Early impediments and first steps

Resistance to connecting politics to the brain has deep historical and intellectual roots. Aristotle ¹¹ analogized the human mind to a blank writing tablet but linked the type of soul that an entity has to its biology (i.e., humans have a rational soul, animals have a sensing soul, and plants have a vegetative soul). During the Renaissance, René Descartes’ ^{Reference Descartes12} philosophy separated the mind from the body. Just after Descartes, John Locke’s ^{Reference Locke13} expansion of Aristotle’s analogy into a full-fledged argument about human nature shifted investigations in the social sciences for centuries away from embodied forms of cognition. ^{Reference Pinker14} Locke’s role in foreclosing a potential function for biology in human nature is ironic given that he was earlier part of an important group investigating the anatomy and architecture of the brain. ^{Reference Zimmer15}

While Sigmund Freud ^{Reference Freud16} was similarly trained in neurology, even his early work emphasized functions of the mind over structures of the brain. Both psychoanalysis and later reactions to it such as behaviorism ^{Reference Skinner17} directed focus away from biological explanations of psychological phenomena. For the psychoanalyst, experiences of the mind and their interpretations were central considerations. For the behaviorist, observations about responses to stimuli were the primary interest, regardless of which biological species happened to be doing the responding. Any potential interplay of biological factors in either of these dominant paradigms was put decidedly aside.

The behaviorist emphasis in the early to mid-twentieth century heavily influenced political science’s agenda, and approaches to political questions were often focused on observable behaviors such as voting or expressions of attitudes rather than exploring a potential role for the brain or biology. ^{Reference Campbell, Converse, Miller and Stokes18} As rational choice theory migrated from economics to politics, ^{Reference Downs19} yet another impediment was laid for understanding how brains might influence political decision-making. Anthony Downs’s five-part definition of rationality, for example, gives no role to the potential emotional, biological, or neurological factors that might bend our political will.

Examinations of a potential connection between biology and politics were made particularly unpalatable for many during the mid-1900s given the role of biology in Nazis’ rhetoric justifying their atrocities. ^{Reference Jackson and Weidman20} However, by the 1970s, some researchers had started using biological methods to investigate political questions. John Wahlke and Milton Lodge ^{Reference Wahlke and Lodge21} employed physiological measures of arousal to investigate political attitudes. Lindon Eaves and Hans Eysenck ^{Reference Eaves and Eysenck22} showed that attitudes demonstrated evidence of biological heredity. And, in what may be the first study of neuropolitics, Roger Sperry and colleagues ^{Reference Sperry, Zaidel and Zaidel23} investigated attitudes toward political figures in both hemispheres of split-brain patients.

In most human brains, the corpus callosum is a bundle of white-matter fibers that connect the left and right hemispheres and enable communication between them. In cases of severe epilepsy, this connection may be severed in order to reduce the symptoms of the disease, resulting in a split brain. Sperry and his colleagues presented a range of stimuli, including photos of Adolf Hitler and Richard Nixon, to only one eye at a time in order to evaluate whether both hemispheres of the brain were capable of reporting attitudes toward these political stimuli. While this type of work garnered a Nobel Prize for Sperry, it is important for neuropolitics as a reminder of the potential gains from using clinical cases and lesion studies in digging into our assumptions about political cognition. Much of our early knowledge about how the brain and mind function together came from studies of patients who had experienced damage to one region or another. While other tools have emerged and will continue to develop, the opportunities to untangle insights from clinical cases remain important.

fMRI and the foundations of cognitive neuroscience

A seminal paper by Patricia Churchland and Terry Sejnowski ^{Reference Churchland and Sejnowski24} included a figure illustrating the tools available at the time that enabled the study of neuroscientific phenomena. The figure shows the temporal ( $x$ -axis) and spatial scales ( $y$ -axis) that those tools covered. Brain lesion studies, for instance, sit in the upper-right-hand side, covering a temporal scale of days to years and a spatial scale of millimeters to centimeters. While the illustration maps the spatial and temporal scales covered by the contemporary methods available in the 1980s, the figure looks like a puzzle with a critical missing piece. There was simply no tool available to study the brain at the temporal scale from seconds to hours and the spatial scale from cubic millimeters to the whole brain. Despite this gap, the authors argued that there was a developing convergence that would undermine the conventional independence of neuroscience and cognitive science. In the following years, the new field of cognitive neuroscience emerged, but progress was dramatically accelerated as the new technique of functional magnetic resonance imaging arose, coming in like the perfect puzzle piece to fill in the hole that had been mapped out in the Churchland and Sejnowski paper.

Magnetic resonance imaging (MRI) devices create extremely powerful fields that provoke distinct patterns of magnetic flux around the protons that make up the hydrogen in the water within our body. Small differences in the polarization of these protons then enable the imaging of structures and differentiation of tissues within the human body. ^{Reference Bandettini25} While MRI creates static images of structure akin to a photograph, fMRI relies on a series of images in sequence, like the frames of a film, that allow researchers to detect change in cerebral blood flow over time.

