The neurobiology of positive emotions

Abstract

Compared to the study of negative emotions such as fear, the neurobiology of positive emotional processes and the associated positive affect (PA) states has only recently received scientific attention. Biological theories conceptualize PA as being related to (i) signals indicating that bodies are returning to equilibrium among those studying homeostasis, (ii) utility estimation among those favoring neuroeconomic views, and (iii) approach and other instinctual behaviors among those cultivating neuroethological perspectives. Indeed, there are probably several distinct forms of positive affect, but all are closely related to ancient sub-neocortical limbic brain regions we share with other mammals. There is now a convergence of evidence to suggest that various regions of the limbic system, including especially ventral striatal dopamine systems are implemented in an anticipatory (appetitive) positive affective state. Dopamine independent mechanisms utilizing opiate and GABA receptors in the ventral striatum, amygdala and orbital frontal cortex are important in elaborating consummatory PA (i.e. sensory pleasure) states, and various neuropeptides mediate homeostatic satisfactions.

Introduction

Recently, scientific study of positive emotions has been receiving increased experimental attention. Extroversion and gregariousness are among the best predictors of subjective well-being and positive affectivity; along with being happier, people who experience high subjective well-being typically have better health outcomes and longevity (Fredrickson, 2004, Seligman and Csikszentmihalyi, 2000). In this paper, we will review the recent scientific evidence on the neurobiology of positive emotions that has emerged from human and animal research.

An abundance of neuroscience evidence indicates that whereas the cognitive aspects of emotions, such as the recognition of happy and sad faces, require neocortical processing, the experiential states of happiness and sadness, as well as the other basic affective states are strongly dependent on sub-neocortical limbic circuitries that we share with the other mammals (Damasio et al., 2000, Liotti and Panksepp, 2004, Panksepp, 1998). Although there are no unambiguous objective indicators of subjectively experienced affective states that commonly accompany emotional and motivational arousals in humans, because of abundant corroborative evidence (Panksepp, 2005), it is becoming more acceptable to entertain that such affective states of consciousness may arise from sub-neocortical brain dynamics that we share with other animals as well.

Throughout the 20th century, there was resistance to entertaining brain/mind entities such as affective states in animals, but modern neuroscience, with its many well-demonstrated neuroanatomical and neurochemical homologies across all mammalian species, now provides ways to evaluate such possibilities in more scientifically rigorous ways than was possible during earlier eras (Panksepp, 1998, Panksepp, 2005). For instance, it is now clear that the neurochemistries of opiate and psychostimulant addictions are organized similarly in the brains of all mammals, and it has been argued that we cannot make sense of such behavioral changes unless we begin to take affective experiences more seriously in the lives of other animals (Knutson et al., 2002, Panksepp et al., 2002; Panksepp et al., 2005). Although there is still a strong and re-vitalized neo-behaviorist tendency to see all behavioral change in animals as arising from unexperienced, pre-conscious brain processes, our goal for this essay is to discuss why it may be wiser to provisionally conclude that ancient forms of consciousness such as affective experiences do exist in the neurodynamics of other animals, and why such brain functions may be all important for behavioral neuroscientists to consider more openly (Panksepp, 2003a, 2005). Animal behaviorists (e.g. Marc Bekoff, Marian Dawkins, and Don Griffin) are increasingly accepting the likelihood that animals experience their lives, and that such issues are essential for discussing animal welfare issues (see McMillan, 2005 for a recent summary of such work). In our estimation, such views are more consistent with the mass of available evidence, and they provide a straight-forward strategy for shedding empirical light on very important neuro-mental functions, critical for advancing understanding that can inform psychiatric practice (Panksepp, 2004). Such issues are almost impossible to study with any neuroscientific precision in human beings. In this paper, we will largely restrict our coverage to anticipatory eagerness and gustatory pleasure, since those are among the best studied and least controversial positive affective responses that are conserved across most mammalian species.

