Section IV: Cognitive Neuroscience

OBJECTIVES

A fundamental goal of neuroscience is to understand how consciousness arises in the human brain and how brain damage alters or destroys it. However, studying consciousness is difficult. Consciousness is a subjective inner experience whose content is not measurable using scientific instruments. What can be measured are what are called “correlates” of consciousness—brain activities or activity patterns that occur when consciousness is present compared to the brain activity when consciousness is lost, such as during coma or sleep. If consciousness is defined as an introspective, linguistic-based, inner thought stream that exists only in humans but not in any other animal, then it can only be studied in humans. This ethically precludes virtually all invasive neurophysiologic recording and manipulation techniques except for notable exceptions such as invasive physiologic recordings carried out during epilepsy surgery.

Neuroscientists and philosophers do not all agree with the materialistic idea that consciousness is created by brain activity. Alternative explanations for consciousness range from quantum mechanics to dualistic religious traditions that hold that a nonmaterial soul is the real seat of conscious. In this chapter, we examine aspects of consciousness that we understand depend on neural activity in the brain, such as differences between highly conscious states like normal wakefulness versus sleep or coma, and alterations in consciousness resulting from brain damage.

HOW DID CONSCIOUSNESS ARISE ON EARTH?

Life began on earth about a billion years after its formation 4.5 billion years ago. This life consisted of unicellular prokaryotes such as bacteria. Roughly a billion and a half years later, eukaryotes arose, cells with nuclei. About a billion and a half years after that, complex animals arose during the Cambrian explosion 500 million years ago, giving rise in a few million years to primitive vertebrates. Mammals arose about 200 million years ago, and primates about 60 million years ago, after the Cretaceous dinosaur extinction. Several hominid lines arose in the last 5 million years, with humans, Homo sapiens, showing up a few hundred thousand years ago.

Brain size and, in particular, brain size in relation to body size have increased significantly in some vertebrates over the eons. Invertebrates such as insects and mollusks have concentrations of neurons called ganglia containing a few thousand neurons. These ganglia differ greatly from each other and bear little resemblance to the brains of vertebrates. The most primitive vertebrate brains, however, such as those of amphibians (eg, frogs) or reptiles (eg, lizards and turtles), have structures very similar to those in mammalian, primate, and human brains, except that nonmammalian vertebrate brains have little or no neocortex, a largely mammalian invention. Mammalian brains are similar to each other in overall structure, cell types, and circuits, being distinguished mostly by the amount and distribution of neocortex. Primates have even more neocortex than most mammals, and humans more than most primates. Somewhere along this evolutionary path, consciousness evolved, and most neuroscientists think it has something to do with the evolution and growth of the neocortex.

BRAIN STRUCTURES MEDIATING CONSCIOUSNESS

Although consciousness and high general intelligence clearly are associated with the neocortex, the existence of these mental abilities clearly still depends on lower brain structures such as the brainstem and thalamus, whose function is necessary but not sufficient for consciousness. In particular, an intact reticular formation is necessary for basic awareness in all vertebrates.

The Reticular Formation

The central nervous system consists of the brain, spinal cord, and cephalic sensory ganglia such as the retina. The earliest vertebrates possessed a complex spinal cord and cephalic sensory ganglia before their brains became very large. The spinal cord is not just a fiber tract; it is also a distributed microcontroller that receives sensory information from the muscles, tendons, joints, and skin and produces complex coordinated motor output for various types of locomotion, ranging from swimming to walking, running, and climbing. The spinal cord also transmits sensory information from sensors in the viscera of the autonomic nervous system that mediate control of respiration, heart rate, digestion, and blood oxygenation. This information projects upward to the phylogenetically oldest part of the vertebrate brain, the brainstem.

The brainstem also receives vestibular, visual, and other information necessary for balance and complex aspects of locomotion. A crucial central organizer of this information is the cerebellum, whose input/output tracts emanate from the pontine region of the brainstem. Half of all neurons in the brain may reside in the cerebellum. The reticular formation is a distributed brainstem neural network that integrates and executes information from both the autonomic nervous system and central nervous system. This distributed network of nuclei and tracts extends throughout the brainstem (medulla, pons, and midbrains) and upward to the diencephalon. The reticular formation regulates states of arousal and consciousness through both ascending axon tracts that modulate a large number of neurons in the brain and descending tracts that influence spinal reflexes and processing. Although the reticular formation is small relative to the rest of the brain, its function is crucial in controlling brain state and homeostasis. Damage to the reticular formation often produces death through loss of respiratory or cardiac control or results in coma due to loss of the ability to maintain an awake brain state. In contrast, it takes damage to much larger areas of neocortex to produce sustained complete loss of consciousness (Figure 19–1).

