This post was originally published on Salon.
Lately, some neuroscientists have been struggling with an identity crisis: what do we believe, and what do we want to achieve? Is it enough to study the brain’s machinery, or are we missing its larger design?
Scholars have pondered the mind since Aristotle, and scientists have studied the nervous system since the mid-1800s, but neuroscience as we recognize it today did not coalesce as a distinct study until the early 1960s. In the first ever Annual Review of Neuroscience, the editors recalled that in the years immediately after World War II, scientists felt a “growing appreciation that few things are more important than understanding how the nervous system controls behavior.” This “growing appreciation” brought together researchers scattered across many well-established fields – anatomy, physiology, pharmacology, psychology, medicine, behavior – and united them in the newly coined discipline of neuroscience.
It was clear to those researchers that studying the nervous system needed knowledge and techniques from many other disciplines. The Neuroscience Research Program at MIT, established in 1962, brought together scientists from multiple universities in an attempt to bridge neuroscience with biology, immunology, genetics, molecular biology, chemistry, and physics. The first ever Department of Neurobiology was established at Harvard in 1966 under the direction of six professors: a physician, two neurophysiologists, two neuroanatomists, and a biochemist. The first meeting of the Society for Neuroscience was held the next year, where scientists from diverse fields met to discuss and debate nervous systems and behavior, using any method they thought relevant or optimal.
These pioneers of neuroscience sought to understand the relationship between the nervous system and behavior. But what exactly is behavior? Does the nervous system actually control behavior? And when can we say that we are really “understanding” anything?
It may sound pedantic or philosophical to worry about definitions of “behavior,” “control,” and “understanding.” But for a field as young and diverse as neuroscience, dismissing these foundational discussions can cause a great deal of confusion, which in turn can bog down progress for years, if not decades. Unfortunately for today’s neuroscientists, we rarely talk about the assumptions that underlie our research.
“Understanding,” for instance, means different things to different people. For an engineer, to understand something is to be able to build it; for a physicist, to understand something is to be able to create a mathematical model that can predict it. By these definitions, we don’t currently “understand” the brain – and it’s unclear what kind of detective work might solve that mystery.
Many neuroscientists believe that the detective work consists of two main parts: describing in great detail the molecular bits and pieces of the brain, and causing a reliable change in behavior by changing something about those bits and pieces. From this perspective, behavior is an easily observable phenomena – one that can be used as a measurement.
But since the beginning of neuroscience, a vocal and persistent minority has argued that detective work of this kind, no matter how detailed, cannot bring us closer to “understanding” the relationship between the nervous system and behavior. The dominant, granular view of neuroscience contains several problematic assumptions about behavior, the dissenters say, in an argument most recently made earlier this year by John Krakauer, Asif Ghazanfar, Alex Gomez-Marin, Malcolm MacIver, and David Poeppel in a paper called “Neuroscience Needs Behavior: Correcting a Reductionist Bias.”
These authors argue that if we want to understand the relationship between brains and behavior, it would clearly be better to study both: the parts of nervous systems as well as the natural behaviors that shaped its evolution and development – the behaviors that helped make the nervous system in the first place. Most neuroscience today places a premium on extremely detailed recordings of the smaller components of nervous systems, such as tagging proteins on cell membranes to better photograph single neurons, or building tiny assemblies of metal pins to measure the electrical activity in a region of the brain. Unfortunately, as the authors note, much less value has been placed on the rigorous and detailed study of behavior. Why is there so little interest in nurturing the study of behavior, and such intense interest in detailing the nervous system?
The authors suspect a twofold problem: relationships are difficult to study, and technology seems more rewarding in the short term. Understanding the relationship between nervous systems and behavior is undeniably hard. Progress can be slow and challenging to evaluate. On the other hand, exciting advanced technology has made it possible to study the components of brains in unprecedented detail. Technical advances come quickly and with clear, data-based measurements, but their methods often favor simple, data-driven verification questions over knotty, more conceptual questions about behavior.
These speedy, seemingly clear rewards have convinced many neuroscientists that studying the properties and interactions of individual cells is the best path towards understanding the nervous system and behavior. But just because certain neurons in a monkey’s brain change their electrical activity when that monkey sees a face doesn’t mean that those neurons can “recognize” a face. They are just one tiny part of the process that the whole monkey uses to recognize faces. In other words, scientists can sometimes forget that “the whole is greater than the sum of its parts,” and that a brain is just part of a human being, not unlike a stomach or a heart. Individual or even groups of neurons cannot see, feel, think, or behave, and yet neuroscientists often talk about them as if they can.
The pull of ‘publish or perish’
The siren call of technology isn’t the only reason why this narrow view of neuroscience seems attractive to researchers. Society, too, creates incentives for researchers to limit their projects. For one, a great deal of money is tied up in developing the technology behind neuroscience tools, and money makes experiments go round. Scientists understandably can’t ignore the politics around money.
Additionally, neuroscience PhD programs have too often become a race to publish as many times as possible, in all the “correct” places and with all the “right” people. In order to survive in academia, neuroscientists often have to pass the particular muster of journal editors, who frequently reject behavioral studies for “not having enough neuro,” as Ghazanfar said in a recent interview, “as if every paper needs to be a methodological decathlon in order to be considered important.”
Addressing these points will require a great deal of change, both in science and in society, and history again provides some clues for how we might be better detectives.
Body as a sensor
In 1973, psychologist and control theory engineer William T. Powers argued that instead of assuming that brains control behavior based on sensory stimuli, it makes more sense to assume that brains adapt behavior to control what stimuli it gets from the world. Put another way, if we consider “the body” (a.k.a. all of our parts that aren’t the nervous system) to be the brain’s sensors, then “the behavior” that we see is the brain moving its sensors so that it can get the input it wants.