Oxygenated blood is pumped to feed the brain, and it has a slightly different magnetic signature when compared with the blood that has been deoxygenated as the brain’s neural tissue absorbs this needed fuel from the supplied blood. The blood-oxygen-level-dependent (BOLD) magnetic signal varies with shifts in the level of oxygen in the blood and changes from image to image as the brain functions and different neural regions require different levels of metabolic activity. By the late 1990s, neuroscience research, such as studies by Paul Broca, ^{Reference Broca26} had shown that certain cognitive functions appeared to be at least partially localized to particular brain regions.

The first wave of fMRI studies suggested that there were changes in metabolic activity in specific brain areas as participants engaged in specific cognitive tasks. Thus, discrete mental functions could be associated with discrete brain structures. The effect on the new field of cognitive neuroscience was dramatic. As Peter Bandettini ^{Reference Bandettini25} and others have shown, dozens, then hundreds, and then thousands of papers were soon being published using fMRI each year.

The temporal resolution of seconds to hours meant that many types of well-developed cognitive science and psychological studies could be replicated while participants were undergoing brain scanning. Although fMRI could not obtain the millisecond precision of contemporary electroencephalogram (EEG) techniques, it benefited from a spatial resolution of millimeters and could provide data for the full depth of the human brain rather than merely inferences from the surface. Additionally, fMRI could be easily and ethically performed with healthy participants rather than requiring the injection of radioactive agents, as in positron emission tomography (PET), or relying on severe brain damage, as lesion studies had. The combination of temporal resolution from seconds to hours, spatial resolution of millimeters across the whole brain, and the very low invasiveness and risks of the method provided an exciting and powerful new tool that filled the gap highlighted by Churchland and Sejnowski. ^{Reference Churchland and Sejnowski24}

As fMRI was developing, I first encountered this research in an article by functional imaging pioneer Marcus Raichle ^{Reference Raichle27} in Scientific American. The initial studies focused on replicating existing cognitive and perceptual studies and reinforcing existing theories of functional localization. A typical study involved having participants perform a new task while being imaged with fMRI and then rescanning them after they had practiced the task extensively. Such practice related paradigms had been employed in measures of cognition all the way back to the early work of F. C. Donders, ^{Reference Donders and Koster28} who measured the amount of time that tasks took in order to infer their cognitive complexity and demands. In a happy confirmation, brain imaging during language-related tasks, such as conjugating verbs in an unfamiliar language, involved the same region that Broca had identified in his lesion studies of people with language difficulties. Results across a wide range of tasks showed that additional oxygenated blood flowed into a region believed to be critical for that task when the task was relatively novel. With practice, the amount of time required for the task diminished and the blood flow to the involved region also decreased. Expertise appeared to demand less mental effort even as performance continued to improve.

My interest in brain imaging was spurred on as I read the work of Philip Converse, ^{Reference Converse and Apter1} John Zaller, ^{Reference Zaller29} and others for the first time as a graduate student at the University of California, Los Angeles (UCLA) in the early 2000s. The repeated conclusion in this literature was that people who knew a lot about national politics thought differently than those who had not attended to it much. To me, this sounded similar to the practice related effects that Raichle had been writing about, and I suggested brain imaging as a potential tool for investigating political cognition. Although he was initially skeptical, Zaller later suggested that I could test at least my broad hypothesis using Donders’s methodology with the first available American National Election Studies dataset to include response latency measures that came out in 1998. While the available data weakly supported the idea of quantifying cognitive differences between political novices and sophisticates using response latency, ^{Reference Schreiber10} a much richer dataset was generously provided by Robert Huckfeldt, ^{Reference Huckfeldt, Levine, Morgan and Sprague30} which bolstered both the central theory and the potential for using fMRI.

Converse’s ^{Reference Converse and Apter1} claims that novices made political judgments with the cognitive equivalent of mental coin flips implied little intellectual effort and thus would suggest very short response times when using Donders’ paradigm. It turned out, however, that people who did not know much about politics took dramatically longer to answer political questions than those who were knowledgeable, ^{Reference Schreiber31} consistent with the extensive literature on practice and rehearsal. ^{Reference Palmeri32} The result functioned as a proof of concept and was sufficient to get me in the door at UCLA’s new Brain Mapping Division.

Traditional cognitive tasks dominated the initial work using functional brain imaging. For instance, early review pieces by Cabeza and Nyberg ^{Reference Cabeza and Nyberg33,Reference Cabeza and Nyberg34} highlighted studies investigating attention, Stroop tasks, language processing, memory encoding and retrieval, and skill learning. Social cognition was nearly entirely absent. The view emerging out of this work would not have led one to believe that man is, by nature, a political animal — but rather that we are roughly similar to the lab rats and pigeons so frequently investigated!

