Opportunities in Neuroscience for Future Army Applications (2009)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

3 Training and Learning This chapter discusses opportunities to expand on training in specific paradigm responses increase trainees’ and improve the Army’s current behavioral approaches to agility in responding to new threats, or does training in training and learning by incorporating recent or emerging probabilistic assessment make them more adaptable? Two advances in neuroscience. The discussion is organized under examples illustrate how each methodology produces ben- five application areas: efits in specific cases. From a behavioral perspective, over­ training reduces time to response by reducing consideration • Evaluating the efficiency of training regimes and of alternatives in favor of an anticipated favorable outcome learning paradigms, to the trained-in response (Geva et al., 1996). By contrast, a • Individual capability and response to training, medical doctor is trained to diagnose any given set of clinical • Monitoring and predicting changes in individual presentations (e.g., presenting symptoms, past individual and performance efficiency, family medical history) in terms of a probabilistic etiology • Soldier selection and assessment, and based on environment, current illnesses in the community, • Monitoring and predicting social and group and other factors. This differential diagnosis strategy helps interactions. focus possible treatment modes on likelihood of outcomes. When sufficient evidence accrues, the most likely cause may The final section summarizes applications of neuro­ not be the most frequent cause. science to Army training and learning in terms of when practical application can be expected: 5 years (near term), Neuroscience-Based Models of Learning 10 years (medium term), or 20 years (far term). Enhancing the utility and predictive power of traditional behavioral and Over the past decade, enormous progress has been made psychological methods by incorporating the insights and toward describing the neurological basis for learning skills tools of neurophysiological monitoring in these and other and procedures. We now have fairly complete models that application areas—but focused on understanding how indi- describe how the brain learns the values of actions (see, for vidual choices are made in nonmilitary contexts—is a princi- example, Niv and Montague, 2009) and uses these values pal goal of neuroeconomics. Results from neuroeconomics to guide future decisions, a process often called reinforce- research are cited frequently in this chapter and Chapter 4. ment learning. Suppose a subject is offered the opportunity to search in one of two locations for a reward on each of hundreds of sequential trials. With repeated sampling, the Evaluating the Efficiency of Training subject learns the relative values of the two locations and Regimes and Learning Paradigms shapes his behavior to maximize his reward. Although ­human Neuroscience offers new ways to assess how well cur- subjects show idiosyncratic behavior under this training rent training paradigms and accepted assumptions about regime, the learning models are now well enough developed learning achieve their objectives. For example, does over- that an individual’s behavior can be well characterized by a single parameter. Given the value of this parameter for an Neuroeconomics is characterized in a current survey of the field as the individual, his future choices can be predicted with accura- convergence of normative models of choice—the province of ­economics— cies approaching 90 percent (Corrado et al., in press). These with the psychological and neurobiological processes (or algorithms) by learning models thus both describe the behavior observed in which individuals (animal and/or human) make choices (Glimcher, in press). humans and animals and predict how a subject, once char- 23 

24 OPPORTUNITIES IN NEUROSCIENCE FOR FUTURE ARMY APPLICATIONS acterized, will behave in dynamic environments (Balleine may be possible to develop specialized interventions aimed et al., 2009). at improving performance in soldiers. Similarly, tremendous advances have been made in under­standing movement generation, including development Subject Populations for Army-Specific Studies of of skills, habits, and automatic performance (Poldrack et Learning and Training al., 2005; Yin and Knowlton, 2006). Moreover, consider an individual who is learning to execute a complex movement A perennial issue for behavioral and neurological testing accurately—for example, tracking a moving target with his is the degree to which experimental findings from a specific finger (Poulton, 1974). We understand quite precisely how sample can be extrapolated to a target population. Many the incentives and feedback provided shape both perfor- of the activities in which soldiers and their leaders engage mance and learning (Newell, 1996; Trommershäuser et al., depend critically on rigorous preparatory training and task- 2009). We also have evidence that the fastest way to train a specific expertise. The majority of behavioral research is movement is not to provide strictly accurate feedback. The performed with subjects who are either patients in a clinical speed and effectiveness of such training can often be maxi- setting or volunteers from a university community (mostly mized by providing feedback regimens that take advantage undergraduate students). A critical question is how far results of this inherent accommodation to variability (Schmidt and based on these study populations of convenience transfer to Lee, 2005; Kording et al., 2007). a soldier population. In cases where the research hypothesis In short, advanced models of reinforcement learning addresses Army-relevant issues directly, the typical research and movement control learning have implications for both subject populations may not be sufficiently representative of training and prediction of learning efficiency. We know in the population to which the Army wants the results to apply. principle how to develop optimal training regimes under In short, neither clinical patients nor university undergradu- many conditions, and we know how to predict agent behav- ates are good surrogates for a soldier. ior with precision under a range of conditions. Both assets As the Army seeks to apply research results from the could be leveraged to improve not only training but also data neuroscientific literature, the extent to which they can be presentation (for situational awareness) and prediction of transferred to a military (specifically, Army) population threat/enemy behavior. should itself be a subject of Army research. In particular, how The combination of neuroimaging tools, cognitive far do results from typical civilian samples represent those neuro­science, and experimental or cognitive psychology has to be expected from an Army population? Although some resulted in the development of models of how the brain may of the research reviewed by the committee has used actual process information. For example, recent accounts of brain soldiers or cadets, for the most part the human subjects in processing that occurs in the dorsolateral prefrontal cortex potentially relevant studies do not compare well to the soldier (PFC) are based on interpreting cognitive control as altering population in cardiovascular fitness, psychological drive to ongoing behaviors in order to adjust to the changing context perform, and learning/training experiences that clearly affect of the environment (Botvinick et al., 2001). The resulting neurobehavioral response—e.g., boot camp, intense training computational models reveal which aspects of cognitive for operational performance, and actual operations. performance are altered as information changes—a field One alternative to constraining Army-usable results of study called “computational neuroscience”—and can be to just the few studies that use soldiers (or even military used to predict performance on behavioral tasks (Brown and cadets) is to seek subject populations that more closely Braver, 2007). resemble Army soldiers in such key characteristics as Computational neuroscience uses mathematical models cardio­vascular fitness, psychological motivation to perform, to study how neural systems represent and transmit informa- and training/learning in immersive, demanding environ- tion. The discipline may be roughly divided into two schools. ments. High-­performance athletes are one such subpopula- The first school uses detailed biophysical models of individual tion, and there is an extensive literature of behavioral and neurons, detailed models of neuronal networks, and artificial neuro­psychological research on them. Appendix C lists a neural network models to study emergent behaviors of ­neural sampling of the research literature from 2001 through 2007 systems. The second school develops signal-­processing on training methods for high-performance athletes: perfor- algorithms, computational models, to analyze the growing mance evaluation/­assessment of athletes in training, includ- volumes of data collected in neuroscience experiments. In ing under stress; social interaction with other athletes; and these computational models, adaptation to new informa- i ­ ssues with performance anxiety and other psychological tion is represented as changes in one or more parameters. issues including depression in ex-­athletes (references 1-109 By combining the models of both schools with information in Appendix C). Several studies have investigated the use from neuroimaging tools (sometimes called “systems neuro­ of mental imagery in training athletes and its effects on science”) and behavioral neuroscience, the possible causes of performance (references 110-123). Performance after mild underperformance and the conditions conducive to improved concussions and determining when the subject can return to performance can be quantitatively constrained. Eventually, it a normal (strenuous) routine is a hot topic for both athletes 