If we define (for vertebrates, at least) awareness as the brain state that permits active movement and response to the environment (as opposed to sleep) and consciousness as a state of awareness possessed only by humans, then the reticular formation is necessary for awareness, which itself is necessary, but not sufficient, for consciousness.

Overview of Cortical Structure

Above the brainstem is the diencephalon, consisting of the thalamus and hypothalamus. The reticular formation projects to the thalamus and modulates its function, whereas the thalamus reciprocally projects to the cortex, forming an even higher order brain system. The hypothalamus is a controller of the autonomic system, receiving projections from viscera sensors via spinal cord tracts and the reticular formation. The integration of autonomic brain states related to hunger, thirst, and reproductive goals with the brain’s motor control hierarchy is through a neural network called the limbic system.

A well-known theory about the evolution of the hierarchical organization of brain structure is the triune theory of MacLean. Although not correct in some details, it still provides a framework for linking the evolution of the brain with the function of major subsystems within it. The 3 components of the MacLean triune system are, in order of evolution, (1) the reptilian complex, (2) the limbic system, and (3) the neocortical system. The reptilian complex is composed of the spinal cord and brainstem, including the cerebellum. This is the dominant brain system in fish and reptilian vertebrates. Next in evolution came the limbic system. This system enables sensory discrimination and motor behavior to be more complex and have more dependence on learning. Figure 19–2A shows a sagittal view of the limbic system with a number of its main structures. Notably, structures such as the amygdala and hippocampus mediate the formation of memories that control future behavioral contingencies based on experience. Many neuroscientists today object to calling the hippocampus limbic because, among other reasons, the hippocampus connects extensively to neocortex, part of the third, neocortical system. Connections between several limbic structures and neocortex are shown in Figure 19–2B.

The neocortical system is the third and phylogenetically newest part of the brain. It composes the largest part of most mammalian brains but is small or minimal in nonmammalian vertebrates such as amphibians, reptiles, and birds. The neocortex is distinguished from 2 older types of cortex, the allocortex and mesocortex (Figure 19–2C), which are associated with the limbic system, by its structure and location. The phylogenetically oldest allocortex (itself composed of archicortex, paleocortex, and periallocortex) is associated with the hippocampus and related structures and some areas of the olfactory system. Allocortex has 3 or 4 layers, versus 6 in neocortex. Mesocortex has 3 to 6 layers and consists of (1) the cingulate cortex immediately above the huge corpus callosum fiber tract that connects the 2 cerebral hemispheres and (2) the parahippocampal gyrus that organizes input projections into the hippocampus.

The neocortex, which composed a small percentage of the total brain in early mammals, became the dominant brain structure in primates and humans. Its 6-layered structure has similar cell types and functional organization throughout the brain. Current theory for mammalian brain function generally postulates that mammalian ancestors prior to the Cretaceous extinction were primarily olfactory animals that used a small primitive neocortex to differentiate and learn complex smells. After the dinosaurs disappeared, mammals radiated into the now unoccupied niches by expanding neocortex for complex processing in visual, auditory, and other domains. An important consequence of adapting a similar neural processing structure for almost all brain processing was that information could then be easily transmitted across the entire brain based on a common underlying neural circuit representation. This intrinsically integrated brain function permitted a high level of unified awareness that then permitted (with other factors) consciousness to arise in humans, as the theory goes.

Functions of Major Neocortical Lobes

The 4 major lobes of the neocortex are shown in Figure 19–3. All mammals have a similar neocortical organization. Mammals considered highly intelligent, such as primates, have more neocortex than other mammals, much of which involves enlargement of the frontal lobes. Apes in general and humans in particular not only have larger brains than other mammals, but also have a larger prefrontal (anterior) area of the frontal lobes. The prefrontal area is where working memory and abstract planning are represented.

