Advancing Neuroscientific Understanding of Brain-Behavior Relationship: A Conversation with Nicole C. Rust
Nicole C. Rust, PhD, is a Professor and brain researcher in the Department of Psychology at the University of Pennsylvania. Her research combines behavioral, neural, and computational approaches to understand the brain’s remarkable ability to remember the things we’ve seen and to determine how memory dysfunction can be repaired. She is currently writing a book focused on the types of understanding that will be required to treat neurological and psychiatric dysfunction. In it, she argues that true progress in our understanding of the brain will require ambitious and unprecedented collaborations between scientists and clinicians (as well as engineers and society), including multidisciplinary conversations of the type that we are having here.
Awais Aftab, MD, is a Clinical Assistant Professor of Psychiatry at Case Western Reserve University. He is interested in conceptual and philosophical issues in psychiatry and is the author of this Substack newsletter.
In this conversation, we both asked each other 3 questions while allowing for a follow-up question or comment.
Aftab: You’ve spent a large portion of your research career studying the neural basis of visual memory. What are some striking or surprising features of the neuroscience of vision and visual memory that have become better recognized in recent decades? Do they offer us any insights that apply more generally to brain functioning?
Rust: As a PhD student, I was drawn to study vision because of its promise to serve as a prototype for understanding brain function in general – in that, it certainly has not disappointed me! Brain researchers studying vision and computer scientists working to build artificial intelligence (AI) have a long and productive history of fruitful interactions, culminating in some of the great AI successes that we have today. It’s been thrilling to watch it happen. But that research program is also constructed around the metaphor that the human brain is an information processing machine that “computes” things. I'm compelled by arguments that we'll need other metaphors for thinking about the brain to understand the systems that go awry in mental disorders.
When I became a professor overseeing my own research program, my pivot to studying memory was targeted, in part, toward pushing the boundaries of understanding vision into some of the brain's more mysterious functions. For instance, we all know what it feels like to have the percept, “I've seen that person before.” When we experience it, we must be tapping into some type of “signal” in our brains — but what exactly are we tapping into? The brain areas that make up the visual system are clearly involved in vision but there really isn't an analogous “familiarity” brain region devoted to this type of memory. For instance, patients with hippocampal damage are not impaired for it. So where exactly is the familiarity signal in the brain and how exactly is it reflected there? In our work, we’ve determined that familiarity is reflected diffusely across the visual system itself, as well as many brain areas downstream. Namely, we know that in the visual system, visual information about what you are looking at is reflected by the pattern of neurons that are firing. That pattern is similar when you look at similar things and different when you look at different things. In those same brain areas, memory about whether you’ve seen it before is reflected by the overall vigor of the response (across all neurons). If you see something for the first time, you get a certain pattern and a vigorous response. If you see it again, you get the same pattern but a less vigorous response (think: all the same neurons fire but 10% less). All the follow-up evidence points to the idea that this is the signal you tap into when you perceive familiarity. From that understanding, we’ve begun to trace out the details of how a visual experience is transformed into a memory.
Rust: If you see something for the first time, you get a certain pattern of neurons that are firing and a vigorous response. If you see it again, you get the same pattern but a less vigorous response. All the follow-up evidence points to the idea that this is the signal you tap into when you perceive familiarity.
Analogous mysteries exist for the types of mental states that are even more relevant for psychiatry, like emotions and mood. When you are feeling happy or sad (or anxious or calm), you must be tapping into some type of signal in your brain – but where exactly is that signal and how exactly does it manifest? While that question is arguably more difficult to study for mood than memory, it’s also arguably even more important to understand. Moreover, one can in principle use similar “decoding” approaches to the ones we are working with for memory to figure it out, and researchers have had some preliminary success with this type of approach. Should researchers successfully pinpoint the neural correlates of mood, that would give them tremendous leverage to better understand what causes mood fluctuations, and it could shed insights into how to intervene with depression.
Aftab: The neuroscience that I encountered during my psychiatric training had a strong reductionistic flavor… it was very much about figuring out the neurotransmitters, the brain areas, the brain circuits, for this or that psychiatric phenomenon. The neuroscience that I discovered subsequently (mostly on my own) — thanks to public-facing scientists such as Lisa Feldman Barrett, Anil Seth, Mark Solms, Luiz Pessoa, Karl Friston, etc., was of a very different flavor. It was very much grounded in computational, predictive, enactive, and complex systems approaches, and understood the mind as arising from complex interactions of multiple parts of the brain with each other, with the environment, and with other brains. The difference was striking to me but I’m also a neuroscience outsider, a clinician peering in. How would you characterize the current state of affairs in neuroscience? How have your own ideas about neuroscience changed over the course of your career?