TRAINING AND LEARNING 25 and soldiers, and some longer-term studies investigating individual brain activity may not be epiphenomenal in the the effects on athletes of multiple concussions have also sense of simply being a developmental outcome that does been published (references 124-143). Studies of training not causally influence individual behavior. Rather, individual and performance issues for female athletes go well beyond patterns of brain activity may reflect (or underlie) the unique the well-researched Female Athlete Triad (references 144- characteristics of individual minds, and they may capture 151). The relationship between athletes’ risk-taking behavior aspects of an individual mind that cannot be obtained using and athletic performance is the topic studied in references conventional behavioral measures or self-reporting. The 152-157. Other relevant topics in this literature include the better we understand the sources of individual variability in relationship between the lifestyle of athletes and changes brain activity through systematic experimentation and analy- in immune function (references 158-160), and the effect of sis, the more fully we can determine whether, and to what music therapy on athletic performance (reference 161) and extent, these individual differences can be used to assess and of caffeine (reference 162). ultimately train that person. While it is important to consider how well the subject populations in the research studies match the population of Individual Variability in Brain Activity interest (here, Army soldiers), the context of the research must also be considered. For example, in testing of futuristic Efforts by neurologists and neuropsychologists to under­ military decision support systems, military subject-matter stand individual differences in brain processes go back to experts perform worse than novice populations, owing to at least the mid-1800s, when the neurologist Paul Broca cultural biases (Graham et al., 2007). When designing a concluded, from examining the common area of damage futuristic system, it may therefore be better to test it using across a group of patients exhibiting similar speech produc- a subject population likely to possess the required skills tion deficits, that speech production could be localized to the (includ­ing constructive attitudes toward novel representa- third convolution of the left inferior frontal gyrus. Around tions), which current military personnel may lack. the same time, the neurologist John Hughlings Jackson argued against a centralized region for speech, basing his opinion on his observations of wide variations in the extent Individual Capability and Response and location of damage in patients exhibiting similar prob- to Training lems in speech perception and wide variations in symptoms Given the increasing technology- and threat-driven in patients with similar damage. More than a century later, dynamic complexity on the battlefield, it has become more Broca’s view of brain organization became the dominant and more critical to optimize the capability of individual paradigm. An example is the Wernicke-Geschwind model, soldiers. One solution to the great differences from one which was developed to explain language function. Today, individual to the next in human capability and expertise is however, the Broca paradigm is often disregarded, largely to tailor human–system interfaces to human capabilities and because of enormous individual variability in the underlying to adapt training regimes to the individual. Although classic brain processes. This variability is evidenced by the fact that experimental psychologists tended to downplay individual a growing number of neurosurgeons painstakingly map out differences in their theories of human cognition, educators individual brains just prior to surgery, after a portion of the have consistently reminded psychologists of the need to patient’s skull has been removed but while the patient is still understand individual differences in cognition and perfor- conscious and responsive. mance (Mayer, 2003). A similar need exists in understanding Researchers using neuroimaging to understand the how the neural systems underlying cognition differ among r ­ elationship between the brain and the mind have recently individuals (Posner and Rothbart, 2005). Understanding e ­ ncountered a similar paradigm shift. Most neuroimaging the neural substrates of individual differences in cogni- studies localize cognitive functions in the brain by conduct- tion can help in characterizing the differences, developing ing a statistical analysis across a group of subjects; this training methods tailored to them, matching individuals to analysis identifies common areas of activation. While this assignments for which they are well suited, and optimizing can be a useful approach to understanding the modular orga- h ­ uman–machine and individual–system interfaces. nization of the brain, it disregards the not-common areas of Recent advances in neuroimaging make this endeavor activation that can be observed at the individual level and that possible (Miller et al., 2002; Miller and Van Horn, 2007). may also be critical for that function in a given individual. Patterns of brain activity as measured by functional magnetic Recent studies have shown that the individual patterns of resonance imaging (fMRI) appear to provide unique identi- brain activity during a memory task are enormously variable, fying characteristics. These are unlike fingerprints, because sometimes with areas of activation that do not even overlap between two subjects (Miller et al., 2002; Miller and Van This is a condition of three syndromes common in high-performing fe- Horn, 2007). Furthermore, such studies have found that indi­ male athletes of all ages, though especially in their teen years, and includes vidual variations in brain activity could not be attributed to disordered eating, amenorrhea, and osteoporosis. An athlete can experience one, two, or all three syndromes in the triad. random noise because the pattern for an individual is stable 

26 OPPORTUNITIES IN NEUROSCIENCE FOR FUTURE ARMY APPLICATIONS over time. Understanding these individual differences in and utilization of information, provides a useful model for brain activity has not only significant theoretical implications understanding how individual differences in cognitive style but also pragmatic implications for the Army in trying to and strategy may affect patterns of activations (Tulving, 2002; under­stand how individuals respond to and remember events Squire et al., 2004). A hypothesis supported by patient studies in battlefield and nonbattlefield situations and in character- and animal models is that the hippocampus is not involved in izing the mental traits of individual soldiers and officers. the permanent storage of information per se but rather serves Individual variability in brain activity could come from to facilitate consolidation of a distributed cortical memory one or two things that are not mutually exclusive. One pos- trace. A principal characteristic of this distributed network is sibility is that individual differences in brain anatomy and that it allows the rapid and flexible formation of multimodal physiology lead to extensive differences in brain activity memories. In this emerging picture of brain activity, episodic (for example, as measured by the BOLD signal in fMRI) retrieval makes use of several distinct brain regions, some of despite the tremendous efforts undertaken to normalize which may be involved in the cognitive processing that is brain spatial variability to a standard spatial representation. peripheral to the actual retrieval of stored information. The Structural differences in brain anatomy and physiology take same individual may, at particular times, differentially engage many forms. It is known, for instance, that the size and shape those brain regions based on the context and strategy of the of individual brains can vary greatly, and genetic markers moment. One potential implication of this architecture is that asso­ciated with this variability have been found (Tisserand the same behavioral outcome—such as an “old” ­ response et al., 2004). There are also well-documented individual on a recognition test—could be based on distinct sets of differences in the tissue structure of specific regions of the information and distinct combinations of ­ neural circuits cerebral hemispheres (cytoarchitectonics) and in the orien- in two different subjects. The recent development of new tation and location of specific fissures and gyri (Rajkowska neuroimaging techniques that enable the systematic study of and ­Goldman-Rakic, 1995). Theoretically, these regions are individual differences in brain activity can be used to improve spatially normalized in fMRI analyses using sophisticated understanding of the optimal brain activity that under­lies, for a ­ lgorithms so that brain regions are consistent across sub- example, optimal performance on the battlefield or in other