In general, the occipital lobe processes vision; the temporal lobe processes audition and visual form information and is important for memory; the parietal lobe processes somatosensory, visually guided movement and auditory location information; and the frontal lobe generates movement. The output of the frontal lobe is primarily via the primary motor cortex at its posterior extent. Posterior to the central sulcus and across from primary motor cortex is the somatosensory area of the parietal lobe.

Control by Thalamus

Buried underneath the neocortex are the thalamic nuclei. Virtually all sensory input projects to primary cortical areas via so-called relay nuclei in the thalamus. Motor control signals from the basal ganglia and modulation of motor commands via the cerebellum and tendon and joint receptors project through the thalamus. Neocortical areas that receive inputs from the thalamus tend to project back to those and other thalamic areas. Thalamic nuclei and the mesocortical cingulate gyrus are like the hub of a wagon wheel and interact with and control the entire neocortex. Damage to these areas has severe impacts on consciousness. Brain imaging studies also show that conscious brain activity always consists of activity in 1 or more sensory neocortical areas, the thalamus, and the prefrontal cortex simultaneously. This organization has been called the “vital triangle” by Cotterill.

The Association Cortices

Vertebrates have the largest brains and appear to be the most intelligent animals, mammals are the most intelligent vertebrates, primates are the most intelligent mammals, and humans are the most intelligent and only conscious primates. The neural basis of this chain is the enlargement of the brain, enlargement of neocortex, enlargement of the frontal lobe, and enlargement of prefrontal cortex. Human and chimpanzee DNA have about 99% overlap. Human brains are a little over twice the size of chimpanzee brains, with much of the added increase in the prefrontal cortex. What is this extra prefrontal neocortex doing that enables consciousness?

Is There Such a Thing as Association Cortex?

Early in the 20th century, tract-tracing histologic studies showed that particular areas of the occipital, parietal, and temporal cortices received direct thalamic input from visual, somatosensory, and auditory areas, whereas lower motor neurons in the spinal cord received direct projections from upper motor neurons in primary motor cortex in the frontal lobe. These came to be called “primary” cortical areas. The size of these areas generally follows the number of sensory receptors in the periphery for sensory areas or the number of lower motor neurons for motor areas. The cortical area sizes correspond to the acuity needed in the ecological niche occupied by various species. For example, moles that dig with their noses have enormous parietal representations of tactile receptors around their mouth and nose.

Beyond the relative cortical allocation based on ecological niche, mammals differ greatly in how much neocortex is not primary, that is, not directly connected to a sensory input or motor output. Early 20th-century behaviorist/associationist ideas about the brain supposed that it was a general, nondifferentiated learning apparatus, the details of whose specific circuit organization outside primary areas was of little consequence in understanding behavior. Nonprimary areas of neocortex were called “association” cortex. The idea was that although it was undifferentiated, the more you had of it, the better for learning and intelligence, and since humans had the most of it, that explained their superior intelligence.

Specific Features of the Association Cortices

At a late 20th-century brain meeting, Jon Kaas famously said that primate association cortex was rapidly shrinking. What he meant was best illustrated in the visual system. Visual thalamus, the lateral geniculate nucleus, projects to the pole of the occipital cortex, called primary visual cortex (V1; also striate cortex and Brodmann area 17). V1 projects to V2 (area 18), which is also visual, and V2 projects to V3, V4, V5, and so on, all with different visual response properties. The retina and visual thalamus respond to light and dark things of certain sizes, V1 responds to oriented edges, V4 to specific color patterns, and V5 to specific motion directions. These are all visual areas, not general association learning tissue.

The same is true in other sensory systems. High-level (as in more synapses away from primary) sensory areas tend to have neurons that respond to more “abstract,” less receptor-response sensory attributes. These higher order areas tend to remain sensory modality specific until very high-level multimodal areas are reached, such as in the parietal lobe, where neurons that are influenced by visual, auditory, or somatosensory input from particular areas of space are found. Neurons in primary motor cortex tend to drive specific muscles in a topologically oriented map of the body. Anterior to primary motor cortex are areas such as premotor cortex and the supplementary motor area whose activation drives groups of related muscles.

The one place in the brain where many neurons are found whose firing (as far as we currently know) has no obvious topologic relation to body location or location on a sensory receptor sheet is the prefrontal cortex. Rather, many neurons in the prefrontal cortex are thought to mediate working memory, forming a transient firing group that can represent any arbitrary constellation of features that exist in the environment that needs to be remembered as a contingency for future action. If we are to differentiate the conscious human brain from the very similar but nonconscious chimpanzee brain, prefrontal cortex is an obvious place to consider.

