There are easily half a dozen different perspectives people bring to the study of emotion: psychological, cognitive, sociological, anthropological, philosophical, and neuroscientific
Given this intellectual diversity and lack of consensus, it is not surprising that there have been many books espousing new theories of emotion. Most of them—certainly most of the theories that make their way into the public conversation—originate in psychology. Such theories are interesting and powerful, but they are often very abstract and hard to falsify by experiment. The perspective of this book is very different. I believe that neuroscience can offer us a way of thinking about emotions that is objective and empirical, and that gives us the tools to begin studying questions about emotion that have been so difficult that some have written them off as unsolvable
Feldman Barrett argued further that anger in humans comes in so many different forms that looking for brain activity that singularly represents this emotion is, effectively, an exercise in futility. From this perspective, efforts to understand anger—or any emotion, for that matter—at the level of neuroscience seem fundamentally flawed, and the nature of emotion itself seems to remain beyond our grasp. What I hope to show you is that such dismissals are too quick. Neuroscience has something real to tell us about how emotions work; we’ve just been going about it in the wrong way. Over the past two decades, that has begun to change with the development of revolutionary new techniques for understanding brain function in so-called model organisms, animals that are bred in the laboratory for research and are amenable to genetic manipulations, like mice and fruit flies. These methods use genes and light to mark, map, measure, and manipulate specific types of neurons in the brain. Unlike brain-scanning experiments, which measure neural activity indirectly, as blood flow to the brain, these new methods can directly measure electrical activity in individual neurons and can map their direct connections to particular cells in other brain regions. These techniques allow specific sets of neurons to be turned on or off at will, to determine how that affects specific behaviors. Unlike brain-scanning experiments, such experiments can distinguish cause from effect. Fittingly, I refer to such experiments as causal neuroscience.
The sad fact is that there hasn’t been a fundamentally new psychiatric drug approved in the last 50 years. All the “new” drugs being released are just variants on the same basic theme—for example, selective serotonin reuptake inhibitors (SSRIs) like Prozac, Paxil, and Lexapro. The reason is that most of the drugs that we do have—like SSRIs—were discovered by chance. Such lucky accidents don’t happen very often, and many people suffer while we are waiting for the next one to occur. We need to be able to have a way to develop new psychiatric therapies by design, through an understanding of underlying disease mechanisms.
if we want to use these new methods to achieve a causal understanding of how emotions like fear and anger are generated by the brain and linked to behaviors like aggression—one that will impact human health—then we need to work on animal models
But now we arrive at a significant problem. How do you measure emotions in animals? Most people use the word “emotion” in everyday speech to refer to “feelings.” Feelings are subjective experiences that we become consciously aware of through introspection. Scientifically, the only way to assess subjective feelings is by verbal report: the researcher asks the subject how they feel, and the subject describes their feelings. Since animals can’t talk, we have no way of knowing what they are feeling—or indeed, if they are feeling anything at all (in the sense that we experience feelings). Subjective feelings are a manifestation of conscious awareness, and there is currently no way to objectively determine whether a non-human animal is conscious. Therefore, if we consider emotions exclusively as “feelings,” we cannot know whether they are an attribute of animals
Darwin’s assumption was fine for his purpose, which was to explain the evolutionary benefit of specific “emotional” behaviors that people and some animals share, like why our eyes widen when we are afraid.i However, for a hard-nosed neuroscientist like me, this assumption is problematic for several reasons. First, if we define emotions as feelings, then, as Tinbergen said, we can’t objectively know if an animal has any emotions at all. Second, if we simply assume, like Darwin, that all animals have emotions, then in order to infer what kind of emotion a given creature is feeling, we have to attribute to that animal the same emotions we would feel under similar conditions. But animals are not little people in furry costumes, and so our intuition may mislead us. For example, if I see my cat roll on her back with her paws in the air when I come home from work, I infer that she is happy to see me—because I would be happy to see me if I were the cat locked in the house alone all day. However, I have no independent, objective way of knowing how or what my cat is feeling, other than by observing her behavior. I can’t both explain her behavior by assuming I know her feelings and decide what she is feeling by observing her behavior—that’s circular logic. Maybe she has simply learned that she can train me to rub her belly by rolling on her back with her paws in the air.
if neuroscience is able to offer a better way of thinking about emotions than we currently have, as I believe it is, then we now know what it must do. First, it should operationally redefine “emotion” in a way that does not require attributing feelings to animals. Second, it needs to be able to distinguish whether a given animal’s behavior expresses any emotion at all or is just an automatic reflex. Third, it must offer a way to determine what kind of emotion the animal is having, without falling back on attributing our own subjective human experiences to it. Finally, it must be able to show us that learning about how emotions work in animal brains can tell us something about how they work in our own brains.
