptual framework (Levy, 1973). These sorts of cultural differences are relevant with respect to affective phenomena that depend on concept use. An example is Russell ’s notion of the metaexperience of an emotion, understood as the categorization of one’s emotion under concepts. A Gidjingali speaker who lacks the concept of fear would not be able to categorize herself under the fear concept (on the other hand, she would be able to categorize herself under the gurakadj concept). This being said, I do not think that cultural differences have any impact on the scientific suitability of folk emotion concepts. An analogy may help. Consider the category we call in English “gold”. Chemists consider it scientifically suitable for chemistry, and they do so because items in the extension of the “gold” category – the extension of a category C is the set of things that fall into C – share lots of



6.  Discrete emotions: From folk psychology to causal mechanisms 143

properties relevant to induction and explanation in chemistry. By contrast, items in the extension of the “jade” category do not, because the category comprises two very different minerals called “jadeite” and “nephrite” which just happen to look alike, and are labeled “jade” simply on account of how they look. These judgments about the scientific suitability of “gold” and “jade” depend entirely on the nature of their extensions in English. The point is that the extension of a category in a given language is not affected by whether other languages have or lack a category inter-translatable with it. Suppose that a speaker of Gidjingali were to describe gold as “gilder”, a concept that includes what in English we call “gold” and “silver”. Quite clearly, this would not make “gold” a scientifically unsuitable category in English. Conversely, suppose that a speaker of Gidjingali were to lack the “jade” category, and distinguished in all cases between “jadeite” and “nephrite”. Once again, this would not make what we call “jade” in English a scientifically suitable category. Russell also remarked that folk emotion concepts bring with them “the assumptions of our religious, philosophical, and intellectual traditions, including such dubious distinctions as that between body and mind or that between reason and emotion” (80). Let us assume for the sake of argument that these are in fact misleading assumptions about emotions. They would have to be accounted for by what Russell calls the descriptive proj­ ect, which aims to reconstruct the “meaning” of folk emotion terms. But these mistaken assumptions do not affect the prescriptive project, which focuses on the scientific study of the events to which the terms refer. As the mistaken popular assumption that if something is water it must be liquid does not stand in the way of the scientific study of water, so the (possibly) mistaken popular assumption that if something is an emotion it must be irrational does not stand in the way of the scientific study of what we call “emotion”. 5.3

Part-whole confusions reinterpreted

Panksepp uses capitalized terms to designate his seven primary systems: seeking, rage, fear, lust, care, grief/panic, and play. These terms preserve orthographic identity with ordinary lower-case English terms such as “seeking”, “rage”, “fear” and so on. What is the capitalization supposed to indicate? Panksepp’s main rationale for using capitalized letters is “to avoid part-whole confusions” (33). Capitalized letters, he says, only refer “to specifiable brain networks that are important parts of intrinsically ambiguous conceptual wholes” (33). This statement is naturally interpreted as follows: the upper-case terms seeking, rage, fear and so on refer to brain networks that are parts of the wholes

144 Andrea Scarantino

designated by the lower-case terms “seeking”, “rage”, “fear”, etc. This would not be a satisfactory rationale for capitalization on two levels. Firstly, if the aim of the capitalized letters were to refer merely to a specifiable brain network, Panksepp’s terminological choice would be infelicitous. Language affords the much clearer expressions “seeking network”, “rage network”, “fear network”, and so on. Secondly, it is not the case that Panksepp’s brain networks are always parts of the wholes designated by the lower-case terms “seeking”, “rage”, “fear”. As pointed out by Russell, there are instances of activation of, say, the seeking network that are not parts of what we call “seeking” in English. Similarly, there are instances of what we call “seeking” in English that do not involve the seeking­ network as a part. I propose another interpretation, according to which seeking, rage, fear, lust, care, grief/panic, and play refer to discrete affective systems that have dedicated brain networks as parts. On this view, Panksepp’s networks are not parts of the “intrinsically ambiguous conceptual wholes” designated by lowercase terms, but rather parts of the seven theoretically motivated coordination systems introduced by Panksepp. Such systems are not folk psychologically defined, but rather defined by a number of characteristics of the brain networks supposed to underlie them (Panksepp 42–44, with my wordings): 1. Primary affective systems are genetically based and their activities are unconditionally triggered by a limited number of sensory/perceptual inputs connected to major life-challenging circumstances 2. Primary affective systems coordinate sets of behavioral and autonomic outputs, including hormonal and immunological parameters, in ways that have proven adaptive in the evolutionary history of the species 3. Primary affective systems gate and modulate incoming sensory inputs relevant to the behavioral and autonomic outputs coordinated 4. The activities of primary affective systems outlast their precipitating circumstances 5. The activities of primary affective systems can be conditionally triggered by neutral stimuli through learning and higher thought 6. The activities of primary affective systems influence, and are influenced by, the activities of the brain networks underlying higher decision making and consciousness 7. Primary affective systems produce affective feelings Panksepp adds that the activities of primary systems “can occur without the last two attributes” (43), which suggests that (6) and (7) are not strictly necessary for



6.  Discrete emotions: From folk psychology to causal mechanisms 145

an affective system to be activated. In light of this caveat, I propose we understand Panksepp’s primary systems as homeostatic property cluster kinds (Section 5.1). On this view, properties (1)–(7) constitute a family of imperfectly co-occurring properties brought about by a subcortical brain mechanism. The problem is that Panksepp did not provide a principled justification for the orthographic identity between the names of his seven primary affective systems and ordinary folk emotion concepts. Why should a “general-purpose appetitive motivational system”, which aims to motivate “animals to find and eagerly anticipate all kinds of resources they need for survival” (34–35) be called seeking? Why should a system “aroused by frustration arising from any attempts to curtail an animal’s freedom of action” and “invigorat[ing] aggressive behaviors” (35) be called rage? Why should a system “designed during brain evolution to help animals reduce pain and the possibility of destruction” (36) and leading to fleeing or freezing behaviors be called fear? Absent a convincing justification, a better strategy would be to use neologisms for the designation of primary affective systems (e.g. the WIS4 affective system, or the threat-coping system). I propose two conditions as individually sufficient to justify orthographic identity. The first is what I call the subset condition. It is satisfied when the capitalized terms apply to a subset of the episodes named by lower-case terms, and nothing else. fear would satisfy the subset condition if all instances of fear were instances of folk psychological “fear”. In this case, Panksepp’s designation would be legitimate, because fear would describe what a particular kind of “fear” is like. The capitalization would simultaneously warn the reader that, even though all instances of fear are instances of “fear”, the converse does not hold: there are instances of “fear” that are not instances of fear. Russell provides one such example – child’s separation anxiety. The second is what I call the significant overlap condition. It is satisfied when the capitalized terms apply to significantly more episodes named by lower-case letters than they apply to episodes to which the lower-case letters do not apply. The significant overlap condition would be fulfilled by fear even if there were instances of fear that are not instances of “fear”, as long as significantly more instances of “fear” than instances of non-“fear” were included in fear. Panksepp’s designation would still strike me as legitimate in this case. This is because most fear episodes, rather than all fear episodes as in the subset condition, would still qualify as “fear” episodes. Admittedly, this second condition is less clear-cut than the first. It relies on a vague notion such as including “significantly more” members of one class rather than another. As a result, there will be borderline cases in which it is unclear whether a “significant overlap” has been achieved. At the same time, the verdict will be clear in many cases. fear, for instance, appears to clearly satisfy the

146 Andrea Scarantino

s­ ignificant overlap condition. Whereas it is easy to find instances of “fear” that are not instances of fear, we would be hard pressed to find more than a handful, if any, instances of fear that are not instances of “fear”. This is to say that the overwhelming majority of cases of fear are also cases of “fear”. On the view I have proposed, this is enough to justify the orthographic identity between “fear” and fear. I emphasize that it is quite possible that, despite including significantly more cases of “fear” than cases of non-“fear”, fear will still comprise a very small number of cases of “fear”. In other words, it is compatible with the rationale for orthographic identity I have offered that most cases of “fear” will not be cases of fear, as long as most cases of fear are cases of “fear”. The fact that certain forms of “fear” are rarely instantiated is not a good reason for lacking a good scientific theory for them. I leave it open for debate whether Panskepp’s seven primary systems all satisfy either the subset condition or the significant overlap condition. 5.4

Heterogeneity is the problem

The real trouble with trying to build affective science around folk concepts lies in the “heterogeneity [of] the extensions of the terms scientists have borrowed from everyday language: emotion, fear, anger, jealousy, mood, and so on” (113). According to what I call the Heterogeneity Hypothesis, instances of folk emotion concepts are too different from one another to participate in the same body of scientific generalizations. If so, no general scientific theory of emotion, or theory of fear, or theory of anger will be forthcoming. Russell has argued for the truth of this hypothesis over almost three decades (cf. Fehr and Russell, 1984). His view is that the majority of emotion theorists still deny – at least implicitly – the Heterogeneity Hypothesis. They believe instead that “[a]ll emotions are produced by (or are) a single type of entity (such as a set of affect programs in the limbic system). That entity produces (or is) an organized pattern of component responses” (81). This assumption extends to lower levels of the emotion taxonomy: all fears, all angers, all disgusts, all shames, all guilts are assumed to be produced by a single type of entity that generates the organized pattern of component responses characteristic of, respectively, fear, anger, disgust, etc. This leads emotion theorists to keep formulating alternative accounts of the allegedly single type of entity that corresponds to emotion, or to fear, or to anger. The striking fact is that they keep failing at this task. For any theory that claims that all emotions are X, or that all fears are Y, or that all angers are Z, counter­ examples can be found (Scarantino, in press).



6.  Discrete emotions: From folk psychology to causal mechanisms 147

The Heterogeneity Hypothesis provides what I consider to be the best explanation for this failure. If the hypothesis is true, the things that fall into folk emotion concepts constitute highly heterogeneous groupings, and there simply is no X such that all emotions are X, no Y such that all fears are Y, no Z such that all angers are Z, and so forth. The Heterogeneity Hypothesis can be further supported by reflections on the origin of folk emotion concepts. Quite clearly, such concepts were not introduced in language to collect homogeneous domains of scientific investigation. Their function was rather to allow speakers to engage in an assorted variety of speech acts, which include describing and predicting states of affairs, making promises, issuing warnings and commitments, broadcasting intentions, offering praise and blame, expressing internal states, faking internal states, recommending or urging courses of action, and so on. It would be purely coincidental, and highly surprising, if the concepts that emerged from these multifarious social practices happened to collect a set of items that are interestingly alike from a scientific point of view. I therefore conclude with Russell that “emotion [or fear or anger etc.] cannot be equated with any one process or component. Emotional episodes are not all of one kind, and they do not all stem from one mechanism dedicated just to emotion [or fear or anger etc.]” (82). The question is: What follows from the heterogeneity of folk emotion concepts?

6. Why do emotion components co-occur? Russell thinks that the heterogeneity of folk emotion concepts calls into question the standard understanding of the causal relation between emotions and their components. Consider the case of fear. “In the standard account”, Russell writes, “fear (or the fear program) made Sally’s heart pound, her palms sweat, her face broadcast danger; fear focused her thoughts on the bear; it made her freeze and then flee” (81). The standard account here exemplified makes two central commitments. Firstly, it is assumed that there is a discrete causal emotion mechanism (or program) responsible for the co-occurrence of the various components (heart pounding, sweaty palms, freezing, etc.). Call this the emotion mechanism assumption. Secondly, it is assumed that the relevant mechanism or program is a “fear mechanism”, which suggests that it is a mechanism common to all cases of fear. Call this is the unique mechanism assumption.

148 Andrea Scarantino

According to Russell, the emotion mechanism assumption and the unique mechanism assumption are both false. I will argue instead that, whereas the unique mechanism assumption is false, the emotion mechanism assumption is true. Now, the absence of a unique fear mechanism follows directly from the Heterogeneity Hypothesis, which both Russell and I subscribe to. One of the key respects in which members of folk emotion categories such as fear are heterogeneous is that there are no causal mechanisms common to them all. However, the absence of a unique emotion mechanism is perfectly compatible with the presence of multiple emotion mechanisms corresponding to different instances of the folk concept. For instance, the fact that there is no discrete emotion mechanism that brings about fear components in all cases of “fear” is compatible with there being an emotion mechanism – call it the fear* mechanism for now – that made Sally’s heart pound, her palms sweat, her face broadcast danger and her legs freeze and then run. Russell’s conviction to the contrary is grounded in an argument from explana­ tory superfluity. His view is that we can provide a satisfactory causal explanation of the co-occurrence of the components without positing any discrete emotion mechanism. My view is that the alternative explanations considered by Russell are not satisfactory. I will conclude that the best explanation for why the components of Sally’s fear co-occurred is still the presence of a discrete – but not unique – causal emotion mechanism. Russell’s alternative “explanation begins with each of the components (her heart pounding, palms sweating, etc). Each component has its own causal chain, which research has gone a fair way in describing. Thus, we need an explanation of how the encounter with the bear resulted in changes in the autonomic nervous system, but that explanation would not include the assumption that fear did it. And so on, through each component that actually occurred during the emotional episode” (82). This formulation appears to beg the question. It is assumed that each component has “its own causal chain”, when the issue is whether or not there exists a “common causal chain” running from an internal emotion mechanism to each component. On the other hand, it is true that there is a cause – potentially the same cause – for each component. Furthermore, since Sally’s fear consists of the joint occurrence of several components, what really needs to be explained is why the components co-occurred, not simply why each of them occurred independently of one another. . Russell rightly emphasizes that in many emotion episodes the correlations among components are much weaker than in prototypical cases (e.g. Sally’s case). A great many non-prototypical­ instances of folk emotion concepts occur by virtue of a few co-occurring components.



6.  Discrete emotions: From folk psychology to causal mechanisms 149

Russell is well aware of this explanatory need, and proposes three nonmutually­ exclusive non-emotional explanatory mechanisms: “(a) features in the environment have a correlational structure, which then creates correlations among components, (b) one component can influence another, and (c) two components are correlated when they are both influenced by the same central mechanism such as attention” (83). Let us consider these candidate mechanisms in turn. According to the first, if two environmental features F1 and F2 are naturally correlated, and they independently cause, respectively, components C1 and C2, then C1 and C2 will correlate despite the absence of a discrete emotion mechanism causing both. Russell gives the following example: “suppose that novel events [F1] are more likely than familiar events to block a goal…Suppose further that goal blockage [F2] elicits an ANS pattern of cardiac acceleration [C2] and that novel events elicit frowns [C1]. If so, cardiac acceleration [C2] and frowning [C1] will be correlated even if no internal process links the two” (Russell­, 2009: 1273). According to the second mechanism, if a component C1 causes a component C2, then C1 and C2 will correlate despite the absence of a discrete emotion mechanism causing both. Russell gives the following example: suppose that making a threat expression [C1] “alters breathing and muscle tension, which in turn alters ANS activity, perhaps cardiac acceleration [C2]. The consequence would be that the threat face [C1] is correlated with cardiac acceleration [C2]” (Russell, 2009: 1273). According to the third proposal, if some mechanism M other than emotion causes component C1 and component C2, then C1 and C2 will correlate despite the absence of a discrete emotion mechanism causing both. Russell gives the following example: “suppose that focused attention [M] produces both muscle tension in the face [C1] and cardiac acceleration [C2]. If so, muscle tension in the face [C1] will be correlated with cardiac acceleration [C2]” (Russell, 2009: 1274). Let us now apply these three mechanisms to Sally’s case. We could argue that the bear was a novel event [F1] which caused Sally’s fear expression [C1], and that this novel event correlated with goal blockage [F2] which in turn caused Sally’s heart to pound [C2]. According to Russell, this would be a candidate explanation for why fear expressions and heart pounding correlate. Furthermore, we could rely on the second mechanism, and suggest that making a fear expression [C1] alters breathing and muscle tension, finally leading to cardiac acceleration [C2]. This would qualify as a second candidate explanation for why fear expressions and heart pounding correlate. Finally, we could appeal to the third proposal, and argue that the attention mechanism M produced both the fear expression [C1] and cardiac acceleration [C2], offering a third candidate explanation for why fear expressions and heart pounding correlate.

150 Andrea Scarantino

The problem with these alleged explanations is that they fail to provide any understanding of why the correlations among components take the form they do. We are told which alternative mechanisms could in principle bring about the components of Sally’s fear, but we are not told what underlies the causal relationships between mechanisms and components. What is left in the dark is precisely what needs to be explained. Why would the novelty constituted by a bear cause both heart rate acceleration – through goal blockage – and the facial expression of fear, rather than something else? Why would fear expressions cause heart rate acceleration, rather than something else? Why would attention shifts cause both heart rate acceleration and the facial expression of fear, rather than something else? On the contrary, positing an internal emotion mechanism – what I have called so far the fear* mechanism – provides a clear functional explanation of why the correlations among components take the form they do. On this proposal, the reason why the novelty constituted by a bear leads to the co-instantiation of heart rate acceleration and a face expression and sweaty palms and freezing and fleeing behaviors – possibly by means of causal relations among the components themselves and attention shifts – is one and the same: all such changes are functional to the avoidance of danger and brought about in a coordinated fashion by an internal mechanism for that very reason.  On Russell’s model, the co-occurrence of the components of Sally’s fear is co­ incidental rather than functional, because his three proposed mechanisms operate independently of one another and are not goal-oriented. Russell appears to explicitly endorse a coincidental explanation when he compares the co-occurrence­ of components of a prototypical emotion episode to the occurrence of a royal flush in poker – a combination of ace–king–queen–jack–ten of the same suit. As Russell remarks, “beyond the rules of the game and statistics, there is no need for an extra mechanism (except in cheating) to explain the very occasional occurrence of a royal flush” (84). But it is far from coincidental that being exposed to a bear is followed by the components of prototypical fear. Whenever we expose Sally, and a great many other animals for that matter, to bears, a very similar set of components will co-occur, over and over again. This suggests the presence of an internal causal . This is not to say that there is always a danger at hand when fear is elicited, nor that fear is always conducive to the avoidance of danger. The claim is simply that danger-avoidance is what the fear system is supposed to bring about. . Russell presents the even more radical analogy between emotions and the élan vital. In this case, Russell seems to lean toward eliminativism, because there is no such thing as élan vital. Discrete emotions do exist, on the other hand, and they are causally important for the reasons I describe in the text.



6.  Discrete emotions: From folk psychology to causal mechanisms 151

mechanism that couples dangerous stimuli with componential changes – the fear* mechanism. This is to say that the “very occasional occurrence” of prototypical fear – the royal flush mentioned in Russell’s example – is not due to the laws of statistics, but rather to the fact that the fear* mechanism rarely encounters stimuli dangerous enough to demand a prototypical fear response. So I conclude that Russell is right in rejecting the unique mechanism assumption, but wrong in rejecting the emotion mechanism assumption. Positing an internal emotion mechanism, I conclude, continues to be the best explanation for the systematic relation between classes of stimuli (e.g. dangerous stimuli) and classes of componential changes (e.g. heart pounding, palms sweating, fear expressions, freezing and fleeing). This being said, I follow Russell in thinking that the standard account of the relation between discrete emotions and emotional components needs to be rejected because it posits the existence of a unique causal mechanism. Furthermore, I have only characterized the fear* mechanism functionally, without saying much of anything about how it is supposed to work. The next section aims to fill these gaps, and propose a reconciliation between Panskepp’s and Russell’s positions on emotional causation.

7. A possible reconciliation? My central proposal is that the fear* mechanism at work in Sally’s case is nothing other than Panksepp’s primary fear system, in combination with learning and cognition. This hypothesis adds to the functional level of explanation an implementational level (Marr, 1982). Rather than simply considering the goal of the mechanism – e.g. avoiding danger – we are now considering a specific proposal on how such goal is physically realized. On this proposal, Sally’s response to the bear is driven by a genetically based subcortical brain network – the fear network – that coordinates sets of behavioral and autonomic outputs in ways that have proven adaptive in the evolutionary history of the species because they have helped animals “reduce pain and the possibility of destruction” (Panksepp, 36). Now, Russell seems to accept that the fear network is involved in Sally’s fear. According to Russell, “Panksepp’s fear circuit…results in momentary freezing, inhibition of ongoing behavior, and changes in Sally’s autonomic nervous system and endocrine system” (112). This admission is quite surprising, because these changes, unlike those accounted for by the three mechanisms I discussed in Section 6, do require positing an extra emotional mechanism – the fear mechanism.

152 Andrea Scarantino

But Russell adds that there are other aspects of Sally’s fear episode that the fear mechanism cannot account for. This suggests that, rather than being ex­ planatorily superfluous as posited in Section 6, the fear system may be explana­ torily insufficient. For instance, Sally must perceive the bear, she must recognize it as a bear, she must produce changes in Core Affect in the direction of high arousal and high negativity, she must enhance perceptual-cognitive processing of the bear, she must search her memory for previous bear attacks, she must search the environment for escape routes, she must formulate a plan about what to do and so on. I think that a reconciliation between Russell and Panksepp on whether discrete emotions are genuine causal mechanisms can be reached at this juncture. The reconciliation hinges on whether Russell can accept that the FEAR system has a global causal reach on the resources of the organism, and on whether Panksepp can accept that the fear system, understood along the lines I suggested in Section 5.3, has a great deal of variability in its operation. Russell argues that Sally’s fear episode includes changes that need to be explained in terms of perception, cognition and memory rather than in terms of the fear mechanism. But the fear mechanism as characterized by Panskepp has a superordinate control function with respect to all organismic sub-systems available to the organism, including perception, memory and cognition. The job of the fear system is to coordinate an open range of resources by gating and modulating incoming sensory inputs relevant to behavioral and autonomic outputs and by engaging in reciprocal interaction with the brain networks involved in learning and higher thought, thereby bringing about secondary and tertiary affects. So while it is true that Sally’s fear episode involves a great many changes that go beyond freezing and inhibition of ongoing behavior, I think such changes can be included within the explanatory purview of the fear system in interaction with other organismic resources. On the other hand, Panksepp should accept that the fear network responsible for coordinating the components of Sally’s fear is not an “important part[] of an intrinsically ambiguous conceptual whole[]” (3) named by the English term “fear”. As emphasized by Russell, there are many cases of fear that have nothing to do with the fear network. I have consequently proposed that we interpret fear as designating a theoretically motivated coordination system – rather than a part of the vernacular notion of fear as suggested by Panksepp. Furthermore, Panskepp’s fear system should explicitly be understood as a homeostatic property cluster kind (see Section 5.3) with a great deal of variability in the combination of components that can instantiate it. I take the presence of variability to be the primary motivator behind Russell’s rejection of the view that



6.  Discrete emotions: From folk psychology to causal mechanisms 153

all emotions, or all fears, or all angers, are produced by (or are) a single type of entity. We should co-opt this insight into theories of discrete emotions, and make room for mechanism-dependent variability. Mechanism-dependent variability consists of the fact that the same causal mechanism can bring about different components in different occasions. This is to be contrasted with across-mechanisms variability, which consists of the fact that there are multiple causal mechanisms, each with its own range of mechanism-dependent variability, underlying episodes of the same folk emotion. Since Panksepp’s fear system has the avoidance of danger as its goal and dangers differ across occasions, we should expect that the fear system will bring about different combinations of components in different circumstances. For instance, if the bear is faraway, the fear mechanism should bring about freezing and some combination C of other components. But if the bear is suddenly looming, the fear mechanism should bring about fleeing and a different combination C’ of components. On this view, the variability emphasized by Russell is not only compatible with the presence of a discrete causal mechanism underlying the co-occurrence of emotion components, but to be expected insofar as such variability is functional to what the mechanism is trying to do – avoiding the specific danger at hand in the specific circumstances at hand. This being said, I agree with Russell that the fear network as such cannot be credited with producing Sally’s meta-experience of fear. This meta-experience requires focused attention and concept use, and it is more likely to occur when she is safely back home than when she is facing the bear. Yet, the absence of the meta-experience of fear is not tantamount to the absence of any subjective experience associated with a fear episode. The subjective experience associated with a fear episode consists of the way it is like to undergo the activation of the fear network in the specific circumstances in which it is activated. Since such activation can involve different combinations of components because of mechanism-dependent variability, I reject Panksepp’s view that the activation of any primary system is associated with a distinctive and intrinsic subjective experience. On the view I am pushing, the activation of the fear system will generate different, context-dependent subjective experiences, depending of which specific behavioral and autonomic outputs are being brought about.

154 Andrea Scarantino

8. Conclusion I have argued that discrete emotions are genuine causal mechanisms with respect to the manifestations commonly associated with them. At the same time, I have agreed with Russell that discrete causal mechanisms cannot be found at the level of folk emotion concepts such as emotion, fear, or anger. The key corollary is that the science of discrete emotions should stop characterizing its objects of investigation in folk psychological terms. Panksepp’s strategy of using capitalized versions of folk psychological terms strikes me as a good way to go, provided either the subset condition or the significant overlap condition are satisfied. I have suggested that a possible way to reconcile Russell’s rejection of standard accounts of discrete emotions with the existence of causal mechanisms is to modify Panksepp’s model to account for the kinds of heterogeneity that Russell has so aptly characterized. Whether this modification can be successfully achieved within the confines of Panksepp’s original theory remains to be seen.

References Boyd, Richard. (1999). Kinds, complexity and multiple realization. Philosophical Studies, 95, 67–98. Fehr, Beverley, & Russell, James A. (1984). Concept of emotion viewed from a prototype perspective. Journal of Experimental Psychology: General, 113, 464–86. Levy, Robert I. (1973). Tahitians. Chicago: University of Chicago Press. Marr, David. (1982). Vision: a computational investigation into the human representation and processing of visual information. New York: Freeman. Pfaff, Donald. (2006). Brain arousal and information theory: Neural and genetic mechanisms. Cambridge, MA: Harvard University Press. Russell, James A. (2003). Core affect and the psychological construction of emotion. Psychologi­ cal Review, 110, 145–172. Russell, James A. (2009). Emotion, core affect, and psychological construction. Cognition and Emotion, 23, 1259–1283. Scarantino, Andrea. (In press). How to define emotions scientifically. Emotion Review. Smith, Kyle S., Mahler, Stephen V., Peciña, Susana, & Berridge, Kent C. (2010). Hedonic hotspots: generating sensory pleasure in the brain. In Morten Kringelbach & Kent C. Berridge (Eds), Pleasures of the brain (pp. 27–49). Oxford: Oxford University Press.

chapter 7

Nothing in mammalian psychology makes sense except in light of primary-process affective capacities Paul Sheldon Davies Department of Philosophy, College of William and Mary, Williamsburg, VA 23187-8795, U.S.A.

In one sense, the views of Panksepp and Russell appear complementary or at least non-conflicting, because the phenomena that each theorist studies are plausibly distinct. Panksepp studies sub-cortical structures and processes implementing primary-process affective capacities shared by all mammals. Russell studies the apparent psychological structure of the full range of emotions in a single primate species. Given the distinctness of these phenomena, we may be tempted to conclude that the methods and claims of both theorists cannot conflict, on the assumption that conflict is possible only when both sides make different claims about the same things. Alas! There is profound disagreement between these two views with respect to the following questions: i. What are the most fruitful methods and orientation for studying human psychology? ii. Do we presently know enough to defend a comprehensive theory of emotions? iii. What sorts of considerations qualify as reasonable evidence for specific claims concerning the emotions? With respect to each of these questions, I will argue that (a) the basic orientation and methods employed by Panksepp are better supported than those employed by Russell, (b) Panksepp’s methods and some of his substantive discoveries force upon us a form of skepticism regarding various capacities of the mind, including much of our emotional lives, and (c) this skepticism casts doubt upon the evidence adduced by Russell for various elements of his theory.

156 Paul Sheldon Davies

The distance between these two views appears too great for meaningful reconciliation. On my interpretation of Panksepp’s view, current knowledge of our primary affective capacities underwrites a skepticism concerning our nonprimary­ emotions such as shame, pride, guilt, and so on. Our present prospects for successfully explaining the full range of emotion episodes appear slender. But on Russell’s view, explaining the full range of emotion episodes is a condition of adequacy on a theory of emotions. Indeed, the three sets of processes posited in Russell’s Psychological Construction are intended to capture every genuine token episode of emotion. Yet it is doubtful that these two approaches can both be correct, in which case the contrast between the two views is stark. Holding out for reconciliation is wishful thinking. As indicated, I argue that Panksepp’s approach drives us toward a form of skepticism that casts doubt on Russell’s theory. The skepticism I defend, however, is mine, not Panksepp’s. I argue that, if Panksepp’s approach to the study of emotions is right then my skepticism follows, but I do not claim that Panksepp endorses any general skeptical thesis (he does not) or that he agrees with me that his view leads us to skepticism (he may not). My skepticism, moreover, is transitory. As knowledge of the human self progresses, we will likely discover the affective bases of all our psychological capacities, forcing us to conceptualize our capacities anew, including those we presently classify as emotions.

1. Ancestral voices in the mammalian mind I begin with a naïve question. If we want to discover the truth about human emotions – if we want to know what our emotional lives are really like – what are the most fruitful methods available to us? One obvious answer concerns our evolutionary history. We know we are related by descent to all extant species of primates, mammals, reptiles, and so on. We also know that lower- and mid-brain structures in humans are homologous with those in other mammal species. There is diversity in mammalian cortical structures but comparatively little in sub-cortical ones, and the relative dearth of diversity among sub-cortical structures is informative. It suggests, for instance, that these structures, which must have been sufficient to sustain the life of . Immediately before introducing the three processes that comprise Psychological Construction, Russell asserts that “each token emotional episode (such as Sally’s encounter with the bear last Saturday) requires an explanation….” (Russell, this volume). . This difference between the two views is by no means exhaustive. It is but one symptom of deep divergence.



7.  Nothing in mammalian psychology makes sense 157

a­ ncestral mammals with little or no cortex, are probably sufficient to sustain the life of all or most extant mammal species. That is, lower- and mid-brain structures in human beings probably implement in us a kind of mind that can function apart from cortical structures. They implement the elemental non-cortical capacities upon which all mammalian cortical capacities evolved; they constitute, that is, the affective foundations of mammalian psychology. An additional answer to my question concerns a few basic observations from neuroscience and psychology, and integrates well with my first answer. Because cortical structures in mammals evolved upon sub-cortical ones, we should expect to find an asymmetric dependence of cortical function on the sub-cortical. And that is what we find. We know that cortical capacities are causally efficacious only when elemental affective capacities are operating, but not vice versa. Rats who have their cortex surgically removed early in life nevertheless exhibit affective coherence later in life. Similar results are observed in humans who fail to develop a normal cortex; despite horrendous deficits in learning and cognition, they exhibit surprising coherence in their affective behavior. We also know that some of our most crucial capacities for social attachments and practical reasoning are lost when certain affective capacities are destroyed. There are, moreover, even more-compelling neuropsychological observations that confirm the hypothesis that the human mind today comprises, in part, an ancestral affective mind. These are the stimulation experiments performed by Panksepp and his colleagues. When we electrically or chemically stimulate neural homologues in animals from two or more species, we observe highly similar effects in systemic processes and in behavior. When we stimulate the area implementing what Panksepp calls the seeking system – roughly, the medial forebrain bundle in the lateral hypothalamus – we observe the same behavioral effects, namely, vigorous searching behavior. When we stimulate the area implementing what he calls the panic system in the infants of any mammal species – roughly, the dorsomedial thalamus, the ventral spetal area, the preoptic area, and sites in the bed nucleus of the stria terminalis – we observe separation distress cries, the

. See Panksepp, Normansell, Cox, and Siviy (1994). . See Shemon, Holmes, and Byrne (1999). . See Damasio (1994). . “The [lateral hypothalamic] continuum, running from the ventral tegmental area (VTA) to the nucleus accumbens, is the area of the brain where local application of electrical stimulation will promptly evoke the most energized exploratory and search behaviors an animal is capable of exhibiting.” (Panksepp, 1998: 145).

158 Paul Sheldon Davies

same cries infants make when separated from their mothers. Similar findings hold for the five other primary process systems studied by Panksepp. The importance of these experiments is difficult to overstate. Three points are relevant. First, these experiments support the claim that the affective effects of lower- and mid-brain homologies have been preserved across mammal species. The evolution of a hefty cortex has no doubt elaborated our ancient capacities, but these ancient affects persist in our psychology and continue to exert their effects. It is in this sense that the foundations of human psychology are properly conceptualized as comprising, at least in part, mammalian affective capacities. Indeed, the point may be cast as a corollary to Theodosius Dobzhansky’s well-known claim that nothing in biology makes sense except in light of evolution. The corollary is that nothing in mammalian psychology makes sense except in light of shared affective capacities. All of human psychology, including our emotional repertoire, must be viewed through the lens of our foundational affective systems. Second, Dobzhansky never claimed that an evolutionary framework was sufficient for understanding all biological phenomena. The knowledge that all life on earth has evolved generates a powerful set of assumptions that any theory in biology must honor but tells us little about, say, the molecular structure of amino acids. Likewise for my corollary. Knowledge of our evolved affective capacities provides a powerful set of assumptions that any theory in psychology must honor but tells us precious little about, for instance, the psychological capacities with which we reason in logic or mathematics. And this, as we will see, is crucial to understanding Panksepp’s view, for it is a consequence of his theory concerning our primary-process capacities that we cannot justifiably claim to know, at least at present, much about our non-primary capacities. Third, the only plausible way to refute Panksepp’s theory is to show that the experiments mentioned above do not confirm the existence of his primary-process systems. That is, if you wish to challenge the claim that there exist several distinct neural systems implementing different primary-process affective capacities, you must face the evidential music provided by Panksepp’s experiments. Or, more modestly, if there is an alternative route by which to refute Panksepp’s thesis, it must include an explanation of why there is no need to refute his stimulation and . “One of the best ways to identify the general locations of panic circuitry is by administration of localized electrical stimulation of the brain…into specific areas. This type of work has now been conducted in a large number of species, including primates, cats, and chickens, and has yielded a remarkably similar picture.” (Panksepp, 1998: 267). And a few lines further on, Panksepp refers to the anatomical structures listed above. . For a general statement of this claim, see Panksepp (1998: 52). . Dobzhansky (1973).



7.  Nothing in mammalian psychology makes sense 159

decortication experiments. In the absence of such refutation or explanation, our best current evidence is that the human mind comprises several ancestral affective systems that are foundational to the whole of human psychology. 2. How we know that we do not know ourselves We know, then, that 1. the evolved foundations of human psychology comprise several primary- process affective capacities. We also know that 2. these primary systems were the ancestral raw materials upon which non-primary­ capacities evolved and 3. the causal efficacy of our current non-primary capacities depends on the ­efficacy of our primary capacities. Knowledge of (1) and (3), in particular, is sufficient to show how little we can justifiably claim to know about our emotional lives. Our emotional lives are subject to an informed skepticism that casts doubt on current attempts to theorize about human emotions in general.10 To illustrate the skeptical thesis, consider another naïve question. How do we come to know which phenomena an adequate theory of emotions must illuminate? What are the sources of knowledge concerning the very phenomena, or what we take to be phenomena, of our emotional lives? We all accept a variety of generic claims concerning our emotions. We all believe, for instance, that most of our experiences include an emotional dimension and that these emotional dimensions affect our states of mind and our actions. We also accept a range of specific claims about our emotions. We all believe, for instance, that we experience shame only when we perceive that we have failed to satisfy some norm of our community. So the naïve question is, What are the sources of these (and other) beliefs concerning the nature of our emotional lives? Whence our presumed knowledge of the phenomena for which we seek a theory of emotions? This naïve question might be naïvely answered as follows. We know about our emotional lives on the basis of our felt experiences, on the basis of observed correlations between our felt experiences and events in the world, and on the basis of 10. Elsewhere I press the same basic skepticism with respect to our alleged capacities for practical reasoning, free will, and legal responsibility (Davies 2009, 2011, forthcoming). In the present discussion, I press the skeptical worry with respect to our presumed knowledge of our emotional experiences..

160 Paul Sheldon Davies

observing other agents. We observe, for instance, that we feel certain ways under certain sorts of conditions; we also observe that we behave in certain ways in certain contexts while experiencing certain feelings; and we observe correlations between patterns of behaviors in other agents under certain conditions, including behaviors that appear to express how those agents are feeling. The naïve answer, then, is that our knowledge of the phenomena of our emotional lives arises from our felt experiences and from reflections upon those experiences (reflections that include observed correlations with behavior and context).11 It is against the backdrop of this naïve answer that my skeptical thesis is best presented. The thesis is that (S) For any human experience comprising an emotional dimension, we cannot justifiably claim to know, from our felt experiences or from reflections on those experiences, the correct nature of the emotion involved.

The “correct nature” of an emotion should be understood as including its triggering causes, neural bases, downstream causal effects in the economy of the organism (especially its effects in various forms of homeostasis), and short- and long-term anticipatory functions. The skeptical thesis, then, is that the presumed phenomena of our emotional lives are an unreliable guide to the correct nature of our emotional lives. And as we will now see, the argument for this thesis rests on the fact that there are demonstrable mismatches between our presumed knowledge of our emotions and the actual facts. The argument consists of two basic steps. First, recall thesis (1) from above, namely, that the affective foundations of the human mind comprise several primary-process affective systems. This is a crucial fact, for it dictates the basic methods with which to frame our initial study of any psychological capacity. We begin by looking for lower- and mid-brain structures that plausibly implement the psychological capacity of interest. Because our initial understanding of our capacity may be off the mark, the process of trying to identify salient brain structures is highly fallible. But a fallible process can also be self-correcting one. If, for instance, none of the candidate neural structures, when stimulated, produce the predicted psychological effects, that is prima facie evidence that we have incorrectly conceptualized the target psychological capacities.12 Having identified plausible neural structures, we then look for homologous structures in other 11. I am not interested in defending this admittedly vague answer to my naïve question. If you think a different answer is closer to the truth, feel free to substitute it into the skeptical thesis (S) in the next paragraph. 12. It is also possible that our initial list of neural candidates failed to include the actual structure implementing the capacity of interest.



7.  Nothing in mammalian psychology makes sense 161

mammals and ­ conduct comparative neurological studies (including Panksepp’s stimulation studies). The same sort of fallibility exists here, but the prospects for experimental self-correction are similarly wide open. Then, having completed our comparative studies, our default expectation is that the results of those studies provide a tentative roadmap for the study of the relevant capacity in ourselves. Not a roadmap for the whole of our emotional lives; just for the affective foundations of our emotional lives that we share with other mammals. This basic methodological framework, however, leads us to the skeptical thesis in (S). It does this by leading us to the discovery of substantial mismatches between our presumed knowledge of our capacities and the actual facts of our capacities. To see this, consider first the seeking system. As Panksepp conceptualizes it, seeking constitutes us as animated, energized, expectant explorers of the world. It is what moves us to be active, practical agents in the world, to notice and investigate our environment. (Think of the vigorous sniffing of a curious rat or the relentless touching, grabbing, and sucking of a curious 18 month-old toddler.) Yet the very capacity that moves us to notice and investigate the world is precisely the kind of thing that we, as active agents, are likely not to notice. This of course makes good evolutionary sense. The structures that implement our experiences as active agents are functionally dedicated to anticipating life challenging events, and it is a safe bet that ancestors who spent much time trying to investigate their capacities for anticipating life’s challenges were not terribly successful at anticipating life’s challenges. (An analogy: We typically do not see the mechanisms that enable us to see, because seeing those mechanisms would have done little for our ancestors’ selective success.) The seeking system, that is, is something our experiences as active agents fail to reveal to us. In particular, nothing in our felt experiences reveals to us the existence of the affective system implementing those felt experiences. We thus are faced with a rather large mismatch between our presumed knowledge of ourselves and the actual facts. A similar point holds for the panic system. It is surprising to learn that the neural system which causes us to panic in life-threatening situations is also one of the main causes of our being the kind of social animals we are. Yet that is ­Panksepp’s hypothesis. The hypothesis is that we do not have a distinct system dedicated to social attachments in the way, for instance, our visual system is dedicated to vision. Instead, the system that helps free us from life-threatening circumstances also minimizes the terror of separation by constantly moving us to form and feed our social attachments. Here the mismatch is not about the existence of the relevant system, as it is with seeking, but rather the functional repertoire of the system. If the panic system works the way Panksepp hypothesizes, we must conclude that there is little or nothing in our experiences as agents that reveals to us the full range of anticipatory functions served by that system.

162 Paul Sheldon Davies

The same general point can be made regarding the other primary-process systems proposed by Panksepp, including fear, rage, lust, care, and play: The anticipatory systemic functions served by these primary systems do not match our presumed knowledge of our own capacities, at least along some very substantial dimensions. We thus may generalize: our experiences as active agents, including our felt experiences and our reflections on those experiences, are not a reliable source of knowledge about the ancient affective systems at the foundations of our psychology. That is my first premise in support of the skeptical thesis (S). But that is just the beginning, for we have ample evidence for the claim in (3), namely, that the functioning of our non-primary systems depends upon the operations of our primary systems. If so, then we may further conclude that our felt experiences and reflections as active agents are a poor source of knowledge of the full range of human emotions, including those that arise from the messy interactions between primary and non-primary systems. To see that (3) is so, notice that our non-primary systems comprise capacities that appear relatively non-affective, such as our ability to solve math problems, but also emotions such as shame and pride that appear to arise from a mix of affective and non-affective mechanisms. The causal dependence of such non-primary capacities on primary affective systems is evident in several ways. (i) As mentioned, we know that our non-primary capacities evolved upon ancestral affective systems; in terms of neural structures, the evolutionary emergence of cortical structures was constrained by antecedent subcortical structures. This historical fact underwrites the rational expectation that the functioning of our cortical capacities is integrally dependent on that of our non-cortical ones. It is perhaps possible that evolving cortical structures have somehow displaced or overridden our ancestral sub-cortical structures, but Panksepp’s stimulation experiments cast doubt on that. (ii) As also mentioned, we know that rats decorticated early in life exhibit coherent affective behavior later in life but lack the normal capacities to learn. The same combination of affective coherence and cognitive deficit is observed in humans who fail to develop a cortex. These studies show that an ancient affective mind continues to function in the absence of some or all cortical structures. And since the converse is not true – since cortical capacities cannot function in the absence of sub-cortical ones – the causal dependence of the cortical is indeed compelling. (iii) As also mentioned, the causal dependence of our actions and practical choices upon our affective capacities is well-supported by the work of Damasio (1994) and his colleagues. These studies show that critical agential capacities are lost when specific sub-cortical sites that implement an affective capacity are damaged. Of course, in humans today, causal interactions between subcortical and cortical structures are far from simple and



7.  Nothing in mammalian psychology makes sense 163

by no means entirely asymmetric.13 That, however, is compatible with the claim in (3) that our non-primary capacities are causally dependent on the efficacy of our primary systems. Further evidence for (3) comes from considering the nature of specific nonprimary emotions. Shame, for instance, appears to be constituted quite directly from central features of some primary affective systems, since the experience of shame can be highly visceral and difficult to extinguish or modulate. We thus might hypothesize that shame occurs when the organism experiences the felt threat of separation distress implemented in the panic system and when that felt experience is triggered by a representation of oneself as failing to meet some social norm. Similar hypotheses regarding other non-primary emotions are similarly plausible, and all reveal the causal dependence of non-primary emotions on our primary-process systems.14 There is, finally, the looming possibility that the affective foundations of our psychology will illuminate and help confirm theories in social psychology that support their own skeptical arguments concerning our knowledge of ourselves. Daniel Wegner (2002), for instance, defends a form of skepticism concerning our knowledge of whether or when we have authored our own actions. Timothy ­Wilson (2002) defends a form of skepticism regarding our alleged knowledge of our reasons for acting. And so on.15 My suggestion is that the effects of Panksepp’s primary process systems may help explain why we are prone to the deficits in selfknowledge that Wegner, Wilson, and others have identified. The basic line of thought is speculative but intriguing. Both Wegner and ­Wilson posit psychological systems that render various aspects of our own actions causally intelligible to us. In addition, both point to significant forms of error that occur in the normal operations of those systems; both systems, under a range of conditions, lead us to think and feel that there exist causal relations when in fact

13. See Damasio et al. (2000). 14. In offering this hypothesis concerning shame, I am not contradicting my skepticism. The hypothesis is that the actual facts concerning shame include the functioning of a primaryprocess­ system that the apparent phenomena of shame do not even countenance. 15. See also the work of John Bargh and colleagues (e.g., Bargh, 2005) on our presumed knowledge of what we are doing and why we are doing it; see Emily Pronin and colleagues (e.g., Pronin, 2007) on our ignorance of the biases with which we judge ourselves or our actions as superior to others; see Joshua Greene and colleagues (e.g., Greene, 2011), as well as Jonathan Hadit and colleagues (e.g., Haidt, 2006), on the effects of emotions on the intuitions which form moral judgments; see Martin Conway and colleagues (e.g., Conway, 2003) on the actual workings of autobiographical memory, some of which are rather surprising; and so on.

164 Paul Sheldon Davies

there is none. My suggestion, then, is that Panksepp’s theory may illuminate and perhaps help confirm the existence of psychological systems dedicated to causal intelligibility. To illustrate, it is a crucial part of Panksepp’s hypothesis regarding the seeking system that it moves us whenever there is a perceived homeostatic imbalance in various organismic systems. More specifically, seeking moves us to render our environment and our actions causally intelligible to us, that we might better anticipate where to go and what to do to re-establish homeostasis. Yet, as Panksepp describes, the seeking system, while moving us to acquire anticipatory causal expectations, sometimes overshoots and generates false causal beliefs. It does this, moreover, in every species that has been studied so far!16 The looming possibility is that the effects of our primary affective capacities will help explain and confirm the skeptical implications of a host of theories in social psychology with their own skeptical implications.17 Here then is where we stand. The skeptical thesis in (S) is compelling, for it is demonstrably true that our felt experiences and reflections on our experiences ­often conflict with the actual capacities of our minds. This is plausible in light of the mismatches discovered by Panksepp and the fact that our non-primary capacities depend so directly on the operations of our primary capacities. We may describe our situation by extending my corollary to Dobzhansky: nothing in mammalian psychology makes sense except in light of our ancient affective ­primary-process capacities and, as we learn about these primary-process capacities, the limits of our knowledge regarding our wider emotional lives come to the fore. We thus must cast a skeptical eye on any theory that aspires to explain the nature of emotions quite generally, especially theories that leave out what is known about the affective foundations of our minds. And this has substantive implications for Russell’s approach to the emotions, as I now explain.

3. In the shadows of doubt Following Oatley and Johnson-Laird (1987), Russell assumes that emotional episodes that take an object – fear of a bear, longing for scotch – are more complex than episodes that appear objectless – chronic grumpiness, free-floating dread. Since Russell aspires to develop a theory of human emotions that begins with the simplest constituents possible, he posits as theoretical primitives the elements

16. Panksepp (1998: 161). 17. For a fuller rendition of this line of thought, see Davies (2011).



7.  Nothing in mammalian psychology makes sense 165

that plausibly constitute such objectless feeling states. With those primitives on the table, he then constructs an elaborate theory that aspires to explain the full range of emotions experienced by human beings. The resulting view has two core components: (i) the dimensional, which comprises two primitives, and (ii) the categorical. 18 After summarizing each component, I consider the implications of my skepticism for this view. At any given moment, intact persons occupy a point (or perhaps an area) along two continua, pain versus pleasure and high arousal versus low arousal. Russell dubs a person’s location along these dimensions her Core Affect, the first theoretical primitive of his theory. By itself, Core Affect has no object. It is also ubiquitous to all living humans; we are always located somewhere along these two continua, whether or not we know it. In normal mature humans, Core Affect is consciously accessible, even if it is not always accessed. As Russell puts it, Core Affect is a “neurophysiological state consciously accessible as the simplest raw (nonreflective) feeling evident in moods and emotions” (2003: 148). The second theoretical primitive is Affective Quality. This is functionally defined as any property of an Object that can trigger a change in Core Affect. The Affective Qualities of a wild bear are whatever properties it has that can elicit a change in a person’s Core Affect. The Affective Qualities of scotch are whatever properties it has that can change Core Affect. And so on. Virtually all the Objects with which we interact are endowed with one or more Affective Qualities. These two primitives combine in various ways, but most centrally they underwrite what Russell calls Attributed Affect, the simplest non-primitive in his theory. Attributed Affect is defined operationally as the attribution of an Affective Quality (e.g., being frightening) to an Object (e.g., the bear) that has changed, or that may change, the perceiver’s Core Affect. Attributed Affect refers, in short, to a state of the person in which her Core Affect is directed toward an Object. That is why it is Russell’s simplest non-primitive. As he points out, the combined effects of Core Affect and Affective Quality can account for some but not all the emotional episodes we experience. They can account, perhaps, for episodes constituted entirely from valence and arousal. Serenity appears to consist in high pleasure and low energy, depression appears to consist in high pain and low energy, and so on. Other emotions, however, such as shame and pride, are difficult to locate in just two dimensions. Something more appears to be involved in these latter cases. To capture this missing element, ­Russell posits the second core component of his theory, the categorical.

18. See Russell, this volume, and also Russelll (2003), for the full picture.

166 Paul Sheldon Davies

The categorical provides greater differentiation of emotions. Or so Russell claims. For any token Attributed Affect – for any token episode in which Core Affect and Affective Quality combine to produce an affective state that has an Object – that token qualifies as a specific emotional episode only if it resembles the content of a socially-based emotion script. Scripts are descriptions of prototypical emotions and are, according to Russell, embodied in the fabric of one’s culture and reflected in one’s natural language. Whether a token Attributed Affect satisfies a specific script is not something the person must recognize. Resemblance can obtain without recognition. Recognition is required, however, for a further component of Russell’s theory, what he calls Emotional Meta-Experiences (EMEs, for short). An EME occurs when the person perceives that she is in a specific emotional state; the Emotional Meta-Experience of, say, fear occurs when she perceives that her Attributed Affect resembles the socially-based script for fear. Emotional Meta-Experiences thus constitute a particular form of self-perception. This component of Russell’s theory is intended to capture the fact that some emotion episodes clearly include the subjective experience of observing one’s current emotional state. Some episodes, but not all. Indeed, on Russell’s view, you can be in the midst of an episode of fear or you can be recouping from an earlier episode of fear and never perceive your own episode of fear. The subjective experience of being afraid is not essential to being afraid: “Thus, the [subjective] experience of being afraid [the perception of one’s own fear, via an EME] can occur late in or even after the emotional episode, or never.” (Russell, this volume) Thus, although recognition is required for EMEs, not all emotion episodes include an EME. One important implication of this view is that, because emotional types depend for their existence on prototypical scripts produced by social forces, they cannot be identified with biological properties. Russell is explicit: The set of events picked out by the English word fear is not a biologically given category. Indeed, the present analysis predicts that there is no neural circuit, peptide, or other biological marker that is unique to fear (or any other discrete emotion). Instead, to categorize is to note a resemblance between observed components and a mental representation, which is here thought of as a mental script.  (2003: 151; italics in original)

Since human cultures vary in the linguistic resources with which emotions are individuated, the range of actual emotional categories may vary across cultures. These linguistic differences, moreover, embody contingencies that may occur during the distinctive histories of different cultures. This leads to a thesis with profound implications concerning the nature of human emotions. If categorical



7.  Nothing in mammalian psychology makes sense 167

judgments based on prototypical scripts are necessary for the existence of emotion episodes and types, and if the origins of prototypical scripts are traceable to cultural contingencies, then the existence and nature of any human emotion is relative to and contingent on the evolving features of a specific culture. Valence and arousal are necessary but not sufficient for emotional episodes; what is additionally required is a resemblance to the contingent, evolving prototypes that exist in the larger culture. I trust it is now clear that Panksepp and Russell are indeed theorizing about quite different phenomena. This, as I say, may tempt us to venture that the two views can be integrated or at least reconciled. However, as I now argue, the disagreements run too deep for that. We cannot accept certain elements in Russell’s approach if we accept the skepticism that emerges from Panksepp’s view. I defend this claim with reference to both core components of Russell’s view. The Dimensional: Consider again Core Affect and compare it to the several primary-process affect systems posited by Panksepp. Since Core Affect is singular and primary-process systems are several, the two views may appear to conflict. Upon inspection, however, the appearance of a conflict might be illusory. After all, Core Affect constitutes a circumplex in which a large range of dimensional states are possible, and this range of possible states might correspond to the variety of states produced by multiple primary-process systems and interactions among those systems. That is, Russell’s dimensions of arousal and valence may be among the regular effects of Panksepp’s several affective systems. Such correspondence is, at any rate, a possibility to be investigated empirically; the appearance of conflict should not be taken at face value. When, however, we turn to the question of evidence, the status of Russell’s Core Affect is called into question. If I am right that Panksepp’s theory of the foundations of mammal emotions gives rise to an informed skepticism regarding our non-foundational emotions, then we should expect that the evidence adduced by anyone aspiring to formulate a general theory of emotions is likely to fall prey to my skepticism. And that is indeed the case for Russell’s theory, as I will now argue. Russell offers at least two categories of evidence for Core Affect, the first from psychometric studies. These studies employ a variety of devices for discerning statistical patterns between subjects’ emotion judgments (based on emotion-related materials such as facial expressions, emotion words, etc.) and proposed theoretical indicators concerning emotions (e.g., the presence of valence and arousal). And, as Russell reports, he and other psychologists find that emotion-related judgments are consistently correlated with the two dimensions that constitute

168 Paul Sheldon Davies

Core Affect.19 These findings, moreover, appear to hold up in cross-cultural studies. This is offered as one category of substantial evidence for Core Affect.20 But we must wonder about the evidential credentials of such studies. If the skeptical thesis in (S) is correct, the skepticism extends to judgments made by subjects in psychometric studies. We may suppose for the sake of argument that, for instance, similarity judgments among English-speaking subjects concerning facial expressions are perfectly correlated with independent estimates of valence and arousal. Of what would that be evidence? We can all agree it would be striking prima facie evidence that, when making similarity judgments about the emotional expressiveness of human faces, subjects are attuned to indicators of pleasure versus pain and high versus low arousal. But if my skepticism is correct, such happy correlations may be irrelevant to the question of what our emotions are actually like. The reason is simple: If our presumed knowledge of the phenomena of our emotional lives is an unreliable source of knowledge concerning the actual facts of our emotions, then that unreliability surely infects the judgments of subjects who, no doubt, depend on their presumed knowledge of our emotional lives in the judgments they produce.21 If so, then even the most vivid statistical patterns among such judgments should not be taken as evidence of the actual facts of our emotions. Such patterns may reveal the relations that we accept as real given our presumed knowledge of our emotional lives, but it is precisely that presumed knowledge that fails to match, in a range of cases, the actual facts of our emotional lives. Russell’s second category of evidence for Core Affect is neurological. The considerations from neuroscience raised by Russell, however, are even more problematic than evidence from psychometrics.22 The problem is that while 19. E.g., Russell (2003: 153). 20. Russell also offers ethnographic evidence (e.g., ibid), but the skepticism that applies to psychometric evidence applies in the same way to ethnographic evidence. 21. That subjects do appeal to such presumed knowledge is something Russell seems not to question: “So far, we have considered core affect on the level of subjective experience. We are, of course, not alone in taking this approach: Whatever the labels attached, self-report measures tell us how people experience their condition.” (Russell and Feldman Barrett, 1999). 22. It appears worthwhile to point out that Russell’s characterization of Panksepp’s view (in Posner­ at al., 2005) overlooks the skeptical implications of Panksepp’s discoveries. Russell claims that “the reigning experimental paradigm in affective neuroscience research posits that emotions can be divided into discrete and independent categories and that specific neural structures and pathways subserve each of these emotional categories.” (Posner et al., 2005: 715). And it is explicitly claimed that Panksepp is among the advocates of this reigning paradigm. But that is false. To be skeptical about our emotional lives is not to posit a theory of emotions; it is to



7.  Nothing in mammalian psychology makes sense 169

Panksepp offers potent neural evidence that discriminates in favor of his theory, nothing in the neural considerations offered by Russell discriminates in favor of his theory. Two points make this clear. First, among all the evidential considerations offered by Russell, those that are consistent with Core Affect are also consistent with several primary-process affective systems. That is, none of the neural studies cited by Russell discriminate in favor of Core Affect as opposed to Panksepp’s several primary systems.23 The reason is simple. As mentioned above, the diversity of possible states within Core Affect is plausibly matched by the diversity of possible states that several, distinct primary-process systems can produce. The thesis that all or most token emotional episodes are characterized by arousal and valence is entirely consistent with the claim that all those token episodes, including their arousal and valence, are the effects of several, distinct primary-process affective capacities.24 Second, and decisively, the same criticism cannot be leveled against ­Panksepp’s view. We may grant that, at the level of arousal and valence, nothing discriminates between a single Core Affect and several primary-process systems, but when we shift our analysis to the level of actual neural structures, the two theories are no longer on a par. The evidence from studies in which specific neural structures are stimulated in organisms from two or more mammal species clearly supports the existence of several primary-process systems. The anatomical structures and neurochemical processes that constitute the seeking system are demonstrably different from those that constitute panic, rage, fear, and so on. This is evidence that Russell does not address, yet it is well-confirmed empirical evidence for precisely the kind of neural differentiation among affective systems on which Panksepp insists. This difference in scientific evidence is decisive, in this sense: The full range of neurological evidence available to us at

posit that, at this point in inquiry, we know too little to advocate a general theory of emotions. Yes, Panksepp posits the existence of several primary-process affective systems but, as I have argued, the discovery of those foundational systems forces us toward a skepticism concerning our emotional lives taken as a whole. Russell and colleagues mistake a theory of the foundations of our emotional lives for a theory of the whole. 23. See Russell (2003), Posner et al. (2005), and Russell, this volume. 24. This point cannot be dodged by insisting that Panksepp’s several primary systems cannot, by themselves, account for the full range of emotion episodes. Two points are relevant. First, the issue under discussion here is the evidence from neuroscience, not the apparent differences between, say, serenity and shame. Second, the skepticism defended above undermines our presumed knowledge of non-primary emotions such as shame. (See n. 26 below.)