THE FRONTAL & PREFRONTAL LOBES

Figure 19–4 shows sagittal views of the 3 principle regions of prefrontal cortex and some of their subcortical connections. On the left is the anterior cingulate (AC), which is older mesocortex. The AC is a master control area that allocates processing resources among various neocortical areas. The center view in Figure 19–4 shows the orbitofrontal cortex (OFC), also called ventromedial prefrontal cortex, which occupies the ventral and medial portion of prefrontal cortex. This cortical area receives low-level sensory input in pathways that parallel the main sensory projections to the thalamus. The OFC interacts with the amygdala, and this system is a kind of emotional working memory that can represent a sensory constellation (eg, the appearance of a snake in the grass) to be associated with an emotionally salient response (escape). On the right of Figure 19–4 is the dorsolateral prefrontal cortex (DLPFC). This large brain area instantiates a working memory for representing sensory constellations arising from either external actual events or imagined, internally generated imagery. This leads to the question of how an increase in DLPFC and, therefore, working memory capacity is associated with consciousness.

Dorsolateral Prefrontal Cortex & Working Memory

Working memory is a concept that replaced the older concept of short-term memory, the ability to remember a list of several items such as phone number digits spoken or presented visually. Short-term memory lasts for only a few seconds without rehearsal, which begins the transference to long-term memory. Rehearsal depends on language, a fundamental component of human intelligence. Working memory extends the concept of short-term memory to include, among other things, the transference of information from long-term memory into a mental “scratchpad” by which this information can be quickly accessed. For example, if you were asked a series of questions about the first car you ever owned, you would bring into this scratchpad long-term memory information about the car’s make, color, and so forth. This would permit answering further questions about the car more quickly.

The DLPFC is the major site for the operation of working memory. DLPFC interacts with the hippocampus and other appropriate area of neocortex to store, activate, and maintain mental memory images. One idea about how a larger DLPFC might support a qualitative increase in intelligence underlying consciousness comes from a memory model postulated by Lisman and colleagues. Their model was motivated by the observation that 2 prominent brain oscillations, gamma and theta, typically occur in a ratio of about 7:1. In their model, each gamma oscillation is like a pointer that enables a particular constellation of neural firing that represents 1 memory item. The Miller limit of 7 memory items occurs because the theta cycle maintains the total memory representation, and only 7 gamma cycles that generate the neural code for 7 distinct items can be loaded or read out each theta cycle.

There are several implications of this model for intelligence and consciousness. Human gamma frequencies range from approximately 40 to 100 Hz, whereas the theta range is 4 to 8 Hz. The gamma/theta ratio varies somewhat in a particular person and between people. Evidence links better working memory to intelligence and indicates a correlation between the gamma/theta ratio and working memory capacity.

There is another, more speculative conjecture that can be made from this theory that suggests a function for a larger DLPFC. This has to do with what is called “chunking.” The human memory limit is about 7 items (plus or minus 2). However, the sets of items can be quite different things. One can remember approximately 7 individual letters. However, one can also remember approximately 7 words, which obviously contain more than 7 letters. Under favorable circumstances, one can remember approximately 7 short sentences. The items in the 7-item memory limit can be complex things as long as they are familiar and can be represented as a unit or “chunk.” What is a chunk, and how complex can it be?

In terms of number of items, human memory capacity is not significantly greater than that of many animals. Many studies have suggested that birds, with brains less than 1/100th the size of human brains, can remember up to 5 or 6 items. It is conceivable that some fundamental aspect of brain architecture, such as the gamma/theta ratio, limits the number of chunks that can be held in working memory for any vertebrate brain, but the complexity of the chunks increases with brain size, particularly the size of the DLPFC. The most complex chunks are enabled by language and linguistic categories.

The Orbitofrontal Cortex & Emotional Memory

The ventral part of the prefrontal cortex is called the orbitofrontal cortex (OFC) or ventromedial prefrontal cortex. Whereas the DLPFC is extensively connected to the hippocampus, the OFC has a similar relationship to the amygdala, which is instrumental in learning about situations of high “limbic” significance, such as those involving physical or social danger. Although most mammals may have in inbred fear of spiders and snakes, one must learn to be wary when encountering flashing red lights or particular facial expressions of one’s boss.

