What, if anything, makes mood fundamentally different from memory?

Whether mood disorders (and mental disorders generally) are “brain disorders” is an interesting philosophical question. Unsurprisingly, the answer depends on how we understand the notion of a brain disorder. It is important to recognize that this question is motivated by daunting epistemic challenges in brain-behavior research. Neuroscientists have been fortunate that classic memory disorders have turned out to be cases of cellular neurology gone awry, with downstream effects on brain circuits and cognitive domains. Mood disorders, in contrast, appear largely to involve brain-body-environment interactions locked into dysfunctional patterns. Historically, neuroscience has lacked the tools to meaningfully articulate such higher-order dynamic interactions, and its current abilities to do so are fairly rudimentary; psychological concepts describe these interactions, albeit in a manner that is often idiosyncratic and socially constructed. Concepts such as narcissism, shame, projective identification—so familiar to psychodynamic clinicians—resist easy articulation in neurological terms. Maybe one day neuroscience will develop the language to adequately describe and explain these higher-order phenomena, but it is not there yet. Neuroscience requires periodic reminders that its current methods are too crude, that it should not be so arrogant as to think it can make psychology and the social sciences redundant. Mood disorders are multi-level phenomena, not categorizable simply as problems of “mental software,” nor as problems of “neurological hardware.” Their scientific understanding requires the “piecemeal integration” of multiple scientific disciplines and a great deal of epistemic humility.

Investigating the nuances of mental health disorders, particularly in mood disorder research, holds immense promise for both basic science and clinical applications. Moving away from rigid diagnostic categories toward a spectrum-based approach can revolutionize our understanding and treatment of these conditions. This shift allows for a more precise diagnosis and personalized treatment plan tailored to individual needs, recognizing that not everyone responds uniformly to the same interventions. By closely tracking behavioral patterns alongside neural activity over time, we can uncover subtle yet significant changes at a granular level. This is the scientific basis of my research program. This comprehensive approach will enable us to fine-tune treatments and potentially offer preemptive measures for people predisposed to mood disorders. Also, examining the dynamic interaction between mood fluctuations and cognitive function, such as memory formation, is essential. Those with major depressive disorder commonly experience cognitive decline, making it imperative to elucidate the relationship between depressed mood and memory consolidation. By doing so, we can gain valuable insights into the underlying mechanisms driving these conditions, thereby leading to more effective interventions and improved patient outcomes.

For me, inroads for mood research in the mental health sciences will come from better understanding the systems these states play a role in. A person’s mood states—happy, angry, anxious—are experiences that serve as both causes and effects. These states influence each other, but they also influence a person’s thoughts and behaviors, and the environment. What I experience when I receive a bad grade in school or kiss someone influences—and is influenced by—my mood states. The systems theory of mental disorders posits that these networks of mood states and related features can form healthy attractor states: Reciprocal interactions and feedback loops keep many of us in a healthy state. An episode of major depression, then, is an alternative stable state in this system. If we take this perspective seriously, the next steps are to map out these systems, by combining dynamic data, collected via smartphones, for example, with appropriate statistical models to study how these systems differ across people, how stable they are over time, and if systems change during transitions. We carry out some of this work in the WARN-D lab.

The most certain path to better treatments for mood disorders depends on insight into their mechanisms—and thus into brain biology. Whether the onset of a depressive episode follows an adverse experience such as loss of a job or seems to come from nowhere, it is grounded in neural mechanisms, as are the ultimate treatment targets, regardless of how treatment is delivered—via psychotherapy, medicine or electrical stimulation. Such assertions do not represent reckless reductionism. Brains support the fitness of all free-living animals, including ourselves, because they are remarkably powerful integrators. To select actions and regulate our physiology in response to threats, rewards or complex social interactions, the brain must synthesize information that comes from each person’s DNA, prior developmental events and their bodies, with current sensory inputs that refine its predictions of what is happening in the present and what will happen in the future. To discover better treatment, we must understand how such mechanisms go awry. Some who accept the brain as the substrate of thought, emotion and behavior—nature’s lesion “experiments” are quite convincing—but nonetheless devalue the brain’s importance, often hold an erroneous model that analogizes the brain to a digital computer that could run any software. They construe the neural hardware as a dumb machine programmed by the important stuff, thought and experience. But the brain is nothing like a digital computer. There is no software independent of biology. Even as neural circuits are computing the current moment and predicting the next, the brain’s structure, including molecules, dendrites and synapses, is changed forever. It is the biology and plasticity of the brain itself that matter for all experience and action and all psychiatric illness. There is nothing left over.