170 Paul Sheldon Davies

this time makes it rational to accept Panksepp’s primary-process systems and reject Russell’s Core Affect.25, 26 The Categorical: Recall that, on Russell’s view, a token Attributed Affect, to qualify as an emotional episode, must resemble an emotion script specified in the natural language of the relevant culture. This view, if defensible, would show that what we regard as emotion episodes are contingent on evolving features of our cultural histories. It would also show that our emotional lives are, to a significant degree, separate from our biological constitution. That is, on Russell’s view, biology thinly constrains our emotions, since only organisms capable of pleasure and pain and capable of different levels of arousal can have emotions. (And if it is possible to implement valence and arousal in non-biological systems, then Russell’s theory of emotions is completely autonomous from biology.) But beyond this thin constraint, nothing in our biology suffices for even a single episode of emotions such as fear or shame or pride. But this categorical component cannot be sustained in light of my skepticism. For if the skeptical thesis in (S) is compelling, then the conclusion we must draw concerning the categorical component is indeed a skeptical one, namely, we are in no position to theorize in an informed manner about our emotional lives in general, including the potential roles of socially-derived categories in the formation of emotions.27 The argument for this skeptical implication has the same structure as my argument for thesis (S). The most compelling available empirical evidence supports the claim that the affective foundations of the human mind comprise several, distinct primary-process systems; yet knowledge of those foundations conflicts with our presumed knowledge of the very phenomena concerning our emotional lives; 25. This requires qualification. As I have just suggested, Russell’s dimensions may be byproducts of Panksepp’s primary-process systems. If so, then we have an explanation of the prevalence of valence and arousal in terms of our affective foundations. This rules out not the importance of valence and arousal, but rather the hypothesis that Core Affect stands as a competitor to – as a theoretical entity distinct from – Panksepp’s primary-process systems. 26. The above comparison of evidence from neuroscience is prohibited by the argumentative strategy employed in Posner et al. (2005). Posner et al. assert (mistakenly) at the outset that Panksepp aspires to a general theory of emotions and then argue that, given ­ Panksepp’s focus on animal studies, his view cannot possibly support a general theory of emotions. Having thus “refuted” Panksepp’s view, the authors proceed to ignore the crucial comparative question, namely, which view best fits all available neural evidence? 27. I take no stand on whether emotion categories depend in some way on socially-based categories. My claim is that we are presently too ignorant to theorize in an informed manner about this question.



7.  Nothing in mammalian psychology makes sense 171

if we cannot reliably specify the phenomena that an adequate theory of emotions must preserve, including the range of emotion episodes that Russell’s categorical component is intended to explain, then we are in no position to formulate rational judgments concerning the adequacy of such a theory. I stress that, if my skepticism is plausible, we are ill-suited to specify even the phenomena that are at issue when we theorize about the non-foundational portions of our emotional lives.28 I am not denying that we take ourselves to know a good deal about our emotional lives – I take it as obvious that most people believe with confidence that they know a good bit about their emotions – but I am claiming that such confidence is subject to a potent epistemic defeater. The defeater derives from the fact that our presumed knowledge about our emotions is often at odds with demonstrable features of our primary-process systems and the further fact that our non-primary emotions are causally dependent on our primary ones. Taken together, these facts show that, for any emotional experience we have, we cannot reliably claim to know, on the basis of our felt experiences, what those emotions are actually like. If so, then it should be obvious that, if we nonetheless insist on theorizing about our emotional lives, especially about the alleged differences among nonprimary emotions, then we are engaging less in scientific inquiry and more in philosophical speculation. Of course, philosophical speculation about the emotions may be fruitful, depending on how far we stray from what is known about our affective foundations. Yet it is hard to imagine a more-distant speculation than the claim that all emotion prototypes are social scripts or that a token Attributed Affect qualifies as a genuine emotion episode only if it resembles some such script. I do not claim that either of these claims is false. My claim is that we are presently in no position to theorize about these claims in an evidentially constrained fashion. That is the upshot of my skepticism, a skepticism based directly on empirical evidence that we do in fact possess.

4. Conclusion It is tempting to imagine Panksepp working from the bottom upward and Russell working from the top downward, someday to meet in the middle and announce a completed theory of emotions. Alas! The two views differ over fundamental 28. In this regard, consider Panksepp’s suggestive speculations (this volume, Chapter 4) concerning laughter, sighs, and empathy in other animals. If these speculations are on track, the identified capacities are likely homologous to capacities in us, and knowing this may force us to revise our understanding of our own capacities.

172 Paul Sheldon Davies

questions concerning the study of the human self. They disagree over the general orientation and methods for studying human psychology; they disagree about whether we presently know enough to defend a comprehensive theory of emotions; they disagree over the forms of evidence they offer and even the sorts of considerations that count as evidence. We fool ourselves to think that these two views can be reconciled. Theorists of emotions must choose between these views or reject both in favor of a third option. But theorists of emotions must also consider the most fruitful modes of inquiry for overcoming my skepticism. On my view, the proper response to our current skepticism is to appeal to methods and findings that have been most fruitful to date. The most robust findings to date come from Panksepp’s stimulation and decortications experiments that support the existence of several, distinct primary-process affective systems. Our best-informed conclusion, at least for the moment, is that nothing in human psychology, our emotional lives included, makes sense except in light of affective neuroscience and, more generally, in light of the full range of facts concerning our shared animal ancestry.

References Bargh, J. (2005). Bypassing the will: Toward demystifying the nonconscious control of social behavior. In R. Hassin, J. Uleman, J. Bargh (Eds.), The New Unconscious (pp. 37–58). New York: Oxford University Press. Conway, M. (2003). Cognitive-affective mechanisms and processes in auto-biographical memory. Memory, 11, no. 2, 217–224. Damasio, A. (1994). Descartes’ error: Emotion, reason, and the human brain. New York: G. P. Putnam’s Sons. Damasio, A., Grabowski, T. J., Bechara, A., Damasio, H., Ponto, L. L. B., Parvizi, J., & Hichwa, R. D. (2000). Subcortical and cortical brain activity during the feeling of self-generated emotions. Nature Neuroscience, 3 (no. 10), 1049–1056. Davies, P. S. (2009). Subjects of the world: Darwin’s rhetoric and the study of agency in nature. Chicago: University of Chicago Press. Davies, P. S. (2011). Ancestral voices in the mammalian mind: Philosophical implications of Jaak Panksepp’s affective neuroscience. Neuroscience and Biobehavioral Reviews, 35 (issue 9), 2036–2044. Davies, P. S. Forthcoming. Skepticism Concerning Human Agency: Sciences of the Self vs. ‘Voluntariness’ in the Law. In N. Vincent (Ed.), Neuroscience and Legal Responsibility. New York, Oxford University Press. Dobzhansky, T. (1973). Nothing in biology makes sense except in the light of evolution. Ameri­ can Biology Teacher, 35, 125–129. Green, J. (2011). The moral brain and how to use it. New York: Penguin Group. Haidt, J. (2006). The happiness hypothesis. New York: Basic Books.



7.  Nothing in mammalian psychology makes sense 173

Oatley, K., & Johnson-Laird, P. N. (1987). Towards a cognitive theory of emotions. Cognition and Emotion, 1, 29–50. Panksepp, J. (1998). Affective neuroscience: the foundation of human and animal emotions. New York: Oxford University Press. Panksepp, J. (2003). At the interface of the affective, behavioral, and cognitive neurosciences: Decoding the emotional feelings of the brain. Brain and Cognition, 52, 4–14. Panksepp, J. (2005). Affective consciousness: core emotional feelings in animals and humans. Consciousness and Cognition, 14, 30–80. Panksepp, J. This volume. In defense of multiple Core Affects. Panksepp, J., Normansell, L. A., Cox, J. K. F., & Siviy, S. (1994). Effects of neonatal decortication on the social play of juvenile rats. Physiology and Behavior, 56, 429–443. Posner, J., Russell, J., & Peterson, B. (2005). The circumplex model of affect: An integrative approach to affective neuroscience, cognitive development, and psychopathology. Develop­ ment and Psychopathology, 17, 715–734. Pronin, E. (2007). Perception and mis-perception of bias in human judgment. Trends in Cogni­ tive Science, 11, 37–43. Russell, J. (2003). Core affect and the psychological construction of emotion. Psychological Re­ view, 110, 145–172. Russell, J. This volume. From a psychological constructionist perspective. Russell, J., & Barrett, L. F. (1999). Core affect, prototypical emotional episodes, and other things called emotion: dissecting the elephant. Journal of Personality and Social Psychology, 76, no. 5, 805–819. Shemon, D. A., Holmes, G. L., & Byrne, P. A. (1999). Consciousness in congenitally decorticate children: developmental vegetative states as self-fulfilling prophecy. Developmental Medi­ cine and Child Neurology, 41, 364–374. Wegner, D. (2002). The illusion of conscious will. Cambridge, MA: MIT Press. Wilson, T. (2002). Strangers to ourselves: discovering the adaptive unconscious. Cambridge, MA: Harvard University Press.

chapter 8

Lessons for affective science from a metascience of ‘molecular and cellular cognition’ John Bickle Departments of Philosophy and Religion, and Psychology, Institute for Imaging and Analytical Technologies (I2AT), Mississippi State University

The disputes between Jaak Panksepp’s “primary affects” neurobiological-cum-evolutionary approach and Jim Russell’s psychological constructionism based now on “Core Affect” look familiar. Their phenomena are emotions rather than cognitions, but that’s no longer novel. And while theirs isn’t quite just another instance of the grizzled old “top-down” versus “bottom-up” research methodology dispute, it’s not too far removed. Where, among the many available “levels” of mind-brain organization, should one direct his or her research efforts? There is no escaping this question’s significant contextual dependence: on one’s background knowledge, and the experimental designs, data collection and analysis procedures, and explanatory approaches one is familiar with, including experimental tools and techniques. All talk of “we” and “our” in answers to these questions – what “we” know about the target phenomenon or its neural mechanisms, about “our” experimental tools and strategies – rings hollow. There are experiments and explanatory insights that Jaak’s lab can do and have, that Jim’s can’t; and vice versa. This is common across all the disciplines that make up mind-brain science, affective or cognitive. And yet the question still looms for everybody: given what I know, and can do in my lab or study, where is the fishing going to be best? I’ll stress two general methodological points in this commentary. Both are based on a metascientific study of a hot field in current neuroscience, molecular and cellular cognition. I’ll explore how my two points answer the editors’ questions #4 (inferring affects from behavioral observations), #7 (relationships between affect and cognition), #9 (the respective roles of biology and psychology in studying affect and emotion), and #11 (prospects for a coherent synthesis in this “splintered” field). Finally I’ll contrast my answers with those offered by Panksepp and Russell. Such is the task of a commentator; the negative things I say below

176 John Bickle

about Panksepp’s and Russell’s target articles should not be taken to indicate that I didn’t appreciate many of their points, didn’t enjoy reading and thinking about their arguments and conclusions, nor didn’t learn a lot about affective science by doing this.

1. Eliminativism as metascience (or eliminativism with a little ‘e’) Both Panksepp and Russell speak disparagingly about “folk” emotion concepts. Though Panksepp uses common English emotion terms to name his primary affects, he capitalizes his usages to distinguish these terms, even syntactically, from their vernacular relatives (e.g., ‘seeking.’ ‘rage,’ ‘fear’). And he clarifies the referents of these technical terms: specifiable brain networks that generate a variety of evolutionarily-provided raw emotional affects. For Russell, a general English vernacular term like ‘emotion,’ and even more specific emotion words like ‘fear,’ play at most an “honorary role” in science, but do no “real scientific work.” This seems to mean that their referents, whatever those may be, don’t figure as mechanisms in real scientific causal-mechanistic explanations. Russell writes: “Thus, we need an explanation of how the encounter with the bear resulted in changes in the autonomic nervous system, but that explanation would not include the assumption that fear did it” (my emphasis). Both Panksepp and Russell are onto an important methodological insight here. It is the insight that should have been the take-home lesson from the eliminative materialists of 1980s philosophical fashion. Eliminative materialists predicted that a developed neuroscience will reveal significant shortcomings in our folk psychology qua explanatory framework, to a degree that folk psychology will be revealed as deeply mistaken, even “radically false.” Unfortunately, in the hands of philosophers, this important methodological point about how science should proceed became an “ontological dispute.” Eliminative materialism became the “ontological hypothesis” that beliefs, desires, intentions, fears, cares, concerns, and the rest of the propositional attitudes (cognitive, conative, and emotive) do not exist; that folk psychology’s “ontology” is in the process of being replaced by neuroscience’s ontology as our best account of “what there is” concerning human behavior. Naturally, this “ontological turn” led philosophers away from the interesting scientific projects that hatched eliminativism, and into territories more familiar to them: theories of reference, term extensions and intensions, semantics, and the ontology of relations like supervenience and realization – in other words, back to all those topics that academic philosophers worry about pathologically, but nobody else cares a lick.



8.  Lessons for affective science from a metascience of ‘molecular and cellular cognition’ 177

Eliminative materialism about folk psychological concepts, kinds, and explanations should have been a metascientifically derived claim about methodology in the mind-brain sciences. That is, it should have been the claim, internal to the actual experimental and explanatory practices in a variety of successful mind-brain sciences, that folk psychological concepts were not being used in the best causal-mechanistic explanations on offer. The metascientific arguments for it should have been a compilation of the kinds of reasons both Panksepp and Russell offer here for why vernacular English emotion words are not and should not be used in serious mind-brain science. This is ‘eliminative materialism’ with a little ‘e,’ and little requisite bombast. It is just a fact of the matter about the role that folk psychological terms and posits play in the actual practices of current mindbrain sciences, from scientific psychology to neuroscience.

2. Silva, Landreth, and Bickle’s Convergent Four, applied to the science of affect Drawing on early suggestions by neurobiologist Alcino Silva, recently Silva, ­Anthony Landreth, and I have developed a metascientific account of the sufficient conditions implicit in actual scientific practice to justify causal-mechanical hypotheses. We draw cases for our metascientific analyses from landmark primary experimental publications and review papers in the relatively new neuroscientific field of ‘molecular and cellular cognition’ (MCC). (Not only was Silva’s earliest scientific work instrumental in kick-starting this field, he also coined its name.) Starting in earnest in the early 1990s, with the application of genetic engineering techniques to mammals (typically mice), this work is now being pursued by over 100 laboratories worldwide, with a professional society (the Molecular and Cellular Cognition Society) with over 2300 current members. Mechanisms of learning and memory have constituted its key experimental successes so far, but work is underway on sensory, motor, attentional, affective, and clinical phenomena: the entire range that composes cognitive and affective science. Since I’ve described MCC research practices and numerous landmark examples in a manner friendly to non-speciaslists (see especially Bickle 2006, 2007, and 2009), and since we go into great detail on this point in Silva, ­Landreth, and Bickle (forthcoming), I’ll be very brief here. MCC research typically involves using state-of-the-science molecular genetic techniques to intervene into the genomes of mammals, usually mice, to increase or decrease the gene expression and subsequent protein synthesis of specific molecules. With regard to learning and memory, typical molecular targets include: cyclic

178 John Bickle

a­ denosine monophosphate­ (cAMP) response-element binding protein (CREB), a transcriptional enhancer for a variety of regulatory and effector genes and proteins; cAMP-dependent protein kinase A (PKA), whose catalytic subunits phosphorylate transcriptional effectors like CREB to activate them; and calcium-calmodulin kinase II (CaMKII), which increases the capacity of postsynaptic glutamate (excitatory) receptors to pass sodium ions (Na+) across the cell membrane and activates “hidden” glutamate receptors in the post-synaptic density. The genes for these specific molecules can be “knocked out” – rendered non-expressive – either permanently, during the embryonic stem cell stage of development, or temporarily in the adult animal using a variety of genetic and pharmacological tricks. Or additional copies of these genes can be inserted into genomes using transgenic techniques, to increase the amount of protein product in specific neurons. Genetically manipulated animals are then tested on a wide range of behavioral measures for specific cognitive functions, and their behaviors are contrasted with those of “wild-type” littermate control animals (who don’t undergo the molecular-genetic intervention). Significant behavioral differences on these measures indicate a role for these genes and their protein products in causal-mechanistic explanations of the specific cognitive function being investigated. Numerous and extensive controls – molecular biological, molecular-genetic, and behavioral – routinely are performed by MCC experimenters, both to insure that the genetic manipulation and subsequent protein synthesis were induced correctly, and to isolate the specific cognitive function from potentially confounding cognitive effects. MCC results over the past decade have been stunning, especially for the molecules mentioned above and others in the intracellular signaling pathways leading from neuron activation to synaptic plasticity. There are now literally hundreds of published MCC experimental results, populating the pages of the best scientific journals. Studying the most influential of these MCC experiments and results metascientifically, Silva, Landreth, and I hypothesize the implicit conditions on experimental evidence deemed sufficient for justifying a causal connection between a proposed cellular or molecular mechanism and a complex, system-level cognitive phenomenon (Silva, 2007; Silva and Bickle, 2009; Silva, Landreth, and Bickle, forthcoming). We’ve isolated four components, which we call collectively the “Convergent Four” (C4). Each component is either a type of experiment or an analysis performed on the results of these experiments. For any hypothesized cause or mechanism A, such as one of the intracellular signaling molecules mentioned above, and a hypothesized effect B, such as an acknowledged behavioral indicator or measure of a specific cognitive function, the C4 (stated starkly) are:



8.  Lessons for affective science from a metascience of ‘molecular and cellular cognition’ 179

1. Positive manipulations: experimental manipulations that increase the probability, extent, or duration of A, and measure the probability, extent, or duration, of B; 2. Negative manipulations: experimental manipulations that eliminate or decrease the probability, extent, or duration, of A, and measure the probability, extent, or duration of B. 3. Non-intervention correlation measures: experiments that measure the correlation of occurrences of A and B, without directly manipulating the probability, extent, duration of either A or B experimentally; 4. Integration: the analysis of results from a collection of experiments, on A and B, and on other phenomena related to both, of types (1), (2), and/or (3), to infer the causal structures among the phenomena studied in those experiments. Integrations can be as simple as conjoining results from a primary experiment (of types (1), (2), or (3)) with those from a single control experiment to rule out another phenomenon, C, as a causal mechanism of B. More complex Integrations are common in experimental literature review papers and meta-analyses. Analysts check for consistency of results across experiments investigating relationships between specific phenomena, or infer the collective causal network structure among a variety of related phenomena based on multiple experimental studies. These structures are often represented in path diagrams, a key type of integration. Our first metascientific hypothesis holds that when all C4 components are met, for a specific molecular mechanism and acknowledged behavioral measures/indicators of a specific cognitive function, scientists judge that a sufficient experimental case has been made to justify a causal mechanistic hypothesis. For example, CREB, in its capacity as a specific gene expression enhancer induced by cAMP and protein kinase A, is a mechanism of the consolidation of a class of short-term memories into long-term form, including numerous hippocampusdependent “declarative” or “explicit” memories. We derived the C4 directly from landmark MCC experiments. But our metascientific C4 hypothesis has broader reach. We further hypothesize that much if not all of causal-mechanistic science (notice we explicitly did not say “all of science”!) implicitly relies on the C4 as sufficient for establishing specific causal-mechanistic hypotheses. Whether this extended C4 hypothesis is justified metascientifically­

. Integration turns out to be a key to our account, and in Silva, Bickle, and Landreth (forthcoming) we dedicate an entire chapter to its metascientific description and role.

180 John Bickle

remains an open question. But assuming for the sake of argument that it is, (something we considered in Silva and Bickle, 2009, and consider further in Silva, Landreth, and Bickle, forthcoming), this result carries important methodological implications for some issues that Panksepp and Russell tussle over in this volume. I turn to these.

3. Bickle versus Panksepp versus Russell on some of the editors’ questions Question #4: How can we infer affect from observations of behavior? MCC studies use animal models – rodent, and more specifically, mouse – of cognitive functions. The key experimental interventions in these studies are ­molecular-biological and molecular-genetic, to increase or decrease a specific protein product. Historical, technical, and ethical limits restrict these interventions in mammals. So we need accepted behavioral measures of cognitive functions in mice to investigate potential molecular mechanisms in positive manipulation, negative manipulation, and neutral correlation experiments. How are these behavioral measures chosen? Simply, based on their historical use, vetting, and experimental control of potential confounds by animal experimental psychologists. The Morris water maze for spatial memory, contextual fear conditioning for one-trial hippocampus-dependent associative memory, object recognition for non-spatial memory, social transmission of food preferences and reduction in stereotypic rodent exploratory behaviors on subsequent exposures for social memory – these are the standard behavioral protocols employed in MCC experiments. They’ve been reliably used for decades in animal experimental learning and memory studies. Control studies have been performed to address every conceivable potential confound anyone has worried seriously about. So what’s the relationship between these observable behavioral measures and the cognitive function under investigation? From a metascientific perspective on MCC, the standard measures in behavioral protocols are observable indicators of the occurrence, strength, and extent of the cognitive function under examination. They are operationalizations of the cognitive function for the purposes of experimental investigations into underlying cellular and molecular mechanisms. Is this dreaded behaviorism, still rattling its chains? No. These accepted behavioral indicators are not offered as analyses of cognitive concepts. They’re rather indicators we can measure reliably, across experiments and labs, which have stood the test of use and revision in light of many controls, and that we are now confident point to a specific cognitive function’s occurrence. They’re now part and parcel of the laboratory search for underlying causal mechanisms for these cognitive functions. If you don’t like one particular behavioral protocol as an indicator for your favorite



8.  Lessons for affective science from a metascience of ‘molecular and cellular cognition’ 181

cognitive function, fine. But if you also want to continue to search scientifically for the cellular and molecular mechanisms of that cognitive function, your only option is to find a better behavioral indicator for it – one that more closely meets your ideas about what that function is. There’s no opting out of using some behavioral measure to indicate that function’s occurrences; at least, there’s no opting out and still claiming to pursue causal-mechanistic scientific explanations of it. So if there are aspects of affect, emotionality, or conscious experiences that you think cannot be operationalized into a set of behavioral measures for pursuing causal-mechanistic explanations via C4 experiments, then one conclusion is inescapable: those aspects are not addressable by causal-mechanistic science. Ruling out all behavioral indicators for some psychological phenomenon rules it outside of causal-mechanistic science tout court. Because that’s just how causal-mechanistic­ science proceeds, according to a meatscience of current MCC. Perhaps there are some philosophers comfortable with this conclusion; but I like to think that most scientists will think that addressing that phenomenon is still worth a try. Note that this MCC-inspired metascientific answer to Question #4 is consonant with some of Panksepp’s answer. His primary affects are aroused by “brain stimulation.” Notice that such experiments are straightforward instances of our C4 “positive manipulations.” A specific primary affect arouses specific “coherent emotional-instinctual behavioral sequences,” which implies that “at a neurethological level, one can probably use instinctual emotional responses as proxies for monitoring distinct affective states.” ‘Proxies’ is a less than crystal clear term, but Panksepp here seems to have something like my metascientific answer to Question #4 in mind. My answer to Question #4 is also in keeping with his numerous remarks about the inevitability of animal research in the quest for causalmechanistic­ explanations of affect and emotion, and his refreshing remark that “we may scientifically understand primary-process emotionality better through animal than human research.” But there’s no weaseling out when it comes to causal-mechanistic scientific investigations of tertiary-level affective processes! If these aspects of affect really are “not yet capable of being monitored empirically other than by verbal subjective self reports,” then they’re not yet amenable to scientific causal-mechanistic explanation. Period. Perhaps we’ll just have to rest content for awhile with Panksepp’s complaining and relief-sighing rats as behavioral indicators for tertiary affective processes, and start investigating their cellular and molecular mechanisms via the C4. At least until better rodent measures for tertiary processes are found. Russell, on the other hand, interprets Question #4 differently, and from a metascientific perspective this difference is revealing. He speaks of using factor analyses on human self-reports of specific emotional and mood states. He notes that numerous such studies and analyses reveal that self-reports display

182 John Bickle

correlations­ between emotional, affective, and mood states along two “large factors”: valence of states, ranging from negative to positive, and amount of arousal of states, from mild to highly charged. In other words, persons who self-report experiencing a particular mood or emotion positive in valence will with high probability also self-report experiencing other affective states positive in valence, and will with low probability self-report experiencing states negative in valence. Similarly for self-reports of particular states with negative valence, low arousal, and high arousal. Russell employs these as the two dimensions of Core Affect. In this sense, “observable behavior” of a variety of domains “can be explained through the operation of an underlying process like Core Affect.” But notice that the resulting explanations are not causal-mechanistic, at least not in the sense that current MCC experiments, and apparently Panksepp’s approach to investigating primary affects, justify. That both, e.g., fear and anger share similar locations near the negative endpoint of the valence spectrum explains nothing, causal-mechanistically, about the specific behaviors we observe in self-report studies. Inferring the specific dimensions of Core Affect from factor analyses of human self-reports is a very different kind of scientific project than is using specific behavioral protocols to indicate the occurrence of a specific cognitive or affective state, in order to investigate potential cellular and molecular mechanisms. The first process involves inferring affect from observable behavior for the purpose of scientific concept con­ struction; the second process involves inferring affect from observable behavior for the purpose of controlled experimental manipulations into potential cellular and molecular mechanisms. These are two very different scientific projects. Question #7: What is the relationship between affect and cognition (conceptualization, intentionality, appraisal, etc.) James McGaugh has been at the forefront of studying the (positive) effects that naturally occurring stress hormones released by emotional arousal, such as epinephrine and corticosterone (in rats), have on memory consolidation. His lab has traced the molecular pathways producing these effects on synaptic plasticity, and the circuitries containing the neurons affected, by doing the appropriate negative manipulation, positive manipulation, and neutral correlation experiments, and integrating the findings, in ways consonant with our C4 hypothesis. They have used numerous behavioral protocols as indicators for different kinds of rodent memory. McGaugh’s work should be embraced enthusiastically by Panksepp, as it consists of well designed causal-mechanistic studies into the “rich broth at the bottom of the evolutionary crucible” of affect. . A good recent review, not overly technical and with many references to the primary experimental literature, is McGaugh and Roosendaal (2008).



8.  Lessons for affective science from a metascience of ‘molecular and cellular cognition’ 183

Of course, McGaugh’s studies have been limited to Panksepp’s primary affects and “the basic mechanisms of memory.” But such is the state of play now. Despite Russell’s speculations about Core Affect and its modular encapsulation from cognition, I’m sympathetic to Panksepp’s assessment: “I personally remain more gratified by the diverse emotional understanding that novelists, film-makers, and other artists of the human affective experiences have provided.” Nobody has viable real scientific – i.e., C4-meeting – evidence for any genuinely causal-mechanistic hypothesis for these experiences, and there’s no reason to pretend we’re on track of something when we’re not. Even the functional neuroimaging results mentioned by Panksepp in this context, intriguing and maybe experiment-guiding as they are, at most and at best only provide C4 neutral correlational evidence. Panksepp rightly sees their limits with regard to causal-mechanistic science, even if he doesn’t articulate his point using the helpful resources of the C4. Until someone figures out how to do the right positive and negative manipulation experiments – and here we shouldn’t forget the very real limits that technical and ethical considerations impose – and integrates the results of those experiments into a plausible story of how those experimental interventions bring about those specific behavioral effects we judge as indicative of “human affective experiences,” crucial kinds of C4 evidence will remain missing in action. No point pretending otherwise. Question #9: What roles do the biological sciences and psychological sciences have in studying affect? From a metascientific perspective on even a “ruthlessly reductive” field like MCC, we can put one nagging worry to rest. Psychologists need not fear that research into the biological mechanisms of cognition or affect will eclipse or replace their science. The C4 makes it clear that psychological investigations provide crucial parts of the very evidential basis that makes for a sufficient experimental case for a cellular or molecular mechanism for a psychological function. Both neutral correlation experiments and many aspects of integration require psychological research. Even the choices of behavioral measures to indicate psychological functions in positive and negative manipulation experiments require psychological justification and continued development. These are jobs for experimental psychologists, not molecular biologists or cellular physiologists. It should thus surprise no one that the top MCC labs worldwide work closely with experimental psychologists and cognitive/behavioral neuroscientists, even though their ultimate goals are the cellular and molecular mechanisms doing the brain’s causal work. But is it really the case that these cross-level interactions amount to biology’s “complementing” psychology, as Russell insists? From a metascientific perspective on ongoing mainstream cellular and molecular neuroscience, is the relationship really one of “compatibility” and “mutual benefit”? I don’t think so – at least, the

184 John Bickle

relationship between biology and psychology doesn’t stop with these. Consider Russell’s own argument for why psychology matters for biology. “My suggestion is that the concept of, for example, fear, doesn’t carve nature at its joints. If so, then biological theories that use that concept are headed for the same fate” (Russell’s emphasis). His lesson seems to be that psychology can save biology from exploratory dead-ends, such as searches for “the central state of fear” in the brain. He points to recent work in which two different behaviors thought to be effects of “the central state of fear” turned out both to be elicited by the same stimulus, but were revealed to be “supported by different brain circuitries” and hence “dissociable.” Russell takes this work to indicate that “the central state of fear” is “too heterogeneous for scientific purposes.” I have no quarrel with that conclusion, but I point out that psychology didn’t find the key dissociation here. Neurobiology did: different brain circuitries. This particular example is hardly a good indicator that psychology matters for neurobiology. Panksepp’s views on Question #9 seem more in keeping with metascientific lessons culled from MCC experiments. But he too makes some remarks that invite challenge. Most “dual aspect monisms” leave unanswered serious questions about the causal status of that second class of aspects. What are they composed of? Where do they come from and how do they come about? What do they in turn do (cause), and how, given their “dual” nature? Muttering darkly about ‘emergence’ or ‘realization’ helps not at all. And while it sounds ecumenical to “call upon all emotion-psychologists” to “consider evidence” about neurochemical transmitter molecules and neuropeptides, we wonder: to what specific psychological ends? What is the psychologist qua psychologist supposed to do with the available ­evidence? Doesn’t “considering this evidence” in any genuine fashion turn psychologists into neuroscinetists? Fortunately, the metascientifically derived C4 offers an answer to what psychologists qua psychologists contribute: psychologists play a crucial role in ­operationalizing cognitive and affective functions into reliable, experimentally accessible behavioral measures and indicators for the animal models we can intervene into. They help design and perform the behavioral measures on neutral correlation, negative manipulation, and positive manipulation experiments. And they contribute to the last stages of integration that achieves a plausible, evidencebased account of how the cellular or molecular effects intervened into ultimately produce the behaviors indicative of the psychological function being investigated. That’s honest work that contributes an important part of the very evidential basis to a sufficient case for a scientific hypothesis about the mechanisms of specific cognitive and affective functions. It is also helpful to note that an even more limited potential “fate” awaits molecular biology. As nanotechnology and biophysical engineering increasingly



8.  Lessons for affective science from a metascience of ‘molecular and cellular cognition’ 185

­ elivers techniques for intervening directly, on the fly, into protein biochemistry, d without having to go through the clumsy stages of gene transcription and protein synthesis, molecular biology’s role in the search for causal mechanisms of cognitive functions will diminish. Psychology’s roles, specified just above, will all remain intact. By then it will be the biochemists and proteonomic engineers who’ll be doing the “manipulation” parts of positive and negative manipulation experiments, into potential biochemical mechanisms for cognition and affect. Ruthless reductionism indeed. Question #11: Considering the diversity of theoretical viewpoints in the scientific study of affective and emotional phenomena, might one hope for a coherent synthesis in this splintered field? More than a decade ago, in a published review of Gazzaniga (1997) – a book of email discussions between Michael Gazzaniga and various luminaries in the then-current field of cognitive neuroscience – I offered the following remarks: “One surprising theme that emerges throughout these interviews is that the interdisciplinary interaction that supposedly characterizes cognitive science is commendable in principle but still a myth in practice” (Bickle, 1999: 220). I quoted extensively from the book to support the first conjunct of that claim, and then continued: “Concerning interdisciplinary interaction in practice, however, these interviews tell a different story” (Bickle, 1999: 220). I followed this with quotes from many of the interviews that were thinly veiled complaints about what scientists investigating other levels of mind-brain organization were delivering. I remarked that these complaints were “hardly what we would expect from an endeavor that considers itself at the cutting edge of interdisciplinary interaction!” (Bickle, 1999: 221). It’s comforting in an odd sense, I suppose, to see in Russell’s and Panksepp’s discussions that my cross-disciplinary assessment from more than a decade ago remains accurate today. Everybody claims to want “interdisciplinary interaction,” but typically that means everybody else kow-towing to one’s own pet projects and concerns. Low and behold! That doesn’t happen. Psychologists track what’s hot in their fields, cognitive neuroscientists in theirs, cell physiologists in theirs, and molecular biologists in theirs. Rarely do these concerns intersect, ­except in either a single lab or at levels of abstraction far removed from actual scientific practice. Russell’s formula for bringing about more actual interdisciplinarity is to follow the little-e eliminativist methodology I discussed earlier. Re-phrasing ­commonsense English emotion words into more specific scientific terminology will “eliminate much of the appearance of diversity in today’s theoretic viewpoints.” From my discussion above, obviously I find Russell’s methodological recommendation helpful, and in keeping with actual scientific practice. But I doubt that it’s ­going

186 John Bickle

to bring us all together. The problem is the appearance of competing causalmechanistic­ accounts of the same phenomenon, at higher and lower scientific levels, and the still widespread avowal to avoid levels “reductionism.” If we really want scientific unity – and I’m no longer convinced that we do – we need some way of erasing the appearance of competing, irreducible different level causalmechanistic explanations of the same data. In the way it divvies up the scientific labor, the metascientifically-derived C4 may provide us with something like that. The C4 may also provide a framework for Panksepp’s request that investigators “situate themselves more clearly at the level at which they are really studying such phenomena.” According to the C4, the jobs differ among investigators “situated at distinct levels,” in terms of what each contributes to the sufficient evidential basis for specific causal-mechanistic hypotheses. Panksepp wonders if the desired synthesis “can only be achieved by broadly trained, empirically-immersed neurophilosophers.” If he’s right about the necessity of such contributors, then none of us should be holding our breaths. Academic philosophy moved quickly away from the “neurophilosophy” that seemed so promising twenty-five years ago. There are still outposts in the profession, training graduate students to do more than just read popular neuroscience undergraduate textbooks, but their numbers are dwindling. Having recently edited the Oxford Handbook of Philosophy and Neuroscience, I can say with some confidence that the field mostly has become another comfortably narrow specialty in the collection of philosophies of the special sciences. Additionally, metaphysics has come back into vogue across philosophy, and nothing coming from that trend is of the slightest help to practicing scientists confronting problems about crosslevel synthesis. Finally, even mirror neurons make their way into an answer to Question #11. Panksepp is a fan. They’re a “poignant and important example” of possibly “new and intrinsic circuits for affective experiences that may be unique for our species.” They have “captured the imagination of all of us.” Well, maybe not all of us. And especially maybe not the very scientists whose attention we’d most expect, if they really are so unique and important: cellular physiologists. In a recent Quick Guide paper in Current Biology, Christian Keysers (2009) addresses the question of “why mirror neurons are so controversial.” He points out that approximately 1200 papers have been published since the mid-1990s discovery of mirror neurons that refer to their features and speculate about their roles in higher functions ranging from empathy and mind reading to autism to politics, aesthetics, and morality. Over that same period, how many publications have reported direct experimental electrophysiological data for the existence and basic cell-physiological properties of mirror neurons? Keysers reports, “about one dozen.” (2009, R972). This discrepancy between serious neuroscience and rampant speculation leaves



8.  Lessons for affective science from a metascience of ‘molecular and cellular cognition’ 187

even a mirror neuron enthusiast like Keysers to conclude that the frequency with which the term ‘mirror neuron’ occurs in the literature should leave us with “some unease.” One would think that the most show-stopping, groundbreaking, theorygenerating discovery in recent cellular physiology would have attracted the attention of … cellular physiologists! Mirror neurons have not. Which leads some of us to wonder if the theoretical and philosophical hoopla that continues to be churned out regarding them might be a little … misplaced?

References Bickle, J. (1999). Review of Gazzaniga, Conversations in the Cognitive Neurosciences. Philosophi­ cal Psychology, 12, 221–223. Bickle, J. (2006). Reducing mind to molecular pathways: explicating the reductionism implicit in current mainstream neuroscience. Synthese, 152, 411–434. Bickle, J. (2007). Who says you can’t do a molecular biology of consciousness? In M. K. D. Schouten & H. de Jong (Eds.), The Matter of the mind. London: Blackwell, Chapter 13. Bickle, J. (2009). Real reduction in real neuroscience: metascience, not philosophy of science (and certainly not metaphysics!). In J. Hohwy & J. Kallestrup (Eds.), Being reduced. Oxford: Oxford University Press, 34–51. Gazzaniga, M. (1997). Conversations in the cognitive neurosciences. Cambridge, MA: MIT Press. Keysers, C. (2009). Mirror neurons. Current Biology, 19, R971–R973. McGaugh, J. L. & Roosendaal, B. (2008). Drug enhancement of memory consolidation: historical perspective and neurobiological implications. Psychopharmacology, 202, 3–14. Silva, A. J. (2007). The science of research: the principles underlying the discovery of cognitive and other biological mechanisms. Journal of Physiology (Paris), 101, 203–213. Silva, A. J. & Bickle, J. (2009). Science of research and the search for the molecular mechanisms of cognitive functions. In J. Bickle (Ed.), Oxford Handbook of Philosophy and Neuroscience. Oxford: Oxford University Press, 71–126. Silva, A. J., Landreth, A., and Bickle, J. (forthcoming). Engineering the next revolution in neuro­ science. New York: Oxford University Press, forthcoming 2012.

. A few minutes perusal of an ISI Web of Science search for ‘mirror neurons’ reveals that this ridiculously skewed distribution of theoretical interest and real scientific investigation has not changed since Keysers’ publication.

chapter 9

Affect as appraisal Meaghan A. Leddy*, Gail C. Robertson* and Jay Schulkin** *American University, Department of Psychology / **Georgetown University School of Medicine, Department of Neuroscience

1. Introduction There have long been intellectual discussions about emotion, as can be seen in the writings of Plato (360 B.C.E./1956), Spinoza (1668/1955), Freud (1924/1960), and Sartre (1943/1973). However, the acceptance of the reality of emotions has fueled numerous debates surrounding the function, structure, location, and components of emotions. Two prominent names within the field of affect and emotion are Jaak Panksepp and James Russell, and both have provided eloquent depictions of their viewpoints. Panksepp, in part, assisted in the revitalization of the study of affect with his impressive book, Affective Neuroscience (1998); Russell’s contribution has been to provide a rational air and meaningfulness to the discussion of affect and emotion. Panksepp and Russell discuss affect as a diffuse entity, and we believe this limits the usefulness of defining the term. Definitions bring us back to concepts and kinds; strictly delineated definitions only go so far, and serve as Platonic seductions, which are of little benefit. Aiming to answer the question “What are emotion and affect?” with single definitions gives the sense that these concepts have transparent, universally accepted definitions, and perhaps even objective measurements, in the way that “density” or “length” do. For example, diverse, ill-fitting definitions have been given to equally controversial topics such as “stress” and “intelligence” that have resulted in decades of intellectual debates over little more than semantics. Thus, we believe that while theory and facts are not universally accepted in the field of affect and emotions, it is beneficial to wait before assigning a priori explanatory labels and definitions to these constructs (Kagan, 2007).

190 Meaghan A. Leddy, Gail C. Robertson and Jay Schulkin

We touch upon each of the questions discussed by the original articles, but, given our interests, focus on the discussion of affect and cognition, aiming to demythologize this topic. We suggest that affect has functionally evolved/ developed­ as an anchor for emotional experience and behavior. It follows that with the evolution of this functionalism, the interconnected dependence of affect and emotion with cognition becomes increasingly evident. We will argue this case, with the added recognition of social influences and perspectives on emotional experience.

2. A perspective on affect and emotion Affect is fundamental, like breathing. What makes affect fundamental is its ability to provide information to an individual about his or her needs. Affect is a quick way to gather information and serves as a starting point for the work of implementing multiple appraisal systems and emotion. Affective changes are consistently expressed by the body, and are often (but not always) available to conscious awareness (James, 1890/1950; Panksepp, 1998; Schulkin, 2004; Stellar & Stellar, 1985). We view affect as a fluid and changing combination of several parts of a larger system: an integrated response to a survival-relevant event, which requires adaptive, cooperative action across levels of neural function (also see Scherer, 1993). Information processing within physiological systems essential for regulatory functioning is not divorced from the impact of bodily experience (e.g., Richter, 1976; Stellar, 1974). We like Russell’s color wheel depiction of his circumplex model, viewing affect as an orientation concerning arousal and valence; one can be in a positively aroused state, a positively calm state, a negatively aroused state or a negatively subdued state (Russell, this book; Russell, 2003). If one encounters an event where he receives negative feedback, the individual’s response will differ based on affect; if the individual is in a negative-calm state, he may experience the emotion of depression, and return to his desk saddened. However, if the individual was starting from a negative-aroused baseline affect, his reaction might include confronting the criticizer and beginning an altercation. In line with Russell’s description (Russell­, this book; Russell, 2003), we believe affect has four possible states defined by valence and arousal dimensions. This is different from emotion in that there are a multitude of possible emotional experiences that fall within any single quadrant (see Figure 1; Russell, 2003). Affect is often unfocused in nature, and serves as a baseline “temperature” from which emotional experience can deviate to a certain extent. For example, if one wakes up with negative affect (i.e., a diffuse negative feeling that is not



9.  Affect as appraisal 191

Positive Relief

Contentment

Joy Excitement

Calm Inhibited

Aroused Regret

Irritation Anger Fear

Depressed

Rage

Hopeles Negative

Figure 1

directed at any individual or event) then every encounter will be colored by that affect and thus even happy emotional experiences may be somewhat dampened; in this way affect serves as an anchor for subsequent emotional experience. It is in such examples that one is able to infer affect from behavior; this person had a negative and active response, they must have been in the negative-aroused quadrant of affect. However, there is not a 1:1 correlation between behavior and affect; for example, the experience and consequences of a certain affective state may differ based on context (Feldman-Barrett and Kensinger, 2010) or even temperament. Concepts relevant to discussions of temperament, like inhibition and control or whether events are discrepant from or adherent to expectations (Hebb, 1946; Kagan, 1997), can influence an affectively-tinged emotional response. For example, that same individual whose affect was negative and aroused may be temperamentally inhibited, and thus may exhibit significant control over his behavioral response, masking both his initial affect and emotional experience. Instead of arguing with his supervisor, he may inhibit that reaction and overcompensate by being extremely kind to his boss. Despite this, emotional expression is typically a good measure by which to determine affect, though it is imperfect. (Darwin, 1872/1965; Stellar & Stellar, 1985). Animals and infants both have affect. What they lack is the finer differentiation of the multiple types of emotion in each of Russell’s quadrants (e.g. anger and fear are both in the negative-aroused sector), that comes with development and the growth/maturation of multiple appraisal systems. These “primordial” affects orient the organism to approach or avoid, seek attachment, etc. It is not to say that these primordial affects are without cognitive appraisal systems, and we disagree

192 Meaghan A. Leddy, Gail C. Robertson and Jay Schulkin

with Panksepp (Panksepp, this book; Panksepp, 1998) that these appraisals are of secondary importance. The innate pieces of affect are pre-linguistic. Once linguistic systems and a lexicon develop within an individual and culture, language, affect, and emotion become intertwined. For example, Wager and colleagues (2008) conducted a meta-analysis and concluded that inferior frontal gyrus (IFG), the pars opercularis (Broca’s area, BA 44), pars triangularis (BA 45) and pars orbitalis on the inferior frontal convexity (BA 47/12 l) are all part of the distributed neural systems that are implicated in perceiving emotion. These areas are also implicated in linguistic capabilities (e.g. Gitelman, et al., 2005). Further evidence for this comes from studies that have shown that reaction times are faster when facial expression and sentence valence match than when they are incongruous (Havas, Glenberg and Rinck, 2007). The authors suggest that this is evidence that emotional systems contribute to language comprehension, demonstrating a link between affect, emotional experience and language. As the brain develops and acquires language, affect and emotion are impacted. This view is similar to Panksepp’s (1998) in that we agree that the breadth of emotion follows the ontogeny of an individual, increasing as the brain develops and other systems mature, such as memory, cognition, and predictive abilities. Some evidence for this comes from the fact that congenitally blind infants (Fraiberg, 1977), children (Roch-Levecq, 2006) and adults (Galati, et al., 1997) can produce several predicted facial action units when portraying emotion, though they rarely exhibit an entire configuration, indicating that some components of emotional experience and expression are acquired through learning. Similarly, studies have demonstrated that 6 month-old infants can discriminate between faces of different emotional valences (Bornstein & Arterberry, 2003), and perhaps even arousal (Flom and Bahrick, 2007). With regard to differentiation, we are in agreement with Russell that the most evolutionarily important components (pleasure-displeasure, which are necessary for reward and learning, and high-low arousal, necessary for approach and avoidance/withdrawal behavior) are critical components that make up all affect (Davidson, 2003; Stellar and Stellar, 1985; Schneirla, 1966). Additionally, Russell’s dimensional approach is appealing because it seems likely that events would exist on a continuum for optimal learning; those things with extremely displeasing consequences are so strong that they result in one-trial learning, whereas more neutral consequences require multiple trials (Rozin, 1998). Similarly, those things with high hedonic value (e.g. opiates) are extremely addictive (e,g., Kelley and Berridge, 2002) and also often require only one use to result in repetitive/addictive­ behavior, whereas more neutral experiences (e.g. smoking a cigarette) require more than one instance to result in dependence. With regards to approach/avoidance/withdrawal, there are some things that evolution has taught us to avoid more than others (e.g. saber-toothed tigers vs. calculus).



9.  Affect as appraisal 193

3. Functionalism The functionalist approach (e.g. Darwin, 1872/1965; Dewey, 1894, 1895; Parrott and Schulkin, 1993) focuses on the utility of affect. Our main question, as functionalists, is: how does affect help one to respond adaptively and survive? To have evolved, affect must have performed functions that enhanced our survival and have been adaptive often enough that those possessing them are more successful than those without. In the functionalists’ view, affect is an interpreter for the environment, anticipating future events and outcomes, and solving problems. In essence, affect serves to orient an individual and guide behavior. In our view, affect and emotional experience are of functional importance in that they influence one’s readiness to act, and in that they establish and/or modify one’s relationship with his environment (Frijda, 1986). This is not to say all affect and emotions are adaptive, appropriate or even rational; these perceptions, or appraisals as we think of them, can be fast or slow and accurate or inaccurate. Emotion and affect may be less than perfect, but no less essential than other forms of information processing.

4. Affect, emotion and cognition Some theorists consider cognitive appraisal as an antecedent to an emotional episode, not as part of the emotion itself. However, we suggest that cognitive processes are an integral component of the emotion process; cognition cannot be separated from emotion, emotion cannot be isolated from affect, and thus the cognitive-affective distinction is misguided (see also Scherer, 1993). Like Russell (this book; 2003), we believe that subjective experience of an emotion (what he calls “meta-experience”) involves an interpretation or perception of one’s own experience. We begin with an orientation towards events, a core appraisal perhaps, which is affectively rich, where differentiation and modification is performed through further cognitive interpretations. This interpretation spans several ingredients, as Russell calls them, including attributions, beliefs, desires, plans, and other components, all of which are cognitive. For us, cognition is an aspect of affect, as opposed to an alternative to it. To divorce affect from cognition is to isolate affect from the intelligence and processing that it requires to perform its function: adaptively preparing an individual for action and learning from events (Izard, 2007). Thus, we go through the world with a sort of cognitive coherence, a pre-existing framework that has validity because it has served us well in the past. The wide range of cognitive abilities is intertwined with affect and emotion, not merely interacting with them. They are part and parcel of affective and emotional experience; however, this is not to say

194 Meaghan A. Leddy, Gail C. Robertson and Jay Schulkin

that the cognitive processes are transparent to the individual, or even under one’s own volition (Rozin, 1998; Schulkin, 2004). Zajonc (1980) stated that affect is likely the first reaction one has to a stimulus, and these affective reactions are automatic and guide information processing and decision-making. For example, Winkielman, Zajonc, and Schwarz (1997) demonstrated that those who went through a rapid positive affect induction (1/250 sec.) prior to seeing a stimulus preferred that stimulus over ones preceded by a negative affect induction. Thus it is evident that affect plays a strong role in judgment. In our view, affect and emotion are components of the rough and ready cognitive heuristics that evolved to help guide daily decision-making­ (e.g. Gigerenzer, 2007; Kahneman, Slovic, Tversky, 1982). This is similar to ­Panksepp’s (1998) discussion of the evolution of affect as a heuristic aimed at anticipatory purposes: supporting survival and providing information (Cabanac, 1992; ­Berridge, 1999; Stellar, 1974). The centrifugal organization of the nervous system and the efferent control of sensory processing (Gallistel, 1980) virtually eliminate the possibility of an individual experiencing something as purely sensory, without also being cognitive (Barton, 2006; Cacioppo et al., 2006; Swanson, 2003). Potentiation of sensory inputs by more central sites can alter what is sensed, and can influence the interpretation/appraisal of behavioral responses to the sensations. Neural activity at any level is determined by inputs from lower levels, but also receives feedback from higher levels. For example, there are a large number of afferents and efferents that run between the amygdala and all of the neocortical sensory association areas, some polysensory regions, the hippocampus, the hypothalamus, the thalamus, and the lower, primitive brainstem (Barton, 2006; LeDoux, 2000; Swanson, 2003). A significant amount of affective/information processing goes on in these regions, and it is likely impossible that such interconnected brain regions could result in such segregated and divided phenomena (i.e., cognition and affect). There are affectively rich circuits implicated in various emotional expressions (Calder, Lawrence, & Young, 2001; Lane & Nadel, 1999). Some may be more integrated than others (e.g., fear and anger vs. fear and joy). This is in line with Russell’s two-dimensional model, where fear and joy would be on opposite ends of valence, and thus are more easily distinguished and likely with less overlapping circuitry. Fear and anger are likely more integrated, and this also follows Russell’s model in that both emotions are negatively-valenced, high arousal states. These circuits are likely not segregated, but are shared among emotions with similar affective bases. For example, the amygdala is integral in many types of fear (Kagan and Schulkin, 1995), but may also modulate anger-like behavior that is primarily a function of other circuits (Calder et al., 2001). The amygdala is also importantly linked to a number of functions other than fear, such as social appraisals­, reward,



9.  Affect as appraisal 195

attention, and appetite (Emery et al., 2001; Davis & Whalen, 2001; Schulkin­, 1991), further highlighting the possible overlap of cognitive circuits in the production of emotions. With the emotion of fear, appraisal mechanisms are partially located in the lateral and basal nuclei of the amygdala and output is instigated in the adjacent central nucleus of the amygdala (LeDoux, 1996, 2000; Swanson & Petrovich, 1998). Connections from the appraisal mechanisms project to the output pathways, and as such appraisals are so linked with emotional experience and expression that it would be difficult to encounter an emotion-inducing stimulus without having a tendency to produce an overt response. Further evidence for the importance of cognition in emotion differentiation comes from findings that suggest that cognitive abilities are involved in parsing out different levels of positively-valenced affect (e.g. Davidson et al., 2003). ­Ekman, Davidson and Friesen (1990) found that the smile that occurs while experiencing enjoyment, the Duchenne smile, is associated with greater left-sided anterior temporal and parietal activation when compared with smiles associated with other positive emotions. The increased relative activation in the parietal region was suggested to indicate that more verbal thinking (a cognitive process) occurs with Duchenne smiles (see Davidson et al., 2003 for a review). However, it is incorrect to view emotion differentiation solely as a result of cognitive abilities, as the Schacter-Singer model of emotion suggests (Schacter & Singer, 1962). Cognition is not the only driving force in affect and emotion; cognition will also be influenced by various changes in physiological subsystems. Thus, individuals who may already have a high baseline of anxiety, hyperarousal and hyperalertness, such as those with post-traumatic stress disorder, may be more sensitive to potential threats than the rest of the population (Kagan, 2007; Rosen and Schulkin, 1998). In such a tightly integrated system it seems improbable that affect, emotion and cognition are separate and discrete phenomena. There is more to emotion differentiation than post-arousal cognitive appraisals and attributions, and as we suggested earlier, affect’s anchoring function plays a part. Appraisal occurs in all emotional responses. Appraisals, either non-emotional­ or emotional, can be fast and automatic such as the recognition of syntax, or the recognition of a happy or sad face (Young, 1998). The fact that the appraisal of both types of stimuli can be rapid suggests that the machinations of information processing are evolutionarily prepared to respond to certain survival-relevant stimuli (Rozin, 1998). This is in line with both Panksepp and Russell in that there are more innate features of affect; where we disagree with Panksepp is his argument that these innate components are non-cognitive. Affect and emotion are no less cognitive than other forms of information processing. Cognition should not be viewed narrowly as higher-level information

196 Meaghan A. Leddy, Gail C. Robertson and Jay Schulkin

processing, but as processing that can have multiple levels from basic habituation and sensitization to volitional, conscious decision-making. The viscera and numerous levels of the neural axis are involved in information processing fulfilling different and overlapping functions with regard to affect and emotions (Rosen and Schulkin, 1998). In our minds, an important step in the future discussion of affect and emotion is to elucidate the role of cognition. However, a first step required is to define what one deems “cognition,” as it seems that some limit it to the more phylogenetically recent regions of the brain.

5. Social influences The social milieu influences emotional experience. For example, it has been suggested that the misinterpretation of physiological arousal, such as tachycardia (increased heartbeat), plays a part in the etiology of anxiety conditions. Finn and colleagues (Finn, Sawyer, and Behnke, 2009) found that physiological reactivity accounted for 73.3% of anxious arousal. However, the same physiological arousal (tachycardia) in an individual from Cambodia is interpreted as a loss of energy due to poor sleep or appetite, resulting in an anxious-dysphoric combination (Hinton et al., 2002). Thus, two individuals experiencing the same objective, physiological arousal interpret this affect differently: one as negative-aroused, the other as negative-inhibited. Similarly, European Americans prefer high arousal positive affective states more and low arousal positive affective states less than Hong Kong Chinese individuals (Tsai, Knutson & Fung, 2006). Tsai and colleagues suggest that this is due to the different affective states idealized by their respective religions, texts (e.g. Christianity and Buddhism; see Tsai, Miao and Seppala, 2007), childhood story­books (Tsai, et al., 2007), interpersonal communication, peer interaction, and magazines (Tsai, 2007). A sociological perspective is useful here. Turner and Stets (2006) discuss five sociological theories of human emotion, which may help to shed light on how affect and emotion differ by culture. For example, dramaturgical theories, which highlight self-presentation, cultural scripts and impression management, emphasize that individuals must manage their emotional expressions when cultural structures and scripts create discontinuity between what emotion is experienced and how culture dictates they must present. For instance, Turner and Stets provide the example of an individual who is sad, but must present as happy. In this situation, the incongruity between experience and expression creates another negative emotional experience, in addition to what was already present; the person in the



9.  Affect as appraisal 197

same example may now experience anger at having to appear happy while feeling sad. This situation likely increases the emotional intensity the person experiences, and requires additional inhibition of their true affective and emotional experience. Cultures with high levels of structural differentiation, high rates of mobility across positions and roles, and mediation of social relations through markets are most likely to create such incongruities (Hochschild, 1983; Shweder, Mahapatra and Miller, 1987). This is not to say that there is no universal affect or emotional experience. Much research has demonstrated the universality of facial expressions of emotions (e.g. Ekman and Friesen, 1971). Recently, Sauter and colleagues (2009) extended Ekman’s work on universal emotional facial expressions to emotional vocalizations. They compared British and Namibian participants and demonstrated that some, but not all, emotional, non-verbal vocalizations are recognized across cultures. British participants were able to match all emotional stories with the corresponding emotional vocalization. However, Namibian participants were only able to pair a story with the following correct emotional vocalizations: amusement, anger, disgust, fear, sadness, and surprise; nonetheless, only Namibian participants were unable to perform better than chance when pairing vocalizations with stories depicting achievement/triumph, relief, and sensual pleasure, even though British participants could do this, and these vocalizations have been validated for Western cultures (Sauter and Scott, 2007). The authors conclude that while several negative emotions have evolved universal signals of various modalities, most positive emotions are communicated via culturally-specific indicators. Sociocultural aspects (i.e., norms, structures, culture and socialization) can certainly impact what affect is preferred and sought, the conditions under which affect is altered, as well as how one experiences and expresses emotion. However, it is necessary to determine not only why there are certain similarities and differences across cultures, but why affective and emotional capacities were selected for in the first place.

6. Conclusion Clearly, biological, psychological and sociological research is needed to clarify the true underpinnings and nature of affect, emotion, cognition and their relation to one another. Some remaining questions that may be answered by further research and/or discussion: In what way is affect related to temperament? How might various models explain blends of emotions? For example, an individual may feel both relief and sadness after a loved one dies from a painful battle with cancer, or may feel anger and sadness after a loved one’s murder.

