CHAPTER 14: Higher Cognitive Function and Behavioral Control

• Executive function, the cognitive control of behavior, depends on the prefrontal cortex, which is extensively developed in higher primates and especially humans.

• Working memory is a short-term, capacity-limited cognitive buffer that stores information and permits its manipulation to guide decision making and behavior.

• Attention permits the selection of relevant information from the enormous welter of sensory inputs. Attention may be effortful or may result “bottom-up” from the appearance of a salient stimulus.

• Attention depends on working memory and on mechanisms that filter sensory inputs for access to working memory.

• Attention and working memory are modulated by drugs that directly or indirectly stimulate dopamine D₁ receptors and noradrenergic receptors. Such drugs, most notably the psychostimulants methylphenidate and amphetamine, are used to treat attention deficit hyperactivity disorder (ADHD).

• ADHD is strongly genetically determined, but it has not yet been possible to identify specific risk genes.

• Obsessive–compulsive disorder (OCD) and Tourette syndrome may represent related abnormalities in the circuitry connecting the prefrontal cortex, striatum, and thalamus. Both are highly heritable conditions, although the specific genes involved remain unknown.

• Declarative or explicit memory, which is the memory of facts and events, and nondeclarative or implicit memory, which includes all other forms such as procedural memory and habits, are the two broad categories of memory.

• Structures in the medial temporal lobe, such as the hippocampus, are particularly important for the temporary storage of declarative memories.

• The striatum plays a central role in procedural and habit memory.

• The amygdala, part of the temporal lobe, and the nucleus accumbens, part of the ventral striatum, are important in emotional memory.

• Long-lasting changes in synaptic efficacy (synaptic plasticity) are thought to be important mechanisms for storing memories.

• Strong emotions enhance memory formation, likely because of the associated activation of diffusely projecting neurotransmitter systems (eg, monoamines, acetylcholine).

• Despite intense interest in finding drugs that enhance memory under pathologic conditions (eg, Alzheimer disease) or normal conditions (eg, aging), no robustly effective agents have yet been developed.

• Social cognition refers to cognitive and emotional processes that underlie social functioning, including social judgments and social affiliation.

• Autism, one of the most strongly heritable neuropsychiatric illnesses, describes a spectrum of disorders characterized by abnormal social function, language impairments, restricted interests, and repetitive behaviors.

The simplest nervous systems sense survival-relevant stimuli in the environment and respond reflexively for self-defense, to obtain nourishment, and to reproduce. As organisms became more complex, an increasing number of neurons were interposed between sensory inputs and motor outputs, creating what has been called “the great intermediate net.” In the human nervous system, this net has reached remarkable levels of complexity, making possible diverse and subtle forms of cognition and behavioral control including many kinds of learning and memory, abstract thought, the creation of fiction, music, and visual art, and the ability to navigate flexibly in uncertain and changing social situations. For the sake of survival, humans retain simple reflexive responses (eg, to noxious stimuli), simple unconscious homeostatic behaviors such as swallowing and control of respiration, and more complex but automatized behavioral responses to salient stimuli (eg, to certain types of threats or to rewards such as attractive foods).

Most human behavior, however, whether automatic or requiring effortful conscious control, whether learned or improvised, results from interactions among sensory stimuli, innate drives (eg, hunger, thirst, sexual arousal), memories, emotions, and diverse types of cognitive processing. This complexity, which brings with it the possibility of flexible and creative responses to sensory stimuli, also creates the potential for conflict among goals and confusion among possible responses. Without coordination among inputs and outputs, and some mechanism to supervise the selection of goals and behavioral outputs, the result could be self-defeating confusion rather than evolutionary advantage.

The first part of this chapter focuses on prefrontal regions of cerebral cortex and their connections with sensory and emotional centers in the brain that exert control over an individual’s behavior under normal and pathologic conditions. The second section covers different types of memory, including explicit memories, which are dependent on hippocampal function, and implicit procedural memories, which are dependent on the striatum. We end with a discussion of social cognition, which is extensively developed in humans and impaired in a range of neuropsychiatric disorders.

Anatomy and Function of the Prefrontal Cortex

The prefrontal cortex is the region that subserves executive function, the ability to exert cognitive control over decision making and behavior. It is also important for anticipating reward and punishment and for empathy and complex emotions. This brain area has grown disproportionately large in our near primate relatives and especially in humans compared with other mammals 14–1. Different subregions 14–2 participate in a large number of distinct circuits that permit the integration and processing of diverse types of information. The prefrontal cortex receives not only inputs from other cortical regions, including association cortex, but also, via the thalamus, inputs from subcortical structures subserving emotion and motivation, such as the amygdala (Chapter 15) and ventral striatum (or nucleus accumbens; Chapter 16). The prefrontal cortex is also innervated by widely projecting neurotransmitter systems, such as norepinephrine, dopamine, serotonin, acetylcholine, and orexin/hypocretin. These diverse inputs and back projections to both cortical and subcortical structures put the prefrontal cortex in a position to exert what is often called “top-down” control or cognitive control of behavior. Because behavioral responses in humans are not rigidly dictated by sensory inputs and drives, behavioral responses can instead be guided in accordance with short- or long-term goals, prior experience, and the environmental context. The response to a delicious-looking dessert is different depending on whether a person is alone staring into his or her refrigerator, is at a formal dinner party attended by his or her punctilious boss, or has just formulated the goal of losing 10 lb. The response to a rattlesnake will differ depending on whether a person is a novice hiker or a herpetologist looking for specimens. Adaptive responses depend on the ability to inhibit automatic or prepotent responses (eg, to ravenously eat the dessert or run from the snake) given certain social or environmental contexts or chosen goals and, in those circumstances, to select more appropriate responses. In conditions in which prepotent responses tend to dominate behavior, such as in drug addiction, where drug cues can elicit drug seeking (Chapter 16), or in attention deficit hyperactivity disorder (ADHD; described below), significant negative consequences can result.

Because the prefrontal cortex does not subserve primary sensory or motor functions, language production, or basic intelligence (eg, arithmetic calculation), it was once thought to be silent or nearly so. As late as the 1950s, prefrontal lobotomy was used to treat schizophrenia and other severe mental disorders and was rationalized, in part, because basic intelligence survived the surgery. However, damage to the prefrontal cortex has a significant deleterious effect on social behavior, decision making, and adaptive responding to the changing circumstances of life. An important role for regions of the prefrontal cortex was demonstrated as early as the 19th century by the case of Phineas Gage 14–1.

Several subregions of the prefrontal cortex 14–2 have been implicated in partly distinct aspects of cognitive control, although these distinctions remain somewhat vaguely defined. The anterior cingulate cortex is involved in processes that require correct decision making, as seen in conflict resolution (eg, the Stroop test), or cortical inhibition (eg, stopping one task and switching to another). The medial prefrontal cortex is involved in supervisory attentional functions (eg, action–outcome rules) and behavioral flexibility (the ability to switch strategies). The dorsolateral prefrontal cortex, the last brain area to undergo myelination during development in late adolescence, is implicated in matching sensory inputs with planned motor responses. The ventromedial prefrontal cortex seems to regulate social cognition, including empathy. The orbitofrontal cortex is involved in social decision making and in representing the valuations assigned to different experiences. It is also implicated in impulsive and compulsive behaviors.