Although not covered in detail, we would also note that the primitive emotional concepts of positive and negative affect (PA and NA, respectively), which are increasingly central to modern psychological analyses of emotional experiences (Lambie and Marcel, 2002, Russell, 2003), are rather general and non-specific ways to conceptualize emotional feelings. It could be argued that PA and NA are merely semantic-conceptual ways to parse the many kinds of ‘good’ and ‘bad’ feelings that the nervous system can construct, and that a neuroscientific analysis must seek endophenotypes that are neurobiologically ingrained affective processes. For instance, a recent personality test designed to evaluate various basic emotional tendencies in humans provides evidence that NA states can be constituted of specific negative feelings such as fear, anger and sadness, while PA can be constituted of specific positive feelings such as those related to playfulness, nurturance and exploratory-seeking urges (Davis et al., 2003). Whether the more general categories of PA and NA simply reflect convenient ways to talk about such conceptual groupings of desirable and undesirable feelings, or whether they have neurobiological realities above and beyond their class identifier status, is an issue that remains to be resolved. Since so much effort in psychology has been devoted to development of general PA and NA concepts (Davidson et al., 2000, Handbook of Affect Science), we will here utilize that scheme as a guide to talking about more specific affective feelings that require a more resolved taxonomy. Although all investigators of such neuro-mental processes surely recognize that they can only be indirectly monitored through the use of various psychological and behavioral measures, it remains to be widely recognized that affective states cannot be scientifically understood without neural analyses.

Positive affective (PA) states can either be scientifically measured via self-report Likert-type rating scales or by examining unconditioned behavioral responses. Thus, the positive emotion of happiness is typically measured by asking subjects to rate their current level of happiness using either a verbally anchored scale or a pictorial one such as the self-assessment manikin in which subjects identify their emotional state by choosing cartoon characters that match their mood states along valence (bad to good mood), arousal, and power/surgency dimensions (Lang, 1995). For example, emotional states elicited by viewing pictures of human babies are rated high on positive valence and moderately on arousal (Lang, 1995). Alternatively, happiness could be measured behaviorally, for instance by the presence of Duchenne smiling, which has been found to be positively correlated with human subjective self-report of positive emotion (Ekman et al., 1990).

In animals, the utilization of conceptual-psychological scales is not possible, and hence all measures have to rely on behavioral analyses. In general, investigators have the option of focusing on the study of unconditioned-instinctual behavioral tendencies or conditioned-learned behavioral changes. Both are useful, but the former may be more useful for a brain systems analysis of the critical neural components, if one makes the simplifying dual-aspect monism assumption that affective feelings may directly reflect the neurodynamics of brain systems that generate instinctual tendencies (Panksepp, 2005). However, the study of learned behaviors is also critical, for that level of analysis provides the opportunity to validate the likely presence of experiential components as measured by various approach and avoidance behaviors, especially conditioned place preferences and aversions to locations where organisms experienced experimental imposed variation of their internal states (studies that are very hard to conduct ethically and logistically in humans).

A great deal of animal work has been conducted on various negative affective processes (Panksepp, 1998), especially fearful behaviors, even though many leading investigators still do not accept the concept of affect as being of any relevance (or even reality) within their neuro-behavioristic ontologies. They commonly ignore abundant work done on other negative emotions, such as separation-distress, which has relied heavily on vocal measures of emotions such as the analysis of separation calls (Panksepp, 2003b), not to mention the variety of positive emotions such as playfulness and other social emotions that exist in animal brains (Panksepp, 1998, Panksepp, 2005). However, it is becoming increasingly clear that the mapping of the separation distress system in animal brains has striking anatomical correspondences to human sadness systems highlighted by PET imaging (Damasio et al., 2000, Panksepp, 2003b), even to the extent that human sadness is accompanied by low opioid tone in the relevant limbic circuits (Zubieta et al., 2003), a principle first revealed through animal brain research (Panksepp, 1981, 2003b). However, to keep this contribution manageable, our coverage here will be restricted to several key exemplars of positive affect.