198 Meaghan A. Leddy, Gail C. Robertson and Jay Schulkin

Though we can agree with various points put forth by Panksepp and Russell, one area in which we are steadfast is our belief that cognition is inherent in affect and emotion. The section of our chapter focusing on cognition’s role in affect and emotion serves as a clear focus for our conclusion. As cognitivists, we view appraisals and interpretations as central to affect and emotion. In this way we are dissimilar from Panksepp. Based on our cognitivist position, two questions emerged regarding Panksepp’s chapter and his postulation that affect and emotion are possible in the absence of cognition. Both questions revolved around the theoretical underpinnings of brain research in affect and emotion. First, Panksepp disagrees with the claim “that cognitions and emotions are totally interpenetrant and that meaningful distinctions between the two types of neuro-mental processes cannot be made” because when animals are decorticated, only cognitive, and not affective abilities are impaired, and it is impossible to evoke affect via neocortical stimulation (Panksepp, this book); is it possible then that neocortical stimulation is a poor model or substitute for what cognition is? Panksepp states that if affect is cognitively mediated, fMRI studies should show positive correlations between experienced emotional arousals and activations of higher brain regions, but they fail to do so. Our second question is: is it possible that brain activity is not a valid measure of cognition and its impact on an individual? This means that is it possible that each thought is depicted by the same amount of activation, but have different salience? In other words, isn’t it possible that a thought, “It is very warm,” be demonstrated by the same amount of brain activity as the thought, “I am a failure,” but with very different consequences/ impacts­? Thus, as we mentioned, it seems that an important first step to quelling this debate may be first to concretely define what one considers “cognition.” This discrepancy regarding the definitions of “cognition” brings us back to our initial argument regarding prematurely defining constructs while theory is lacking and facts are inconsistent (e.g. Kagan, 2007). Arguments regarding semantics mask the larger problems of lacking theory and objective facts that could help to clarify the field. It is easy for discussions to become viciously circular, with no clear answer or resolution in sight. We have discussed affect, emotions, and appraisals as adaptive, and serving the function of orienting an individual to his environment. Thus, affect, emotion and cognition are fluid and interconnected with each other; they are not frozen structures. As such, research should focus on the practical consequences and functions of affect and emotion, and should investigate the extent to which our innate and developed appraisal systems are modifiable, and methods by which this may be done.



9.  Affect as appraisal 199

References Barton, R. A. (2006). Primate brain evolution: Integrating comparative neurophysiological and ethological data. Evolutionary Anthropology, 15, 224–236. Berridge, K. C. (1999). Pleasure, pain, desire, and dread: Hidden core processes of emotion. In Well-being: The foundations of hedonic psychology, D. Kahneman, E. Diener & N. Schwarz. (eds.). 525–557. New York: Russell Sage Foundation. Bornstein, M. H., & Arterberry, M. E. (2003). Recognition, discrimination and categorization of smiling by 5-month-old infants. Developmental Science, 6, 585–599. Cabanac, M. (1992). Pleasure: The common currency. Journal of Theoretical Biology, 155, 173–200. Cacioppo, J. T., Hawkley, L., Ernst, J. M., et al. (2006). Loneliness within a nomological net: an evolutionary perspective. Journal of Research in Personality, 40, 1054–1085. Calder, A. J., Lawrence, A. D., & Young, A. W. (2001). Neuropsychology of fear and loathing. Nature Reviews Neuroscience, 2 (5), 352–363. Darwin, C. (1872/1965). The expression of emotions in man and animals. Chicago: University of Chicago Press. Davidson, R. J., Scherer, K. R., & Goldsmith, H. H. (Eds.). (2003). Handbook of affective neuro­ science. New York: Oxford University Press. Davis, M., & Whalen, P. J. (2001). The amygdala: Vigilance and emotion. Molecular Psychiatry, 6 (1), 13–34. Dewey, J. (1894). The theory of emotion: (1) Emotional attitudes. Psychological Review, 1, 568–569. Dewey, J. (1895). The theory of emotion: (2) The significance of emotions. Psychological Review, 2, 13–32. Ekman, P., & Friesen, W. V. (1971). Constants across cultures in the face and emotion. Journal of Personality and Social Psychology, 17, 124–129. Ekman, P., Davidson, R. J., & Friesen, W. V. (1990). Duchenne’s smile: Emotional expression and brain physiology II. Journal of Personality and Social Psychology, 58, 342–353. Emery, N., Capitanio, J., Mason, W., Machado, C., Mendoza, S., & Amaral, D. (2001). The effects of bilateral lesions of the amygdala on dyadic social interactions in rhesus monkeys (Macaca mulatta). Behavioral Neuroscience, 115, 515–544. Feldman Barrett, L., & Kensinger, E. A. (2010). Context is routinely encoded during emotion perception. Psychological Science, 1–5. Finn, A. N., Sawyer, C. R., & Behnke, R. R. (2009). A model of anxious arousal for public speaking. Communication Education, 58, 417–432. Flom, R., & Bahrick, L. E. (2007). The development of infant discrimination in multimodal and unimodal stimulation: The role of intersensory redundancy. Developmental Psychology, 43, 238–252. Fraiberg, S. (1977). Insights from the blind: Comparative studies of blind and sighted infants. New York, NY: Basic Books, Inc. Freud, S. (1924/1960). The ego and the id. New York: Norton. Frijda, N. H. (1986). The emotions. Cambridge, MA: Cambridge University Press. Galati, D., Scherer, K. R., & Ricci-Bitti, P. E. (1997). Voluntary facial expression of emotion: Comparing congenitally blind with normal sighted encoders. Journal of Personality and Social Psychology, 73, 1363–1379.

200 Meaghan A. Leddy, Gail C. Robertson and Jay Schulkin

Gallistel, C. R. (1980). The organisation of action: A new synthesis. Hillsdale, NJ: Erlbaum. Gigerenzer, G. (2007). Gut feelings. New York: Viking Press. Gitelman, D. R., Nobre, A. C., Sonty, S., Parrish, T. B., & Mesulam, M. M. (2005). Language network specializations: an analysis with parallel task designs and functional magnetic resonance imaging. Neuroimage, 26 (4), 975–985. Havas, D. A., Glenberg, A. M., & Rinck, M. (2007). Emotion simulation during language comprehension. Psychonomic Bulletin & Review, 14 (3), 436–441. Hebb, D. O. (1946). The nature of fear. Psychological Review, 53, 259–276. Hinton, D., Hinton, S., Um, K., Chea, A., & Sak, S. (2002). The Khmer ‘Weak Heart’ Syndrome: Death from palpitations. Transcultural Psychiatry, 39, 323–344. Hochschild, A. R. (1983). The managed heart: Commercialization of human feeling. Berkeley, CA: University of California Press. Izard, C. E. (2007). Basic emotions, natural kinds, emotion schemas, and a new paradigm. Per­ spectives on Psychological Science, 2(3), 260–280. James, W. (1890/1950). The principles of psychology. Mineola, New York: Dover Publications. Kagan, J. (1997). Temperament and the reactions to unfamiliarity. Child Development, 68 (1), 139–143. Kagan, J. (2007). What is emotion? History, measures, and meanings. New Haven, CT: Yale University Press. Kagan, J., & Schulkin, J. (1995). On the concepts of fear. Harvard Review of Psychiatry, 3 (4), 231–234. Kahneman, D., Slovic, P., & Tversky, A. (Eds.). (1982). Judgment under uncertainty: Heuristics and biases. New York: Cambridge University Press. Kelley, A. E., & Berridge, K. C. (2002). The neuroscience of natural rewards: relevance to addictive drugs. The Journal of Neuroscience, 22, 3306–3311. Lane, R. D., & Nadel, L. (1999). Cognitive neuroscience of emotion. New York: Oxford University Press. LeDoux, J. E. (2000). Emotion circuits in the brain. Annual Review of Neuroscience, 23, 155–184. LeDoux, J. E. (1996). The emotional brain. New York, NY: Simon & Schuster. Panksepp, J. (1998). Affective neuroscience: The foundations of human and animal emotions. Oxford: Oxford University Press. Parrott, W. G., & Schulkin, J. (1993). Psychophysiology and the cognitive nature of the emotions. Cognition and Emotion, 7, 43–59. Plato. 360 B. C. E. (1956). Symposium. (B. Jowett, Trans.) London, England: Prentice Hall. Richter, C. P. (1976). The psychobiology of Curt Richter (E. M. Blass, Ed.). Baltimore: York Press. Roch-Levecq, A. C. (2006). Production of basic emotions by children with congenital blindness: Evidence for the embodiment of theory of mind. British Journal of Developmental Psychology, 24, 507–528. Rosen, J. B., & Schulkin, J. (1998). From normal fear to pathological anxiety. Psychological Re­ view, 105 (2), 325–350. Rozin, P. (1998). Evolution and development of brains and cultures: some basic principles and interactions. In M. S. Gazzaniga & J. S. Altman. (Eds.). Brain and mind: Evolutionary per­ spectives. Strassbourg: Human Frontiers Science Program. Russell, J. (2003). Core affect and the psychological construction of emotion. Psychological Re­ view, 110, 145–172. Sauter, D., Eisner, F., Ekman, P., & Scott, S. K. (2009). Universal vocal signals of emotion. In N. Taatgen, & H. Van Rijn. (Ed.). Proceedings of the 31st Annual Meeting of the Cognitive Science Society, 2251–2255. Cognitive Science Society.



9.  Affect as appraisal 201

Sartre, J. P. (1943/1973). Being and nothingness. New York: Washington Square Press. Schacter, S., & Singer, J. E. (1962). Cognitive, social, and physiological determinants of emotional state. Psychological Review, 69 (5), 379–399. Sauter, D. A., Calder, A. J., Eisner, F., & Scott, S. K. (2009). More than one kind of happiness: can we recognize vocal expressions of different positive states? Motivation and Emotion, 31, 192–199. Sauter, D. S., & Scott, S. K. (2007). More than one kind of happiness: can we recognize vocal expressions of different positive states? Motivation and Emotion, 31, 192–199. Scherer, K. R. (1993). Studying the emotion-antecedent appraisal process: an expert system approach. Cognition and Emotion, 7, 325–355. Schneirla, T. C. (1966). Behavioral development and comparative psychology. Quarterly Review of Biology, 41, 283–303. Schulkin, J. (2004). Bodily sensibility: Intelligent action. New York, NY: Oxford University Press. Schulkin, J. (1991). Sodium hunger: the search for a salty taste. New York: Cambridge University Press. Shweder, R. A., Mahapatra, M., & Miller, J. G. (1987). Culture theory: Essay on mind, self and emotion. Cambridge, UK: Cambridge University Press. Spinoza, B. (1668/1955). On the improvement of the understanding. New York: Dover. Stellar, E. (1974). Brain mechanisms in hunger and other hedonic experiences. Proceedings of the American Philosophical Society, 118, 276–282. Stellar, E., & Stellar, E. (1985). The neurobiology of reward. Berlin: Springer-Verlag. Swanson, L. W. (2003). Brain architecture: Understanding the basic plan. Oxford: Oxford Unviersity Press. Swanson, L. W., & Petrovich, G. D. (1998). What is the amygdala? Trends in Neurosciences, 21 (8), 323–331. Tsai, J. L. (2007). Ideal affect: cultural causes and behavioral consequences. Perspectives on Psy­ chological Science, 2, 242–259. Tsai, J. L., Knutson, B., & Fung, H. H. (2006). Cultural variation in affect valuation. Journal of Personality and Social Psychology, 90, 288–300. Tsai, J. L., Louie, J., Chen, E. E., & Uchida, Y. (2007). Learning what feelings to desire: Socialization of ideal affect through children’s storybooks. Personality and Social Psychology Bulletin, 33, 17–30. Tsai, J. L., Miao, F. F., & Seppala, E. (2007). Good feelings in Christianity and Buddhism: Religious differences in ideal affect. Personality and Social Psychology Bulletin, 33 (3), 409–421. Turner, J. H., & Stets, J. E. (2006). Sociological theories of human emotions. Annual Review of Sociology, 32, 35–52. Wager, T. D., Feldman Barrett, L., Bliss-Moreau, E., Lindquist, A. K., Duncan, S., Kober, H., et al. (2008). The neuroimaging of emotion. In M. Lewis, J. M. Haviland-Jones, & L. Feldman Barrett. (Eds.). The Handbook on Emotion, 249–267. New York: The Guilford Press. Winkielman, P., Zajonc, R. B., & Schwarz N. (1997). Subliminal affective priming resists attributional interventions. Cognition & Emotion, 11, 433–465. Young, A. W. (1998). Face and mind. Oxford: Oxford University Press. Zajonc, R. B. (1980). Feeling and thinking: Preferences need no inferences. American Psycholo­ gist, 35, 151–175.

chapter 10

What should theories of emotion be about? James R. Averill University of Massachusetts, Amherst

In their chapters in this volume, both Russell and Panksepp are frustratingly reasonable. Over the years, they have heard and are now prepared to meet almost any argument that might be raised against their positions; and when not prepared with a counterargument, they simply rule the objection off limits. Thus, at the beginning of Russell’s article, he readily admits that “biological and social-cultural­ levels of analysis remain to be added.” Panksepp makes similar claims with respect to psychological and social-cultural levels of analysis. It is hardly fair to criticize persons for not doing what they make no claims of doing. Panksepp (1998, this volume) makes a threefold distinction between what he calls primary-process emotional circuits (to which he believes the notion of “basic emotions” should be reserved), secondary-process emotions, and ­tertiary-process­ emotions. Primary-process emotions include fear, panic/ grief, rage, lust, care, play, seeking, and possibly others yet to be delineated. They are hard-wired into the nervous system at the subcortical level, and they have homologues in all mammalian species. Secondary-process emotions are elaborations on the primary, based on simple associative learning, e.g., classical and instrumental conditioning. They also can be found throughout the mammalian world. Tertiary-process emotions are human phenomena, shaped by complex cognitive (neocortical) and social influences, with language being an important mediator. Russell indicates that the differences between him and Panksepp are “difficult to pinpoint.” I agree. Their positions are more similar than different. Russell’s Core Affects correspond roughly to Panksepp’s primary-process emotions (e.g., both kinds of experience are noncognitive and phylogenetically primitive), and his Meta-emotional experience corresponds to Panksepp’s tertiary-process emotions. The major difference is that Panksepp’s “bottom-up” neurophysiological approach leads to the identification of discrete neuro-emotional systems, whereas Russell’s “top-down” psychological approach leads to a dimensional organization

204 James R. Averill

of Core Affects. That is an important difference, certainly, but as Panksepp indicates, he would have little difficulty accommodating Russell’s dimensions, provided relevant neurophysiological substrates can be identified; conversely, Russell indicates that he can locate Panksepp’s primary-process emotions within the two dimensional space formed by pleasantness and arousal. But even without such accommodations, the two approaches are not necessarily contradictory. Categorical and dimensional schemes can be found in many areas: they serve different purposes, and the strengths of neither need be sacrificed in a rush at synthesis. Warren Austin, the first US Ambassador to United Nations, is supposed to have remarked that “Jews and Arabs should sit down and settle their differences like good Christians.” It was an attempt at humor with a serious point. The ­IsraeliArab dispute is not simply a disagreement over land. It is fundamentally a religious dispute that cannot be settled by an outsider, especially one from a different religious perspective. In a similar vein, I want to say to Panksepp and Russell, Why don’t you sit down and settle your differences like good empiricists? But that would be to misunderstand the nature of their disagreements, which are more theoretical than empirical. Theoretical allegiances have some of the same qualities as religious commitments. I will therefore not even try to reconcile the positions of Panksepp and Russell, especially since my own commitment is toward meta (tertiary) emotional experiences. I will proceed as follows: first, I will offer a few thoughts on reading the chapters by Panksepp and Russell; I will then offer a defense of meta (tertiary) emotional experiences, and why they should be taken seriously by any scientific (not just humanistic) theory of emotion.

1. Some thoughts on reading Panksepp Failure to distinguish among levels of analysis, Panksepp contends, is “the ultimate source of many perennial, seemingly unresolvable, debates in the psychology and philosophy of emotion.” This contention, I believe, attributes greater obtuseness on the part of many psychologists and philosophers of emotion than is warranted. For example, would any of the participants in the seemingly unresolvable debates change his or her position having once read Panksepp? And if not, why not? At best, Panksepp’s diagnosis is only partially correct. More important than a failure to distinguish among levels of analysis, I believe, is the tendency of authors to assume that their preferred level is fundamental, and hence to claim hegemony over the others. In spite of claims to the contrary, Panksepp seems to do just that, as when he claims that “emotional theories that fail to deal with the evidence for their [primary-process emotions] existence may be deemed to be profoundly incomplete – perhaps premature, short sighted or outdated from a neuroscience



10.  What should theories of emotion be about? 205

perspective.” That may be true from a neuroscience perspective. But what about other perspectives? Panksepp also notes that most critics do not dispute “the validity of [his] animal findings, just their generalizability.” I am one of those critics, at least with regard to emotions as experienced in everyday affairs. For example, how often is the rage circuit activated in ordinary episodes of anger, which typically do not involve aggression, especially not physical aggression (Averill, 1982)? I can not imagine how one would even address this question empirically, given the practical and ethical constraints on doing the relevant neurophysiological research with humans. There is, however, more than one way to generalize. Rather than clarifying typical emotional experiences, Panksepp’s findings may help explain – and perhaps ultimately lead to better treatment of – some emotional disorders. He notes, for example, that activation of the fear circuit simpliciter may result in free-floating anxiety. Panic attacks, which are often confused with anxiety attacks, he further speculates, are a form of separation distress that result from activation of the panic/grief system. Analogous observations may apply to the other neural circuits identified by Panksepp. For example, sudden aggressive outbursts in young children may result from uninhibited activation of the rage circuit. Depending on the history of the child and other presenting symptoms, the condition may be diagnosed as AD/HD (attention deficit/hyperactivity disorder), bipolar (manic/depressive) disorder, or autism, among other possibilities. Or consider the case sometimes referred to as “the cable car named desire.” In 1964, Gloria Sykes, a 24 year-old devout Lutheran from Michigan, moved to San Francisco. One day as she was riding a cable car, a failure of the mechanical gripping system caused the car to careen down a hill. Sykes hit her head on a pole and received what at first appeared to be minor injuries. Sykes claimed, however, that the accident triggered an uncontrollable desire for sex. Within a short time after the accident, she had sex with 100 men, 50 within one week. She sued the city for $500,000. Many people considered the suit to be an abuse of the legal system, but her lawyer successfully argued that the accident caused her to suffer from an unusual form of post-traumatic stress disorder. Eventually, Sykes was awarded $50,000 in damages. Brain injuries can be subtle and difficult to detect with a typical neurological exam. Rather than PTSD, it is plausible that Sykes’ inordinate sexual urges stemmed from chronic activation of her lust circuit, or a disruption of normal inhibitory mechanisms. Similar observations could perhaps be made with respect to Panksepp’s care, play, and seeking systems, although in these instances the “disorder” might be more subtle and less likely to come to the attention of medical or legal authorities.

206 James R. Averill

2. Some thoughts on reading Russell Classification is propaedeutic to any science, for it helps delineate what the ­science is about. Not surprisingly, therefore, some sciences have subdisciplines devoted specifically to classification, such as nosology in medicine and taxonomy in ­zoology. Psychology has no equivalent subdiscipline, at least for normal behavior. One reason for this lack interest, I believe, is simple boredom. William James (1890) remarked that he “would as lief read verbal descriptions of the shapes of the rocks on a New Hampshire farm” as read again classic works describing emotional experiences (p. 448). We can sympathize with James; however, his dismissal was, I believe, premature. No science can proceed without some delineation of its subject matter. Classification also raises important philosophical issues, such as realism, about which I will have more to say later (see Zachar, 2001, 2006). As a first approximation, most emotion theorists distinguish two broad classes of affective phenomena: (a) background moods or experiences, as when we wake in the morning feeling happy and energetic or depressed and lethargic; and (b) discrete emotions such as anger, fear, grief, and love, to mention but a few. Each class is heterogenous within itself, and the dividing line between the two is vague. Nevertheless, it is an important distinction. Panksepp’s primary-process emotions and Russell’s Core Affects fall within the first class; whereas their tertiary-process emotions and meta-emotional experiences fall within the second class. Core Affects. Russell considers Core Affects to be preconceptual (and hence relevant to the experiences of infants and infrahuman animals), continuous (we are always in one state of Core Affect or another), and free floating (without an object). It follows logically that Core Affects cannot be described in the same terms as discrete emotions, which are subsumed under concepts, such as “anger,” “fear,” etc.; are episodic, rather than continuous; and typically take objects, that is, are not free floating. How, then, are we to conceive of that which is preconceptual? By ordering them within a two-dimensional space, with pleasant-unpleasant and aroused-unaroused representing the two axes. Russell acknowledges that other theorists believe that both pleasantness and arousal may be broken down into qualitatively different types. But he rejects such a “multiplicity view.” I don’t find his rejection convincing. A brief historical excursion may illustrate one reason for my reservations. Russell notes that his Core Affects bear some resemblance to Wundt’s (1897) tridimensional theory of emotion. Wundt was the leading academic psychologist of his day and is generally credited with founding the first experimental laboratory for psychological research. He suggested that feelings (the presumed building blocks of more complex emotions) could be ordered within a dimensional



10.  What should theories of emotion be about? 207

space formed by three bipolar factors: pleasantness-unpleasantness (Lust-Unlust), excitement-inhibition (Erregung-Hemmung), and tension-relaxation (SpannungLösung). Introspective reports under controlled laboratory conditions soon suggested that the dimensions of excitement-inhibition and tension-relaxation could be eliminated; careful introspection by trained observers indicated that these two dimensions were often confused with feelings of pleasantness-unpleasantness; were associated with other sensory attributes through past experience and hence were not independent; or could be reduced to bodily sensations and hence were not feelings as a separate category of subjective experience (Titchener, 1908). Only pleasantness-unpleasantness seemed to remain as an unalloyed dimension of feelings. But additional research suggested that it, too, might be reducible to organic sensations (Nafe, 1924). As a unique experience, pleasantness appeared less like a self-evident “raw” feel and more like a Cheshire cat; a product of the imagination that disappears on close introspection. The conclusion I draw from these early introspective studies is that pleasantness is not the unitary experience that Russell presumes. Of course, it may be that all pleasant experiences have common neurophysiological elements. Even if that is true (a possibility that Panksepp seems to doubt), it would not mean that pleasantness is a unitary experience. To draw an analogy, many diseases include an increase in body temperature, and hence involve some of the same physiological mechanisms. This should not, however, obscure important differences among otherwise distinct conditions. I do not mean to imply that a dimensional approach is without utility. One of the most widely used assessment techniques in psychology, the Semantic Differential, involves rating scales for the Evaluative (e.g., “good-bad”) and Arousal (e.g., “active-passive”) dimensions discussed by Russell, along with a Potency (e.g., “strong-weak”) dimension (Osgood, May, & Miron, 1975). Perhaps a dimensional approach is the most practical way to measure the kind of preconceptual states that Russell calls Core Affects. However, we should not confuse a method of measurement with that which is being measured (operationism notwithstanding). Let me turn now to Russell’s approach to discrete emotions. Psychological constructionism. Constructionist approaches to emotion come in biological, psychological, and social varieties, although the term is often restricted to the latter two. All varieties reject essentialism, that is, they assume that the objects of enquiry comprise components, no one of which is necessary or sufficient for the whole. It follows that any constructionist position must (a) account for the way the semi-independent components cohere to form a recognizable pattern or syndrome, and (b) explain the origin and function of that syndrome within a broader context (e.g., survival of the species, individual, or society).

208 James R. Averill

Russell labels his position “psychological constructionism.” I should state up front that my own position is “social constructionism.” My views are thus closer to Russell’s than to Panksepp’s. The major difference lies in where we look for the source of patterning among the components of an emotion, and the origin and function of emotional syndromes. As is fitting with his emphasis on psychological construction, Russell looks to features within the individual and the immediate situation that might create correlations among component reactions. Yet, he cannot ignore social factors entirely. Thus, he notes that a pattern of responses may be categorized as a particular emotion when it resembles a prototype or script for that emotion. This raises an obvious question: What is the source of the prototype or script that people use to categorize (and, indeed, to help construct) discrete emotions? People are not free to invent their own emotions any more than they are free to invent their own language – not if they wish to be understood. From a socialconstructionist perspective, widely shared and named emotions are organized by social beliefs and rules (norms), and the emotions serve social functions, sometimes even at the expense of the individual. Grief is a good example (e.g., Averill, 1979; Averill, & Nunley, 1988). This does not mean that social and psychological construction are independent processes (or that either is independent of biological processes). The social norms of emotion allow a good deal of improvisation. What societies construct, individuals can reconstruct. Take romantic love. During the long course of Western history, conceptions of love – and associated behaviors – have undergone significant transformations. It is difficult to attribute any transformation to a single person, as opposed to the accumulation of minor changes by many individuals over extended periods of time. Nevertheless, the poet Dante (1265–1321) played a seminal role in the development of romantic love as it has passed along to us, with many modifications along the way (Averill, 1985; Oatley, 2007). Perhaps the last major reconstruction of romantic love occurred during the “sexual revolution” of the 1960s. The current debate over gay marriage may indicate yet further transformation, or at least expansion of the concept. However that may be with respect to love, when a reconstruction proves beneficial for the individual or group, we may speak of emotional creativity (Averill, 2005); when maladaptive, of neurosis (Averill & Nunley, 2010). I will not pursue further the difference between psychological and social con­structionism, for as the above observations suggest, each presumes the other­. Rather, I want to explore a kind of dismissive attitude that both Russell and ­Panksepp seem to share with regard to everyday emotional experiences.



10.  What should theories of emotion be about? 209

3. In defense of meta (tertiary) emotional experiences In trying to gain a broader perspective on the chapters by Panksepp and Russell, I have read between their lines. There, I sense a dismissive attitude toward everyday (meta or tertiary) emotional experiences. I may, of course, be attributing to them thoughts they do not hold. Yet, even if not held by Panksepp and Russell, I believe a dismissive attitude is sufficiently common among emotion researchers to deserve extended comment. (Recall the earlier quote by James comparing “classic” works on emotion to descriptions of rocks on a New Hampshire farm.) Both Panksepp and Russell express skepticism about the value of ordinary language for a science of emotion. I agree – but only to a point. Terms from ordinary language, and the concepts they index, refer to molar behavior occurring within a context; they cannot be used to refer to underlying processes, whether psychological or physiological. A neural circuit, for example, cannot be characterized as angry or fearful, any more than a molecule can be characterized as red or yellow. Recognizing this fact, Panksepp uses capital letters when referring to his primary-process circuits, thus distinguishing his use of a term (e.g., fear) from its lower-case counterpart in ordinary language. This is a good first step, but it is only a way station along the route we must ultimately travel. In the final analysis, neural circuits must be described in neurological terms, not quasi-psychological terms. When we reach that point – admittedly, still a long way off – how will we deal with anger, fear, love, and the myriad of other emotions recognized in ordinary language (not just English, but in everyday speech, whatever the language)? In an earlier essay, Russell (1998) drew comparisons between astronomy and astrology, on the one hand, and scientific and folk psychologies of emotion, on the other. “An astronomer,” he writes, “studies the stars, an anthropologist studies different people’s beliefs about stars, such as a belief in astrology” (p. 423). The folk-psychology of emotion is akin to astrology, Russell implies, whereas the scientific study of emotional episodes is presumably more akin to astronomy. Considering how different patterns of response get categorized as one kind of emotion or another (as anger, say, or fear), Russell (2008) compares the task to the grouping of stars into constellations. “No grouping of stars into constellations proved useful in advancing astronomy, to the development of a deep scientific understanding, for the simple reason that constellations are not causal entities” (p. 424–425, emphasis added). Panksepp (this volume) makes a similar point. He acknowledges that a thorough understanding of emotions must involve psychological and sociocultural as well as neurophysiological levels of analysis, but he leaves no doubt at which level the field ultimately needs to be grounded. “Affect studies,” he states, “lie at the very fulcrum of the long-standing debate between objective-scientific and

210 James R. Averill

i­ ntrospective-humanistic understanding of our nature as living, feeling organisms.” It is a revealing contrast. Panksepp does not disparage humanistic approaches to understanding tertiary-level emotions; he simply implies that such approaches cannot be “objective-scientific.” By contrast, “at lower MindBrain levels we are bound to be closer to the nomothetic (universal) processes laid down early in mammalian evolution.” In short, does the study of everyday emotions properly have a place in science, or does it belong in the arts and humanities, as implied by Panksepp, or cultural anthropology, as suggested by Russell? This, it seems to me, is the central issue raised by both Panksepp and Russell. Let me begin with Panksepp’s assertion that lower MindBrain processes “are closer to the nomothetic (universal) processes laid down early in mammalian evolution.” The Oxford English Dictionary (online edition) defines nomothetic as “relating to or concerned with the study or discovery of the general laws underlying something.” As Panksepp implies, scientific laws are general within their range of applicability – for example, all mammalian species. One way to achieve such generality is for each uneliminable term in a law to be defined “classically,” that is, in terms of necessary and sufficient conditions (Hull, 1976). (Another way is for each term to be defined so vaguely that it may be stretched to cover almost any contingency – but, then, the “law” would not be nomothetic in any meaningful sense.) Mathematical entities are prime examples of classically defined concepts. Consider what it means to be a square. Any four-sided figure with equal sides and internal angles of 90 degrees each is a square, and any figure without one or another of these characteristic is not a square. There are no in-between cases. Classically defined entities have traditionally been referred to as “natural kinds,” in spite of the fact that they seldom exist in nature except under ideal conditions (e.g., a perfect vacuum in the case of mass and uniform motion). The important point is that they are invariant over time and place. Tertiary-process emotions are not invariant: they vary from one culture to another, from one individual to another, and from one occasion to another. It follows that tertiary-process emotions cannot be nomothetic in Panksepp’s sense. This does not mean, however, that everyday emotions have no proper scientific role to play. On the contrary, I would like to suggest that any theory that neglects tertiary-process emotions is bound to be (in words that Panksepp used in a different context) “profoundly incomplete – perhaps premature, short sighted or outdated.” Two historical analogies will illustrate the role that tertiary-process emotions might play in a science of emotion. The first analogy involves some parallels between conceptions of emotions and diseases; the second, between emotions and biological species.



10.  What should theories of emotion be about? 211

Analogy 1: Emotions and diseases In the nineteenth century, some “radical” physiologists (Hughlings Jackson, among others) proposed that disease concepts could be eliminated from medicine (Temkin, 1968). They noted that the same physiological processes are involved in both health and disease. A “mumps process,” say, need not be invoked to explain the symptoms of that condition. When a person falls ill, ordinary physiological processes become exaggerated or disorganized in ways that interfere with other vital processes. If the interference is sufficiently common or important to be recognized by society, the condition may be identified as a disease. There is nothing more to it. The reasoning here is correct, but the conclusion is, I believe, a reductio ad absurdum, as Hughlings Jackson himself soon recognized. Neither “disease” as a generic category nor specific diseases such as mumps, measles, and cancer, have special status as physiological constructs. Nevertheless, it is difficult to imagine the progress of medicine without adequate diagnostic criteria for various disease syndromes. In concluding his review of the history of classification in medicine, Temkin (1968) observed that: Only when the medical categories of health and disease and the conceptual tools for grasping them, viz. diagnosis, prognosis, indication, and therapy are brought to bear upon a science does it acquire the nature of a medical science. (p. 19)

What are the parallels with the study of emotions? Duffy (1934) made an argument with respect to emotions similar to that made by Jackson and his colleagues with respect to diseases. So, too, does Russell. Following suggestions by William James, Russell notes how different aspects of an emotional syndrome – perception, physiological change, expressive reactions, etc. – can be peeled away and turned over to other aspects of psychology. After such a peeling-away process is complete, nothing of the emotion would be left that a theory of emotion would uniquely be about. Again, I believe the reasoning to be sound, but we must be cautious about the conclusions we draw. The study of emotion may involve the whole of psychology, as Russell suggests, but it does not thereby cease to be a science of emotion. Scientific theories do not exist divorced from reality; they are about something. And what they are about determines the kind of theory they are. To paraphrase Temkin, quoted earlier, only when the conceptual tools afforded by science (including, but not only, neurophysiology and psychology) are brought to bear on everyday emotional experiences will there be a science of emotion.

212 James R. Averill

Analogy 2: Emotions and species Most people agree that the theory of evolution, originated by Darwin, is among the most important in science. And although the theory concerns the origin of species, no biological species enters the theory as a theoretical entity. For example, there is no law describing evolutionary processes that includes an uneliminable reference to dodo birds, or to any other extinct or extant species. This is because, according to Hull (1976), if individual species “are spatiotemporally localized and laws of nature must be spatiotemporally unrestricted, then it follows that no law of nature can make uneliminable reference to an individual [species]” (p.189). Hull goes on to note that: From Darwin and Wallace to Levins and Lewontin, the laws which have been proposed for the evolutionary processes have been couched in completely general terms. They do not concern particular genes, gametes, organisms, colonies, populations, or species but theoretically significant kinds; e.g., dominant genes, (p. 189) organisms which reproduce sexually, and founder populations. 

Prior to Darwin, species were generally viewed as fixed categories in nature. A species, it was believed, could be defined in terms of essential features, and an organism either was or was not a member of a particular species. As long as species were conceived of as natural kinds, the evolution of species was inconceivable. One of the major innovations of Darwin was to offer an alternative conception, namely, species as sets of interacting (interbreeding) individuals (Mayer, 1972). Since individuals members of a species can vary one from another, species – being nothing more than aggregates of individuals – can evolve depending on which individuals reproduce (assuming some degree of heritability). Darwin continued to study finches, barnacles, and humans even after he began to doubt their status as “natural kinds.” As discussed earlier, something similar occurred in medicine. Even after some physiologists came to doubt the theoretical utility of disease concepts, they continued to talk about and study measles, mumps, tuberculosis, and the like. One reason that concepts like “species” and “disease” can not be eliminated is practical: we need such mid-level concepts to communicate efficiently, and to organize research and treatment programs. A more important reason is theoretical: Many of the questions that we want to ask are geared to this level of analysis; put differently, the original referents for our theories – what we want to explain – are species and diseases, and these remain the referents even as understanding of underlying processes advances. Similar considerations apply, I believe, with respect to tertiary (meta) emotions: Despite doubts about emotions as theoretical entities, everyday episodes of fear, anger, sadness, etc., remain what our theories of emotion are about.



10.  What should theories of emotion be about? 213

None of this means, of course, that emotions themselves do not change as circumstances change. On the contrary, as we saw in the case of romantic love, ostensibly the same emotion can change dramatically over time. Like species, emotions arise, evolve, and sometimes become extinct. An example of the last is accidie, a kind of religious lethargy that plagued monks during the middle ages, but that disappeared with the rise of the reformation (Harré, 1986). It is not just individual emotions that undergo change. The entire category of emotion has also undergone changes during the course of Western history (Averill & Sundararjan, 2006). As noted earlier, all constructionist approaches (biological, psychological, and social) reject essentialism and thereby open the possibilities for change. On that point, I agree with Russell when he speculates that advances in science may alter the way we conceive of emotions. But let us take a lesson from Darwin: If he had not accepted species as entities worthy of study, he would not have developed a theory to explain their origins by natural selection. Analogously, if we do not take everyday emotions seriously, we will never develop a theory to explain their origins and functions – and to further the possibilities for their change.

4. Emotional causality and realism Implicit in the discussion thus far are two assumptions that deserve further consideration, namely, meta (tertiary) emotions can be causes and they are real. These two assumptions go to the very heart of a science of emotion. Zachar (2006) has argued that everyday emotions can “be called ‘real’ with respect to their causal role in interpersonal interactions” (p. 130). I agree, in part, I must admit, out of self-interest. Much of my own research has been devoted to the analysis of ­everyday emotional experiences, including anger, grief, romantic love, fear, hope, solitude, happiness and aesthetic and spiritual experiences. If such emotions are not real, their analysis amounts to little more than fiction. As Panksepp might put it, such analyses would belong more in the arts than science. Fortunately, logic more than self-interest leads me to believe that everyday emotions can be both causal and real.

Emotions and causality In several places in his chapter, Panksepp reiterates that only a neurophysiological approach can lead to the kind of “causal understanding” that is truly explanatory. Like Panksepp, Russell emphasizes the importance of causal understanding, but unlike Panksepp, he does not limit causes to the neurophysiological level of

214 James R. Averill

analysis. Any component of an emotional syndrome can be analyzed in terms of its own causal chain (antecedents and consequences). What is not scientifically reasonable, according to Russell, is to treat meta-emotions themselves as causes. “My way,” he asserts, “stems from my not using everyday emotion terms as scientific terms, but simply as folk concepts, and from not thinking of emotions as causal entities.” The type of cause implicit in the analyses of both Panksepp and Russell is traditionally known as an efficient cause, the paradigm case of which is one billiard ball hitting another. Emotions are not like billiard balls; on that we can all agree. But might there not be other ways of thinking of emotions as causal entities? Think of gravity. In what sense does gravity “cause” an apple to fall, the tides to rise, and the planets to orbit the sun? Classical (Newtonian) mechanics was never as mechanistic as is often portrayed; and contemporary (quantum) mechanics is even less so. Perhaps it would be best to do away with causal talk altogether; that has been tried, but without notable success. Evidently, the notion of cause serves too many purposes to be lightly abandoned. A better strategy, therefore, might be to expand the notion of cause so that it includes more than efficient causes. Panksepp himself hints at such an expansion when he suggests that primary-process emotions might resemble “attractor landscapes” in chaos theory. I have a less esoteric suggestion, namely, treat emotions as dispositional causes. Let me begin with an everyday example from Ryle (1949): Why did the glass break? Because it was hit by a stone (an efficient cause) and because it was brittle (a dispositional cause). In this example, only the efficient cause is of interest, since the brittleness of glass is already known. So, let me take another example. Why did the airplane crash? Because wind buffeted its wings (an efficient cause) and because the metal of a wing was fatigued (a dispositional cause). In this example, the efficient cause is of little interest, for being buffeted by wind is normal for an airplane in flight. On the other hand, the metal of the wing should not have been fatigued (a condition not unlike the brittleness of glass). Of course, we seldom stop with a dispositional explanation. Why had the metal become fatigued? But the same is true of efficient-cause explanations. Why did the stone hit the glass? The point at which we stop asking the question, Why?, has less to do with the nature of the cause as with other concerns: how to prevent or encourage recurrence or, more abstractly, how to make the unintelligible intelligible. Following Ryle (1949), I have argued that emotional states can best be conceived of as episodic (relatively short-term, reversible) dispositions (e.g., ­Averill, 1991). In making that case, I relied on two features of everyday emotional episodes. First, most episodes outlast any given response (e.g., anger can last for hours and grief for months, whereas facial expressions and physiological reactions, say, may



10.  What should theories of emotion be about? 215

last for only minutes, if that). Second, at any moment in time during an episode a person may respond in a variety of different ways, or not respond at all, depending on situation and his or her temperament. I would now add a third reason, namely, as episodic dispositions, emotions can count as causes. An analysis of emotions as episodic dispositions is, however, incomplete. As Russell points out, the attribution of an emotional state depends, not on the presence of a particular response, but on the extent to which the pattern of responses matches more or less the cultural prototype for an emotion. But pattern-matching­ is not, by itself, sufficient to distinguish a “real” emotion, say, from its feigned counterpart. More important than the pattern of responses is social legitimation. To take a nonemotional example, no matter how effective a person might be at practicing law, he or she will not be considered a lawyer without the proper credentials. It might seem odd to speak of emotional performance as being “credentialed,” but it is quite common to question the genuineness of a person’s emotional reactions on personal or situational grounds, regardless of how well the reaction fits a prototype. In short, emotional concepts have an irreducible social component. Reflecting this fact, I have suggested that emotions can also be conceived as transitory or transitional social roles (1980, 1991). Emotions typically occur when a person’s normal (rational, deliberate) coping resources are taxed, sometimes in a positive direction, but more commonly in a negative way . If the taxing situation is common enough, or serious enough, society provides bridging mechanisms (social roles) that allow adjustments to be made. Emotional roles are transitory in the sense that they are typically enacted for only short periods of time; they are transitional in that they allow a transition from one normal (premeditated, deliberate) state of affairs to another. The two conceptions – emotions as episodic dispositions and as transitional social roles – are complementary. When in an emotional state, a person is disposed to act in ways consistent with the relevant role. However, the locus of causality differs in each case. The first (emotion as an episodic disposition) focuses on the individual; the second (emotion as a transitional social role) focuses on society. Two examples may help to illustrate the importance of a social level of analysis. Briggs (1970) maintains that the Utku, members of an Eskimo tribe that live along the Canadian arctic shore, do not have a concept for anger, nor do they express an emotion equivalent to anger. This is because, Briggs suggests, the Utku exist in a precarious environment that demands close cooperation. Anger might be a disruptive influence, and hence a threat to survival. Critics have suggested that the Utku do – or must – experience anger, but that they suppress the emotion for the good of the group. To a certain extent, this is a matter of definition. The Utku, like people everywhere, must struggle to

216 James R. Averill

overcome obstacles, and they occasionally show “childish” bad temper. But anger is not any kind of struggle, whether good or bad tempered. It is a mode of coping with interpersonal conflict that has been insititutionalized as a transitional social role in Western cultures, with its own entry requirements, privileges, obligations, and restrictions (Averill, 1982; in press). As a second example, consider liget, an emotional syndrome considered basic by the Ilongot, a head-hunting people who live in Northern Luzon, the Philippines. Liget shares some features in common with anger, such as the struggle to overcome obstacles. But it is a very different emotion. One manifestation of liget is the taking of a head, a feat that is an occasion for celebration, for it helps establish a young man’s standing within the community (Rosaldo, 1980). Returning for a moment to the Utku, to say that they experience anger but fail to recognize the fact is, I believe, equivalent to saying that Westerners experience liget but fail to recognize the fact. A more reasonable interpretation is that anger and liget have been institutionalized within their respective cultures as distinct emotional roles, whereas no equivalent emotion has been institutionalized among the Utku. Even during times when beheadings have been common in the West, as during the French Revolution (when, during one particularly bloody 47-day period, 1,376 individuals were guillotined), the associated celebratory emotion was not liget. To conclude, whether conceived of as episodic dispositions (an individual variable) or as transitional social roles (a social variable), everyday emotions can be treated as causes in their own right. By implication, tertiary (meta) emotions are also real. But in what sense real?

Realism and tertiary (meta) emotional experiences I raise the issue of realism with some misgivings: It stretches my mind, especially as it concerns physical reality (which I have no doubt is in some sense real, although I am not sure in what sense I have no doubt), and it exceeds the limits of my professional competence (I am a psychologist, not a philosopher). Nevertheless, I cannot avoid the issue entirely, at least as it concerns the reality of tertiary (meta) emotional experiences. I do not question that Panksepp and Russell believe that everyday emotional phenomena are real; they say so explicitly. Still, reading between their lines, I have some misgivings. By referring to everyday emotional experiences as “tertiary,” Panksepp implies that they are twice removed from the primary processes on which theories of emotion must ultimately be “grounded.” Similarly, Russell’s concept of meta-emotional experience suggests something beyond to scope of normal science. This suggestion is reinforced by the comparison



10.  What should theories of emotion be about? 217

Russell (2008) draws between astrology and folk-theories of emotion. Our ideas about the ways stars are configured may lead us to act in certain ways; but as Cassius observed (in Shakespeare’s Julius Caesar), “It is not in the stars to hold our destiny but in ourselves.” At the risk of putting words in his mouth, Russell might add with respect to emotions, “It is also not in ourselves.” That is, our ideas about emotions may influence the way we behave; also, various reactions during emotion (e.g., certain physiological changes) may involve their own causal chains. However, as discussed in the previous section, Russell does not believe there is anything about emotions per se (no unique feeling or affect program) that causes us to act emotionally. I agree that many of our folk-beliefs about emotions, like astrological beliefs, border on the superstitious. Moreover, increases in scientific knowledge may help make those beliefs more useful, that is, their obejcts pragmatically more real. Realism comes in a bewildering variety, of which pragmatic realism is only one. I pursue it here because I believe it has particular relevance to the way we conceive of the emotions. As an explicit philosophical movement, pragmatism was initiated by Charles Sanders Peirce, William James, and John Dewey, although it had ample antecedents, including Darwin’s theory of evolution and what Putnam (1995: 42) has referred to as the primacy of practical reason in Kant’s philosophy (the italics are Putnam’s, not Kant’s). For the most part, I will limit discussion to William James and Richard Rorty. James and Rorty may not be the best representatives of Pragmatism, but they are the ones I know best. Before getting to them, however, let me continue with a few observations on Darwin and Kant. Kant is justifiably considered the grandfather of cognitive psychology. Standing tradition on its head, so to speak, Kant argued that theoretical knowledge is possible only by filtering sensory experience through a system of cognitive constructs or schemas, some of which (like space, time, substance, causality) are not acquired through experience, but are priori to, or preconditions for, experience. (Try, he asked, to imagine an object outside of space, or space completely devoid of objects.) Kant’s formulation drove a gap between the world we know (phenom­ ena) and the world as it exists “in itself ” (noumena). In a classic paper, Lorenz (1962) argued that the theory of evolution could bridge the gap between Kant’s phenomena and noumena, and also explain why some ways of sensing and thinking are a priori. Briefly stated, a priori mental structures are part of our biological heritage; they reflect (but do not copy or mirror) the environment to which they are adaptations. Lorenz used a noncognitive example to illustrate his point. The hoof of a horse is a product of the terrain in which the horse evolved. Metaphorically, the horse’s hoof “knows” or is “true” to

218 James R. Averill

its environment without being a facsimile of that environment. And so it is with our cognitive structures or “organs of thought”; to the extent that they are products of our evolutionary past they are not only a priori to currently living persons, they also allow us to “know” the reality to which we are adapted. Biological evolution is not the only source of a priori “knowledge.” Social evolution has now replaced biological evolution as the primary determinant of how we think, feel, and act. Societies evolve and change over time. From a logical (or analogical) point of view, Darwin’s variation-and-selection model of evolution is applicable to social change. This fact was recognized by the founders of Pragmatism, who also adopted a variation-and-selection model of social evolution (see, for example, James, 1896/1956, p. 216). To sum up these preliminary observations, we are not born into this world as blank slates upon which experience can write just any script. We are, in a sense, pre-adapted to think and feel in ways that were beneficial during the long course of hominin evolution (roughly six million years). In all but the simplest matters, however, our biological propensities can be modified, if not completely transformed. Put differently, socialization can write over, albeit never completely erase, the scripts encoded in our genes. (See Park and Huang, 2010, for a review of evidence that immersion in a culture can influence brain structure and neural function; whether or to what extent such influence can reach “down” to the ­primary-process circuits discussed by Panksepp is an open – and perhaps unanswerable – question.) Part of the environment to which humans must adapt is the stock of knowledge and practices passed on from previous generations in the form of culture. James (1907/1955) put the matter as follows: “ideas (which themselves are but parts of our experience) become true just in so far as they help us to get into satisfac­ tory relation with other parts of our experience” (p. 49, emphasis in original). In other words, if a new way of responding is to gain traction it must preserve, James asserted, “the older stock of truth with a minimum of modification, stretching them just enough to make them admit the novelty, but conceiving that in ways as familiar as the case leaves possible” (p. 50). Pragmatism is, however, more forward than backward looking. To quote James (1907/1955) again, Pragmatism is an “attitude of looking away from first things, principles, ‘categories,’ supposed necessities; and of looking towards last things, fruits, consequences, facts” (p. 47, emphasis in original). It has long been a tradition in the West to ground knowledge and action on a foundation that is both atemporal and universal. (The efficient cause model attributed to classical mechanics, discussed in the previous section, has become one such foundational principle adopted by many social scientists.) Looking away from first things means abandoning that quest in favor of what James called a “radical empiricism.”



10.  What should theories of emotion be about? 219

Pragmatists have been accused of relativism. The accusation is warranted if by relativism is meant that truth depends, in part, on cultural context, that is, on an existing body of beliefs and practices. The accusation is not warranted if by relativism is meant that no criteria exist to help distinguish the better from the worse, the real from the not-real. The problems posed by relativism have been addressed by Rorty (1982, 1989). As we have seen, truth from a pragmatist perspective is inextricably tied to an existing body of thought and practices. This means, according to Rorty, that we may speak of truth within the Western tradition, a legacy of ideas bequeathed to us by such thinkers as Plato, Newton, Kant, Marx, Darwin, Freud, Dewey, and Milton. (These are a few of the persons mentioned by Rorty, 1982: 173; others would illustrate his point as well.) Needless to say, many false starts have been made and blind alleys pursued, sometimes at a terrible cost in human suffering. Rorty’s point is simply that there is no way to step outside of culture to obtain a neutral vantage point. Thus, we can with equal justification speak of truths within the Indian, Chinese and other cultural traditions. The extrapolation of Rorty’s observation to the “truth” or reality of emotional experiences is relatively straight forward. Emotional phenomena always exist in cultural contexts. We can, of course, gain perspective by studying multiple cultures. Cross-cultural research does not, however, lead to some transcultural reality; rather, it opens the mind to new possibilities, new adaptations that we may not have imagined before. The way to avoid ethnocentrism, Rorty argues, is to have only tenuous and flexible criteria for holding beliefs, as opposed to standards that are ostensibly rooted in some transcultural imperative, whether an “ultimate reality” or “universal human nature.” By “universal human nature” I mean a set of assumptions (mostly implicit) about what it means to be human. Some of these assumptions are methodological. For example, within the laboratory tradition espoused by Panksepp there are certain well established ways of thinking about human nature, especially with respect to the issue of causality; and similarly, for the cognitive mechanistic tradition adopted by Russell. Other assumptions are substantive. Most relevant to our present concerns, emotions are often depicted as an integral part of human nature. That may be more science fiction than science (cf. the novel, Invasion of the Body Snatchers, and the television series Star Trek), but the two are not unrelated in terms of underlying attitudes. Be that as it may, as the human analogue of an “ultimate reality” in physics, the idea of a “universal human nature” goes beyond such legitimate scientific notions as “species specific” characteristics. And what does all this have to Panksepp’s and Russell’s formulations? The relation, I must admit, is tenuous. As noted earlier, my primary concern in this section is with a skeptical attitude toward the theoretical usefulness (“reality”)

220 James R. Averill

of tertiary (meta) emotional experiences, in contrast to primary-process emotions (Panksepp) and Core Affects (Russell). The latter constructs are based on laboratory research, and hence rest, presumably, on more solid scientific ground. Nevertheless, it is they, not the former, that are steps removed from everyday experience. Tertiary (meta) emotions, I want to argue, are as real and as scientifically legitimate as primary-process neural circuits and Core Affects. With that said, let me return to the general argument, namely, that it is no more legitimate to found a science of emotion on a presumed universal human nature (however conceived) than it is legitimate to found a science of matter on a hypothetical real world, that is, a world that our ideas, as they advance, come to match more and more accurately. This is not to deny the existence of an independent reality. On the contrary, there are multiple realities; or, what amounts to much the same thing, there are multiple ways of responding to what is real – and, pragmatically, some of those ways may be better than others. The above observations apply even more to psychological than to physical reality. For the most part, physical objects do not change because of the way we conceive of them. For example, the planets did not adjust their motions to conform to the shift from a pre-Copernican (geocentric) to a post-Copernican (helio­centric) world view. We humans are different. We do adjust our behavior to conform to how we believe it is, or should be. Emotions – especially so-called “basic” emotions – have sometimes been taken to be exceptions to this generalization; they, presumably, are “mirrors of the soul,” true reflections of our original (authentic) human nature. However, as ample cross-cultural research exemplifies, human emotions appear in a disconcerting variety of ways, regardless of the level of analysis. Arguments to the contrary are based as much on prior assumptions as on data. Pragmatism posits a human nature that is flexible and subject to change – or at least reinterpretation – as circumstances require. This is naive, according to some critics. No one would deny that flexibility is a desirable attribute. However, we still need criteria for distinguishing the better from the worse. One does not want, for example, to keep an open mind about the mass killings that occurred in Nazi Germany, Cambodia under Pol Pot, and Rwanda, or about the many lesser atrocities – often attributed to emotions – that mar the daily lives of people around the world. The issue is thus not the need for criteria, but how to arrive at appropriate criteria in any given case. In this regard, Rorty asks us to rely on a “free and open encounter” between competing ideas, and “informed agreement” on which practices might better fit the circumstances. This advice, it might be noted, corresponds loosely to a variation-and-selection model of social change: Free and open encounters encourage variation, and informed agreement serves as a selection mechanism. Rorty (1989) makes the analogy explicit when he advocates



10.  What should theories of emotion be about? 221

“a nonteleological view of intellectual history, including the history of science.” Such a view, he claims, “does for the theory of culture what the Mendelian, mechanistic, account of natural selection did for evolutionary theory” (p. 16). But we are still left with the question: How might we evaluate the results of such a social evolutionary process? Rorty’s rejection of realism has been criticized as removing the motivation, as well as the foundation, for scientific research (not to mention moral behavior). If one adopts Rorty’s view, searching for truth (or goodness) might seem as auspicious as hunting snarks. To such criticism, Rorty replies that he is not an anti-realist. Rather, he questions the contemporary relevance of both realism and anti-realism, as well as such ancillary dichotomies as objective versus subjective, cognition versus emotion, and science versus art. These distinctions may have use (meaning) in some contexts, but as abstract categories they have solidified into dogmas that now hinder rather than facilitate progress. Obviously, by “progress” Rorty does not mean a closer approximation to some transcendental “reality”; rather, for Rorty, progress means adopting more creative and tolerant approaches to problems, both scientific and civic.

5. Concluding observations What are we to conclude from this brief excursion into issues of causality and pragmatic realism, as well as the earlier analogies between emotions, on the one hand, and diseases and species, on the other? Simply this: Treat tertiary (meta) emotions as scientifically relevant. Whether conceived of as episodic dispositions or transitional social roles, they have causal efficacy, and they are real. To answer the question posed in the title to this article, they are what theories of emotion should ultimately be about. This does not mean that Panksepp should do anything different than what he currently does, nor that Russell should do anything different than what he currently does. But between these bottom-up and topdown approaches there is a vast middle ground that has been relatively neglected by emotion researchers. If we treat everyday emotions as scientifically relevant, our theories will correspond better to the realities of our emotional life. However, the gain will not be purchased without a price. Individual emotions, such as anger, fear, love, etc., will lose their nomothetic status; no law of biology or psychology will contain uneliminable references to everyday emotions. But we can still have theories of emotion. Referring back to the earlier quotations by Temkins (1968) and Hull (1976), these theories will not contain reference to particular emotions but to “conceptual tools” (Temkins) or “theoretically significant kinds” (Hull). Perhaps Panksepp’s primaryprocess neural circuits and Russell’s Core Affects will be among those conceptual

222 James R. Averill

tools, as well as such other topics as appraisal, expressive reactions, prototypes and scripts, prereflective and reflective (higher-order) experience, transitional social roles, and emotional intelligence, to mention a few possibilities. The past few decades have seen an explosion of research on emotion; clearly, however, much remains to be done, tempered by a healthy skepticism of received wisdom, whether couched in terms of folk-beliefs or scientific dogma.

Acknowledgement It is unusual these days for an editor of an academic book to actually edit. Peter Zachar is a noteworthy exception. His comments on an earlier draft of this chapter were both detailed and insightful. Needless to say, whatever errors and misrepresentations remain are my responsibility alone.

References Averill, J. R. (1979). The functions of grief. In C. Izard. (Ed.). Emotions in personality and psy­ chopathology (pp. 339–368). New York: Plenum. Averill, J. R. (1982). Anger and aggression: an essay on emotion. New York: Springer-Verlag. Averill, J. R. (1985). The social construction of emotion: with special reference to love. In K. Gergen & K. Davis. (Eds.). The social construction of the person (pp. 89–109). New York: Springer-Verlag. Averill, J. R. (1991). Emotions as episodic dispositions, cognitive schemas, and transitory social roles: steps toward an integrated theory of emotion. In D. Ozer, J. M. Healy, Jr., & A. J. Stewart. (Eds.). Perspectives in personality (Vol. 3a, pp. 139–167). London: Jessica Kingsley Publishers. Averill, J. R. (2005). Emotions as mediators and as products of creative activity. In J. Kaufman & J. Baer. (Eds.). Creativity across domains: faces of the muse (pp. 225–243). Mahwah, NJ: Erlbaum. Averill, J. R. (2011). Ten questions about anger that you may never have thought to ask In P. Farzaneh. (Ed.). Multiple facets of anger: getting mad or restoring justice? (pp. 1–25). New York: Nova Science Publishers. Averill, J. R., & Nunley, E. P. (1988). Grief as an emotion and as a disease. Journal of Social Is­ sues, 44, 79–95. Averill, J. R., & Nunley, E. P. (2010). Neurosis: the dark side of emotional creativity. In A. Cropley, D. Cropley, J. Kaufman, & M. Runco. (Eds.). The dark side of creativity, (pp. 255–276). New York, NY: Cambridge University Press. Averill, J. R., & Sundararajan, L. (2006). Passion and qing: intellectual histories of emotion, West and East. In K. Pawlik, & G. d’Ydewalle. (Eds.). Psychological concepts: an international historical perspective (pp. 101–139). Hove, UK: Psychology Press. Briggs, J. L. (1970). Never in anger: portrait of an Eskimo family. Cambridge, MA: Harvard University Press.