The output of the OFC is also qualitatively different from that of the DLPFC. DLPFC working memory is highly consciousness and language dependent, requiring attention, and is often associated with linguistic maintenance via subvocal rehearsal. The OFC, in contrast, signals the “sense” of danger or inappropriateness by what are colloquially called “gut feelings” of unease, termed by Damasio as somatic markers. Intuitive gut feeling are generated by a fast, subcortical system termed the low road by LeDoux whose output may begin generating avoidance behavior before conscious awareness of the source of the feeling. To paraphrase the famous William James quote: “I see a bear, I run, and then I am afraid.”

The OFC system exists in all mammals, but it is particularly highly developed in social mammals such as primates, presumably to deal with complex aspects of social rank in primate species. Enlargement of the OFC in humans, compared to other primates, increases the sophistication and differentiation of environmental cues associated with threat or risk. Driving a car is an example of a situation of constant, largely unconscious management of responses to a host of environmental cue threats ranging from the behavior of other cars to road signals. These threats are obeyed below the level of conscious awareness, with the sophistication of threat assessment and response enabled by expansion of the OFC.

The Anterior Cingulate Cortex

The cingulate cortex, although phylogenetically older mesocortex, remains an important controller of neocortical processing. This is particularly true of the AC. This structure receives inputs from most of the neocortex and the thalamus. Its most important output is to the entorhinal cortex, part of the hippocampal formation. The AC compares goals, set largely by the DLPFC, to current progress toward those goals, as relayed by sensory information controlled by the thalamus. Cells in the AC are activated by task difficulty and errors. The AC is also instrumental in the perception of pain and the anticipation of pain. Lesions of this area reduce pain perception.

LANGUAGE, LATERALITY, & CONSCIOUSNESS

There are numerous difficulties associated with the scientific study of consciousness. The most discussed of these, of course, is the fact that consciousness is a personal experience not directly accessible to instrumental measurement. The best neuroscience can do at present is to look at the various borderlines between neural states or systems that have it, versus those that do not. This includes, primarily, differences between conscious humans and near relatives that are not conscious, and between conscious and unconscious states in humans.

Language & Consciousness

The biggest problem in determining the substrate of consciousness based on species differences is that there is only 1 species, humans, that is generally agreed to have full consciousness. Of all the differences between conscious humans and all other nonconscious species, the 1 difference that stands out is language. As Helen Keller said, “Before my teacher came to me, I did not know that I am. I lived in a world that was a no-world. I cannot hope to describe adequately that unconscious, yet conscious time of nothingness. I did not know that I knew aught, or that I lived or acted or desired. I had neither will nor intellect. I was carried along to objects and acts by a certain blind natural impetus. I had a mind which caused me to feel anger, satisfaction, desire.”

Consciousness is highly impaired without language, which all typically developing humans acquire purely by exposure with little explicit instruction. Some nonhuman primates can learn hundreds of words and even generate novel word pairs, but none have mastered the grammar of any language, even with extensive instruction. It is hard to argue against the proposition that human consciousness is derived from language.

One thing that language might contribute to consciousness is the ability to have episodic memory. Episodic memory is distinguished from semantic memory in that episodic memory and learning can arise from single events, whereas semantic memories are typically associated with the world knowledge semantic system in the brain established mostly through repetition and association. Episodic memory allows single-trial learning and learning by observation and verbal instruction. It must depend significantly on the ability to categorize experience through language and generate rules from categorized experience that help to make predictions about future outcomes. Many neuroscientists would argue that no animals other than humans have episodic memory. In this scheme, language-enabled episodic memory is a significant element of human consciousness. The planning ability it supports has obvious survival utility. Evolution has not produced any nonconscious human zombies lacking language; all normally developing humans have language, episodic memory, and consciousness.

Left & Right Brains

To the extent that human consciousness depends on language, it must also depend on the distinct capability of the left versus the right side of the brain (where most control of language resides in virtually all right-handers and a majority of left-handers). Yet, no significant functional differences are known to exist between the right and left hemispheres in overall structure, cell types, or neural circuit architecture. In fact, the right hemisphere competently develops normal language in some left-handers and in many cases of left hemisphere damage in infancy.