From cognitive neuroscience to social cognitive neuroscience

Studies of racial perception and moral judgment were among the first social neuroimaging papers to gain attention. Elizabeth Phelps and colleagues ^{Reference Phelps, O’Connor and Cunningham35} found that activity in the amygdala correlated with measures of implicit racial bias. Alexandra Golby and colleagues ^{Reference Golby, Gabrieli, Chiao and Eberhardt36} showed that patterns of activity in memory and face-related regions of the brain mirrored difficulties that whites had in recalling black faces. And Joshua Greene and colleagues ^{Reference Greene, Sommerville, Nystrom, Darley and Cohen37} showed that differences in how people judge two seemingly equivalent moral problems were reflected in differences in brain activity. Soon, enough work was coalescing around social observations and brain imaging to enable a new field of social cognitive neuroscience to emerge onto the scene as a full-fledged entity. ^{Reference Ochsner and Lieberman38}

The first international Social Cognitive Neuroscience Conference was held in 2001, bringing together a Tower of Babel-like collection of senior academics from a range of disciplines who struggled for meaningful dialogue. A Google search for “social cognitive neuroscience” would yield 53 hits that year. ^{Reference Brooks39} While that initial conference was characterized by senior scholars from fields such as primatology, economics, and neuroscience with vastly different disciplinary views of terms as basic as trust, it was just a few years later that New York Times columnist David Brooks ^{Reference Brooks39} described the meeting as “The Young and the Neuro.” After receiving the nod from senior scholars, younger scholars quickly converged on a common language and were, for example, running neuroscience studies of monkeys playing the behavioral economist’s games, a dramatic example of E. O. Wilson’s ^{Reference Elshtain40} consilience as multiple fields converged on shared theories and methodologies. This consilience manifested in the founding of the journals Social Cognitive and Affective Neuroscience and Social Neuroscience.

In those early days, the literature of neuroscience experiments broadly relevant to politics was extremely limited, which led to difficulty with experimental design and data analysis. My original fMRI studies, the first to combine functional brain imaging and political science, were run in 2001. At the time, the work was conducted on a 3-tesla (T) (a household refrigerator magnet is 0.005T) MRI scanner that had been specially modified to enable functional imaging, and the data analysis took a 64-processor supercomputer that filled an entire floor three days to converge. By the end of the decade, fMRI scanners were readily accessible at many universities around the world, and that same data analysis could be run in three days on a consumer iMac.

Social cognitive neuroscience advanced rapidly, with noticeable differences in flavor and approach between American and European research. ^{Reference Lieberman41} Europeans developed work on the theory of the mind, mirror neurons, and empathy. ^{Reference Singer42} In contrast, American researchers tended to be more individually focused with studies of self-knowledge, emotion regulation, and implicit attitudes.

Soon enough groundwork had been laid for the journal Political Psychology to commission a special issue on the intersections of neuroscience and political science. From this early work to the present, discussions of neuropolitics have focused on both promise and peril as repeated themes. While the measurement millennium prophesied by Converse in the opening quote had arrived, substantial and warranted criticism of fMRI’s ability to measure subjective states persisted. As Raichle noted in a commentary to the special issue, the “tasks of interest are always associated with activity changes in large networks of brain areas with task-unique spatial and temporal properties.” ^{Reference Raichle43} Even in the best circumstances, converting data about task-related changes in brain activity into meaningful psychological insights, much less political processes, would continue to be a challenge.

Neuropolitics: Initial findings

The special issue of Political Psychology in 2003 staked out a number of central themes that continue to be at the heart of investigations in neuropolitics today. John Cacioppo and Penny Visser ^{Reference Cacioppo and Visser44} highlighted the difficulty of transcending many levels of analysis. Elizabeth Phelps and Laura Thomas ^{Reference Phelps and Thomas45} cautioned against the tendency to view results from neuroimaging as somehow more definitive than traditional psychological methods. And Matthew Lieberman and colleagues ^{Reference Lieberman, Schreiber and Ochsner46} illustrated how well-established theories of brain function have implications for political phenomena and sketched out how cognitive neuroscience could be used to both generate theories of political behavior and test those theories. As reviews and critiques of neuropolitics have appeared in the literature in the years since, these early words about the area’s promises and perils continued to be echoed.

As neuropolitics transitioned from theoretical work to an experimental agenda, three themes in experimental design quickly emerged. One trend was exploring political attitudes with tasks in which participants would report political attitudes in a manner roughly similar to the long tradition of answering survey questions in political science research. A second theme in research was for investigators to look for variations in neural activity as participants viewed faces. A third theme involved the study of mechanisms of basic cognition and perception to identify distinctions between liberals and conservatives.

My own dissertation reflected the first two tendencies, with experiments asking participants to report their political attitudes and others asking them to view faces or visual scenes. The first experiment showed the effect of experience regarding national politics on the cognitive and neural mechanisms involved in political attitudes, in particular the brain’s default mode network ^{Reference Schreiber47} (a network of brain regions that is highly active when a person is at rest). Drew Westen and colleagues ^{Reference Westen, Blagov, Harenski, Kilts and Hamann48} likewise showed the same brain regions activated as politically knowledgeable participants answered political questions. Westen ^{Reference Westen49} extrapolated from these findings in his well-received popular book, The Political Brain.

