How to read a brain (or not)

By Saren H. Seeley, Ph.D.

One of the first social neuroscience research studies that I remember reading in graduate school was an older functional MRI study showing that physical and emotional pain had overlapping neural representations – meaning that many of the brain areas involved in physical pain were activated when researchers scanned the brains of people experiencing social pain, such as a recent breakup or being excluded by teammates.

The idea of physical and social pain overlapping in the brain makes a kind of intuitive sense, considering how many idioms we have for distress at the hands of others: rejection leaves us “brokenhearted”; we “ache” for an absent partner; harsh words “bruise” our feelings. Humans are social mammals who rely on caregivers for a relatively long time, compared to other species. Community, friends, and loved ones continue to play a prominent role throughout our entire lifespan, and humans are particularly sensitive to anything that might threaten those connections.

Broken social connection undeniably hurts. Still, the idea that we could see social pain as co-opting the brain’s physical pain system turned out to be controversial (in a low-stakes, academic-niche kind of way). Critics claimed that the idea relied too heavily on an observation that both social and physical pain activated the same segment of the anterior cingulate cortex during functional MRI studies.

The anterior cingulate cortex, shown in yellow, is a band of grey matter that runs along the top of the corpus callosum (a thick, C-shaped bundle of nerve fibers connecting the left and right hemispheres of the brain).

Two brain regions that showed up in these social pain studies were the dorsal anterior cingulate cortex and the anterior insula. Both were active during physical pain – specifically, the emotional aspect of pain that makes it so much more than just an unpleasant physical sensation. So both physical and social pain have “distress” in common – but why else might the brain represent physical pain and the sting of rejection or heartbreak in apparently the same place?

Think of the infinite number of experiences that we can potentially have in a lifetime, and the myriad of tasks that the brain performs every day. If there was a part of the brain exclusively dedicated to that and only that experience or function, we’d rapidly run out of different brain structures long before hitting puberty. Yet, the brain continues to become shaped by new experiences and learn into old age, so some of its functions must have to share.

The assumption that each area in the brain has a one-to-one relationship with a certain task or function reflects a common research pitfall known as reverse inference. Reverse inference occurs when researchers look at brain scan data, observe activity in a particular part of the brain, and – based only on that single observation – infer that people were thinking, feeling, or doing something specific during their scan.

Reverse inference is especially problematic for brain regions that have become famous for one or two things. The amygdala is one such brain region. It is a pair of almond-shaped structures at the base of the brain that has received a great deal of attention for its role in fear, threat, and other emotions that most of us would prefer to avoid. Many people think that we can improve well-being by learning to dampen the amygdala’s activity. This may be true in some cases, for example in PTSD where individuals’ brain system for detecting danger becomes hyper-reactive and hyper-vigilant (much like an over-sensitive car alarm set off by a passing pedestrian). But the amygdala actually has a much more complex role in our lives. It contributes to diverse functions like learning, memory, input filtering, feeding behaviors, maintaining equilibrium in our bodies (homeostasis), and even other, more desirable types of emotions, like surprise or happiness.

Like any other part of the brain, the amygdala does not accomplish all of its tasks alone, but rather through crosstalk with other brain regions. We can imagine each part of the brain as a node in a network, with connections running from one brain region to another and information flowing through the connections. If every single time we saw amygdala activation on a functional MRI image, we would be performing reverse inference, if from that fact alone, we announced “Aha! Given that we see activity in the amygdala, participants in the study were obviously feeling afraid.”  

Every research paper ends with a discussion section, in which the researchers put their results into context to make sense of what they found, name any limitations to the research that readers should consider, and make an argument for what it all means. When I write discussion sections for my own research, I predictably hit a point of existential despair midway through where I begin to question if we can know anything at all from functional MRI data. After all, any brain region I might be looking at could potentially be doing an overwhelming number of other things in addition to what I’m hypothesizing it is doing in my study.

However, there are ways to find a signal in all the noise.

One of the best ways is to triangulate the functional MRI data with other sources of information. If the experiment is designed to elicit fear (such as showing a scary movie scene to participants), it is more likely that amygdala activation reflects something fear-related. We increase our confidence in the “fear” interpretation if the participants report that they did indeed feel terrified during the scan.

We can further probe whether the “fear” interpretation holds by comparing how the amygdala reacts to a related, but distinct condition. Here, we might examine amygdala activity during scary movie watching versus watching a movie scene that is surprising, but not at all scary. If we see a similar level of amygdala activation in both “fear” and “surprise” conditions, that disproves our idea that amygdala activation is just about fear. However, if we see significantly more amygdala activation during “fear” versus “surprise”, then perhaps the amygdala we see in our scans does indeed reflect fear. 

Another way to get more context is to look for patterns of activation across multiple brain regions, instead of focusing on a single part of the brain. Looking for patterns across the brain has revealed that the function (an emotion, a type of thinking, a response to the environment) that we are studying often relies on a network of multiple brain regions acting in tandem to turn “up” and “down” at the right times.

In the case of pain research, a 2014 study used a form of artificial intelligence known as machine learning to zoom in on the hundreds of individual voxels (think 3D pixels) making up the image of the dorsal anterior cingulate. The researchers found that there were two tiny groups of neurons that separately encoded social and physical pain, but these two groups of brain cells blurred together so that they appeared to overlap if you looked at activity in the whole dorsal anterior cingulate. Neuroimaging research has benefited from more mathematically sophisticated ways of analyzing data that provide a closer look at how brain activity is organized.

As a grief and trauma researcher, I find the concept of social mammals having a “neural alarm system” for broken connections fascinating. If this system can alert us to dangers of physical damage and social disconnection alike, that offers an intriguing starting point for thinking about how neurobiology shapes the way we experience events, like the death of a loved one, or a collective trauma in the community. Though I never fully made up my mind about who won the social pain/dorsal anterior cingulate cortex debate, understanding the reverse inference problem taught me to detect overblown conclusions in both the academic literature and science journalism, and to avoid that mistake in my own research.

 

Saren H. Seeley is a Postdoctoral Research Fellow in Psychiatry at the Icahn School of Medicine at Mount Sinai. She completed her PhD in clinical psychology at the University of Arizona and the Pittsburgh VA Medical Center, following undergraduate studies at CUNY Hunter College. Her research at ISMMS aims to identify neural mechanisms of adaptation to trauma and loss, with a current focus on (1) PTSD resilience in a cohort of World Trade Center first responders, and (2) developing a computational model of predictive processing in complicated grief. When not in the lab, she prefers to spend time in the dance studio or walking/running/hiking anywhere outdoors.

Edited by Denise Croote, PhD

 

References

Eisenberger, N. The pain of social disconnection: examining the shared neural underpinnings of physical and social pain. Nat Rev Neurosci 13, 421–434 (2012). https://doi.org/10.1038/nrn3231

Eisenberger, N. I. (2015). Social pain and the brain: Controversies, questions, and where to go from here. Annual review of psychology66, 601-629. https://doi.org/10.1146/annurev-psych-010213-115146

Poldrack, R. A. (2011). Inferring mental states from neuroimaging data: from reverse inference to large-scale decoding. Neuron, 72(5), 692-697. https://doi.org/10.1016/j.neuron.2011.11.001

Woo, C. W., Koban, L., Kross, E., Lindquist, M. A., Banich, M. T., Ruzic, L., ... & Wager, T. D. (2014). Separate neural representations for physical pain and social rejection. Nature communications, 5(1), 1-12.https://doi.org/10.1038/ncomms6380



Saren H. Seeley