10.  What should theories of emotion be about? 223

Duffy, E. (1934). Emotion: an example of the need for reorientation in psychology. Psychologi­ cal Review, 41, 184–198. Harré, R. (1986). An outline of the social constructionist viewpoint. In R. Harré. (Ed.). The social construction of emotions. New York, NY: Basil Blackwell. Hull, D. L. (1976). Are species really individuals? Systematic Zoology, 25, 174–191. James, W. (1890). The principles of psychology (Vol. 2). New York: henry Holt and Company. James, W. (1955). Pragmatism. Cleveland: meridian Books. (Original work published 1907). James, W. (1956). The will to believe and other essays in popular psychology and human immor­ tality. New York: dover Publications. (Original work published 1896). Lorenz, K. (1962). Kant’s doctrine of the a priori in the light of contemporary biology. General Systems Yearbook, 7, 23–30. Mayer, E. (1972). The nature of the Darwinian revolution. Science, 176, 981–989. Nafe, J. (1924). An experimental study of the affective qualities. American Journal of Psychology, 35, 507–544. Oatley, K. (2007). Dante’s love and the creation of a new poetry. Psychology of aesthetics, creativ­ ity, and the arts, 1, 140–147. Osgood, C. E., May, W. H., & Miron, M. S. (1975). Cross-cultural universals of affective meaning. Urbana, Ill: University of Illinois Press. Panksepp, J. (1998). Affective neuroscience: the foundations of human and animal emotions. New York: Oxford University Press. Panksepp, J. This volume. In defense of multiple core affects. Park, D. C., & Huang, C-M. (2010). Culture wires the brain: A cognitive neuroscience perspective. Perspectives on Psychological Science, 5, 391–400. Putnam, H. (1995). Pragmatism: an open question. Cambridge, MA: Blackwell. Rosaldo, M. Z. (1980). Knowledge and passion: ilongot notions of self and social life. Cambridge, UK: Cambridge University Press. Rorty, R. (1982). Consequences of pragmatism. Minneapolis: University of Minnesota Press. Rorty, R. (1989). Contingency, irony, and solidarity. New York, NY: Cambridge University Press. Russell, J. (2008). In defense of a psychological constructionist account of emotion: reply to Zachar. Journal of theoretical and philosophical psychology, 28, 423–429. Russell, J. This volume. Temkin, O. (1968). The history of classification in the medical sciences. In M. M. Katz, J. O. Cole, & W. E. Barton. (Eds.). The role and methodology of classification in psychiatry and psycho­ pathology (Public Health Service Publication No. 1584) (pp. 11–20). Washington DC: U.S. Government Printing Office. Titchener, E. B. (1908). The tridimensional theory of feeling. American Journal of Psychology, 19, 213–231. Wundt, W. (1897). Outlines of psychology (C. H. Judd, Trans.). New York: Gustav E. Stechert. Zachar, P. (2001). Psychiatric disorders are not natural kinds. Philosophy, Psychiatry, and psy­ chology, 7, 167–182. Zachar, P. (2006). The classification of emotion and scientific realism. Journal of Theoretical and Philosophical Psychology, 26, 120–138.

chapter 11

Valence, reductionism, and the ineffable Philosophical reflections on the Panksepp–Russell debate Ronald de Sousa Department of Philosophy, University of Toronto

One of the difficulties inherent in interdisciplinary work is that we approach the subject equipped with different vocabularies. In my writing I have seldom if ever used the word ‘affect’, at least not without quotation marks. A worse difficulty is that we do use the same words, and are not always aware of the extent to which our usages differ and carry incompatible assumptions. In that light, as a philosopher eavesdropping on psychologists and neuroscientists, I feel I should begin with a general disclaimer. Panksepp and Russell may find it hard to locate both their areas of agreements and their differences, but I am not unlikely to get them both wrong. Mostly I shall be raising questions which may simply illustrate that peril. This will also serve as my excuse for being skimpy in my specific comments on the particular questions posed.

1. Some questions about valence I begin, then, with some general considerations. Let me note first that there is, at least apparently, a good deal of agreement all round. Both Panksepp and I can agree with Russell that emotions are not “a single kind of entity or process,” and that the word ‘emotion’ “does not point to an affect program, perception of bodily reaction, Core Affect, arousal, appraisal, or attribution.” Faced with this diversity, Russell’s programme involves approaching each token of emotion – a specific episode of fear, for example – armed with different tools belonging to different disciplines within psychology and brain science. But in order to account for the specifically emotional character of the experience, Russell says

226 Ronald de Sousa

“Core Affect remains.” (R7). Core Affect is a specific kind of affect, which he uses in “a narrow sense limited to private subjective conscious feelings.” Beside Core Affect, affect in general also includes “affective quality and emotional meta-experience.” (R10). Core Affect is, it seems to me, uneasily poised between an observational and a theoretical entity. It is observational, in the sense that it is described as referring to a specific kind of affect, which Russell uses in “a narrow sense limited to private subjective conscious feelings.” But it is also a theoretical term, in that it postulates some sort of process or mechanism, underlying what we commonly think of as our emotional life, and that is entirely reducible, in Russell’s view, to two dimensions: valence, and arousal. Arousal, both Panksepp and Russell agree, may well rest on relatively uncontroversial physiological measures; valence, on the other hand, has been subjected to quite a bit of criticism. Though Panksepp insists that his approach to emotions is “bottom up” and based largely on animal studies, his view of valence echoes that of Robert Solomon, self-described phenomenenologist and existentialist, who objected that no single dimension of valence could do justice to the complexity and diversity of positive and negative emotions in the great variety of contexts in which they can be experienced and assessed. This raises the question, what determines the theoretical adequacy of a “dimension” such as valence? Strictly speaking, to say that two parameters are situated in orthogonal dimensions is to say that each is logically independent of the values of parameters in other dimensions. In the case of valence and arousal, this doesn’t seem quite true, since extreme arousal may switch a positive or “pleasant” valence to a negative or “unpleasant” one. But perhaps that is a mere quibble: in psychology, dimensions are unlikely to be entirely clean. More disturbing is the question of whether the quality of a particular form of stimulation must be deemed to constitute a specific and irreducible kind of valence, or whether we can think of it as involving a component of positive or negative valence added to a purely qualitative component. I’m not quite sure what is the correct answer here. But the idea that quality is intrinsically linked to valence has some curious consequences. Consider the question of whether it is possible for us to disagree about whether the taste of caviar is pleasant or not. I say it’s delicious, and you find it revolting. On the face of it, there are two plausible explanations for our difference. One is that both of us have exactly the same qualitative experience, but that it differs in valence for you and me. Yet there is something plausible about the following retort. “Obviously caviar does not taste the same to you as it tastes to me; if it did, then you would find it delicious.” Logically speaking, it seems at . Numbers in brackets preceded by R or P refer to the manuscript pages of Russell’s or ­Panksepp’s target articles, respectively.



11.  Valence, reductionism, and the ineffable 227

least possible that the component of valence is separable from the quale. But from the phenomenological point of view it may seem equally compelling to view the pleasantness or unpleasantness of a given sensible experience as intrinsic to the very quality of that experience. The question seems to be a philosophical one in the pejorative sense of the term, meaning one which affords no reliable method of empirical resolution; for though one might be tempted to adduce the separation of the aversive from the qualitative components of pain (Grahec, 2007) as an argument for the separability of quale from valence, the cleavage remains arbitrary from the strictly phenomenological point of view – insofar, that is, as we are concerned with the subjective character of emotional experience. Still, the fact that this involves something of an arbitrary terminological decision makes it just the sort of thing on which we philosophers should defer to brain science. Russell cites a number of authorities in support of the view that valence is both independent of arousal, at least if we stay away from the extremes, and unitary in the sense of taking in both pleasure and pain. Obviously I have no expertise on this question, but I find it intriguing that Panksepp is inclined to reject this unitary notion. Given his claim to be working bottom up, I would have thought that he would prefer a view on which the most elementary parameters, deriving from the simplest mechanisms, were used like bricks, as it were, to construct more complex structures and processes. What counts as bottom-up, however, may itself be in dispute. For while Panksepp uses the term to mean that he favours looking at psychological phenomena as stemming from subcortical structures and brain chemistry, Russell claims to work “bottom-up” in an even more radical sense, aiming to explain the workings of those structures in terms of general properties of more elementary constituents of brain tissue. Philosophers, perhaps like psychologists, divide into two classes of intellectual temperament, depending on whether they instinctively favour reductionist solutions, or find them intellectually repellent. On the whole, I’m with the reductionists; constructivism appeals to me because it seems sensible to suppose that complicated things and processes are built out of simpler elements. I am therefore happy to endorse a reductionist scheme on the face of it, providing that it identifies some clearly existing mechanism or mechanisms, and is able to build convincing models of the complex phenomena it is intended to illuminate. Is such a view plausible with respect to valence? The reasons Bob Solomon adduced against it rest largely on the felt incongruity of classifying “positive” or “negative” experiences together as “the same” when they arise out of wholly disparate situations. One of his arguments, for example, is that pain and pleasure cannot form a single continuum, because pains, by and large, are localizable in the body whereas pleasures are not. This is certainly plausible at the phenomenological level, but it is a weak argument against constructionism For one thing, pain is

228 Ronald de Sousa

only one sort of suffering, which can also, like joy, be purely mental. Furthermore it seems to be vulnerable to empirical discoveries such as that recently claimed for some work on the neural correlates of the pain of rejection, which involves some of the very same brain processes as simple localizable physical pain; to the extent that both can be abated by opioids and even Tylenol (MacDonald, 2005). In a commentary on MacDonald, Panksepp argued that the system involved is specifically the PANIC system, noting that “[t]he correspondence between the brain regions activated during human sadness and those activated during animal separation distress suggests that human feelings may arise from the instinctual emotional action systems of ancient regions of the mammalian brain.” (Panksepp, 2005). But of course this doesn’t imply that some of the very same mechanisms and chemicals involved in responses to physical pain aren’t also involved, which might, though it wouldn’t necessarily, explain the apparent kinship between physical and mental pain. So the advantage here seems to lie with Russell. Another of Solomon’s objections is that valence can’t be identified with the contrast of “approach = positive” and “avoidance = negative,” for some “negative” emotions such as anger impel us to move forward rather than retreat. One might add that a strong positive emotion such as aesthetic appreciation might motivate us to take a step back in order to get a better overall look. Positive and negative valence cannot, therefore, be explicated in terms of any simple pair of behavioral dispositions. On this point, the advantage remains with the pluralist. In an article that sympathetically surveys Solomon’s arguments, Jesse Prinz has suggested that we can find a coherent and consistent sense for the notion of valence if we identify it with the reward function. A positive valence will then be attributed to those experiences that result in a greater probability of a response’s recurrence, and the negative valence will correspondingly be attributed to those experiences that lower that same probability (Prinz, 2010). I don’t know whether this suggestion will fly, but I do think it is the right sort of suggestion for someone who is committed to a reductionist program.

2. Reductionism What is a reductionist program? Should one be committed to reductionism, or is it, on the contrary, something to be shunned? On this contentious topic I want to make just two remarks. Broadly interpreted, reduction is no more and no less than the fundamental project of science. It rests on the methodological presupposition that seeks to understand the working of complex phenomena in terms of their parts or components and the arrangement or interactions of those parts or components.



11.  Valence, reductionism, and the ineffable 229

So why has reductionism acquired such a bad name? I see two reasons, one good and one bad. But even the good reason is based on a misunderstanding and so it’s bad too. The simply bad reason, which William Blake and John Keats shared with a number of modern self-described anti-reductionists, derives from the feeling that understanding is incompatible with admiration, respect, awe, or even enjoyment: it is the idea that to understand a phenomenon is effectively to destroy it. As Keats saw it, by “unweaving the rainbow”, you will have ruined its beauty. This is a simply bad reason, because the constituents of any physical object are ultimately all the same – namely atoms, or subatomic particles – so it makes no sense to claim that the worth of a physical object depends on the nature of its constituents. About the potentially good reason, I am less confident. The suggestion is this. In identifying the constituents of something we value we may fool ourselves into thinking that this is all there is to it and forget that we are likely to have omitted some components and misunderstood some of their connections. My reason for holding that this is also based on a misunderstanding is that no scientist would ever claim to have exhausted the possibilities in the quest for the ultimate constituents and mechanisms that govern some important process. What antireductionists­ object to, then, is not reduction itself but premature closure. I think we should also take account of a third factor, which is neither a good reason nor a bad one: this is what I have already referred to as intellectual temperament. Some people just like to contemplate complex wholes; others just like to take them apart. Now in the case of the disagreement between Panksepp and Russell, I am tempted to think that temperament may have something to do with what is going on. I find it puzzling, however, because in some ways both figures appear to be committed to a reductionist approach in the inoffensive sense that I characterized above. In this respect I would expect to find them on the same side of the ideological divide, jointly contrasting with someone like Bob Solomon, an existentialist committed to the essential indeterminacy, unpredictability, and ultimately ungraspable nature of human reality. Panksepp, at least, does seem to think scientific temperament, or at least scientific approach, may underlie their differences. He writes that “the disagreement…. may amount to little more than the fact that the two of us have been approaching a difficult problem that interests us from two seemingly incommensurate perspectives – with Jim arriving at core issues from a top-down perspective, where the study of human words and concepts has led the way, and with me proceeding from a bottom-up view, where one probes ancient brain territories that are critical for the emotional-affective life of all mammals.” (P 28). But this just illustrates the unclarity of the term “bottom-up” remarked on above. For on the face of it one might construe their positions as exactly reversed. Russell might

230 Ronald de Sousa

be seen as placing the bottom deeper down than does Panksepp; on this view, the structures elaborated by the latter’s research are already in some relevant sense “constructed.” That, at least might come to mind when Panksepp emphasizes the diversity of human predicaments and of the types of responses we might have evolved to face them. Thus he reminds us that “there are many things that need to be done in this world for survival, and raw affects, in order to have optimal veridical ‘signaling’ value for learning, would need to come in many varieties. Otherwise, ‘everything else’ in affective life would have to arise from simple reward and punishment linked learning.” (P12). Indeed, but why shouldn’t it? That seems to be precisely what Prinz (2010) suggested. Isn’t it possible that all these other things arise, precisely, from associations built up in terms of the primitive mechanisms of positive and negative reward? One more example. In favor of specificity of primitive affect, Panksepp mentions food deprivation experiments: “No human in dire straits would experientially confuse hunger and thirst, and this is not just a conceptual act unique to humans” (P13). But recalling my little thought experiment about the caviar it will surely be plain that this is quite irrelevant to the issue at hand. In the case of thirst or hunger, it’s a plausible hypothesis that something other than mere discomfort, something that signals the nature of the present need, is associated with the discomfort in such a way as to guide behavior and specify an intentional object of desire. (In one of his early works, Bertrand Russell – not Jim Russell – floated the idea that the object of desire is simply whatever will actually put an end to the discomfort associated with it (Russell, 1921: 72). This was an unpromising theory, implying as it did that a sharp blow to the head would bring fulfillment to any of our multifarious longings.) Lacking the required expertise in brain science, I find it difficult to judge between Russell and Panksepp on the question of the nature of appropriate primitives. On general biological principles I see no reason to suppose that evolution has not facilitated the development of relatively modular processes, implemented, as Panksepp has argued that some basic emotional syndromes are implemented, in behavioral, neural, anatomical, and hormonal processes working together. Panksepp need not and does not deny that some emotional processes are learned. In the past I’ve referred to “paradigms scenarios”: these are the episodes that stand as defining identifiable items in our adult emotional repertoire. These develop in individual ways as a result of the individual experiences to which individuals are subjected in the course of their development. On any account, their development must arise from the concatenation of the most primitive, innate dispositions for “primary process” responses in Panksepp’s sense, or “Core Affect” in Russell’s scheme, in coherent sequences that we later make sense of in terms of our emotional vocabulary: a paradigm scenario is perhaps akin, though proceeding­



11.  Valence, reductionism, and the ineffable 231

from different intellectual ancestry, to the “script laying out a series of subevents (the components) in a temporal and causal order.” (R 20). It is in that sense that ­Russell’s constructionist conception of the complex phenomena we refer to as emotions are built up out of the elementary processes of “core affect”. But what is the sort of the “coherence” we attribute to emotional episodes?

3. Coherence and construction Russell’s contention is that the coherence in question is due to observer-relative construction (R8). For example, “in the mental prototype for fear, all the components cohere, but in reality, they rarely do.” That’s obviously true; and it seems to be a very general fact about all biological phenomena that they seldom meet all the criteria we set up when we try to define what comes under a given type. In terms of individual members of a species, Aristotle used to characterize outliers as ‘terata,’ or monsters, giving rise to an entire domain of investigation called teratology, or “monster-science.” The term, though not the concept, subsists in modern medicine; but from the standpoint of evolution, while outliers can still be characterized as poor examples of a given type for the sake of convenience, they are also potentially the very stuff of evolutionary change. If all our ancestors had been perfectly true to type, we would all be unicellular organisms. All this, however, is actually beside the point. For biology, while not requiring uniformity or perfect conformity to any types, is still perfectly well able to identify certain functions as objectively real, using the “aetiological” account of teleology. On this account, the function of the heart is likely to be the circulation of the blood rather than the production of rhythmic sounds. Both are actual effects of the heart’s activity, but only the former is plausibly supposed to have given our ancestors an adaptive advantage. Bearing this in mind, it is perfectly sensible to ask whether some of our learned responses are in effect the result of structures built up not merely during individual development but in the longer term by natural selection. And as I understand Panksepp’s proposal, the latter is just what is likely to be the case if there are indeed in the brain seven identifiable circuits. I find much plausibility in Panksepp’s view of the centrality of the seven functions distinguished in his scheme. I note, in particular, that three of them, namely the PANIC, LUST, and CARE systems, are central to something we call “love”. That there is at least a modicum of modularity attaching to those systems goes a long way to explaining the way that the demands of the variety of states we call love tend to clash and conflict. Otherwise, why couldn’t the ubiquity of Core Affect, which is supposed to determine a single point on the map at any given time (a point I shall take up in a moment), guarantee whole-heartedness at all times?

232 Ronald de Sousa

Russell, however, seems to be arguing that these “systems” are “constructed”. In my own mind, I must confess I’m not entirely clear as to what is involved in such a construction. I can see at least two significantly different senses the word might bear here. One might be said to be objective, the other subjective. In the subjective sense, ‘construction’ would refer to the fact that the concepts or phenomena constructed do not exist independently of the way they are constructed or construed by observers. They are effects of perspective and categorization that go beyond what exists objectively in the world. Russell’s analogy of constellations suggests just such a subjective interpretation: “Beyond the laws of physics and statistics, astronomy does not need an additional explanation for the appearance of constellations” (R8). But that can’t be quite right. It is not that constellations are explained by physics and statistics, but rather that constellations do not exist at all, independently of the labels that are applied to certain visible objects. They are entirely arbitrary ways of grouping some visible objects in the sky whose actual individual positions and appearance are indeed explained by physics and statistics. The reality of constellations belongs entirely to the realm of the psychological. The appearance of constellations must therefore be understood in terms of the psychology of perception, not of physics. Furthermore, the analogy to emotions seems rather weak in the face of the wealth of functional cohesion that is to be found in named emotions. That cohesion, whatever its origin in ontogeny or phylogeny, in brain organization or in learning, has no analogue in constellations. The other way of understanding construction, which I have called objective, is modeled not on our capacity to project patterns onto stars, but on the way that a house or wall is built out of bricks. Much of what Russell writes suggests this more robust, objective sense of construction. This, for example: “Emotional episodes are real events, whether or not they are labeled or categorized. They exist out there in the real world as the referents of our concept of emotion.” (R22). Core Affect, with its two dimensions, represents a basic capacity of the mammalian brain which is concretely involved in the elaboration of complex responses. We recognize every token of these responses as broadly speaking emotional.

4. Levels of analysis Another issue on which there is superficial agreement, but perhaps also lack of mutual understanding, is the question of levels of analysis. Panksepp insists on the importance of this in two ways. I’ve already noted that he characterizes the difference between himself and Russell as stemming in part from contrasting “bottom-up” and “top-down” approaches. He also insists on the importance of



11.  Valence, reductionism, and the ineffable 233

distinguishing three levels of affective processes, of which only the tertiary corresponds to complex emotions elaborated not only by personal and cultural influences but also by fully cognitive processing. In a general way this is clearly important and makes excellent sense. When trying to define in more detail what is at stake, however, I find myself wondering how these two “distinctions of level” relate to two others that philosophers have found it profitable to attend to. The first was introduced by David Marr (1982). It identifies three levels of analysis: the computational, the algorithmic, and the implementational. At the computational level, we are concerned with the nature of the overall problem to be solved. In the case of an emotion, we can think of this in terms of the function served by the emotion in the context of a person’s life situation and crucial concerns. At the lowest or implementational level, we would be looking at the physiological, chemical, and physical detail of the way the emotion is actually instantiated in the brain. That level itself can give rise to further distinctions that open up in a quasi-fractal way, in the sense that implementation of a given emotion can itself be thought of as involving computational problems, algorithmic solutions, and finer grained physical implementation. If Panksepp’s circuits are real, they themselves might be characterized as representing the coarser-grained algorithmic level, insofar as the neural circuitry is organized in logical ways and are in turn implemented by the chemical action of neurotransmitters, electrical potentials, etc. But this is somewhat speculative on my part, for it is difficult to pin down what exactly constitutes the algorithmic level in biological systems that don’t literally make use of systems of representation. In AI systems, the importance of the algorithmic level is easy to explain. It can be illustrated by reference to the procedures used to perform arithmetical operations such as multiplication. Such algorithmic procedures consist in specific series of well-defined operations on classes of symbols identified in terms of their role in a representation. This presupposes certain conventions of representation on which the operations can be effectively carried out. Comparing Arabic with Roman numerals, for example, makes it easy to see why the familiar multiplication algorithm requires us to use the former rather than the latter notation. The reason is that it relies on the possibility of carrying digits representing higher orders of magnitude so that they can be added to the next leftmost column. Roman numerals, not being based on decimal notation, afford no mechanical way of doing this. Among natural biological systems, the genetic system presents the only clear case of algorithmic operations, where we understand what is going on in terms of digital “coding” genes, embodied in “words” specifying amino acids or regulating the action of other genes. In the case of the brain and its implementation of emotional states, my poacher’s status as a philosopher does not afford me enough

234 Ronald de Sousa

knowledge to speculate on what might actually constitute the best interpretation of the algorithmic level. But I have a hunch that clarifying this question might be profitably attended to. Marr’s distinction isn’t the only one of its kind. A closely related but slightly different distinction is Dennett’s distinction between three “stances”: the intentional, the design, and the physical. Again, the top stance identifies problems to be solved; more specifically, it attributes to an agent intentional states, and particularly beliefs and desires, in terms of which an agent deliberates. Clearly, if this level is to be identified with anything in Panksepp’s scheme, it must be with the tertiary processes. When the phenomena resist explanation in terms of the intentional stance, Dennett counsels dropping down to the design stance, from which we can see how certain systems, organs or functions are doing just what they are supposed to be doing in their own terms, but yield the wrong answer to an overall practical problem because the current circumstances are not the ones for which the “module” in question was “designed”. It is easy to find examples of this sort of thing in our emotional life. Jealousy, for example, motivates notoriously counterproductive scenes that seem calculated to drive away a lover; at the same time, however, the emotion of jealousy, whether designed at the phylogenetic or at the ontogenetic level, seems clearly “designed” to serve the larger purpose of binding to oneself a mate. No such function can be ascribed to a sudden and unprecipitated endogenic depression. The most appropriate level of explanation in this case would therefore be the “physical” – a story concerned entirely with the excess or deficit of this or that neurotransmitter in this or that area of the brain. On the Eleven Questions, my comments will be both sparse overall and uneven in length. What is affect? What is its structure? Is affect fundamental, and if so, what makes it so? I note that Russell quotes with apparent approval a formula from Berridge & Winkielman (2003) (in turn borrowing from Zajonc): “affect, or the property of being good/bad.” This suggests that in his usage, affect is simply valence, and inherits the problems of that term as well as its advantages, as I sketched them above. Affect can be assimilated to the minimal condition of caring. In that sense it is fundamental to all value. If there were nobody to care, nothing would be of value. There is inevitably a threat of circularity when we try to explain the relation between caring, positive or negative affect, pleasantness, valence, and other such words. All have been associated with valence, but if there are important differences between them we will need to know which one is closest to valence per se, and what additional ingredients account for the different nuances carried in the (1)



11.  Valence, reductionism, and the ineffable 235

others. In any case, the least we can try to do is to regiment these various terms in such a way as to differentiate their functions. If Core Affect designates whatever it takes to care about anything, to want anything, to respond emotionally to anything, to experience, as Russell puts is, “agony or ecstasy,… excitement or lethargy” (P7), then indeed there is such a thing as Core Affect: but that does not in itself settle the question of whether we can distinguish some more complex structures that are biologically pre-adapted to perform a subset of the many functions we think of as emotional. It is a plausible general principle that complex organisms have evolved complex functions by biasing the organism’s learning facility to favour certain patterns over others.. This would lead us to expect that many emotions, and specifically those that have been identified as “basic”, will exhibit a high degree of coherence and consistency in their various subjective and objective manifestations. This brings up the next question. (2) How does affect relate to prototypical emotional episodes? Prototypical episodes are those that have names because they represent, in a vast space of possible affective responses, those that have been set up by both genes and the learning of paradigm scenarios as pertaining to commonly recurring life situations of adaptive importance. For Russell, prototypicality is a feature of our classification, not of nature: “prototypical cases are not necessarily more natural than mediocre cases” (R23). And Core Affect serves a function, but not necessarily the same function in all cases. My own reflections on this have already been set out. In this context they amount to agreeing, with a pinch of salt, with Panksepp’s take: it may still be that prototypical cases have no metaphysical priority, any more than prototypical birds are more worthy of the names than ostriches. But there could be an adaptive explanation for the existence of innate structures that make it easier for the life experience of individuals to pull their responses in towards the prototypical. To what extent that is so is an empirical matter that I am again ill-equipped to judge. Panksepp tells us that “[t]he evidence suggests raw emotional feelings are closely enmeshed perhaps isomorphic with the neural circuits that engender … the socalled ‘prototypical emotional episodes’” (P14), and I find that plausible; in particular, I find persuasive the consideration that “[p]ractically all [such raw feelings] are obtained only from stimulating subcortical regions of the brain, homologously organized in all mammals, and never from stimulating higher neocortical areas that enrich our lives with tertiary-process cognitive mentality.” (P15). I also find very significant the caveat that follows: “(this should give us pause to worry about how the study of human words relates to primary-process affects)”.

236 Ronald de Sousa

What kind of affect might be had by other mammals and/or human infants? How does such affect relate to the affect of adult humans? On this I again find Panksepp’s answer very plausible: “very many kinds of affect, and about the same for all mammals at a primary-process level. However, cognitive enrichment proceeds very rapidly after birth.” (P15). And again, “the primaryprocess emotional affects are foundational for the emergence of the more complex emotional-affective experiences throughout human life, where they can be combined into the seeming infinity of social-emotional nuances from admiration to envy and Schadenfreude.” (P16). Furthermore “even basic homeostatic/sensory affective states such as disgust may be used to mold higher order emotional responses such as social disgust and disdain through social learning.” Indeed, research on micro-expressions in responses to low offers in the Ultimatum game indicate that there is activation of the patterns of facial expression associated with physical disgust (Chapman, Kim, Susskind et al., 2009). Russell is right to wonder whether consciousness necessarily goes with the behavioral and other observable manifestations of affect, since the functions of response etc. that he lists as plausible causes of the adaptive value of affect can easily be envisaged to work without consciousness. So in attributing affect to infants and animals one doesn’t have much to go on except for the argument from analogy between one brain and another. “[T]otal elimination of higher neocortical regions of the brain at birth, leaves the affective vitality of animals intact” (R15). Panksepp here clearly disagrees: “there is a mountain of human brain imaging evidence indicating that a large variety of very modest anxiety-related stimuli arouse the higher reaches of the primary-process fear system in the amygdala” (P17). Is it real disagreement? Russell stresses the great variety of circumstances in which one speaks of ‘fear’: but if the same circuits are implicated, that tells us something about the building blocks of those various forms. It suggests, in particular, that there might genuinely be identifiable structures (not necessarily cortical) that are affected by both somatic and purely mental instances of fear, just as the ‘disgust’ circuits are implicated in the detection of both literal and metaphorical pollution. (3)

(4) How can we infer affect from observation of behavior? This is a big one, in theory, but perhaps not so crucial in practice because we do it easily enough. “Core Affect plays a role in selecting the goal and the plan to achieve it.” (R29). But how exactly can it do that, if at any particular point it is just a point on the circumplex space? A plausible idea is that it does this because valence is, as Prinz suggests, a matter of positive or negative reward. But the way this gets manifested in behaviour will generally involve cognitive functions, as in calculations of ways and means. So “playing a role in” does not mean ‘determine’.



11.  Valence, reductionism, and the ineffable 237

Is affect pre-linguistic? How can we study what is pre-linguistic? How does language alter affect? Sure, on any account affect has to be present pre-linguistically. We can agree with Russell that “Core Affect preceded language and phylogenesis and … ontogenesis and … continues to operate on a daily basis independent of language” (R35). But surely we need to qualify that last phrase: that Core Affect continues to operate regardless of linguistic activity does not mean it is impervious to linguistic activity. Our mood, notoriously, can shift radically on a word. And the experimental evidence cited from Berridge and Winkielman (2003) involves priming with linguistic cues unconsciously perceived. Conversely, our mood can influence what is salient in what we hear, as well as what we are able to understand and accept in what we hear. So we need to understand how language functions, both ways, in influencing and being influenced by affect, but not necessarily through conscious channels. Panksepp notes: “Primary-process affects have to be pre-linguistic for them to make any neuroevolutionary sense – namely that very similar ‘rewards’ and ‘punishments’ are critically important tools for living and learning in all mammals. This, of course, causes great mischief in emotions studies since we have no option but to talk about such BrainMind functions with our most sophisticated higher brain functions” (P19). Thus we are necessarily in a paradoxical position in speaking of the ineffable. In what might be called the ideology of emotions, the question often arises, though it is raised neither by Russell nor by Panksepp, about the extent to which the experience of emotion is ineffable. Briefly to advert to this question will reveal some of the diversity of issues that are lurking in the question about the relation of language to emotion. Language has many functions and affords several classes of illocutionary acts. Consider just four important classes: to “eff ” can be to refer to, to express, to describe, and also what has been called the special function of “critical communication” (Isenberg, 1949). Reference is unproblematic: a claim that a certain experience is ineffable cannot possibly mean that it is not possible to refer to it since that would make meaningless the very claim itself. As for expression, although there has been a great deal of discussion of the extent to which facial expression of emotion is either innate or universal, I don’t know whether the grunts, groans and moans emitted in spontaneous expression of emotions have been studied in such a way as to reveal the degree to which they issue from coherent and universal as well as at least relatively distinct systems in the brain, corresponding perhaps to Panksepp’s circuits. If they are found to be more stably correlated with the activation of these subsystems than facial expressions have proved to be, then even Russell might accept a level of structure that is both basic and yet more elaborate than his two-dimensional Core Affect. (5)

238 Ronald de Sousa

As for description, it cannot, by definition, arise in animals or pre-linguistic children; in some cases, it seems that emotion description elaborates and refines emotional experience (Frijda and Sundararajan, 2009); but many people complain that many emotional experiences, including experiences of an aesthetic nature such as that of music, cannot be adequately described and are more likely to be blunted than enhanced by the attempt to verbalize them. It would be an interesting area of research to make more precise what distinguishes those circumstances in which verbalization has a enhancing effect from those in which it has a negative effect. The fourth use of language, critical communication, looks like a description in the service of reasons to like or dislike something. But as Isenberg has shown it is better construed as a way of directing attention to features of a work of art or perhaps of an experience aesthetically envisaged, in such a way as to make a desired response more likely: If you direct your gaze to this part of the picture, if you observe the intertwining of these particular lines, you will likely “get it”. The reason I find this use of language important in this context is that it illustrates one way in which conscious experience and verbalization interact so as to modify and enrich both. Therapists undoubtedly make use of this sort of interaction in the process of “working through” a neurotic symptom such as jealousy or panic attacks. (6) Is affect always conscious? Everyone agrees, it seems, that the answer is negative, in part on the basis of a distinction between what is available to consciousness and what is being attended to. We can be affected by a great many stimuli the occurrence of which we are unable to report on. We might say that this results in an emotion, of which only the awareness, which could come in retrospect, constitutes the feeling. Panksepp’s answer is, sensibly, that it depends what you mean. “Consciousness is just like any other evolved aspect of BrainMind, with multiple levels of processing. To my way of thinking, three levels need to be considered at minimum – anoetic, noetic, and autonoetic consciousness” (P21). The third is what some people refer to as metacognition; the second is informed by conceptual knowledge of the world, but not reflected upon. Panksepp further makes a very intriguing distinction between intensity and power: “the ‘power’ of an emotion in the mental apparatus, which can be described more vividly perhaps by how much it fills the mind, as compared to ‘arousal’ which is the felt internal intensity of an emotional feeling” (P22). It’s

. This point was suggested by Peter Zachar, who also provided countless other improvements to the present essay.



11.  Valence, reductionism, and the ineffable 239

not clear whether he thinks that this might be another dimension that should be added to Russell’s circumplex model. For his part, Russell writes: “I believe we can speak of unconscious Core Affect but not of unconscious emotional meta-experience” (R36). Clearly, then, he too agrees that the notion of consciousness is a very slippery one. Something can be available to consciousness even though we are not in fact attending to it: that is sufficient to warrant that there is a referent to phrase, ‘unconscious affect’. But I find the claim just quoted slightly confusing in virtue of the fact that early on ­Russell defined affect, and specifically Core Affect, in terms of experience. “It is thus misleading to say we are conscious of or not conscious of the sadness, as if we were detecting or not the entity that causes the pattern. To experience sadness is to interpret the pattern of ingredients as sadness” (R39). There are actually two ideas here. One is that unconscious emotion is affect that is not being attended to; the other is that it unconscious emotion is just affect identified in terms of an inappropriate pre-existing pattern. The second idea has received a good deal of attention in philosophy. One example comes from the feminist literature in discussions of the way that depression can be reconfigured by being newly seen as anger. The first idea, about affect not attended to, is plausible in itself, but raises doubt in my mind about Russell’s claim that “a person is always in some state of Core Affect, just as a particular color falls at some point in the color space” (R12). The reason I find this confusing is that since there can be marginal, unattended items in the fringes of my awareness, you would expect the core-affective components of those items to differ from one another. So unless the one state of Core Affect is some sort of sum of all those, this suggests that we can, after all, be on several points of the map at once. Since Core Affect can be unconscious, it can’t be a single conscious sum of all one’s focal and fringe concerns that determines one’s unique state. What is the relationship between affect and cognition (conceptualization, intentionality, appraisal etc.)? This question is closely related to the last two. Russell writes: “A person’s Core Affect is often influenced by cognitively processed information and is often embedded within an intentional state” (R39). This is quite right; but I don’t completely understand what Russell means by the claim that it does not need any conceptualization “to do its work” (R40). Its work surely depends on current circumstances, and current circumstances are relevant only if they are interpreted. The exception is, of course, the case of undirected moods. I also agree that Core Affect or, I would say more generally, emotions, are to some degree informationally encapsulated in ways that are reminiscent of perception. But the account of the ways in which emotions are both encapsulated and also sometimes permeable seems (7)

240 Ronald de Sousa

unsatisfactory. Russell writes that “Core Affect is isolated from general information, with the massive exception of information currently seen through window of consciousness.” (R41). Notoriously, however, consciousness is by no means sufficient to alter affect. Sometimes, concentrating consciously on the aspect of a situation that should deliver us from anger or resentment simply has no effect. Is affect altered or filtered through culture? In what ways? Are there universal aspects of affect? Russell stresses the weakness of claims for basic emotions, whether interpreted as quasi-modular brain circuits, as Panksepp argues, or in some more abstract way as atomic building-blocks for more complex emotions. Panksepp, by contrast, holds that “affective neuroscience supports a view that an enormous amount of complexity is built into the spectrum of affective feelings – hunger, thirst, delight, disgust, the Big Seven, etc. – that are our birthright.” (P27). Apart from my own incompetence to judge, the evidence here seems inherently difficult to assess because of the many confounding variables: display rules as well as paradigms and norms that underlie the meta-experience of emotions can be expected to lead to a great deal of diversity, regardless of the amount of variety in biological underpinnings.

(8)

What roles do the biological sciences and psychological sciences have in studying affect? Russell is as sensible as concise on this: biology needs psychology, and conversely. I see no major disagreement in Panksepp here, nor do I have any. (9)

(10) What role might Panksepp’s primary affects play in Russell’s model of psychological construction? What implications might Russell’s model of psychological construction have for Panksepp’s construct of primary process affects? Both Russell and Panksepp are rightly irenic on this. They grant a good measure of agreement and complementarity. But they might go even further. In his questions to Panksepp, for example, Russell emphasizes the diversity in vocabulary, and presumably in experience, that exists between similar emotion concepts in various cultures. He asks, for example, whether there is a specific quale associated with each one of Panksepp’s circuits: “is it exactly as specified by an English word such as fear, or does it include the Gidjangali gurakadj, the Ifaluk metagu, the Pintupi ngulu or the Pintupi nginyiwarrarringu? What are the properties of the subjective experience that Panksepp hypothesizes?” (R10). We all apparently agree about the cultural relativity of specific complex emotions, and the importance of the



11.  Valence, reductionism, and the ineffable 241

ways in which we construe the circumstances of any particular emotional experience, which itself is going to be highly susceptible to cultural influences. The quale involved in any particular emotional experience, will undoubtedly be specific to that experience in context. Panksepp’s scheme implies that where the same basic syndrome is aroused, there will be a common quality; but he’s not committed to that quality exhausting the character of the experience in context, and it’s hard to see how one could identify and disentangle the various strands involved in the felt quality of a complex emotional experience. Let me close with a brief comment on Panksepp’s suggested “potentially critical experiment” (P29). As I understand it, this would rest on the existence of known chemical agents associated with positive affect, such as endogenous opioids and dopaminergics. If the action of these chemicals is uniform over different centres in the brain, particularly those known to be involved in higher cognitive functions, then one would expect that damaging any of these centres will affect all: “If damage to those higher brain regions reduces general ‘reward’ and ‘punishment’ functions in the brain, Russell’s viewpoint will have been supported. If such research does not yield coherent evidence for global positive and negative affects, the concept of a shared supramodal network would be weakened, albeit perhaps not terminally disconfirmed. Conversely, mine will be negated if those kinds of damage markedly reduce all ‘reward’ and ‘punishment’ effects across different ‘reinforcing’ objects throughout the brain.” (P29). My comment on this is that to the extent that this would be “potentially critical,” it may confirm my hunch that the differences are not as large as both writers, despite their irenic disclaimers, continue to believe they are. I see no theoretical reason to think that at some level of analysis, something like Core Affect might not be present throughout the brain, supported by neurotransmitters that are potentially to be found anywhere. What the neurotransmitters do belongs to the physical level of implementation. Conversely, if neurotransmitters are distributed evenly in the brain, I don’t see why that isn’t compatible with there being relatively specialized circuits the function of which is determined, as I suggested above, at the “design” level.

References Berridge, K. C., & Winkielman, P. (2003). What is an unconscious emotion? the case for unconscious ‘liking’. Cognition and Emotion, 17, 181–211. Chapman, H., Kim, Susskind, J., & Anderson. A. K. (2009). In bad taste: Evidence for the oral origins of moral disgust. Science, 323, 1222–6.

242 Ronald de Sousa

Frijda, N., & Sundararajan, L. (2009). Emotion refinement: A theory inspired by Chinese ­poetics. Perspectives on Psychological Science, 2 (3), 227–241. Grahec, N. (2007). Feeling pain and being in pain. Cambridge MA: MIT Press. Isenberg, A. (1949). Critical Communication. Philosophical Review, 54 (4). MacDonald, G., & Leary, M. R. (2005). Why does social exclusion hurt? The relationship ­between social and physical pain. Psychological Bulletin, 131 (2), 202–233. Marr, D. (1982). Vision. San Francisco: Freeman. Panksepp, J. (2005). Why does separation distress hurt? Comment on MacDonald and Leary (2005). Psychological Bulletin, 131 (2), 224–230. Prinz, J. (2010). For valence. Emotion Review, 2 (1), 5–13. Russell, B. (1921). The analysis of mind. London: George, Allen and Unwin.

chapter 12

Functional and empirical presuppositions in Russell and Panksepp Neural predispositions of affect Georg Northoff Canada Research Chair for Mind, Brain Imaging and Neuroethics Michael Smith Chair for Neuroscience and Mental Health Research Unit Director University of Ottawa Institute of Mental Health Research (IMHR) Royal Ottawa Healthcare Group University of Ottawa

1. Introduction Emotions are a crucial component of our lives and our selves. Everything we perceive and experience in our world is coloured by affect and emotions. We perceive specific events in association with specific emotions, we experience other persons in an emotional way, and we experience our own self emotionally. In short, affect and emotions are central to all human experience. This is clearly recognized by both Jaak Panksepp and Jim Russell. Based on the centrality of affect and emotions, Jaak Panksepp (1998, this volume) developed a neuroscientifically-based theory of primary process affects as raw emotional feelings which he associates with evolutionary ingrained subcortical circuits. Jim Russell (this volume) shifted from an earlier Psychological Construction Theory of emotions to the assumption of what he calls “Core Affect” as a basic and foundational unit (or building block) of any specific emotional feeling. While Panksepp’s concept of primary process affect overlaps at least conceptually (and also to some degree empirically) with Russell’s concept of Core Affect (see especially Russell’s commentary on Panksepp), they are not the same. My aim is to uncover and reveal some of the silent and hidden presuppositions in both Russell’s and Panksepp’s account in order to better understand the shared ground upon which their concepts of primary process affects and Core Affect are based. In particular, I aim to focus on functional and empirical presuppositions. In addition to pointing out their respective presuppositions, I aim to demonstrate

244 Georg Northoff

where they left questions and issues open and unanswered, and also to provide some hypotheses to begin to answer these questions. My ultimate aim is to show how our brains must operate in order to predispose and enable Core Affect and primary process affect. In other words, what makes any ‘basic affective unit’ \ possible and thus be necessary for Panksepp’s and Russell’s theories to be empirically plausible.

2. Functional presupposition: ‘Affective assignment’ and ‘double transformation’

Functional presupposition Panksepp distinguishes between three distinct kinds of primary process affects, homeostatic, sensory, and emotional affect. Homeostatic affect provides information about the body and thus interoceptive stimuli, sensory affect is related to exteroceptive stimuli, and emotional affect is associated with the brain (or BrainMind as Panksepp says) and hence with what one may call ‘neural stimuli.’ These distinctions make it clear that primary process affect is linked with stimuli generally, and more specifically with stimuli of different origins, be they of bodily(i.e., interoceptive), environmental, (i.e., exteroceptive), or neural origin. Hence, primary process affect must be somehow assigned to stimuli since otherwise Panksepp could not associate primary process affect with stimuli of such different origins. I call such association of stimuli with affect or primary process affect ‘affective assignment’ meaning that a stimulus of whatever origin can be assigned affect. Analogous to Panksepp, Russell must also presuppose affective assignment though in a slightly different way. He does not associate Core Affect itself with a specific type of stimulus since unlike Panksepp he does not speak of sensory, homeostatic or emotional Core Affect. Instead, Core Affect is continuously present independent of the presence or absence of particular stimuli. One though has to mention that Russell seems to refer here only to the absence of exteroceptive stimuli since he does not explicitly talk about interoceptive or even neural stimuli in this context. This means that it cannot be excluded that Core Affect may be related to the assignment of affect to either neural or interoceptive stimuli which would mean that Russell’s concept of Core Affect would also presuppose what I call affective assignment. In contrast to affective assignment with regard to neural and interoceptive stimuli, Russell explicitly refers to the assignment of affect to exteroceptive stimuli when he describes the transition from Core Affect to emotional episodes and emotional meta-experience. In the moment when the continuously present Core Affect is related to an episodically occurring exteroceptive stimulus, an emotional episode and meta-experience may occur. This however is possible only if the Core



12.  Functional and empirical presuppositions in Russell and Panksepp 245

Affect is linked and thus assigned to the exteroceptive stimulus thus presupposing what I here call affective assignment. While both Panksepp and Russell seem to presuppose the assignment of affect to stimuli, the exact functional mechanisms that enable and predispose such affective assignment remain unclear in their accounts. What functional mechanisms are necessary to enable and predispose the transformation of a non-affective stimulus into an affective one? I call this the ‘non-affective-affective transformation.’ The ‘non-affective-affective transformation’ raises the question how it is possible that a stimulus is suddenly associated with either Core Affect or primary process affect. It is especially worthwhile to consider that the stimulus as it begins neural processing, be it of interoceptive, exteroceptive or neural origin, must be non-affective. Hence ‘non-affective-affective transformation’ raises the question: what kind of functional mechanisms and neural input must the brain must provide in order for a stimulus to assigned affect. The question about the ‘non-affective-affective transformation’ raises another issue. Both Panksepp and Russell consider affect to be essentially subjective rather than objective. Panksepp refers to primary process affect as subjective by describing it as an ‘internal experience’ while Russell describes Core Affect as subjective in the sense of a private experience. Hence, Panksepp distinguishes internal from external and Russell private from public when they characterize Core Affect or primary process affect as subjective rather than objective. One should need to make a conceptual remark here. The meaning of the term subjective refers here only to the experience of affect, it does not say anything about the underlying neuronal mechanisms that may well be objective. This raises the question how affective assignment makes it possible to transform the originally objective stimulus be it interoceptvie, exteroceptive or neural into a subjective one. Hence, ‘non-affective-affective transformation’ is not limited to transforming a non-affective into an affective stimulus but with transforming the objective into a subjective stimulus. I therefore speak of ‘objective-subjective transformation.’ ‘Objective-subjective transformation’ raises the question: how the functional mechanisms related to neural input enable the transformation of an objective stimulus into a subjective one such that, in conjunction with ‘nonaffective-affective transformation’ it can be subjectively and thus internally and privately experienced. Taken together, affective assignment can be characterized by what I call ‘double transformation’, ‘objective-subjective transformation’ and ‘non-affectiveaffective­ transformation.’ While both Panksepp and Russell presuppose such ­double transformation they unfortunately remain silent about the exact functional ­mechanisms that enable and predispose it. I therefore want to fill this gap by proposing specific functional mechanisms that may underlie both ‘objective-subjective transformation’ and ‘non-affective-affective transformation’ (see Figure 1).

246 Georg Northoff

Non-affective-affective transformation: Interaction of any stimuli with the body’s interoceptive stimuli

Non-affective stimulus

Neuronal non-conscious state

Affective stimulus Neural coding in terms of differences

Objective stimulus

Conscious state

Subjective stimulus

Objective-subjective transformation: Interaction of any stimuli with the brain’s resting state activity, e.g., neural stimuli

Figure 1

‘Objective-subjective transformation’ Both Russell and Panksepp seem to presuppose some kind of intrinsic stimuli to be crucial in generating affect. Russell does so by explicitly distinguishing Core Affect from extrinsic stimuli and related emotional episodes, while Panksepp argues that neural activity in the subcortical circuits is not dependent upon extrinsic stimuli, i.e.,exteroceptive stimuli. This means that both must presuppose some kind of intrinsic activity for the generation of affect. What could such intrinsic activity be? One may assume it is that activity that can be observed in the absence of any extrinsic stimulation by either intero- or exteroceptive stimuli. Intrinsic means then that the origin of that activity must be traced back to the brain itself as distinguished from body and environment. One may refine such intrinsic activity as the brain’s resting state activity, or that activity in the brain in the absence of any intero- and exteroceptive stimuli (see ­Northoff et al., 2010). And it is such resting state activity as intrinsic activity that can be observed in all brain regions be it the cortical regions, the sensory cortex, or ­subcortical regions (see Northoff et al., 2010). The fact that resting state activity is present throughout the whole brain means that there may already be some neural interactions between the different brain regions within the resting state itself. For instance, the resting state activity level in the sensory cortex may interact with



12.  Functional and empirical presuppositions in Russell and Panksepp 247

the resting state activity level in the subcortical regions so that one may want to speak of rest-rest interaction. And there is further interaction. As soon as an inter- or exteroceptive stimulus enters the brain it interacts with the brain’s resting state activity level thus yielding what can be called rest-stimulus interaction (Northoff et al., 2010). Such rest-stimulus interaction may be specified according to the stimulus type either rest-interceptive stimulus interaction or rest-exteroceptive stimulus interaction (which in the following I will describe as rest-intero and rest-extero interaction. How do the three types of interaction, rest-rest, rest-intero, and rest-extero relate to affective assignment and more specifically to the non-affective-affective transformation and the objective-subjective transformation? The resting state activity level is different not only between different regions but even more importantly, between different persons. The same stimulus encounters a different brain in different persons meaning it must interact with a different resting state activity level. This means that rest-stimulus interaction individualizes the stimulus and adapts its processing according to an individual brain’s resting state activity level. Due to the individual resting state activity level and its impact on rest-stimulus interaction, the public stimulus is thus processed in a very individual and ultimately private way. By individualizing and privatizing the stimuli with respect to the brain’s actual resting state activity level, the originally objective stimulus is transformed into a subjective one. Hence what I called above objective-subjective transformation may correspond on a functional level to rest-stimulus interaction. Any processing of the stimulus be it of interoceptive or exteroceptive origin in relation to the brain’s resting state activity level (and hence its neural stimuli) may privatize and individualize the stimuli, and thereby transform it from an objective to a subjective one. Russell and Panksepp may now want to argue that this accounts only for half of the story. Panksepp may say that this leaves emotional affects as based on the stimuli from the BrainMind itself and hence its neural stimuli out; this may be so because rest-stimulus interaction concerns only the interaction with interoand exteroceptive stimuli. Hence, my assumption of rest-stimulus interaction corresponding to objective-subjective transformation may well account for what ­Panksepp calls homeostatic affects and sensory affects, but not emotional affects. Russell may want to make an even stronger point. My assumption of reststimulus interaction misses Core Affect altogether because Core Affect IS neither related to interoceptive nor exteroceptive stimuli, but precedes both kind of stimuli which becoming relevant only in emotional episodes. Hence, my assumption that rest-stimulus interaction corresponds to objective-subjective transformation may hold for emotional episodes and emotional meta-experience but not for Core Affect itself.

248 Georg Northoff

This however is to neglect that the brain’s resting state can interact with itself, as for instance the resting state activity level in the subcortical circuits interacting with cortical regions. There may thus be what could be called ‘rest-rest interaction’ where the neural stimuli of one particular region’s resting state may interact with those of another region. Recent imaging data show that such rest-rest interactions do indeed occur (see Northoff et al., 2010, for recent reviews). In the case of such rest-rest interaction, the resting state activity level of one network is set against that of another network. This entails that the former may be privatized and individualized relative to the latter, thereby being transformed from a purely objective state into a subjective state. Hence, rest-rest interaction may entail objective-subjective transformation in the same way rest-stimulus interaction does. More generally, any interaction between different stimuli that include the brain’s intrinsic neural stimuli and hence its resting state activity level as one of the interacting variables may be prone to privatization and individualization, and hence to objective-subjective transformation.

‘Non-affective-affective transformation’ How about the second feature of affective assignment, that OF non-­affectiveaffective­ transformation? What functional mechanisms correspond to the transformation of a non-affective stimulus into an affective one? Panksepp (this volume) himself gives one hint in this direction. He considers primary process affect to be intrinsically valuative (in a wider sense as not being restricted to reward) in that it mirrors the value of environmental, bodily, and neural information for the organism. The question is how such value can be generated, and what kind of functional mechanisms are necessary in order to value stimuli of different origin, exteroceptive, interoceptive, or neural? In order for stimuli of various origins to be valued for the organism, they must be related to the organism itself, including its body and brain. More specifically, exteroceptive stimuli from the environment need to be related to the brain’s neural stimuli leading to rest-extero interaction and the body’s interoceptive stimuli leading to intero-extero interaction. The same holds for interoceptive stimulim which need to be related to the brain’s neural stimuli thus requiring rest-intero interaction. Finally, as demonstrated above, the brain’s resting state activity itself may be valued when rest-rest interaction occur. How does affect enter these various interactions? Russell tells us that Core Affect is continuously present even in the absence of exteroceptive stimuli. Unlike exteroceptive stimuli which arise more episodically, there is continuous interoceptive input and thus continuous rest-intero interaction in the brain. Due to the continuous presence of the body, continuous interoceptive input and subsequent



12.  Functional and empirical presuppositions in Russell and Panksepp 249

continuous rest-intero interaction cannot be avoided. One may consequently consider rest-intero interaction as one possible candidate functional mechanism that corresponds to what Russell describes as Core Affect. It is by the continuous neural processing of the body’s interoceptive stimuli against the brain’s neural stimuli that affect may be generated. Hence, interaction of the interoceptive stimuli with the neural stimuli may transform the originally non-affective interoceptive stimulus into an affective one. The hypothesis is thus that rest-intero interaction may correspond on the functional level to the non-affective-affective transformation and thus to what Russell described as Core Affect. However, there is not only Core Affect but also emotional episodes (Russell­) or sensory affect (Panksepp) in relation to exteroceptive stimuli. How can extero­ ceptive­ stimuli be assigned affect and how can they undergo the non-affective­affective­ transformation? Very simple. They may be linked to interoceptive stimuli resulting in an intero-extero interaction. They would thereby be valued, which in turn would lead to a non-affective-affective transformation with the subsequent assignment of affect. Hence, one may consider the interaction of stimuli of ­various origins with specifically interoceptive stimuli from the body as a necessary condition for the non-affective-affective transformation. This may apply to the brain’s neural stimuli with rest-intero interaction which then leads to what Russell described as Core Affect and Panksepp as homeostatic affect. It may also apply to exteroceptive stimuli with intero-extero interaction that may then result in what Russell characterized as emotional episodes and Panksepp as sensory affect. One may now be puzzled. I characterized objective-subjective transformation by the interaction of any kind of stimulus with the brain’s resting state activity, i.e., its neural stimuli, so that any kind of rest-stimulus interaction will do the job. And I considered the interaction of any stimulus with interoceptive stimuli from the body as being necessary for the non-affective-affective transformation. Hence, both transformations, objective-subjective and non-affective-affective are characterized by interactions with different stimuli, the brain’s neural stimuli and the body’s interoceptive stimuli. As on a psychological level where affectivity and subjectivity co-occur, nonaffective-affective and objective-subjective transformations also co-occur in the ‘normal’ case. There is interaction with the body’s interoceptive stimuli (e.g., intero-extero interaction), and there is interaction with the brain’s resting state and thus its neural stimuli, (e.g., rest-intero and rest-extero interaction). This means that affectivity and subjectivity are co-constituted, which is reflected in both Panksepp and Russell definitions of affect by. If one interaction takes over at the expense of the respective other, the co-constitution between affectivity and subjectivity may become dysbalanced. This is, for instance, the case in schizophrenia where rest-intero and rest-extero interactions may be reduced leading to an

250 Georg Northoff

abnormal loss of subjectivity (Northoff and Qin, 2010). While the reverse seems to be the case in depression, where rest-intero interaction seems to predominate over intero-extero interaction (Alcaro et al., 2010).

3. Empirical presupposition: Regional specificity and neural predisposition

Empirical presupposition Panksepp considers various subcortical regions and circuits (see above for details) as the neural basis of primary process affect. He characterizes these subcortical circuits by specific features that concern the neural activity pattern, their inputs and outputs, and their modulation by and of cortical regions (see above and below for further details). By means of these specific neural properties, the subcortical circuits are supposed to enable and predispose the constitution of primary process affects. In contrast, Panksepp claims, cortical regions are more involved in enabling and predisposing cognitions. He thus considers subcortical regions and their specific features as providing the neural predisposition (see below for further discussion of this concept) of primary process affect. Analogous to Panksepp, Russell also singles out specific regions or a specific neural circuit to correspond to what he describes psychologically as Core Affect. Since Core Affect is characterized by the pleasure-displeasure and arousal dimensions, Russell assumes the pleasure circuit and the arousal circuit are central in constituting Core Affect. The pleasure and the arousal circuits comprise several subcortical regions and extend to cortical regions. Unlike Panksepp though, ­Russell does not specify the physiological and functional features of these circuits and their specific relationship to Core Affect (his theory is more psychological than neural). Despite some differences in the determination of the exact subcortical regions and their specific physiological and functional features, both Russell and Panksepp assume that their respective circuits enable affect. This however the raises a question about what this neural predisposition must look like in order to produce affect and thus how it bridges the gap from the neural level to the mental level of subjective affective experience. Unfortunately neither author provides a satisfying explanation how that gap can be bridged. What exactly predisposes and enables the assumed subcortical circuits to yield and constitute affect rather than for instance mere behaviour or cognition? This question includes two distinct parts. First, the question of regional specificity: What enables and predisposes the subcortical circuits (rather than ­ cortical



12.  Functional and empirical presuppositions in Russell and Panksepp 251

regions) to constitute Core Affect or primary process affect rather than mere behaviour or cognition? In addition to the question about regional specificity, the transformation of a neuronal into a non-neuronal conscious affective state needs to be addressed: What enables and predisposes the subcortical circuits to transform its neuronal and non-conscious (seemingly physical) states into non-neuronal and conscious (apparently mental) states?