Kristine Knutson and colleagues ^{Reference Knutson, Wood, Spampinato and Grafman50} investigated reactions to well-known political faces using the implicit association test and fMRI. Michael Spezio ^{Reference Spezio, Rangel and Alvarez51} and his collaborators looked at how we could use fMRI to understand the role of negative attributions toward losing candidates and the pattern of neural activation that corresponds with those negative perceptions. Jonas Kaplan and colleagues ^{Reference Kaplan, Freedman and Iacoboni52} also studied reactions to political faces, finding activity in regions often engaged in cognitive control when participants viewed candidates from the opposing party.

However, the first significant controversy arose soon after, as some of the same team members described results from an fMRI study they had performed using faces of candidates in the upcoming 2008 presidential race in a New York Times op-ed. ^{Reference Iacoboni, Freedman and Kaplan53} An immediate response signed by a number of prominent cognitive neuroscientists and psychologists criticized the inferences made from the data and its lack of peer review. ^{Reference Aron, Badre and Brett54} As I discuss later, it might possible to infer the subjective mental states of individuals as they react to political candidates, but doing so requires a complex set of conditions that were not met in the research described in the op-ed.

While reactions to political questions or the faces of political candidates have prompted a number of interesting neuropolitics studies, some of the most fascinating findings have emerged when nonpolitical tasks have been used to differentiate individuals on opposite sides of the political spectrum. Although political scientists have a long tradition of measuring responses to political stimuli based on information processing paradigms of political decision-making, ^{Reference Converse and Apter1,Reference Campbell, Converse, Miller and Stokes18,Reference Zaller29,Reference Axelrod55} looking for fundamental differences in how the brain processes the world was motivated by theoretical insights coming from psychology and cognitive neuroscience. ^{Reference Lieberman, Schreiber and Ochsner46} David Amodio and colleagues, ^{Reference Amodio, Jost, Master and Yee56} for instance, found differences in how the brains of liberals and conservatives reacted to a version of the go/no-go task.

My colleagues and I were able to accurately differentiate liberals and conservatives using brain activity in response to a gambling task, even though the observable gambling behavior was identical. ^{Reference Schreiber, Fonzo and Simmons57} A series of studies coming from the University of Nebraska’s political science lab likewise demonstrated first that there were physiological differences in reactions to disgusting stimuli ^{Reference Oxley, Smith and Alford58} and later that the differences in brain reactivity were large enough that ideology could be correctly classified based on brain data after a single disgusting image. ^{Reference Ahn, Kishida and Gu59} Academy Award–winning actor Colin Firth was a coauthor of one of the studies in this vein after commissioning research on the political brain. This work showed that liberals and conservatives could be reliably identified by differences in the size of brain regions implicated in the other studies. ^{Reference Kanai, Feilden, Firth and Rees60}

The converging results in even these first few neuropolitics papers are noteworthy but perhaps not surprising. Regions in the default mode network reliably appeared throughout the first studies in social cognitive neuroscience as individuals performed tasks involving the contemplation of themselves or their roles in social dynamics. ^{Reference Lieberman61} Regions such as the amygdala, basal ganglia, and insula were investigated precisely because they had been implicated in a range of decision-making tasks that researchers believed might be relevant to political cognition. ^{Reference Spezio, Adolphs, Crigler, MacKuen, Marcus and Neuman62} Nonetheless, it was heartening to see coherent threads emerging as the low-hanging fruit was sampled from the tree of neuropolitics knowledge.

Neuropolitics and its neighbors

Although the first wave of neuropolitics research produced consistent findings, this line of inquiry has not led to the intensity of publications seen in the intersection of neuroscience with neighboring disciplines. Neuroeconomics has produced hundreds of papers in top journals and an entire academic society (the Society for Neuroeconomics) with annual meetings. Neurolaw has similarly generated a cottage industry of law review articles, books and textbooks, a number of conferences, and large funding commitments from the MacArthur Foundation. ^{Reference Jones, Marois, Farah and Greely63} Why does neuropolitics appear to lag in development?

Lawyers and judges have had to deal with cases related to brain function since ancient times. Issues of mental capacity and intent can be seen as influencing legal reasoning back to the days of stone tablets and the Code of Hammurabi. As cognitive neuroscience provided new tools and insights, legal theorists immediately grasped the implications for both broader debates and specific cases. While more than 1,000 neurolaw articles have been published exploring these implications, ^{Reference Jones, Marois, Farah and Greely63} only a small portion have actually involved experimental research. ^{Reference Aharoni, Vincent and Harenski64} Thus, the nature of the discourse in neurolaw has centered on theory building rather than theory testing.

Neuroeconomics, in contrast, was able to connect quickly to the foundations laid by behavioral and experimental economics. Behavioral games such as the dictator and ultimatum game fit nicely within the constraints of fMRI experimental design. ^{Reference Krueger, Grafman and McCabe65} The same economic task can be performed repeatedly inside the scanner with variations in partners, incentives, or costs. While many of these games have clear implications for debates in political science, ^{Reference Dawes, Loewen and Schreiber66} political scientists and political psychologists often have been interested in experiments with a greater emphasis on naturalistic realism ^{Reference Ansolabehere and Iyengar67} or connections to the current political context.