jects, but some differences have been shown to exist even demanding environments and task contexts. after extensive spatial normalization. Other structural differ- To improve the assessment of individual soldiers, ences may also exist, despite the fact that brain development r ­ esearch programs should systematically investigate the is generally universal (Kosik, 2003; Rakic, 2005). structural and cognitive factors that may account for the How anatomical differences affect cognition is not well extensive individual variability that has been observed in understood, but there are intriguing possibilities. For example, brain activity across normal subjects using fMRI. Three basic the size and location of the planum temporale is thought to questions need to be answered: affect language processing (Hutsler et al., 1998). Other recent studies have linked individual differences in white-matter • Why do individuals’ patterns of brain activity differ connectivity to individual differences in cognition (Baird so much from each other? et al., 2005; Ben-Shachar et al., 2007). Understanding the • Can individual differences in brain activity be extent to which structural differences account for variability a ­ ccounted for by differences in anatomy or physiol- among individuals in brain activation will greatly enhance our ogy? Can information about individual brain activity knowledge and understanding of individual minds, allowing be indicative of limitations and constraints in the human–system interfaces to be better tailored. kinds of cognitive strategies and skills that a particu- The second possible source of variability in brain activ- lar individual is capable of or tends to engage in? ity is individual differences in cognitive styles, abilities, and • Can individual differences in brain activity be strategies. These differences, which appear important to how a ­ ccounted for by differences in cognitive style and individuals perform many cognitive tasks, often can be cor- strategy? Can information about individual brain related with significant differences in BOLD activity, despite activity be used to assess the thought processes of procedural efforts to constrain and control the psychological individuals engaged in a variety of tasks? state when test subjects are performing the same experimental task. For example, a recent study found that brain activations Identifying Conceptual Change in Individual Learning underlying a standard memory task were extremely variable from subject to subject. These differences extended well The potential utility for the Army of monitoring indi­ beyond spatial normalization or relatively small differences vidual differences in brain activity to track and evaluate in the location of brain structures (Miller et al., 2002). A sig- individual learning can be illustrated by recent work in cor- nificant portion of this variability correlates with differences relating differences in brain activity with success in assimi- in retrieval strategies (Donovan et al., 2007). lating a basic concept of physics that is nonintuitive. Teachers Episodic memory, which relies on an extensive and researchers have found that beginning physics students h ­ ippocampal-cortical network for the consolidation, storage, retain a naive “impetus” theory of motion that differs from 

TRAINING AND LEARNING 27 the fundamental concepts of force and motion central to the body and mind, which in response initiate complex cogni- Newtonian physics. Many physics courses may be needed tive and affective coping strategies. Research on expedition before the Newtonian concepts become a student’s “natural” members, soldiers, elite athletes, and competitors in extreme way of seeing the world (Clement, 1982; McCloskey, 1983; athletic events provides substantial evidence that exposure to Mestre, 1991). extreme situations profoundly affects performance. Different A number of studies have examined the degree to which environmental stressors can place different demands on indi­ major changes (through either education or experience) in viduals or groups exposed to them. For example, exposure the way people view the world can be captured using fMRI. to extreme cold during an Antarctic expedition may result in Recently, fMRI has been used to compare the brain activity social deprivation, whereas exposure to combat may result patterns of students who had taken no high school or col- in affective overload. However, beyond these differences in lege physics with the patterns of students who had taken at response that correspond to differences in the environmental least five college-level physics courses (Dunbar et al., 2007). stressors, individual cognitive and affective responses to the Students were shown similar movies of two balls falling at same stressor vary just as widely. either different rates or at the same rate. The students were Neuroscience offers some distinct advantages over asked whether the movie they viewed was consistent or not behavioral assessment or self-reporting for assessing and with their expectations. The researchers were particularly even predicting how an individual’s baseline performance interested in movies where balls of different sizes (and there- is affected prior to, during, and following exposure to a par- fore different apparent mass) either fell at different rates (as ticular environmental stressor. One key limitation of standard expected in an “impetus physics” view) or fell at the same self-reports and observer reports is their limited ability to rate (as expected in a “Newtonian physics” view). The fMRI predict future behaviors. Moreover, a number of studies have data showed that a particular brain region associated with shown that individuals do not always report their current error detection, the anterior cingulate cortex, was activated psychological, mental, or emotional status accurately (Zak, in the nonphysics students when they saw the balls of dif- 2004). Although there is still no single neural measurement ferent sizes drop at the same rate, while the same region was tool that can unequivocally replace self-reporting, the cur- activated in physics students when they saw the balls drop at rent tools can depict an individual’s status more fully, as a different rates. Conversely, neither group showed the char- complement to, rather than a replacement for, self-reporting acteristic error detection activity when the movie they saw and behavioral assessments. Moreover, recent insights ­enable fit the expectations of their engrained physics concepts. The researchers to parse cognitive and emotional processes into researchers concluded that the physics students had indeed more basic modules such as attention, working memory, fully assimilated the Newtonian concepts. In the context of cognitive control, and others. These modules can be assessed Army applications of neuroscience, this example illustrates e ­ fficiently and quantitatively by linking behavioral para- how fMRI could be used as an indicator of whether soldiers digms to measurements made with electroencephalography, have learned and assimilated key concepts fully enough to fMRI, or other imaging modalities. act on them instinctively. Finally, behavioral tasks that have been developed r ­ ecently for use with these imaging modalities are parametric in the sense that the imaging results can be used to quantify Monitoring and Predicting changes in the degree to which performance is altered. Quantifying Individual Performance Efficiency the degree of underperformance (the performance deficit This section begins by outlining some conditions in relative to the individual’s baseline) is crucial to designing which neuroscience approaches may be useful in assessing and administering countermeasures. (Chapter 5 discusses in changes in the performance efficiency of individual soldiers more detail some countermeasures to stressors that degrade that are due to the stresses of extreme environments and com- soldiers’ cognitive performance.) bat operations. Advantages and disadvantages of potential A major challenge for neuroimaging used in this way neuroscience approaches are illustrated with a few examples. is to determine its sensitivity and specificity for monitor- Finally, short-, medium-, and long-term opportunities for ing performance in extreme environments. Thus far, most Army R&D on these approaches are discussed. As discussed in Chapter 2, sensitivity and specificity have rigorous definitions that apply here. Thus, “sensitivity” measures the proportion of The Effects of Environmental Stressors on the actual positive cases that a test identifies as positive. Mathematically, it Individual Performance is the ratio of true positive test results to the sum of the true positive tests and the false negative tests (false negatives should have been positive). Extreme environments, including but not limited to “Specificity” measures the proportion of actual negative cases that the test combat environments, are characterized by the high demand identifies as negative. It is the ratio of true negative test results to the sum they place on physiological, affective, cognitive, and social of the true negative tests and the false positive tests (false positives should processing in the individuals exposed to them. In short, the have been negative). The aim of having a test with both high sensitivity and high specificity is to identify all the positive cases efficiently while also stressors present in extreme environments strongly perturb distinguishing positive from negative cases. 