The key intellectual adjustment involved letting go of the idea that emotions consist exclusively of subjective feelings. Rather, they are internal, central states of the brain that can exist independently of whether the owner of that brain has any conscious awareness of them or not. We know anecdotally that even we humans can sometimes have emotions that we are not consciously aware of but which our friends and loved ones can infer from our body language or facial expression. And there are laboratory studies that have provided evidence of unconscious emotions in humans as well. If emotions can exist independently of consciousness in people, then they may exist in animals as well, whether or not we consider those species to be “conscious.” This doesn’t mean that I think that cats, dogs, and other animals don’t have subjective feelings; it just means that I don’t need to answer that question scientifically in order to study how their brains generate emotion states.
You may have a subjective feeling of being starved or sated, but the feeling is not what controls your behavior. Your brain state controls your behavior, and your subjective feelings are your conscious experience of your brain at work—the brain’s perception of its own internal state
In other words, emotions are internal states that control how the brain’s input is converted into its output, like a supervisor directing workers how to connect calls at an old-fashioned telephone switchboard. Externally visible behavior is one such output, or “readout,” of the internal emotion state. But there are other measurable readouts that can occur internally, such as changes in heart rate, blood pressure, or hormone levels. Subjective feelings—our conscious awareness of these internal states—are just another such readout. However, they are neither the only one nor the essential one. Therefore, by considering emotions as functional internal states, we can study how the brain controls emotions in animal models without having to assume or figure out whether animals do or don’t have subjective feelings.
We refer to these meta-behavioral properties as emotion primitives. These primitives include properties such as persistence (emotional behaviors tend to outlast their inciting stimulus, whereas reflexive behaviors terminate when their stimulus disappears), scalability (emotional behaviors can increase or decrease in intensity as the intensity of the underlying state changes, whereas reflexes are typically all or nothing), and generalization (the same emotion state can be triggered by different stimuli and, once evoked, can alter an animal’s responses to other stimuli). We have proposed additional such emotion primitives, and these will be described in Chapter 2. Importantly, we postulate that emotion primitives are properties both of internal brain states and of the externally observable behaviors that express those states. Therefore, if an animal responds to a particular stimulus in a way that exhibits these meta-behaviors, its behavior may reflect an underlying internal state rather than simply a “mindless” reflex.
You can also think of emotion primitives as evolutionary building blocks of emotion, in much the way that the internal combustion engine, transmission, and carburetor are building blocks of a car. Just as those devices had to be invented before they could be combined to create an automobile, emotion primitives could have appeared in evolution before full-fledged emotions evolved
IN THIS BOOK, I will describe our efforts to apply the concept of emotion primitives by using it to investigate defensive and aggressive internal states in both mice and fruit flies. We did this to determine whether this approach would generalize across different emotion states in the same animal (e.g., “fear” versus “anger” in mice) and whether it would generalize across similar states in different animals (e.g., “fear” or “anger” in mice versus those states in flies).
What was going on in the tiny brains of these little animals, each fly smaller than a grain of rice? The amateur observer in me wanted to believe—as Darwin did—that I was witnessing an example of fear in flies: that the shadow repetitively moving overhead had evoked a mounting state of anxiety, which gradually trumped hunger and induced the flies to stop feeding, and eventually to flee in terror from what they perhaps perceived as a predatory bird circling overhead, waiting to strike
A word here on the difference between a reflex and an emotion state. By “reflex,” I mean a genetically specified neural pathway that automatically connects a particular sensory stimulus to a particular motor response—something like how when you’re
operating a marionette, jiggling a wire (the stimulus) causes a particular limb of the marionette to move in a particular manner. Activating that pathway produces that response, and only that response
In contrast to such hardwired, stimulus-response reflex systems, behaviors controlled by internal emotion states involve what psychologist Kent Berridge has called an intervening variable, something in the brain that lies at the interface between the sensory input and the motor output
These intervening variables are the product of information gathered from multiple input pathways and combined to produce an internal “driving force” that can vary in intensity.1 The intensity of that driving force, together with other influences, determines which behavior is chosen from among a number of possible options. Internal states offer more flexibility than stimulus-response reflexes: they can respond to any of several different stimuli (or “fan-in,” to use a term originally derived from electrical engineering), and they can produce any of several different outputs (“fan-out”)
In the case of flies and their escape behavior, a large amount of research has looked inside the brain to identify a reflexive neural pathway controlling jump responses to
looming visual stimuli: it involves a (relatively) huge neuron, called the “giant fiber,” that basically connects the fly’s eyes almost directly to its legs and can trigger a jump in response to a shadow within a few milliseconds
Come to think of it, is there any argument against explaining the behavior of birds, gazelles, and cats by the same type of non-emotional reflex mechanism as well? This conundrum gets to the heart of the problem of studying emotions in animals: How can you know, objectively and scientifically, whether the animal is showing an “emotional” response to a certain stimulus or situation, or just a hardwired stimulus-response reflex? In fact, most attributions of emotions to animals have been based on anthropomorphic inference: if the animal is doing what I would do in a given situation, and if I would feel a certain way in that situation, then the animal must be feeling that way, too.