Regional specificity Let me start with the first question, that of regional specificity. What enables and predisposes the subcortical circuits rather than sensory and other cortical regions to constitute affect rather than mere behaviour or cognition? In what follows I will discuss four of the six features by means of which Panksepp characterizes the subcortical circuits (Panksepp, this volume). The first feature concerns the connections, the afference and efferences, of subcortical regions. Subcortical regions show much higher connections to those regions processing interoceptive stimuli, e.g., brain stem and midbrain, than to ones associated predominantly with exteroceptive stimuli, e.g., sensory cortex. This distinguishes them from cortical regions where the connections to regions involved in exteroceptive processing, predominate over those associated with interoceptive processing. Hence, the connectivity pattern of the cortical regions seems to favour rest-extero interaction which distinguishes them from the predominant rest-intero interaction in the subcortical circuits. Considering the previous section, the different connectivity patterns of subcortical and cortical regions predisposes them functionally to different kinds of stimulus-stimulus interactions. Due to its predominant interoceptive connectivity, subcortical neural activity may be determined strongly by rest-intero interaction. As noted, the the involvement of interoceptive stimuli, leads to the subsequent non-affective-affective transformation and hence to affect. Since, in contrast, interoceptive connectivity is less prominent and replaced by exteroceptive connectivity, cortical regions may not be as involved in non-affective-affective­ transformation and the subsequent constitution of affect. Instead, they may rather be involved in constituting cognition which, speculatively, may be related to the rest-extero interaction rather than the rest-intero interaction. The second feature characterizing the subcortical circuits is, according to Panksepp, the coordination of behavioural and autonomic output. Autonomic and non-autonomic stimuli of the body are coordinated in the subcortical circuits so that the body is processed as a homogeneous whole rather than in distinct parts. Regions strongly involved in coordinating the autonomic and motor

252 Georg Northoff

f­ unctions of the body are the PAG and closely related regions like the Colliculi and the Tectum. Functionally, such coordination may correspond to interaction between different interoceptive stimuli in the gestalt of intero-intero. What does such intero-intero interaction entail for the constitution of primary process affect? We associate primary process affect with the body as a whole in our subjective experience; i.e., primary process affects are not ascribed, attributed or associated with a specific body part. This holds also for Russell’s concept of Core Affect. How does intero-intero interaction distinguish the subcortical circuits from the cortical regions? Due to their lack of (direct) interoceptive input, intero-intero interaction and hence the coordination of the various bodily autonomic and behavioural functions remains impossible for these cortical regions. However, they have strong exteroceptive connectivity and input from the five different sensory modalities. This prones the cortical regions for extero-extero interaction. Rather than coordinating the distinct aspects and functions of the body into a whole, they may rather coordinate and integrate the distinct aspects of the external world into a homogeneous whole. Does this mean that neural processing in cortical regions remains affectfree? Since as Panksepp himself points out in his fifth and sixth characteristic (see above) the reciprocal connectivity between cortical and subcortical regions enables reciprocal modulation between both. In other words, the affect assigned to interoceptive stimuli on the basis of rest-intero interaction in subcortical regions can be transferred to the rest-extero interaction in cortical regions. Following Russell, this subcortical-cortical interaction with the transference of affect from intero- to exteroceptive stimuli may correspond to what he describes as the transition from Core Affect to emotional episodes and meta-experience. Panksepp describes a third feature of subcortical circuits, namely their direct gating and modulation of inputs. This seems to correspond very much to what I here described on the functional level as rest-intero interaction, i.e., a gating and modulation of the interoceptive stimuli by the actual resting state activity level of the subcortical circuits. As I said, such gating and modulation and thus restintero interaction predisposes and enables the subcortical circuits to allow for objective-subjective transformation. Analogously, cortical regions may be prone to rest-extero interaction and hence to the privatization and individualization of exteroceptive stimuli so that both body and environment are experienced (and accessed) in a subjective rather than objective way. A fourth feature of subcortical circuits concerns the timing of their neural activity. Panksepp argues that subcortical circuits reverberate considerably longer than their precipitating circumstances are active, and he assumes neuropeptides play an essential role in such temporal extension. Such temporal extension by



12.  Functional and empirical presuppositions in Russell and Panksepp 253

neuropeptides affects the constitution of primary process affect and, consequently, our subjective experience of them remains open. We do indeed experience primary process affects like fear, joy, etc. in a prolonged way that reaches and stretches well beyond the actual trigger. Hence, Panksepp’s hypothesis of neuropeptide-based temporal extension of subcortical neural activity may well correspond on the subjective-experiential level to the prolonged experience of affect. Could the neuropeptide-based temporal extension also account for Russell’s assumption of the continuous presence of Core Affect? Certainly, this is not excluded. However, what Panksepp seems to have in mind is an episodic occurrence whose corresponding neural activity is temporally extended by the release of neuropeptides. Hence, to characterize Russell’s Core Affect by Panksepp’s neural mechanisms of temporal extension would be to confuse episodic occurrence and continuous presence. How then can we account for the continuous presence of Core Affect? I hypothesize that the continuous presence of Core Affect may be due to the continuous interoceptive input and its continuous interaction with the subcortical resting state activity, i.e., continuous rest-intero interaction. The body is always there and consecutively sends continuously interoceptive inputs to the brain leading subsequently to continuous rest-intero interaction. Since restintero­ interaction is associated with the consecutive constitution of affect, the latter should also be continuously present in for instance the gestalt of Core Affect or mood, which is exactly the case. How is the timing of neural activity in cortical regions distinguished from timing in subcortical regions? In contrast to the continuous interoceptive input of the subcortical regions, the exteroceptive input to the cortical regions may not be as continuous, as for instance, while sleeping. This may lead to a more episodic exteroceptive input in cortical regions which is reflected in the episodic nature of emotional episodes What remains to be explored, however, is how such differential timing of the intero- and exteroceptive inputs, continuous versus episodic, into subcortical and cortical regions shapes and determines their respective neural activity pattern. Hence, what Russell (this volume) called coherence on a functional-psychological level and what neurally may be called synchronization might differ in subcortical and cortical regions due to the temporal patterns of their respective predominant inputs.

Neural predisposition of consciousness (NPC) We are now ready to tackle the second question: What enables and predisposes the subcortical circuits to transform its neuronal and non-conscious seemingly physical states into non-neuronal and conscious apparently mental states?

254 Georg Northoff

Let me start with the concept of consciousness. Russell says that Core Affect is accessible to consciousness, but need not be conscious. Panksepp, in contrast, characterizes primary process affect by anoetic consciousness, one that does not yet involve any knowledge. Despite these differences, which may be more terminological than substantial (see above), both authors share the assumption that affect is somehow related to consciousness. What exactly do I refer to when I say that affect is somehow related to consciousness? I here do not single out a specific form of consciousness as for instance phenomenal or reflective consciousness. Instead, I denote by the term consciousness the principal possibility of a state becoming conscious whether it is actually unconscious, pre-conscious or conscious. It is more the predispositionor the principal possibility rather than the actual realization that I here target by the concept of consciousness. This meaning of consciousness is rather broad since it includes the unconscious, the pre-conscious, and the conscious with all three showing the ­ predisposition towards­ consciousness. Such a broad meaning resembles what the philosopher Searle (2004) calls the ‘principle consciousness’ or the principal possibility of a state becoming conscious. The ‘principle consciousness’ must be distinguished from non-consciousness. An example is neuronal states as such which have no principal possibility of becoming conscious as neuronal states; we do not experience our neuronal states as neuronal states in our consciousness but rather we experience what we call phenomenal or mental states (See also above with Panksepp touching upon this issue). The term principle consciousness in this sense comes close to the one that Panksepp likes to use, ‘core consciousness.’ What does the distinction between principle consciousness and the nonconsciousness­ entail for the characterization affect? As pointed out by both ­Russell and Panksepp, affect can become principally consciousness and must thus be characterized by principle consciousness. Their corresponding neuronal states though do not have access to consciousness as neuronal states and must thus be characterized by non-consciousness. Now we have the conceptual tools to again raise our initial question. What enables and predisposes the subcortical circuits to transform its neuronal and non-conscious states into non-neuronal and principally conscious states? Let me start with the transformation of neuronal into non-neuronal states. We remember, neural processing in subcortical (and cortical) regions may be characterized by various interactions: rest-rest; rest-intero; rest-extero; and intero-extero interactions. By neurally processing and setting one specific stimuli, for instance an interoceptive stimuli, against another one, the features associated with that stimulus itself are lost; this means that the stimulus-induced activity related to that particular stimulus independent of its interaction with others can no longer be traced back in our brain’s neural activity.



12.  Functional and empirical presuppositions in Russell and Panksepp 255

What is then processed in our brain are relations between different stimuli rather than single stimuli and their respective stimulus-induced activities. If, however ,the single and isolated stimulus-induced activities are lost, we remain also unable to perceive their corresponding neuronal states. All we perceive are then relations between different (admittedly hypothetical) stimulus-induced activities, which in turn makes it impossible to access the neuronal state as neuronal state. Hence it may be that the neural processing of relations between different stimuli rather than of specific isolated stimuli is what prevents us from accessing the neuronal states as neuronal states – thereby transforming them into non-neuronal­ states. This however accounts only for one part of the equation, the transformation of neuronal into non-neuronal states. The much harder part is to account for the transformation of non-conscious into principally conscious states? We remember, the neural processing of stimuli does not only occur in the gestalt of interactions­, but also in relation to the brain’s resting state activity and its neural stimuli. This in turn leads to the objective-subjective transformation and the consecutive constitution of subjectivity (see above). Subjectivity, e.g., private and internal experience of stimuli, is one hallmark of conscious states and hence of the principle consciousness as described above. Another hallmark feature of the principle consciousness is its reference to non-neuronal states, and hence phenomenal or mental states. This means that if subjectivity and non-neuronal states can be accounted for, we are able to understand the transformation of non-conscious­ to principally conscious states. Hence, combining the transformation of neuronal into non-neuronal states with the objective-subjective transformation, we are well able to account for the transformation of non-conscious into principally conscious states. What does this mean for Core Affect and primary process affect? Since Core Affect and primary process affect are assumed to involve the neural processing in terms of relations between different stimuli and against the brain’s resting state activity level, they are subject to both neuronal-non-neuronal and to non-conscious­-principally conscious transformations. Such a model is compatible with Russell’s account of Core Affect as pre-linguistic and accessible to consciousness and Panksepp’s description of primary process affect as conscious in an anoetic sense. What is also clear is that primary process affect or Core Affect may be situated right at the border between neuronal and non-neuronal states as well as between non-consciousness and principal consciousness. Core Affect may thus be the border station or transition point where both transformations take place and it may be therefore that their description by Panksepp and ­Russell oscillates between unconscious and conscious. This situates Core Affect or primary process affect also right at the border between empirical and epistemic domains something that seems to be mirrored in Panksepp’s (commentary on Russell) concept of ‘neuro-psycho-epistemology’ which he does not explicate

256 Georg Northoff

further though. One should be careful. Russell’s characterization of Core Affect as pre-linguistic is not the same as describing it as unconscious. The former, the prelinguistic nature, does not entail the latter, e.g, unconscious. However, both concepts highlight the relevance of a level beneath the higher-order cognitive level of consciousness and language. Finally, it should be mentioned that I do not presuppose the subcortical or cortical circuits as the neural correlates of consciousness (NCC), describing the sufficient conditions of phenomenal (or reflective) consciousness as specific forms of consciousness (Koch, 2004). Instead, I here only target the necessary conditions of consciousness in a broader sense, as principle consciousness. Hence, I target different conditions, i.e., necessary and not sufficient, as well as a different and broader concept of consciousness. Due to the fact that I target only necessary conditions that enable and predisposebut not actually realize consciousness (as in the case of sufficient conditions), I here speak of ‘neural predispositions.’ And since I focus on principle consciousness rather than on a specific form of consciousness, I speak of the ‘neural predisposition of consciousness’ (NPC) as distinguished from the ‘neural correlates of consciousness’ (NCC). I claim that the NPC is particularly relevant in the case of affective functions, i.e., Core Affect or primary process affect, and that without the NPC we will not be able to understand the rather intricate relationship between affect and consciousness advocated by both Panksepp and Russell.

References Alcaro, A., Panksepp, J., Witczak, J., Hayes, D. J., & Northoff, G. (2010). Is subcortical-cortical midline activity in depression mediated by glutamate and GABA? A cross-species translational approach. Neurosci Biobehav Rev., 34 (4), 592–605. Koch, C. (2004) The quest for consciousness. A neurobiological approach. Oxford/New York: Oxford University Press. Northoff, G., Qin, P., & Nakao, T. (2010). Rest-stimulus interaction in the brain: A review. Trends in Neuroscience, in press. Northoff, G., & Qin, P. (2010). How can the brain’s resting state generate auditory hallucinations? A resting state hypothesis. Schizophrenia Research, in revision. Panksepp, J. (1998). Affective neuroscience. Oxford: Oxford University Press. Panksepp, J. This volume. In defense of multiple core affects. Russell, J. This volume. From a psychological constructionist perspective. Searle, J. R. (2004). Mind. Oxford/New York: Oxford University Press.

chapter 13

Comparison of affect program theories, appraisal theories, and psychological construction theories Agnes Moors Ghent University

The water between the emotion theories of Russell (this volume) and Panksepp (this volume) seems deep. The aim of this chapter is to structure the theoretical landscape in the hope of gaining a better understanding of how these theories differ from each other as well as from appraisal theories (e.g., Lazarus, 1991; Scherer, 1984). This should help us decide which empirical research we could design to test these theories. Emotion theories differ in their answer to several questions. Some questions are about the definition of emotion. Others are about the cause of emotion. I discuss both types of questions in turn. After that, I examine which differences await an empirical answer and what kind of research is suitable for providing such answers.

1. Definition of emotion A concept can be used to refer to a set or to an element or instance of that set. An instance can be defined (a) by pointing at it (i.e., ostensive definition) or (b) by detailing the components (i.e., a componential definition). For example, an instance of a car can be defined by pointing at one, or by detailing components such as wheels, engine, tank, and dashboard. A set can be defined (a) by listing all instances or subsets (i.e., extensional definition) or (b) by stating the necessary and sufficient conditions for a thing to be an instance of the set (i.e., intensional definition). Intensional definitions rarely contain all the necessary and sufficient conditions. Usually, authors focus on the few conditions or criteria that distinguish the set from certain other sets. This kind of intensional definition can be dubbed a differential definition. It is good to keep in mind that the criteria figuring in differential definitions depend on the to-be-compared sets. For example, to distinguish­

258 Agnes Moors

the set of cars from the set of bikes, one may put forward the criteria that cars have four wheels and an engine. To distinguish it from the set of flowers, one may put forward the criteria that cars are non-living objects made out of steel. In extensional definitions in which subsets are listed, the choice of the subsets can be based on various features. Subsets of cars can be formed on the basis of brand, color, size, and horsepower. The feature chosen as the basis for creating subsets depends on the context. Size matters when transporting a big family. Color matters when organizing a traditional wedding procession. In the remainder of this section, I examine the questions (a) what are the components of instances of emotions (i.e., componential definition), (b) how to distinguish the set of emotions from other sets (i.e., differential definition), and (c) how to create subsets within the set of emotions (i.e., one type of extensional definition). I discuss the answers given by three (families of) emotion theories: affect program theories (e.g., Ekman, 1972; Izard, 1977; Panksepp, 1998, 2000; Tomkins, 1962), appraisal theories (Frijda, 1986; Lazarus, 1991; Roseman, 1984; Scherer, 1984), and Russell’s (2003, this volume) psychological construction theory. Panksepp’s theory can be considered a special type of affect program theory. I plead the cause of appraisal theories.

Componential definition An instance of an emotion can be considered in a broad or a narrow sense. The proponents of all three theories agree that an emotion in the broad sense (also termed emotional episode) consists of various components such as (a) a cognitive component, referring to processing or appraisal of a stimulus, (b) a motivational component, referring to action tendencies, (c) a somatic component, referring to central and peripheral physiological activity, (d) a motor component, referring to behavior (e.g., facial and vocal expressions, gross actions), and (e) a subjective component, referring to experience or feelings. Both Russell (this volume) and Panksepp (this volume) use the term of affect to refer to the feeling component. Most theorists conceive of feeling or affect as the reflection of (aspects of) the other components of the emotion into (first or second order) consciousness. Some theorists, however, leave the possibility open that some feelings are unconscious (e.g., Scherer, 2004, 2005a; Prinz, 2004 ) or do not require “neocortical readout” (Panksepp, 2007). I should point out that I use the term “component” as shorthand for “change in a component”. There is some disagreement about the exact number and nature of the components to include in the emotion. For example, Parrott (2007) omits the component of physiological responses (because all other components have a central



13.  Comparison of theories 259

physiological basis) and adds a component of emotion regulation. Some theorists argue that the cognitive component refers not only to (changes in) appraisal but also to (changes in) the operation of cognitive faculties like attention and memory (cf. Prinz, 2004). Russell (2003) supplements the standard components of appraisal, somatic responses, and motor behavior with the components of Core Affect, categorization, and emotional meta-experience. Core Affect is a combination of the variables of valence and arousal. It has a somatic side (it is defined as neurophysiological activity) and a mental side (the conscious experience of valence and arousal). The categorization of core affect produces what Russell calls emotional meta-experience, which is not the same thing as what other theories call emotional experience. Emotional meta-experience is more a cold thougt than a hot feeling. An emotion in the narrow sense consists of a subset of the components described above. Again, there is disagreement about the exact number and nature of the components that are part of the narrow definition of emotion. For example, James (1890) identified emotion in the narrow sense with the feeling component whereas Frijda (1986) identified it with the motivational component plus the feeling component. Others equate emotion with the sum of the somatic component, the motor component, and the feeling component (e.g., Izard, 1972). Still others seem to include all the components of the broad definition in the narrow definition of emotion (Clore & Ortony, 2000). In the remainder of this chapter, I use the term emotion in the broad sense because there is more agreement on the composition of emotion in the broad than in the narrow sense.

Differential definition The three theories discussed here distinguish the set of emotions from other sets in different ways. According to affect program theories, all and only emotions have a type of neural activity called affect program. The affect program of an emotion is situated in the central part of the somatic component and is put forward as the cause of several other components (motivational, peripheral somatic, motor) in the emotion. The problem is that all mental phenomena have concomitant neural activity somewhere in the brain. The question thus arises which neural activity counts as an affect program and which does not. According to Panksepp (this volume), a neural circuit counts as an affect program if direct stimulation of the circuit serves as a reinforcement or punishment in operant conditioning procedures. This already excludes a host of phenomena, but it does not exclude episodes of hunger and pain. It seems that affect program theories also rely on the additional criterion that all and only emotions have

260 Agnes Moors

one of a fixed set of survival-related functions (e.g., preserving safety, seeking new opportunities for goal fulfilment, and caring for offspring; Panksepp, this volume) or deal with a fixed set of species-constant problems (e.g., loss, danger, frustration; Ekman, 1999a). Each function corresponds to a distinct affect program. The problem with this criterion is that the exact composition of the set of functions is somewhat arbitrary. According to appraisal theories, appraisal is a necessary or (at least) typical condition for something to be an emotion. Appraisal is a process in which a stimulus is evaluated according to several appraisal variables such as goal relevance, goal congruence, coping potential, and agency. For example, a noise in the hall at night can be appraised as goal relevant, goal incongruent, difficult to cope with, and caused by someone else. In addition to being a necessary or typical part of emotion (in the broad sense), appraisal is also seen as the cause of the other components in the emotion. Appraisal theorists do not, however, think that appraisal is a sufficient condition for something to be an emotion. Appraisal of a stimulus only develops into an emotion when the output of this process is “goal relevant”. When the output is “goal irrelevant” (or relevant for a goal of minor importance), no emotion ensues (Moors, 2007). In other words, the criterion that distinguishes emotions from other phenomena is not the presence of appraisal, but the output or content of appraisal. Russell’s (this volume, 2003; Russell & Barrett, 1999) psychological construction theory refuses to point to any one component that differentiates the set of emotion from other sets. Emotion is a folk concept with no scientific value. The fact that Russell does talk about the set of emotions (or emotional episodes) indicates that he acknowledges that there is a feature that is shared by the instances of this set. As I understand it, the criterion for inclusion that Russell uses is “those phenomena that people categorize as emotions”. Emotions are collections of components grouped together in people’s minds on the basis of shared criteria. Discovering these criteria is one of the modest tasks of emotion research. Russell’s criterion for inclusion is objective in that it can be determined in an observerindependent manner which phenomena people categorize as emotions. On the other hand, the criterion is subjective in that the instances that end up in the set are dependent on people’s categorizations (they are observer dependent). It may be recalled that Core Affect or even a change in Core Affect is not specific to emotions. Core Affect is necessary for emotions, but not sufficient; it is ever present (Russell & Barrett, 1999). The presence of or change in Core Affect can thus not be considered as a criterion for demarcating the set of emotions from other phenomena.



13.  Comparison of theories 261

Extensional definition One type of extensional definition is a list of subsets of emotions. Emotion theories have proposed various criteria for dividing the set of emotions into subsets. I discuss the criteria for differentiation proposed by affect program theories, appraisal theories, and constructivist theories. Each of these criteria can be linked to an emotional component. Affect program theories divide the set of emotions according to the content of the central part of the somatic component. Subsets of emotions are created on the basis of the specific affect program involved. Several affect program theorists (e.g., Ekman, 1992a) also think that differences in affect program translate into differences in peripheral physiological activity (i.e., peripheral part of the somatic component) and differences in facial expressions (i.e., part of the motor component). As a consequence, the specific pattern of physiological activity and the specific facial expression also serve as criteria for differentiation. The subsets proposed by affect program theories are often referred to as basic emotions. Some affect program theorists posit that basic emotions fill the entire set of emotions (e.g., Ekman, 1994). They present each basic emotion as a family in which there is room for various shades. For example, anger also covers frustration, rage, and fury. Other affect program theorists hold that the set of emotions contains non-basic emotions in addition to basic ones. They depict non-basic emotions as blends or elaborations of basic ones (Damasio, 1999; Prinz, 2000). For example, contempt has been presented as a blend of anger and disgust (Plutchik­, 2001). Jealousy has been presented as an elaboration of anger in that it is anger about a particular event (i.e., another person has what you want). Many affect program theorists who distinguish between basic and non-basic emotions hold that non-basic emotions have the same affect program(s) as the basic emotions from which they derive. Some theorists leave room for non-basic emotions that do not have an affect program (cf. Griffiths, 2004). Some affect program theorists (e.g., Ekman, 1992a) think their subsets collapse with vernacular subsets of emotions like anger, fear, sadness, happiness, surprise, and disgust. Others reject a complete overlap with vernacular subsets and choose stipulative labels. For example, Panksepp (this volume) chooses the labels seeking, rage, fear, lust, care, grief/panic, and play to point at the central part of the somatic component in each of the subsets of emotions, but he has no labels for the subsets themselves. Appraisal theories form subsets on the basis of appraisal variables such as goal relevance, goal congruence, coping potential, and agency. These make up the content of the cognitive component. Appraisal variables can be treated as discrete (e.g., Roseman, 1991) or continuous (e.g., Scherer, 1994). For example, goal

262 Agnes Moors

r­ elevance can be treated as a discrete variable with the values “goal irrelevant” and “goal relevant”, or it can be treated as a continuous variable with an infinite number of values ranging from goal irrelevant to goal relevant. When appraisal variables are treated as discrete variables, their values combine to produce a large but finite number of subsets. For example, when the variables of goal relevance, goal congruence, coping potential, and agency are treated as discrete variables with two values each, their values combine to produce 16 subsets of emotions. When appraisal variables are treated as continuous variables,with an infinite number of values for each variable, their values combine to produce an infinite number of subsets of emotions. It may be noted that some appraisal theorists do not allow all possible combinations of appraisal values to occur. This is because they attempt to map appraisal patterns onto the subsets of emotions put forward by affect program theories or onto a set of vernacular subsets. For example, Lazarus (1991) does not have an emotion subset that belongs to the pattern with the values goal relevant, goal congruent, low coping potential, and self-agent, because there is no vernacular emotion subset that corresponds to it. Psychological construction theory differentiates emotions on the basis of combinations of valence and arousal, two continuous variables that characterize the content of the feeling component and the central part of the somatic component. The values of these variables combine to produce an infinite number of subsets. Constructivist theorists admit that the variables of valence and arousal do not capture all the variation that is present in the set of emotions. Additional variation may come from any of the other components. To summarize, the criteria for differentiation put forward by the three theories can all be linked to components. The criterion of affect program theories is linked to the central part of the somatic component, that of appraisal theories to the cognitive component, and those of constructivist theories to several components. Some criteria are variables (e.g., valence, arousal, appraisal variables) whereas other criteria are not easily captured in variables (e.g., affect programs). When criteria are variables and lead to a large or infinite number of subsets of emotions, the subsets are most efficiently described by pointing at the variables. It is easier to list a limited number of variables than a large or infinite number of subsets of emotions. When criteria are not variables (e.g., affect programs) and lead to a small number of subsets of emotions, it is equally efficient to list the criteria and the subsets. Disagreements among theories have to do with the boundaries of the category of emotions and the way in which to internally organize the set of emotions into subsets. Regardless of whether theorists think that the entire set of emotions as well as subsets of emotions can be distinguished with objective or (inter)subjective criteria, they agree that the instances in the set thus obtained show variation.



13.  Comparison of theories 263

There are many types of variation. I focus on three types. Emotions feel good or bad, and active or passive (Type 1 variation). People also have (or report) feelings of the so-called basic type (like anger and fear; Type 2 variation) and the so-called non-basic type (like guilt and jealousy; Type 3 variation). The same three types of variation can be found in the behavior-related components (motivational, peripheral somatic, and motor). Emotions can involve activity or passivity and the preparation and/or execution of approach or avoidance behavior (Type 1 variation). They can also involve preparation and/or execution of behaviors mentioned for so-called basic emotions (e.g., fleeing and fighting; Type 2 variation) and the so-called non-basic ones (e.g., repairing, Type 3 variation).

2. Emotion causation The previous section examined questions about the definition of emotion such as what are the components of an instance of emotion and how to distinguish and subdivide the set of emotions. The present section examines questions about the cause of the components in the emotion. The three emotion theories agree that (in normal cases) the components are caused by a stimulus (external or internal) that is somehow processed by the organism. In other words, a stimulus is the remote cause whereas the process(es) in the organism is(are) the proximate cause. The theories differ with regard to the kind of causal process(es) that they put forward. Affect program theories propose affect programs. Appraisal theories put forward appraisal. Russell (this volume) argues that there are separate processes for each component and that none of these processes is specific to emotions. I discuss seven differences between the processes of the three theories. (1)  Theories differ with regard to the component in which they situate their causal process(es). Appraisal belongs to the cognitive component whereas affect programs belong to the central part of the somatic component. Appraisal (in appraisal theories) and affect programs (in affect program theories) instigate action tendencies and peripheral physiological responses, which prepare and support motor responses. Aspects of all these components are reflected in the feeling component. Given Russell’s (this volume) proposal that there are separate processes for each component, each of the processes belongs to a different component. ­Russell does not entertain a priori assumptions about causal influences among these components, but he does not exclude them either. (2)  Theories differ with regard to the level of analysis at which they cast their causal process(es). I follow Marr’s (1982) proposal that any process can be described at (minimally) three levels of analysis: a functional level, an algorithmic

264 Agnes Moors

level, and an implementational level. At the functional level, a process is described as a relation between an input and an output. At the algorithmic level, a process is described in terms of the mechanisms that translate input into output. Examples of mechanisms are rule-based computation, the activation of learned associations, and the activation of harwired associations. At the implementational level, a process is described in terms of the brain structures or circuits involved. The three levels are related in that every functional process is based on some mechanism and some kind of neural activity. Yet, distinctions at one level of analysis do not map neatly onto distinctions at another level. For example, one functional process can be based on different mechanisms and different neural circuits. Affect programs are processes described at the implementational level of analysis. They are cast in neural and/or chemical terms. It is important to note that Marr’s (1982) three levels of analysis (functional, algorithmic, implementational) do not align with Panksepp’s (this volume) three levels of processing (primary, secondary, tertiary). Panksepp’s levels can all be situated at Marr’s algorithmic level. Roughly, Panksepp’s primary mechanism corresponds to the activation of hardwired associations, his secondary mechanism to the activation of learned associations, and his tertiary mechanisms to rule-based computations. Each of these mechanisms has a neural basis to be described at the implementational level. Panksepp has the assumption that primary mechanisms map onto subcortical areas and tertiary mechanisms onto neocortical areas. Appraisal is a process described at the functional level of analysis. It takes a stimulus as its input and produces values for one or more appraisal variables (e.g., goal relevance, goal congruence, coping potential, and agency) as its output (Moors, 2010; Reisenzein, 2001). A handful of appraisal theorists (Leventhal & Scherer, 1987; Smith & Kirby, 2001) have also addressed the algorithmic level. The mechanisms that they propose are rule-based, associative (i.e., activation of learned associations), and sensory-motor (i.e., activation of hardwired associations). In Russell’s psychological construction theory (this volume), some processes are cast at the implementational level (e.g., the processes producing Core Affect are defined as neurophysiological activity) whereas others are presumably cast at the functional level (e.g., the appraisal process takes the stimulus as input and produces appraisal values as output; the categorization process takes Core Affect and the stimulus as input and produces emotional meta-experience as output). . Another assumption is that affect is tied to the primary mechanism and cognition to the tertiary mechanisms. If the tertiary mechanism is at all capable of eliciting emotions, it is only because it is accompanied by activation of the primary mechanism. Tertiary processing is cognitive elaboration on top of primary processing.



13.  Comparison of theories 265

(3)  The causal chain between the stimulus and the emotional components can be split into two parts. The first part deals with the stimulus whereas the second part deals with translating the output of the first part into the remaining components of the emotion. Appraisal deals with the stimulus; hence it covers the first part. Affect programs cover the second part (Moors, 2009; see also Barrett, Ochsner, & Gross, 2007). Theories differ with regard to the part in the stimulus-to-components chain that they focus on. Appraisal theories talk about appraisal whereas affect program theories talk about affect programs. Thus, the natural focus of appraisal theories is on the first part whereas the natural focus of affect program theories is on the second part. This being said, some appraisal theorists have also addressed the second part and some affect program theories have also addressed the first part. Indeed, some appraisal theories have formulated hypotheses about how the output of the appraisal process influences the other components in the emotion. A first hypothesis (e.g., Lazarus, 1991; Figure 1, upper panel) is that all appraisal values are integrated in a pattern and that this pattern determines the emotion subset that is at stake. After this has been determined, a cluster of components that is typical for the emotion subset is elicited. For example, a sudden noise in the hall yields an appraisal pattern typical for “fear”. In turn, this pattern generates action tendencies, peripheral physiological responses, behavior, and feelings that are also typical for fear. In this hypothesis, the influence of the appraisal pattern on the other components is driven by a common process. This process may even take the form of an affect program. Hence, appraisal theories that adopt this hypothesis are compatible with affect program theories. A second hypothesis (Scherer, 2005b; Figure 1, lower panel) is that all appraisal values have a separate influence on the other components in the emotion. Each of these influences is supported by a separate mechanism so that many mechanisms are operating at the same time. For example, the appraisal value “goal incongruent” influences one piece of the person’s action tendencies, responses, and feelings, whereas the appraisal value “low coping potential” determines another piece of each of these components. Together, these pieces shape the components in the emotion. Appraisal theories that adopt this hypothesis are incompatible with affect program theories. In line with appraisal theories adopting the first hypothesis, some affect program theories (e.g., Ekman, 1992a; Matsumoto & Ekman, 2009) raise the possibility that affect programs are preceded by appraisal. Affect programs translate the . Some appraisal theorists (Frijda & Parrott, in press) argue that the influence of appraisal on the somatic and motor components is not direct, but indirect in that it goes via the motivational component.

266 Agnes Moors

Appraisal Action tendencies

Goal relevance Goal congruence

Emotion subset

Physiological responses

Coping potential

Behavior

Agency

Feelings

Appraisal Goal relevance

Action tendencies

Goal congruence

Physiological responses

Coping potential

Behavior

Agency

Feelings

Figure 1.  Appraisal theories that are compatible (upper panel) and incompatible (lower panel) with affect program theories

output of appraisal processes into the other components of the emotion. Affect programs are depicted as innate locks that are triggered when the right keys are inserted. The keys are specific appraisal patterns or schemes. In this version of affect program theory, appraisal occupies the first part in the stimulus-to-components chain whereas affect programs occupy the second part. In Russell’s (this volume) psychological construction theory, each component has its own causal chain which should be studied separately. Therefore, this theory does not accommodate easily to the framework in which a first step deals with the stimulus and the second step translates the output of the first step into the other components. (4)  Theories differ with regard to whether they hold that the emotional components are caused by a common process or by separate ones. According to affect program theories, the components in an emotion are caused by a common process, the affect program. There is a limited set of affect programs and each gives rise to its own cluster of values on the other emotional components. In appraisal theories, appraisal is put forward as the common cause of the remaining components in the emotion. Appraisal may be a single process at the functional level of analysis, but it covers multiple processes (i.e., mechanisms) at the algorithmic level of analysis (rule-based computation and the activation of



13.  Comparison of theories 267

learned and hardwired associations, Leventhal & Scherer, 1987). In addition, appraisal occurs for several variables (e.g., goal relevance, goal congruence, coping potential, agency) and some appraisal theorists have argued that each variable is processed separately (Scherer, 1984; but see Smith & Lazarus, 1993). Furthermore, it is likely that the various mechanisms underlying appraisal rest on different neural processes. In conclusion, appraisal may or may not be a single process at the functional level, but it does not correspond to a single process at the algorithmic level, nor to a single process at the implementational level. This being said, in a given instance of emotion, appraisal may be based on a single process at the algorithmic level (e.g., activation of learned associations) and perhaps a single process at the implementational level. It may further be recalled that some appraisal theories have not only addressed the first part but also the second part in the stimulus-to-components chain. Appraisal theories that are compatible with affect program theories (e.g., Lazarus, 1991; Figure 1, upper panel) assume that after the appraisal process is completed, the resulting appraisal pattern influences the other emotional components via a common mechanism, like an affect program. Appraisal theories that are incompatible with affect program theories (e.g., Scherer, 2005b; Figure 1, lower panel), on the other hand, assume that each appraisal variable exerts its influence on each of the remaining components via a separate mechanism so that many mechanisms are operating at the same time. In the latter case, the emotional components are not caused by a common process. Russell (this volume) argues against the existence of a single process (or type of process) that is the common cause of the other emotional components. In his view, each component is caused by its own process, and the study of each component should be handed over to a different branch of psychology. According to Russell, the components are not completely independent but they are more weakly linked than commonly assumed. This view is different from the scenario proposed by appraisal theories that are incompatible with affect program theories (cf. supra; Figure 1, lower panel). In that scenario, each appraisal variable exerts its influence via a separate mechanism, yet each of these mechanisms has a common influence on the other components in the emotion. The appraisal theories that are incompatible with affect program theories parse processes according to appraisal variables whereas Russell parses processes according to components. In other words, Russell proposes component-specific processes. There is a caveat. It can be questioned whether the links between the components are really that weak. It seems that appraisal does exert a strong influence on action tendencies and responses, and that these do have a strong impact on feelings. Regardless of whether a cluster of these components deserves the label “emotion” or not, the components in the cluster may still entertain strong causal

268 Agnes Moors

relations. I suspect that Russell’s suggestion that processes causing components are component specific stems from a failure to distinguish between causal processes and underlying processes. A causal process precedes a component and it is cast at the same level of analysis as the component, whereas an underlying process occurs simultaneously with the component and is cast at a lower level of analysis. An underlying process simply is the component described at a lower level of analysis. To clarify the distinction, consider the example that putting a kettle on the stove causes the water to boil. It is appropriate to say that putting the kettle on the stove causes the water to boil because putting the kettle on the stove occurs temporally prior to the boiling of the water and both events are described at the same level of analysis. The faster movement of the water molecules, on the other hand, is better conceived of as an underlying mechanism than a cause. It is a lower-level description of the boiling of the water. Returning to the case of emotions, I think it makes sense to claim that the processes underlying emotional components are component specific (e.g., neural processes underlying behavior may be specific to behavior and neural processes underlying feelings may be specific to feelings) but not that the processes causing components are component specific (e.g., if appraisal causes feelings, the process causing feelings is not specific to feelings). (5)  Theories differ with regard to their position regarding the specificity of the proposed causal processes or their outputs for the set of emotions or for some subsets of emotions. Specificity has to do with necessary and sufficient conditions. A is specific for B when A is a necessary and sufficient condition for B. In affect program theory, the set of affect programs is specific for the set of emotions (Ekman, 1994). Affect programs are necessary and sufficient causes of emotions. In addition, each instance affect program is specific for one subset of emotions. Each instance affect program is necessary and sufficient for one subset of emotions. It is not clear whether affect program theories have the additional assumption that the various affect programs are modular, in the sense that they do not share parts. In any case, the claim that affect programs are specific to subsets of emotions does not imply the claim that they are modular. In appraisal theory, appraisal is a necessary (or at least typical) but not a sufficient condition for something to count as an emotion (cf. supra, section on definitions). For example, when a noise in the hall at night is appraised as goal irrelevant, no emotion ensues. Thus, appraisal is not specific for emotions; it can also give

. This is akin to Crisp and Warfield’s (2001) distinction between diachronic and synchronic explanations. Both are explanations, but only the former are causal explanations. Both Russell and Panksepp tend to treat underlying processes as causal processes, situating all the causal forces at the implementational level of analysis.



13.  Comparison of theories 269

rise to non-emotional phenomena. The mechanisms underlying appraisal at the algorithmic level are not specific for emotions either. Rule-based and associative mechanisms, for example, are standard ingredients of dual mode models in all sorts of domains (cf. Sloman, 1996; Rouder & Ratcliff, 2004). In appraisal theory, it is not appraisal as such that is specific to emotions, but appraisal that produces the output “goal relevant” (cf. supra). In addition, each appraisal pattern is specific for each subset of emotions. Russell (this volume) not only argues that the processes that cause the emotional components do not have strong causal links (i.e., are component specific, cf. supra), but also that none of these processes is specific for emotions. Feelings that are part of emotions are caused by the same process(es) as feelings that are not part of emotions, and the same goes for categorization, appraisal, physiological responses, and behavior. (6)  Theories differ with regard to the number of causal processes they postulate, and more importantly, the total number of outputs of these processes. The total number of outputs of processes is related to the type of variation (Types 1, 2, and 3) that the processes explain. More outputs yield more variation. Affect program theories postulate a limited set of affect programs (e.g., ­Panksepp, this volume, proposes seven) and each has one output, that is, one value for each component. Each cluster of values corresponds to one subset of emotions. In affect program theories, subsets correspond to basic emotions. Thus affect programs account for Type 2 variation. Affect program theories refer to the notions of families, blends, and elaboration based on other processes than affect programs to account for Type 3 variation. Appraisal theories postulate one appraisal process, but this process has many possible outputs, as many as there are possible appraisal patterns (i.e., combinations of values on appraisal variables). There is variation among individual appraisal theories regarding the number of appraisal variables postulated (cf. Scherer, 2001) and the number of possible values for these variables (a few vs. an infinite number depending on whether the variables are treated as discrete or continuous). As a consequence, appraisal theories vary with regard to the number of appraisal patterns they consider possible. The number of possible appraisal patterns corresponds to the number of possible subsets of emotions. Appraisal theories that postulate a small set of discrete appraisal variables end up with less

. These outputs can be cast in a broad sense. For example, the output “withdrawal behavior” can refer to all kinds of concrete actions (fleeing, fighting) that have the aim of eventually decreasing contact with the stimulus. So, affect program theories are not committed to one concrete action per emotion subset.

270 Agnes Moors

emotion subsets than appraisal theories that postulate a large set of continuous appraisal variables. It should be noted that appraisal theories that are compatible with affect program theories tend to limit the set of possible appraisal patterns to those that correspond to basic emotions (cf. supra). In these theories, appraisal accounts for Type 2 variation. Like affect program theories, they have to resort to the notions of families, blends, and elaboration based on other processes than appraisal to account for Type 3 variation. In contrast, appraisal theories that postulate a large/infinite set of appraisal patterns causing a large/infinite set of subsets of emotions naturally account for substantial parts of Type 3 variation in addition to Type 2 variation. But even these theories may have to invoke elaboration for some forms of Type 3 variation (e.g., jealousy). All appraisal theories naturally account for Type 1 variation (valence and arousal). Variation regarding valence comes from the appraisal variables of goal congruence (in all appraisal theories) and intrinsic valence (in some appraisal theories). Variation regarding arousal comes from the appraisal variable of goal relevance (the more relevant a stimulus, the more arousal). In Russell’s theory, the number of possible outputs of the categorization process is limitless. Episodes can be categorized as “emotional,” but also as “negative,” “anger,” or “jealousy.” These outputs, however, do not correspond to types of emotional experience (i.e., the feeling component), but to types of emotional metaexperience­. The processes producing Core Affect can explain Type 1 variation in the feeling component. The processes producing the behavior-related components are responsible for all three types of variation in these components. (7)  Theories differ with regard to the mechanisms they propose on the algorithmic level of analysis (rule-based computation, activation of learned associations, and activation of hardwired associations) and hence whether the types of variation that they explain are computed on-line, based on learning, or hardwired, Affect programs are said to be hardwired, localized in the subcortical part of the brain, and homologous to the affect programs of other mammals. Thus, Type 2 variation is hardwired. Elaboration responsible for some forms of Type 3 variation is performed by associative and rule-based mechanisms located in the neocortical part of the brain and is based on learning and computation. In addition, Panksepp (2007) suggests that Type 1 variation (at least for valence) is the result of a (neocortical) categorization process based on learning. As mentioned, appraisal theories that address the algorithmic level propose that all three mechanisms (rule-based, associative, sensory-motor) can underlie appraisal. Leventhal and Scherer (1987) suggested that all appraisal variables can be processed by all three mechanisms. Other appraisal theorists tend to map particular appraisal variables onto particular mechanisms. For example, Reisenzein



13.  Comparison of theories 271

(2009) suggested that appraisal variables like goal relevance and goal congruence rely on hardwired processes. This view resonates with attempts to link appraisal variables to anatomical brain structures (e.g., amygdala as a relevance detector, Sander, Grafman, & Zalla, 2003; e.g., amygdala as a novelty or uncertainty detector, cf. Barrett, 2006a). Russell (this volume) has not given many details about the mechanisms underlying the components. He does suggest, however, that most components do not stem from the activation of hardwired or even learned associations, but that they are computed or constructed on the spot (and this is what the term “construction” in “psychological construction theory” refers to, Russell, personal communication). Core affect is the only component that can be based on the activation of hardwired associations; hence, Type 1 variation is the only type of variation that can stem from hardwired associations.

3. Empirical research Now that similarities and differences between constructivist theories, affect program theories, and appraisal theories have been discussed, it is time to consider whether and how the differences can be tested empirically. Similarities and differences pertain to definition and causation. With regard to definition, the three theories agree that an instance of an emotion consists of various components (componential definition). The theories differ regarding (a) the criteria they propose for demarcating the set of emotions (differential definition), and (b) the criteria they propose for differentiating subsets within the set of emotions (extensional definition). These differences involve an element of choice and therefore cannot be put to an empirical test. This being said, it is recommended that theorists choose criteria that are neither over- nor under-inclusive for the phenomena they have in mind (whether these correspond to the vernacular set of emotions or not) and that they make their criteria explicit. Another remark is that affect program theories and appraisal theories happen to distinguish and internally differentiate the set of emotions on the basis of their purported causal process. Thus, the viability of their definitions depends on empirical evidence for the existence of these processes. With regard to causation, the three theories agree that (in normal cases) stimuli are somehow processed and that the output of this process is somehow translated into the components of an emotion. The theories differ with regard to (a) the component(s) in which they locate the causal forces, (b) the part in the stimulusto-components chain that they capitalize on (first vs. second), and (c) the level of analysis at which they describe their causal process(es) (functional vs. ­algorithmic

272 Agnes Moors

vs. implementational). These are merely differences in focus. Empirical research should concentrate on disagreements among theories. Theories disagree about (a) whether components in an emotion are caused by a common process or by separate ones, (b) whether the causal process(es) or its (their) outputs is (are) specific for the set of emotions or subsets of emotions, (c) the total number of causal processes and/or the total number of outputs of these process(es) – this is related to the type of variation that the process(es) account for –, and (d) the mechanisms they propose at the algorithmic level and and hence the origin of different types of variation. I wish to emphasize that these are separate questions that require separate research programs. Empirical evidence for theories’ answers to these questions can be direct or indirect (Ortony & Turner, 1990). Each type of evidence presents its own difficulties.

Direct evidence In this paragraph, I discuss the kind of direct evidence needed for settling each of the issues of disagreement discussed above. (1) Direct evidence for the claim that the components in an emotion are caused by a common process consists in demonstrating (a) the occurrence of a single process, (b) the occurrence of the remaining components in the emotion, and (c) a causal relation among the process and the components. (2) Direct evidence for the specificity of a causal process for its effect (e.g., an emotion or anger) consists in demonstrating (a) that the process causes the effect, and (b) that the process is necessary and sufficient for the effect. (3) Direct evidence for the existence of a limited number of processes producing one output each (cf. affect program theories) vs. the existence of one process producing a large number of outputs (cf. appraisal theories) consists in (a) identifying the postulated process(es), (b) registering the outputs, and (c) demonstrating a causal relation among process(es) and outputs. (4) Finding direct evidence for the involvement of a mechanism at the algorithmic level is difficult if not impossible because mechanisms cannot be observed directly. Hence, it is difficult to determine what kind of evidence would constitute direct evidence for the hardwired, learned, or computed origin of a process. Some theorists seem to think that a process is hardwired when it can be associated with a particular brain area. Attempts to obtain direct evidence face several problems. A first problem is that researchers have to agree on what to consider an instance of an emotion. As mentioned, the criterion that affect program theories propose for calling something an emotion is that it should be caused by an affect program. The criterion that appraisal theories propose for calling something an emotion is that it should be caused by an appraisal process that produces the output “goal relevant”. By



13.  Comparison of theories 273

delineating the set of emotions on the basis of their cause, claims about the cause of emotions may become circular. For instance, if emotions are defined as those that are caused by an affect program, investigating whether the components in an emotion are caused by a single process comes down to investigating whether components that are caused by an affect program – which is a single process – are caused by a single process. Affect program theories and appraisal theories need to give us additional criteria to determine whether an emotion has occurred. A second problem is that it is difficult to determine what constitutes a single vs. several processes. It can be argued that, at the functional level, a single process corresponds to a single function. But one function can often be split into several sub-functions. For example, the process of appraisal can be split into subprocesses­, each of which has the function of determining the value for a separate appraisal variable. At the algorithmic level of analysis, it is even more difficult to determine what constitutes a single mechanism. Not only can mechanisms be split into sub-mechanisms; they are also unobservable. Hence, each postulated mechanism can be replaced by a combination of alternative mechanisms. Processes described at the implementational level of analysis are difficult to parse as well. When a pattern or network of neural activity is identified, it is difficult to determine whether it constitutes a single process (e.g., an affect program) or several processes. Patterns are composed of parts and it is possible that each of the parts constitutes a separate process. If affect program theorists would be more precise about what kind of neural activity counts as a single neural process, it would be easier to investigate their claims. A third problem when seeking direct evidence is that causal relations cannot be observed directly. This gap can be bridged to some extent by conducting experimental studies set up to demonstrate dissociations. A process is considered as a cause of an effect when (a) the presence of the process results in the presence of the effect and (b) the absence of the process results in the absence of the effect (i.e., a dissociation). It is not considered as a cause when both the presence and absence of the process result in the presence of the effect (i.e., no dissociation). A process is considered as a specific cause for an effect, when double dissociations are demonstrated. A process is considered as a specific cause for an effect when in addition to the single dissociation described above, presence of the process results in the absence of other effects. To illustrate, the set of affect programs is considered as a specific cause for the set of emotions when (a) activation of

. It may be noted that in experimental practice, causes are treated as necessary conditions in sufficient sets (this is close to Mackie’s, 1975, characterization of a cause as an insufficient but necessary part of a set that is itself unnecessary but sufficient).

274 Agnes Moors

an affect program results in an emotion, but not in non-emotional states, and (b) not activating an affect program results in the non-occurrence of an emotion. Instances of affect programs are considered as specific causes for subsets of emotions when, for instance, (a) activation of the affect program of fear causes the remaining components of fear but not those of other subsets like anger, and (b) activation of the affect program of other subsets like anger causes the remaining components of anger but not those of fear. The burden of proof is not only on affect program theories. Direct evidence for the claim that appraisal values are specific for parts of the other components in the emotion also requires demonstrating double dissociations. For example, if the appraisal value “goal incongruent” is a specific cause for a furrowed brow, this appraisal value should always result in a furrowed brow but not in other facial action units. In addition, no other appraisal value should result in a furrowed brow. Likewise, direct evidence for Russell’s claim that processes are component specific also consists in finding double dissociations. Processes that cause actionrelated components should not cause the feeling component and processes that cause the feeling component should not cause action-related components.

Indirect evidence In addition to direct evidence, researchers have adduced indirect evidence for their claims. Indirect evidence is based on inferences that follow from the claims. I give three examples. (1) An inference from the claim that components in the emotion are caused by a common process is that, in the default case, these components should cohere in time and intensity. (2) An inference from the claim that there is a limited set of affect programs leading to a limited set of outputs is that, in the default case, there should be consistency in each of the components across a broad range of stimuli. All stimuli must pass through a limited set of channels. An inference from the claim that the appraisal process produces a large/infinite amount of outputs is that there should be substantial variation in each of the ­components

. Only a few appraisal theories that are incompatible with affect program theories (e.g., Roseman­, 2001; Scherer, 2009) have fully worked-out hypotheses about relations between appraisal values and parts of the other components and they do not all claim that appraisal outputs are specific for parts of the other components. Another remark is that specificity on the level of parts is not in itself incompatible with specificity on the level of wholes. For theories that assume that the emotions in each subset have a unique part, evidence for the specificity of a process for that part also counts as evidence for the specificity of the process for the subset. For example, if Panksepp’s affect programs are specific for certain actions (i.e., parts), this is not a problem as long as it is assumed that each subset of emotions is characterized by a specific action.



13.  Comparison of theories 275

(3) An inference from the claim that affect programs are hardwired is that these should lead to universal signatures in expressive behavior (cf. Ekman, 1992b). Attempts to obtain indirect evidence faces several problems (for comprehensive reviews, see Barrett, 2006a, 2006b; Ekman, 1999a, 1999b; Ortony & Turner, 1990; Russell, 1994). A first problem is to know how much coherence, consistency/variation, and universality counts as evidence for coherence, consistency/variation, and universality. A second problem is that the default case rarely obtains in real life. The process causing emotional components always interacts with other processes such as regulation processes. This may dampen coherence, consistency, and universality. A possible solution is to keep the need for regulation low by removing external observers from the scene or to manipulate the need for regulation. A third, more serious problem is that indirect evidence falls prey to the fallacy of affirming the consequent. Coherence need not result from a common process; it may also result from the stimulus (independent of appraisal or an affect program). This problem can be circumvented by direct activation of the process in the brain (cf. Panksepp, this volume). Consistency need not result from a limited set of processes or outputs; it may also result from a limited range of stimuli that is generally encountered. Universality need not result from the hardwired nature of the causal process; it may also result from species-constant learning (Ekman, 1999a).

4. Conclusion The aim of the present chapter was to analyse the complex debate between proponents of affect program theories, appraisal theories, and constructivist theories of emotion. Starting from common ground among the theories (i.e., the premises that emotions in the broad sense are collections of components, that components are caused by process/es, and that different types of variation exist), this analysis revealed several axes on which the theories differ. By disentangling the points of agreement and disagreement, it becomes possible to identify the empirical questions. Each of these questions requires a separate research program and presents its own set of difficulties.

Acknowledgement Preparation of this chapter was supported by Methusalem Grant BOF09/01M00209 of Ghent University.

276 Agnes Moors

References Barrett, L. F. (2006a). Are emotions natural kinds? Perspectives on Psychological Science, 1, 28–58. Barrett, L. F. (2006b). Solving the emotion paradox: Categorization and the experience of emotion. Personality and Social Psychology Review, 10, 20–46. Barrett, L. F., Ochsner, K. N., & Gross, J. J. (2007). On the automaticity of emotion. Chapter in J. Bargh (Ed.), Social psychology and the unconscious: The automaticity of higher mental processes. New York: Psychology Press. Clore, G. L., & Ortony, A. (2000). Cognition in emotion: Always, sometimes, or never? In R. D. Lane & L. Nadel (Eds.), Cognitive neuroscience of emotion (pp. 24–61). New York: Oxford University Press. Crisp, T. M., & Warfield, T. A. (2001). Kim’s master argument. Noûs, 35 (2), 304–316. Damasio, A. R. (1999). The feeling of what happens: Body and emotion in the making of con­ sciousness. New York: Harcourt Brace. Ekman, P. (1972). Universals and cultural differences in facial expressions of emotion. In J. Cole (Ed.), Nebraska Symposium on Motivation 1971, (Vol. 19, pp. 207–283). Lincoln, NE: University of Nebraska Press. Ekman, P. (1992a). An argument for basic emotions. Cognition and Emotion, 6, 169–200. Ekman, P. (1992b). Are there basic emotions? Psychological Review, 99, 550–553. Ekman, P. (1994). All emotions are basic. In P. Ekman, & R. J. Davidson (Eds.), The nature of emotion: Fundamental questions (pp. 15–19). Oxford, UK: Oxford University Press. Ekman, P. (1999a). Basic emotions. In T. Dalgleish and T. Power (Eds.), The handbook of cogni­ tion and emotion (pp. 45–60). Sussex, U.K.: Wiley. Ekman, P. (1999b). Facial expressions. In T. Dalgleish and T. Power (Eds.), The handbook of cognition and emotion (pp. 301–320). Sussex, U.K.: Wiley. Frijda, N. H. (1986). The emotions. New York: Cambridge University Press. Frijda, N. H., & Parrott, W. G. (2011). Basic emotions or ur-emotions? Emotion Review, 3, x–x. Griffiths, P. E. (2004). Is emotion a natural kind? In R. C. Solomon (Ed.), Philosophers on emo­ tion (pp. 233–249). Oxford and New York: Oxford University Press. Izard, C. E. (1972). Patterns of emotions. New York: Academic Press. Izard, C. E. (1977). Human emotions. New York: Plenum Press. Lazarus, R. S. (1991). Emotion and adaptation. New York: Oxford University Press. Leventhal, H., & Scherer, K. R. (1987). The relationship of emotion to cognition: A functional approach to a semantic controversy. Cognition and Emotion, 1, 3–28. Mackie, J. L. (1975). Causes and conditionals. In E. Sosa (Ed.), Causation and conditionals. Oxford, UK: Oxford University Press. Marr, D. (Ed.). (1982). Vision: A computational investigation into the human representation and processing of visual information. New York: Freeman. Matsumoto, D., & Ekman, P. (2009). Basic emotions. In D. Sander & K. R. Scherer (Eds.), Ox­ ford companion to affective sciences (pp. 69–72). Oxford, UK: Oxford University Press. Moors, A. (2007). Can cognitive methods be used to study the unique aspect of emotion: An appraisal theorist’s answer. Cognition and Emotion, 21, 1238–1269. Moors, A. (2009). Theories of emotion causation: A review. Cognition and Emotion, 23, 625–662. Moors, A. (2010). Automatic constructive appraisal as a candidate cause of emotion. Emotion Review, 2, 139–156.



13.  Comparison of theories 277

Ortony, A., & Turner, T. J. (1990). What’s basic about basic emotions? Psychological Review, 97, 315–331. Panksepp, J. (1998). Affective neuroscience: The foundations of human and animal emotions. New York: Oxford University Press. Panksepp, J. (2000). Emotions as natural kinds within the mammalian brain. In M. Lewis & J. M. Haviland-Jones (Eds.), Handbook of emotions (2nd ed., pp. 137–156). New York: Guilford. Panksepp, J. (2007). Neurologizing the psychology of affects: How appraisal-based constructivism and basic emotion theory can coexist. Perspectives on Psychological Science, 2, 281–296. Parrott, W. G. (2007). Components and the definition of emotion. Social Science Information, 46, 419–423. Plutchik, R. (2001). The nature of emotions. American Scientist, 89, 344–350. Prinz, J. (2000). Getting to the heart of emotions. New York: paper delivered at American Philosophical Association. Prinz, J. (2004). Gut reactions: A perceptual theory of emotions. Oxford, UK: Oxford University Press. Reisenzein, R. (2001). Appraisal processes conceptualized from a schema-theoretic perspective. Contributions to a process analysis of emotions. In K. R. Scherer, A. Schorr & T. Johnstone (Eds.), Appraisal processes in emotion (pp. 187–201). New York: Oxford University Press. Reisenzein, R. (2009). Emotions as metarepresentational states of mind: Naturalizing the ­beliefdesire theory of emotion. Cognitive Systems Research, 10, 6–20. Rouder, J. N., & Ratcliff, R. (2004). Comparing categorization models. Journal of Experimental Psychology: General, 133, 63–82. Roseman, I. J. (1984). Cognitive determinants of emotion: A structural theory. Review of Per­ sonality & Social Psychology, 5, 11–36. Roseman, I. J. (1991). Appraisal determinants of discrete emotions. Cognition and Emotion, 5, 161–200. Roseman, I. J. (2001). A model of appraisal in the emotion system: Integrating theory, research, and applications. In K. R. Scherer, A. Schorr, & T. Johnstone (Eds.), Appraisal processes in emotion: Theory, methods, research (pp. 68–91). New York: Oxford University Press. Russell, J. A. (1994). Is there universal recognition of emotion from facial expression? A review of the cross-cultural studies. Psychological Bulletin, 115, 102–141. Russell, J. A. (2003). Core affect and the psychological construction of emotion. Psychological Review, 110, 145–172. Russell, J. A., & Barrett, L. F. (1999). Core affect, prototypical emotional episodes; and other things called emotion: Dissecting the elephant. Journal of Personality and Social Psychol­ ogy, 76, 805–819. Sander, Grafman, & Zalla (2003). The human amygdala: An evolved system for relevance detection. Reviews in the Neurosciences, 14, 303–316. Scherer, K. R. (1984). On the nature and function of emotions: A component process approach. In K. R. Scherer & P. Ekman (Eds.), Approaches to emotion (pp. 293–317). Hillsdale, NJ: Erlbaum. Scherer, K. R. (1994). Toward a concept of “modal emotions”. In P. Ekman & R. J. Davidson (Eds.), The nature of emotion: Fundamental questions (pp. 25–31). Oxford, UK: Oxford University Press.