Because functional brain imaging typically relies on repeated measures of the same task condition, successful neuropolitics studies must involve stimuli that are realistic enough to be politically meaningful for the question at hand but also novel enough that each repeated instance of the task is engaging for the participant. Given severe constraints and significant variation among citizens in their knowledge of national politics, ^{Reference Delli Carpini and Keeter68} experimental design is even more complicated when the requirements of functional imaging are added to the mix. As a consequence, many successful neuropolitics experiments have involved variations on tasks that have already been well tested in more traditional psychological contexts.

Neuroeconomics has also benefited from a convergence in research agendas. While many economists were questioning the axioms of rational choice theory, psychologists and cognitive neuroscientists were pushing their understanding of basic decision-making processes. Thus, the levels of analysis of primary concern to these researchers were relatively proximate. The interdisciplinary collaboration and funding necessary became routinized as coordinated efforts such as the Society for Neuroeconomics expanded.

Political science and neuroscience have had a much smaller space of intersection. Questions at the heart of a lot of social cognitive neuroscience are adjacent to issues that are of central to political science, but a critical mass of crossover has yet to be achieved. Political science and political psychology have often focused on phenomena at a national or international scale, while much of social cognitive neuroscience has focused on dyadic or at most small group dynamics.

An interesting and productive counterexample is studies of racial attitudes. Some of the earliest work in social cognitive neuroscience focused on race, ^{Reference Phelps, O’Connor and Cunningham35} with a sizeable literature emerging in its wake ^{Reference Amodio69} that had exciting implications for political science. Racial attitudes have been an enticing area of inquiry for these approaches precisely because central concerns about implicit attitudes or pressures for social conformity converged with methodological strengths and theoretical foundations for understanding how automatic mental processes operated. The dynamic and highly contextual nature of racial attitudes ^{Reference Schreiber and Iacoboni70} along with the well-developed and creative experimental designs in this area suggest a ready source for inspiration that prospective researchers in neuropolitics should be investigating as we look for workable frameworks that can be adapted for addressing central questions in political science.

Fortunately, as research in adjacent disciplines expands, so does the potential for fruitful crossover with neuropolitics. As a recent review by Sekoul Krastev and colleagues ^{Reference Krastev, McGuire and McNeney71} notes, relatively few neuropolitics experiments have leveraged the reward paradigm that is at the center of much research in neuroeconomics. Economic theory’s emphasis on individual utility and, in particular, on monetary payoffs contrasts with political values that are often distal from the individual making the decision. Justice, equality, patriotism, and other central political values can heavily influence discourse and individual decision-making, and it is an open question whether expressions of these values rely on the same neural foundations found in neuroeconomics. As the points of contact with other related fields increase, new developments in methodologies and research paradigms will enhance the potential for cross-pollinated research in neuropolitics.

While neuropolitics has not yet had the influence seen in related disciplines, the potential identified early ^{Reference Tingley72} remains intact. Insights have been gained in our understanding of prejudice, motivated reasoning, attitudes, and ideology. ^{Reference Jost, Nam, Amodio and Van Bavel2,Reference Mendez73} Early speculation on the distinct roles of automatic and controlled processing in political cognition ^{Reference Lieberman, Schreiber and Ochsner46} have been borne out as research has advanced. ^{Reference Tusche, Kahnt, Wisniewski and Haynes74} Moreover, promising leads about the interplay of policy goals and political identity are emerging. ^{Reference Jenke and Huettel75}

While discussions about the intersections between biology and politics have been controversial historically, mind-sets have begun to shift. Findings about the biological heritability of political attitudes that were largely ignored in an earlier era ^{Reference Eaves and Eysenck22} have become some of the most cited works ^{Reference Alford, Funk and Hibbing76} in contemporary political science decades later. A number of researchers have investigated evolutionary roots of political behavior. ^{Reference Hibbing, Smith and Alford77,Reference Haase and Starling-Alves78} And while no one academic society devoted solely to neuropolitics has emerged, this work can far more easily find outlets in a variety of professional associations, divisions, interest groups, and scholarly journals that cater to work at the intersection of biology and politics more broadly. ^{Reference Peterson and Somit79} Thus, while few political scientists would believe that their field is really a subdiscipline within biology, ^{Reference Rosenberg80} many are far more open to the potential value for understanding biology and the brain in the context of politics than was previously the case.

