A Neuroscientist’s Perspective on Debates about the Nature of Emotion

One System for All Emotions: The Limbic System Concept

In the middle of the 20^th century, neuroscientist Paul MacLean attempted to provide an all-encompassing theory of the emotional brain (MacLean, 1949, 1952, 1970). Drawing on the anatomical wisdom of the day, MacLean situated emotion in areas of the medial (presumed older) cortex (e.g., hippocampus, cingulate gyrus) and cognition in the lateral (presumed newer) cortex. The medial cortex and related subcortical forebrain areas were said to form the paleomammalian brain, which contrasted with the later emerging cortex, which MacLean called the neomammalian brain (also known as the neocortex). Both the limbic system and neocortex were viewed as novel forebrain structures present in mammals but not in other vertebrates, and that were layered on top of the reptilian brain (the basal ganglia and brainstem). According to MacLean, specific emotions (specific feelings like fear, love, and anger) are products of the limbic system, whereas cognitive functions (thought and language) are functions of the neocortex.

MacLean’s ideas of old and new cortex were based on the anatomical wisdom of the day. But this anatomical foundation of the limbic system theory has been discredited by modern comparative neurobiology, which has shown that the two regions that MacLean argued were novel mammalian structures both have precursors in reptiles and birds (Butler & Hodos, 2005; Nauta & Karten, 1970; Northcutt & Kaas, 1995; Swanson, 1983). Further, MacLean’s notion that the architecture of paleomammalian (limbic) cortical areas is ill-suited to cognition, and is instead ideal for irrational emotions, is also unfounded. The hippocampus, which MacLean thought of as the centerpiece of the limbic system, emerged as being essential for important cognitive functions, such as declarative memory (Eichenbaum, 2004; Squire & Kandel, 1999) and spatial mapping of the environment (O’Keefe & Nadel, 1978). Finally, there is little evidence that areas identified as components of the limbic system function as a unified emotion system (see Kotter & Meyer, 1992; LeDoux, 1987, 1996).

Still, the idea that there might be circuits in the mammalian brain that are relatively specialized for functions typically referred to as “emotional” and are conserved to some extent throughout mammals, including humans, continues to be relevant. The question is, how should we account for these functions?

Different Systems for Different Emotions: Basic Emotions in the Brain

Basic emotions theory argues that certain emotions are hard-wired in the brain, are expressed the same and recognized universally around the world, are part of human nature, and are conserved to some extent across mammalian species (Damasio, 1994, 1999; Ekman, 1999; Izard, 2007; Panksepp, 1998, 2005; Tomkins, 1962). Although the emotions listed by different authors as basic emotions differ somewhat (Ortony & Turner, 1990), the fundamental idea is that human feelings, defined by introspective experience and labeled with common language emotion words (usually English words), reflect evolutionarily conserved circuits (affect programs).

Some neuroscientists have attempted to identify brain areas and circuits that account for basic emotions. Panksepp has the best developed brain-based basic emotions theory (e.g., Panksepp, 1998, 2005). He argues that the way to capture innate basic emotions in their pure form is when they are aroused artificially by direct stimulation of brain areas where those operating systems are most concentrated. On the basis of such results Panksepp proposed separate circuits involved in seven basic emotions: seeking, fear, rage, lust, care, panic, and play. Each circuit controls specific emotional behaviors (e.g., the fear circuit controls defensive behavior, the lust circuit sexual behavior) and is also responsible for specific feelings (e.g., fear, lust, anger). The relation of innate conserved circuits to feelings is particularly important for Panksepp, who notes that affective consciousness is an “intrinsic function of the brain, shared homologously by all mammalian species” (Panksepp, 2005, p. 30). Although his basic emotions are not the same as those in most psychological theories, as noted before, psychological theories do not have all the same categories either.

Thus, for Panksepp, the existence of conserved circuits for emotional behavior means that feelings too are conserved across mammals. However, evidence that these circuits are responsible for the behaviors proposed relies on results from electrical brain stimulation, a method that is problematic since it indiscriminately activates both neurons and axons in an area and, among other things, cannot be used to precisely identify the areas responsible for the behaviors that result. But this is not the main problem. Regardless of the validity of the conclusions for the relation of the circuits proposed to emotional behavior (some of which may be valid and some not), there is no direct evidence that these circuits contribute to conscious feelings in animals (how could there be?). As indirect evidence, Panksepp cites findings from studies in which electrical stimulation has been delivered to brain areas in humans (supposedly the same brain areas that are responsible for basic emotional behavior in animals). However, such methods are also highly problematic when applied to the human brain (even more problematic than in animals since they are necessarily cruder in human studies, and are only used as part of diagnosis/treatment in patients with pathological brains). For a review of the problems involved in using electrical stimulation findings to reveal circuits dedicated to specific basic emotions and their corresponding feelings in the human brain, see Barrett et al. (2007).

At the same time, we also need to be careful not to go too far and reject the idea that there are specific circuits of emotional behavior. For example, Barrett argues that the lack of emotion specificity in fMRI findings showing activations to emotional stimuli (usually faces) challenges the idea that there is emotion specificity in the brain (Barrett, 2006). There are two responses to this critique. First, it is not at all clear that emotional face stimuli are the most appropriate stimuli for studying highly conserved neural circuits. Second, behavioral functions are localized not at the level of brain areas, which is the level of resolution fMRI can reveal, but at the level of microcirucits within brain areas.