278 Agnes Moors

Scherer, K. R. (2001). Appraisal considered as a process of multilevel sequential checking. In K. R. Scherer, A. Schorr, & T. Johnstone (Eds.), Appraisal processes in emotion (pp. 92–120). New York: Oxford University Press. Scherer, K. R. (2004). Feelings integrate the central representation of appraisal-driven response organization in emotion. In A. S. R. Manstead, N. H. Frijda, & A. H. Fischer (Eds.). Feelings and Emotions: The Amsterdam Symposium (pp. 136–157). Cambridge, UK: Cambridge University Press. Scherer, K. R. (2005a). Unconscious processes in emotion: the bulk of the iceberg. In P. Niedenthal­, L. Feldman-Barrett & P. Winkielman (Eds.), The unconscious in emotion (pp. 312–334). New York: Guilford. Scherer, K. R. (2005b). What are emotions? And how can they be measured? Social Science Information, 44, 695–729. Scherer, K. R. (2009). Emotions are emergent processes: They require a dynamic computational architecture. Philosophical Transactions of the Royal Society: B, 364, 3459–3474. Sloman, S. A. (1996). The empirical case for two systems of reasoning. Psychological Bulletin, 119, 3–22. Smith, C. A., & Lazarus, R. S. (1993). Appraisal components, core relational themes, and the emotions. Cognition and Emotion, 7, 233–269. Smith, C. A., & Kirby, L. D. (2001). Toward delivering on the promise of appraisal theory. In K. R. Scherer, A. Schorr & T. Johnstone (Eds.), Appraisal processes in emotion (pp. 121– 138). New York: Oxford University Press. Tomkins, S. S. (1962). Affect, imagery, consciousness: Vol. 1: The positive affects. New York: Springer.

chapter 14

Final remarks James A. Russell

I thank Jaak Panksepp, Peter Zachar, Ralph Ellis, and the commentators for an informative, challenging, and fun exchange of ideas. The commentators advanced the field by pursuing the initial issues and by proposing new ideas. Space limitations and my own limitations preclude me from commenting on many of these ideas, however much I agree or disagree. Instead, I focus on the disputes regarding my own account. I briefly clarify a dozen points and then take up seven issues at greater length.

1. Brief points 1. Some commentators saw me as carrying the flag for one discipline or another. Disciplinary boundaries are arbitrary historical accidents. Despite its name, “Psychological Construction” is a framework with room – indeed, a need – for philosophy, neurobiology, genetics, psychology, sociology, linguistics, anthropology, and others. 2. The word emotion continued to infuse mischief into this exchange. Emotion appeared as word, as concept, as disposition, as occurrent event, and as a class of events. Even as an event, emotion is ambiguous: emotion as an observable pattern of components, as inferred cause of that pattern, or as one of the components. Often, I simply don’t know what a writer means by emotion. 3. I am not calling for a “theory of emotional episodes.” It might appear that I snuck the concept of emotion in through the back door by saying that emotional episodes need to be explained. Not guilty. Emotional episodes are a heterogeneous bunch, and no one theory – short of all of psychology – will explain them. All the same, these episodes, like all human episodes, must be explained, and I sought to sketch one way to do so that is consistent with Psychological Construction.

280 James A. Russell

4. Core Affect and Psychological Construction are different ideas. They are not dependent on each other, nor is one an alternative to the other, nor is one based on the other. I have not shifted from one to the other. I just happen to believe both of them. Psychological Construction is a broad framework, a perspective, a research program. Within Psychological Construction, one could propose various specific theories, concepts, and hypotheses (Gendron & Barrett, 2009). Core Affect is one such specific concept, but is not essential for Psychological Construction. My proposal of Core Affect grew out of work I did as a graduate student (Mehrabian & Russell, 1974), which evolved into a circumplex model of affect (Russell, 1980), which evolved into Core Affect (Russell, 2003). Psychological Construction grew out of work done with my graduate student, Beverley Fehr, on the concept of emotion (Fehr & Russell, 1984) and out of a review of ethnographic reports on emotion concepts in various societies (Russell, 1991). 5. Psychological Construction and Social Construction are different ideas. Both are “constructionist” in rejecting essentialism and seeking to understand how emotional episodes can be built out of simpler constituents. Psychological Construction is unique in its claim that a token emotional episode is constructed at the time of its occurrence. 6. Psychological Construction is meant in de Sousa’s objective sense. Emotional episodes are real and are constructed in what he called a quasi-fractal hierarchy. The whole emotional episode is constructed on the fly out of simpler constituents, which, in turn, are constructed out of even simpler constituents, and so on, down to the most minute level. 7. Psychological Construction‘s claim that an emotional episode is the product of mechanisms not specific to emotion is misunderstood surprisingly often. This approach is common in much of modern psychology, as when remembering is no longer considered the product of one memory process. The idea is to replace a form of explanation popular long ago, as when a benevolent deed was explained by the benevolence faculty, a thought by the thinking faculty, and fighting by the fighting instinct. We don’t get far explaining an observed emotional episode by an unobserved emotion of the same name. 8. Northoff distinguished exteroceptive, interoceptive, and neural stimuli. As he surmised, I assume that all three types influence Core Affect. He then raised a key question: how is the source assigned to the Core Affect? This question was once central to the study of emotion, but is surprisingly neglected today. My proposal was called Attributed Affect (Russell, 2003). 9. Northoff noted that I remain silent on the question of the “objective – subjective’ transformation. Guilty. I have no solution to the “hard problem of consciousness.”



14.  Final remarks 281

10. Leddy, Robertson, and Schulkin claimed: “Much research has demonstrated the universality of facial expressions of emotions (e.g. Ekman and Friesen, 1971)” (p. 197). This claim is disheartening. Much has been learned on this topic in the last 40 years. At a minimum, the surrounding controversy should be cited (Fridlund, 1994; Russell, 1994; Russell, Bachorowski, & Fernandez Dols, 2003; Russell & Fernandez Dols, 1997). I believe that the claim does not stand up to scrutiny. 11. De Sousa wrote that if, based on brain imaging data, the same subcortical circuit is found in all cases of, say, fear, then “that tells us something about the building blocks of the form” (p. 236). Meta-analyses of brain imagining studies have not found the same subcortical circuit in all cases of fear (Lindquist et al., in press). 12. My central objection to any basic emotion theory is empirical, not philosophical. Although some data are consistent with the theory, many are not, and the prognosis is poor. Table 1 summarizes the problems. Much research has shown that actual emotional episodes differ from what we assume. Table 1.  Empirical evidence on predictions from basic emotion theory Prediction

Evidence

There is a unique neural pattern corresponding to each basic emotion.

Different cases of a basic emotion involve different neural circuits, and each neural circuit is involved in different types of emotion.

Each basic emotion produces a unique facial expression.

Failure to find convincing evidence that emotions produce facial expressions of emotion.

The facial expression of a basic emotion is easily and universally recognized.

Failure to find convincing evidence that people in different societies recognize the specific emotion allegedly signaled.

Each basic emotion produces a signature Failure to find convincing evidence of a unique pattern in the autonomic nervous pattern for each emotion in the autonomic nervous system. system. Each basic emotion produces a qualitatively different subjective experience.

Failure to find separate factors corresponding to basic emotions in studies of self-reported emotional experience.

People lexicalize the universal qualia of basic emotions.

Different languages lack a one-to-one correspondence between emotion terms.

Each basic emotion produces an action or action tendency.

Failure to find a class of actions common to instances of a given emotion.

Each basic emotion produces a pattern of co-occurring components.

Dissociation rather than predicted associations among manifest components.

282 James A. Russell

2. Davies’ skepticism and his interpretation of the two accounts Davies’ Skeptical Thesis is skeptical of folk psychology. From personal experiences and informal everyday observations of others, “we cannot justifiably claim to know … the correct nature of the emotion involved“ (p. 160). The “correct nature of an emotion should be understood as including its triggering causes, neural bases, downstream causal effects in the economy of the organism (especially its effects in various forms of homeostasis), and short- and long-term anticipatory functions” (p. 160). Later I consider the implications of this thesis, but some version is a background assumption of my framework. His Skeptical Thesis casts doubt on whether the extension of the everyday word emotion is the most useful domain for a scientific theory, whether fear names the mechanism causing its components, and so on. I accept his Skeptical Thesis because folk psychology is not science; folk psychology is not based on systematic controlled observations that test falsifiable hypotheses. Panksepp circuits. Davies offered a different reason to accept the Skeptical Thesis: Panksepp’s seven circuits. Davies further concluded: “our best current evidence is that the human mind comprises several ancestral affective systems that are foundational to the whole of human psychology” (p. 159). He later called this a “fact” (p. 160). In presenting the Panksepp circuits, Davies’ foundationalism went beyond what the scientific evidence warrants. Panksepp circuits may play a role in some emotional and non-emotional episodes, but so do other neural processes. Consider the role of Panksepp circuits in primary, secondary, and tertiary emotional episodes. I again use fear as an example. a. Primary emotional episodes are very rare in humans and far from the prototypes of the emotions. I know of no data on the relative frequency of primary, secondary, and tertiary types. To stimulate gathering such data, I speculate that the relative proportions are, respectively, 1%, 9%, and 90% of emotional episodes. The fear circuit does indeed produce behavioral inhibition and changes in the peripheral nervous system and endocrine system. A primary fear episode is an unconditioned (unlearned) reflex: an unconditioned stimulus (US) produces an unconditioned response (UR). In rats, the US for this reflex might be a chemical (one associated in evolutionary history with a predator), pain, sudden noise, sudden movement, or an open space. The reflex need not involve percepts (seeing the predator), cognitions (appraising the predator as dangerous), planned action (finding an escape route), or Emotional Meta-experience (feeling afraid of the predator). Does the fear circuit produce any affect, that is, any conscious feeling? And, if so, what feeling? Panksepp claimed that it does and wrote that the ­feeling



14.  Final remarks 283

is “fear itself ” (p. 124). Two problems. First, we were given no characterization of the feeling “fear itself.” The everyday word fear presents problems, but at least we have some idea what it means. “fear itself ” is deliberately not-fear, but what is it? Second, I see no evidence that the fear circuit alone produces any conscious feeling. Reflexes, behavior, physiological change, and information processing can occur without consciousness. Most neural circuits operate non-consciously. Consider blindsight: when a person’s primary visual pathway is impaired such that the person has no conscious experience of seeing something, an older pathway can allow some processing of visual information that guides subsequent behavior but without conscious awareness of seeing. Perhaps a Panksepp circuit operates without conscious awareness in a similar fashion. b. Secondary emotional episodes are also rare and far from the prototypes. They are conditioned (learned) reflexes. Some involve a simple conditioned stimulus and the fear circuit. Others are more complex; examples are cases in which the Conditioned Stimulus is multi-modal, in which the Conditioned Response is not the same as the Unconditioned Response, and in extinction. More complex cases might involve neural structures beyond those in the fear circuit (Kim & Fanselow, 1992). In humans, fear is difficult to elicit through conditioning (Rachman, 1977). c. Tertiary emotional episodes are what we think of as typical emotional episodes. The fear circuit, even if involved in most tertiary episodes, produces parts (behavioral inhibition, changes in the peripheral nervous system and endocrine systems) rather than the whole. Tertiary emotional episodes involve perceptual-cognitive processes (appraising the bear as dangerous; anticipating a tax audit), a planned behavioral response (finding an escape route and taking it; hiring a lawyer), and an Emotional Meta-experience (feeling afraid). Many involve complex appraisal and attribution processes, goals, intentions, and plans. Some even involve the imagination, judgments of morality, and goals that are not our gene’s goals. Hence, many neural circuits are involved. Meta-analyses of brain imaging studies show that each type of emotion involves many neural structures and each structure is involved in many types of emotion (Lindquist et al., in press; Murphy, Nimmo-Smith, & Lawrence­, 2003; Phan, Wager, Taylor, & Liberzon, 2002). Pankseep in his second chapter wrote that higher cognitive processes must be added to the fear circuit to produce a tertiary episode. Even for the parts produced by the fear circuit, the variety of autonomic patterns seen in fear episodes suggests that other circuits are involved as well. (Recall evidence that an escapable danger produces a different autonomic pattern than does an inescapable one. This difference is likely due to a circuit involving the ventral medial prefrontal cortex; Maier & Watkins, 2010.)

284 James A. Russell

Cortex. Davies emphasized subcortical circuits in part because he assumed a limited role for the cortex in what he called (but did not define) “affective coherence.” He claimed affective coherence among decorticate humans; as evidence, he cited only Shewmon, Holmes, and Byrne (1999). They presented case studies of four congenitally decorticate children who showed greater than expected functioning. These children had some cortical tissue, but very little. These were rare cases in which, despite their diagnosis, the children showed social responding; recognition of familiar others, toys, and pets; functional vision; associative learning; and responsiveness to music. Emotion was not formally assessed, but from casual observations during visits, physicians reported startle reflex, crying, smiling, giggling, irritability, unspecified facial expressions, and unspecified emotional vocalization. I see little evidence here of “affective coherence.” These casual observations are just what Davies’ Skeptical Thesis is rightly skeptical about. Psychological construction. Davies’ comments on my account were based on a misinterpretation. He wrote, “on Russell’s view, explaining the full range of emotion episodes is a condition of adequacy on a theory of emotions” (p. 156). I proposed no such condition because I do not think that a theory of emotions is a viable scientific project. Emotional episodes need to be explained, because all human episodes need to be explained, but a “theory of emotions” won’t do it. Davies wrote that Russell “posits the second core component of his theory, the categorical” (p. 165). His statement is unclear, but he seems to believe that, on my account, the occurrence of an emotional episode requires a categorical judgment as to what emotion it is. Perhaps he assumed incorrectly that there must be some means within my framework to divide emotions into natural kinds. There isn’t. Of course, token emotional episodes are sometimes – but not always – categorized by a scientist, a witness, or by the person undergoing the episode. This is simply a fact. The same episode can be categorized in different ways by different observers. When categorization occurs, it may or may not play a role in the emotional episode. (Watching from afar, a witness to Sally’s encounter with the bear might categorize her as afraid, but doing so would play no role in Sally’s emotional episode. A nearby witness might categorize her as afraid, shoot the bear, and thus alter the course of Sally’s episode.) Emotional episodes can occur with or without categorization of that episode. When someone categorizes his or her own emotional episode, the person has an Emotional Meta-experience (Davies’ EME), the self-perception of having a specific emotion. An EME is not necessary for an emotional episode to occur. Indeed, Davies provided one of several quotations from my chapter to this effect: The EME “can occur late in or even after the emotional episode, or never”



14.  Final remarks 285

(p. 91 and p. 166). Yet, curiously, Davies went on to portray me as claiming that “categorical judgments … are necessary for the existence of emotion episodes and types” (p. 167). This statement is a misunderstanding of what I said. Emotional episodes can occur with or without an EME. When an EME occurs, it may or may not play a role in the rest of the emotional episode. Some profoundly moving emotional episodes occur without an EME. Davies criticized my account for implying “that what we regard as emotion episodes are contingent on evolving features of our cultural histories. It would also show that our emotional lives are, to a significant degree, separate from our biological constitution” (p. 170). Indeed, I did claim a role for culture, but separating culture from “biological constitution” is a false dichotomy. If “biological constitution” is the currently functioning biological system rather than just DNA, then culture is not separate from biology. Culture influences individuals by influencing their biological system. Our emotional lives cannot be separate from our biological constitution. Davies would likely be equally skeptical of my ideas even if he correctly interpreted them. He was seeking the culture-free foundation for emotion rather than an account of the final product and how it comes about. I don’t believe that he has the right foundation, but, in any case, the foundation is not the final product. One of the most moving experiences of my life was a visit to Chartres cathedral. It could not exist without its foundation, and some may choose to study that foundation. But look at what’s above it. Affect and the skeptical thesis. Davies’ Skeptical Thesis implies that what people tell us is useless as evidence in the study of topics they know little about. Asking ordinary folks about pleasure’s neural circuitry or its downstream effects on homeostasis is no way to do neuroscience. Davies argued that the Skeptical Thesis challenges all evidence based on what people tell us and hence evidence I cited in support of my account of affect (conscious experience). But asking folks if they find a stimulus in front of them pleasant or unpleasant is very useful because their own conscious experience is something they do know about. For the study of immediate conscious experience, evidence based on what people tell us is essential and legitimate. Reports of subjective experience cannot be taken simply at face value, but they do yield reliable evidence that can be used to test specific hypotheses. In this way, scientific accounts of affect have made progress. Current accounts embed affect in a broader framework that includes antecedents, behavior, psychophysiological changes, neural changes, and so on. When done properly and when limitations are taken into account, reported conscious experiences are systematic controlled observations that test falsifiable hypotheses and thus are part of science.

286 James A. Russell

Davies interprets the Skeptical Thesis as challenging all evidence based on what people tell us. If so, this thesis would challenge much of psychology, including the study of sensation, perception, and large parts of cognition. It would challenge any study of consciousness. Sweeping skeptical theses, such as those implying the non-existence of motion (Zeno’s paradox), the non-existence of mind, or the non-existence of everything but mind may be interesting puzzles, but are not useful when they are roadblocks to scientific progress.

3. What’s wrong with folk emotion concepts? Concepts are necessary to science, and many emotion theorists borrow everyday English concepts – emotion, fear, anger, etc. – for their theories. Both Panksepp and I offered arguments against this practice. Scarantino accepted our conclusion, but rebutted some of our arguments. Scarantino examined each of our arguments, asking whether folk emotion concepts are scientifically suitable or not. This is not the most useful question. A science often begins with what turn out to be unsuitable concepts. Science can progress despite faulty assumptions, simplified theories, and poor concepts. Science typically begins by borrowing everyday folk concepts, but then modifies, redefines, hones, and, often, replaces them. Our job is to make improvements. Scarantino agreed that folk emotion concepts lack clear definitions. That is, they are not defined by individually necessary and jointly sufficient features. Thus, cases exist that cannot be decided as members or non-members; borders are fuzzy. Scarantino argued that unclear definition and borderline cases do not rule out folk emotion concepts as suitable for science. Still, unclear definition and borderline cases are not desirable properties in science. Not knowing what are the members and non-members of the category of fear, for example, makes testing hypotheses about fear indeterminate. Lack of clear definition and fuzzy borders are properties of concepts not of the objects or events they are concepts of. The concepts of ghosts and goblins lack clear definitions and have borderline cases, but not because of properties of actual ghosts and goblins. Other languages often lack exact equivalents to the folk emotion concepts named in English. Scarantino argued that culture/language-specificity does not rule out the English concepts as suitable scientifically to capture, ironically, universal human emotions. Still, it would be pure coincidence if English got it right, and all languages that categorize emotions differently got it wrong. Culture/language­ differences also help us see that our English concepts are not the only possible way to conceptualize the events loosely referred to as emotions.



14.  Final remarks 287

Our folk concepts bring with them assumptions of our religious, philosophical, and intellectual traditions, such as the dubious distinction between mind and body or that between reason and emotion. Scarantino argued that this feature too does not make these concepts unsuitable for science. Still, simplistic faulty assumptions are not desirable in science, and we must work to rid our accounts of them. Such assumptions may focus our attention on some aspects of the world, blind us to others, or even lead us to false conclusions. Because the extension of each folk emotion concept is heterogeneous, no account will apply to all and only cases called emotion or fear or anger. Scarantino agreed that heterogeneity renders folk emotion concepts fatally flawed for scientific purposes. Rather than dividing concepts into those suitable and those unsuitable for science, we must take a hard look at the actual properties of the concepts we use in our accounts and search for ways to improve them or find better ones. Our everyday folk emotion concepts have served affective science for decades, and much progress has been made. Scarantino and I agreed that these concepts are a source of recurring difficulties. They have outlived their usefulness. They have undesirable properties. We are not bound to use them. They are not inevitable. We can do better.

4. Eliminating emotion All of us have emotional experiences and plainly see emotions in others. We need to develop a scientific theory about such happenings. Those who have questioned the scientific usefulness of the concept of emotion have thus been accused of denying the obvious existence of emotions and called anti-emotion theorists. In 1933, Meyer suggested abandoning the concept of emotion as a scientific tool. He drew a parallel to physiologists abandoning the four humors or physicists the four elements. He called for the courage to throw off our “terminological shackles” (p. 296) and predicted that by 1950 the term emotion would have passed out of scientific psychology, much like the term will. Worcester (1933) defended keeping the concept of emotion by pointing out that emotions really exist. Duffy (1934) defended Meyer and countered Worcester by pointing out that “abandonment of a term does not imply abandonment of the study of phenomena loosely referred to by that term” (p. 103). I side with Meyer and Duffy against Worcester. The problem can be resolved by distinguishing a concept from the objects/events it is a concept of. Neither of these is the other. Constructing one is not constructing the other. Eliminating

288 James A. Russell

one is not eliminating the other. My proposal is parallel to science’s treatment of superlunary objects. Superlunary object is no longer a scientific concept, and yet all objects beyond the orbit of the moon nevertheless exist and need to be explained. Averill compared the call to eliminate the concepts of emotion, fear, etc. to calls once heard to eliminate the concepts of disease and species from biology. The words disease and species are used in science today, but the concepts have evolved from their origins in everyday speech. The folk taxonomies of diseases and species are being eliminated from science as they evolve into the taxonomies of modern medicine and biology. Because they were allowed to evolve, the scientific concepts of disease and species are in much better shape today than are emotion concepts. The general term disease remains problematic, but concepts of some specific diseases have made progress. In modern medicine, some diseases are now defined by a specific causal agent – a bacterium, virus, etc. (You have poliomyelitis if and only if your symptoms are caused by the poliovirus.) The same symptoms from a different cause are not poliomyelitis. The concept of the disease of poliomyelitis is therefore verified if and only if a specific causal agent is discovered for the symptoms. In parallel, if a causal agent (such as an affect program or a peptide) for all and only cases of fear were discovered, then my argument against the concept of fear too would be defeated. If the causal agent were found in a proper subset or highly overlapping set of fear cases, and only there, then the word fear might be retained, provided we allow the concept to evolve. But no such causal agent has been found, and there is great resistance to allowing emotion concepts to evolve. The concept of species has evolved greatly, with pre-Darwinian versions very different from the concept of the same name in modern biology. There are biological mechanisms that continue species and that keep different species separate from each other (mechanisms that sometimes fail). Identification of a member of a given species is typically made by observing superficial features, and mistakes are uncommon but possible. What defines the species in biology today is not, however, the observable features, but DNA. The set of all tigers may lack superficial properties that constitute a classical definition (a 3-legged, 1-eyed, tail-less tiger painted pink remains a tiger). All tigers share enough DNA to make them tigers and to separate them from other species. Because species can be defined through DNA, biological taxonomy continues to evolve. Emotional episodes exist, and the concept of emotion exists. Thus, I agree with Averill when he wrote, “Despite doubts about emotions as theoretical entities, everyday episodes of fear, anger, sadness, etc., remain what our theories of emotion are about” (p. 212). At the same time, following Meyer, Duffy, and others, I suggest we stop using the concepts of emotion, fear, etc. as tools for science. My



14.  Final remarks 289

statements on this topic created confusion, perhaps because of my analogies to constellations and élan vital. Such cases follow a pattern. Constellations: (A) The concept of an astrological constellation can be (indeed, has been) eliminated as a scientific concept from astronomy. (B) The stars that make up a constellation continue to exist, as does the geometric pattern among those stars when seen from our planet. (C) The folk concept of an astrological constellation continues to exist; it did and continues to influence the thoughts and actions of individuals and societies. This folk concept can be studied by, for example, historians. Élan vital: (A) The concept of élan vital can be (indeed, has been) eliminated as a scientific concept from biology. (B) The basic processes of life (metabolism, reproduction, etc.) continue to exist. (C) The folk concept of élan vital (even if not so named) can be found to exist in the minds of young children and some adults in many societies; this concept influences their thoughts and actions. The folk concept of élan vital can be studied by, for example, developmental psychologists. Emotion: (A) The concept of emotion can be eliminated (and I suggest we do so) as a scientific concept from affective science. (B) All the token emotional episodes continue to exist as a set of component processes. (C) The folk concept of emotion remains and influences thoughts and actions. It can be studied by, for example, cognitive psychologists and cultural anthropologists. One complexity exists in this case that does not exist with astrological constellations and élan vital. The folk concept as a concept can be part of and influence other parts of an emotional episode. In this case, the folk concept and its role are the subject matter of affective science rather than scientific tools. On elimination, I follow Bickle. Bickle pointed to a philosophical movement in the 1980s, Eliminative Materialism, that raised the spectre that folk psychology might be mistaken. (What? Our hunting and gathering ancestors might have made some mistakes?) Unfortunately, this movement became an “ontological dispute” on the existence of psychological states and processes. Bickle suggested a revised “eliminative materialism with a small e” as a metascientific rather than ontological issue on what concepts are useful to science. The question of causation can now be clarified. In eliminating the scientific use of the concept of emotion, I was eliminating emotion conceptualized as an entity that caused the components that make up the emotional episode. I meant to rule out using Sally’s fear, for example, to explain her heart beating, her palms sweating, her fleeing the bear, her conscious experience of fear, and so on. Averill countered that emotions do have causal powers, in that Sally’s fear can frighten her friend, her anger later that day can start an argument with the

290 James A. Russell

friend, and so on. Emotional episodes are real and have real effects. Still, a more fine-grained analysis is more helpful: The emotional episode consisted of a pattern of components in a context. Her friend perceived, not fear or anger directly, but the pattern of components in their context and used the folk concepts of fear and anger to interpret those components. So, here’s my analysis: the components, their pattern, the context, and the folk concepts influenced the friend’s thoughts and actions. Adding to that analysis one more event called fear or anger adds no new information or explanatory power.

5. Scarantino’s fear* mechanism Many readers likely share Scarantino’s belief that we can save the commonsense notion of emotions as causal entities while bypassing the problems with everyday folk concepts. His central question was, “Are discrete emotions genuine causal mechanisms with respect to the components commonly associated with them? … With respect to fear, for instance, we can ask: Does it bring about the heart rate changes, action tendencies, facial expressions and feelings commonly associated with it?” (p. 135). I answered no, Scarantino yes, provided we replace fear with a set of mechanisms such as fear*. The fear* proposal. Scarantino’s fear* is Panksepp’s fear with three revisions. First, fear* does not account for all and only cases called fear in English. The heterogeneity of the extension of the everyday folk concept of fear requires this revision. Scarantino said that heterogeneity rules out what he called the unique mechanism assumption – that is, one mechanism unique to all and only fear episodes – but does not rule out multiple causal emotion mechanisms, with different mechanisms producing different fear episodes. He wrote of “the presence of multiple emotion mechanisms corresponding to different instances of the folk concepts” (p. 148). So, different sets of token fear episodes stem from different fear mechanisms: presumably fear**, fear***, etc. Unfortunately, he does not tell us how many such mechanisms there are, what they are, which fear episodes each is meant to explain, or why these new mechanisms are limited to producing subcategories of fear. The number of such mechanisms is important: replacing fear with a vast number of distinct categories would not be an interesting scientific theory. It is not clear what the justification is for these new mechanisms being named by a term related to the English word fear, or emotion for that matter. For all we know, fear* etc. could be just the mechanisms I’ve suggested in Psychological Construction. Such questions are left for further research. Although Scarantino seems to defend fear as a causal mechanism, he does so by abandoning the concept of fear.



14.  Final remarks 291

Second, fear* alone does not cause the resulting components, but the rest of the brain and body are involved. This suggestion is surely correct. The brain and the rest of the body are the cause of all human episodes and therefore of fear episodes. Third, Panksepp’s fear circuit is the executive that coordinates all the other processes. Scarantino wrote that the fear mechanism “has a superordinate control function with respect to all organismic sub-systems available to the organism, including perception, memory and cognition. The job of the fear system is to coordinate an open range of resources by gating and modulating incoming sensory inputs relevant to behavioral and autonomic outputs and by engaging in reciprocal interaction with the brain networks involved in learning and higher thought, thereby bringing about secondary and tertiary affects” (p. 152). This revision is an empirical claim and likely false. The fear circuit has known specific effects and influences other parts of the brain, but it does not function as an executive. In short, the fear* proposal fails to specify just what the fear* mechanisms are (other than that they involve the brain and rest of the body) and what events they account for. Scarantino wants to keep a theory of emotions as causal mechanisms while ridding that theory of current concepts that specify just what the mechanism is and what events the mechanism explains. Traditionally, emotion, fear, ­anger and so on fulfilled both these roles. Scarantino therefore needs replacements for the folk concepts. Fear* is not fear, but what is it? What is its extension? (My proposal assumes that we have no need of emotions as causal mechanisms and therefore no need to specify the events they cause.) Emotions as causal mechanisms. Even if his fear* proposal has conceptual problems, Scarantino made a valid point: Problems with current folk emotion concepts cannot rule out a new theory with new concepts. Scarantino’s new theory is based on what he called the emotion mechanism assumption. That assumption is two claims: (A) each emotional episode is caused, and (B) the cause is an emotion. I agree with Claim A, but B is problematic. A does not entail B. It is unclear how B can be formulated without the folk concept of emotion or just what is the “emotion” in Claim B. Scarantino appears to abandon and then resurrect the everyday concept of emotion. Perhaps “emotion” is something other than the everyday concept, but what? Claim B is vacuous if “emotion” is just the emotional episode, whatever happens to causeany one token emotional episode, or the entire brain and body. Scarantino conflated Claim A with Claim B. For example, he thought that I had admitted the existence of an emotional causal mechanism for fear. In my fictional story of Sally and the bear, the Panksepp fear circuit produced some of the components of Sally’s emotional episode. Scarantino found this “admission

292 James A. Russell

quite surprising” (p. 151) because he saw my doing so as positing a causal fear mechanism – Panksepp’s fear circuit. But there is no reason to call this circuit the emotion or the fear mechanism, and Panksepp did not. Calling it fear does not make it so or add any information to its known effects. ThePanksepp fear circuit produces freezing, inhibition of behavior, and changes in the autonomic nervous and endocrine systems. The fear circuit is indeed a causal mechanism, but its effects are not limited to fear or even to emotions. My surprising admission is nothing more than Claim A. It is not Claim B. Still, Claim B – that emotions cause emotional episodes – has a strong intuitive appeal, which we need to diagnose. Scarantino articulated these intuitions well. In his first argument, he re-examined the story of Sally and the bear. Most people have a strong intuition that fear made Sally react the way she did. On my account, however, each component must be described separately and each has its own causal story in which fear plays no part. Once each component is explained, there is no need to posit fear as the cause. My explanation, he argued, makes co-occurrence of the various components and their coordination a mere coincidence. His explanation for co-occurrence and coordination is that fear (or fear*) did it. Co-occurrence and coordination are real and not coincidental, but fear or fear* are not useful explanations. Four points: First, on co-occurrence, the Sally-and-the-bear story was selected (created actually, but set that aside) to match closely the prototype script for a fear episode. Imagine that we select a Royal flush, and then ask why all the cards are of the same suit. Of course, the specific Royal flush we examine was produced in a causal chain. But, the causal chain was not aimed at a Royal flush. The answer to the question of why is not some Royal-flush-causing mechanism churning out cards of the same suit. Cards of are the same suit because we selected a card hand with that property. We see various components co-occur in prototypical cases of fear because we are looking at prototypical cases. Second, co-occurrence depends on how the components are defined. When the components are described broadly enough, they frequently co-occur. For example, if the action component is defined as “changing one’s relation to the environment,” then that component is almost always occurring and therefore co-occurring with other components. A person is almost always engaged in appraisal processes and behavioral processes, and the autonomic nervous system is always active. So, defined broadly enough, the components are almost always co-occurring, not only in fear. When components are defined narrowly, then fear provides no explanation. If the action component is defined as running, we know that fear does not always involve running. If the appraisal is defined as seeing danger, we know that danger does not always result in fear. What if fear* replaces fear? Perhaps the hypothesis of fear* assumes that there is a point



14.  Final remarks 293

between broad and narrow that defines components in such a way that they co-occur in all and only fear* episodes. Until fear* is defined, this possibility is difficult to evaluate. Third, on coordination, coordination among the components of Sally’s reaction was not a coincidence. They were coordinated in response to her specific circumstances, understanding of the situation, resources, background, skills, and so on. Her autonomic nervous system prepared her to run, but doing so is not unique to fear: similar changes would occur running from the bear, running furiously into battle, running a race, running to burn calories, running to entice someone to follow, or running for fun. Her perceptual processes were coordinated with her behavior, but such coordination is a general feature of perception and behavior. Coordination is a general property of animal behavior. Fourth, co-occurrence and coordination in Sally’s reaction to the bear do not imply (a) a single cause of all co-occurring components, (b) a single mechanism coordinating them, (c) that the same processes cause and coordinate other instances of fear or fear*, or (d) that it is meaningful to think of the coordinating mechanism as an emotion. (Of course, if “single mechanism” is taken to be the entire brain, then there is a single mechanism, but its name is mindbrain rather than fear orfear* or emotion.) Because folk concepts of emotion and fear have been abandoned in Scarantino’s account, the coordinating mechanisms he suggests could be just the mechanisms I suggest, with the only difference being their name. Scarantino’s second argument for Claim B concerned relations among components not in a selected particular case, but for all cases. He wrote: “Positing an internal emotion mechanism, I conclude, continues to be the best explanation for the systematic relation between classes of stimuli (e.g. dangerous stimuli) and classes of componential changes (e.g. heart pounding, palms sweating, fear expressions, freezing and fleeing)” (p. 151). In this passage, Scarantino relied on our intuition that there are strong systematic correlations across cases to be explained. When he wrote that my account fails “to provide any understanding of why the correlations among components take the form they do” (p. 150), he is assuming that the correlations are large. Our mental script for fear and other emotions leads us to anticipate strong correlations. That is why the empirical results on this issue are so surprising. Correlations between components turn out to be much weaker than anticipated. Components are not completely independent, but the weak correlations that exist can be explained without positing an “internal emotion mechanism.” And I offered suggestions for doing so. (Might the correlations be higher within the set of fear* episodes? Of course, new narrower categories could be defined such that correlations are higher, but the usefulness of doing so remains to be seen.)

294 James A. Russell

Conclusion. Scarantino’s proposal abandons folk concepts as scientific tools and essentialist assumptions about those concepts – as does mine. Scarantino’s logical point is well taken: From the properties of current everyday emotion words and from accumulated evidence now available, we cannot disprove anew theory with new concepts. That new theory would specify the mechanisms that cause the thoughts, feelings, and action components of newly defined classes of token emotional episodes. Still, to turn this logical point into a viable scientific theory, it would have to be specified in more detail through research. The possibility of new concepts embedded in a new theory is not a defense of current concepts and current theories. In the end, Scarantino’s proposal entails abandoning current assumptions and replacing them through empirical research – as does mine.

6. Affect and core affect Consciousness is as natural as life, but has proven even more difficult to study scientifically. Emotion has been similarly difficult. So, the combination of the two – emotion in consciousness, which Panksepp and I agreed to call affect – cannot be easy. I offered three concepts about affect: Emotional Meta-experience, perception of affective quality, and Core Affect. All three are hypotheses, subject to change as the evidence comes in. If evidence dictates, each of these three might be replaced with other concepts within Psychological Construction. Conversely, each could be placed within a different broad framework. The history of science suggests that all the specific hypotheses I have offered within Psychological Construction will eventually be replaced. Emotional Meta-experience and perception of affective quality are both hypothesized to be perceptions, and so we can at least bring theories of perception to bear on them. Core Affect is special in that it is the one process that is not handed over to another branch of science. It is the most primitive of the three. It cuts across the traditional distinctions of feeling, mood, and emotion. Because it is typically experienced in combination with other ingredients, it is elusive. It proved in this exchange to be the most controversial of the three. Core Affect is a name. I coined the name to be able to express a specific concept. Our discussion is not clarified by using the name Core Affect for other concepts. Panksepp and others took the word core to be an empirical claim, perhaps meaning foundational, important, primitive, necessary, causal, or innate. If so, one could dispute the claim and point to evidence that hunger or something else is core. But, the word core in Core Affect is not a claim. Core Affect is ever-present (and Northoff had some interesting thoughts on that feature) and therefore occurs outside as well as inside emotional episodes.



14.  Final remarks 295

When inside, it provides the hedonic tone and subjective sense of energy to an emotional episode. Still, although Core Affect can be part of an emotional episode, it is not equivalent to, not a necessary feature of, not sufficient for, not always foundational to, an emotional episode. There is more to an emotional episode than Core Affect. For example, Core Affect plays a role in selecting goals and plans during an emotional episode. De Sousa remarked that “playing a role” is not “determining.” Exactly. The conscious feeling of Core Affect is one feeling, even though Core Affect is described in terms of two dimensions. De Sousa may be right that the two dimensions are not completely independent of one another, especially at the extremes. Still, neither dimension can be reduced to the other (Kuppens et al., submitted), and so the descriptive model of Core Affect is two-dimensional. One dimension of Core Affect, pleasure-displeasure, is commonly called va­ lence. De Sousa reminded us of the many uses of the word valence. By valence, I do not mean Core Affect’s moral value, adaptive value, etc. I do not equate valence of Core Affect with approach-avoidance or with reward-punishment, although they are empirically linked. Rather, by valence of Core Affect, I mean a simple nonreflective feeling good or bad. So, I agree with de Sousa on the many uses of the word valence, but my use was limited. Even restricted to a good or bad feeling, valence is controversial. I conceptualized valence as a single bipolar dimension, because this conceptualization best accounts for available data (Yik, Russell, & Steiger, 2011). There are other hypotheses and the jury is still out. Some researchers hypothesized and offered evidence for a bivalent view in which a pleasant feeling can co-occur with an unpleasant feeling (Norman et al., in press). The debate between bipolarity and bivalence has advanced research on affect in interesting ways. On my account, pleasure is a single process. Thus, all pleasures are at bottom drawing on that one process. This proposal is controversial. Averill proposed qualitatively different types of pleasure. De Sousa suggested that if unconscious Core Affect is possible, then qualitatively different types are also possible. ­ Panksepp proposed that each circuit produces a qualitatively different pleasure. ­ Ortony, Clore, and Collins (1988) offered three types, depending on the object of the emotion. Or there could be an uncountable number of qualitatively different pleasures: the pleasure of tasting caviar, the pleasure of listening to Gould play Bach, and on and on. So, is there one pleasure or many qualitatively different types? I find it more parsimonious to think of the source of the pleasure as one thing and the pleasure itself as another. The source and the pleasure combine to form a percept: Attributed Affect. A percept such as the pleasure of caviar is one gestalt experience, but composed of the source, the pleasure, and an attribution linking the two. Attributed Affect is an intentional state in the philosophical sense and

296 James A. Russell

more complex than Core Affect. The distinction between the source and the pleasure is needed to allow pleasure to serve its function of being a common currency, a means of comparing the costs and benefits of qualitatively different sources and options (Cabanac, 1971). De Sousa pondered the pleasure of tasting caviar: is pleasure an intrinsic part of that experience or is it a separate atom added to other atoms to form a molecule. The concept of Attributed Affect suggests that de Sousa’s two options are not mutually exclusive. Pleasure can be an atom that is added to other atoms to form a new single token experience. The final experience – the pleasant-taste-ofcaviar – is phenomenologically a single experience, but all the same, a gestalt, a meta-experience, the pleasantness of which is “intrinsic to the very quality of that experience” (p. 227).

7. Appraisal theories Moors distinguished two types of appraisal theory. The first type (portrayed in her Figure 1 top panel) is inconsistent with Psychological Construction but consistent with most basic emotion theories. (One exception is Panksepp whose primary and secondary emotions have no appraisal process.) Most early appraisal theories were of this first type, as are those basic emotion theories to which has been added an appraisal step (Ekman, 1992). In a theory of this first type, appraisal is described as a set of appraisal variables (goal relevance, agency, evaluation, etc.) used to analyze the emotion-eliciting stimulus. These theories posit a limited number of patterns among the appraisal variables. Each pattern, in turn, triggers a different discrete emotion (typically specified by an English word). Thus, differences in appraisal are what differentiate among different types of emotion. The immediate cause of an emotion is an appraisal, different individuals can appraise the same stimulus differently, and appraisal is necessary for an emotion. The emotion is the cause of the patterned set of observable behaviors and physiological changes. Thus, a causal chain is envisioned: stimulus, appraisal, emotion, and behavioral and physiological changes. Because appraisal is necessary, any feedback mechanism that alters the emotion must operate via re-appraisal. Moors portrayed the second, often more recent type of appraisal theory as inconsistent with basic emotion theory, and I find it complements Psychological Construction. (Appraisal theorists might not agree with me, however.) Leddy, Robertson, and Schulkin’s version, for example, resonates nicely with Psychological Construction. Moors’ portrayal of this type (Figure 1 bottom panel) shows no discrete emotion mediating between the appraisal and the behavioural-physiological­ components. It shows not a set of hypotheses but all possible ­connections between



14.  Final remarks 297

appraisal variables and behavioural-physiological components – much as I have done when characterizing how components are linked. There is no limit to the number of patterns among appraisal variables. Appraisals are not necessary for an emotional episode, but are “typical.” Some emotional episodes occur without the appraisal, and some appraisals occur without an emotional episode. Appraisal theories of the second type are compatible with Psychological Construction. The two differ only in focus. I suggested processes (such as Core ­Affect, somatosensory feedback, context, and planning and goal setting) that are not processing of the eliciting event. Emotional episodes can occur with no appraisal and are influenced by events, such as drugs, other than the eliciting event. Such claims are consistent with appraisal being typical rather than necessary. Appraisal theorists emphasize appraisals and have offered detailed hypotheses and supporting data. I emphasize components, one of which is an appraisal, and I have not offered details. I wrote of sensory-perceptual-cognitive processing of the eliciting event, which includes not only the appraisal variables, but also anticipation, attention, attribution, categorization, interpretation, perception of affective qualities, remembering, simulation, and so on. I find no clear boundary between appraisal and other cognitive processing of the eliciting event. Once “necessary” is replaced with “typical,” I see no reason appraisal theorists would in principle need to deny any of my proposals. Moors’ suggested substantive differences between this second type of appraisal theory and my proposal dissolve upon close inspection. For example, whereas these appraisal theories have each appraisal variable potentially influencing all behavioral components and I never mentioned that possibility, I should have. I did not specify causal mechanisms for the various components, but I hope to someday. Appraisal theories anticipate stronger links between appraisal variables and behavioral outcomes than I do, but strength is a matter of degree and an empirical question, the answer to which need not change either framework. I said that we need to establish links empirically before assuming them in our theories, but having specific hypotheses is valuable.

8. Social construction Averill defended a social constructionist view. Social Construction must be distinguished from Psychological Construction, and, in certain formulations, the two could be inconsistent. Any assumption that emotional episodes are pre-assembled­, through, for example, social norms, roles, concepts, or language would be at odds with my account. All the same, the two can also be formulated as complementary, as Averill did.

298 James A. Russell

Averill noted that my account did not include a social-cultural level of analysis. Guilty. I am working on it. Components of emotional episodes are influenced by concepts, assumptions, beliefs, practices, roles, norms, feeling rules, ideals, and so on – all of which are influenced by culture. The individual confronts a social ecology: a series of events, situations, and relationships that vary from one society to the next. The individual lives in a social reality: laws, money, universities, and borders between countries exist only in the sense that a group of people agree that they exist. Averill found my analogy with astronomy and astrology dismissive of everyday emotional episodes. Not guilty. I do distinguish between emotional episodes and beliefs about emotional episodes. In my analogy with astronomy and astrology, the science of actual emotional episodes is like the real science of astronomy, not the pseudo-science of astrology. Astrology and beliefs about emotion are the subjectmatter of a different branch of science. It is for those who study beliefs. Averill took my term Emotional Meta-experience to be dismissive. Not guilty. By meta, I did not man epiphenomenal. I meant that the experience of fear, for example, is not a simple quale, unanalyzable and universal. Rather, it includes other experiences, such as a perception of the antecedent, somatosensory sensations, plans and desires to behave, etc. It is not beyond science, not a secondary or inferior topic, and not less real than Core Affect. Averill defined emotions as temporary social roles and dispositions. The set of emotions is too heterogeneous to be captured by either definition. All the same, temporary social roles are one of the mechanisms that shape many emotional episodes. Averill’s work on this topic has been pioneering. As to dispositions, the word emotion sometimes refers to a disposition. Love and hate, for example, are prototypical emotions and are mainly – but not always – dispositions rather than occurrent events. We can also use fear, anger, and other emotion words to refer to dispositions, in the sense of a tendency to respond in certain ways as circumstances arise; and tendency is established by a summary of past responses. For example, to say that Jim is afraid of spiders is to say he tends to think, feel, and act in certain ways when spiders appear. We say so because, in the past, he has frequently reacted to spiders that way. In analyzing occurrent emotional episodes, however, I have not seen the usefulness of thinking of them as dispositions. Disposition as summary is not relevant to a token occurrent event. Still, we might say that Sally being frightened last Saturday morning is a disposition. That is, she was disposed to act, think, and feel in a way that resembles a fear script. This analysis is roughly equivalent to the one I offered, except that calling it a disposition runs the danger of implying that the disposition is a causal entity inside Sally.



14.  Final remarks 299

When Averill wrote that social norms of emotion allow improvisation, we’re back to Psychological Construction. Averill wrote that what societies construct, individuals reconstruct. If so, we need to understand how the individual improvises the specific token emotional episode. (The word improvises in the last sentence might suggest a conscious deliberate process, but of course I mean no such thing.) Even given a script, real actors on a real stage have latitude and must construct their character. A social-cultural analysis is needed to describe those structures that exist in the society in which the individual resides. But something like Psychological Construction is needed to describe how those structures are incorporated into the token emotional episode that the individual improvises.

References Cabanac, M. (1971). The physiological role of pleasure. Science, 173, 1103–1107. Duffy, E. (1934). Is emotion a mere term of convenience? Psychological Review, 41, 103–104. Ekman, P. (1992). An argument for basic emotions. Cognition and Emotion, 6, 169–200. Ekman, P., & Friesen, W. V. (1971). Constants across cultures in the face and emotion. Journal of Personality and Social Psycholog, 17, 124–129. Fehr, B., & Russell, J. A. (1984). Concept of emotion viewed from a prototype perspective. Jour­ nal of Experimental Psychology: General, 113, 464–486. Fridlund, A. J. (1994). Human facial expression: An evolutionary view. San Diego, CA: Academic. Gendron, M., & Barrett, L. F. (2009). Reconstructing the past: A century of ideas about emotion in psychology. Emotion Review, 1, 316–339. Kim, J. J., & Fanselow, M. S. (1992). Modality-specific retrograde amnesia of fear. Science, 256 (May), 675–677. Kuppens, P., Tuerlinck, F., Russell, J. A., & Barrett, L. F. Submitted. The relationship between valence and arousal in subjective experience. Lindquist, K. A., Wager, T. D., Kober, H., Bliss-Moreau, E., & Barrett, L. F. In press. The brain basis of emotion: A meta-analytic review. Behavioral and Brain Sciences. Maier, S. F., & Watkins, L. R. (2010). Role of the medial prefrontal cortex in coping and resilience. Brain Research, 52–60. Mehrabian, A., & Russell, J. A. (1974). An approach to environmental psychology. Cambridge, Mass.: MIT Press. Meyer, M. (1933). That whale among the fishes – the theory of emotions. Psychological Review, 40, 292–300. Murphy, F. C., Nimmo-Smith, I., & Lawrence, A. D. (2003). Functional neuroanatomy of emotions: A meta-analysis. Cognitive, Affective & Behavioral Neuroscience, 3, 207–233. Norman, G. L., Norris, C. J., Gollan, J., Ito, T. A., Hawkley, L. C., Larsen, J. T., Cacioppo, J. T., & Berntson, G. G. In press. Current emotion research in psychophysiology: The neurobiology of evaluative bivalence. Emotion Review. Ortony, A., Clore, G. L., & Collins, A. (1988). The cognitive structure of emotions. NY: Cambridge University Press.

300 James A. Russell

Phan, K. L., Wager, T., Taylor, S. F., & Liberzon, I. (2002). Functional neuroanatomy of emotion: A meta-analysis of emotion activation studies in PET and fMRI. NeuroImage, 16, 331–348. Rachman, S. (1977). The conditioning theory of fear-acquisition: A critical examination. Be­ haviour Research and Therapy, 15, 375–387. Russell, J. A. (1980). A circumplex model of affect. Journal of Personality and Social Psychology, 39, 1161–1178. Russell, J. A. (1991). Culture and the categorization of emotion. Psychological Bulletin, 110, 426–450. Russell, J. A. (1994). Is there universal recognition of emotion from facial expression? A review of the cross-cultural studies. Psychological Bulletin, 115, 102–141. Russell, J. A. (2003). Core affect and the psychological construction of emotion. Psychological Review, 110, 145–172. Russell, J. A., Bachorowski, J. A., & Fernandez Dols, J. M. (2003). Facial and vocal expression of emotion. Annual Review of Psychology, 54, 329–349. Russell, J. A., & Fernandez-Dols, J. M. (Eds.) (1997). The psychology of facial expression. New York: Cambridge University Press. Shewmon, D. A., Holmes, G. L., & Byrne, P. A. (1999). Consciousness in congenitally decorticate children: Developmental vegetative states as self-fulfilling prophecy. Developmental Medicine and Child Neurology, 41, 364–374. Worcester, D. A. (1933). In defense of the whale – emotion is at least a term of convenience. Psychological Review, 40, 478–480. Yik, M., Russell, J. A., & Steiger, J. H. (2011). A 12-Point circumplex structure of Core Affect. Emotion.

chapter 15

My reflections on commentaries and concluding perspectives Jaak Panksepp

I thank all the commentators for sharing their perspectives, and for Peter Zachar’s synopsis of their reflections. I will address a few key issues that may help bring our thinking into some kind of consilience. My main research goal has been to provide some understanding of the foundational neural networks that, to the best of our knowledge are homologous in all mammals because of shared genes from common descent. Such an analysis of mammalian emotionality may be essential for lasting progress in the field, but it will not answer the higher-order issues many commentators discussed. Hopefully the dialog between Jim Russell and myself, along with the many thoughtful commentaries and editorial reflections, will bring us nearer to a shared appreciation of the key issues: (i) clear enunciation of the relevant empirical questions that need to be asked; (ii) specification of the best empirical ways to answer those questions; and thereby (iii) clarification of how it all happens in the brain, along with an appreciation of the limits and flaws of our various experimental and conceptual efforts. Before addressing the commentaries, I would briefly review my own framework for studying how subjective affective experiences arise from objective emotional arousals. The central empirical and conceptual issue that has motivated me is a scientific understanding of the evolved neural infrastructures of our affective lives – especially how emotional feelings are constituted by brain activities. My work assumes that scientific clarification of phenomenal consciousness requires brain research on appropriate neural networks – without that, our primal feelings of anger (rage), fear (fear) and grief (panic) – the various intrinsic negative emotional feelings of our lives – cannot be understood. The same goes for our feelings of general desire (seeking), sexual urgency (lust) social nurturance (care) and playfulness (play) – gifts of nature that stir people and animals toward the positive affective poles of mental existence. Because of my evolutionary convictions, I chose to study the primal neural controls of natural (instinctual) emotional behaviors in other animals, hoping insights about the human condition would

302 Jaak Panksepp

emerge. I think many have, but nothing close to a complete picture. I remain surprised, that such bottom-up approaches are not as widely deployed in psychology as are top-down strategies. But that is because primary-processes are inaccessible in standard human research. With few exceptions, those who have worked on animal emotions and motivations have also been little interested in discussing or studying the nature of primal affective experiences. Fortunately, increasing numbers of human investigators have thought it was an important problem, and Jim Russell has been among the most creative thinkers in the field (e.g., Damasio, 1999; Denton, 2006). But that work has yielded largely descriptive and correlative studies, which cannot access “constitutive” issues – namely the topic that interests me most. Thus, our conceptual perspectives simply have not overlapped until Jim Russell postulated a Core Affect process in the brain. However, I believe that cross-species functional brain research remains the major approach to understanding how affects are constituted within the mammalian BrainMind (e.g., Hess, 1957; Panksepp, 1971, 1998; Siegel­, 2005). And that requires bottom-up thinking. A causal analysis is simply not achievable either through human brain imaging or semantic-conceptual-cultural­ top-down oriented approaches. Clearly, the brain is an evolutionarily layered organ, and primary processes are elaborated by lower, subcortical brain systems. That has been clear for a long time. Jim Russell’s work is at a very different level, where the abundant data and concepts have no clear relevance for understanding the neural nature of affective experiences. Hence, in this concluding section I will primarily focus on my own position and the reflections of the commentators. With regard to understanding the maddeningly difficult affective components of emotional arousals, for me the break-through facts, and insight, were that wherever in the brain one evoked coherent emotional behavior patterns with localized subcortical brain stimulation (i.e., via generation of “unconditioned emotional responses”), one could demonstrate that animals treated those evoked states as “rewards” and “punishments” – the primary initial empirical way to access experiential states in other animals. Indeed, this discovery seems to be as close to a lawful relationship in emotion studies as currently exists. A reasonable corollary of this finding is that such emotional “rewards” and “punishments” may modify learning largely because neurally-based affective shifts are probably foundational for the “reinforcement” processes of the brain. Surprisingly, this remains a minority view among those interested in the neural basis of emotional learning. To this day, most behavioral neuroscientists use concepts such as “the process of reinforcement” and “reward prediction errors” (Schultz & Dickinson, 2000) with little discussion of affective issues in the animals they study. Few explicitly express any interest in the nature of experience, which diminishes needed bridges between the animal and human work.



15.  My reflections on commentaries and concluding perspectives 303

Most of the controversies that have come to my doorstep during my career have reflected my use of “basic emotion” terminologies and the difficulties that investigators have with the notion that raw affective experiences may arise largely from primary-process action-generating states of the nervous system. Most researchers seem to believe that all experiences must reflect higher-brain functions and do not arise from subcortical action networks that are presumably deeply unconscious. In any event, most who are thinking about the nature of experiences are working at tertiary-process levels, where experience is more easily linked to sensory than to motor/action processes (leading to perception-action visions of the brain). I suspect that in a hierarchically organized organ like the brain we may need to understand the foundational importance of certain types of primaryprocess action urges for allowing higher brain mechanisms to operate – namely action-perception controls that guide the developmental transitions from raw affective experience to learning and thinking. Indeed, with our increasing appreciation of the evolutionary layering within the BrainMind we can envision that many of the apparent disagreements among Russell and myself, perhaps also the modest ones with commentators, reflect people working and thinking at quite different levels of the BrainMind axis. My position, I believe evolutionarily and empirically supported, is that what happens at the primary-process level will surely inform what happens at secondary- and tertiary-process levels, especially early in organismic development. In contrast, top-down controls – from tertiary to primary – only emerge through cognitive brain maturation along with many poorly understood epigenetic processes. Thus, to have intellectual skirmishes between my version of primary-process affective neuroscience and Jim Russell’s tertiary process research program on human emotional concepts makes little sense from my empirically constrained point of view. The positions are incommensurate. Thought-mediated emotional processes and understandings are bound to be dramatically more complex than evolutionarily mediated ones. However, without clear visions of primaryprocess­ evolutionary and neuroscientific underpinnings, tertiary-process­ levels of understanding may remain forever debatable, and may generate discourse at cross-purposes­ (as I think it long has). Perhaps these qualifications will help readers to decipher why I respond to many of the fine ideas of commentators the way I do. In the following section I will attempt to focus on the reflections of the eight commentators, and will order my remarks in the approximate sequence of “criticality” used in two senses of the term: (i) raising the types of issues that all emotion researchers need to consider, and (ii) ones that do not agree with my position. Thus, in line with the intrinsic ambiguity of vernacular language, the most “critical” commentaries will come first, followed progressively with those that are increasingly supportive of my overall position. As a result (and to

304 Jaak Panksepp

minimize­ overlap) my reflections will typically get shorter as we proceed. I will also tend to focus on issues that interest me most, and are of greatest importance from my perspective. It is not possible to deal with point-by-point issues in the limited space available.