Working Memory

Executive function depends on working memory, a short-term, capacity-limited cognitive buffer that maintains a representation of sensory information. Working memory permits the integration and manipulation of this information to guide thought, emotion, and behavior. Working memory can be demonstrated in a range of mammals and has been studied extensively in nonhuman primates as well as humans. Findings drawn from primate research have been used to examine the mechanisms of working memory and its role in executive function. A classic experiment involved the placement of bilateral lesions in the prefrontal cortex of a chimpanzee and subsequent testing of the animal for delayed spatial responses. The chimpanzee watched as a piece of food was placed under one of several opaque containers. After a brief delay, the animal was allowed to choose a container. Unlesioned control animals uniformly chose the container with food, whereas lesioned animals made random selections. After several additional experiments were performed to determine the types of cognitive deficits involved, it became apparent that basic sensory and cognitive functions were intact in the lesioned chimps. What the chimpanzees lacked was the ability to maintain an internal representation of the food and its significance. These chimps needed ongoing sensory stimulation to track the food.

Our understanding of working memory has expanded considerably since these initial experiments. Subsequent studies, for example, have examined the electrical activity of particular neurons in prefrontal cortex in experimental paradigms that require working memory. In one such study, a monkey was conditioned to fix its eyes on a central point on a video screen 14–3. Subsequently a box was displayed briefly in one of eight areas on the screen. After a 3- to 6-second delay, the central fixation point was removed from the screen, and the monkey was trained to shift its gaze to the area where the box previously had been displayed. This study enabled the identification of neurons in prefrontal cortex that are specific for the region of the screen where the box was displayed. Such neurons become more active during the delay phase of the task and return to baseline levels of activity when the gaze returns to the area in which the box appeared. The increased activity of neurons in the prefrontal cortex during the delay phase of a task is one signature of working memory. This activity appears to provide an internal representation of the box even when it is not visible.

The pharmacologic manipulation of working memory has been the focus of considerable investigation, in part because of the working memory deficits that characterize schizophrenia (Chapter 17) and in part because working memory may be impaired in ADHD and in several other conditions, including severe stress.

Mild dopaminergic stimulation of the prefrontal cortex enhances working memory; in contrast, higher levels of stimulation profoundly disrupt this function. Because stress is known to increase dopaminergic transmission from the ventral tegmental area to the prefrontal cortex, the actions of dopamine in this brain region may explain why low levels of stress can enhance performance in working memory tasks, whereas higher levels of stress can disrupt performance. These findings are consistent with evidence that working memory depends on an optimal level of stimulation of D₁ dopamine receptors (see also Chapter 17). D₁ agonists have, therefore, been studied as enhancers of working memory, although it has not yet been possible to generate such agonists devoid of troubling side effects such as nausea and vomiting.

Manipulation of the norepinephrine system also affects working memory. For example, α₂-adrenergic agonists such as clonidine and guanfacine (14–4; see Chapter 6) enhance working memory, a finding that may explain the utility of these agents in the treatment of ADHD. The selective norepinephrine reuptake inhibitor (NRI), atomoxetine 14–4, is approved for the treatment of ADHD, although its effects on working memory per se have not been established. Atomoxetine does not appear to be distinct in its therapeutic properties from older tricyclic NRIs (eg, desipramine) that are approved for the treatment of depression (Chapter 15), although it does have milder side effects.

Therapeutic (relatively low) doses of psychostimulants, such as methylphenidate and amphetamine 14–4, improve performance on working memory tasks in individuals with ADHD and, at higher doses, in normal subjects. Positron emission tomography (PET) demonstrates that methylphenidate decreases regional cerebral blood flow in the dorsolateral prefrontal cortex and posterior parietal cortex while improving performance of a spatial working memory task. This suggests that cortical networks that normally process spatial working memory become more efficient in response to the drug. Both methylphenidate and amphetamines act by increasing synaptic levels of dopamine, norepinephrine, and serotonin, actions mediated via the plasma membrane transporters of these neurotransmitters and via the shared vesicular monoamine transporter (Chapter 16). Based on animal studies with micro-iontophoretic application of selective D₁ dopamine receptor agonists (such as the partial agonist SKF38393 or the full agonist SKF81297) and antagonists (such as SCH23390), and clinical evidence in humans with ADHD, it is now believed that dopamine and norepinephrine, but not serotonin, produce the beneficial effects of stimulants on working memory. At abused (relatively high) doses, stimulants can interfere with working memory and cognitive control, as will be discussed below. It is important to recognize, however, that stimulants act not only on working memory function but also on general levels of arousal and, within the nucleus accumbens, improve the saliency of tasks. Thus, stimulants improve performance on effortful, but tedious tasks, probably acting at different sites in the brain through indirect stimulation of dopamine and norepinephrine receptors.

Although the animal studies described above involve very simple mental representations, working memory is at the heart of many complex mental tasks in higher mammals. When viewed as a rudimentary form of abstraction, working memory can be seen as crossing from the realm of the brain into that of the mind, where internal representations of the external world are consistent and reliable. The ability to think in abstract terms presumably allows humans to create a sense of identity, to establish goals, and to plan for the future.

Attention

In the mass of sensory stimulation by which an animal is bombarded at all times, it must have mechanisms to select the information that is most relevant for its particular situation and ultimately its survival. Selection of relevant information is the role of attention. Once attended to, information gains access to working memory, and can thereby be used to plan appropriate responses. In humans, responses may be simple, but may also involve complex trains of cognition, the activation of emotional circuits, and production of elaborate behaviors. It must be emphasized that much important sensory information is processed by the brain unconsciously, and need not be made conscious to elicit significant responses. For example, subliminal processing of fearful faces can activate physiologic aspects of the fear response in humans without ever reaching consciousness (Chapter 15). However, information that is processed consciously is, at some point, attended to and entered into working memory.

Attention is not a simple unitary function. It may be commanded by “bottom-up” sensory information such as a stabbing pain or the sudden appearance of a loud noise or bright flashing light. The concept of attention also includes effortful “top-down” processing involving the prefrontal cortical circuits that connect with other specialized brain regions. For instance, we can purposefully allocate attention (eg, to a particular spatial location, which involves parietal cortex), we can pay selective attention to specific features of the world (eg, the color of an object, which involves the inferior temporal cortex), and we divide our attention, suppress distractions, and concentrate (ie, sustain attention over time).

In one model, four basic cognitive processes are considered to be the building blocks of attention 14–5: (1) working memory, as mentioned earlier, is required to exert top-down control over those sensory representations that will be attended to and for effortful direction of “the searchlight” of attention. Working memory acts by guiding orientation, such as the direction of the body, the head, or the eyes, and also produces signals that influence the sensitivity of neural circuits that represent information. (2) The allocation of attention, for example, to a point in space, can be shown by functional neuroimaging in humans and by physiologic recordings in monkeys to alter activity in relevant brain regions or relevant neurons. Thus, working memory can exert sensitivity control over neural representations and thereby influence the selection of representations that are attended to. Patients with ADHD have difficulty sustaining attention or ignoring distractions, suggesting problems with sensitivity control. (3) Diverse sensory information is subjected to salience filters so that irrelevant information does not gain access to working memory. Patients with psychotic disorders including schizophrenia attend to irrelevant and to hallucinatory stimuli, suggesting a failure of filtering processes. Salience is determined, for example, by innate or learned predictors of threat or reward or by the appearance of rare or high-intensity stimuli. (4) Sensory representations that pass salience filters are subjected to competitive processes that select the strongest signals for access to working memory.