28 OPPORTUNITIeS IN NeUROSCIeNCe fOR fUTURe ARMY APPLICATIONS imaging studies have revealed intriguing results on a group- tion of individuals whose performance in future missions averaged level. But a key goal of a neuroscience-based is at high risk of deteriorating, as well as the identification approach to quantifying behavior and performance is greatly of design factors for military equipment. The example in improved ability to predict the future behavior of individuals Box 3-1 illustrates the potential of neuroscience methods to (individual-specific prediction). It is not enough to predict make such predictions. the distribution in performance of a group (or parameters of that distribution such as the average performance or a soldier selecTioN aNd assessmeNT “normal range”). The goal is to predict which individuals will perform well and which will not, and even why individuals Currently the Army selects most of its individual sol- perform as they do. Recent results from neuroimaging diers using two basic tests, both administered by the Depart- studies indicate that such predictions are possible. Most ment of Defense: the Armed Services Vocational Aptitude imaging studies have demonstrated large effect sizes, which Battery (ASVAB), for selecting enlisted personnel, and the supports the idea that differences among individuals and at Scholastic Aptitude Test (SAT), for selecting officers. These different times for the same individual may be large enough tests are used for assessing both an individual’s fitness for to be meaningfully measured. the Army and his or her suitability for specific assignments. The next step is to determine if neuroimaging or other Neither test assesses personality traits or neuropsychological neuroscience approaches that combine sensitivity and high traits of applicants. specificity can be used to generate quantitative predictions of There has been little attention to how the Army could individual behavior in response to environmental stressors. select, from a general pool of applicants, those individuals To be more useful as a predictor of performance than cur- who, by inherent or acquired ability, would add value to a rent approaches, neural monitoring methods such as neuro- particular unit. Even for assignments that require expensive imaging need to track performance states closely, including and specialized training, the Army knows little about a considering whether or not the individual is self-reporting a candidate’s neuropsychological traits before that individual change in performance and whether or not the self-reporting starts training. Little is known about how to identify soldiers is as accurate as objective measures of performance. To be whose individual traits would enhance the performance of useful for identifying (and eventually predicting) perfor- a unit to which they could be assigned. Indeed, for many mance deficits that are due to extreme environments, the high-performance assignments, one cannot state with cer- neural monitoring methods must do more than distinguish tainty which psychological or behavioral traits correlate with poor-performing individuals from normal performers. They superior performance. For example, the training for an attack must also consistently distinguish altered activity from nor- helicopter pilot costs about $225,000, but candidates for this mal activity in specific brain structures of those individuals training are selected today largely on their expressed interest who subjectively report performing at normal levels but in becoming an attack helicopter pilot rather than on their whose performance has deteriorated by objective measures. ability or fitness for this high-value assignment. The last-mentioned capability will enable the identifica- At present, the Army has low washout rates even for BoX 3-1 Predicting Future Behavior in extreme environments Imaging techniques could be used to detect individuals who are at high risk for experiencing deterioration of performance on future Army missions. Paulus et al. (2005) used fMRI to scan the brains of 40 methamphetamine-addicted men who had been sober for 3 to 4 weeks. This imaging technique can map brain regions involved in specific mental activities. Scans were performed while the men were involved in a decision-making task to identify brain regions stimulated by the task. Approximately 1 year later, the researchers correlated the fMRI results with subsequent drug abuse in the 18 men who relapsed and the 22 who remained abstinent (drug-free). In the scans made at the beginning of the study, the scientists observed low activation patterns in the brains of some of the men in structures that are known to participate in making decisions. These regions were the right middle frontal gyrus, the right middle temporal gyrus, and the posterior cingulate cortex. Lower activity in these structures correlated with early relapse to methamphetamine use. The scans also showed that reduced activa- tion in the insula and the dorsolateral prefrontal, parietal, and temporal cortices correlated with drug relapse in 94 percent of cases. By comparison, significant activation in these same regions correlated with nonrelapse in the 86 percent of the men who remained abstinent after 1 year. Thus, the initial brain scans provided an indicator of which individuals were at greatest risk of relapsing and which were at least risk. Furthermore, these differences in activation pattern show both specificity and sensitivity as a predictive indicator for risk of relapse. 

TRAINING AND LEARNING 29 difficult jobs because trainees are free to repeat the train- Advanced Soldier Selection Tools ing sequence until they are able to meet minimum perfor- For many high-value soldier assignments, the ASVAB mance requirements. Less than 5 percent of helicopter pilot may only weakly predict how long it will take to train a trainees, for example, wash out of training. This approach recruit, how likely the recruit is to complete a tour of duty, obviously increases training costs. It may also reduce unit or even whether he or she will be effective in the assign- performance; a soldier who requires three times longer than ment. The Army recognizes the poor predictive power of the average to achieve minimal standards of proficiency during ASVAB and has commissioned ARI to construct secondary flight training may or may not be a good helicopter pilot. At selection tests for a number of high-value specialties that the moment, however, the Army has no validated means of call for a substantial investment in training. There are now measuring how this approach affects unit performance. specialized selection tests for helicopter pilots, improvised explosive device (IED) detection experts, Special Forces, Current Enlisted Soldier Selection drill sergeants, and recruiters. In general, ARI has developed these specialized selection tests as psychological tests. The ASVAB, which is administered to roughly 500,000 By developing and using these tests as selection tests— applicants each year, is a computer-administered multiple- even if only to a limited extent—the Army has in fact imple- choice test with two components: the Armed Forces Qualifi- mented a screening process for those interested in certain cation Test (AFQT) and five subject area tests. The goals of high-value assignments. For example, the helicopter pilot the AFQT are to predict the likelihood of service completion test allows some recruits to become helicopter pilots and by the applicant and to predict overall performance. The excludes others from the required training. The question, AFQT has four parts: word knowledge, paragraph compre- then, is not whether applicants should be screened, at least hension, arithmetic reasoning, and mathematics knowledge. for demanding and high-value positions, but rather how well AFQT scores are percentile rankings that have been normed the screening instruments do their job. To date there appears to the ASVAB applicant pool and are correlated with the to have been little effort to ensure that these tests have both likelihood that the applicant will complete a 2-year tour of specificity and sensitivity with respect to predicting actual duty. The higher the score, the more likely the applicant is performance. to complete that tour. Interestingly, the AFQT prediction of To illustrate the state of practice in Army selection test- tour completion is most accurate for applicants who hold a ing and highlight the challenges where neuroscience-based traditional high school diploma. The AFQT score is a poor approaches can help, the discussion will focus on the Army’s predictor of tour completion for applicants who did not current test for helicopter pilots, the Alternate Flight Aptitude graduate from high school or who have a graduate equiva- Selection Test (AFAST) and the test proposed as its replace- lency diploma. ment. After taking the ASVAB, Army recruits who want to To address the limitations of the AFQT, the Army train as helicopter pilots are invited to take the AFAST, which R ­ esearch Institute (ARI) has developed an alternative test is a computer-based test. The Army uses the AFAST score for candidates who did not complete high school. This to assess the suitability of the recruit for flight training. Two “noncognitive” test, called the Assessment of Individual factors limit the effectiveness of the AFAST as a screening Motivation (AIM), was designed to assess conscientiousness, tool. First, the security of the AFAST has been compromised; stress tolerance, and openness to new experiences. The AIM answers to the test can be found on the Internet. Second, in combination with a measure of body mass index is called according to the briefing ARI gave to the committee, the the Tier Two Attrition Screen and is now in limited use as a AFAST has almost no predictive ability. complement to the ASVAB. In recognition of the limitations of the AFAST, ARI was The second part of the ASVAB consists of five tests of tasked with developing an alternative, the Selection Instru- the applicant’s factual knowledge in general science, auto- ment for Flight Training (SIFT). As an instrument for soldier mobile and shop information, mechanical comprehension, selection, the SIFT model is based more closely than AFAST electronics information, and assembling objects. The scores on the selection techniques used in private industry, although on these tests by soldiers who have successfully completed it is still limited to conventional behavioral testing. The SIFT a tour of duty in a particular area of specialization are used is a secure test that measures, among other things, a number as a benchmark or score profile that is associated with suc- of personality features, perceptual speed and accuracy, and cessful tour completion in that area of specialization. The flexible intelligence. It has almost triple the predictive accu­ score profiles are used by recruiters and recruits as guides racy of the AFAST for helicopter pilot selection, based on to a recruit’s likely job performance when the recruiter and subsequent pilot performance and aptitude evaluations by recruit together select an area of specialization. Essentially, peers and instructors. The SIFT score has not yet been cor- the score profiles from the five subject tests are derived by related with in-theater performance of helicopter pilots who simple linear regression. Their actual predictive power is low compared with that of best practices in the vocational Lawrence Katz, research psychologist, Army Research Institute, briefing assessment and training community. to the committee on April 28, 2008. 

