The following is a guest post by Rob Chavez.
If I’m ever asked ‘what was a defining moment of your career?’, I can think of a very specific instance that has stuck with me since my early days as a student in social neuroscience. I was at a journal club meeting where we were discussing an early paper using fMRI to investigate facial processing when looking at members of different racial groups. In this paper, the researchers found greater activity in the amygdala for viewing black faces than for white faces. Although the authors were careful not to say it explicitly, the implication for most readers was clear: The ‘threat center’ turned on for black faces more than white faces, therefore the participants may have implicit fear of black faces. Several students in the group brought up criticisms of that interpretation revolving around how the amygdala is involved in other processes, and we started throwing around ideas for study designs to possibly tease apart alternative explanations (e.g. lower-level visual properties, ambiguity, familiarity) that might also account for the amygdala activity.
Then it happened: The professor in the room finally chimed in. “Look, these are interesting ideas, but they don’t really tell us anything about racial bias. I don’t really care about what the amygdala does; I just care what it can tell us about social psychology.” Even in the nascent stages of my career, I was somewhat flabbergasted. Who wouldn’t want to know everything they possibly could about the thing they are using to draw inferences, especially when that thing is part of the source of mechanism? For me, this event marked a turning point where I began to think of neuroscience less as a method for answering psychological questions and started thinking of it as a multidisciplinary endeavor to which psychology has much to contribute.
Now as a card-carrying social neuroscientist, when I attend conferences, such as the Society for Personality and Social Psychology meeting, I am frequently asked what neuroscience can contribute to our understanding of psychology that we don’t already know from behavioral studies, which are often more flexible, less noisy, and much, much cheaper to run. However, contributing to psychological theory or outperforming behavioral predictions are often not the proximal goals for researchers using neuroimaging methods. Instead, much of the interest in social neuroscience stems from the potential of applying insights from psychology (and other disciplines) to better understand how cognitive and social phenomena are represented in the function and structure of the brain for its own sake, and not simply using the brain as a tool or methodology. I believe that these efforts help us refine the link between these levels of analysis and, frankly, are interesting in their own right.
To be fair to the professor at the journal club, there may be a reason that people hold the view that the brain can simply be used as a method or a tool. Many early researchers using neuroimaging were not originally trained in neuroscience per se but instead transitioned over to it from using other kinds of psychophysiological methods. As such, there are understandable reasons why many researchers doing psychophysiological work don’t have much of a motivation to care as deeply about the underlying physiological process itself. For example, if a researcher is doing a study measuring electrodermal activity, chances are that they don’t care very deeply about sweat (and possibly don’t really care about sympathetic nervous system activity), but rather use it as an indicator of emotional arousal. Put differently, nobody assumes that sweat is the origin of arousal or believes that the fingertip is the organ responsible for the seat of the mind.*
This is not true for the brain, and things start to get even more complicated quickly. Even if you want to just use fMRI amygdala activation to be a marker of threat or fear, the path to do so is not as clear as in other physiological measures. (The amygdala is not even a single structure but rather a collection of functionally distinct nuclei, each with its own functional tuning and connectivity profile to other parts of the brain.) I believe that the way many have been taught to think about measures of brain function and structure has been conflated with some of these more peripheral measures in other parts of the body that are obviously not ‘the source’ of the mind. As such, it doesn’t make much sense to ask how psychological processes are represented in skin conductance in the same way as asking how they are represented in the brain (even if using relatively crude and indirect tools like fMRI). Thus, the common criticism of some neuroimaging work that “we already know that the mind happens in the brain” is shortsighted. Yes, but how, when, and at what level of granularity? However, this perspective is not without its challenges.
One of ways in which brain imaging can be useful to psychologists is to know when activation of a particular brain region or network is indicative of a specific psychological phenomenon. However, in the context of neuroimaging, the term reverse inference is a bit of a dirty phrase.** When someone accuses an fMRI researcher of engaging in reverse inference – drawing conclusions about what psychological process is involved given the activation of a brain area – it is usually a criticism. However, reverse inference is indeed one of the overarching goals of how cognitive and social neuroscience inform psychology in general. We want to know when we can make sound inferences about the psychological processes involved based on neuroimaging metrics in a given task or under certain conditions. Although this is major goal of this endeavor, it is only one of them and is often a distal one. What people ought to be criticizing is premature, decontextualized, or otherwise incomplete reverse inference that overreaches on the conclusions drawn from these methods – Does amygdala activation really mean ‘threat’? Are there other processes involved that might explain it? Does that depend on the particular stimuli being used? Is it a single part of the amygdala or several in concert? Even if replicable, how confident are we that the paradigm being used is representative of the possibility space of reasonable paradigms that could have been used instead? – We have acknowledged for a long time that there is almost never a one-to-one mapping of activity in a single brain region to a single psychological process. Tackling the issues of how then to meaningfully relate psychological processes to the brain is what many of us are working on right now.
However, it’s hard to accomplish these goals while pushing the envelope of psychological theory simultaneously. The collective expectation that cognitive and social neuroscience experiments have an obligation to contribute to our understanding of complex psychological phenomena (and not vice versa) is often too premature to be definitive, much less revolutionary. Moreover, I feel strongly that the insistence that we frame the interpretation of our results in ways that placate this expectation has led many, otherwise cautious, researchers to take unwarranted liberties into the dark side of reverse inference. Ironically, this may have has spurred many criticisms of overreach from cognitive and social neuroscience from others within the broader psychology community. Just as it took years for psychometricians to gain an understanding of how measurement and the scope of our inferences make up the scaffolding upon which we can build psychological theory, it is my hope that a more mature understanding of the intersection between neuroscience and psychology can offer analogous insights. However, given the overwhelming complexity of each of these domains, these efforts will take time.
I sometimes like to think of cognitive/social neuroscience as more of an applied field like, say, psychology of law where researchers are using what we understand about cognition and behavior to inform how the processes are deployed in the legal system – except in our case it’s how they are deployed in the brain. If we were talking to a psychology of law scholar, we would never say to them “I don’t care about the law; I just care about what it can tell us about psychology”, because the inferential arrow is not pointed in that direction. For many questions, I think psychology has more to offer neuroscience than the other way around. I am excited to be a part of this endeavor and use my knowledge of both domains to try and build a stronger and more fruitful bridge between them.
At the end of the day, neuroscience is going to move forward whether psychologists want to come along or not. And just as they have in many ‘big data’ domains, engineers working in neuroscience have already started asking questions about psychological phenomena without psychologists’ input. It seems to me like psychologists not only should want representation at the neuroscience table but also recognize that psychology is needed to have a comprehensive understanding of the brain; the engineers will not be able to figure it out without it. I see the work of researchers in my subfield as attempting to fill that chair to some degree. I hope others join us in not only appreciating the beauty of the brain, but also in recognizing the extent to which psychologists’ understanding of the mind and behavior is an essential for contributing to our understanding the very organ that gives rise to them and the challenges therein.
* To be clear, I am not saying that brain imaging is better than psychophysiology for making inferences about psychological phenomena. On the contrary, psychophysiological measures are often clearer and less expensive than their brain imaging counterparts. However, if you care about inference about the neural systems themselves, psychophysiology often can’t say as much about that (with some exceptions, like pupillometry and locus coeruleus activity).
** Because we cannot directly measure most phenomena of interest, almost all psychological measures – including things as simple as reaction times – are technically reverse inferences. Moreover, reaction times involve engaging volitional actions in motor cortex via a cascade of spatiotemporal events in the rest of the brain that are critical for understanding and making appropriate responses for the task at hand. There are many cogs in the machine, and in psychology there is no such thing as a free inference.