The monoamine neurotransmitters and orexin have a critical permissive role in attention by regulating arousal (Chapters 6 and 13). Performance on cognitive tasks has different optimal levels of arousal depending on the degree of effort required. This is captured in simplistic form by the Yerkes–Dodson principle 14–6 that expresses the relationship of arousal to performance for specific tasks as an inverted U-shaped curve. Such a curve also captures the effects of pharmacologic agents that influence arousal, ranging from psychostimulants and caffeine (Chapter 13) among those drugs that increase arousal to β-adrenergic antagonists and benzodiazepines that might decrease maladaptive arousal in an extremely anxious person. Beyond these general permissive effects, dopamine (acting primarily via D₁ receptors) and norepinephrine (acting at several receptors) can, at optimal levels, enhance working memory and aspects of attention. Drugs used for this purpose include, as stated above, methylphenidate, amphetamines, atomoxetine, and desipramine. Modafinil is effective in improving both arousal and attention; it is believed to act (like methylphenidate) by inhibiting the dopamine transporter, although this is not established with certainty (Chapter 13).

Attention deficit hyperactivity disorder

ADHD is characterized by symptoms in three dimensions of behavior: inattention, impulsiveness, and hyperactivity. Hyperactivity may be absent, in which case the term attention deficit disorder (ADD) may be used. These dimensions are continuous with normality, but, when severe, ADHD produces significant impairment. ADHD can be a profound obstacle to success in school or work, despite normal intelligence. It also increases the risk for substance abuse and accidents, as well as for comorbid depression, anxiety disorders, and conduct disorders.

ADHD begins in childhood, often very early, but is typically diagnosed when children enter school. While symptoms often remit during teen years, a substantial fraction of individuals with ADHD remain symptomatic in adulthood. Based on current diagnostic criteria, worldwide prevalence ranges between 3% and 5%, but widely divergent diagnostic practices in different regions of the world produce varying local estimates. Indeed, the diagnosis and treatment of ADHD varies widely in the United States, with far higher rates in affluent, suburban communities compared with those in poorer, inner city areas.

ADHD is highly heritable, with genetic factors comprising ~75% of the risk. However, like all major psychiatric disorders, including depression, bipolar disorder, and schizophrenia, ADHD is genetically complex, making the identification of risk alleles very challenging. Elucidating genetic factors that contribute to ADHD risk awaits large-scale genome-wide studies.

ADHD can be conceptualized as a disorder of executive function; specifically, ADHD is characterized by reduced ability to exert and maintain cognitive control of behavior. Compared with healthy individuals, those with ADHD have diminished ability to suppress inappropriate prepotent responses to stimuli (impaired response inhibition) and diminished ability to inhibit responses to irrelevant stimuli (impaired interference suppression). Such deficits have been documented as well in functional MRI studies.

The ability to suppress prepotent responses is thought to require the action of frontal-striatal-thalamic circuits 14–7. A series of parallel loops connect the prefrontal cortex with specific regions of the basal ganglia and, via the thalamus, project back to prefrontal cortex. These loops are thought to be involved in the initiation and control of motor behavior, attention, cognition, and reward responses. Functional neuroimaging in humans demonstrates activation of the prefrontal cortex and caudate nucleus (part of the dorsal striatum) in tasks that demand inhibitory control of behavior. Subjects with ADHD exhibit less activation of the medial prefrontal cortex than healthy controls even when they succeed in such tasks and utilize different circuits.

Early results with structural MRI show a thinner cerebral cortex, across much of the cerebrum, in ADHD subjects compared with age-matched controls, including areas of prefrontal cortex involved in working memory and attention. Longitudinal studies comparing ADHD and control populations find that cortical thickness tends to normalize during adolescence and young adulthood at least in some patients. These results, while still preliminary, suggest that ADHD might involve an abnormal rate of cortical maturation rather than a unique pattern of cortical development. It is hypothesized that residual symptoms in adulthood might result from incomplete normalization, but at this point the data are not available. Volumetric differences have also been reported for several subcortical structures.

As suggested by the previous discussions of the pharmacology of working memory and attention, ADHD can be treated with low doses of psychostimulants, such as methylphenidate and amphetamine, which remain by far the most effective treatments available. The most widely used treatments are sustained-release preparations of methylphenidate that compensate for its short half-life, or mixtures of amphetamine derivatives with different half-lives to provide both early and extended treatment during the day. The main side effects of stimulants are appetite suppression, growth delay, and insomnia. Atomoxetine or other NRI antidepressants such as desipramine represent an alternative for patients who do not tolerate stimulants, although these drugs are not as effective as stimulants in most cases. The α₂-adrenergic agonists, clonidine and guanfacine, show beneficial effects in a subset of patients, but their use is not well supported by clinical trial data. Children and adults with ADHD also benefit from behavioral treatments aimed at enhancing top-down control of behavior.

Stimulant prescriptions have engendered controversy because these drugs are potentially abusable (when the drugs are used at higher than therapeutic doses), because they are used even in young children (as early as age 3), and because they may be used with benefit by individuals with only mild or even absent symptoms, in which case they are being used for cognitive enhancement rather than treatment. Controversy notwithstanding, there are extensive clinical trial data that demonstrate that stimulants effectively reduce ADHD symptoms, albeit with little data bearing on long-term academic and employment outcomes. Several longitudinal studies suggest that untreated ADHD is associated with elevated risk of substance abuse and conduct disorders, as stated earlier, and that stimulant treatment decreases that risk. Two factors may be at play. First, supervised use of stimulants at therapeutic doses may decrease risk of experimentation with drugs to self-medicate symptoms. Second, untreated ADHD may lead to school failure, peer rejection, and subsequent association with deviant peer groups that encourage drug misuse.

Obsessive–Compulsive Disorder Spectrum Disorders

Obsessive–compulsive disorder (OCD) is characterized by obsessions (intrusive, unwanted thoughts) and compulsions (highly ritualized behaviors intended to neutralize the anxiety and negative thoughts resulting from the obsessions). Individuals with OCD experience the obsessions as unwanted and even nonsensical, but powerfully insistent. Attempts to resist the performance of compulsions result in high levels of anxiety. Typical symptom patterns are remarkably similar across cultures, such as repetitive hand washing, sometimes hours a day to the point of skin damage, to neutralize fears of contamination, or repeatedly checking the front door because affected individuals cannot be sure the door is locked. OCD often begins in childhood or teen years, although later onset can occur. It is estimated to have a heritability of 40% to 50%, although specific genes that contribute to this risk have not yet been identified.