1. Reflections on the commentaries Andrea Scarantino offers a fine and in-depth analysis of how terms and concepts in emotion studies may be optimally deployed to minimize conceptual confusions. I basically agree with most of his analysis, although I am tempted to quibble with many details. I would like this first response to be a general rejoinder to most of the commentaries. As Zachar puts it “Scarantino agrees with Russell that the scientific search for single homogeneous essences that are causally responsible for discrete emotions has been a failure.” I also agree but am rather puzzled that this perspective even needs to be raised, for practically all “basic emotion” approaches have been premised on behavioral/psychological analyses and they certainly have not even tried to discuss or search for “single homogeneous essences”. Indeed, I have never imagined that I, or anyone else, was searching “for single homogeneous essences” (where “single” – in the world of neuroscience – must be deemed an irretrievably problematic concept). However, being the only main neuroscientific representative of the illusory homogeneous “tribe” of “basic emotion” investigators, I trust no one is suggesting my research program has been a failure. Surely, all ongoing work needs to be critiqued from both empirical and conceptual perspectives, so I deeply welcome Scarantino’s reflections on the various primary-process terminological/conceptual issues I have raised (but I simply can’t be tempted to use terms such as the “the WIS4 affective system”; my commitment to practical communicatory clarity as opposed to any attempt at impossible precision was made a long time ago (Panksepp, 1982). Let it suffice that I am perhaps the first investigator of brain emotional networks to recognize he was working at the primary-process level. In that role, I hope I wisely selected a heuristic nomenclature that can facilitate further thinking at all relevant levels, from primary to tertiary to philosophic, and which can promote higher levels of analysis, including psychiatric (we may seem a long way off, but some work in that direction has begun – e.g., Burgdorf, et al., 2011; Coenen, et al., 2011; Panksepp, 2004; Watt & Panksepp, 2009). My perspective is as follows: I do not believe there is a unidimensional ­essence-type relationship between the labels I have chosen to describe at least seven distinguishable BrainMind emotional/affective processes and underlying brain networks. Obviously, all emotional systems have many shared processes,



15.  My reflections on commentaries and concluding perspectives 305

including several general arousal regulating systems (e.g., norepinephrine and serotonin), and a host of other shared substrates within the subcortical reticular networks essential for generating distinct types of emotionality. That there have to be some shared underlying systems for all emotional functions should be a foundational assumption to which all subscribe, yielding maybe even something as subtle as the core-self (Panksepp, 1998). Talking clearly about functional systems within massive brain complexities is surely a challenge. Perhaps I should have avoided the vernacular completely in preference for some totally artificial (i.e., intrinsically meaningless) terminology, but I chose not to for several hopefully good reasons (discussed since Panksepp, 1982): (i) Premature artificial terminologies tend to become idiographic – unique to individual investigators, with no universal appeal. (ii) Artificial terminologies have little didactic impact, and hence do not promote thinking about the underlying issues, without layers of potentially artificial conceptual analysis. (iii) Finally, I was most interested in the role of primal emotional imbalances for understanding psychiatric disorders, and hence preferred terms that could facilitate such linkages. The capitalization of terms is the cue that I deployed to help assure that the sins of vernacular usages – i.e., all human words are cultural inventions, subject to mereological fallacies when used by scientists – were not as evident as they currently are in the non-neurological versions of “basic” emotion approaches (which Jim Russell has cogently criticized). Time will tell whether these labels will have any traction, but I do not agree that my choice of primary-emotion system terminology places my work in any type of narrow essentialist camp. As we all should know, such circuits show intrinsic inter-individual and inter-“species” allelic variability, as well as various poorly understood epigenetic modifications. However, these systems are as “real” as the very similar bodily organs and skeleto-muscular structures of vertebrates across many species. I simply need to visit the Comparative Vertebrate Anatomy Museum of our college to keep in mind the diversities among crossspecies homologies, and how little we really know about even the fine anatomy of brain systems. Still, it is remarkable how much we can functionally generalize across species despite the obvious complexities and divergences. One can easily conclude that in medicine, physiology and biochemistry we have learned more about the human body from studying the bodies of other mammals. Such generalizability is surely in line with modern evolutionary theory, hopefully without any of the simpleminded species essentialism that has been aptly criticized in Richard Boyd’s “homeostatic property cluster model” as well as in many other modern philosophical critiques such as those of David Hull, Paul Griffiths and Robert Wilson, to name a few. To paraphrase Gertrude Stein: “A heart is a heart is a heart” (“things are what

306 Jaak Panksepp

they are”)…despite endless differences in detail. My terminology reflects Shakespearian recognition that “A rose by any other name is still a rose” much more than any catalog of detailed differences can achieve. Words do allow us to conceptually “image” and certainly imagine the vast complexities that lie ahead – e.g., the many inter-systemic interactions (see, Figure 2 of Panksepp, 1982, and Figure 2.5 of Panksepp, 1998), few of which have been empirically addressed (although historically and philosophically they have been extensively discussed). Of course the selection of terms can often freeze our thinking. I trust that my choice of terminology sustains the communicatory robustness needed for pursuing explanatory integrity (i.e., searching for new primary-process neuroempirical findings in affect sciences) while not impeding the flexibility needed to discuss key human issues in psychiatrically meaningful ways. I thank Scarantino for raising such important conceptual issues, but trust that as he becomes more conversant with my empirical work, he may agree not only with my choice of approaches but also the suggested labels. I suspect we all agree that development of standard nomenclatures in this pre-verbal realm of mental life is a challenge we must face together… otherwise we will continue to talk past each other. In any event, I sought to anchor critical concepts in demonstrable brain functions. As we come to understand primary emotional processes of mammalian brains (with many overlapping sub-components), we can begin to entertain some linkages to higher issues. For instance, our parsing of affective space, and the resulting cross-species work, has already promoted new ideas for re-balancing human affective symptoms in autism, depression, and explosive anger-type “disorders” (e.g., Moskal, et al., 2011; Panksepp, 2005a; Panksepp & Watt, 2011). There are comparable signs of such cross-species fertilizations emerging in the study of oxytocin effects in humans (for perspective taking, see Panksepp, 2009), and in translating our early understanding of opioid mediation of the psychic pain arising from social loss (Panksepp, et al., 1978), to more complex human feelings and social relationships (e.g., Eisenberg, 2011). I trust that all of us in this cross-disciplinary and multi-dimensional discussion can agree that the scientific legitimacy of the cross-species primary-process analysis of mammalian emotions is critical for any foundational understanding of human emotions. We simply can’t study such issues in any neuroscientific depth in human beings. Conversely, we cannot study such issues in any psychological-subjective depth in animals. But we can work synergistically to figure out how diverse affective “reward” and “punishment” principles operate in human brains, and how they generate affects (Panksepp, 2005b, 2007) through neuroscientific analyses of non-human animals. Scarantino’s superb analysis tells us much about the difficulties that we should not evade and the dilemmas we cannot currently solve.



15.  My reflections on commentaries and concluding perspectives 307

Agnes Moors provides useful systematics for guiding the discussions among investigators taking different perspectives – from cognitive “constructions” and “appraisals” to “affect programs”. The utility of this approach is that it can provide equal weight to all interacting levels of analysis without prioritizing any, which is essential for a reasonably comprehensive top-down approach. However, from a neuro-evolutionary bottom-up approach, such as mine, a key question is whether higher-order systemic functions can coherently operate without support of the more primal levels? I doubt it. Primary-processes are of unique importance in understanding higher developments. My understanding of the nervous system is that the earlier foundational levels always provide a platform for what comes later. Without clear visions of evolutionary foundations, cognitive activities that reflect higher levels of processing cannot be well understood. Hence, my view is that the fundamental organization of the lower levels does have some priority in mental coherence. Thus, when Moors suggests that the appraisal and constructivist approaches have the resources to address more emotional states, she is conceptually correct. But this can lead to a scientifically disadvantageous situation where conceptual understanding, working within semantic hermeneutics, substitutes for adequate neuro-empirical knowledge. Scientific constructivism may remain empirically crippled unless investigators learn to link their more complex conceptual levels more fully to evidence-based primary-process perspectives. We all should agree that an understanding of the full package is bound to be better than any of the individual levels of analysis! With respect to Moors’s extensive discussion of interactions among levels of control, I would ask: (i) Can a coherent understanding of subcortical emotional systems be achieved without the higher levels of analysis? My answer is yes, based, for instance, on the study of animals radically neo-decorticated soon after birth (e.g., Panksepp, et al., 1994). (ii) Can a coherent understanding of the higher order aspects, of emotional life be achieved without a solid understanding of the subcortical foundational issues? My answer is no, based on how severe damage to lower brain regions, but not vice versa, can compromise consciousness. Obviously, we are rarely scientifically well positioned to understand the full complexity of many interacting systems, and in a sense that is the project of constructivism. Why it has sought to wage any battles with my perspectives, whose goal is to address delimited questions with potentially widespread ramifications, puzzles me. I work within the intrinsic constraints of cross-species empirical approaches, and perhaps should not generalize to humans where so little primary-process understanding has been achieved. But my goal was understanding the evolutionary foundations of human emotions, and enough work has now been done to give us confidence that we are “just” another species of mammal at the primary

308 Jaak Panksepp

emotional­ and affective realms. Despite that fact that good primary-process research is incredibly hard to do in humans, I think we can be confident that such primal manifestations of affect operate within our higher levels of mentation. Leddy, Robertson, and Schulkin provide their own overall vision of the emotional universe. They recognize, as does every investigator of emotions, that in the final accounting emotions need to be understood as interactive processes at many levels. They also highlight especially the interactive power of primary and secondary processes, as do I, as well as their influence on diverse tertiary cognitive processes. Perhaps it needs no more emphasis that, during development, ­bottom-up controls prevail while with maturation, top-down controls become ever more influential. But that is a critical perspective for anyone interested in higher emotional processes in humans to consider. Too many working “at the top” tend to ignore bottom-up issues. Although primary and secondary processes cannot be neatly separated in the intact brain, some clear separations can be achieved in simplified preparations (e.g., radical neo-decortication and various disconnection syndromes as can be achieved through lesions of various basal ganglia such as radical amygdalectomies). Indeed, the view that emotional unconditioned responses (e.g., arousal of the fear and panic system) are both of critical importance in the emergence of distinct anxiety disorders and aversive state conditioning needs much more attention than it has yet received (e.g, Pankepp, 2011; Panksepp, et al., 2011). If I have ever intimated that “cognitive … appraisals are of secondary importance” compared to “primordial affects” in humans, I wish to retract that as a valid descriptor of my perspective. My position is simply that subcortical affect generators lead to much more intense feelings on their own (i.e., relatively pure affects) than feelings generated within the neocortex (i.e., networks that integrate appraisals and object-relations). It is easy to conflate such nuances, based apparently on our potential differences in what should be deemed to be “cognitive” as opposed to “affective.” I see all “cognitions” as being critically dependent on information-processing based on sensory/perceptual inputs coming into the nervous system from outside the brain, while primal emotional affects reflect intrinsic within-brain potentialities of instinctual brain networks, concentrated subcortically. Of course, every emotion theorist accepts that affective and cognitive processes interact. Indeed, some see them as constituting one BrainMind category. But in my reading of functional neuroscience, they are not the same species of mind. I suspect at global conceptual levels, we all have more agreements than disagreements. Of course, the goblins are in the empirical details. So any apparent differences between my views and those of Schulkin and his students may simply



15.  My reflections on commentaries and concluding perspectives 309

reflect their desire for an overall synthetic view, from brain to culture, while my analytical concern is simply how do we decode the neural nature of primal emotional affects, upon which much of the rest of our emotional life is based. Jim Averill. I am sympathetic to his holistic cognitive view of emotions. What Jim describes so well are the nuanced moments of our lived lives. Few of these interesting issues are capable of being addressed in animal models for obvious reasons. The most evident one is that most other animals do not speak; hence we have no access to their subjective “object relations”. Although Averill is comparably critical of Russell’s approach as he is mine, I would simply plead forbearance again. I have no battle with the full multi-tiered, culture-penetrated vision of our everyday emotions that Averill so ably portrays. As emotional people, living in a complex world, this is a position we all can accept. And yes, excellent descriptive science can be done at this level, but little good neuroscience with any explanatory impact. Some neat brain imaging can now be done, but that level of understanding is comparable to views of the cosmos provided by Galileo’s telescope, by no means the resolved views of “deep space and time” provided by the Hubble. Comparably, scientific descriptions of our ­everyday emotions typically pale phenomenologically in comparison to the subtle textures provided by great novelists. If Averill is calling for more good phenomenological research, I am completely with him. I wish Francisco ­Varela’s vision were already part of a “pychoethology” that human investigators were pursuing (Panksepp, 1999, 2000). But that kind of research, by itself, will never take us to the evolutionary foundations of emotions that scientifically interest me most, but it could easily be interfaced with neuroscience psycho-pharmacologically­ and psycho-endocrinologically. Among the tertiary-processes of human mind, Averill’s view provides as resolved a phenomenological strategy as we are likely to have for a long time, but we do also need to develop rigorous neuro-psycho-phenomenological tools in human research that can empirically interface primary-processes to the subtleties of human affective experience. Perhaps Averill’s critique is more pertinent to the work Jim Russell pursues than to mine. In any event, I trust that Averill appreciates not only the limits of a primary-process neuroscientific analysis of cross-mammalian emotions, but also the potential such understanding may have for revealing our deeper nature, which flows into endless rivulets and diversities through both cultural and interpersonal good fortunes and vicissitudes. John Bickle is renowned for his advocacy of “ruthless reductionism” in the BrainMind sciences. Zachar provides a fine synopsis of Bickle’s analysis and I have little to add. I agree with the importance of causal-mechanistic analysis of all psychological processes, and the primary-process ones, I think, are most attainable. Only

310 Jaak Panksepp

a few psychological processes can currently be pursued with robust mechanistic strategies, and as John emphasizes, studies of learning and memory have yielded prominent success stories, by no means close to complete. My goal has been to achieve comparable analysis of a more difficult problem – an understanding of primal emotional affects. This has been motivated by a desire to provide fundamental knowledge to promote a better understanding of the affective infrastructures of many psychiatric disorders that cause such abundant human misery. My premise is that if we can understand the primal emotional-affective mechanisms, we should be able to more readily develop new mind-medicines that will be more effective, more specific, and less problematic/dangerous than the ones currently in use (Burgdorf, et al., 2011; Moskal, et al., 2011; Panksepp & Watt, 2011). Perhaps Bickle is right that our search for causal mechanistic explanations of psychological processes is not enhanced by folk psychological concepts, but I have found them useful. I would not have sought to understand any of the 7 emotional primes unless I felt their force in my own life. Indeed, our understanding of brain opioid-based social attachment (bonding) mechanisms and the genesis of depression have been facilitated by understanding that the lack of social support generates a form of psychological pain (deep loneliness) that may be part of the depressive cascade (Watt & Panksepp, 2009; Panksepp & Watt, 2011). I see no critical need to proceed toward a “ruthless reductionism” (i.e., eliminating mind from the equation) that currently well describes biological psychiatry, where diagnostic categories and brain changes seem to suffice, while the changing affective experiences of humans do not. I don’t think we can discard the psychological parts of our nature if we are to have either a full or adequate understanding of psychiatric disorders. I think neuroscience is currently at a point where it really does not have any better measures of global circuit functions than the coherent “instinctual” behaviors of organisms. And here psychology and neural function come together dramatically. My big point is that the evidence substantially and consistently indicates that raw emotional-affective experiences arise from these very incompletely understood but diverse “reward” and “punishment” network properties as they operate on a subtle neural fabric of self-representation (Panksepp, 2007), perhaps a suitable neural substrate for Russell’s Core Affect space. I think Bickle has backed off from his original eliminative materialism position considerably as he correctly defends the perspective that behavioral measures can be used effectively to decode how affective processes operate in animal brains. Affective states are real brain functions, and indeed, perhaps “emergence” is not quite the best concept to use here. But “emergence”, for me, means little more than a dual-aspect monism perspective, which recognizes that psychological states are not reflections of just neuron-only properties, or sub-neuronal properties, but of



15.  My reflections on commentaries and concluding perspectives 311

massive networks that generate types of visually evident organismic coherence that has intrinsic (anoetic, affective) psychological meaning for animals. This holistic view allows me to accept the traditional viewpoint that, obviously, neuroscience can only study parts of phenomena – at present the whole is too large to really tackle in anything that approaches a neuroscientifically credible and comprehensive way. This makes the traditional psychological analyses that many commentators above advocated a very important part of the overall agenda, as long as one does not see one level of analysis substituting for another. Both Bickle and I realize that higher-order analyses do not diminish the importance of proceeding with deep causal analyses, wherever possible. In my estimation, understanding the neural nature of various primal emotional feelings are currently low-hanging fruit in consciousness studies, as long as we can get past the illusions (perhaps even de­ lusions) of ruthless reductionism and provincialism in neuroscience. Ronald deSousa provides a useful philosophical analysis of the role of overall description and scientific reductionism in the study of emotions. I suspect we are pretty much on the same page. I think the role of a scientific analysis is ultimately to provide some understanding of the actual underlying mechanisms of the brain. This in no way diminishes the importance of overall descriptions such as those advocated by Averill and others, but the power of science only comes to the forefront when we begin to deeply understand the phenomenology of the feelings we so easily discuss verbally. In the arena of emotion studies the most fascinating topic for me is raw affective experience. Before deep-scientific inquiries can be pursued, we need many synergistic conversations about the diverse manifestations of such processes in the real word (as Averill as well as Schulkin and colleagues emphasized). But to profoundly understand what we are talking about when it comes to affective feelings, we need to descend to a causal-manipulative level of analysis – namely as close to the BrainMind mechanisms as we can manage. I don’t understand why so few recognize this as “a” (perhaps “the”) foundational question of emotions studies. From the vantage of empirical strategies, the approaches of Jim Russell and myself do not really overlap. But our interest in affect does. We simply disagree on some conceptual and strategic issues, but that is a perennial human condition when investigators come from totally different intellectual traditions. We don’t even seem to disagree whether we need to descend to the neural-mechanistic level in emotion studies to understand the affective processes that humans experience, but Russell (perhaps deSousa also), because of their unique intellectual backgrounds, must pass those projects on to others. Wisely, deSousa encourages us to see the importance of such projects, so his perspective is harmonious with my life’s work.

312 Jaak Panksepp

Georg Northoff, a close colleague, provides systematic heuristic visions of how the various neuro-conceptual entities that Russell, I and everyone else are forced to use when our empirical stream of knowledge dries up. Indeed, this useful commentary deserves a full-length commentary in itself. However, this cannot really be done without delving into Northoff ’s recent book (Northoff, 2011), which I have done. However, going into those issues may be outside the domain of our current discussion. Suffice it to say that the type of analytic-synthetic thinking that Northoff provides is desperately needed to make progress on the actual neurodynamics that constitute affective-subjective experiences. None of us can yet claim much success at that level of analysis, but we know the types of experiments that are required: Characterization of global neural network activity patterns as affective mind processes that are aroused at specific times, by specific stimuli, in specifiable brain systems. Those levels of analysis have simply not been achieved yet with great success (for one of our attempts, see Bekkedal, et al., 2011). Whether such global network analyses would eventually support a Core Affect vision of some type of global experiential space laid out in valence and arousal coordinates, or more discrete varieties of affective experiences, or both, is simply unknown. On the neuroempirical ledger, the basic emotion view is currently more fleshed out for both humans (Damasio, et al., 2000; Vytal & Hamann, 2010) and animals (Panksepp, 1998, 2005a), than a dimensional Core Affect view. But that may change when Core Affect theorists devise constitutive ways to test their ideas neurologically. However, to do that well, we may need better analyses of the archaic withinbrain and interoceptive body mapping that occurs in mind-generating regions deeper than the well-known “Penfield maps” in primary motor and sensory neocortices. As Northoff notes: “We associate primary process affect with the body as a whole in our subjective experience” and this will surely be more than just sensory information coming from the outside surfaces of the body, perhaps even more than the additional information from inside the skeletomuscular and visceral inputs to the brain.There may be ancestral visceral bodily maps within the brain itself. That could serve as a generalized space for Core Affects or specific emotions, or both. Northoff gives us visions of how to think about many such subtle issues. Although Northoff does not highlight the evolutionary layerings within the brain, it is implicit in his analysis and, I assume, a fact of mental life. The concept of nested hierarchies, well developed by Northoff and his colleagues (Northoff, et al., 2011), provides a systematic and heuristic way to talk about the massive interpenetrations of primary, secondary and tertiary MindBrain processes. In short, Northoff provides useful ways to carry forward a synergistic empirical analysis of Core Affect and basic-emotion system concepts in ways that may one day help empirically resolve some of the issues discussed here.



15.  My reflections on commentaries and concluding perspectives 313

Paul Davies, instead of describing his own views about emotions, provides the most “critical” analysis of the current debate itself. He emphasizes the shortcomings of the top-down constructivist position advanced by Russell, and casts a skeptical eye over our presumed knowledge about the causes of our behaviors, that affective neuroscience has put on the table through its bottom-up approach. He explicitly favors my approach and proceeds to highlight how Jim Russell’s and my positions seem irreconcilable. But perhaps we will find they are not, once ­Russell convinces neuroscientists how to study the properties of Core Affect. For instance, I have long been convinced that all other primal positive emotion circuits can share, to some degree, the positive euphoria of the seeking system. Further, I wish to entertain at least one key point of consilience based on the fact that Russell and I may be truly discussing two very different aspects of emotionality: I am seeking to provide an empirical, neuroscientific description of primary-process cross-species emotions, based empirically on conjoint behavioral and affective (i.e., “reward” and “punishment” analyses), responses to artificial (i.e., brain stimulation) provocations. In contrast, Jim Russell studies verbal emotional concepts strictly in humans. We can all agree that my data is based on the study of more ancient brain networks, namely deep subcortical systems that are justifiably conceptualized as primary-processes of the mammalian brain. Russell’s work is based on neocortical brain networks that allow humans symbolic communication. Clearly, Russell’s data is based on tertiary-process functions of the human BrainMind. This is an enormous difference, and must mean that at an empirical level, we are truly studying very distinct BrainMind processes, albeit they may be strongly linked at an ontological level. Davies is correct in stating that my view of BrainMind organization is more falsifiable than Russell’s. That is largely because Russell’s work has not directly included any explicit analysis or discussion of the possibilities of brain research. But even if it had, we would be studying very different regions of the brain – language requires neocortex. As far as we know primal emotional affects do not. Thus, ­ Russell is empirically only justified in reaching provisional conclusions about neocortical functions, even though he may hypothesize about subcortical ones. And if he did, then he would need to consider negating the types of data and interpretations upon which my position is based. Although I agree with his overall analysis, Davies steps into some factual “quicksand” when he claims that “there is diversity in mammalian cortical structures but comparatively little in sub-cortical ones.” It is more appropriate to readjust the verbiage somewhat – to claim that there is greater plasticity and hence functional diversity in mammalian cortical functions, as compared to greater genetic specification of sub-cortical functions. The anatomical and neurochemical diversity in both higher and lower brain regions across species is substantial:

314 Jaak Panksepp

In cortical regions it is a matter of amounts and distribution of neocortical tissues in various subregions of the hemispheres depending on an animal’s bodily specialization (e.g., the star-nosed mole has a massive cortical representation of the nose compared to humans), even though the cortical columns in all mammals are remarkably similar. In subcortical regions, it is a matter of diversity in absolute and relative numbers and types of neurons, degrees of interconnectivity, and specific neurochemistries of specific subcortical regions. What is clear is that the neuroanatomical/neurochemical control of intrinsic psychological functions (raw affects) are more genetically homologous in subcortical than in neocortical regions. In contrast, within neocortical regions, functional plasticity is much more extensive. To the best of our knowledge, neocortical functions are not as tightly genetically dictated, but remain much more “open” to developmental programming. The prototype example is the fact that visual cortex develops just fine if the normal projection areas for vision in the occipital cortex are surgically eliminated prior to birth (e.g., Sur & Rubinstein, 2005). In contrast, if one damages the seeking system early in life, the future life of young rats is definitively compromised (Almli & Golden, 1974). There are many other examples like these. Many types of subcortical damage are catastrophic for normal mental and physical life. Substantial neocortical damage is not, albeit creatures with total decortication, whether human or other animals, can only thrive in “sheltered workshops” since they are quite “stupid” – i.e., raw affects can make animals seem emotionally “bright” but they know not, nor apparently have much forethought or anticipatory care about, what they do. Davies wisely points out that the primal powers of the affective mind can make shambles of our perception of cognitive-deliberative control of our behaviors. Our primal emotions do have a mind of their own (intentions-in-action), and it is a powerful mind that can coordinate with cognitive decision-making (intentions-to-act). But it can also generate behavioral outcomes that are perplexing to our higher minds. Affect trumps cognition, when the going gets tough.

2. Concluding remarks Once again, I thank the editors, Jim Russell and the commentators for participating in this conversation. The one thing we can be certain about in this new and popular scientific endeavor – the emotional/affective sciences – is that our ignorance is much vaster than our secure base of knowledge. We can also be sure that there are abundant bottom-up and top-down “circular” paths of causality in the BrainMind that have hardly been discussed or scientifically documented. Jim Russell is providing a useful description of our psychological affective space that,



15.  My reflections on commentaries and concluding perspectives 315

as far as we know at the present time, resides only in the neurosymbolic cortical conceptual maps. I work on the primary-process emotional/affective systems of mammalian and avian brains, where a diversity of affects were evolutionarily constructed in deep time, which allows us to seek homologies with even our coldblooded kin, including invertebrates (Huber, et al., 2011). Accordingly, I find it hard to imagine that such valuable survival devices as various positive and negative affects are not the birthright of many, many classes of animals. For anyone to claim that such processes do not exist in animals (e.g., Rose, 2002), is Cartesian foolishness. We may differ much more dramatically in our cognitive abilities than in the state-control systems that were designed to automatically/ affectively encode and thereby implicitly “anticipate” our survival concerns. These systems do not think, but they can channel our thoughts. When they and the many habits they promote control our behavior, we are expressing no “free-will”. However, when we rotate behavioral options in our higher “mind’s eye” I think some of our choices deserve that royal term. In any event, that is “free will” enough for me. A key issue to consider is the degree to which Core Affect (CA) and PrimaryProcess (PP) emotion theories are based on empirical evidence as opposed to “theory theory” type of thinking, which means the “natural” tendency of humans to postulate causes for their experiences. In my estimation, CA is largely a “theory theory” type of synthesis, at least so far. Indeed, even though earlier meta-analysis could not provide strong evidence for either PP or dimensional views of emotions (i.e., the foundational tenet of CA theory), as noted earlier, a more sophisticated recent meta-analysis has offered solid evidence for differential limbic neural representations of five emotional processes in higher limbic and paleocortical regions (Vytal and Hamann, 2010), a finding also observed in deep subcortical regions, as confirmed with PET imaging (e.g., Damasio, et al., 2000). Of course, a critical issue is the degree to which CA and PP can yield additional testable predictions as well as differential ones. My approach has, I believe, scored reasonably well on the former compared to Core Affect theory, partly because it has been around for much longer and it is fundamentally neurological. So let me just focus on three future differential predictions: i. If Core Affect theory is correct, it would predict that humans stimulated in, for instance, fear and rage circuits, as could be done in Periaqueductal Gray (PAG), hypothalamus and amygdala, would report an undifferentiated negative affect, which depending on one’s cultural labeling history, would be described as some type of “fearfulness/anxiety” by some people and “anger/ irritability” by others. My primary-process affect theory would predict that the former descriptors would apply dramatically more for fear stimulation locations and the latter to rage locations.

316 Jaak Panksepp

ii. If my version of PP theory is on the right track, it would predict that animals stimulated at the two above sites would readily discriminate the two types of stimulation, while CA theory would predict great difficulty in animals forming discriminations when the two sites were used as positive and negative conditioned stimuli (CS+’s and CS−’s) in behavioral jargon. iii. With regard to neurochemistries, very similar predictions would be made if one administered neuropeptides that facilitate fear (e.g., Corticotrophin Releasing Factor) or rage (e.g., Substance P) into the ventricular system of humans. There are many others, but the key point I wish to make is that perhaps only neurobiological types of experiments and predictions can provide really distinctive experiments to discriminate among CA and PP perspectives. Further, it is noteworthy that CA and PP seem to envision the interaction of cognitions and affects differently, even as both recognize their inter-penetrance. PP suggests a variety of ways to distinguish such major categories of mental experience, with primal affects being distinctly more intrinsic and ancient in brain organization and evolution than the capacity to think. CA tends to see them more as different facets of higher mental processes, with cognitive activities being rather important, perhaps essential, for differentiating affective states. Both views, I assume, would assert that a better understanding of affects will have important practical consequences. When we begin to understand these complex systems that interact with many other BrainMind functions (e.g., Scherer, 2009), we may hopefully have some lasting impact on psychiatric thought. PP has been explicit about such issues (Panksepp­, 2004, 2006, 2010). For instance, if we work out the neurochemistries of the seeking and panic/grief systems, we should be able to translate such knowledge to useful new concepts for treating psychiatric disorders (e.g., Coenen, et al., 2011; Panksepp & Watt, 2011). I am especially pleased that some of the neurochemistries of affect that we have revealed by studying gene expression patterns in the brains of happy animals have already yielded new potential medications for the treatment of depression (Burgdorf, et al., 2011), and perhaps autism (Moskal, et al., 2011) with a single agent that has just proceeded into FDA Phase 2A clinical testing (namely GLYX-13, a glycine receptor partial agonist). This said, I have no disagreements with Jim Russell’s empirical work and admire his devotion to understanding emotions. He is as interested in the affective life as I am. However, because of his own educational and life trajectories, he came to CA theory from a very different intellectual tradition than the one that led me to a PP approach. His approach is strongly premised on an appreciation of our highly plastic, culturally-based human nurture, at which the human



15.  My reflections on commentaries and concluding perspectives 317

neocortex is superbly adept. Thus, Jim’s empirical work proceeds completely at the tertiary-process level of psychological analysis, and mine largely at the primary-process level. When he published his seminal paper on Core Affect in Psychological Review in 2003, I was thrilled by the concept, eager to learn more. I promptly sent him a congratulatory note, pointing out how close our views might be. I suggested that we should write a paper of interdisciplinary consilience together – a statement on how our two views could synergize. After more reading and especially the analysis provided by Paul Davies, I am no longer confident that could be readily done – as Paul put it “Panksepp and Russell are indeed theorizing about quite different phenomena” and that now rings true. Anyway, in 2003, Jim was intrigued by the proposal. But as he told me then, he had “other fish to fry.” I suspected who those other fish might be – perhaps a few more famous “basic emotion theorists” among whose ranks I have been repeatedly enjoined by those who argue against the existence of basic emotions in humans(e.g., starting most prominently with Ortony & Turner, 1990). But my laboratory work is quite different – deliberately, and hopefully, focused on the primary-process emotional affects of all mammalian/avian brains. The perception that there are fundamental empirical disagreements between Russell and myself is a fiction. I trust we will find ways to integrate both CA and PP approaches, perhaps as Davies noted, “the two views can be integrated or at least reconciled.” At present, I think this could be achieved if CA theory were simply cast as a tertiary-process theory-theory about the cortical aspects of higher-order human emotions, and not at all a theory of affective life across species. Still, I do see the possibility that all affects share a common fundamental substrate of neurosymbolic organismic representation – a core self (Panksepp, 1998, 2005b, 2007) – upon which and through which a variety of feelings are woven. For now, I think if we keep our own preferred levels of BrainMind analysis clearly in view, basing our arguments on the best cross-disciplinary and crossspecies data available, we could conjointly promote a lasting synthesis in this troubled field. That might serve scientific psychology, and new evidence-based humanistic, philosophical and psychiatric views, very well indeed. And thereby, hopefully, promote better tools and therapeutic approaches in biological psychiatry and psychotherapy, and ultimately, synergistic ways to understand our nature and nurture conjointly, both psychologically and neuroscientifically (e.g. Panksepp­ & Biven, 2012; Solms & Panksepp, 2012; Zellner, et al., 2011).

318 Jaak Panksepp

References Almli, C. R., & Golden, G. T. (1974). Infant rats: effects of lateral hypothalamic destruction. Physiology & Behavior, 13, 81–90. Bekkedal, M. Y., Rossi, J., 3rd, & Panksepp, J. (2011). Human brain EEG indices of emotions: Delineating responses to affective vocalizations by measuring frontal theta event-related synchronization. Neuroscience and Biobehavioral Reviews, 35, 1959–1970. Burgdorf, J., Panksepp, J., & Moskal, J. R. (2011). Frequency-modulated 50kHz ultrasonic vocalizations a tool for uncovering the molecular substrates of positive affect. Neuroscience & Biobehavioral Reviews, 35, 1831–1836. Coenen, V. A., Schlaepfer, T. E., Maedler, B., & Panksepp, J. (2011). Cross-species affective functions of the medial forebrain bundle-Implications for the treatment of affective pain and depression in humans. Neuroscience & Biobehavioral Reviews, 35, 1971–1981. Damasio, A. R. (1999). The feeling of what happens: Body and emotion in the making of con­ sciousness. New York: Harcourt Brace. Damasio, A. R., Grabowski, T. J., Bechara, A., Damasio, H., Ponto, L. L., Parvizi, J., & Hichwa, R. D. (2000). Subcortical and cortical brain activity during the feeling of self-generated emotions. Nature Neuroscience, 3, 1049–1056. Denton, D. (2006). The primordial emotions: the dawning of consciousness. New York: Oxford University Press. Eisenberg, N. I. (2011). The neural basis of social pain: findings and implications. In MacDonald, G. & Jensen-Campbell, L. A. (Eds.) Social pain:  Neuropsychological and health implications of loss and exclusion (pp. 53–78). American Psychological Association, Washington, DC. Hess, W. R. (1957). The Functional Organization of the Diencephalon. New York: Grune & Stratton. Huber, R., Panksepp, J. B., Nathaniel, T., Alcaro, A., & Panksepp, J. (2011). Drug-sensitive reward in crayfish: An invertebrate model system for the study of SEEKING, reward, addiction, and withdrawal. Neuroscience & Biobehavioral Reviews, 35, 1847–1853. James, W. 1884. What is emotion? Mind, 9, 188–205. Moskal, J. R., Burgdorf, J., Kroes, R. A., Brudzynski, S. M., & Panksepp, J. (2011). A novel NMDA receptor glycine-site partial agonist, GLYX-13, has therapeutic potential for the treatment of autism. Neuroscience and Biobehavioral Reviews, 35, 1982–1988. Northoff, G., Wiebking, C., Feinberg, T., & Panksepp, J. (2011). The ‘resting state hypothesis’ of major depressive disorder: A translational subcortical-cortical framework for a system disorder. Neuroscience and Biobehavioral Reviews, 35, 1929–1945. Northoff, G. (2011). Neuropsychoanalysis in practice: Brain, self and objects. New York: Oxford University Press. Ortony, A., & Turner, T. J. (1990). What’s basic about basic emotions? Psychological Review, 97, 315–331. Panksepp, J. (1971). Aggression elicited by electrical stimulation of the hypothalamus in albino rats. Physiology & Behavior, 6, 321–329. Panksepp, J. (1982). Toward a general psycho-biological theory of emotions. Behavioral and Brain Sciences, 5, 407–422. Panksepp, J. (1998). Affective neuroscience: The foundations of human and animal emotions. New York: Oxford University Press.



15.  My reflections on commentaries and concluding perspectives 319

Panksepp, J. (1999). Emotions as viewed by psychoanalysis and neuroscience: an exercise in consilience, and accompanying commentaries. Neuro-Psychoanalysis, 1, 15–89. Panksepp, J. (2000). On preventing another century of misunderstanding: toward a psychoethology of human experience and a psychoneurology of affect. Neuro-Psychoanalysis, 2, 240–255. Panksepp, J. (Ed.). (2004). A Textbook of Biological Psychiatry. Hoboken, NJ: Wiley. Panksepp, J. (2005a). Affective consciousness: Core emotional feelings in animals and humans. Consciousness & Cognition, 14, 19–69. Panksepp, J. (2005b). On the embodied neural nature of core emotional affects. Journal of Con­ sciousness Studies, 12 (8–10), 158–184. Panksepp, J. (2006). Emotional endophenotypes in evolutionary psychiatry. Progress in NeuroPsychopharmacology & Biological Psychiatry, 30, 774–784. Panksepp, J. (2007). Affective consciousness. In M. Velmans & S. Schneider (Eds.), The Black­ well companion to consciousness (pp. 114–129). Malden MA: Blackwell Publishing, Ltd. Panksepp, J. (2009). Primary Process Affects and Brain Oxytocin. Biological Psychiatry, 65(9), 725–727. doi: DOI 10.1016/j.biopsych.2009.02.004 Panksepp, J. (2010). Affective neuroscience of the emotional BrainMind: Evolutionary perspectives and implications for understanding depression. Dialogues in Clinical Neuroscience, 12, 533–545. Panksepp, J. (2011). The neurobiology of social loss in animals: Some keys to the puzzle of psychic pain in humans. In MacDonald, G. & Jensen-Campbell, L. A. (Eds.) Social pain:  Neuropsychological and health implications of loss and exclusion (pp. 11–51). American Psychological Association, Washington, DC. Panksepp, J., & Biven, L. (2012). The archaeology of mind: Neuroevolutionary origins of human emotion: Norton, W. W. & Company, Inc. Panksepp, J., Fuchs, T., & Iacobucci, P. (2011). The basic neuroscience of emotional experiences: The case of FEAR and implications for clinical anxiety in animals and humans. Applied Animal Ethology, 129, 1–17. Panksepp, J., Normansell, L., Cox, J. F., & Siviy, S. M. (1994). Effects of neonatal decortication on the social play of juvenile rats. Physiology & Behavior, 56 (3), 429–443. Panksepp, J., Vilberg, T., Bean, N. J., Coy, D. H., & Kastin, A. J. (1978). Reduction of distress vocalization in chicks by opiate-like peptides. Brain Research Bulletin, 3, 663–667. Panksepp, J., & Watt, J. (2011). Why does depression hurt? Ancestral primary-process separation-distress (PANIC) and diminished brain reward (SEEKING) processes in the genesis of depressive affect. Psychiatry, 74, 5–14. Rose, J. (2002). The neurobehavioral nature of fishes and the question of awareness and pain. Reviews in Fisheries Science, 10, 1–38. Russell, J. A. (2003). Core Affect and the psychological construction of emotions. Psychological Review, 110, 145–173. Scherer, K. R. (2009). The dynamic architecture of emotion: Evidence for the component process model. Cognition and Emotion, 23, 1307–1351. Schultz, W., & Dickinson, A. (2000). Neuronal coding of prediction errors. Annual Review of Neuroscience, 23, 473–500. Siegel, A. (2005). The neurobiology of aggression and rage. Boca Raton, FL: CRC Press. Solms, M., & Panksepp, J. (2012). What is neuropsychoanalysis? Trends in Cognitive Sciences, 16, 6–8.

320 Jaak Panksepp

Sur, M., & Rubenstein, J. L. R. (2005). Patterning and plasticity of the cerebral cortex. Science, 310, 805–810. Vytal, K., & Hamann, S. (2010). Neuroimaging support for discrete neural correlates of basic emotions: a voxel-based meta-analysis. Journal of Cognitive Neuroscience, 22, 2864–2885. Watt, D. F., & Panksepp, J. (2009). Depression: an evolutionarily conserved mechanism to terminate separation-distress? A review of aminergic, peptidergic, and neural network perspectives. Neuropsychoanalysis, 11, 5–104. Zellner, M. R., Watt, D. F., Solms, M., & Panksepp, J. (2011). Affective neuroscientific and neuropsychoanalytic approaches to two intractable psychiatric problems: Why depression feels so bad and what addicts really want. Neuroscience & Biobehavioral Reviews, 35, 2000–2008.

chapter 16

Concluding observations Comparisons, contrasts, and some important convergences Ralph D. Ellis and Bill Faw Clark Atlanta University / Brewton-Parker College

In this concluding chapter, we attempt to do two things: Flesh out some of the important neurophysiological underpinnings of the emotional phenomena discussed by both Russell and Panksepp; and consider some crucial theoretical implications of these discussions, focusing especially on points of convergence in the two perspectives. Section 1 focuses primarily on the first endeavor, and Section 2 primarily on the second. Both authors of this chapter take responsibility for both tasks, although Faw is more responsible for the first, and Ellis for the second.

1. Working from both ends toward each other This book asks Panksepp to look from his own work at the comparative-biology level toward Russell’s work at the verbal-psychological level – and for Russell to look from his work at the verbal-psychological level at Panksepp’s work at the comparative level – and for each to assess the compatibilities and incompatibilities of the other’s work. This is a difficult but important thing for each of these researchers to do. As an analogy of what we are requesting of them both, consider the First Intercontinental Railway created in the United States from 1863 to 1869. The Central Pacific Railroad crew began laying track near San Francisco, California, heading East. The Union Pacific Railroad began laying track at Council Bluffs, Iowa, heading West. They met and linked tracks at Promontary Summit, Utah. This combined Pacific Railroad was later joined to the Eastern Railroad tracks to span the United States from the Atlantic to the Pacific. The Central Pacific and the Union Pacific were able to pull off this amazing task (especially for 1869) to meet in Utah and to link their tracks, only because they, at every moment of their efforts, were specifically aiming for each other. Imagine the infinitesimal odds of

322 Ralph D. Ellis and Bill Faw

the two track-laying efforts ever meeting, or even passing through the same states, if each had laid its tracks according to its most convenient trails – even if one generally moved east and the other generally moved west! If one assumes at least a modest doctrine of “scientific realism,” scientific research is different from track-laying across the Rocky Mountains, in that there are supposed regularities and even “laws” to be discovered in nature – while there are no natural paths in geography, other than rivers and mountain passes. And yet the odds of “meeting each other” in the middle – from opposite ends of levels of scientific analysis – are still slim, unless there is some “aiming for each other” – of which scientific emotion research has known very little. Panksepp and the few others currently working at the primary level of emotion science (dealing with brain areas involved in unconditioned emotional responses) have been working on that for its own sake. Because Panksepp’s proposed emotion circuits link primarily to secondary brain mechanisms (those involved in conditioned emotional responses, such as in Joseph LeDoux’s work) and tertiary (cognitive processing) brain areas, Panksepp has at least looked at the other levels. The vastly larger number of researchers, like Russell, currently working at the tertiary level of emotion science – much at the “high tertiary” linguistic and cultural levels – are following the paths which open up to them, without much constraint for “aiming for” specific primary-level biological findings. Russell, however, has started to change that. This book allows two of the leading researchers to look hard at the other camp and to speculate as to how each might work his way to the other – in some neuroscience version of Promontary Summit, Utah. This does not mean that all of science, or even all of affective science, needs to probe only in the relatively “straight and narrow” pathways which allow it to link to other existing findings within its field. In fact, most of the research of science is done – and perhaps should be done – in a piecemeal fashion: finding and exploring empirical nuggets wherever they might be found; and then letting the rare grand theorists or science textbooks discover what broader pictures are suggested by the isolated jig-saw puzzle pieces. But it is helpful along the way to work on “grand theory.” Indeed, while many scientific data come from chance empirical discovery, other data come from theory-driven research to test hypotheses and to join other data points.

“Affect,” “emotion,” and “discrete emotions” Russell points to three types of theories in “emotion” research. (1) The traditional view treats “emotion” as a super category or a “natural kind,” as well as treating specialized emotion systems such as fear and anger as “natural kinds.” (2) LeDoux (1996) represents researchers who reject “emotion” as a super category or natural



16.  Concluding observations 323

kind, but deal with specific emotion systems – especially fear – presumably as natural kinds. (3) Russell represents researchers who not only reject “emotion” as a super category, but also reject the consideration of specific emotion systems. Once you explain each of the components in a given emotional episode, there is no super category nor are there specific emotions to explain. It is not clear whether Panksepp should be put in the first or second option, above. He holds to there being discrete emotional responses (and so is clearly not in the third group with Russell), but may not be making a case for “Emotion” as a super category. His use of “affective” is much wider than “emotion” and his “emotional circuits” are basically “motivational circuits.” “Affect” seems to be equivalent to “feelings” or “experience,” with at least three categories of “affects”: sensory affects, homeostatic affects, and emotional affects. These three types of affect may have evolved from a unitary “go and stop” message in invertebrates. It may be because of his own broader use of “affect” that in their exchanges in this book Panksepp initially misinterpreted Russell’s basic concept of Core Affect, as suggesting that all affects can be defined by the two dimensions of valence (­pleasure/ displeasure) and arousal. Panksepp rightly asserted the absurdity of defining both intense hunger and intense thirst as only “intense displeasure” – so that one, presumably, could not differentiate (in Russell’s presumed model) whether one was very hungry or very thirsty – or very fearful. Surely, Panksepp suggested, there is something else added to the experience of hunger and thirst. Russell quickly retorted that of course one can differentiate hunger and thirst. Russell’s position is that emotions include other elements besides what he calls “Core Affect” – and those other elements involve the way emotions, like other experiences, refer to environmental or imaginary scenarios in the world, which experiences are felt to be “about.” Knowing what a feeling is about – its intentional meaning – requires more than Core Affect (as we will discuss below). Moreover, Russell is limiting his claims about Core Affect to “emotional” feelings. But ­ Russell needs to explain how he can limit his Core Affect to “emotional” feelings (and not hunger or thirst), when (a) he sees Core Affect as ever-present­; when (b) he has such questionably-emotional descriptions as calm, tired, lethargic, tense, serene, comfortable, fatigued, bored, drowsy and jittery in the two-dimension­ Core Affect space defined by valence and arousal; and when (c) Russell rejects the category: “emotion.” Russell wants to eliminate the term “emotion” as a needless group term, but throws it in to differentiate the mixedbag of Core Affect from non-emotional feelings. Russell finds correlations to be weak among the components of emotional episodes, as measured at the cortical-psychological level – while Panksepp finds much tighter correlations among brain stimulations and emotional responses, as measured at the sub-cortical biological level. This seems reasonable, given the

324 Ralph D. Ellis and Bill Faw

greatly increased “degrees of freedom” of responding at the psychological level. This raises the problem of reaching conclusions regarding “basic emotions” from much higher levels of analysis.

Are emotions rooted in motivational systems? Panksepp labels his “seven fundamental emotional urges” seeking, rage, fear, lust, care, grief/panic, and play. Panksepp makes a profound but controversial move from motivational systems to emotions. It is profound in that Panksepp roots all “basic” emotional feelings and responses in motivational systems of comparative biology; so that there is no danger of seeing emotional science in terms influenced by social folk concepts. It is also profound in showing the “social roots” of even “basic” emotions. But, Panksepp’s move from “motivational” to “emotional” systems is quite controversial, because seeking, parental care, and play are not easily seen as emotional systems; and because it is not self-evident what distinguishes this set of motivational systems from hunger, thirst, self-defense, hierarchy setting, or even the need for air, maintaining body temperature, and electrolyte balance. From a mainline affective science perspective, there seems to be an arbitrary nature to his “Big Seven.” Panksepp’s big seven. Instead of detailing each motivational/emotional system, we will raise some implications and questions deriving from them. Some of the comments and questions on these derive from the much fuller description of the anatomy and chemistry of these systems in his 1998 book. seeking. In terms of Russell’s two dimensions of valence and arousal, it is tempting to see Panksepp as having seven “valence” systems and one “arousal” system. But seeking is a major source of arousal as well as of valence. It may be the source for libido; in addition, electrical stimulation of the dopamine seeking system activates the most intense emotional experiences. It is the general-purpose appetitive motivational system; essential for the other basic emotion systems to be effective. seeking is associated with an “emotional tone” of interest, enthusiasm, exhilaration, anticipation, and euphoria. In distinction, endogenous opioid systems seem to underlie consummatory pleasures – as found in various other circuits. Panksepp states that this seeking circuit leads to “depression” when underactive and “mania” when overactive. But this differs from the way psychiatrists typically use the term “depression.” When this seeking system is underactive, one has apathy, which is not the same as depression. Apathy means severe lack of motivation to do, think, or feel anything, which is present in only some forms of “depression.” In fact, a profound temporary or permanent lowered ­functioning



16.  Concluding observations 325

of dopamine systems – or the blocking of their feed to posterior and anterior cingulate and related neo-cortical areas – seems to be the cause of Akinetic Mutism (Damasio, 1999), where a person has, at most, a passive form of conscious awareness but no engagement with the world – through active thought, emotion, or action. Many would also characterize the overactive state differently from the way Panksepp does. Overactivity of the VTA-to-nucleus accumbens dopamine circuit has been strongly linked to the “positive signs” of schizophrenia, such as speeded up thinking and behavior, auditory hallucinations, delusions, and paranoia. Mania­ is more identified with excessive norepinephrine than with excess dopamine. Identifying changes in activation with apathy and acute schizophrenia – rather than with depression and mania – demonstrates the “chief motiva­ tional” character of this circuit, rather than a “chief emotional” character. rage/anger and fear. Rage and fear are not responses to external events per se, but responses to those external events curtailing freedom of action – curtailing the seeking system. The frontal lobe matches top-down expectations with the bottom-up results of seeking, and then activates down-traveling rage and/or fear nuclei in the amygdala, which then activate neighboring circuits through the hypothalamus to the periaquaductal gray (PAG) in the brainstem, which then activates autonomic and motor response areas of the brainstem. Rage or fear responses can be triggered by stimulation along their respective paths. Russell’s critical point, that there are many behavior patterns associated with fear, does not negate the basic escape behaviors emitted by stimulation of fear areas. There seem to be basic responses at Panksepp’s level, but considerable flexibility of response at Russell’s level. play. Panksepp is probably known best in the popular press because of his experiments in which he tickles rats and makes them “laugh” – a high pitch 50 kHz chirping which may be related to the innate laughing response in human children. Rough and Tumble (RAT) play is the source of “joy” and is linked to somatosensory information processing in the midbrain, thalamus, and cortex; which are very different from the seeking circuits and circuits in other motivational systems. While a certain amount of activation of dopamine seeking circuits seems crucial for play, actual seeking behavior inhibits play. The neurotransmitter systems involved in play also differ from those of seeking – with some overlap in opioid systems. For mammals and birds, playing is the key developmental social system, helping develop skills in dominance-testing, courtship, parenting, and constructive aggression, including experience in losing. love/lust. Oxytocin, vasopressin, and opioids are involved in lust. While male and female sexuality have distinct brain control systems, most of these circuits are totally different from those involved in seeking and play. For delightful reasons, the areas involved in erections eventually converge into the pathway of

326 Ralph D. Ellis and Bill Faw

the seeking system; but opioids seem to be more involved in sexual pleasure than is dopamine. care. Panksepp’s characterization of the care system seems to include the experiences and behaviors of dependent minors as well as of their caregivers. Both opioids and oxytocin are activated by various pleasurable pro-social activities, which may promote the subtle feelings of acceptance, nurturance, and love. panic/grief. While Panksepp does not make a broad generalization, for the most part his findings tend to suggest that rage and fear are basically the contrary systems for seeking; while panic/grief is the contrary system for the social motivations/emotions of play, lust, and care. Corticotrophin Releasing Factor (CRF) helps drive the panic/grief system, while Oxytocin and Prolactin inhibit it. The panic system appears to arise close to physical pain-induction­ areas. Separation calls are among the “verbal reports” of this system.

Is there experience at the primary level? Russell has his own three-level schema in relation to the components of emotional episodes: (1) not available to consciousness; (2) available to consciousness but currently not attended; and (3) attended and conscious. For Russell, many com­ ponents of emotional episodes can operate at the unconscious level. He believes that the basic affect experience which he calls Core Affect goes back and forth between the second and third levels. In contrast, Panksepp agrees with Freud that affects are “always experienced.” More precisely, for Freud, affect is always “con­ scious.” It is not clear whether Panksepp equates his “experienced” with Freud’s “conscious” – or whether he agrees with Russell’s concept that Core Affect can be non-attended. We would also raise the question as to whether “non-attended” automatically means “un-conscious.” Some would argue that there is a middle category of “unattended conscious.” Many researchers in affective- and consciousness-science assume that the lower brain areas are incapable (in themselves) of producing conscious experience, with the possibilities of conscious experience rising only through mechanisms in the cortex – and, some would say, even the pre-frontal cortex. Panksepp, by contrast, is more in agreement with Damasio (1994), who shows evidence that there is “core consciousness” as long as lower brain areas are intact, with the cortex only adding increasing sophistication to the consciousness. Along with defining basic emotional systems in terms of basic motivational systems, Panksepp makes big leaps from evidence to conclusions in the following two related assertions: (1) affective feelings “arise from” (or “are constituted substantially from”) the same subcortical neurodynamics as do emotional responses; and (2) there is



16.  Concluding observations 327

conscious experience of emotions even at very low brain levels. For this he mentions decorticate children, who develop without any functioning cortex. They have all of the emotional-instinctual urges, perhaps even more intensely. However they cannot learn new behaviors (which are dependent upon secondary and tertiary structures). While Panksepp cannot prove that they also have basic emotional experiences, he seems to believe that they do. Panksepp believes we can use instinctual emotional responses as proxies for monitoring distinct affective states: with decorticate children, normal infants, and other mammals. If we accept verbal reports, then we should accept natural emotional vocalizations (with rats: 22 kHz vocalizations report negative affects; while 50 kHz chirps report positive affects) and behaviors in non-human animals. Panksepp may be right in these leaps, but Russell makes a strong argument for questioning them. Russell is much more willing to allow that there may be hard-wired patterns of emotional neural functioning and behavior, than to allow for the conclusion that seven somewhat-distinct emotional areas lead to somewhat-distinct conscious emotional experiences. While Panksepp is the biological psychologist of the pair, Russell is closer to the mainstream in at least this one point – in suggesting that his two levels of emotional experience (Core Affect and Emotional Meta-Experience) are mediated by tertiary prefrontal and anterior cingulate structures. Even if we side with Panksepp on this point – we should still recognize how controversial it is.

Russell’s components of emotional episodes Russell is not attempting to construct a “model of emotion,” but a model of the ways that emotional episodes might be analyzed; which is by looking at the emotional experience and the various physiological and behavioral responses involved – as distinct components. Russell stresses that fear episodes do not stem from a single mechanism dedicated to fear. Panksepp agrees with the need of the other components at the secondary and tertiary levels, but would still have a single mechanism for fear at the primary process level. Some of the detail of the following narrative draws from Russell’s comprehensive 2003 article. Russell distinguishes his theory from component theories which see a specific pattern or sequence for those components. Affective science probably does need to shake loose from the traditional assumptions that there are clearly discrete emo­ tions, specifically-emotional components, and generally set causal sequences of these components in emotional episodes. Russell and his colleagues are doing that job well. However, there is the danger that Russell’s strong need to shake loose from positing causal sequences among components will discourage the search for what

328 Ralph D. Ellis and Bill Faw

causal links there really are among components. For instance, Russell writes about the sequence of components in one emotional episode being like the constellation Orion (or a full house in poker), and the sequence in another episode being like the constellation Leo (or four aces in poker). This may leave the reader wondering: “Who in his right mind would bother to look for causal links among the stars in the constellation or the cards in the deck?” While Russell pulls back from both the “constellation” and “playing cards” analogies – as suggesting totally arbitrary and random links – the fact that he leads with such analogies (before pulling back) seems to risk overstating the element of independence among components. In fact, it is quite important to look for the causal links among components, such as the following. What are the triggers in any given species’ environment or homeostatic imbalances that tend to evoke which kinds of responses? What conditions heighten the need to attribute the arousal internally or externally – or to certain objects? How do perceptions and expectations link to become apprais­ als? How do appraisals activate amygdala-hypothalamic-PAG pathways? How do evaluations of ability to cope, to defend, or to escape change the firing of these pathways? By what mechanism does thinking of a peaceful scene change one’s Core Affect? By what mechanism can “putting on a happy face” lead one to feel happier? And by what mechanism might emotional experience (William James, 1884) precede emotional responses – and by what mechanism might it follow them (Walter Cannon, 1929). We will look briefly at several of Russell’s “components” of emotional episodes and raise some questions along the way. Core Affect (CA). Core Affect (CA) is the most elementary simple primitive affective feeling – represented by one point in the space defined by two orthogonal dimensions of Valence/hedonic (pleasure/ displeasure) and Arousal (activation/ deactivation). CA is a psychological primitive but a biological complexity; we are always in a state of CA, and we only have one CA at a time. The conscious side of CA is in primary, simple, non-reflective consciousness. Russell speculates that because Core Affect monitors one’s state of pleasure and arousal in order to set goals, it would be useless in totally reflexive creatures. Thus CA likely co-evolved with behavioral flexibility: perhaps in all land dwelling vertebrates. Pushing Russell a bit, one wonders if an unconscious version of CA might be present in species and human infants who do not have conscious CA. Affective Quality (AQ). Core Affect is a “feeling inside oneself,” while Affective Quality is a property attributed to the thing perceived; isolated from any judgment of reality of the object. It is typically quick and automatic, but can be deliberate. In many cases, Affective Qualities arose because objects have evoked certain Core Affects in us; but we store the AQ of objects into memory. “Mixed feelings” in a situation are due to there being several objects with different AQs – not evidence that there is more than one CA. Russell hypothesizes that Affective



16.  Concluding observations 329

Qualities can also be defined as points in space formed by the Valence and Arousal dimensions. It is not clear why our attribution of qualities to objects and situations would be identical in dimensions to our emotional experience. Is this just a convenient formulation? Antecedent Event and Attributed Affect. Attributed Affect occurs when one realizes that one’s Core Affect has changed and rightly or wrongly attributes that change to its Antecedent Event: which may be external events, bodily feelings, memories, or trains of thought. Part of that attribution stems from the Affective Qualities of these events and objects. If our CA is overwhelmed by a single event, we have one of those “rare prototypical episodes.” We may see a lion as having the affective qualities of being both “beautiful” and “potentially threatening.” If the lion actually threatens us and our Core Affect changes accordingly, we will likely “attribute” our change in CA to the lion. Other times our Core Affect is so little related to specific events or objects that we find it hard to attribute our CA to a specific object. Appraisals. In addition to possibly attributing the change in CA to some perceived “cause,” one can judge the “reality” of the event, and appraise the degree of threat or reward and one’s ability to cope with such changes. Russell sees these appraisals, attributions, and judgments as normal cognitive processing – not at all dedicated to “emotional” processing. Physiological and Expressive Changes. Russell asserts that there are no nonverbal expressive signals or patterns of autonomic nervous system activity unique to “emotion” in general or to any proposed “discrete emotion.” Panksepp differs sharply here regarding primary process emotional circuits. Stimulating fear, rage, or lust basic circuits lead to specific expressive signals, Autonomic Nervous System (ANS) activity, and vocalizations. Panksepp would maintain that Russell’s assertion is true only at higher levels of responses – because of response flexibility given by secondary and tertiary processing. Instrumental Actions. Russell asserts that there is no separate class of “emotional behavior.” Any behavior can be recruited. However, we might still question whether there are special patterns of behavior (in the fight, flight, fright categories) in emotional responses. In complex human interactions, “running away” is distinguishable from “backing down in the face of a political interlocutor’s arguments,” but both may be instances of a broader class of “backing away from danger” responses. Emotional Meta-experience (EME). Russell’s concept of EME can be a helpful notion in many contexts. As the term suggests, Emotional Meta-Experiences are reflections upon emotional experiences – specifically reflections upon one’s immediate or past Core Affects in light of the Affective Qualities of objects and events present, cognitive appraisals made, and of our visceral and behavioral responses

330 Ralph D. Ellis and Bill Faw

to those events. Core Affect gives an episode its subjective feelings of valence and energy. An EME builds upon that. While Core Affects may be fully conscious or unattended, EMEs are reflectively-conscious experiences, but can be ambiguous or “wrong.” It is because of EMEs that we label our experience as “emotional” and, often, as constituting a specific emotion, such as “fear.” In this way, EME sees the resemblance between one’s current state and a mental script. EME is not an introspection and naming of internal events – as in a Freudian perspective – but a self-perception and categorization of one’s state. Folk Psychology supplies our EME with a number of prototypical emotions; but there is really a continuum of degree-of-prototypicality within emotions. Russell makes a seemingly strange assertion that prototypical cases are not necessarily more natural than mediocre cases. In fact, that they are rarer than mediocre cases. This claim may seem plausible if “prototypical” is taken to imply a cartoonlike stereotype or an unusually simple example that might be used to stand for a usually-more-complex category – for example where a yellow smiling-face icon is used as the “prototype” for a smile. Most smiles do not look exactly like the yellow smiling-face icon. But in ordinary language, it seems more appropriate to say that emotional responses are seldom “stereotypical” than “prototypical.” Their proto­ typicality suggests bell curves, with prototypical at the apex (with the highest frequency) and then with lessening frequencies of outliers in both tails. It seems more accurate to say that “absolutely complete emotional sequences” are rare; but the word “prototype” does not have that same “complete” sense in ordinary language. Emotion- and Affect-regulation. Through Emotional Meta-Experience, one categorizes oneself as afraid, etc, placing ones state and situation within social norms and rules, and is thus “regulating the emotion.” This differs from “affect regulation,” which constitutes attempts to change the Core Affect – such as seeking out an Object of known affective quality. This is a useful distinction. Panksepp’s Own List of Components. In this volume, Russell asks Panksepp what produces the other objective parts of an emotional episode in addition to Panksepp’s circuits. Panksepp lists the secondary learning level and the tertiary verbal level with autonoetic consciousness, mental time-travel, mind-mirroring­, and so on. Panksepp also lists his own components in emotional responses: (1) unconditional sensory inputs; (2) coordinated physiological and behavioral outputs; (3) gating of inputs; (4) positive feedback; (5) cognitions instigating emotions; (6) emotions control cognitions; and (7) affect arises from activity of the whole system. Panksepp is willing to acknowledge Russell’s EME, involving perceptual and cognitive processes. Several times Panksepp says that his biggest problem with Russell’s current model is the two-dimensional nature of Russell’s Core Affect, to which we now turn.