Aftab: The neuroscience that I encountered during my psychiatric training had a strong reductionistic flavor. The neuroscience that I discovered subsequently was very much grounded in computational, predictive, enactive, and complex systems approaches…
Rust: Indeed, you still see that strong reductionist flavor reflected in many contemporary neuroscience textbooks. The psychiatrist and Nobel laureate Eric Kandel channeled much of that ethos in his 1998 paper, “A new intellectual framework for psychiatry.” In it, he proposed that all psychiatric phenomena (including the benefits of psychotherapy) could be traced back to changes in genetic expression. But he left the details of how the brain gives rise to the mind unspecified, “The great challenge for biology and psychiatry at this point is to delineate that relationship in terms that are satisfying to both the biologist of the brain and the psychiatrist of the mind.” In the 25 years since Kandel’s paper, filling in that gap has proven challenging.
One illustration of the challenges involved focuses on arguably the most elusive function of the human brain — consciousness. It turns out that we are only conscious of a small fraction of the information that activates our sensory receptors (like the photoreceptors in our eyes) and one big goal is to understand what happens in our brain to support our conscious percepts. In the same year that Kandel proposed his framework, another prominent brain researcher, Christoph Koch, speculated that the answer would involve a small set of neurons with special, intrinsic properties, like the selective expression of certain genes to regulate the production of a particular neurotransmitter or a neuron's pattern of electric firing. The idea that the origins of human consciousness might be traced back to genetic expression aligns with the sentiments spelled out by Kandel.
A decade or so later, Koch along with many others began to advocate for proposals about consciousness that are vastly more complex — in fact, Koch’s new favored theory, integrated information theory (IIT), is so complicated that only a handful of researchers claim to understand it. Formalized by Giulio Tononi in 2004, IIT claims that consciousness is a certain type of information called phi. The gist of IIT is that a system is conscious to the extent that its whole generates more information than its parts, and phi captures the degree to which this happens. Notably, there are no genes or molecules in the theory (at least not yet). While it still proposes that the brain gives rise to the mind, by definition it relies on emergent properties to get there. It's an enormous jump in complexity relative to earlier theories about consciousness, but that jump also comes with a cost — one of the primary concerns about IIT is that many of its details are currently untestable, or as the saying goes, the theory is not even wrong. Researchers continue to refine and develop it.
Rust: There’s a strong sense that the path forward involves appreciating that the brain is a dynamical, complex system with emergent properties. At the same time, brain research absolutely needs theories that make testable predictions to move forward…
This example reflects what brain research is up against today. There’s a strong sense that the path forward involves appreciating that the brain is a dynamical, complex system with emergent properties. At the same time, brain research absolutely needs theories that make testable predictions to move forward and in the case of complex systems, they are (as their name suggests) “complex.” Fortunately, researchers are having some success developing and testing theories of that type, not just for consciousness but also for many other functions of the brain and mind that are also relevant to mental disorders. A few examples of sizable resources invested in that direction include the new division of the Allen Institute focused on Neural Dynamics and the Simons Foundation’s Collaboration on the Global Brain. I expect that we'll see some really amazing discoveries “emerge” in this space (wink wink) in the next 25 years.
Aftab: What are your thoughts on the whole “are mental disorders brain disorders?” debate? Any important disagreements from how I approach the issue (as outlined in my article for Philosophical Psychology)?
Rust: I’ve thought a lot about this question (in the context of a book I'm writing). As I see it, the literature on this topic is a mixture of unnecessary confusion and profoundly interesting questions; I'll highlight one of each type in ways that complement your article (which I appreciated).
The first thing to know is that all sides of the debate agree that all mental function is mediated by the brain. That is, some might suspect that mental function emerges from brain function, but everyone agrees that changes in the mind cannot happen without physical changes in the brain. This is thus not a debate about materialism versus dualism; rather, the issue comes down to a debate about the best level to explain and intervene — is it the brain, the mind, or some combination?