There is a high degree of comorbidity between OCD and Tourette syndrome and related tic disorders: about 20% of OCD patients have a chronic tic disorder, while about 50% of Tourette patients have or will develop OCD. OCD is also often accompanied by depression or by body dysmorphic disorder (relentless obsessions about one’s supposed bodily deformities). Tourette syndrome and body dysmorphic disorder are thought to be related etiologically or pathophysiologically to OCD and are sometimes described as part of an OCD spectrum.

Tourette syndrome is characterized by motor and phonic tics that wax and wane in severity over time. Tics are habitual movements or vocalizations that appear suddenly and that mimic fragments of normal behaviors. They usually begin during childhood, and may start as early as age 3. Patients with Tourette syndrome may also have OCD, ADHD, and other behavioral disorders characterized by poor impulse control. While stress can exacerbate the symptoms of Tourette syndrome, as seen for most neurologic and psychiatric disorders, there is no evidence that stress per se contributes to the pathogenesis of the illness. Indeed, Tourette syndrome is estimated to be highly heritable (~60%), with no bona fide risk genes yet identified.

Frontal-striatal-thalamic circuits are thought to be involved in implicit learning that consolidates fragments of motor behavior, speech, or thought into smooth, but flexible routines, for example, the ability to drive one’s car home without effortful attention, or the ability of an actor to deliver his or her lines without thinking. These well-learned assembled routines are often described as habits. The striatum and other nuclei of the basal ganglia are thought to recode both cortical (eg, cognitive and motor) and subcortical (eg, emotional and motivational) inputs into sequences, sometimes called “chunks” that permit appropriate cues to release automatic responses.

Structural MRI studies of OCD patients show reduced gray matter volume in the medial frontal gyrus, medial orbitofrontal cortex, and other regions. Functional imaging studies of OCD have reported increased activity in components of cortical-striatal-thalamic circuitry including the orbitofrontal cortex and the caudate nucleus. Similarly, in Tourette syndrome, increased activity has been observed in other regions of the prefrontal cortex as well as in the caudate. These and similar observations, combined with increasing basic knowledge of the functioning of basal ganglia circuits (see below), have led to the hypothesis that OCD and Tourette syndrome result from abnormal activity in different cortical-striatal-thalamic loops inappropriately releasing unwanted intrusive thoughts (OCD) or fragments of motor behavior (Tourette). Consistent with this hypothesis, genetic deletion of certain key synaptic proteins (eg, SLITRK5 and SAPAP3) in the striatum of mice elicits dramatic excessive grooming, which at least phenocopies symptoms of OCD. Whether such mechanisms contribute to OCD in humans remains unknown.

Despite the hypothesized similarities in pathophysiology and the frequent co-occurrence of OCD in Tourette patients, the pharmacologic treatments differ. OCD is treated with selective serotonin reuptake inhibitors (SSRIs) such as fluoxetine or sertraline, or with tricyclic antidepressants that have relative selectivity for the serotonin transporter such as clomipramine (Chapter 15). OCD is also responsive to cognitive behavioral therapies focused on stopping unwanted thoughts and rituals, presumably acting to facilitate top-down cognitive control. The doses of SSRIs required for OCD are usually higher than those for depression, and the onset of therapeutic benefit is often even slower (up to 10 weeks). SSRIs are also effective in the treatment of body dysmorphic disorder. Deep brain stimulation, targeting the anterior limb of the internal capsule (which includes the nucleus accumbens), is also approved for the treatment of severe OCD.

In contrast to OCD, Tourette syndrome is most often treated with both first- and second-generation D₂ dopamine receptor antagonist antipsychotic drugs at lower doses than those used to treat schizophrenia. α₂-Adrenergic agonists (eg, clonidine and guanfacine) are also effective in many patients. These drugs are not efficacious in treating OCD. Such differences in treatment response suggest that ascending serotonin and dopamine projections may modulate parallel but separate cortical-striatal-thalamic loops quite differently.

Our ability to learn and remember is such an integral part of our lives that it is difficult to function with any significant memory impairment. Memory permeates existence, from the mundane (Where did I put my keys?) to the profound (Who am I? What are my beliefs?). It is only a slight exaggeration to say that our memories are who we are. Memory depends on the precise functioning of healthy neurons. Age, trauma, malnutrition, and genetic factors can diminish our ability to store and recall information, and severe forms of memory impairment, as seen in common dementias such as Alzheimer disease, can have tragic consequences. This section considers the neural mechanisms that underlie learning and memory.

Declarative and Nondeclarative Memory

Learning and memory are closely related to one another, but the terms are not interchangeable. Learning refers to the process by which behavior can be changed as a result of experience or practice. Memory refers to the ability to recall what has been learned. There are several different types of memory that can be divided into two major classes termed declarative and nondeclarative 14–8. Declarative memory, which also is known as explicit memory, refers to the process that enables us to recall facts or events. Memorizing the Gettysburg Address and remembering where you parked your car this morning involve declarative memory. This type of memory also may be thought of as conscious memory or as memory that is capable of being verbalized. In contrast, nondeclarative memory, also known as implicit memory, can be thought of as subconscious or unable to be verbalized. The term nondeclarative memory is used broadly and may refer to several learning processes, including those that involve procedural skills and the formation of habits.

Procedural memory is a type of nondeclarative memory involved in learning skilled behaviors; it is used, for example, when an individual learns to type or to play tennis. Procedural memory is distinct from a declarative memory of the actual training episode; for instance, typing is distinct from remembering when you learned how to type. The formation of habits, or the gradual acquisition of behavioral tendencies that are specific to a set of stimuli, also involves nondeclarative memory. Habit learning is exhibited during operant conditioning, as when an animal learns to perform a task to obtain some reward, such as food or a drug of abuse. Significantly, declarative and nondeclarative memory processes can be experimentally distinguished, are mediated by distinct neural substrates, and are affected by different types of diseases 14–8.

The Medial Temporal Lobe System

In 1953, a patient referred to as H.M. underwent a bilateral resection of the medial structures of the temporal lobe—including the anterior two thirds of the hippocampus—as a desperate measure to treat intractable epilepsy. This surgery had tragic consequences that caused H.M. to become one of the best known neurologic patients. After his operation, H.M. had severe anterograde amnesia: he lost his ability to store new memories. Although he could recall the names of people he had known for many years before his operation, he could not remember the names of individuals he met postoperatively, including those of nurses and doctors whom he would meet repeatedly over a period of several months. In addition to his inability to form new memories, H.M. lost memories of events that had occurred up to a decade before his surgery. Yet his ability to recall more remote events remained intact and his I.Q. test scores rose (from 104 to 118), an improvement that can be attributed to postsurgical control of his seizures. Moreover, H.M. retained some memory capacity; with great exertion he was able to remember facts, such as short strings of digits, for approximately 15 minutes. However, as soon as he was distracted, such facts would vanish from his memory. Thus, H.M. did have the ability to store facts for brief periods of time, that is, he had normal working memory. However, he appeared unable to transfer a fact from working memory to long-term memory, which enables facts to remain accessible for long periods of time.