Many psychometric models of positive emotions conceptualize positive affective states as originating from approach and various consummatory behaviors. From this perspective, all emotions can be categorized on a two or three-dimensional Cartesian planes, with an approach-avoidance valence dimension and an arousal dimension (Cabanac, 1979, Knutson et al., 2002, Lang, 1995, Russell, 2003; see Fig. 1), and often a third, and less well-conceptualized dimension of surgency or power of a feeling. For example, joy would be categorized as approach+high arousal whereas sensory postprandial pleasure would be approach+low arousal. Approach based psychometric models of positive emotion have also been used to conceptualize stable PA traits states such as subjective well-being (Davidson et al., 2000, Panksepp et al., 2002) as well as the personality trait of extroversion which is associated with high levels of PA (Diener, 1998). However, as already noted, both positive and negative affect could be broken down into a variety of specific emotions, affording a more refined view of the many distinct species of affect that may exist in the brain (Davis et al., 2003; Panksepp, 1998).

From this perspective, what is needed in the study of affect is a more resolved taxonomy that hopefully will eventually map onto distinct brain systems. For instance, it could be argued that positive and negative affect are simply conceptual class identifiers rather than ‘natural kinds’ and it is easy to envision many distinct categories of positive affects—for instance, (i) those that emerge from homeostatic bodily need states which are alleviated by specific sensory–motor consummatory activities, (ii) those that reflect emotional action processes (e.g. play and investigation), and (iii) those that emerge as general background feelings related to various forms of satisfaction, distress, and relief (Ostow, 2004, Panksepp, 2004, Panksepp and Pincus, 2004). So far neuroscience has had relatively little to say about such issues, except for the recognition that many of these feelings are substantially due to activities of the limbic system (MacLean, 1990) and specific sub-systems coursing through those sub-neocortical regions of the brain (Panksepp, 1998). Since most of that work comes from the study of laboratory animals, where discussion of internal subjective states continues to be shunned, it is an understatement to say that the nature of affective experience remains a conceptually and empirically underdeveloped territory. Future developments along these lines will require investigators to be willing to take objective behavioral measures, especially of spontaneous behaviors (e.g. emotional vocalizations) and parallel preference and aversion paradigms, as potential indicators of affective experiences.

An emerging human research tradition does, in fact, attempt to infer PA from unconditioned behaviors that are correlated to self-reported PA states. Perhaps, the classic example is the Duchenne smile, also known as the felt smile (Ekman et al., 1990). Such approaches suggest that from a felt smile one can infer PA due in part to the strong positive correlation between Duchenne smiling and subjective self-report of positive affect in humans (Ekman et al., 1990). However, it is increasingly recognized that such measures may lack validity in many adult human studies, because of various cognitive-instrumental ways humans regulate their affects, including cultural display rules that commonly make such indicators fuzzy signals of emotional feelings. However, such instinctual displays may be more veridical readouts of internal affective states in infants who are not yet skilled in cognitive control of their behavioral displays, since they have not fully assimilated cultural expectations.

In fact, human infants exhibit certain patterns of oral-facial behaviors exclusively to sweet solutions that adults find highly palatable (Rosenstein and Oster, 1998, Ganchrow et al., 1983). Adult primates as well as rodents also exhibit these hedonic taste reactivity reactions exclusively to highly palatable solutions (Berridge, 2000, 2004). Positive emotional vocalizations that are exhibited during anticipation of rewards, as well as during pro-social interactions such as conspecific reunion as well as rough and tumble play, have also been used as a behavioral index of positive affective states in primates (Jürgens, 1976, Jürgens, 1998, Jürgens, 1979) and rodents (Knutson et al., 2002, Panksepp et al., 2002). Many of the conclusions derived from the study of these unconditional behaviors have been validated by corresponding place preference and place aversion paradigms (Burgdorf et al., 2001b).