The ability of mental health professionals to deliver treatments that effectively relieve mental anguish pales next to the success that other areas of medicine have achieved. I think part of the problem is that the medical model may not be as applicable to mental disorders, or at least not to the mental part of such disorders. Medicinal treatments are especially useful in altering behavioral and physiological symptoms. But mental anguish is subjective, conscious suffering. That does not mean that conscious experience is not physical. It just means that it involves higher-order circuits that operate independently of those that control behavioral and physiological responses. It is not surprising that medications that alter behavioral and physiological responses in animals do not do much to relieve mental anguish in humans. Possibly medications developed in human studies would do a better job. But only if the researchers accept that changing a person’s behavior or physiology is not going to do the trick. The target has to be subjective experience itself. The core problem of mental anguish might be best treated with psychosocial approaches. Medications can also have a role in helping to ease process, but not as a cure or primary treatment.

My lab studies synaptic plasticity processes in memory, as well as in the treatment of mood disorders. I think of memory and mood as manifestations of brain function that rely on processes from neurons and their synaptic contents. These two distinct processes may be embedded within the same neuronal synaptic network yet have distinct features. Forming memories requires encoding precise bits of information in our synaptic networks. Precision, long-term storage and retrievability are essential features of memory. Mood, in contrast, is a global regulator of synaptic networks that does not necessarily influence the specific information encoded but rather determines how and when it is encoded and retrieved, and what actions we choose to take in response to this information. In essence, memory is the information, whereas mood is how we use and react to this information. Therefore, these two processes may arise from the same synaptic networks in the brain but rely on their distinct abilities to process information.

Memory and mood are clearly similar in the sense that they are defined by our subjective experience and can have powerful effects on how we think, feel and behave. But I see at least three ways in which they differ, which are directly relevant to how we can most successfully study them. First, memory is (at least often) tethered to the objective world, whereas mood is wholly subjective; whether I went to Astroworld as a child is a matter of fact, but whether I am happy today is simply a matter of my subjective experience of happiness. This external verifiability of memory has made it much more tractable to study, and particularly to develop models in nonhuman animals. Second, memory poses a much clearer set of computational problems, such as recognition (differentiating previously experienced stimuli from similar but novel stimuli) and recall (retrieving details about previous experiences). This has led to the development of successful computational models of memory that have made it possible to computationally characterize neural circuits for memory. The computational problem that mood solves, and whether it even makes sense to analyze mood in computational terms, are much less clear. Finally, with memory, we may have simply gotten lucky with regard to functional anatomy. The fact that limited lesions to a small set of brain structures in the medial temporal lobe can result in punctate memory disorders has provided a platform for subsequent analysis and modeling of those neural circuits. There does not appear to be any similar high-value target for mood. Instead, it seems that mood disorders can result from lesions across a widespread set of brain regions. The “localize and decompose” strategy may have worked well for memory, but mood may require a different strategy, such as one grounded in complex systems theory.

Memory and mood both have genuine causes within and outside of the brain, but they also differ in important ways. In particular, it is worth considering whether mood states have more psychological causes and characterizations (in contrast to biological and neurobiological causes), especially when compared with memory. One main reason for this is evidence that moods can be successfully and strongly influenced by “self-interventions,” such as when people alter their own thoughts and emotions as a way of changing their behavior. Cognitive behavior therapy, talk therapy, and mindfulness and meditation all fall into this category. Though memory can also be influenced by psychological factors, it seems somewhat less subject to self-intervention. And memory has other features that make it easier to connect to neurobiological-level causes—memory is more straightforwardly conceived of as a coherent, functional system, whereas “mood” is often an umbrella term for disparate states, such as depression, anxiety and fear, many of which gain scientific attention as dysfunction or disease, rather than functional systems. Further exploring these similarities and differences is important for supporting advances in both.