Despite his difficulty in storing facts, H.M. was capable of procedural learning as evidenced by his ability to learn how to trace an object while viewing his hand in a mirror. Although such a task is quite tricky, H.M. learned to perform this task as quickly and as well as control subjects. However, after repeating the task many times he was completely unable to recall whether he had ever previously performed the task. This demonstrated that procedural learning can occur independently of declarative learning and that declarative memory, but not procedural or other types of nondeclarative memory or working memory, is critically dependent on at least one of the brain structures that had been removed during H.M.’s surgery.

Patients such as H.M., who have well-defined brain lesions with otherwise good health, are relatively rare. Thus, our knowledge of the brain structures that are crucial for declarative learning has been drawn primarily from animal experiments. It may seem problematic to attribute declarative learning to animals that cannot talk to human experimenters. However, it is possible to distinguish “learning what” (declarative memory) from “learning how” (nondeclarative memory) in nonhuman primates and rodents. One method of testing declarative memory in animals involves the use of the “delayed nonmatching to sample task.” During this task, a test subject is shown an object that is subsequently taken away. Next, two objects—a new object and the original object—are presented. To receive a reward, the subject must choose the object that has not been seen before. Consequently, to choose correctly, the subject must be able to retain a declarative memory of the original object between the two presentations. Normal monkeys perform this task well. Monkeys with bilateral medial temporal lobe lesions perform at chance levels, except when the intervals between presentations are short. Extensive lesion studies have revealed that the specific crucial structures underlying the ability to store declarative memories are the hippocampus, the subicular complex, and the entorhinal cortex 14–9. Importantly, these findings are consistent with what has been learned from the small number of human patients with lesions restricted to medial temporal lobe structures.

Does the medial temporal lobe store our declarative memories? The answer is that it probably does not, at least not permanently. As previously mentioned, H.M. did not lose all of his declarative memories. He was able to recall his name and his date of birth, and he appeared to have accurate and normal memories of events that had occurred 10 or more years before his operation. Results of animal experiments are consistent with H.M.’s experience. In one type of experiment, animals are trained in a task, and at varying times thereafter undergo bilateral damage to the hippocampus. Retention of the task is then assessed shortly after the surgery. These studies reveal that such damage causes a retrograde amnesia of days to months depending on the time interval between the training and the surgery. Experiments such as these and observations of human amnesiacs have led investigators to postulate that structures in the medial temporal lobe coordinate a process by which memories or associations are gradually imprinted on the cerebral cortex. How this occurs has not been determined, but it is believed that inputs from the medial temporal lobe to specific cortical regions gradually reorganize synaptic connections in the cortex until memories are somehow stored and available for recall. Recent work suggests that some of this consolidation of declarative memories in the cortex occurs during sleep.

In recent years, reconsolidation of declarative memories has been described. This is a process whereby a previously consolidated memory is recalled and then updated. Such updating, while strengthening a memory, can also render it less accurate than earlier versions. The molecular and cellular basis of memory consolidation and reconsolidation are thought to be similar (eg, both of them require new protein synthesis), but the detailed mechanisms remain poorly understood (see below).

The Striatum and Habit Memory

In contrast to declarative memory, nondeclarative memory does not appear to be supported by a single brain region, but by diverse neural circuits that underlie the behavior that comes to be modified by learning. For example, the basal ganglia play a central role in the acquisition of stimulus–response associations that, as described above, can ultimately be reorganized into automatic behavioral repertoires or habits. Stimulus–response associations result from the gradual modification of neural circuits that comprise the basal ganglia. Evidence that the caudate nucleus and putamen influence stimulus–response learning comes from lesion studies in rodents and primates and from neuroimaging studies in humans and from studies of human disease.

In Parkinson disease, the dopaminergic innervation of the caudate and putamen is severely compromised by the death of dopamine neurons in the substantia nigra pars compacta (Chapter 18). Patients with Parkinson disease have normal declarative memory (unless they have a co-occurring dementia as may occur in Lewy body disease). However, they have marked impairments of stimulus–response learning. Patients with Parkinson disease or other basal ganglia disorders such as Huntington disease (in which caudate neurons themselves are damaged) have deficits in other procedural learning tasks, such as the acquisition of new motor programs.

Synaptic Plasticity: How Do Neurons Remember?

Whether we consider declarative or nondeclarative memory, a key question is how do the properties of neurons change in a manner that allows new memories to form. A major mechanism by which this occurs is synaptic plasticity, which refers to the ability of neural activity itself to modulate the strength of synaptic communication. Indeed, as discussed in Chapter 5, it is now well established that in virtually all brain regions different patterns of activity can bidirectionally influence synaptic strength by eliciting various forms of either long-term potentiation (LTP), a long-lasting increase in synaptic strength, or long-term depression (LTD), a long-lasting decrease. In the context of learning and memory, NMDA receptor–dependent LTP has received the most attention, and we will briefly review its properties and the evidence of its functional importance. Molecular mechanisms underlying LTP and LTD are covered in Chapter 5.

LTP and memory formation

LTP was first described in the hippocampus where it was shown that repetitive, brief (1-second) high-frequency stimulation of axons sufficient to strongly depolarize postsynaptic cells caused a long-lasting increase in the size of the postsynaptic response lasting for hours and in some cases even weeks 14–10. Not only can LTP be elicited in the hippocampus, a structure known to be important for learning and memory, but it also has properties that seem ideal for a synaptic mechanism that underlies the storage of new information: it can be elicited rapidly by a short stimulus (high-frequency stimulation) and exhibits both input specificity and associativity. Input specificity indicates that LTP occurs only at synapses that are stimulated by a given pattern of afferent activity, and not at adjacent, unstimulated synapses on the same postsynaptic cell. Limiting the number of synapses on a given cell that are modified by afferent activity greatly increases the storage capacity of the neural circuit in which that cell participates. The property of associativity refers to the fact that the generation of LTP requires coincident presynaptic and postsynaptic activity. Importantly, a weak input that is incapable of significantly depolarizing the cell is potentiated only if it is active at the same time that the postsynaptic cell is depolarized by a strong input.

These properties of LTP derive from the fact that it is triggered by a large Ca²⁺ influx through postsynaptic NMDA receptors and that activation of these receptors requires coincident presynaptic release of glutamate to depolarize the postsynaptic neurons, which relieves the Mg²⁺ block of the NMDA receptor channel (Chapter 5). Thus, LTP does not occur at synapses unless they are active when the postsynaptic cell is strongly depolarized. Because a neuron generally has thousands of inputs and several of these must be active simultaneously to cause significant depolarization of the neuron, the firing of a single or small numbers of inputs does not generate LTP.

A great deal of experimental evidence supports the requirement of NMDA receptor activation and Ca²⁺ influx for the induction of this major form of LTP. As discussed in Chapter 5, this NMDA receptor–mediated rise in Ca²⁺ activates intracellular signaling pathways (in particular, Ca²⁺-activated protein kinases), which increase synaptic strength by increasing the number of AMPA receptors in the postsynaptic membrane. The enzymatic alteration of preexisting synaptic proteins by means of phosphorylation is sufficient to change synaptic efficacy for relatively short periods of time. Therefore, the maintenance of LTP for days or weeks involves additional mechanisms. LTP is associated with the structural enlargement of synapses and possibly the growth of new dendritic spines. Such changes in synapse structure can explain how the initial increase in synaptic strength becomes permanent since larger synapses in general are stronger synapses and would allow the new “information” encoded by LTP to become more permanent.