While that may be a perfectly good conclusion to draw when thinking about how to keep our beloved pets happy, it’s not science. If we really want to know whether animals have emotions or are just very fancy robots with hardwired reflexes that “look” emotional, we need to make every effort to rule out alternative interpretations. In science, it’s not sufficient to just choose the interpretation you prefer and ignore the inconvenient or less palatable alternatives. You have to show, by experiment, that the alternatives are not valid explanations. As my late mentor and colleague Seymour Benzer once said to me, “The greatest sin in science is wishful thinking.”
Why does it matter, anyway, whether fly behavior incorporates fear, or any emotions at all? It matters because it determines whether researchers can use fruit flies as a subject to study how animal brains generate emotion states. For well over a century, fruit flies have been a workhorse model organism for figuring out fundamental mechanisms in biology, from mapping the position at which different genes (units of heredity) are located along a chromosome (a cellular structure that contains a string of different genes) to decoding the genetic blueprint for embryonic development. That’s because they’re small, easy and cheap to grow, have large numbers of progeny and a two-week generation time, and have only 4 pairs of chromosomes (compared to 20 in mice and 23 in humans). Yet they obey the same principles of heredity, development, and evolution that apply to us. Because they are compact and easy to work with, flies tend to yield solutions to basic biological problems faster than we can figure them out in furry animals like mice or rats, let alone people. The fruit fly genome was sequenced years before the human genome, and most fly genes have counterparts (called homologs) in humans. Many of those genes are relevant to human diseases. For that reason, flies have been an important biological stepping-stone to finding medically important genes in humans
Howard Hughes Medical Institute investigator Gerald M. Rubin headed a team that generated a fly “connectome”—an electron-microscopic map of the synaptic connections of every single neuron in the fly brain. The only other organism for which a complete connectome exists is the nematode worm C. elegans, but that creature has only 302 neurons, compared to about 100,000 in a fruit fly, and its behavioral repertoire is more primitive than that of flies
At one level, trying to figure out how the brain—any brain—works without such a map is like trying to understand how New York City public transportation works without a map of the subway system and bus lines. It’s going to be many years before we have a complete connectome for the mouse brain, and probably decades before we have one for humans.
To use an architectural analogy, the Taj Mahal and the cathedral at Chartres both contain arches although the buildings’ styles are
totally different. In the same way, studying fly brains can reveal basic principles of neuronal circuit architecture used by many organisms.
Two observations that seemed inconsistent with the reflex hypothesis initially caught my eye. The first was that not all the flies responded to the shadow the first time they were exposed to it
a reflexive response should, by definition, be the same every time.
I realized that this feature, which I call scalability, may be a general feature that can be applied to other emotions as well
I also noticed that the flies kept moving around the perimeter even after I stopped the auto-swatter. In other words, their reaction persisted long after the termination of the stimulus that triggered it
Like scalability, persistence, too, is a general feature of many emotional responses. Once an emotion is “excited,” it can take a while to dissipate. For example, fear reactions often outlast the stimulus that triggered them
Over time, my colleague Ralph Adolphs and I began to piece together more of these basic attributes of emotion states from our observations and the work of others. For example, emotions have an intrinsic valence: they can be positive or negative
Another important feature of emotions is that they generalize: the same emotion can be evoked by different stimuli, and a given emotion can affect behavior in different contexts
the concept of emotion primitives is not a new theory of emotion. It does not try to say what emotions are or aren’t, or to hypothesize how they are produced by the brain or the mind.3 Instead, it is intended to be a different, and useful, way of thinking about emotions as functions carried out by the brain, and it allows researchers to study them experimentally in various types of model organisms, using sophisticated neuroscience methods that cannot be applied in humans. It is useful because it should help us to explain behavior by understanding the mechanisms the brain uses to generate the internal state that drives the behavior. If our understanding is correct, we should be able to predict what will happen to an animal’s behavior if we causally intervene in its brain and change its chemistry or its circuitry.