With regard to the sensory aspects of positive affects, various pleasures reflect the capacity of certain stimuli to return the body to homeostasis. Although there are many historical antecedents to this idea from Plato onward, the modern discussion goes back to the work of Michel Cabanac (1971). For example, a warm stimulus would be experienced as pleasurable by a cold individual, with the magnitude of the pleasure being proportional to the ability of the stimulus to return the body to homeostatic conditions. This has been referred to as sensory alliesthesia (Cabanac, 1992). Positive emotional responses are also thought to reflect the relative utility of eliciting stimulian old idea (e.g. Bindra, 1978, Young, 1966) that fits modern neuroeconomic principles (Shizgal, 1997, Panksepp et al., 2002).

Many studies have found that the rewarding effects of addictive/euphoric drugs are mediated by sub-neocortical systems as indicated via self-injections studies as well as conditioned place preference studies (McBride et al., 1999). Opiate agonists that bind preferentially to μ-opiate receptors are euphorgenic, whereas kappa selective opiates generate negative affect in humans (Schlaepfer et al., 1998). When injected directly into the brain, μ-opiate agonists (morphine, endomorphin 1, DAMGO) produce conditioned place preference consistently when injected into the ventral tegmental area, nucleus accumbens, periaqueductal gray and lateral ventricle (Bals-Kubik et al., 1993, Olmstead and Franklin, 1997, Terashvili et al., 2004), whereas kappa selective agonists (U50,488H) produce conditioned place aversion when injected into many of these same brain regions (Bals-Kubik et al., 1993). Cocaine produces conditioned place preferences and self-administration when injected directly into the accumbens, prefrontal cortex, and olfactory tubercle (Goeders and Smith, 1993, Gong et al., 1996, Ikemoto, 2003), whereas amphetamine place preference seems restricted to the accumbens and ventral palladium (Carr and White, 1986, Gong et al., 1996). Alcohol and nicotine produce place preference and self-administration when injected into the ventral tegmental area (Laviolette and van der Kooy, 2003, Rodd et al., 2004). In sum, most of the drugs that are euphorigenic in humans are rewarding when injected directly into the ventral striatum or its cortical afferents in rats.

A number of peptide systems have been implicated in positive affective states since the peptides have been rewarding when given peripherally or microinjected directly into the rodent brains. Neurotensin and CART, peptides closely associated with brain dopamine, exhibit place preferences when injected into the ventral striatum (Glimcher et al., 1987, Kimmel et al., 2000). Neuropeptide Y, a peptide involved in food intake and alleviating anxiety, yields place preference when injected into accumbens and perifornical hypothalamic nuclei, with its rewarding effects being partially dissociable from its hyperphagic effects (Brown et al., 2000). Oxytocin, which is involved in creating social bonds and the pleasures of social contacts shows place preference when given peripherally (Liberzon et al., 1997), but it has been difficult to see this effect centrally, even though in unpublished work we find socially induced place preference to be facilitated by oxytocin (Panksepp, 1998). These studies implicate a number of sub-neocortical circuits in the generation of affective states.

Before we proceed, let us briefly emphasize why investigators should entertain the likelihood that affect largely has a sub-neocortical locus of control. At present, many investigators and theoreticians remain skeptical about the fundamental role of sub-neocortical systems in the elaboration of affect, partly because they feel consciousness is only a characteristic of the higher heteromodal cortical functions in humans and perhaps several other highly cerebrated species. Modern brain imaging studies which demonstrate results contrary to the evidence long provided by animal brain research, tend to highlight the arousal of many higher cortical regions in the emotional-cognitive processes aroused by exteroceptive stimuli (for summaries, see Lane and Nadel, 2000).