The prolonged enhancement of synaptic strength likely requires the activation of particular genes and new protein synthesis. Although the critical gene products are unknown, the cAMP pathway and its regulation of the transcription factor CREB (Chapter 4) have been strongly implicated in the long-term maintenance of LTP and also in the maintenance of long-term memory. In diverse invertebrate and vertebrate species, for example, manipulations that decrease functioning of the cAMP pathway or CREB have been shown to impair certain memory tasks, whereas manipulations that enhance CREB function have the opposite effect.

An important question is how altered gene expression, which occurs in the nucleus of the cell body, leads to the maintenance of potentiation only in previously stimulated synapses, and not in the neuron’s thousands of other synapses. One possibility is that the modification of preexisting synaptic proteins that occurs immediately after the induction of LTP might somehow tag these synapses so that the newly synthesized proteins generated through the stimulation of gene transcription are targeted to these synapses and not to adjacent synapses on the same cell.

Although LTP displays characteristics that might be expected of a memory mechanism, it is experimentally difficult to directly link LTP per se to a behavioral memory. Based on the hypothesis that LTP underlies learning and memory, two major predictions can be made. First, the prevention of LTP should be able to prevent or at least disrupt the establishment of memories. Consistent with this prediction, genetic manipulations in mice that lead to the absence of the NMDA receptor in various regions of the hippocampus cause profound deficits in spatial learning. Conversely, mutant mice with enhanced NMDA receptor function in this brain region, achieved through the overexpression of NMDAR_2B receptor subunits, which increases receptor-mediated currents (Chapter 5), perform better than wild-type mice in memory tasks. Furthermore, when injected into the hippocampus of rats 1 day after they had been trained in a spatial memory task, agents that reverse LTP caused the loss of the memory. These findings strongly suggest that LTP in the hippocampus is required for at least some forms of learning and memory known to be dependent on this brain region.

A second prediction is that if LTP underlies learning and memory, it also should occur during the formation of memories. However, its presence during memory formation is not easy to demonstrate since it is very difficult to isolate the small proportion of neural connections that are responsible for a given learning event. Nonetheless, LTP-like phenomena have been observed to occur in parallel with learning in the hippocampus and in the amygdala during a form of learning termed fear conditioning. As described in Chapter 15, one form of conditioned fear is produced in experimental animals when a neutral tone is paired with an electric shock. This fear conditioning involves the amygdala, which receives glutamatergic inputs that undergo NMDA receptor–dependent LTP. Consistent with a role for LTP, blocking NMDA receptors in the amygdala prevents the development of the conditioned fear response. A subsequent critical experiment, in which auditory-evoked responses in the amygdala were recorded while an animal underwent a conditioning procedure, demonstrated that fear conditioning induces LTP in the auditory pathway to the amygdala 14–11. Taken together, these findings make an impressive case for a connection between memory processes and LTP.

Other forms of neural plasticity

Although NMDA-dependent LTP remains the synaptic mechanism most commonly thought to mediate learning and memory, other forms of synaptic plasticity are also likely to play important roles. These include other forms of LTP as well as several different forms of LTD (Chapter 5). Evidence suggests, for example, that a form of LTD observed in the cerebellum is critically important for mediating certain types of motor learning. As well, LTD may be important for reversing LTP because, without such a mechanism, LTP might completely saturate synaptic strength at large numbers of synapses and thereby limit the storage capacity of any given neural circuit. Indeed, computational models of neural networks have demonstrated that the ability to bidirectionally modify synaptic strength, that is, utilizing both LTP and LTD, greatly increases the power of these circuits to store and retrieve information.

Synaptic plasticity (LTP and LTD) is the best understood mechanism by which neuronal responsiveness can be regulated. However, there is evidence that nonsynaptic plasticity (also referred to as whole cell or homeostatic plasticity; Chapter 2) may also be important, particularly for more generalized forms of learning and memory such as emotional memory versus memory of a specific event. Nonsynaptic plasticity comprises changes in the electrical excitability of neurons that are generalized to many synapses or perhaps the entire cell and not localized to a particular dendritic spine. For example, altered levels of expression of ion channels in a neuron or even changes in the subtypes of ion channels expressed in response to some stimulus would be expected to either increase or decrease that neuron’s excitability by a host of inputs. Similarly, general growth or retraction of dendritic arborizations would have significant effects on neuronal responsiveness that would not be limited to specific synapses. It is easy to imagine how such nonsynaptic plasticity might mediate changes in the responsiveness of neurons to certain types of environmental exposures, for example, tolerance or sensitization to a drug of abuse, which can be viewed as a type of memory (Chapter 16). Further work is needed to determine whether nonsynaptic plasticity also contributes to memory of nonpharmacologic, that is, behavioral–environmental stimuli.

A detailed understanding of the physical changes in the nervous system that are responsible for learning and memory will require a great deal of additional research. However, the elucidation of some of the basic mechanisms underlying LTP and LTD and other mechanisms of plasticity has focused attention on what for many decades was believed to be an intractable problem. Thus, it seems likely that major discoveries are on the horizon and that during the next decade new pharmacologic agents will be developed to facilitate learning and memory by targeting proteins that have been shown to be involved in neural plasticity.

Beyond LTP: why we remember what we remember

It is impossible to remember everything. Therefore, an important goal for research on learning and memory is to understand the factors that allow us to remember certain events or associations and forget others. One factor that can enhance memory is extreme emotion, positive or negative. Almost everybody will remember a car accident or where they were during a public tragedy such as 9/11. This type of declarative memory, which is enhanced by strong emotion, is often called flashbulb memory. In general, the stronger the associated emotion when a memory is encoded, the more vivid is the memory. This cognitive enhancement of declarative memory, which is hippocampal-dependent, also requires an intact amygdala. Evidence suggests that such enhancement can be inhibited, without associated amnesia, by β-adrenergic antagonists (eg, propranolol). This observation is consistent with the well-established promotion of certain forms of LTP in the hippocampus and amygdala by β-adrenergic receptor activation, which involves stimulation of the cAMP–CREB pathway.

Strong emotion also produces amygdala-dependent emotional memories (see Chapter 15). This form of mostly nondeclarative memory can be observed even in individuals with hippocampal lesions who have no conscious recall of the circumstances under which the memory was acquired. Emotional memories are usually activated by cues that are reminiscent of the original stimulus and that produce physiologic and behavioral responses appropriate to the emotion. In the case of fear, such responses include activation of the sympathetic nervous system and may result in a freezing response in prey species. Conversely, in strongly rewarding situations, responses elicited by emotional memory lead to approach behavior and physiologic preparation for activities such as eating, drinking, or sex. Both the declarative and nondeclarative components of memories when encoded in the context of strong emotional stimuli have survival value: a vivid memory of a dangerous situation helps an organism avoid similar situations in the future, and memory of a rewarding situation helps an organism obtain future rewards.