Elegant studies of the difference in language competence between the right and left hemispheres were done by Sperry, Gazzaniga, and colleagues in so-called “split-brain” patients. These were people in whom the large fiber tract connecting the 2 hemispheres (the corpus callosum) was transected to prevent the spread of epileptic seizures between the 2 hemispheres. Visual stimuli were presented to either the left or the right visual field for intervals too short for any eye movements to occur, which isolated these stimuli to the right or left side of the brain, respectively. These patients could not make verbal reports about stimuli isolated to the right side of the brain. Moreover, the ability (intelligence) of the right side of the brain to integrate aspects of visually presented stimuli into a coherent whole was very deficient compared to that of the left hemisphere.

Gazzaniga concluded from these studies that the isolated (by callosal transection) left hemisphere, which is about the same size as a chimpanzee brain, has normal language competence and consciousness. The isolated right hemisphere was deficient in language and in any many aspects of consciousness. Therefore, brain size alone cannot be the cause of the consciousness difference between humans and similar species such as chimpanzees. Some developmental process that occurs universally in humans, but only in humans, enables the left side of the brain to produce language and consciousness.

Sleep

Sleep is the daily transition between consciousness and unconsciousness that virtually all humans experience daily. Sleep itself is complex, composed of rapid eye movement (REM) and non-REM categories. Non-REM sleep is further divided into 4 different stages that differ in electroencephalogram (EEG) brain waves and physiologic attributes such as muscle tone. Most dreams occur during REM sleep, during which brain waves resemble those during wakefulness, including activity in motor cortex, although actual movement is suppressed via the parabrachial pathway.

Sleep and sleep cycles are controlled by the brainstem reticular formation (particularly the pons parabrachial area), several nuclei in the basal forebrain, and the hypothalamus, via the neurotransmitters acetylcholine (ACh), norepinephrine, and serotonin. The release of ACh, which occurs during REM sleep, increases EEG amplitudes but decreases EEG synchronous activity. The neurotransmitters norepinephrine and serotonin are also involved in the control of sleep. Norepinephrine is released by neurons in the brainstem locus coeruleus nucleus whose axons project widely (but diffusely) throughout the brain. Norepinephrine release increases cortical activation and EEG enhancement but, unlike ACh, does not induce REM sleep. Serotonin is released from axons whose somas are in the Raphe nucleus in the brainstem. Like norepinephrine, serotonin is important for the maintenance of the stages of non-REM sleep.

REM sleep EEG activity is so similar to that of normal waking consciousness that REM sleep is sometimes called paradoxical sleep. The thalamus and other subcortical structures control the progression of neural activity across the brain during sleep, but with brain activity generated more by internal sources than sensory input, which is highly attenuated during all stages of sleep. Sleep, particularly REM sleep has been hypothesized to have restorative and memory consolidation functions. REM sleep appears to be particularly important for the consolidation of memory because REM sleep deprivation has negative effects on learning. After moderate sleep deprivation, the proportion of REM to non-REM sleep increases. Prolonged sleep deprivation can produce psychosis, a disturbance of normal consciousness.

NEUROLOGIC DEFICITS & CONSCIOUSNESS

Another way of investigating the neural basis of consciousness is to examine how brain damage affects particular aspects of consciousness. Several disorders of visual consciousness give some interesting clues about the neural circuitry that supports it.

Blindsight

There are many things humans see, in the sense of processing information from visual input, of which we have no awareness. We are not aware of the blind spot in each eye where there are no photoreceptors at the optic nerve head. We are not aware of the letter configuration at the locations to which our eyes saccade during reading, despite the fact that these saccades are optimally accurate. Conscious vision is mediated by retinal projections to the thalamus, whereas the retinal projections to the superior colliculus and other retinal recipient zones process information in an unconscious manner.

A particularly interesting dissociation of conscious from unconscious visual processing is that of so-called “blindsight,” studied by Weiskrantz. Patients who have extensive loss of primary visual cortex in the occipital lobe (V1, or striate cortex) report no awareness of stimuli presented in the corresponding visual field. However, when forced to guess whether an object was presented there or move their eyes to the object location, they perform far better than chance. Thus, behaviorally, they can use visual information presented in their blind hemi-field despite having no conscious awareness of any stimuli presented there.

It is interesting that humans alone are especially dependent on the thalamus–occipital lobe pathway for vision. Other primates with similar lesions in the V1 occipital cortex exhibit far less visual dysfunction than humans, with a loss of some visual acuity and form discrimination but less symptoms of actual blindness.