For instance, when we wish to cross the street, we turn our head so that our eyes can look for any oncoming traffic; when we’re in a conversation with a quiet talker, we lean our bodies in and direct our ears towards the speaker to better hear their words. In all these situations, we have a goal, and we move our bodies in order to get the sensory input that lets us achieve our goal.
Powers’ framework grants that there can be movements without a purpose, but for a movement to be considered behavior it must have a purpose. Thus if we want to study behavior, we need a way to distinguish between behavior and all possible movements. How can we do this?
A new way forward
This is where the study of animals, moving freely in their ecological niches, becomes crucial to validate any neuroscientific study of behavior. While we can safely assume that one of the goals of every living creature is to not die, we know from our own experience as humans that survival alone is hardly satisfying: we want, in addition to not dying, to feel safe, healthy, productive, fulfilled.
Why do factors as subjective as our preferences matter in the quest to understand the nervous system? Powers emphasized that any theory or model of the nervous system must be consistent with our subjective experience of living with one. Thus, detailed descriptions of these two factors – our subjective life experiences, and the behaviors of free humans and animals – are just as important as detailed descriptions of nervous systems, because they teach us about the goals of the brain under our study, and thus allow us to pick out behaviors from movements.
Neuroscience isn’t entirely lacking in the kind of detective work that Powers advocates. Neurologist Oliver Sacks’ meticulous notes on his neurological patients’ case histories revealed extraordinary insights into some of the most puzzling aspects of brain function, such as proprioception (knowing where your body parts are without having to look at them), aphasia (the inability to comprehend or express thoughts through language), and blindsight (the ability to respond to visual stimuli without consciously seeing). Many of Sacks’ patients were referred to him after they were deemed too difficult to diagnose by other methods. Another example is zoologist Konrad Lorenz’s extensive use of photography and film to research the evolution of behavior. He used these films not just for analyzing behavior but also to share his observations more fully while teaching or giving talks; if a picture is worth a thousand words, a film can convey things that escape verbal description.
More of this kind of detective work is urgently needed. The stark differences in how different neuroscientists understand behavior causes enormous confusion and frustration, especially among incoming PhD students. In the current training of neuroscientists, most PhD programs do not discuss the history or goals of modern academic neuroscience. But if we are trying to understand the relationship between thing A (nervous systems) and thing B (behavior), we would do well to study both things with close detail. And if we don’t want to be hindered by potentially false assumptions, we need to be clear about proposed relationships, because that very first assumption will decide exactly what we label as “data” versus “noise”.
Lessons from history
The past shows us where neuroscience has been and how we can move forward. We first assumed that nervous systems control behavior, then realized flaws in that framework. If the nervous system is just driving behavior, the way a person can drive a car, then studying the parts of the person won’t explain how the car works, or why the person is driving in a particular direction. This idea also recalls the endlessly regressive homunculus argument, where the hard questions – “how does this behavior happen?” – are delayed by creating one black box inside another, one mystery just beyond the grasp of the one right in front of you.
Perhaps it’s time for neuroscience to try out a new paradigm. Even if there is no doubt that neuroscience needs sophisticated technological tools, we need equally sophisticated models of how the many parts of nervous systems work together to make our movements serve specific goals. Already, we can sometimes become lost in the struggle to interpret the sea of data delivered by our tools. It’s time to turn toward new models of the nervous system based on an adjusted set of assumptions.
We asked other neuroscientists to respond with some commentary to this article. In a very small way, this is how peer-review works in scientific journals. We wanted to give you a taste of what scientific discussion looks like! If you want to know more, feel free to contact the scientists directly via Twitter.
Benjamin Bell: This is a really interesting walk through the foundations of neuroscience. But I somewhat disagree with one of the authors’ main assertions, where they claim that researchers focus primarily on the molecular components of the nervous system at the cost of a careful analysis of behavior.
In my experience, much of modern neuroscience actually works off the behavioral model: researchers carefully study a specific behavior in a model animal and then perturb certain regions or components of its brain and observe the effects on behavior. Just for an example, basically the entire field of sleep research is based on this model.
Still, I do agree with the authors that this still takes the shape of a reductionist approach. In order to really understand the relationships between neural mechanisms and behavior, we have to carefully examine only one small aspect of a behavior in the lab. But in nature, this action would be many times more complex and involve much more functionality than we are equipped to measure currently. I fully agree that reductionism in the lab is losing the forest for the trees, but I suspect in order to truly understand the brain, we will need to first break it down into the smallest of cause-effect components possible, and only afterwards we can begin to understand the larger frameworks of the mind.
Kim responds: We’re not saying manipulating the brain is not useful, only that we should know what our assumptions are.
We are concerned with the assumption that changes in behavior are caused by perturbing certain regions/components of the brain, which implies that the brain doesn’t respond and change when an external force perturbs it. This perspective on behavior doesn’t account for what we know about the evolution and development of the nervous system, in particular phenomena such as neuroplasticity.
The current “behavioral” methods try to assign “function” to circuits by assuming those circuits are fixed and rigid, but this is actually a very important difference between nervous systems and computers, our current favorite analogy for the brain. We now know that nervous systems change and adapt their own structure and physiology all the time. If our goal is to understand behavior by manipulating neural circuits, we need to be much more strict about the assumptions underlying our manipulations, because we aren’t the only ones making manipulations — the nervous system is also making manipulations on itself.
I completely agree that if “behavior” means any movement output of the organism, studying behavior in the wild would be entirely intractable. We believe that behaviors are movements that serve a purpose for the organism, and the defining aspect of a behavior is not the specific motor output but the goal.
This post was originally published on Salon.