However, emotional memory can be maladaptive in some situations, for example, when it is activated under the pathologic circumstances that characterize posttraumatic stress disorder (PTSD; Chapter 15) or drug addiction (Chapter 16). Most research on the neural substrates of emotional memory has focused on the amygdala in studies of aversive stimuli and on the ventral striatum in studies of reinforcing stimuli. Currently, a major challenge is to understand how these subcortical structures interact with more traditional memory structures, such as the hippocampus and prefrontal cortex, to produce fully integrated emotional memories, and how such circuits malfunction in pathologic conditions. The view that pathologic emotional memories may contribute to disorders such as PTSD and drug addiction has led to the novel idea of enhancing extinction of these memories during treatment. Rather than simply undoing the earlier memory, extinction is now thought to represent a second learning event on top of the underlying memory. Current clinical trials are testing whether stimulation of pathologic memories in patients with PTSD or drug addiction (eg, by use of virtual reality of combat or drug-taking situations) coupled with drugs that promote new memory might enhance extinction, or perhaps modify the pathologic memory through reconsolidation, and lead to clinical improvement. One such drug used for this purpose is D-cycloserine, a partial agonist at NMDA glutamate receptors. There is also evidence in animal studies that glucocorticoids may exert similar actions.

A decrement of cognitive abilities is a hallmark of normal aging, a phenomenon observed across mammalian species 14–12. Unlike the situation for Alzheimer disease or other dementias, which are characterized by neuronal death (Chapter 18), normal aging is thought to be caused by more subtle changes in synaptic morphology and function in prefrontal cortex, hippocampus, and presumably other memory-related brain regions. A role for estrogen has been implicated in protecting against such age-associated deficits in females. Understanding the molecular basis of age-related changes in neuronal morphology and cognition is a major goal of current research.

Pharmacologic Regulation of Memory

There has been intense interest in developing medications that correct the cognitive impairments commonly seen during aging or in many neurologic and psychiatric illnesses, in particular, schizophrenia and Alzheimer disease, not to mention medications that promote cognitive function in normal individuals, so-called cognitive enhancers. Numerous strategies are currently under investigation and several show promise in preclinical studies; however, no drug mechanism has yet been definitively validated in humans.

Among the neurotransmitters with the greatest influence on memory is acetylcholine. In both humans and animals, muscarinic cholinergic antagonists, such as atropine and scopolamine, produce memory deficiencies at low doses and a more profound disruption of memory and consciousness known as delirium. Delirium may be caused by any of several common psychiatric medications that have muscarinic cholinergic antagonist properties, such as tricyclic antidepressants (Chapter 15) and first-generation antipsychotic drugs (Chapter 17). Delirium is particularly common among the elderly, where it can be confused with dementia 14–1.

In the mammalian brain, cholinergic nuclei—most notably the nucleus basalis, or Meynert’s nucleus, in the basal forebrain—project heavily to the neocortex and hippocampus (Chapter 6). The release of acetylcholine by these nuclei generally promotes depolarization or rhythmic firing of postsynaptic neurons through the activation of muscarinic and nicotinic cholinergic receptors. Carbachol, for example, a muscarinic cholinergic agonist, produces rhythmic, synchronized firing of hippocampal CA1 pyramidal neurons. This synchronized firing occurs at a frequency of approximately 5 to 7 Hz and is described as a theta rhythm. The precise role of theta rhythm in memory storage is not completely understood; however, when it is induced in a hippocampal slice preparation, weak stimuli that ordinarily do not produce potentiation can evoke LTP. One current view is that cholinergic activity decreases the threshold of neural activity necessary for learning. Nicotinic receptors are similarly implicated in hippocampal function and in the formation of memory.

This body of knowledge led to the introduction of acetylcholinesterase inhibitors (eg, donepezil and tacrine) for the treatment of memory deficits in Alzheimer disease. However, the efficacy of these agents is weak at best (Chapter 18). Current efforts are focused on specific muscarinic or nicotinic receptor agonists or positive allosteric modulators (PAMs). For example, animal studies suggest that agonists or PAMs at nicotinic α₇ homomeric receptors or at muscarinic M₁ or M₄ receptors can improve learning and memory. Indeed, xanomeline, a PAM with some selectivity for M₁ and M₄ receptors, showed early promise in improving cognition, but its development was halted due to peripheral side effects. Efforts to generate more selective agents continue. An alternative approach would be the development of selective antagonists or negative allosteric modulators (NAMs) of M₂ muscarinic receptors; since these receptors serve as autoreceptors (Chapter 6), their antagonism would be expected to enhance cholinergic neurotransmission. Clinical studies are now under way to validate such mechanisms in humans.

Several other mechanisms are being considered to improve cognitive function under pathologic and normal conditions. Drugs that promote glutamate function might be expected to enhance learning and memory, based on the essential role of glutamate in synaptic plasticity as discussed earlier. D-Cycloserine, a partial agonist at the glycine site of the NMDA glutamate receptor, is under clinical evaluation as discussed above. So-called AMPAkines, drugs that are PAMs of AMPA glutamate receptors (Chapter 5), show promise in animal studies and are also in clinical trials. Agonists or antagonists at particular metabotropic glutamate receptors might also be expected to promote glutamate-mediated synaptic plasticity and thereby enhance learning and memory.

The ability of dopamine D₁ receptor agonists, in prefrontal cortical circuits, to promote working memory was discussed earlier in this chapter. Similarly effective in these brain regions in some animal models are protein kinase A (PKA) activators or phosphodiesterase (PDE) inhibitors; these agents would be expected to mimic the activation of D₁ receptors, which signal via the G_s–cAMP–PKA pathway (Chapters 4 and 6). Such agents might also be expected to promote memory formation in hippocampal and other circuits, where they would mimic the promemory effects of β-adrenergic receptors, which couple to this same signaling pathway. The procognitive effects of stimulating the cAMP pathway are consistent with this pathway’s activation of CREB, a mediator of long-term memory as discussed earlier. Rolipram, an inhibitor of PDE4, shows moderate cognitive-enhancing activity in preclinical studies, although it has unacceptable side effects (eg, nausea and vomiting). Accordingly, there has been great effort to develop more selective inhibitors of PDE4 subtypes, which might be devoid of these side effects, as well as inhibitors of several other PDE subtypes highly expressed in memory-related circuits in the brain.

Based on the ability of high-potency, short-acting benzodiazepines (eg, midazolam and triazolam) to disrupt learning and memory, there has been interest in developing inverse agonists at the benzodiazepine site of the GABA_A receptor as cognitive enhancers. While full inverse agonists would cause seizures (Chapter 5), it is possible that weak partial inverse agonists might be sufficiently safe and effective. This strategy has not yet been tested in humans. Histamine H₃ antagonists or inverse agonists, which would be expected to enhance central histaminergic neurotransmission (Chapters 6 and 13), have also been considered to promote cognition, but clinical trials have not yielded positive results. Finally, there is interest in the potential of adenosine receptor antagonists in promoting learning and memory (Chapters 8 and 13).

While most efforts are being devoted toward developing drugs that enhance memory, there are a small number of clinical situations where disruption of memory is advantageous. The high-potency, short-acting benzodiazepines mentioned in the previous paragraph are useful in the context of medical procedures (eg, colonoscopy, dental surgery) in part because they disrupt memory. Moreover, as previously indicated, it appears that β-adrenergic antagonists such as propranolol may be useful in preventing memories of traumatic events (Chapter 15).