Rust: … all sides of the debate agree that all mental function is mediated by the brain. That is, some might suspect that mental function emerges from brain function, but everyone agrees that changes in the mind cannot happen without physical changes in the brain. The issue comes down to a debate about the best level to explain and intervene…
To illustrate, consider the phrase: “insomnia causes fatigue.” We all agree that this causal chain is mediated by the brain. Namely, the brain needs sleep to prevent it from entering the brain state that produces the mental state of fatigue. We also agree that in this case, thinking about the brain (and its ion balances and neural circuits) isn't a good route to the most obvious intervention for fatigue: get some sleep.
Many of the arguments on the “mental disorders are not brain disorders” side of this debate boil down to an extension of this straightforward idea: once mental variables start interacting in complex ways to “cause” mental disorders, it can be more useful to think about causal interactions at the mental (as opposed to brain) level, like the insomnia example. This point is not so profound — we call the neurodegenerative disease Huntington's a “brain disorder” as opposed to a “subatomic particle disorder” after all. Similarly, some of my colleagues prefer to be called “Psychologists” and others “Neuroscientists” based on the level that they emphasize in their research.
The position here is quite simply that the brain will not always be the singular best level to explain the causes of mental disorders or think about interventions. My impression is that nearly all researchers who use the terms “mental disorder” and “brain disorder” interchangeably agree with this position. Even Kandel (who speculated that all changes in mental function involve changes in genetic expression, as described above) explicitly championed this position in his 1998 paper. So, I wonder: why are some making the consensus and obvious position appear profound and complicated? Are we really discussing concepts here? Or are we simply arguing over the modifier “brain” (in “brain disorder”) in the absence of any good options for this situation?
That said, there are in fact some other more subtle (and interesting) issues at play here as well. One is a question about whether mental states can have a top-down causal influence on brain states. Answering that question comes down to tricky issues about what you mean by the word “cause” (which one would think would be straightforward but turns out not to be). One view is that the answer “yes” would clash with known physics, where higher levels can emerge but not have independent causal influences; as such, it is deemed unlikely. Another view is that under some definitions of causality (namely the interventionist account widely adopted by the medical community), top-down mental influences are a reasonable way to think about things. This is where the conversation begins to get really interesting, I think. I wonder: from a purely practical perspective, which of these two ways of thinking will best lead us to new and better interventions in the longer term?
Aftab: Your analysis is quite accurate and reveals that there is more in common among proponents of these views than is often acknowledged. An aspect of this debate that tends to generate controversy relates to the level at which things can be said to have gone wrong. Conceptualizing mental disorders as brain disorders tends to encourage thinking of brain processes themselves as dysfunctional, in ways that can be characterized using neuroimaging, neurochemistry, EEG, etc. In contrast, mental disorders may be dysfunctional in ways that are only characterized as such with reference to behaviors and their context, even though these behaviors are mediated by the brain. For instance, anxiety that is out of proportion to the context, or coping strategies that outlive their adaptive value, or beliefs that violate norms of rationality, etc. It isn’t quite accurate to say that the brain is “normal” in such instances – I find it difficult to maintain that a depressed brain or a delusional brain is a healthy, functional brain! — but it also isn’t quite accurate to say that the brain processes themselves are dysfunctional. It’s more like, as Anneli Jefferson points out, that the brain is dysfunctional in such instances but only in an inferential sense with reference to behavioral dysfunction. This sort of nuance has been difficult for clinicians and patients to grasp though, and hence the pushback against the idea that in mental illness it is brain itself that is malfunctioning.
Aftab: An aspect of this debate that tends to generate controversy relates to the level at which things can be said to have gone wrong…
Rust: This is a great point. Figuring out how to best think about what is going wrong is still an open question. To move forward, we need to conceptualize what the answers might look like and do so in a manner that resonates with everyone–patients, clinicians, and scientists alike. How do we achieve this? These notions are nontrivial but they are not impenetrable. For instance, first-year university students grasp the insomnia example presented above and appreciate the gist of how it generalizes to mental disorders. I suspect that a bigger challenge is the natural human tendency to presume that whatever we were drawn to focusing upon first (brains vs. mental states) is the thing that “matters” most, and after that, we put our fingers in our ears so we can focus on details about why and how. Perhaps more of us need to engage in conversations “across the divide,” like the one we are having here.