Our understanding of mood at a computational and neurobiological level lags behind our understanding of memory. I think this is largely because of mood’s subjective nature. We can ask an animal which picture they have seen before, but we can’t easily ask an animal if they are happy. Can we just use neural and physiological measurements instead? Not quite. A person can be wrong about whether they have seen a picture before, but they can’t really be wrong about whether they are happy, no matter what our measurements say. That doesn’t mean we can’t ask people what they are feeling and use mathematical models to predict what they will say. This approach has shown that happiness depends on whether you are doing better than expected recently, and that sometimes learning matters more for happiness than reward (you might have a hobby where this finding applies). It turns out that happiness can sometimes go up and down in much the same way in people with and without major depression, a disorder in which diagnosis depends on subjective reports. There’s no reason we can’t better understand mood at both a psychological and biological level, and I’m optimistic that this improved understanding will lead to some of the new treatments we so badly need.

Mood states are complex and multidimensional, so therapies for mood disorders could benefit from traversing several levels. Our research has found that variations in a person’s mood can be decoded from their brain activity patterns. We measured the depression and anxiety subscales of mood with self-reports and developed machine-learning models to decode them. Though our models were trained in a personalized fashion, they often recruited some common brain regions for decoding, such as the orbitofrontal cortex. Evidence suggests that electrically stimulating some brain regions (for example, the orbitofrontal cortex) may alleviate mood symptoms. Combined, these findings suggest that certain brain activity patterns can read out mood states, and that regulating these brain activity patterns may provide a way to regulate mood symptoms in mood disorders. That is why we are excited about the possibility of developing brain-computer interfaces as alternative personalized therapies for those in whom no current therapy is effective. These brain interfaces could potentially read-out the patient’s mood symptoms, which can then be used to optimize electrical stimulation patterns using machine-learning models. This may allow us to tailor neurostimulation therapies to a patient’s own needs and provide a potential personalized therapy for mood disorders at the brain level.

As a neuropsychiatrist, I can explain some causes of my patient’s memory disorder, and I can usually offer effective treatments for their mood disorder—but the converse is not satisfyingly true. How will we fill those gaps? First, we might need to recognize a fallacious dichotomy: either the disorders are fundamentally similar, meaning both are brain disorders, or fundamentally different, meaning one is of the brain and the other of the mind. Both models are useful, but both are wrong (like all models). This ancient question plagued philosophers from Buddha to Descartes. But it was more relevant when epilepsy was attributed to evil spirits, all lung disorders were called “pleuritis,” and the pineal gland was considered the seat of the soul. Medicine has since moved on, pragmatically adopting empiricism over Descartes’ rationalism. It would be absurd to debate whether COVID-19 and lung cancer are fundamentally similar or different. Similarities and differences are both obvious. Perhaps more importantly, we have vastly different treatment approaches for COVID-19 and lung cancer, but also commonalities—we can reduce mortality globally by vaccinating against infection, raising nicotine taxes and improving health-care access. So are mood disorders caused by social factors or biological factors? The answer is “both.” But the question is wrong. I won’t live to see the full chain of causality for either mood or memory disorders. For now, we can make progress by mapping which parts of that chain can be targeted in which situations to reduce real suffering in real people.

Enthusiasts and skeptics of a biological explanation of human behavioral traits face twin challenges. For us skeptics, the trick is to avoid ghost-in-the-machine dualism. No soul, no free will, no mind, unless it is unambiguously rooted in physical, evolved organic reality. Enthusiasts must avoid tautological materialism. I get it: all thinking, wanting and believing is done with the brain, and if all you have to say is that therefore aphasia and religious devotion are all equally “in your brain,” fine, but it gets you nowhere. All assertions that a complex behavior is “biological” should be paired with an example of something that is not. In 1998, I imagined two silent people: a person who had a stroke in the left frontal lobe that resulted in a deep aphasia, and a monk who had taken a vow of silence as a matter of religious devotion. No dualism: Both silences are in their brains, somehow. Why do they seem so different? The answer is mostly about entities and the language we use to define and describe them. Aphasia, as a category, corresponds to a class of neurological events resulting from stroke. There is no class of neurological events that corresponds to devotional silence.