Social cognition refers to cognitive and emotional processes that explicitly underpin social processes, such as social judgments or social affiliation. Other cognitive processes, such as language, that enable social interaction are considered outside the purview of social cognition or of social neuroscience more broadly. Humans are a social species and, as such, are constantly emitting social signals, both consciously (effortfully and automatically) and, more commonly, unconsciously. Similarly, we receive and automatically process a large number of social signals throughout the day. Along with our biases and nonsocial sensory inputs the social signals we receive influence diverse social judgments. These include: (1) judging the likely mental states and motives of others (eg, by observing the direction of their gaze and their other actions); (2) forming empathy, that is, not only judging the other’s feelings from the outside but also feeling the emotion the other seems to feel; (3) forming moral and esthetic judgments about the actions and motives of others; and (4) becoming aware of how one’s own internal thoughts and emotions are influenced by others.

Signals that communicate emotional states have potent effects on social behaviors, although at present the best experimentally characterized example of the effect of emotional signals comes from outside the social domain. Faces, and most especially eyes, that express emotion can influence not only social emotions such as pride, shame, guilt, trust, or jealousy but also basic “survival emotions” such as fear, anger, happiness, and disgust. For example, fearful faces, which might signal a threat, activate the amygdala in functional imaging experiments (Chapter 15). Interestingly, a fearful face can still elicit amygdala activation and a change in the autonomic nervous system even if it is presented for a very short period and then followed by a “mask,” a neutral face. Subjects who participate in such experiments have no conscious recall of having seen the fearful face, despite their brains having clearly processed the information.

Much of our information on human social cognition is based on human brain injuries and on functional imaging with social tasks. As one example, the ventromedial prefrontal cortex appears to be activated when thinking about the mental states of both self and others, and patients with lesions of the ventromedial prefrontal cortex have deficits in social emotions such as shame, guilt, and empathy. Emerging evidence suggests that social cognition is highly dependent on prefrontal cortex, and on the integration of more purely cognitive processing with emotional information coming from the amygdala, hypothalamus, and ventral striatum. So-called mirror neurons have been proposed to be part of the neural basis of social cognition. Demonstrated in rodents and nonhuman primates, mirror neurons fire when an animal sees itself performing a task or another member of the species performing that same task. Mirror neurons are best characterized in the primate inferior frontal gyrus, although they have been observed in several brain regions. Much further work is needed to determine whether mirror neurons are in fact related to social cognition.

There is as yet little specific pharmacology associated with social cognition with the exception of experiments using the peptides oxytocin and vasopressin (Chapter 7). These are evolutionarily closely related nonapeptides that cross the blood–brain barrier when given intranasally. In different species both have physiologic functions based on their release from the posterior pituitary (neurohypophysis; Chapter 10) and central functions based on their actions within the hypothalamus, amygdala, and other structures. Within the brain, these peptides regulate anxiety and social behavior. In humans, vasopressin increases anxiety, acting on circuits within the amygdala (thus, vasopressin V_1a receptor antagonists are being studied as possible antianxiety agents), and oxytocin decreases anxiety acting on different cells in the amygdala. In rodent species, oxytocin is well known to increase social recognition, pair bonding, and affiliation (Chapter 7). In normal humans, oxytocin nasal spray has been shown to increase trust in laboratory-based tests, and its potential efficacy in the treatment of autism spectrum disorders is currently under evaluation. We are just embarking on the beginning of the exploration of the neural substrates and pharmacologic manipulation of social emotions, with all of the complex therapeutic and ethical implications waiting to be addressed.

Autism

Autism describes a spectrum of disorders in which social cognition is powerfully disturbed. It is characterized by abnormal social function, by language delay (or in severe cases absence of language function), and restricted interests and repetitive behaviors. Autism is highly genetically influenced—it is 80% to 90% heritable (see below)—and the diagnosis is made in early childhood. The disorder ranges greatly in severity from very-high-functioning individuals (Asperger syndrome) with normal language and intellectual ability to children who also suffer from intellectual disability (mental retardation), seizures, or other neuropsychiatric symptoms. Some individuals with autism have remarkable islands of cognitive strength.

The restriction of behaviors and interests in autism has been hypothesized to reflect abnormal functioning of the cortical-striatal-thalamic loops discussed above. Far more investigation is needed to confirm this theory. Whatever the pathogenesis, however, this aspect of autism can lead to significant impairment and can warrant both behavioral and pharmacologic interventions. Autistic people forced to break routine can be angry, anxious, and extremely stubborn. Severely affected individuals may have repetitive behaviors that are self-injurious, which are nonetheless extremely difficult to terminate. In these cases, when behavioral interventions fail, pharmacologic agents are used empirically, for example, SSRIs to control anxiety or low-dose second-generation antipsychotic drugs to control highly problematic behavioral outbursts. Current clinical trials are targeting several mechanisms that have been implicated in autism-related behaviors in animal models, such as oxytocin and insulin-like growth factor-1 (IGF1).

The core deficit in the social cognitive abilities of individuals with autism is thought to be an inability to form empathy. This deficit makes it difficult or impossible for the autistic person to attribute mental states to others and thus to make sense of their motives and actions, resulting in severe impairments of social interaction. Unfortunately, there are currently no pharmacologic treatments for this disabling aspect of autism spectrum disorders.

Recent work is beginning to identify genes that cause or predispose individuals to autism spectrum disorders. Loss-of-function mutations in MeCP2 (a methylated DNA-binding protein) cause Rett syndrome (Chapter 4), while loss-of-function mutations in fragile X mental retardation protein (FMRP) cause fragile X syndrome, which often presents with autism-like symptoms (Chapter 5). Growing knowledge of the physiologic function of FMRP has led to clinical trials of mGlu5 receptor antagonists or GABA_B receptor agonists in patients with fragile X syndrome, with mixed results reported to date. Mutations or copy number variations (CNVs) in genes encoding several synaptic proteins (eg, neuroligins, neurexins, shank3) are associated with a small number of cases of autism. Interestingly, neuroligins are postsynaptic proteins that bind to the presynaptic neurexins, interactions that are believed to specify neuronal connections in the developing brain. Likewise, shank3 is part of the postsynaptic density found at excitatory synapses and regulates synapse development and function (see 5–1 in Chapter 5). These advances are enabling better animal models of autism spectrum disorders; thus, mice with mutations in the aforementioned proteins show abnormalities in social behavior. It is hoped that such models will lead eventually to a better understanding of the neurobiologic basis of the disorders and to more rational and effective pharmacotherapies. Beyond these very rare genetic mutations of strong effect, many additional genes have been found in large genome-wide studies to be associated with increased risk for autism but to contribute only a very small amount to the overall risk for the syndrome. A surprising finding of this work is that some of the same genetic variations that contribute to risk for autism also do so for schizophrenia and bipolar disorder, syndromes with very different clinical features and age of onset (Chapter 17). These findings underscore the genetic complexity of neuropsychiatric disorders and the likelihood of a genetically driven diathesis for a range of brain abnormalities, the specific nature of which depends on many other genes and on nongenetic (environmental or random epigenetic) factors.