“Psychiatry at the Margins” is a newsletter about exploring critical, philosophical, and scientific debates in psychiatric practice and the psy-sciences.
Rust: Many foundational brain/mind researchers regard philosophy as unhelpful (although I note that I am not among them). You obviously (also) see that differently. In broad strokes, how do you think philosophy can help navigate the path forward?
Aftab: In a 1992 article, the philosopher Mary Midgley draws an instructive parallel between philosophy and plumbing. She writes:
“Plumbing and philosophy are both activities that arise because elaborate cultures like ours have, beneath their surface, a fairly complex system which is usually unnoticed, but which sometimes goes wrong. In both cases, this can have serious consequences. Each system supplies vital needs for those who live above it. Each is hard to repair when it does go wrong, because neither of them was ever consciously planned as a whole. There have been many ambitious attempts to reshape both of them, but existing complications are usually too widespread to allow a completely new start... [However, a striking difference is that] When the concepts we are living by function badly, they do not usually drip audibly through the ceiling or swamp the kitchen floor. They just quietly distort and obstruct our thinking.”
I think most people who spend time studying the challenges facing brain-behavior research appreciate at some level that we have been immersed in concepts that have quietly distorted and obstructed our thinking, although opinions differ tremendously on which concepts are problematic and in what way! Causality, reductionism, representation, computation, etc. are probably concepts where a lack of clarity has been historically problematic. A recent paper by Westlin et al. “Improving the study of brain-behavior relationships by revisiting basic assumptions,” with Lisa Feldman-Barrett as senior author, generated some really interesting discussion in the neuroscience community on social media, and illustrates (to me, at least) how much room there is for us to challenge assumptions that have been taken for granted by many.
When it comes to brain-behavior research, we are inevitably operating with theoretical assumptions that are guiding and organizing research. The choice is between whether we are going to try to pretend that these assumptions do not exist, or we are going to engage with them with the conceptual rigor that is needed. In the context of psychiatric training, I have argued in support of the idea of “conceptual competence” which tries to do something quite similar and expects clinicians to acquire this sort of competence. By conceptual competence, I refer to the health professional’s transformative awareness of background conceptual assumptions (held by clinicians, patients, and society) and their influence on various aspects of clinical practice, research, and education.
Rust: My impression is that we haven’t figured out the right ways to think about many problems in brain/mind research. Rather, it often seems like we are working in frameworks akin to Ptolomy’s wonky descriptions of planetary motion (under the misguided assumption that the planets revolve around the earth). If that analogy holds, we are waiting for a Copernican revolution to unlock the path forward (by discovering something equivalent to the realization that things all make much more sense if the planets, including the earth, revolve around the sun). That’s honestly not a criticism: understanding the brain and mind, and their dysfunctions, is among the most complex challenges humanity has ever faced and we’ll no doubt stumble a bit before we figure it out.
Rust: … we haven’t figured out the right ways to think about many problems in brain/mind research. Rather, it often seems like we are working in frameworks akin to Ptolomy’s wonky descriptions of planetary motion (under the misguided assumption that the planets revolve around the earth).
Copernicus began with the suspicion that something about Ptolomy’s model wasn’t right, wondering “whether there could perhaps be found a more reasonable arrangement of circles, from which ... everything in itself would move uniformly.” Are there one or more aspects of brain/mind research for which you have a suspicion that things aren't quite right (even if you don't yet know of better ways to think about them)?
Aftab: My understanding is that Ptolemy’s model was fairly successful in accounting for planetary positions (and predicting things like solar and lunar eclipses) and enjoyed near universal support among scholars until it was displaced. It is difficult to think of anything comparable in the brain-behavior sciences. There is no unifying model of brain-behavior relationship that is empirically successful and is supported by scientific consensus. Different approaches to research are often dominant at different times, but that dominance is not typically because of compelling evidence; it is usually driven by the shortcomings of the previously dominant approach and by the enthusiasm that this or that new approach will deliver will the answers. I may be wrong, but that’s the impression I have as a clinician in a field that is adjacent to neuroscience.