30 OPPORTUNITIES IN NEUROSCIENCE FOR FUTURE ARMY APPLICATIONS took it prior to training. Unfortunately, there are no definite standard neuropsychological tests packaged for easy analysis plans to gather the data to establish this correlation, a final and administration. A tester first selects a number of test sets assessment that is critical for such a tool. from the battery. The test sets might, for example, be selected Despite its advantages over the AFAST, the SIFT to gain a baseline assessment of the neuropsychological never­the­less lacks any neuropsychological measurements, functioning of an individual or to allow the state-specific nor does it look at neurochemical or genetic indicators of assessment of neuropsychological function. For example, the neuro­psychological traits. ARI does not have expertise in ANAM can be used to assess mental function after a concus- these neuroscience-based measurements. Enhancing the sion, after a period of sleep deprivation, or after exposure to SIFT by giving it the ability to identify established neuro­ pharmacological agents. ANAM test sets can also be selected psychological correlates of the personality and cognitive to obtain a clinical assessment of medical disorders such as traits characteristic of highly successful helicopter pilots Parkinson’s disease or Alzheimer’s disease. would greatly improve its value as a screening instrument. The high incidence of IEDs on the contemporary For example, research has found that individuals whose battlefield has heightened the interest of the Army’s mission hypothalamo-pituitary axes are highly reactive to stress are commanders in ANAM. A number of commanders have unlikely to complete Navy SEAL training (Taylor et al., instituted predeployment baseline testing for all their troops 2006, 2007). However, such information is not yet used by with a subset of ANAM well-suited for identifying neural the Army in its recruit assignment process. function changes caused by traumatic brain injury. Such Even after advanced selection instruments have been testing ­ allows for comparing an individual’s neurological developed, implementing them throughout the Army will be functioning before and after exposures that increase the risk a challenge. Recruiting stations throughout the United States of traumatic brain injury, such as the blast from an IED. The must be able to administer specialized selection tests like ANAM support center estimates that over 50,000 troops were the SIFT, whose implementation cost has been estimated at screened before deployment in 2008. roughly $200,000. Although this implementation cost is less than the cost of training a single helicopter pilot, ARI has not Summary: Status of Soldier Selection and Assessment yet succeeded in getting the SIFT implemented generally in and the Potential for Neuroscience-Based Improvements the recruitment process. A similar project is under way to improve the test instru­ The ASVAB, which the Army currently uses to assess ment used for identifying trainees in explosive ordnance the fitness of most candidate recruits for the Army and disposal. This project, which is conducted jointly by ARI and their suitability for specific assignments, does not assess the Joint Improvised Explosive Device Defeat Organization, personality traits or neuropsychological traits. Even in the was not evaluated by the committee because it involves clas- case of high-value assignments for which the Army must sified information. make a substantial investment in training, trainees are free to repeat the training sequence until they can meet minimum performance requirements, and the Army knows little about Neuropsychological Testing in the Army: a candidate’s neuropsychological traits before the training The Automated Neuropsychological Assessment Metrics begins. The only Army testing instrument that contains sig- To address limitations in the ASVAB, the Army has nificant elements of neuropsychology and cognitive neuro­ d ­ eveloped a complementary test that attempts to assess a few science is the Automated Neuropsychological Assessment basic behavioral traits considered valuable for completing a Metrics (ANAM). ANAM evolved from a series of neu- tour of duty, such as conscientiousness, stress tolerance, and rological assessment tools initially developed during the openness to new experiences. For a number of high-value Vietnam war. The goal was to develop behavior-based tests specialties, the Army uses screening tests, but the specific- for determining whether the cognitive capability of a soldier ity and sensitivity of these tests have not been evaluated as had been impaired by exposure to a chemical agent and for predictive of performance in the specialties for which they testing protective agents for side effects. A group of such are used. Even when a screening test that incorporates con- tests was consolidated by a multiservice group into what ventional behavior-based factors—for example, the SIFT for later become the ANAM. In its current form, ANAM is a helicopter pilot training—has been developed to replace an computerized neuropsychological battery of 30 test sets, older test with much less predictive power, the Army has not some of which have not been normed. ANAM support and incorporated the improved test in its recruitment process. development are currently under the direction of the Center The problem is that even improved tests like the SIFT do for the Study of Human Operator Performance at the Uni- not measure the neuropsychological traits that neuro­science versity of Oklahoma. research has been identifying. Based on the progress seen An ANAM assessment ranges from simple tests of when neuropsychological indicators are applied in voca­ r ­ eaction time to dual-task interference tests, which are useful tional settings, adding fairly simple neuropsychological for assessing executive function. The test sets are essentially testing to the current mix of soldier assessment techniques 

TRAINING AND LEARNING 31 appears likely to improve soldier selection immensely, activity: electrocardiography, electromyography, galvanic particularly for high-value assignments where the training skin ­ response, eye tracking, and genotyping. Still other investment is high and individual performance is critical to methods used in social neuroscience include drug infusion the accomplishment of a unit’s mission. studies, comparisons with patients who have neurologi- The ANAM, which is currently deployed and being cal disorders or focal brain lesions, and comparison with administered to assess neural function, seems a natural animal models. starting point for expanding the Army’s selection of testing Although it is difficult to measure brain activity in a instruments. As more and more soldiers are screened with laboratory setting during natural social interactions, tech- ANAM before being deployed in combat theaters, the Army niques that approach a natural interaction have been devel- is building a large experience base that, if captured in an oped. For example, Montague developed a technique called analytically useful medium, could provide essential feedback “hyperscanning” to simultaneously measure brain activity for improving recruit selection processes. If the Army can using fMRI while two or more people interact (Montague et identify soldiers who excel in their areas of specialization, al., 2002). More simply, brain activity can be monitored in it could use their ANAM test data to identify characteristics one person while he or she interacts in real time with one or common to high performers but less likely to be found in more individuals whose brain activity is not measured (see, lower-performing individuals in that area of specialization. for example, Eisenberger et al., 2003). Many studies use a The ANAM data could thus be used to develop a set of tools computer-simulated “person” with whom the study subject for relating neuropsychological assessments to mission interacts without knowing the interaction is with a computer- performance. based simulation. This approach appears to activate the same brain regions associated with actual social behavior (see, for example, McCabe et al., 2001). Monitoring and Predicting Social and Group Interactions Relevance of Social Neuroscience to the Army The Scope of Social Neuroscience There are numerous psychosocial factors that contribute Most of a soldier’s actions involve other people, includ- to stress in the armed forces, including unpredictability of ing fellow soldiers, commanders who are giving orders, the danger, concern about resiliency and sustaining performance enemy, and noncombatants. Recent findings in social neuro- in combat, inability to control situations, separation from science can inform and improve Army training, tactics, and