The future of neuropolitics

Although fMRI has enabled an explosion of research in cognitive neuroscience and related disciplines over the years, it has some rather severe limitations, especially in relation to neuropolitics. For one, equipment costs run into the millions of dollars and require a significant staff with technical expertise to maintain the delicate apparatus and run the machine. Participants in fMRI experiments must lie very still at the center of a large superconducting magnet. As a consequence, the possibilities for the actual social interactions that underpin politics are severely constrained, and the types of stimuli and tasks that may be used are quite limited. The result is an experimental set-up far from the naturalistic environments that can boost external validity in political psychology research. ^{Reference Ansolabehere and Iyengar67} Additionally, while fMRI filled the methodological gap identified in the late 1980s, it is still restricted in terms of temporal resolution and causal inference due to fundamental characteristics of the method. The sibling disciplines to neuropolitics are taking advantage of alternative approaches to fMRI in order to sidestep its limitations and neuropolitics will need to follow suit as it develops.

Investigating disruptions in brain activity

While a variety of imaging methods all seek to make inferences about mental function by looking at brain activity, another approach is to investigate disruptions of activity. The famous case of Phineas Gage, ^{Reference Macmillan89} who suffered trauma to the left frontal lobe and experienced significant personality changes, and the aphasic patients of Broca, ^{Reference Broca26} who suffered damage to the region named after him, provided foundational insights into cognitive processes through disorders in brain function. In my dissertation, ^{Reference Schreiber47} I proposed using patients suffering from anterograde amnesia to test hypotheses from Milton Lodge and others ^{Reference Lodge, McGraw and Stroh90} regarding “online” impressions of candidates. Jason Coronel and colleagues ^{Reference Coronel, Duff and Warren91,Reference Coronel92} later carried out this research, showing that amnesic patients who are unable to recall previous information about the political issue positions of candidates can nonetheless have their attitudes about the candidates altered by the same information they cannot recall having seen.

Just as they have been vital for neuroscience since its earliest days, patient populations can provide powerful opportunities for future neuropolitics research. The tragedy of brain-injured Vietnam veterans has led to decades of important insights, with studies of veterans who sustained damage to their prefrontal cortex adding to prior work. Findings have provided important insights concerning theory of mind, goal-directed social behavior, moral judgment, emotional intelligence, stereotypes, and a wide range of social cognition. ^{Reference Raymont, Salazar, Krueger and Grafman93} Since the development of the Vietnam head injury study, a number of projects have gathered brain injury patients and benefited from their participation in a wide range of studies. Researchers in neuropolitics should benefit from collaborations with neurologists and neuropsychiatrists who work with patient populations and are interested in understanding the political implications of brain injury. ^{Reference Mendez73,Reference Arciniegas and Anderson94}

Beyond brain injury, a wide range of patient populations can yield insights for questions at the heart of political science. Individuals with borderline personality disorder (BPD), for instance, have a stronger tendency to idealize or devalue. As a consequence, such individuals have a more difficult time restoring reciprocity in an iterated trust game, ^{Reference King-Casas, Sharp, Lomax-Bream, Lohrenz, Fonagy and Montague95} have stronger reactivity in the medial prefrontal cortex to experiencing social exclusion, ^{Reference Ruocco, Medaglia, Tinker and Ayaz96} and possess stronger negative stereotypes regarding social out-groups. ^{Reference Dunbar97} With half of all Americans experiencing an episode of mental illness during their lifetime, ^{Reference Kessler, Berglund, Demler, Jin, Merikangas and Walters98} understanding the implications for political attitudes and political behaviors is critical. By investigating the particular ways that specific disorders alter our political psychology we can enhance our understanding of the mental and emotional processes that underlie our political world. Additionally, benefits can obtain for the patient populations, as the iterated trust game appeared strongly diagnostic for BPD, a notoriously difficult condition to diagnose. ^{Reference King-Casas, Sharp, Lomax-Bream, Lohrenz, Fonagy and Montague95}

Transcranial magnetic stimulation (TMS) is a technique that can temporarily disrupt the function of a specific brain region through the introduction of a focused magnetic field. In this way, TMS mimics the effect of a lesion, but only for a short time and without lasting effects. The approach is particularly useful for testing hypotheses about the specific contribution of a brain region to a broader mental process. For instance, Liane Young and colleagues ^{Reference Young, Camprodon, Hauser, Pascual-Leone and Saxe99} used TMS to disrupt the right temporoparietal junction (RTJP) in order to show that undermining the function of that area also impaired participants’ ability to take account of mental states when making moral judgments. Although RTJP had been implicated previously in functional imaging studies looking at our ability to infer the mental states of others, showing that disruptions to the region also impair that ability strengthens the case. As with other types of lesion studies, TMS enables a much stronger set of inferences about causation than is typically available with functional imaging approaches. However, TMS is limited to the regions of the brain closest to the scalp, thus deeper brain areas are inaccessible.