In psychiatric science, the shift from psychoanalysis to a neo-Kraepelinian framework in the 1970s is a good example of this. One of the ideas implicit in the neo-Kraepelinian framework is that there are discrete disease entities in psychopathology and that validators (such as genetics, neuroimaging markers, course of illness, treatment response, etc.) would eventually converge to reveal their existence; a DSM-style classification was seen as a useful starting point for this undertaking, subject to iterative improvement and increased convergence onto the hypothesized disease entities. It has become obvious since the early 2000s that no such convergence of validators is evident. We have spent decades studying psychopathology looking for categorical disease entities, and the project has failed to find specific biomarkers or neurobiological signatures corresponding to major psychiatric syndromes. Psychopathology, instead, appears to be dimensional (as far as symptom distribution at the population level is concerned) and neurobiological relationships appear to be transdiagnostic (cutting across many traditional syndromes), so researchers are now having to figure out how to navigate this landscape.
I think the history so far is quite instructive of the dangers of premature closure and a near-exclusive focus on a single set of ideas for scientific research. People keep looking for a “paradigm shift” in psychiatric science, but what we need is a rigorous and open-minded consideration of a wide range of hypotheses. We need multiple lines of inquiry, because we don’t know enough about the basic mechanisms of psychopathology to put all our eggs – research funding, resources, respectability – in one basket, so to speak.
Aftab: People keep looking for a “paradigm shift” in psychiatric science, but what we need is a rigorous and open-minded consideration of a wide range of hypotheses. We need multiple lines of inquiry, because we don’t know enough about the basic mechanisms of psychopathology to put all our eggs – research funding, resources, respectability – in one basket…
Here are some aspects of brain-behavior relationship, especially in the context of psychopathology, that I think nudge us towards revisiting conventional ideas:
Lack of clear separation between “normal” psychological phenomena and psychopathology (neurophysiological and neurocognitive mechanisms likely don’t respect our judgments of psychopathology)
Lack of clear separation between different forms of psychopathology (Are there natural kinds in psychopathology? What would a scientific understanding of psychopathology look like in their absence?)
Lack of specific neurological or bodily signatures for psychological (e.g. emotions) or psychopathological constructs (Is there simply too much heterogeneity, degeneracy, and multiple realization? Or are these constructs socially constructed in some manner?)
High rates of response with placebo intervention in clinical trials (We can take chronic and treatment-resistant patients, say for deep brain stimulation for depression, and sham intervention still ends up doing a fairly decent job… why?)
High rates of treatment failure for established clinical treatments (what mechanisms account for such tremendous variation?)
The predictive coding/processing paradigm of brain functioning seems to me to be an attractive candidate in explaining some of these discrepancies.
Rust: If nudged to offer up a bit of healthy constructive criticism for foundational brain/mind researchers, how would you nudge them? In other words, are there things you'd like to see the nuts-and-bolts arm of brain/mind research do, emphasize, or acknowledge a bit differently?
Aftab: I’ve touched on some of this in the previous answer as well. My impression is that neuroscience researchers can at times be quite dismissive of the complexities of psychological life that aren’t amenable to quantification. I’m thinking of the sorts of complexities that psychodynamic clinicians are particularly attuned to and make productive use in the context of psychotherapy. In addition, the ways we measure psychological traits or behavioral phenomena are very imperfect indicators of what’s going on and leave a lot to be desired. We tend to think any measurement is better than no measurement, but I worry that in some instances, measurement of psychological-behavioral phenomena done poorly may be worse than no measurement because of what it obscures in the process. As one example, the 2022 paper “Revisiting the theoretical and methodological foundations of depression measurement” by Eiko Fried and colleagues was quite sobering in its conclusion that “the measurement of depression rests on shaky methodological and theoretical foundations.” It’s difficult to uncover the neurocognitive mechanisms of depression when we are measuring the construct so poorly.
Aftab: I worry that in some instances, measurement of psychological-behavioral phenomena done poorly may be worse than no measurement because of what it obscures in the process.
Another thing we need to pay more attention to, I think, is variation in brain-behavior relationships. We focus too much on average differences (between groups, categories, etc.) and not enough on the mechanisms that underlie variability. What should neuroscientific explanations look like in a world where there is often greater neurobiological variability within groups than there is between groups we are interested in differentiating? The environmental and social context in which brains are operating is also crucial here. How much of what we are interested in isn’t in the brains per se, but in the interactions of the brains with their environments, or in the interactions between two brains?