Neglect

Neglect is another type of loss of visual consciousness, usually due to damage to the right parietal lobe (which receives input from the left visual hemi-field). Patients with neglect tend to ignore objects on their left side, shaving only the right side of their face, for example. This is despite the fact they can freely move their eyes to place objects in the nonneglected right visual hemi-field. When asked to describe from memory a familiar place from a particular point of view, patients with neglect may also fail to describe any objects that would be seen on their left side. However, when asked to describe the scene from the opposite point of view, they neglect objects they described previously that now would appear in their left hemi-field.

THEORIES OF CONSCIOUSNESS

There are nearly as many theories about consciousness are there are neuroscientists and philosophers. The following discussion samples a few ideas that have recently received considerable attention or that can be related to psychophysical or physiologic data. The section on the quantum brain discusses some consciousness theories at the intersection of neuroscience and physics.

Priming & the Unconscious Brain

Most people have some idea of the existence of priming from the phenomenon of subliminal advertising. The most well-known example is of the insertion of 1 frame out of every 24 frames per second in a movie that says, “buy popcorn.” The briefness of the presentation (<50 milliseconds) is such that people are not consciously aware of its existence. Nevertheless, in carefully controlled psychophysical experiments, it has been shown that such presentations have small but statistically significant influence on subsequent choices or the speed of processing of prime-related versus unrelated stimuli.

One of the most interesting aspects of this type of experiment is that because the prime image is a short text phrase, it must have been processed through the language system to generate semantic meaning, even though the subject is not conscious at all of its existence. Priming is just one example of data that indicate that most human brain processing is unconscious. Which of all the underlying unconscious brain activity rises to consciousness depends on both top-down and bottom-up attention mechanisms. These mechanisms must maintain neural activity for at least about half a second in the vital triangle of the thalamus, sensory cortex, and frontal cortex as a prerequisite of conscious awareness.

Illusory Conjunctions & Working Memory

Ann Treisman and colleagues performed a number of experiments whose basic format is as follows. A set of 3 letters each of a different color, such as a red X, blue S, and green T, are flashed briefly on a projection screen. The stimulus duration is adjusted to obtain a criterion level of correct reports by the subjects. The experiment is then changed by the addition of a “mask” or noise figure immediately after the set of letters. The mask appears to “erase” the early visual system eidetic representation of the stimulus. A type of error now occurs that did not occur in the first experiment, which is that subjects now sometimes report letters correctly that were in the display, but with the wrong color (that of a different letter). This switching of the colors in the report between letters in the display is called an illusory conjunction.

Illusory conjunctions of this type are interesting because we know that the visual system processes some aspects of form and color with different neurons in different brain areas. That is, there are visual neurons in cortex that respond well to particular colors but are agnostic about shape, and visual neurons that respond to shape independent of the color of the shape. Thus, in this experiment, the stimulus activated color neurons that respond to red, blue, and green and shape neurons that respond to the X, S, and T shapes. The DLPFC maintains some “neural image” of this activity after the stimulus flash. How does the brain know which activated color neurons go with which activated letter neurons in order to report it correctly after the mask? This has been called the “binding problem”: how activity in different areas of the brain associated with multiple objects can be sorted or bound so that encoded properties of each object are associated with that object and no others. This ability clearly depends on conscious working memory. One mechanism proposed for mediating this type of working memory is synchronous firing.

Synchronous Firing & Brain Oscillations

Most of the discussion in this chapter has been about the location of brain substrates that support consciousness. However, few neuroscientists believe that consciousness is located in any specific part of the brain or is dependent on any particular type of neuron. Consciousness is a process for which some particular brain areas appear to be necessary, but not sufficient. Considerable brain activity can persist under the influence of many anesthetics, but the pattern of activity no longer supports consciousness.

Sensory neurons produce trains of action potentials whose pattern is modulated by the presentation of some stimulus. It is well known that such patterns are noisy; that is, repeated presentations of the same stimulus produce slightly different spike patterns with each presentation. This suggests that the spike response of sensory neurons consists of a stimulus-associated component that is perturbed by random noise that varies for each specific presentation. The implication of this is that the information in neural spike trains must be in the average overall rate of firing that corresponds to the stimulus-generated component, with the exact spike pattern conveying little information because, if the spike pattern is not the same for repetitions of the same stimulus, it is unlikely to enable distinguishing between different similar stimuli.