Modern brain imaging, especially with fMRI may be yielding deceptive findings, at least for understanding the nature of affect: Investigators, by using perceptually driven methodologies, are typically visualizing the cognitive components of emotional processing rather than core affective states. fMRI does not appear to have the ideal characteristics to detect true affective changes (which have a time-course which is not well suited for fMRI parameters). PET is much better; indeed Damasio et al.'s (2000) study picked up patterns that are rarely evident in fMRI studies. The ever increasing imaging of unconscious emotional processes is pertinent for understanding the sensory–perceptual aspects of emotional stimuli, not affective states. Indeed, most investigators that use such methodologies fail to evaluate potential affective changes empirically, which may compromise the generality of conclusions concerning the affective dimensions of experience, which may be most relevant for clarifying psychiatrically relevant emotional issues. Thus, although fMRI may be a fine tool for identifying brain cognitive correlates of emotional experiences (generated very rapidly to specific sensory or cognitive contingencies), as currently used, it is a blunt tool for analyzing how affect is generated in the brain. Since emotional affects emerge relatively slowly (e.g. sadness), they may not be as readily linked temporally to precipitating stimulus events as are emotional perceptions.

Fortunately, PET imaging is better suited for visualizing affective responses of the brain, and an increasing number of experiments using PET during the past few years have been more concordant with the animal data then fMRI studies. Perhaps, the most compelling evidence comes from Damasio et al. (2000), who asked individuals to achieve deep, existentially experienced feeling states of anger, fear, sadness and happiness via personal reminiscences. When subjects truly experienced those feelings, radioactive water was infused and PET images were constructed. The results affirmed abundant arousals in sub-neocortical brain regions, accompanied by substantial reductions of blood flow in many higher brain areas, suggesting a narrowing of information processing in neocortical systems during intense emotional states (Liotti and Panksepp, 2004).

Various studies have also highlighted the importance of sub-neocortical regions in human affective experiences such as air hunger (Liotti et al., 2001), the taste of chocolate (Small et al., 2001), the appetite for various rewards including winning money (Knutson et al., 2001a, Knutson et al., 2001b), the sex-specific appeal of pretty faces (Aharon et al., 2001), the pleasure of musical peak experiences (Blood and Zatorre, 2001), male sexual arousal (Redoute et al., 2000) and orgasmic pleasure (Holstege et al., 2003). All of these studies report arousals of various subcortical brain areas implicated in the generation of affect by animal research, as well as those mesocortical zones, especially orbitofrontal, anterior cingulate and insular cortices that MacLean (1990) originally highlighted in his Limbic System concept (which has been increasingly attacked by a growing number of cognitive neuroscientists more accustomed to working on the higher informational functions of the brain). It is claimed that the Limbic System is not a coherent anatomical or functional entity (which can be said for any region of the brain), without realizing that the concept was originally used to designate visceral regions of the brain that are critical for elaborating emotionality, a general conclusion that continues to be supported by modern research.

The extended viscerally focused brain regions known as the limbic system, descending deep into the medial diencephalon and upper brainstem, do appear to comprise the fundamental neuro-geography of spontaneous emotional behavior and affective experience in both humans and other mammals (and probably birds and reptiles also). These systems are regulated and further parsed by higher cortical activities, but aside from the role of mesocortical areas like the insula in the pleasures and displeasures (e.g. disgust and pain) of certain sensations, there is little evidence that higher neocortical regions are essential for generating affective experiences that accompany emotional arousal, even though they are be essential for the cognitive memories associated with those states. Accordingly, we should be devoting more effort to studying the details of basic emotional systems in appropriate animal models (e.g. as summarized in Panksepp, 1998). These systems, which appear to be homologously organized in all mammals, are largely inaccessible for causal human research.

In this vein, we should also recall that emotional feelings have typically been much easier to activate in humans through stimulation of sub-neocortical circuits that mediate the instinctual emotional behaviors in our fellow animals, than through higher brain stimulation (for reviews see Heath, 1996, Panksepp, 1985). A most recent striking example was Bejjani et al.'s (1999) observation of sudden onset of depression by stimulating midline diencephalic structures near the subthalamic nuclei, and mirth by stimulating the nucleus accumbens (Okun et al., 2004). In sum, the evidence is substantial for a sub-neocortical locus of control for the generation of experienced affects that accompany various emotional states and consummatory responses.