16.  Concluding observations 331

Consideration of dimensions of Russell’s Core Affect Psychometric Research. Russell builds much of his model on the twin hypotheses that both one’s moment-by-moment Core Affect (CA) and the Affective Qualities (AQ) one attributes to situations and objects can be described by single points in an area defined by the two dimensions of Valence and Arousal, with Valence as the X-axis and Arousal as the Y-axis. Russell and the many researchers with “bipolar views” contrast their approach to “multiplicity views,” where there are many qualitatively different types of valences and/or of arousal. Panksepp’s Big-7 basic emotion circuits represent a multiplicity view with, presumably, at least seven types of valence, and perhaps several types of arousal. Logic of Psychometric Research. Russell’s concept of two-dimensional Core Affect stems from a half century of psychometric research, in which participants are asked to imagine how they would feel in a certain situation; to react to sounds or photos; to describe their current mood; or to remember a past emotion. Then factor analyses (FA) are run on the ways in which certain words or numerical scores appear together in the narratives. Then certain “dimensions” or “polarities” or “continua” are identified mathematically which meet a specified level of statistical significance or account for a specified proportion of the variance of this correlation. The strength of the final “factors” relates to the degree to which the factors are orthogonal-to (independent-from) each other. Then it is up to the researchers to name those dimensions and thus “explain” these patterns. Russell reports that these factor analyses on reports on emotional episodes have remarkably consistently found two broad dimensions, to which different researchers give different names related to Valence and Arousal. Then researchers have to name the two poles of each dimension; generally something like “pleasure” versus “displeasure”; and “activation” versus “deactivation.” Subjectivity in Naming Dimensions and Poles. There is considerable room for subjectivity in this naming. For instance, some major personality tests define their “introversion” pole in terms of traits of being introspective, self-reflective, and liking to think and meditate; while others define their “introspection” pole more in terms of being shy and having social anxiety. Thus, if the first sense of “introversion” is measured in a fair test, it might be better to have “extroversion” and “introversion” as two separate dimensions – so that one could be high on both – instead of two ends of a single dimension – so that answers that count for “extroversion” count against “introversion.” The notorious Spence-Eysenck controversy raged for years, with the two camps finding apparently conflicting orthogonal (or nearly orthogonal) dimensions because of the uncertainty and overlap of what was being measured by their respective “anxiety” and “introversion” scales.

332 Ralph D. Ellis and Bill Faw

Core Affect Dimensions Derived from Emotional Meta-Experience Reports. In addition to the serious biological problems in assuming that there is a single valence dimension and a single arousal dimension to Core Affect, there is a serious (but unrecognized) methodological problem in Russell’s specific approach to defining Core Affect dimensions. Participants reporting on their current mood or current reaction to sounds or photos – or asked to describe a past emotion from memory, or to anticipate how they might feel in a certain situation – are all reporting out of highly-reflective Emotional Meta-Experience, rather than out of their non-reflective Core Affect. And yet the reports are used as the main way of defining the dimensional structure of Core Affects. Russell anticipates related criticisms that nibble away at the edges of this problem: that his psychometric method “relies too much on words” and “relies on selfreport.” Russell’s responses are that “Core Affect is the state reported, rather than words”; and that until we have better objective indicators, the mental side now gives us the most straight-forward access to Core Affect. However, neither of these responses acknowledges that EME, not Core Affect, is the “state reported” and that EME does not at all give us straight-forward access to Core Affect. Add to this the criticism (which Russell discusses) that “dimensional accounts” fail to differentiate between discrete emotions; such that anger, fear, jealousy, disgust, contempt, embarrassment, guilt stress, acute grief, and envy are all “unpleasant – high-activation” states, which fall close to each other. And yet the difference is not just a little more or less displeasure or arousal. Then Russell seems almost to “give away the store” by acknowledging that “all of these might occur with an identical Core Affect” – that they are not differentiated by pleasure and arousal; but that is okay, because Core Affect is “only a part” of the final EME experience. Pulling these acknowledged and unacknowledged criticisms together one gets the following incongruities: – The mountain of data on which the two dimensions of Core Affect are based come from hundreds of experimental transcripts all of which (presumably) are really self-reports out of highly-reflective Emotional Meta-Experience trials. – In EME people generally experience anger, fear, disgust, and grief as feeling very different from each other. – Yet their EME verbal reports on these experiences might all be assigned the same scores on these two dimensions of valence and arousal. – Now, if anger, fear, disgust, and grief can seem so different in EME experiences, then mathematical models which then place them at the same spot must be seriously questioned. – Despite this, these dimensional scores from EME protocols are treated by the experimenters as applying to the presumed basic Core Affect of the persons.



16.  Concluding observations 333

– If Russell’s dimensions are such a questionable measure of emotional experiences at the EME level, why would they be more accurate at the Core Affect level; – especially if one then tried to explain that EME adds to “emotional experience” in addition to that given us by CA. – Meanwhile, presumably we have no independent data based on Core Affect transcripts. Panksepp’s Take on Dimensional Affect Models. Panksepp says he welcomes tertiary-level research. For instance, he would welcome massive open-ended surveys of emotional experiences, to see what is universal and what is cultural. Both Panksepp and Russell point to some moves in that direction. However, Panksepp’s biological commitment suggests a helpful wrinkle for such trans-cultural research, in pointing out that some cultural differences derive from genetics. Meanwhile, Panksepp is no stranger to factor analysis. He has developed and tested the Affective Neuroscience Personality Scales (2003), where he found at least six distinct entities in emotional-affective space: three positive (seeking, care, play) and three negative (fear, rage, and panic/grief) – how did “lust” get left out? Notably, his factor analysis suggested overall positive and negative affect superfactors, which could be explored as to possible integration with Russell’s views. However, having a positive super factor and a negative super factor is still quite different from having positive-and-negative (or pleasure and displeasure) as opposite ends of a single factor – which is Russell’s claim. Despite his own super factors, Panksepp does not believe that the simple dimensional theories, such as Russell’s, take into account what happens at the primary process level of emotional responses – where Panksepp works. Perhaps dimensional theories of CA work at tertiary-process level – perhaps new affective circuits exist in the neocortex where there might be positive and negative paths. But Panksepp does not see them at this time. We would point out that Davidson (2000) and others find hemisphere differences in positive versus negative and/or in approach versus avoidance emotional responses. Russell’s Take on Arousal. The vertical or Y-axis dimension in Russell’s models of Core Affect and Affective Quality is generally called Arousal, with “activated” and “deactivated” as the two poles. Arousal is identified as ranging from sleep, drowsiness, various stages of alertness, to frenetic excitement. The “feeling” is one’s sense of mobilization and energy. Russell acknowledges that the combination of Arousal and Valence leads to a range of states which are not covered by “emotion”; so it is not clear what differentiates such “feelings” from body temperature, hunger, and the like.

334 Ralph D. Ellis and Bill Faw

Russell mentions the early introspective psychologists as introspecting dimensions of Arousal and Valence. For instance, Wilhelm Wundt (a century ago) had three dimensions: pleasant/unpleasant, tension-relaxation, and excitementcalm. Is Russell’s “arousal” more like “tension-relaxation” or “excitement-calm,” or a hybrid of the two? Should there be two or more “arousal” dimensions? ­Russell acknowledges the criticism that “arousal” is ambiguous. Russell states that “arousal” is not a single physiological process, but from a poorly inter-correlated network of the Autonomic Nervous System (ANS). However, that very statement shows that Russell’s dimension of “arousal” is even more ambiguous than he seems to realize, because the Autonomic Nervous System is only one of several different nervous system circuits that have been linked to the concept of “arousal.” This is the adrenaline rush visceral type of arousal, feedback-from-which, James and Lange proposed as the major source of the “conscious experience of emotion.” Part of Russell’s definition of “arousal” suggests this ANS adrenaline-rush arousal: a feeling of mobilization and energy. But the other part of his definition of “arousal” does not: arousal which moves the brain from sleep, drowsiness, various stages of alertness, to frenetic excitement. This part suggests, instead, the “cortical arousal” system, which is quite distinct from the ANS. In his writing in this book, Russell cites D. Pfaff ’s 2006 suggested “generalized arousal system,” which is (in Russell’s words) fundamental to alertness, attention, orientation, behavior, and moods and feelings. This clearly seems to be the cortical-arousal system. Panksepp’s Views on Arousal. Panksepp says that Russell’s dimension of Arousal “is not troublesome.” Panksepp says that Pfaff points to diverse brain systems which contribute to emotional and cognitive arousal, involving (in our terms) systems involved in “cortical arousal” (acetylcholine, dopamine, histamine, ­ peripheral-norepinephrine, serotonin, and orexin) and systems involved in “autonomic nervous system arousal” (epinephrine/adrenaline and centralnorepinephrine­). Panksepp reports that Pfaff prefers a unitary arousal concept at the psychological level, but is empirically undecided. There actually seem to be several related “cortical arousal” functions (Faw, 2000). Roughly, Acetylcholine seems crucial for cortical arousal during dreams as well as waking; orexin and histamine for waking; dopamine is crucial for the “seeking” system and motor control; norepinephrine for vigilant attention; and serotonin in emotional stability. Panksepp suggests that these different arousal circuits may also regulate more specific types of affective valences: acetylcholine and CRF for negative arousal; dopamine and orexin for positive arousal; and norepinephrine for both – suggesting a lack of orthogonal independence between arousal and valence dimensions. In addition, Panksepp suggests that his dopamine seeking system is a major source of energy – as well as having a specific positive valence.



16.  Concluding observations 335

Russell’s Formulations on Valence. Russell builds much of his model on the twin hypotheses that both Core Affect (CA) and Affective Qualities (AQ) can be described by single points in an area defined by the two dimensions of Valence and Arousal. He also calls the valence dimension hedonic, and labels the two poles: pleasure and displeasure. His valence dimension was derived, along with his arousal dimension, from the factor analysis mentioned above – based on a half century of guided affective self-reports. Together the “hedonic tone” and “subjective sense of energy and mobilization” constitute one’s Core Affect at any given moment. He reports that all languages have terms for “I feel good” and “I feel bad,” suggesting that they are semantic primitives. He suggests that good/bad and arousal are likely universal semantic dimensions, while “emotion” and discrete emotions such as “fear” are not universal or semantically primitive. Russell searches for “universals” in languages and cultures – while Panksepp searches for “universals” in mammal anatomy; a clear sign of the opposite extremes at which their research is done. Russell anticipates the criticism that the term “pleasure” is ambiguous; that it might relate to evaluation, liking, positive-negative, approach-avoidance, valence, utility, or hedonic tone, not all of which are identical. He clarifies some rival dimensions in emotion research, by noting that “approach” and “avoidance” are not a dimension of “affect” but of “behavior”; while “evaluation” and “liking” are aspects of his Attributed Affect or Affective Quality, rather than of Core Affect. In his writing in this book, Russell cites K. S. Smith and colleagues (2010) in their identification of the “pleasure” and “hedonic” circuit, which presumably underlies “pleasure of all sorts.” Russell’s describing this in terms of nucleus accumbens and ventral pallidum clearly identifies it as the same dopamine circuit which Panksepp sees as his seeking circuit – although his inclusion of parabrachial pons suggests a mixture of systems (perhaps related to pain). Russell makes a passing reference to S. Leknes and I. Tracey (2008 – without citation) to the effect that the neural mechanisms for pleasure and pain are similar. Panksepp’s Views on Valence. Panksepp’s work suggests that “pleasure” is complex for a different reason than Russell’s suggestion above. If each of Panksepp’s Big 7 motivational/emotional systems does represent a discrete basic emotion at the primary processing level, then Panksepp would suggest that there are at least seven different types of valence. But it is more complex than that. Panksepp distinguishes several circuits that contribute to “pleasure”: with the motivational systems of seeking, lust, care, and play yielding very different kinds of “pleasure” – each with distinct brain circuits and neurotransmitters. In the other direction, motivational systems of fear and rage have parallel but not identical pathways, and Panic has a completely different pathway.

336 Ralph D. Ellis and Bill Faw

Panksepp suggests that Russell might be proven correct in terms of dimensions, if endogenous opioids, dopamine, and (perhaps) serotonin are all found to have similar roles in positive affect – if their reinforcing properties operate through a homogenous shared higher mechanism for positive affect. Then those with “negative” effects (acetylcholine, CRF, CCK, and glutamate) would have to be shown to have a unified mechanism. If damage to those higher regions reduced general “reward” and “punishment” – it would fit with Russell. But even if those clusters of more positive circuits or more negative circuits were found to have synchrony, it is important to note that Panksepp’s Seven Circuits are not consistently examples of “pleasure” or “displeasure. The fact that over-stimulation of the seeking circuits leads to schizophrenic symptoms suggests that the dopamine circuit cannot be reduced to “pleasure” – and too much play, lust, and even care (“smothering” or “being smothered”) is aversive. In addition, anger and rage can feel very pleasurable at times – as can even fear (as in sky diving) and panic (by some masochists). They are primarily “motivational” circuits with “emotional” experiences involved. The most “collapsing” of the pleasure circuits that Panksepp’s research might allow is to distinguish between ap­ petitive (dopamine system) and consummatory (endogenous opioids) pleasure. Panksepp concedes that perhaps primordial negative and positive affect may have existed before evolution of more discriminating affects – but that there is little evidence of that. In Panksepp’s analysis, not only is there more than one pleasure and more than one displeasure system, but “pleasure” and “displeasure” are clearly not opposite ends of a single “dimension,” which would be implied by labeling a “hedonic: pleasure/displeasure” dimension. One of the major lessons that Olds and Milner (1954) learned from their original brain stimulation studies is that the circuits that seem to underlie “pleasure,” and thus would lead to self-stimulation, are distinct from the circuits that seem to underlie “displeasure,” and thus would lead to lever pressing to turn off stimulation. They found this to be a significant refutation of simple drive theories that saw pleasure as the cessation of displeasurable drives, or simple hedonic theories that might see displeasure as merely the cessation of pleasure. Moreover, Panksepp in his Affective Neuroscience presents extensive evidence that self-stimulation can result from either pleasure or from fulfillment of the seeking system, and that there are anatomically distinct areas of the hypothalamus whose stimulation leads to the two types of reward. Panksepp wants to distinguish the satisfaction of the seeking system from “pleasure.” “Consciousness” Pathways. Russell cites M. L. Kringelbach (2010) as suggesting that the two purported biological circuits underlying Core Affect may be blended and brought into consciousness by the orbital prefrontal and anterior cingulate gyrus, which serve as the global workspace integrating access to



16.  Concluding observations 337

c­ onsciousness. This proposal seems more persuasive than the possibility that there are specific “arousal” and “pleasure” circuits – especially if anterior insula are taken into consideration. However, Panksepp makes a good case for the possibility of at least primary process conscious experience at the sub-cortical level, based on brain stimulation studies and the emotionality of decorticate children. This is also consistent with Damasio’s (1999) view that “core consciousness” is present as long as brain stem and lower limbic areas are intact, although knocking out more and more cortical areas renders the content of consciousness less and less sophisticated. Priority of the Biological Approach? This commentary began with the analogy of the First Intercontinental Railway created in the United States, whereby the Central Pacific Railroad (heading East) and the Union Pacific Railroad (heading West) met and linked tracks at Promontary Summit, Utah. This was an analogy of what it might take for an affective science effort starting at the biological level, like Panksepp’s, to “meet” – be compatible with – an affective science effort starting at the verbal level, like Russell’s. However, this analogy suggests that both tracklaying efforts were of equal importance and equally “authoritative.” That is, if they did not meet, it could be either railroad’s fault. Neither railroad was the one that the other had to meet! Does this part of the analogy hold up when one “railroad” is a biological approach and the other is a psychological approach? Is starting from the tertiary “correlational” psychological level of equal importance and authority as starting from the “causative” biological level, or is there supremacy to the biological approach? Both Panksepp and Russell seem to believe that robust research must be done both on the higher philosophical/psychological levels and on the lower biological levels. The classic formulation of this bifurcated research is that ultimate “definitions” and “explanations” of psychological events “point downward.” What Panksepp calls the “Affective Neuroscience thesis” – that all pre-linguistic universals of the mind can only be defined coherently through an understanding of the underlying neurological processes – is really the thesis of a broadlyphysicalist scientific method. On the other hand, we must remember that the psychological research on which Russell’s approach relies is every bit as empirical as any biological study. If psychological phenomena occur at all, then physicalists must presume that they can be equated with some combination of physical occurrences. And all physical occurrences both give and receive causal impact. So we cannot say that psychological phenomena have no causal impact, if at the same time we are assuming that they are equivalent with some combination of physical occurrences. Moreover, we often are able to identify causal relations at a higher level of process even

338 Ralph D. Ellis and Bill Faw

when we have not understood the lower level mechanisms involved. We knew for millenia that fire can cause water to boil, without understanding what was going on at the molecular or atomic levels – and we still know very little about these phenomena at the level of quarks and superstrings. (See Ellis & Newton, 2010 for further discussion of the causal relations between lower level and higher level processes in the brain and consciousness.) How do these reflections relate to the Russell-Panksepp “transcontinental railroads”? Russell claims that his Core Affect (CA) is psychologically primitive and cannot be analyzed further without venturing into biology. Thus Russell hopes, eventually, to show how his Core Affect construct emerges from brain activities, and the evolutionary evidence of that. Panksepp believes his own work can help in that, but urges Russell and others working at the psychological level to frame questions to be linked to causal brain research – with the assumption that biological analysis might provide new perspectives for conceptualizing CA. In regard to this great dialog between Russell and Panksepp, where the “metal wheels hit the tracks” is the point where Russell asserts that the dimension of valence and arousal are “biological primitives,” but that the blended Core Affect is a “psychological primitive.” Most of Russell’s earlier work on the various components of emotional episodes can be analyzed at a fairly high “level.” But his Core Affect – and the possible circuits underlying valence and arousal – ventures into biology; and biology might radically alter the psychological constructs.

2. Important convergences between Russell and Panksepp Up to this point, there has been a great deal of emphasis in this volume on the conflicts between Russell’s dimensional approach and Panksepp’s “natural category” approach to emotional phenomena. But it is extremely instructive to note the remarkable number of convergences between the two theories – especially when these convergences, resulting from careful thinking and research along two entirely different lines, lead in directions that significantly diverge from often-unquestioned assumptions that many would regard not only as traditional wisdoms of the sciences of emotion, but often even as unquestionable assumptions. Our concluding section will focus on these Russell-Panksepp convergences emerging from this dialogue. First, are there such things as “basic emotions,” and if so, in what sense? If there are such animals, to what extent are they important in determining the overall affective life? And in either case, how do we account for the robust unity, regularity, and intentional meaning of many affective processes and their physiological and behavioral concomitants? Russell and Panksepp are re-opening these questions in exciting ways.



16.  Concluding observations 339

Many psychologists and neurophysiologists assume that certain emotions are “basic” in a dual sense: (1) we come into the world biologically hardwired to exhibit them in response to certain kinds of predetermined stimuli; and (2) all other emotional feelings, no matter how nuanced, are “based on” these “basic” ones through associative learning, through either operant or classical conditioning, and by means of combining or “blending” the subjective properties of the basic emotions in various ways, with various degrees of intensity. Both Russell and Panksepp seriously call into question many facets of this simplistic view, in different ways, and offering different approaches with which to replace the simpler view, based on their differing levels of analysis (neurophysiological on the one hand, psychological on the other). Russell seems to reject both points (1) and (2), while Panksepp regards point (2) as an oversimplification, while calling for a careful rethinking of point (1), which he could accept only with important qualifications. Among the interesting points of congruence is the way both prompt a rethinking of these kinds of basic assumptions. For Russell, no one category of emotions, such as fear or anger, is any more “basic” than any other, and in fact the categories “fear” and “anger” themselves are somewhat arbitrarily selected as unitary categories, let alone as “basic.” Instead, what we have are cases that people in folk taxonomy call “fear” and “anger” because they seem to be “prototypical” cases where certain stereotypic features happen to coincide in a marked way – certain kinds of objects with certain behavioral tendencies, motivational goals, and subjective qualities – but these cases are no more frequent or “typical” than any other, and except for the two hardwired dimensions of valence and intensity, there can be an infinite variety of different objects, goals, and subjective descriptions of different kinds of emotions, none any more “basic” than the others. What is basic is a constantly changing “core affect,” and the objects, aims, and behavioral tendencies that we associate with these changes can vary widely. It is important to realize that “core affect” for Russell does not refer to anything like “basic emotion” in its classic sense. Core affect refers only to the valence and intensity of how we feel without any regard to intentional referents, causal attributions, triggering stimuli, or even behavioral concomitants. These other elements are external appendages to core affect, and the addition of these completely independent and separable elements is what can allow the whole conglomerate – including core affect – to add up to folk psychological concepts like “fear” and “anger.” Any of the elements of fear and anger can occur without the others, and the same core affects may or may not be associated with any particular causal attributions, triggering stimuli, systemic bodily responses, or even behavioral outcomes. The folk psychological emotion words are analyzable into more elementary components, one of which is core affect. Fear, anger, or lust per se may

340 Ralph D. Ellis and Bill Faw

be “basic emotions” in traditional parlance, but they are not by any means “core affects” in Russell’s sense. Panksepp accepts point (1) of the “basic emotion” view as described above, with some significant qualifications, but it is important to notice that, like Russell, he rejects many aspects of point (2). Panksepp’s research does indicate that certain brain emotion systems are hardwired with a tendency for certain neural patterns to trigger certain bodily responses, and for the neural patterns to be triggered in turn by both bodily and environmental circumstances. But Panksepp thoroughly rejects the behaviorist notion that emotions are geared toward maximizing pleasure, optimizing consummatory reinforcements, or attaining outcomes. The emotional systems drive behavior because we are wired to behave in those ways – not because the stimulus “causes” a response, and not in order to achieve some reward extraneous to the behavior of the specific emotional system being activated. Moreover, the behavioral tendencies triggered by the emotional systems are general tendencies that leave a good bit of flexibility in terms of overt behavioral outcomes and, in cortically sophisticated animals like humans, there is a genetically pre-ordained plasticity to the cortical components of the cognitive processes that are triggered and energized by the activation of any of the evolutionarily inherited emotional brain systems. For Panksepp, there is also a fair amount of variability in the response of the basic brain emotion systems themselves, because the basic patterns are natu­ ral kinds rather than logical kinds. Like other natural phenomena such as tidal waves, earthquakes, or the formation of ice crystals, they have a strong tendency to occur in certain patterns when they do occur, but there can be some degree of variation in the way things play out. Throughout Panksepp’s extensive research and analysis, we see that one of the important variations for the way emotions trigger behavior, in primates and other higher mammals, is that emotions tend to be hardwired in such a way as to cause them to be mediated to some extent by the interaction of cortical and subcortical processes, so that what gets energized by the activation of an emotion system is, to a certain extent, a process of cognitively executing learned or even thought-out behavioral responses to a situation, rather than only a set of pre-programmed overt behavior patterns. To be sure, associative learning through classical conditioning plays a role, but there are also genetically hardwired feedback loops between these cognitive and emotional processes. A clear example is the way Kim et al. (2007) and Lungu et al. (2007) find that prefrontal and anterior cingulate processes can modulate amygdala activation in fear and anger, while at the same time George Bush et al. (2000; no relation to the U.S. president elected in that same year) find that amygdala activation can dampen prefrontal contributions to critical thinking. These findings are consistent with



16.  Concluding observations 341

Pankepp’s proposal that some degree of flexibility and adaptability in our emotional responses is already hardwired, and is not completely attributable to simple classical conditioning, although Panksepp, like Russell, does allow considerable importance to the role of learning. Panksepp goes into a good bit of detail on the way the balances of excitatory and inhibitory neurotransmitters play intricate roles in these processes. This too is consistent with the idea that cortical, limbic, and subcortical interactions can display subtle variation and flexibility. Another important qualification to the “basic emotion” view emphasized by Panksepp (and in agreement with Russell) is that emotions are not merely caused by an external stimulus, or even our own representation of the stimulus. Instead, the organism itself is already motivated to energize certain patterns, prior to the presentation of a stimulus. The clearest example is the seeking system, which must be energized and activated prior to the finding of an interesting stimulus. The organism is not simply caused to have the goals that it has by something external to itself, but also and much more importantly by its own complex homeostatic and dynamical internal processes. To a great extent, the goal is set prior to the presentation of the stimulus, and one of the reasons certain kinds of stimuli are deemed relevant to the organism’s emotional interest is that they either facilitate or thwart those already ongoing goals. So stimuli do not simply “cause” emotional episodes to occur. They usually are only part of what triggers the episode, along with endogenously regulated and energized mostly-subcortical homeostatic processes that are always already purposefully underway. The external stimuli are typically identified as salient only because they often are the most visible part of what appears to us as triggering an emotional episode. Russell and Panksepp agree that non-human mammals have consciousness, but Russell questions whether we can infer from an animal’s behavior or neurophysiology that it subjectively feels emotions similar to ours, especially given that even humans can display emotional behavior without conscious feeling. Panksepp’s position here hinges on the analogy between the human and other mammals’ emotional brain systems, which Panksepp finds to be almost identical anatomically, and consistently analogous in terms of the functioning and distribution of neurotransmitters. If brain function B correlates with conscious emotion C in humans, and B is the same in other mammals as in humans, then Panksepp’s conclusion is that C must be very similar in other mammals. But, as some of the commentators in this volume have argued, we must not neglect the contributions of the neocortex to the way conscious states in general feel – and thus there might be inter-species disanalogies as well as analogies when it comes to the way emotions consciously feel. Panksepp generally speaking grants this point to a limited degree, but also reminds us that, even in humans, there can be conscious emotional feelings even where large segments of the neocortex have been disabled.

342 Ralph D. Ellis and Bill Faw

The point of congruence between Russell and Panksepp here is that the degree of interspecies similarity must be resolved empirically, not a priori. Russell and Panksepp converge on the issue of the complexity of relationships between brain function and specific emotions, or conscious states generally. Russell uses the example that the amygdala is not simply correlated with fear responses. Ablation of the amygdala, for instance, leaves an unconditioned fear response intact, and of course the amygdala modulates other functions besides fear. Panksepp is well aware and in agreement with this point. In fact, he emphasizes that the amygdala should not be considered the “seat of fear,” but rather that the role of the amygdala in fear is mostly to learn which stimuli to associate with fear, whose substrates reach much deeper into the subcortex, especially the PAG. Although there are crucial differences in Russell’s and Panksepp’s use of the term “core affect,” there is also an important similarity. Both seem to be pointing to the aspect of emotion that is neurophysiologically separable from the attribution of emotion to particular triggering stimuli. In the case of the amygdala, for example, core affect is the aspect of brain functioning that would have been similar even if the animal had had a different learning history, and therefore had responded with the same core affect to completely different stimuli or environmental situations. Since the role of the amygdala is more important in learning which stimuli should trigger fear than in the actual feeling of fear (which can occur even with amygdala ablation), it follows that the amygdala plays less of a role in the core affect than does the PAG, which contributes to the feeling regardless of the animal’s learning history or cortical function. So to this extent, Panksepp’s notion of core affect is pointing us in a direction similar to Russell’s – i.e., to the distinction between brain functions needed for the actual feeling of a given emotion, versus the functions needed for attribution of the feeling to specific stimuli and situations. The fact that Panksepp also emphasizes that we come into the world already pre-programmed to exhibit an emotion such as fear toward certain highly specific stimuli does not contradict this point. With experience and cortical development, the range of objects and situations that elicit a given emotion will change, and we will also learn to inhibit or otherwise manage the emotion in various contexts. But all of this learning involves different brain functions from those needed for the core affect per se. On that point, Russell and Panksepp are substantially in agreement. And part of the take-home message here is that we cannot simply observe through brain imaging that certain brain areas “light up,” and then infer from this that the ones that “light up” are the ones needed for feeling that particular emotion. Moreover, as Panksepp also warns us, many popular brain imaging techniques such as fMRI are based on changes in blood flow, whereas much less



16.  Concluding observations 343

change in blood flow is needed for subcortical than cortical function. So reliance on fMRI tends to lead to a cortical bias, which can further blur the distinction between core affect and the relevant associational functions. This distinction is equally important under both Panksepp’s and Russell’s definitions of “core affect,” in spite of the differences mentioned above. As a result of the interactions between Panksepp and Russell in this volume, we learn that they are not as far apart as originally perceived on the question as to whether all human experience can be reduced to only two dimensions. As Russell clarifies, this is not in fact his position, contrary to some of the popularized versions of the dimensional view that one sometimes hears in non-professional or in overly-simplified discussions. The difference here seems to hinge partly on what is different in the different meanings of “core affect” intended. What Panksepp means by “core affect” includes more than what Russell means by the same term, as indicated above. Panksepp is referring to what his research suggests are “natural kinds” – clusters of behaviors and brain events such that, while there is no logical necessity that they must occur together, nonetheless, given the structure of the organism and its relations to its environment, there is naturally a strong tendency for those elements to go together. But in spite of this difference, both are on the same track in suggesting that the phenomenology of actual emotional feelings includes more than the two dimensions of pleasantness and intensity. For Russell, those are the two dimensions of core affect, but a phenomenally experienced emotional feeling includes more than core affect; as his discussion of hunger shows, it includes many sensations and associations besides just the core affect. All those associations contribute to the intentionality, the experienced “aboutness,” or the “of-ness” that characterizes intentionality, and in experience those cannot be cleanly separated from the sensations that go with them. The crucial point here, in which, again, Russell and Panksepp tend to converge, is that we should be cautious in making oversimplified assumptions about the intentionality – the intentional meanings and referents – of an emotional experience or episode. The following distinctions, in fact, are crucially underscored by Russell’s “constructivist” approach, and are important even for those, like Panksepp, who want to guard against extreme contructivism: First, we should not automatically equate the perceptual object that we associate with a concurrent emotional feeling or behavior with the “cause” of the emotion. The way language is used in much of the emotional literature tends to suggest such an unwarranted equivalence. An obnoxious motorist might trigger rage, but in most cases he is neither the cause nor the intentional object of the rage, even though he may be the intentional object of the perception that triggered the rage. The cause of the rage may be extremely complex, including the

344 Ralph D. Ellis and Bill Faw

c­ umulative ­thwarting of motivations that have nothing to do with the motorist who cut me off. And the intentionality of the rage – that to which the rage refers – typically consists of the people or circumstances that have thwarted those motives, and which again have nothing to do with the motorist. Moreover, as both Panksepp and Russell point out, I may remain largely un­ conscious of both the intentional aboutness of the emotion and the people or circumstances that actually thwarted my motives and thus precipitated the rage. If I do equate the motorist with either the cause or the intentional aboutness of the emotion, I am usually in a condition that, in another context, Lambie (2008) has referred to as a “lack of emotional awareness” (see also Ellis, 2005, 2008b) – i.e., a misunderstanding of the actual meaning and/or content of one’s own emotional experience. As Russell emphasizes, such a lack of self-awareness is very typical and widespread in everyday interpretations of one’s own emotional experience. In a more extreme form, this lack of emotional awareness would become alexithymia – the inability to know what one is feeling or what the feelings mean (Ellis, 2008a). Sundararajan (see Ellis 2008b) reports a case in which a man understood intellectually that he must have murdered his father out of anger, but had not consciously felt any anger toward his father. Acknowledging these distinctions also warns us against an oversimplified account of the causation of the actual brain mechanisms in an emotional expereience. Oversimplified accounts often begin this way: “First we perceive an object….” Already, the temptation here is to see the perceived object as the first item in a sequence of causes and effects. But the failure of my employer to give me a pay raise, in spite of my obvious merit, involves a series of events that had already taken place long prior to the perception of the motorist cutting in front of me. The causation of the emotional episode was already underway, and in many instances actually contributed to the selection of a triggering stimulus; thus the concurrent perceptual object that served as trigger may typically play little causal role. More important, it is usually counterproductive and oversimplified to think of the perception of the triggering event as the initial event in a chain of neurophsiological causes and effects. The brain correlates of the thwarting of motives, as well as the motives themselves, all were already underway before the trigger presented itself, and in many instances may even have caused the directing of attention to the trigger. In fact, there can be many cases in which a triggering object is merely imagined without any environmental change causing the imagined trigger. The clearest instances are in cases of cortical micro-stimulation, in which direct stimulation of the cortex causes the subject to have the vivid image of some particular object such as a musical melody, along with accompanying emotional feelings. And it is well known that perceptual brain areas must be activated to facilitate the forming



16.  Concluding observations 345

of the image. So we should be cautious in attributing causal roles to perceptual processes in the producing of an emotion or emotional episode. It is interesting to notice, then, that a volume that began with a contrast between two divergent ways of conceptualizing emotional processes also ends up not only with interesting convergences, but in fact with important points of agreement that run against the grain of much of the traditional wisdom. The fact that such contrasting approaches converge in these cutting edge directions may be a further indication that we should all take them seriously in our own work.

References Bush, G., Luu, P., & Posner, M. I. (2000). Cognitive and emotional influences in anterior cingulate cortex. Trends in Cognitive Sciences 4 (6), 215–222. Cannon, W. (1929). Bodily Changes in Pain, Hunger, Fear and Rage: an Account of Recent Re­ searches into the Function of Emotional Excitement. New York: D. Appleton. Damasio, A. (1999). The Feeling of What Happens: Body and Emotion in the Making of Con­ sciousness. San Diego: Harcourt. Davidson, R. J. (2000). The neuroscience of affective style. In The New Cognitive Neurosciences, 2nd Ed; Michael S. Gazzaniga, Editor-In-Chief. MIT Press: Cambridge, Mass. Ellis, R. D. (2005). Curious Emotions: Roots of Consciousness and Personality in Motivated Action. Amsterdam/Philadelphia: John Benjamins. Ellis, R. D. (2008a). The phenomenology of alexithymia as a clue to the intentionality of emotion. In Louis Charland and Peter Zachar (Eds.), Fact and Value in Emotion, pp. 181–192. Amsterdam/Philadelphia: John Benjamins. Ellis, R. D. (2008b). In what sense is “rationality” a criterion for emotional self-awareness? Consciousness and Cognition, 17, 972–73. Ellis, R. D., & Newton, N. (2010). How the Mind Uses the Brain. Chicago: Open Court. Faw, B. (2000). Consciousness, motivation, and emotion: Biopsychological reflections. In R. D. Ellis & Natika Newton (Eds.), The Caldron of Consciousness: Motivation, Affect and SelfOrganization. Amsterdam: John Benjamins. James, W. (1884). What is an emotion? Mind, 9, 188–205. Kim, S. H. & Hamann, Stephan (2007). Neural correlates of positive and negative emotion regulation. Journal of Cognitive Neuroscience, 19, 776–798. Lambie, J. (2008). On the irrationality of emotion and the rationality of awareness. Conscious­ ness and Cognition, 17, 946–971. LeDoux, J. (1996). The Emotional Brain: The Mysterious Underpinnings of Emotional Life. New York: Simon & Schuster. Lungu, O. V., Tao, L., Waechter, T., Willingham, D. T., & Ashe, J. (2007). Strategic modulation of cognitive control. Journal of Cognitive Neuroscience, 19, 1302–1315. Olds, J., & Milner, P. (1954). Positive reinforcement produced by electrical stimulation of septal area and other regions of the rat brain. Journal of Comparative Physiological Psychology, 47, 419–427. Panksepp, J. (1998). Affective Neuroscience: The Foundations of Human and Animal Emotions. Oxford: Oxford University Press.

346 Ralph D. Ellis and Bill Faw

Pfaff, D. (2006). Brain Arousal and Information Theory: Neural and Genetic Mechanisms. Cambridge, Mass.: Harvard University Press. Russell, J. A. (2003). Core affect and the psychological construction of emotion. Psychological Review, 110/1, 145–172. Smith, K. S., Mahler, S. V., Peciña, S., & Berridge, K. (2010). Hedonic hotspots: Generating sensory pleasure in the brain. In M. L. Kringelbach & K. C. Berridge (Eds.), Pleasures in the Brain. Oxford: Oxford University Press.

Index

A acetylcholine  43, 46, 64, 66, 121, 334, 336 affect program  9, 14–16, 23–26, 81–82, 217, 225, 257–263, 265–275, 288, 307 Affective Neuroscience Personality Scales  67, 73, 333 affective quality  85, 89–90, 92, 97–98, 117, 124, 137–139, 165–166, 226, 294, 328, 330, 333, 335 aggression  99, 109, 114–116, 133, 205, 222, 318–319, 325 alexithymia  344–345 algorithmic level  25, 233–234, 263–264, 266–267, 269–270, 272–273 amygdala  34, 50–51, 69, 73, 111–112, 115–116, 130, 133, 194–195, 199, 201, 236, 271, 277, 315, 325, 328, 340, 342 anger  7–8, 14–15, 17, 21, 26, 32, 35–36, 47, 55, 65, 68–69, 79–80, 82, 85, 88, 90, 92, 94, 100–101, 105–106, 110, 113–114, 131, 133, 135, 137–139, 146–147, 154, 182, 191, 194, 197, 205–206, 209, 212–216, 221–222, 228, 239–240, 261, 263, 270, 272, 274, 286–291, 298, 301, 306, 315, 322, 325, 332, 336, 339–340, 344 anoetic experience  56–57, 77, 124, 127, 238, 254–255, 311 anterior cingulate  34, 87, 127, 325, 327, 336, 340, 345 anxiety  15, 36, 50–52, 55, 69, 73, 78, 92–93, 99, 108, 115, 127, 130, 145, 195–196, 200, 205, 236, 308, 315, 319, 331 appetitive  34, 76, 145, 324, 336

appraisal  18–19, 24–26, 58, 82, 84–85, 91, 103, 115, 137, 182, 189–191, 193–195, 198, 201, 222, 225, 239, 257–278, 283, 292, 296–297, 307 arousal  2, 6, 10, 13, 22–23, 32, 34, 36–38, 41, 43, 46, 51–52, 54, 57–58, 60, 64–65, 82, 86–89, 95–97, 100–101, 108, 110–112, 115–116, 119–121, 127, 130, 132–133, 137–138, 167–170, 194–196, 199, 206–207, 225– 227, 312, 323–324, 328–329, 331–335, 337–338, 346 attention  13, 27, 50, 54–55, 83, 87, 89, 91, 101, 104, 112, 116, 131, 149–150, 153, 186–187, 195, 205, 238–239, 259, 287, 297, 308, 334, 344 attributed affect  165–166, 170– 171, 280, 295–296, 329, 335 auditory  cortex 40, 256, 325 autonoetic experience  56, 71, 77, 123–124, 127, 238, 330 autonomic nervous system  42–43, 73, 82–83, 88, 92, 106, 108–109, 111–113, 115–116, 119, 129, 139, 144, 148, 151–153, 176, 251–252, 281, 283, 291–293, 325, 329, 334 B basal ganglia  34, 308 basal-ganglia thalamic loops  40 basic affects  23, 39–42, 44, 58, 120, 326 basic emotion  7, 10–11, 15, 26, 39, 50, 67, 69, 79, 94, 106, 119–120, 122, 261, 277, 281, 296, 303–304, 312, 317, 324, 331, 335, 339–341 behaviorism  2–3, 16, 31, 70, 180

bottom-up approach  4, 23, 29, 65, 89, 91, 116, 175, 203, 221, 227, 229, 232, 302, 307–308, 313–314, 325 brain stem  34, 251, 337 Broca’s Area  60, 192 C care  10, 33, 35, 37–38, 49, 54, 67, 78, 125, 136, 138, 143–144, 162, 203, 205, 231, 234–235, 261, 301, 314, 324, 326, 333, 335–336 causality  19, 115, 213, 215, 217, 219, 221, 314 circumplex model  6, 8, 29, 86, 115, 117, 173, 190, 239, 280, 300 conditioned place preference (CPP)  44, 47, 75 conditioned responses  44, 64, 67, 75, 98, 107, 283, 316, 322 conditioning  32, 39, 52, 68, 72–73, 77–78, 107, 110, 133, 136, 180, 203, 259, 283, 300, 308, 339–341 consciousness  1, 5, 24, 27–29, 31, 39, 46, 48, 56–57, 73–78, 87, 90–92, 96–99, 101–104, 110, 112, 114–118, 122–124, 126–127, 129–131, 133, 238– 240, 253–256, 300–301, 307, 311, 318–319, 326, 328, 330, 336–338, 341, 345 constructivism  227, 277, 307 core affect  2, 8–11, 13, 17–19, 22–24, 26, 30–31, 51, 65–67, 77, 84–92, 94–105, 110–113, 119– 121, 124–126, 130–132, 137–139, 165–170, 182–183, 225–226, 230–232, 235–237, 239–241, 243–256, 259–260, 264, 270– 271, 294–298, 312–313, 326–333, 335–336, 338–339, 342–343

348 Categorical versus Dimensional Models of Affect

core self  50, 76, 317 cortex  34–35, 40, 43, 60, 67, 71, 73–75, 77, 87, 112, 157–158, 162, 246, 251, 283–284, 299, 314, 320, 325–327, 344–345 cortical micro-stimulation  344 cortical processes  4, 10, 16, 24, 34–35, 40, 43, 50, 56, 59, 61, 73, 75–76, 112–115, 126, 155– 157, 162, 172, 236, 246, 248, 250–254, 256, 284, 313–315, 317–318, 323, 325, 334, 337, 340–344 Corticotrophin Releasing Factor (CRF)  36, 64, 66, 121, 316, 326, 334, 336 culture  12, 17, 22, 29, 61–62, 70–71, 78, 80, 92–93, 105–106, 117, 133, 139, 142, 166–167, 170, 192, 196–197, 201, 210, 218– 219, 221, 223, 240, 285–286, 298, 300, 309 curiosity  35, 47, 65 D decision-making  34, 194, 196, 314 decortication  38, 59, 76, 159, 173, 308, 314, 319 depression  4, 35, 61, 72–74, 85–86, 99, 165, 190, 234, 239, 250, 256, 306, 310, 316, 318–320, 324–325 dimensions and dimensional model  6, 8, 13, 17–18, 39, 45–46, 48, 55, 57–58, 63, 67, 117–121, 125, 131, 165, 167, 192, 194, 203–204, 206–207, 237, 295, 306, 312, 315, 330–333, 338, 343 discrete emotions  14–15, 79, 99–100, 135, 137–139, 150–154, 206–208, 277, 290, 304, 322, 327, 332, 335 discrimination studies  53–54, 58, 123 disgust  7–8, 50, 63, 105, 113, 146, 197, 236, 240–241, 261, 332

displeasure  2, 6, 22, 86–88, 90, 94–95, 97, 100, 110–111, 132–133, 137–138, 192, 250, 295, 323, 328, 331–333, 335–336 disposition  215, 279, 298 dopamine  35, 46, 54–55, 64, 66, 74, 76, 78, 121, 324–326, 334–336 dopaminergics  66–67, 241 dual-aspect monism  63, 310 E eliminative materialism  18–19, 176–177, 289, 310 emergence  20, 31, 38, 49, 62, 76, 162, 184, 236, 308, 310 emotional episodes  8–9, 12, 15–16, 30, 47, 81–85, 88, 91–111, 147–148, 164–166, 231–232, 244–249, 252–253, 277–285, 280–299, 323–328, 331, 338, 344–345 emotional meta-experience  85, 90–92, 95–96, 98, 101–103, 105–106, 111–114, 124, 129, 137–139, 166, 226, 239, 244, 247, 259, 264, 270, 282–284, 294, 298, 327, 329–330, 332 emotional vocalizations  52, 54–55, 123, 197, 327 empathy  52, 70, 72–74, 104, 171, 186 envy  49–50, 77, 236, 332 epigenetic effects  38, 62, 77, 303, 305 essentialism and essence  9, 15–16, 27–28, 31, 88, 106, 193, 207, 213, 280, 294, 304, 305 estrogen  36–37, 78 evolution  2, 21–22, 29, 32, 36–37, 39, 46, 70, 72, 75–76, 96, 119–120, 125, 145, 158, 172, 190, 192, 194, 199–200, 210, 212, 217–218, 230–231, 316, 336 explanatory gap  63 expression  1, 7–8, 30, 62, 73, 77, 140, 149–150, 177, 179, 191– 192, 195–196, 199, 236–237, 261, 277, 281, 299–300, 316

F factor analysis  17, 19, 67, 99–100, 333, 335 fear  7–8, 13–15, 33, 35–36, 38, 41, 43, 48–53, 65, 67–69, 73– 77, 79–83, 85–86, 88, 90–92, 94–96, 99, 101–102, 105–115, 124, 127, 129–131, 135–154, 162, 164, 166, 169–170, 176, 180, 182–184, 191, 194–195, 197, 199–200, 203, 205–206, 209, 212–213, 221, 225, 231, 236, 240, 253, 261, 263, 265, 274, 281–283, 286–293, 298–301, 308, 315–316, 319, 322–327, 329–330, 332–333, 335–336, 339–340, 342, 345 feeling  1, 4, 31, 41, 43, 46, 51, 58, 64, 72–73, 80, 84–88, 91, 94–96, 99, 102, 104, 110, 112, 124, 126, 129–130, 132–133, 136–138, 140, 160, 165, 172, 190, 197, 200–201, 206, 210, 217, 223, 229, 238, 242–243, 258–259, 262–263, 270, 274, 276, 282–283, 294–295, 298, 318, 323, 328, 332–334, 341–345 fMRI  56, 73, 75, 125, 127, 133, 198, 300, 342–343 folk psychology  81, 83, 135, 176, 282, 289, 330 forebrain  34, 39, 72, 157, 318 freezing response  24, 36, 85, 107, 109, 111, 136–137, 145, 147, 150–153, 292–293 frontal cortex  34–35, 67, 326 functionalism  190, 193 G glutamate  36, 66, 178, 256, 336 grief  3, 33, 37, 53, 67, 91, 136, 138, 143–144, 203, 205–206, 208, 213–214, 222, 261, 301, 316, 324, 326, 332–333 H homeostasis  160, 164, 282, 285 homeostatic affects  41, 46–47, 49, 53, 66, 120, 247, 323



homeostatic property cluster model  14, 305 homologies  39, 48, 125, 158, 305, 315 hunger  13, 41, 44–46, 49, 63, 93, 121, 129, 131–132, 136, 201, 230, 240, 259, 294, 323–324, 333, 343, 345 hypothalamus  29, 34, 111, 157, 194, 315, 318, 325, 336 I implementational level  151, 233, 264, 267–268, 273 infants  12, 37, 49, 52, 59, 76–77, 95–96, 98–99, 105, 157–158, 191–192, 199, 206, 236, 318, 327–328 information processing  43, 190, 193–196, 283, 325 inhibitory neurotransmitters 341 instinct  22, 280 intentionality  12, 19, 28, 58, 103, 182, 239, 343–345 J jealousy  26, 50, 80, 90, 94, 110, 113, 139, 146, 234, 238, 261, 263, 270, 332 L language  12, 18, 23, 27, 32, 33, 53, 55–56, 58, 60, 66–67, 71, 80, 92, 94, 98, 100–102, 106, 110, 113, 117–118, 124, 133, 140, 143–144, 146–147, 208–209, 237–238, 286, 297, 303, 330, 343 laughter  4, 37, 51–52, 54–55, 75–76, 171 Law of Effect  45, 99 learning  3, 29, 32, 34–36, 38–39, 41–47, 50–51, 53, 56, 58, 65, 68–72, 91, 114, 123, 132, 136, 151–152, 192–193, 235–237, 270, 284, 291, 302–303, 310, 330, 339–342 levels of analysis  9–10, 20–21, 27, 33, 39–40, 63, 65, 69, 79,

Index 349

111, 119–120, 203–204, 209, 232–233, 263–264, 304, 307, 312, 324, 339 levels of control  32, 45, 50, 307 libido  34, 324 limbic system  5, 34, 81, 127, 146, 315, 337, 341 longitudinal neural networks  34 lust  33, 36, 38, 45, 49, 54, 131, 136, 138, 143–144, 162, 203, 205, 207, 231, 261, 301, 324– 326, 329, 333, 335–336, 339 M mania  85, 324–325 maternal care  35, 37, 78 Medial Forebrain Bundle (MFB) 34–35, 72, 157, 318 midbrain  34–35, 51, 74, 251, 325 Minnesota starvation experiments  45 mirror neurons  20, 60–61, 70, 186–187 modules  44, 71, 117, 234 motivational systems  324–326, 335 music  55, 75, 85, 87, 91, 104, 158, 238, 284 N natural kind  2, 9, 11, 15, 17, 21, 26, 28, 30, 72, 74, 200, 210, 212, 223, 276–277, 284, 322–323, 340, 343 neocortex  56, 67, 70–71, 308, 313, 317, 333, 341 networks (neural)  32–34, 39, 41–44, 46, 49–51, 61, 65–66, 72, 76, 120, 137, 143–144, 151–153, 176, 291, 301, 303–305, 308, 311, 313, 320 networks (emotional)  32, 42–44, 49–50, 304 neuropeptides  35–36, 43, 45, 48–49, 64, 76, 121, 184, 252–253, 316 noetic experience  56, 77, 124, 127, 238 norepinephrine  43, 46, 64, 121, 305, 325, 334

nucleus accumbens  34–35, 74, 87, 157, 325, 335 nurturance  37, 65, 301, 326 O occipital cortex  40, 314 opioids  35–37, 66–67, 127, 228, 241, 306, 310, 324–326, 336 orbitofrontal cortex  34 orexin  35, 64, 121, 334 oxytocin  35–37, 49, 77, 306, 319, 325–326 P panic  3, 15, 33, 37–38, 49, 51, 53, 67, 80, 92, 132, 136, 138, 143–144, 157–158, 161, 163, 169, 203, 205, 228, 231, 238, 261, 301, 308, 316, 319, 324, 326, 333, 335–336 Periaqueductal Gray (PAG)  34, 51, 69, 72, 74, 111, 252, 315, 325, 328, 342 PET  50, 55, 69, 109, 125, 133, 185, 300, 315 phenomenology  27–28, 311, 343, 345 play  4, 7, 10, 12, 19, 28, 33, 37–38, 49, 53–55, 64, 67, 76, 80, 95, 101–102, 107, 113, 123, 131–132, 136, 138, 143–144, 162, 173, 176–177, 183–184, 203, 205, 210, 240, 252, 261, 282, 284–285, 295, 301, 319, 324–326, 333, 335–336, 340–341, 344 pleasure  2, 6, 10, 22, 26, 35, 72– 73, 86–90, 94, 96–97, 99–101, 108, 110, 115–117, 132–133, 136–138, 285, 295–296, 299, 323, 326, 328, 331–333, 335–337, 340, 346 positive affect  34–35, 37, 45–46, 54, 66, 123, 133, 194, 241, 278, 318, 327, 336 post-traumatic stress disorder 195, 205 pre-linguistic affect  12, 41, 53–55, 63, 100, 110, 192, 237–238, 255–256, 337

350 Categorical versus Dimensional Models of Affect

primary process (affect/emotion) 5, 9–13, 16, 23, 31–34, 38–46, 48–54, 56–62, 64, 67–71, 107, 119–125, 136–137, 158, 162–164, 167–172, 203–204, 214, 220–221, 230, 235–237, 240, 243–245, 248, 250–256, 306–309, 312, 319, 327, 329, 333, 337 prolactin  37, 326 prototypes (prototypicality)  17, 88, 93, 140, 167, 171, 222, 282–283 prototypical emotional episodes  8, 12, 30, 47, 92–93, 95, 173, 235, 277 psychiatry  30–31, 72–74, 76–78, 116, 126–127, 199–200, 223, 310, 317, 319 psychological construction  12, 17, 24–26, 28, 30, 64, 77, 79, 81–84, 107, 117, 130, 154, 156, 173, 200, 208, 240, 243, 257–258, 260, 262, 264, 266, 271, 277, 279–280, 284, 290, 294, 296–297, 299–300, 319, 346 Q qualia (quale)  106, 281 R rage  33, 35–36, 38, 47–49, 53, 67, 136, 138, 143–145, 162, 169, 176, 203, 205, 261, 301, 315–316, 319, 324–326, 329, 333, 335–336, 343–345 realism  10–11, 22, 27–28, 30, 206, 213, 216–217, 221, 223, 322 reinforcement  35, 90, 259, 302, 345 ruthless reductionism  18, 185, 309–311

S sadness  8, 26, 37, 69, 94, 100–103, 105, 135, 139, 197, 212, 228, 239, 261, 288 secondary process affect  308 seeking  3, 6, 14, 33–35, 38, 47, 49, 53–54, 67, 69, 76, 80, 84, 111, 136, 138, 143–145, 157, 161, 164, 169, 176, 203, 205, 260–261, 273, 280, 285, 301, 313–314, 316, 318–319, 324–326, 330, 333–336, 341 self-reports  6, 10, 17, 26, 51–54, 67–68, 138, 168, 181–182, 332, 335 self-stimulation  3, 35, 38, 48, 54, 76, 336 separation distress  3, 15, 37, 157, 163, 205, 228, 242 serotonin  43, 46, 62, 64, 66, 74, 78, 121, 305, 334, 336 simulation  98, 200, 297 social bonding  3, 35, 37 social construction  222–223, 280, 297 social roles  21, 215–216, 221–222, 298 somatosensory cortex  40, 49, 91, 112, 297–298, 325 subcortical processes  2, 5, 16–17, 24, 27, 32, 37–38, 40–43, 46, 48, 54, 56, 59–60, 65, 73–76, 87, 108, 113, 120, 122, 126–127, 136, 145, 151, 162, 172, 203, 227, 235, 243, 246–248, 250–254, 256, 264, 270, 281, 284, 302– 303, 305, 307–308, 313–315, 318, 326, 340–341, 343 Substance P  36, 316 supramodal network  66, 241

T temperament  23, 62, 191, 197, 200, 215, 227, 229 tertiary process  32, 39–40, 69–70, 136, 303 thalamus  34, 112, 157, 194, 325 thirst  41, 45–46, 49, 63, 121, 131–132, 230, 240, 323–324 tickling  37, 54 top-down  29, 55, 64, 89, 91, 116, 175, 203, 221, 229, 232, 302–303, 307–308, 313–314, 325 trigger  55, 165, 253, 296, 328, 340, 341–344 U ultrasonic vocalizations  36, 72, 74, 126, 318 unconditioned emotional responses  120, 302, 322 unconscious  16, 23, 38, 47, 57–58, 72, 76–77, 101–103, 115, 118, 137, 172–173, 239, 241, 254–256, 258, 276, 278, 295, 303, 326, 328, 344 universal  7–8. 12, 14, 21–22, 30, 32, 61, 63, 74, 78, 92, 101, 105–106, 108, 122, 133, 197, 200, 210, 218–220, 223, 237, 240, 275–277, 281, 286, 298, 300, 305, 333, 335 V valence  2, 10, 22–23, 26, 46, 57–58, 64, 85, 100, 108, 110, 115, 119–121, 125, 138, 167–170, 225–228, 234, 236, 242, 259, 262, 270, 295, 299, 312, 323– 324, 328–335, 338–339 vasopressin  36, 325 Ventral Tegmental Area (VTA) 34–35, 157, 325