family, recovery from injury, death or injury of comrades, leadership for dealing with these social interactions. even the anticipation of returning to civilian life. As the Social neuroscience, also called social cognitive neuro- neurophysiological conditions that correlate with and per- science, investigates the neurophysiological basis for social haps underlie the behavioral and psychological (“mental behaviors. Topics examined are those traditionally studied in experience”) responses to these and other psychosocial social psychology, including attraction to others and attach­ factors are uncovered, it will become easier to understand ment, altruism, speech recognition, affiliation, attitudes both their detrimental (stressing) and beneficial (stimulat- toward other individuals and groups, empathy, identification ing) consequences. This section highlights several of many of others, kin recognition, cooperation and competition, self- applications where research using the methods of social regulation, sexuality, communication, dominance, persua- neuroscience is likely to produce results of value to Army sion, obedience, morality, contagion, nurturance, violence, training and learning. and person memory. Social neuroscience research integrates Using the methods of social neuroscience, tasks with information about a person’s physiological state, social con- structured, well-defined parameters of interaction can be text, experience, and cognition to understand his or her social used to gauge an individual’s commitment to group goals behaviors. Studies in social neuroscience draw heavily on while studying the brain processes associated with indi- findings from affective neuroscience, as many studies have vidual goals. When the group in question is an Army unit shown that social behaviors have a strong affective (emo- or a surrogate for one, this approach can provide an objec- tional) component (Ochsner and Lieberman, 2001). What tive measure of effective training and an individual’s likely social neuroscience adds is explicit attention to brain activity performance as a team member. By adding factors to the and neurophysiology when studying social behaviors. structured interaction—for example, time constraints on The techniques used to measure brain activity ­directly decisions, sleep deprivation, environmental stressors, or during experiments involving social tasks include mag- gunshots and other disturbances typical of combat—group netic resonance imaging generally and fMRI in particular, cohesion under stress and the neuropsychological correlates computerized tomography, electroencephalography and of the observed behaviors can be examined for insights into magneto­encephalography, positron emission ­ tomography, whether and why cooperation is sustained or fails. Adding transcranial magnetic stimulation, and event-related neuroscience methods and understanding to established p ­ otentials. Other techniques are used as surrogates for brain techniques for training and evaluating group interactions and 

2 OPPORTUNITIeS IN NeUROSCIeNCe fOR fUTURe ARMY APPLICATIONS unit performance can improve the diagnostic and predictive that can be linked to neural state, including biomarkers such value of current practices. as the small biomolecules involved in brain functioning that Behavioral measurements obtained through the use of are relatively easy to monitor. Recent research has shown that structured interactions, such as measurements based on game when an individual receives an intentional and tangible signal theory constructs, can predict how soldiers in a particular unit of trust from another individual, the brain releases one of these will interact and respond during combat as well how they will biomolecules, the neuropeptide oxytocin (OT) (Kosfeld et al., interact and respond within their unit or with other “friendly” 2005; Zak et al., 2005). OT has been shown to reduce anxiety groups. Behavioral measurements can be structured to (Heinrichs et al., 2003). It has been known for some time emphasize cooperation, competition, punishment, or a blend that OT is associated with bonding to offspring and spouses. of all three. Many structured interactions, or games, have Moderate stress also induces the release of OT, but fear and been formulated and studied in the context of behavioral eco- great stress inhibit it. By knowing this brain target, training nomics for small groups (two to four players). A challenge could be redesigned to induce OT release. for the Army will be to extend the results to larger groups in Another aspect of social neuroscience relevant to leader- contexts that reflect different points along the full spectrum ship training concerns the neurophysiological correlates that of warfare. The tasks can be modified to better relate to a are being identified for the cognitive constructs by which we field operation, while incorporating the basic structure of a interact with other persons as conscious agents. This attri- formal game whose behavioral and neurological aspects have bution of a “mind” to others is referred to as the theory of previously been established and confirmed. mind (ToM), whereby in order to understand another person Social neuroscience research on leadership is a nascent and respond to his or her behavior, an individual assumes effort that could inform the training of both officers and the the other person has a conscious mind directing his or her soldiers under their command. For example, characteristics observed behavior (see Box 3-2). of leaders under stress can be studied with the methods of A large number of studies have allowed brain processes social neuroscience as can the impact of a specific leader’s related to ToM attributions to be localized to specific cor- characteristics on the performance of those he or she leads. tical regions (Gallagher and Frith, 2003). Neurophysical Behavioral game theory uses insights from human behavior monitoring techniques can be used to determine whether to improve decisions (Camerer, 2003), and, as discussed these regions of the brain are activated during Army-relevant above, the neural correlates of effective decision strategies activities—for example, battle simulation training—to opti- are now being mapped. One direct application of this work mize soldier learning and retention. The brain’s ToM regions would be to teach leaders some of the findings from experi- also appear to be active during moral judgments (Young et ments in behavioral game theory in order to improve their al., 2007). Neural processes in these regions may affect deci- awareness of the factors that influence their own decision sions soldiers make when they apply the Uniform Code of making behavior as well as that of others. Understanding Military Justice to combatants and noncombatants. Thus, the the behavioral and psychological factors in how choices are techniques of social neuroscience can be adapted to monitor made, including the brain mechanisms that support effective trainee responses in unaccustomed or problematic situations, choices, can contribute to improving the performance of to improve training, and to test retention of leadership, group Army officers and the training regimens for soldiers. dynamics, and moral judgment skills. As described in Chapter 8, an important trend in long-term Social neuroscience methods can also enhance research is the continuing discovery of performance indicators Army-relevant research on how group dynamics affect BoX 3-2 Theory of mind ToM, a construct that was introduced by Premack and Woodruff (1978) to characterize the mental ability of higher apes, refers to the ability of individuals to attribute independent mental states to self and others in order to explain and predict behavior (Fletcher et al., 1995). This approach has been particularly important to characterize cognitive development in children (Frith and Frith, 2003) and dysfunction of cognitive development in autism and autism spectrum disorders (Happé et al., 1996) as well as in psychosis (Doody et al., 1998). More recently, several conceptual connections have been made between ToM and other important psychological constructs. For example, it appears that the neural representation of self (Happé, 2003) and, more generally, self-generated beliefs (Leslie et al., 2005) are closely related to the ability of ToM. This proposition is supported by some imaging studies that suggest the importance of the medial prefrontal cortex as part of both self-relevant as well as ToM-related processing (Wicker et al., 2003). Therefore, ToM is an important construct that can be used to examine one’s ability to infer mental states, related to self, and beliefs but more important is accessible to experimental modulation using neuroscience approaches. 