An alternative to TMS is transcranial direct current stimulation (tDCS), which uses a relatively simple device to create weak electrical currents that modify neural excitability and activity. ^{Reference Nitsche, Cohen and Wassermann100} The method has a lot in common with TMS in terms of its potential and limitations. However, the technology is much simpler and inexpensive, and thus it can be used affordably on multiple participants at the same time. For example, in one study, researchers were able to use tDCS simultaneously on six participants who were playing ultimatum games with a series of 12 partners. ^{Reference Knoch, Nitsche, Fischbacher, Eisenegger, Pascual-Leone and Fehr101} Compared with a placebo condition, those who had their right dorsolateral prefrontal cortex activity disrupted by tDCS were more likely to accept unfair offers. Note, however, that some researchers have questioned the replicability of tDCS as a measure. ^{Reference Horvath, Forte and Carter102}

One important limit common to TMS and tDCS is that deep brain structures are out of reach. However, a new technique looks to transcend this problem without the need for neurosurgery. By applying a high-frequency oscillating electrical field at multiple locations on the scalp, it is possible to drive activity in deep neural tissue without disrupting cortical structures. Although it currently has been demonstrated only with a mouse model, researchers expect to apply it to humans. ^{Reference Grossman, Bono and Dedic103} As a consequence, neural activity in the amygdala or hippocampus could soon be altered temporarily in neuropolitical studies to help isolate the role of these deep structures in a variety of political experiences, processes, and behaviors.

Moving forward, neuropolitics will also want to take advantage of developmental approaches to political psychology. Although historically important work has been done on the development of political attitudes and behaviors in children, ^{Reference Hess and Torney-Purta104,Reference Jennings and Niemi105} this area of inquiry has been significantly undervalued in political science in recent years, ^{Reference Sapiro106} especially compared with the broader work in contemporary developmental psychology ^{Reference Kagan107} and developmental social neuroscience. ^{Reference Baron-Cohen, Tager-Flusberg and Lombardo108} For instance, while the biological heritability of political attitudes is quite stable across the adult lifespan, early political attitudes are entirely dominated by the effects of socialization. ^{Reference Hatemi, Funk and Medland109} The evidence from genopolitics then leads to contradictory expectations about what might be occurring in the development of children’s political brains.

Another important direction for the future will involve combinations of the previously described approaches. Francesca Happé and Uta Frith ^{Reference Happé and Frith110} have argued for deeper theorizing about atypical social cognition from a developmental neuroscience view. For example, the fact that racial stereotyping, but not gender stereotyping, is impaired in children with Williams syndrome ^{Reference Santos, Meyer-Lindenberg and Deruelle111} provides further confirmation of an evolutionary psychology account of the origins of racial discrimination as built on mechanisms for coalitional cognition, ^{Reference Kurzban, Tooby and Cosmides112} which are impaired in Williams patients.

Because of the lower costs of some of the methods mentioned, it is also possible to use them simultaneously and with multiple participants. Simultaneous imaging of multiple participants is known as hyperscanning; when combined with the freedom associated with running subjects outside of an fMRI scanner, it can lead to far more naturalistic experiments. In a study of social ranking behaviors, pairs of participants were examined using fNIRS, EEG/ERP, and response times, ^{Reference Balconi and Vanutelli113} enabling multiple measures of mental processes. Other studies have used functional imaging approaches with both neurotypical and clinical populations (such as BPD) to understand how patients with these conditions compare with typical individuals in a range of social cognition tasks. ^{Reference King-Casas, Sharp, Lomax-Bream, Lohrenz, Fonagy and Montague95,Reference Ruocco, Medaglia, Tinker and Ayaz96} As neuropolitics research develops and becomes more sophisticated, presenting converging lines of evidence across methodologies will be critical for fostering confidence that findings are not mere artifacts of the experimental procedure or lab situation.

Finally, while fMRI devices with a magnetic field strength of 3T are now relatively common, more powerful machines have come on the scene in recent years. Ultra-high-field devices, such as 7T fMRI scanners, are still relatively new, but initial studies have demonstrated benefits relevant to the future of neuropolitics. In particular, regions such as the amygdala have been difficult to image given the spatial resolution of typical 3T fMRI studies and other constraints, but with 7T fMRI significantly smaller voxels (the 3D version of pixels) can be patterned, enabling far greater precision given the spatial and temporal trade-offs. ^{Reference van der Zwaag, Da Costa, Zürcher, Adams and Hadjikhani114} The greater precision and sensitivity of 7T has led to an increasing number of research centers using these devices. However, the time needed to acquire each slice means that imaging the whole brain can provide a challenge ^{Reference Vu, Jamison and Glasser115} and additional artifacts from the imaging process must be managed. ^{Reference Beisteiner, Robinson, Wurnig and Hilbert116} As researchers continue to develop proficiency with the nuances of 7T imaging, many of the hurdles are being overcome and the new technology is yielding promising results. ^{Reference Yoo, John, Farquharson, Cleary and Wong117}

Conceptualizing experiments: ROI and whole brain analysis

While the new measurement tools bode well for a future of neuropolitics research in which the costs are much lower and the experimental conditions far more naturalistic, an important fundamental shift is coming in terms of new ways of conceptualizing experiments. Two particular advances involve the use of machine learning for data analysis and the paradigm of neuroforecasting. Even while utilizing familiar equipment, these techniques provide qualitatively different ways of viewing the enterprise of neuropolitics.