Rust: Great to see you bring up the 2022 Revisiting paper; it had a big impact on me as well. That paper draws from the philosopher Hasok Chang’s book “Inventing Temperature” to focus on the question: How do we figure out how to measure something that we don't yet understand, like temperature? (Or depression?) In it, Chang describes what he calls “epistemic iteration,” which is a fancy way for saying that we take our best guess about how to go about it and we iterate our way toward a good solution. Chang applies the idea to the history of how humanity figured out how to measure temperature in the 17th century, starting from our bodily sense of warmth and coolness, through the creation of rudimentary thermometers, and eventually to sophisticated theories about temperature such as thermodynamics.
Rust: What can we learn from the history of temperature to make analogous progress toward understanding depression? Chang’s thesis is that it doesn’t really matter where we begin as long as we continue to iterate.
What can we learn from the history of temperature to make analogous progress toward understanding depression? Chang’s thesis is that it doesn’t really matter where we begin as long as we continue to iterate. The gist is that we will get the measurements wrong before we get them right, but we will get things right eventually (and crucially, starting somewhere is the only way in). But he also emphasizes that epistemic iteration is not a specific prescription of what should happen next in a given field—the considerations for temperature and other things (like depression) will be different.
In that light, I'm intrigued to hear you voice concerns that some instances of measurement might be worse than none (because of what it might obscure). To unpack it a bit more, are you concerned that we cannot (or somehow will not) iterate our way out of the obfuscation? I know that some, including the psychiatrist Ken Kendler, have voiced concerns that if we begin epistemic iteration from a position that is “not even in the ballpark,” we might get stuck in something akin to a local minimum when the true answer lies elsewhere. However, to combat that, Kendler acknowledges that a good approach to is to tackle big questions like depression from multiple perspectives (a principle known as pluralism). I'm very curious to hear more about your own concerns around epistemic iteration as the path forward for depression.
Aftab: I’m a big fan of epistemic iteration too, and I like to think of it as the way forward. When I say that some instances of measurement might be worse than none, I have something different in mind. In the early stages of scientific inquiry, our measurement methods will be limited, imperfect, and error-prone, and that’s ok as long as we recognize that and proceed with humility and pluralism. It becomes a huge problem when we begin to think that our measurements have more validity than they actually possess, and when we are so enchanted by the scientific trappings of the measurements that we cannot attend to all the complex, messy aspects that are not being quantified. Epistemic humility and explanatory pluralism are essential here, and psychiatric science — especially in its interface with clinical practice — has often displayed neither of these virtues. Some diagnostic manuals and rating scales have acquired an authority, dominance, and ubiquity of use that far exceeds their scientific merits, drowning out other forms of conceptualization and measurement.
Aftab: Epistemic humility and explanatory pluralism are essential here, and psychiatric science — especially in its interface with clinical practice — has often displayed neither of these virtues.
Philosopher C. Thi Nguyen has a wonderful lecture for the Royal Institute of Philosophy about “value collapse” where he discusses how overly clear measures of a complex, multifaceted, hard-to-capture phenomenon can narrow our attention and make us neglect aspects that are not being measured. “Hyper-clarity represents a value as finalised — and the world as if there were nothing else to learn from it about what really matters.” This sort of hyperclarity has been evident in psychiatric research (Thomas Insel: “Terms like “depression” or “schizophrenia” or “autism” have achieved a reality that far outstrips their scientific value.”1) and clinical practice (Ken Kendler: “What especially bothered me was that many residents thought that once you had studied the DSM criteria for depression or schizophrenia, you knew pretty much all you needed to know about these clinical syndromes… This attitude drove me a bit crazy as if the publication of the most recent DSM edition covered everything important in the rich and diverse clinical descriptive tradition of our field.”2) but it has been especially egregious when it comes to patient communication and public understanding, where psychiatric diagnostic constructs (and measures) have been communicated to them as if they were validated natural kinds with little to no acknowledgment of their tentative, fallible nature. I once heard Kendler say in a lecture, “Psychopathology is an immature, faddish science” and I loved it so much that I try to say it every time I give a public-facing talk because I think we owe the public this sort of clear acknowledgment.
Rust: These are such important points. I completely agree.
Aftab: Professor Rust, thank you for this conversation!
Psychiatry at the Margins is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.
This was posted by Insel in 2012 in a blog post titled “Words matter” on the NIMH’s Director’s Blog. The post is no longer available on the NIMH website but you can see the post here archived by the Internet Archive)