TRAINING AND LEARNING 33 an individual’s ability to process information and make of possible connections if every node is fully linked with choices based on that information. Most army operations every other node in the net, but information can pass rap- involve soldiers trained to operate as a team. However, in idly between any two nodes in the net.) This question also most studies—fMRI, positron emission tomography, and arises when attempting to maximize the efficiency of road m ­ agneto­encephalography—of neural information process- networks, and such research might reveal that in some cases ing the subjects are isolated from other persons, and the study adding a communications channel would actually impede is not designed to assess differences in performance related overall information flow. to the influence of other team members. With the increasing importance of network-centric Summary concepts in planning for the future Army, an important ­issue is the performance of individuals operating as a de facto Neuroscience techniques (neuroimaging, ­physiological team that results from being connected by communications measures, biochemical assays of brain function) can be links. Social neuroscience methods and insights can help to used to measure the training status of individual soldiers. answer two questions raised by these network-created team Specifically, these techniques can be used to determine situations: (1) functional status, (2) recovery time, or (3) level of training. The sensitivity, specificity, and accuracy of these • Do soldiers respond more efficiently to perceived a ­ pproaches are unknown and require further research. Thus, h ­ uman–human communication than to machine– a primary short-term goal for the Army should be to conduct h ­ uman communication? this research, which will allow for assessing the training • What is the most efficient size (number of members) status of soldiers. These efforts should last a relatively short needed by a particular team to respond to a specific time (<5 years). threat? There is emerging proof of principle that neuroscience techniques could be useful not just for short-term predictions Recent findings suggest that a person interacting with ­another about outcomes and behaviors but also for predicting the person over a communications link has a different pattern of performance, behavior, and potential of individuals over the brain area activation than a person interacting with a com- long term. It is not clear which targets would be most valu- puter. In the human–human situation just the dorsolateral able for the Army—for example, the best target population PFC is likely to be stimulated, whereas in machine–­human and what outcomes the Army should be most interested in. interaction the medial PFC also becomes activated (McCabe Close collaboration between neuroscience laboratories and et al., 2001). Army leadership would help to develop a common agenda Based on this work, neuroimaging experiments could that would benefit both communities. First, the Army would indicate whether military operators are receiving informa- improve its ability to predict performance and, second, the tion from a computer-based system or from another person. laboratories would obtain further proof of principle of the In these experiments, which would combine behavioral practicality of their methods. A collaborative effort could be monitoring and neuroimaging, key variables would be the viewed as a long-term initiative with high potential payoff nature (human or machine) and number of communication over the next 10 years. channels. One hypothesis is that the responses of team mem- bers may vary significantly if, under information overload, References they begin to attend in different ways to communications perceived to come from human counterparts and to com- Baird, A.A., M.K. Colvin, J.D. Van Horn, S. Inati, and M.S. Gazzaniga. 2005. Functional connectivity: Integrating behavioral, diffusion tensor munications perceived to come from computers, depending imaging, and functional magnetic resonance imaging data sets. Journal on the degree of confidence they have in the two sources. A of Cognitive Neuroscience 17(4): 687-693. competing hypothesis is that the team members will in fact Balleine, B.W., N.D. Daw, and J.P. O’Doherty. 2009. Multiple forms of value operate more efficiently if their sense of social interaction learning and the function of dopamine. Pp. 367-387 in ­Neuro­economics: with other humans increases their trust in the information D ­ ecision Making and the Brain. P.W. Glimcher, C.F. Camerer, E. Fehr, and R.A. Poldrack, eds. New York, N.Y.: Academic Press. received and encourages them to be concerned about the Ben-Shachar, M., R.F. Dougherty, and B.A. Wandell. 2007. White likelihood of low-probability, high-consequence events. m ­ atter pathways in reading. Current Opinion in Neurobiology 17(2): To explore optimal team size for these de facto teams, 258-270. experiments could be framed to examine how individuals Botvinick, M.M., T.S. Braver, D.M. Barch, C.S. Carter, and J.D. Cohen. prune their information load to concentrate on what they 2001. Conflict monitoring and cognitive control. Psychological Review 108(3): 624-652. perceive to be essential. For example, does a small-world Brown, J.W., and T.S. Braver. 2007. Risk prediction and aversion by ante- network structure emerge, in which there are highly inter­ rior cingulate cortex. Cognitive, Affective, & Behavioral Neuroscience connected groups with sparse connections between groups? 7(4): 266-277. (In this type of network structure, each node has a low Camerer, C.F. 2003. Behavioral Game Theory: Experiments in Strategic number of connections compared with the total number Interaction. Princeton, N.J.: Princeton University Press. 

34 OPPORTUNITIES IN NEUROSCIENCE FOR FUTURE ARMY APPLICATIONS Clement, J. 1982. Students’ preconceptions in introductory mechanics. Mayer, R.E. 2003. Learning and Instruction. �������������������������� Upper Saddle River, N.J.: American Journal of Physics 50(1): 66-71. Merrill. Corrado, G.S., L.P. Sugrue, J.R. Brown, and W.T. Newsome. 2009. McCabe, K., D. Houser, L. Ryan, V. Smith, and T. Trouard. 2001. A func- The trouble with choice: Studying decision variables in the brain. tional imaging study of cooperation in two-person reciprocal exchange. Pp. 463-480 in Neuroeconomics: Decision Making and the Brain. P.W. Proceedings of the National Academy of Sciences of the United States Glimcher, C.F. Camerer, E. Fehr, and R.A. Poldrack, eds. New York, of America 98(20): 11832-11835. N.Y.: Academic Press. McCloskey, M. 1983. Naive theories of motion. Pp. 229-324 in Mental Donovan, C.L., R.E. Mayer, and M.B. Miller. 2007. Individual differ- Models. D. Gentner and A.L Stevens, eds. Hillsdale, N.J.: Lawrence ences in fractional anisotropy and learning strategy predict variability Erlbaum Associates. in brain activity during episodic encoding and retrieval. Society for Mestre, J.P. 1991. Learning and instruction in pre-college physical science. Neuroscience, 2007 Abstracts. Available at http://www.abstractsonline. Physics Today 44(9): 56-62. com/viewer/viewAbstract.asp?CKey={74138971-902E-45F7-80BF- Miller, M.B., J.D. Van Horn, G.L. Wolford, T.C. Handy, M. Valsangkar- 12CACEC7A65E}&MKey={FF8B70E5-B7F9-4D07-A58A-C1068F Smyth, S. Inati, S. Grafton, and M.S. Gazzaniga. 2002. Extensive DE9D25}&AKey={3A7DC0B9-D787-44AA-BD08-FA7BB2FE9004} individual differences in brain activations associated with episodic &SKey={75963B82-6391-4C70-8E91-F80E73368F20}. Last accessed retrieval are reliable over time. Journal of Cognitive Neuroscience August 25, 2008. 