The traditional approach to brain imaging focuses on two strategies for data analysis. The first, region of interest (ROI) analysis hypothesizes that a specific brain region is involved in the cognitive task of interest and tests that hypothesis by extracting data from that region to compare activity in task and control conditions. This approach derives from the localization hypothesis that is at the core of much neuroscience work, namely, that there are discrete regions of the brain that are engaged in specific mental activities. By measuring the activity in a particular region, we can then infer the engagement of a specific mental process. The ROI approach is most powerful when used in a context in which there are strong reasons to believe that a particular region is involved in a particular cognitive task. One problem is that researchers are often unclear which mental tasks or brain regions might be involved in the phenomena of interest. By limiting investigation to a specific region, one might miss critically important activity in other areas.

This limitation can be circumvented by a second approach, referred to as whole brain analysis. In both ROI and whole brain analysis, data might be collected for the entire brain at once; the difference is in the analysis of the data collected. With whole brain analysis, the strategy is to run statistical comparisons for each of the brain’s thousands of voxels, then correct for many simultaneous analyses and set a statistical threshold for the reliability of the clusters of activity that result. In this way, researchers can discover unexpected engagement with the task regardless of where in the brain it occurs. One downside is that this type of analysis often leads to “see what lights up” experimental designs that are driven by overly vague theory that “the brain is probably doing something different when you do this task of interest than when you do not.” The resulting activity might then be subject to post-hoc interpretations that get reframed in writing up the hypothesis test.

In some cases, an ROI analysis is carried out after the whole brain analysis in a manner that can trigger inference problems, and in particular, puzzlingly high correlations, similar to the issue of selecting on the dependent variable (choosing cases based on a criterion and then using the cases as evidence of the outcome you’re selecting on) familiar in political science. ^{Reference Vul, Harris, Winkielman and Pashler118} While some papers are overzealous in their analyses of unexpected results from whole brain analyses, there can also be a problem of sweeping a puzzling finding under the rug or sticking it in supplemental materials without further mention (see the discussion in Schreiber and Iacoboni ^{Reference Schreiber and Iacoboni70} ).

A more fundamental issue arises from the conceptual framework of localization. While there is considerable evidence of the connection between particular brain regions and specific mental processes, the brain is also extremely interconnected and mental processes are often distributed. Thus, looking at one voxel, one anatomical region, or one activation cluster can give a false impression of how much, and what parts, of the brain exactly are engaged during the processing of a task. Rather than focusing on specific regions, clusters, or networks within the brain, machine-learning approaches often leverage all of the data from a whole brain scan in order to model the phenomenon of interest.

Discussion

The use of neuroscientific methods to ascertain the political attitudes and behaviors of individuals and large portions of the population, however, should give us pause. For one, the broader implications of these rapidly materializing techniques need to be developed. Ideological purity has often been demanded by both authoritarian regimes and (on occasion) democracies attempting to root out perceived enemies from within. How would the purges of Joseph Stalin, Mao Zedong, or even Joseph McCarthy have differed if they had had the tools of neuropolitics at hand? Will the selections of party nominees or Supreme Court justices someday be influenced by brain scans? As distasteful as these possibilities may sound, how long will it take until the neuromarketing techniques that are currently influencing other sectors of society come into full play in the political domain? Trends would indicate not that long.

Already, some are engaged in practical neuropolitics, but the ethical implications and the threats to democratic deliberation are woefully underappreciated — and undertheorized. Days after a recent New York Times article highlighted the use of neuropolitics methods to sway a Mexican political campaign, ^{Reference Randall135} the governing party vowed to stop using the approach. ^{Reference Malkin and Randall136} The founder of one of the firms involved claimed, “We cannot read minds and of course we cannot change them.” However, the developing work in neuropolitics suggests that these aims are increasingly within reach. The reporting intimates that campaigns around the world have already started using these tools, including for presidential elections in the United States.

Neuropolitics, a field studying the most complex thing known to humans that likely evolved because of the demands created by exactly the context we want to understand it operating in, is still in its infancy. ^{Reference Schreiber, Crigler, MacKuen, Marcus and Neuman4} An analogy from biology illuminates where the field currently stands. The $C$ . elegans worm is a model species with 302 neurons in its tiny brain that have been laboriously mapped out. ^{Reference Varshney, Chen, Paniagua, Hall and Chklovskii137} But even then, we are far from being able to fully understand the function of this worm’s brain, much less the role its brain plays in its very simple social life. ^{Reference Wade138} The human brain, in contrast, has around 86 billion neurons and 100 trillion synapses. It is the result of a 3-million year long cognitive arms race within the species to compete for survival while paying the ever-increasing metabolic costs to feed this device that converts glucose into love and hate, poetry and war. As our understanding of the brain is developing, old notions that our brains are somehow hardwired to perform in a certain predestined manner are giving way to a realization that they are wired to be flexible precisely because of the pressures of being a political animal. In the long run, the field has the potential to fundamentally change the way we understand human nature. But we are just taking the first steps in that journey.