This discovery told us that dopamine is required for agitated flies to calm down. We also found that administering extra dopamine to the flies made them calm down faster.
Interestingly, children with ADHD are often treated with a drug called Ritalin (methylphenidate), which increases dopamine levels in the brain. Moreover, like some children with ADHD, our flies with the mutation also had learning disabilities, as shown by a simple memory test. It is commonly believed that children with ADHD have learning disabilities because they are easily agitated and cannot sit still long enough to concentrate. Surprisingly, however, we found that the learning deficit in our mutant flies was not caused by their hyperactivity; rather, these two behavioral abnormalities were due to separate functions of the dopamine receptor in two different brain regions
Thus, the learning disabilities in ADHD may be causally unrelated to the hyperactivity. These findings may in the future suggest better ways to treat the different symptoms of ADHD, and they illustrate how studies of emotion primitives in fruit flies can uncover brain mechanisms that are shared between flies and humans
Behaviors like aggression and freezing are external observables, and easily measured. In contrast, anger and fear are internal emotion states that are much harder to measure directly.
Why not just make the neuroscience of emotion an off-limits topic in animals, and study it in humans using brain-scanning techniques? After all, with humans, we can put them in an MRI scanner, watch their brains light up, and ask them if they feel afraid or angry. Isn’t that enough to understand how the brain controls emotions? Unfortunately, that’s not enough—at least not if we want ultimately to develop new medicines for psychiatric disorders that are based on a mechanistic understanding of
brain function and dysfunction, in the way that insulin was developed as a treatment for diabetes through an understanding of what the pancreas does. While brain-scanning studies are certainly useful—they can give us a global picture of activity throughout the brain—they are fundamentally correlational, and correlation does not prove causation. Without causation, it is difficult to understand what the mechanism might be—think about trying to understand how an automobile engine works simply by watching it while a car is running, and observing that it cycles more rapidly when the auto accelerates.
The only way to prove that activity in some region of the brain is the cause of a given behavior is to physically access that region, switch the activity on or off, and see whether the behavior is also switched on and off. And those kinds of experiments can, for better or for worse, be done only in animals, not in humans—for both technical and ethical reasons
emotions can be thought of as internal states of the brain.1 Subjective feelings in humans are not the essence of such states but, rather, one of their “readouts,” or expressions; emotion states are also expressed by externally observable things, such as behavior (Figure 2-1B), as well as by bodily responses such as heart rate, pupillary dilation, and increases in stress hormone levels. In this view, an animal does not have to have any subjective experience (“feeling”) of its internal emotion state in order to be in such a state. Indeed, there is evidence that under some conditions, even humans can be unaware of their emotion state.
According to this view, “feelings” arise from our brain’s perception of its own internal state, of which we have a conscious awareness
At some level, this disagreement is about the meaning of the word “emotion” when it is used in a neuroscientific context. Adolphs and I use the term “emotion” to refer to a type of internal brain state while being clear that this usage does not imply that the state is necessarily accompanied by subjective feelings (although it could be)
something different from its colloquial sense in the context of their research, it will confuse the general public. “If you tell people that flies have ‘emotions,’” he has said to me, “they will think you mean that flies have feelings”—which is not what I mean. I would counter that if you tell people that one cannot use the word “emotion” when talking about animal behavior and brain mechanisms, people may think you mean that these animals are just little pre-programmed robots—which is not what LeDoux means
Emotions and feelings are two separate phenomena, and two separate but related problems
If the neuroscience community had collectively decided in the early days of research that “vision” should be defined exclusively as the subjective experience of visual percepts—such as the feeling of seeing redness—then we would only ever have been able to study “vision” in humans. In that case, most of these important advances in our understanding of vision (how we see things) would never have occurred.
By analogy, I would argue that emotion is to feelings as vision is to the subjective experience of seeing red. The field of vision research has shown us that we can learn a great deal about how the brain works if we set aside the problem of conscious awareness and study the nuts and bolts of a neural process in animal models. And the nuts and bolts of how the brain processes visual stimuli are still incompletely understood, even after 50 years of research by many investigators. By comparison, the study of emotion is about 50 years behind the study of vision. The field has a lot of catching up to do in comparison to other areas of neuroscience research. But we can and will move forward if we aim to understand emotions as internal states of the brain rather than as purely subjective experiences requiring conscious awareness.