14(8): 1200-1214. Doody, G.A., M. Gotz, E.C. Johnstone, C.D. Frith, and D.G. Cunningham Miller, M.B., and J.D. Van Horn. 2007. Individual variability in brain activa- Owens. 1998. Theory of mind and psychoses. Psychological Medicine tions associated with episodic retrieval: A role for large-scale databases. 28(2): 397-405. International Journal of Psychophysiology 63(2): 205-213. Dunbar, K., J. Fugelsang, and C. Stein. 2007. Do naïve theories ever go �������������������������������� Montague P.R., G.S. Berns, J.D. Cohen, S.M. McClure, G. Pagnoni, away? Pp. 411- 450 in Thinking with Data. M. Lovett and P. Shah, eds. M. Dhamala, M.C. Wiest, I. Karpov, R.D. King, N. Apple, and R.E. Hillsdale, N.J.: Lawrence Erlbaum Associates. Fisher. 2002. Hyperscanning: Simultaneous fMRI during linked social Eisenberger, N.I., M.D. Lieberman, and K.D. Williams. ��������������������� 2003. Does rejection i ­ nteractions. Neuroimage 16(4): 1159-1164. hurt? An fMRI study of social exclusion. Science 302(5643): 290-292. Newell, K.M. 1996. Change in movement and skill: Learning, retention, and Fletcher, P.C., F. Happé, U. Frith, S.C. Baker, R.J. Dolan, R.S.J. Frackowiak, transfer. Pp. 393-429 in Dexterity and Its Development. M.L. Latash and and C.D. Frith. 1995. Other minds in the brain: A functional imaging M.T. Turvey, eds. Mahwah, N.J.: Lawrence Erlbaum Associates. study of “theory of mind” in story comprehension. ����������������� Cognition 57(2): Niv, Y., and P.R. Montague. 2009. Theoretical and empirical studies of 109-128. learning. Pp. 331-351 in Neuroeconomics: Decision Making and the Frith, U., and C.D. Frith. 2003. Development and neurophysiology of men- Brain. P.W. Glimcher, C.F. Camerer, E. Fehr, and R.A. Poldrack, eds. talizing. Philosophical Transactions of the Royal Society B: Biological New York, N.Y.: Academic Press. Sciences 358(1431): 459-473. Ochsner, K.N., and M.D. Lieberman. 2001. The emergence of social cogni- ������������������������������������ Gallagher, H.L., and C.D. Frith. 2003. Functional imaging of “theory of tive neuroscience. American Psychologist 56(9): 717-734. mind.” Trends in Cognitive Sciences 7(2): 77-83. Paulus, M.P., S.F. Tapert, and M.A. Schuckit. 2005. Neural activation pat- Geva, N., R. Driggers, and A. Mintz. 1996. Effects of ambiguity on strategy terns of methamphetamine-dependent subjects during decision making and choice in foreign policy decision making: An analysis using com- predict relapse. Archives of General Psychiatry 62(7): 761-768. puterized process tracing. Paper presented at the 30th North American Poldrack, R.A., F.W. Sabb, K. Foerde, S.M. Tom, R.F. Asarnow, S.Y. Peace Science Society Meeting, October 25-27. Houston, Tex.: Rice Bookheimer, and B.J. Knowlton. 2005. The neural correlates of motor University. skill automaticity. Journal of Neuroscience 25(22): 5356-5364. Glimcher, P. In press. Neuroeconomics, general: History. In New Encyclo- Posner, M.I., and M.K. Rothbart. 2005. Influencing brain networks: Implica- pedia of Neuroscience. Oxford, England: Elsevier. tions for education. Trends in Cognitive Sciences 9(3): 99-103. Graham, H.D., G. Coppin, and M.L. Cummings. 2007. The PRIM: Extract­ Poulton, E.C. 1974. Tracking Skill and Manual Control. New York, N.Y.: ing expert knowledge for aiding in C2 sense & decision making. 12th Academic Press. International Command and Control Research and Technology Sym- Premack, D., and G. Woodruff. 1978. Chimpanzee problem-solving: A test posium, Newport, R.I. Available at http://web.mit.edu/aeroastro/labs/ for comprehension. Science 202(4367): 532-535. halab/papers/The_PRIM.pdf. Last accessed July 14, 2008. Rajkowska, G., and P.S. Goldman-Rakic. 1995. Cytoarchitectonic defini- Happé, F. 2003. Theory of mind and the self. Annals of the New York tion of prefrontal areas in the normal human cortex: II. Variability in Academy of Sciences 1001: 134-144. locations of areas 9 and 46 and relationship to the Talairach coordinate Happé, F., S. Ehlers, P. Fletcher, U. Frith, M. Johansson, C. Gillberg, R. system. Cerebral Cortex 5(4): 323-337. Dolan, R. Frackowiak, and C. Frith. 1996. ‘Theory of mind’ in the brain: Rakic, P. 2005. Less is more: Progenitor death and cortical size. Nature Evidence from a PET scan study of Asperger syndrome. ������������ NeuroReport Neuroscience 8(8): 981-982. 8(1): 197-201. Schmidt, R.A., and T.D. Lee. 2005. Motor Control and Learning: A Behav- Heinrichs, M., T. Baumgartner, C. Kirschbaum, and U. Ehlert. ������ 2003. ioral Emphasis. Champaign, Ill.: Human Kinetics. Social support and oxytocin interact to suppress cortisol and subjec- Squire, L.R., C.E.L. Stark, and R.E. Clark. 2004. The medial temporal lobe. tive responses to psychosocial stress. Biological Psychiatry 54(12): Annual Review of Neuroscience 27: 279-306. 1389-1398. Taylor, M.K., G.E. Larson, A. Miller, L. Mills, E.G. Potterat, J.P. Reis, Hutsler, J.J., W.C. Loftus, and M.S. Gazzaniga. 1998. Individual variation of G.A. Padilla, and R.J. Hoffman. 2007. Predictors of success in basic cortical surface area asymmetries. Cerebral Cortex 8(1): 11-17. underwater demolition/SEAL training—Part II: A mixed quantitative Kording K.P., J.B. Tenenbaum, and R. Shadmehr. 2007. The dynamics of ���������������������� and qualitative study. Naval Health Research Center Report No. 07-10. memory as a consequence of optimal adaptation to a changing body. Springfield, Va.: National Technical Information Service. Nature Neuroscience 10(6): 779-786. Taylor, M.K., A. Miller, L. Mills, E.G. Potterat, J.P. Reis, G.A. Padilla, Kosfeld, M., M. Heinrichs, P.J. Zak, U. Fischbacher, and E. Fehr. 2005. and R.J. Hoffman. 2006. Predictors of success in basic underwater Oxytocin increases trust in humans. Nature 435(7042): 673-676. d ­ emolition/SEAL training—Part I: What we know and where do we go Kosik, K.S. 2003. Beyond phrenology, at last. Nature Reviews Neuro­science from here? Naval Health Research Center Report No. 06-37. Springfield, 4(3): 234-239. Va.: National Technical Information Service. Leslie, A.M., T.P. German, and P. Polizzi. 2005. Belief-desire reasoning as a process of selection. Cognitive Psychology 50(1): 45-85. ���������������������������������� 

TRAINING AND LEARNING 35 Tisserand, D.J., M.P.J. van Boxtel, J.C. Pruessner, P. Hofman, A.C. Evans, Yin, H.H., and B.J. Knowlton. 2006. The role of basal ganglia in habit and J. Jolles. 2004. A voxel-based morphometric study to determine formation. Nature Reviews: Neuroscience 7(6): 464-476. individual differences in gray matter density associated with age and Young, L., F. Cushman, M. Hauser, and R. Saxe. 2007. The neural basis of cognitive change over time. Cerebral Cortex 14(9): 966-973. the interaction between theory of mind and moral judgment. Proceedings Trommershäuser, J., L.T. Maloney, and M.S. Landy. 2009. The expected of the National Academy of Sciences of the United States of America utility of movement. Pp. 95-111 in Neuroeconomics: Decision Making 104(20): 8235-8240. and the Brain. P.W. Glimcher, C.F. Camerer, E. Fehr, and R.A. Poldrack, Zak, P.J. 2004. Neuroeconomics. Philosophical Transactions of the Royal eds. New York, N.Y.: Academic Press. Society B: Biological Sciences 359(1451): 1737-1748. Tulving, E. 2002. Episodic memory: From mind to brain. Annual Review Zak, P.J., R. Kurzban, and W.T. Matzner. 2005. Oxytocin is associated with of Psychology 53: 1-25. human trustworthiness. Hormones and Behavior 48(5): 522-527. Wicker, B., P. Ruby, J.-P. Royet, and P. Fonlupt. 2003. A relation between rest and the self in the brain? Brain Research Reviews 43(2): 224-230.