Thus, while I disagree with Panksepp about some (though not all) of the details of the specific circuits involved in emotional behavior (on methodological grounds), and especially disagree about the relation of circuits in rodents to human feelings (on methodological and philosophical grounds), I share with Panksepp the conviction that there are innate emotion circuits that are conserved in mammals, including humans.

One Emotion at a Time

Most neuroscientists today who explore brain functions related to emotion in animals tend to be more interested in circuits that contribute to individual kinds of emotional states, as measured within specific kinds of well-controlled behavioral tasks, than in simultaneously accounting for many emotions (i.e., all basic emotions, however defined). For example, some researchers focus on fear, others on aggression, others on reward, and so on. For each such behavioral category, progress is being made in exploring the contribution not so much of gross brain areas as of systems of microcircuits, each consisting of specific neurons interconnected by specific synapses and with communication within and between such neurons identified in terms of specific molecules regulated by genetic and epigenetic factors.

When I wrote The Emotional Brain (LeDoux, 1996), I emphasized work on fear. I did this primarily because there was and still is no clear definition of emotion, as the present target articles demonstrate. I therefore decided to focus on one emotion that there was good agreement about (most theories accept fear as an emotion), and that I happened to have researched quite a bit. The basic idea was that such a focus would allow progress on a well-defined aspect of emotion, even if it failed to solve the problem of “emotion” in general.

But what do I mean when I say I study “fear” in animals. There are two important things to note. First, I am not studying “fear” in the general sense. I am studying those aspects of fear that can be modeled by Pavlovian defense (fear) conditioning, a procedure in which an emotionally neutral stimulus comes to elicit defense responses (freezing and supporting physiological changes mediated by the autonomic nervous system and the endocrine system) (e.g., Johansen, Cain, Ostroff, & LeDoux, 2011; LeDoux, 2000, 2008; Maren, 2005). This may seem to be a very narrow focus, but in fact it is a pretty good model of how specific stimuli, through experience, acquire threat value and come to elicit defensive responses. It also seems to be relevant to some psychiatric conditions such as specific phobic and post-traumatic stress disorder (PTSD) (Bouton, Mineka, & Barlow, 2001; Grillon, Southwick, & Charney, 1996; Shalev, Ragel-Fuchs, & Pitman, 1992). Second, I am not studying “fear” in the sense of feelings or affective consciousness. I am studying how the brain detects and responds to threats (LeDoux, 2008).

Fear conditioning is a good example of a task that has been used to explore one aspect of one emotion with the exquisite tools available in modern neuroscience. Some of the findings will be described in what follows.

Relevance to the Target Articles

Dixon’s (2012) interesting article concludes that the term “emotion” does not refer to something that constitutes a natural kind. Challenges to the natural kind status of emotion have come up many times, and take various shapes (e.g., Barrett, 2006; Izard, 2007; Scarantino, 2012). The version Dixon prefers is that emotion should be divided into categories, such as basic and higher cognitive emotions (Griffiths, 1997, 2004), that might be natural kinds.

Scarantino (2012) also is skeptical of the idea that “emotion” is a natural kind, but goes further than Dixon in arguing that even the category of basic emotions is not a natural kind since the specific emotions that constitute basic emotions are themselves not natural kinds. But he does not go as far as Barrett (2006) in rejecting the entire notion of basic emotions. Instead, he feels that some emotions are basic and do form a natural kind set, but that a new conceptualization of what basic means is needed.

Mulligan and Scherer (2012) argue that emotions are, at a minimum, psychological kinds. They give several criteria that identify components that occur in and define what emotions are psychologically. A key component is appraisal. The way situations are appraised allows the coordinated activity of the various components of emotion to operate in a way that produces specific emotions. Appraisal and coordinated activity are key concepts in their proposal.

How does all this relate to the brain models discussed before? Any approach that accepts the validity of the basic emotions concept, namely the idea that basic emotions form a limited and natural set of emotions, is most compatible with a basic emotions approach to brain mechanisms, where each basic emotion has its own neural representation. The one-size-fits-all approach of the limbic system is not going to work in a basic emotions theory since it assumes one system for all emotions. The emotions-one-at-a-time approach also won’t work very well as an accompaniment to a psychological approach that is centered on basic emotions since it does not compare and contrast across emotion systems to define the set of emotions that are basic. So both the Dixon (2012) and Scarantino (2012) pieces would seem to be most compatible with a basic emotions approach to the brain.

On the other hand, something like the limbic system theory could be compatible with the Mulligan and Scherer (2012) idea of coordination across components of a broad system. However, given my critique of the limbic system concept before, I think it would be best to look for other, more grounded mechanisms. The basic emotions approach might work since, in their model, emotions often start with an appraisal, and basic emotions systems can be thought of as having an appraisal mechanism at their front end.

But further thought suggests that the basic emotions approach might not be quite right for either the Scarantino (2012) or the Mulligan and Scherer (2012) model. Let’s look at why in some detail by making things very concrete. This will be done by considering fear and the brain in some detail.

Fear (Defense) and the Brain

As noted above, much of what has been learned about the fear or defense system comes from studies of Pavlovian defense (fear) conditioning. Once a conditioned threat is created as a result of a neutral stimulus being associated with danger, the conditioned stimulus will elicit defense responses via a well-established set of pathways in the brain that have been implicated in fear by the consistency and convergence of conclusions across a variety of methodologies (electrolytic and neurotoxic lesions; electrophysiological recordings; anatomical tracing; pharmacological, molecular, and genetic manipulations; functional imaging) (see Johansen et al., 2011; LeDoux, 2000; Maren, 2005; Phelps, 2006; Rodrigues, Schafe, & LeDoux, 2004).

We often hear that the amygdala is involved in fear or other emotions. But this is way too crude a conclusion. Very restricted components of the amygdala (subpopulations of neurons in subareas of amygdala areas) are involved. Only by applying the techniques of modern neuroscience can such conclusions be drawn.

Information about the learned (conditioned) threat is transmitted from sensory processing systems to the lateral amygdala, where the stimulus is assessed (appraised). Through the conditioning process, the stimulus has acquired the ability to activate lateral amygdala neurons in such a way as to initiate a cascade of subsequent processing in the amygdala and beyond. Specifically, by way of connections (direct and indirect) from the lateral nucleus to the central nucleus, and from there to specific downstream areas that control different response modalities, behavioral, autonomic, and endocrine responses are triggered in the effort to cope with the danger. Particularly impressive has been the degree to which modern neuroscience methods have been able to implicate specific neurons within subareas of amygdala nuclei (i.e., subareas within the lateral and central nuclei) and the molecular mechanisms that mediate communication within and between these neurons (e.g., fear learning depends on synaptic plasticity in neurons in the superior part of the dorsal subnucleus of the lateral nucleus, and the expression of fear behavior involves specific neurons within the medial division of the central nucleus).

In addition to connections to output areas, the central nucleus has connections to brainstem arousal systems. When these are activated, neuromodulators (norepinephrine, dopamine, acetylcholine, serotonin, etc.) are released widely in the brain, including in cognitive processing regions in the cortex. The areas affected are put on alert (are aroused in a nonspecific way). That is, their cells are primed for information processing. Various amygdala areas also project directly to cortical areas and can influence processing this way as well. A global state of readiness results in which attention, perception, memory retrieval, and memory formation are all facilitated.

The overall effects are thus twofold. Defensive behavior is elicited and the brain is globally activated. Because this global activity is occurring simultaneously in widespread areas involved in diverse functions, you could say that activity is coordinated in the brain, as Scherer has argued (Scherer, 1984), and could even say that brain activity is monopolized by emotional arousal, as I have argued (LeDoux, 2002).

The term motive state has sometimes been used to refer to global brain states associated with emotion and motivation (Bindra, 1969; Hebb, 1949; Morgan, 1943). Emotive state might be a better term. Such states likely underlie (or at least make significant contributions to) what we call feelings or subjective emotional experiences in humans. But a motive or emotive state is a physical state of the brain that, unlike feelings, can potentially be measured in humans and animals alike.

One way of interpreting the foregoing is that fear is a basic emotion mediated by a specific circuit built into the brain and thus qualifies as a natural kind. But I wouldn’t put it that way. I would instead say that there are phenomena that are talked about under the topic of fear that are products of specific circuits in the brain, but that these don’t qualify as a single unified state of fear in the usual sense of the term. Specifically, there are a variety of phenomena that fall under the rubric fear that have different neural requirements. For example, defense responses elicited by conspecific or predator odors, which are innate fear stimuli, require different circuits (Choi et al., 2005; Markham, Blanchard, Canteras, Cuyno, & Blanchard, 2004; Motta et al., 2009) than those just described for learned cues. States where harm is possible but unpredictable (more like anxiety than fear) involve still different circuits (Davis, Walker, & Lee, 1997). Completely different circuits are likely to be required to fear falling in love or to fear fear itself (the latter two states are more like the cognitive emotions that Dixon [2012] discussed). Ekman’s notion of emotion families (Ekman, 1999) might work as a psychological grouping of these diverse phenomena, but it is unlikely that the various phenomena have much in the way of a unified biological basis.

Conclusion

Because of the heterogeneity of such a seemingly simple and “basic” emotion as fear at the neural level, I doubt that a basic emotions approach to the brain will work. In fact, I believe the only approach that works is a one-emotion-at-a-time approach, with the term emotion not referring to common language notions like fear, anger, or joy, but to states that contribute to survival (defense, reproduction, energy homeostasis, etc.) (LeDoux, 2012). The states should be studied in terms of specific kinds of experimental tasks designed for the purpose of identifying specific circuits and microcircuits, much in the way that learned and unlearned fear has been studied. This may be unsatisfying from the point of view of an immediate, all-encompassing theory of emotion, but in the long run might provide a basis for a viable natural kinds approach to emotional phenomena (if not emotion itself) centered on brain circuits.