This view of spike pattern irrelevance changed with the finding from dual- and multi-electrode recordings that, depending on the stimulus and attention state of the nervous system, sometimes neurons fire synchronously despite the noise. Synchronous firing means that the spikes in one neuron occur at about the same time as those in another neuron more often than chance would predict due to the overall mutual firing rates. Synchronous firing could solve the binding problem if, during different cycles of an ongoing brain EEG rhythm such as the gamma rhythm, the constellation of neurons that encoded properties of 1 object fired synchronously on 1 particular cycle, or cycle phase, whereas those associated with other objects fired synchronously on different cycles or phases. Attention and consciousness could then be mediated by control of neural synchronous firing across large areas of the brain with readout neurons being sensitive to spikes from multiple inputs occurring synchronously. The existence of stimulus-dependent and attention-dependent neural synchrony is beyond dispute, but the idea that synchrony mediates attention or consciousness is controversial. Some neuroscientists have suggested that firing epochs called bursts, characterized by very high firing rates for tens of milliseconds, mediate attention and conscious control effects, with synchrony being a by-product of burst firing.

The Quantum Brain

There has been considerable theorizing in the past 2 decades about the relationship between consciousness and the outcome of some quantum mechanics experiments. The fact that decisions made by conscious observers determines the outcome of quantum experiments has led to the suggestion that consciousness itself creates reality and is not reducible to, or identical with, the neural firing patterns going on in the brain.

The argument is best illustrated with the double-slit experiment, as pointed out by Feynman. Light can behave as a wave. If light passes through 2 closely spaced slits, an interference pattern will exist on some detector after the slits that is predictable from the wavelength of light and slit widths and spacing. However, if the intensity of the light is reduced until only 1 photon at a time is sent toward the slits, the pattern of hits on the detector after multiple single-photon transits is still an interference pattern. Did each single photon go through both slits and interfere with itself? According to current quantum physics, a probability function, described by the Schrödinger wave equation, passes through both slits, and the photon exists in a superposition of all allowed states until observation collapses the wave function into a particular outcome. The observation occurs in the system that consists of the mechanical detector and the human observer. The question is, where in this system does the collapse occur? Is it when the wave function interacts with the detector, or when the result of this interaction is consciously observed?

The difficulty is embodied in the Schrödinger’s cat paradox. A cat is placed in a sealed box with a quantum device that has a probability of one-half of releasing a poison gas and killing the cat in the next hour. After the hour has passed, is the cat dead or alive before the box is opened? The mainstream so-called Copenhagen (Bohr) interpretation of this situation is that the cat exists in a live/dead superposition until the box is opened and the state of the cat is observed by a conscious human.

Physicists attempted to resolve the single photon passing through 2 slits paradox by using another detector, such as a polarizer, to tell which slit the single photon went through. The existence of the which-slit detector, however, eliminates the interference pattern at the photon detector screen. If a measurement can be made of which slit the photon passed through, the impact pattern is what would be expected if the slit that the photon did not pass through were not present. An experiment can be set up in which a which-slit measurement is made after the photon has passed through the slits, and this measurement still destroys interference pattern. The observation by a human of some quantum properties of a quantum system can alter the state of the system everywhere, including in its past. In so-called entangled systems, a consciously observed measurement in 1 place on 1 particle can change the state of an entangled particle at the far end of the universe instantaneously.

Given that the measurement collapses the wave function, the debate in physics is over where the collapse occurs. Most physicists believe that the detector causes the wave function collapse by a process called decoherence. However, because the various parts of the detector apparatus must obey the laws of quantum mechanics, it is a problem to decide where in the detector apparatus the decoherence occurs, and a significant number of physicists place this point in human consciousness. Theories about how this might occur in brains, such as in neural microtubules, are too complex to discuss within the space available in this treatment. Almost no one understands or is happy about this situation. As Feynman supposedly said, “Anyone who says that they understand quantum mechanics does not understand quantum mechanics.”

Consciousness & Free Will

Two of the most fundamental questions in all of science are as follows: (1) Is consciousness causal or merely an epiphenomenal by-product of mechanistic brain operation? (2) Does free will really exist? The implications of the answers to these questions extend far beyond science into religion, ethics, law, and societal governance. These questions are intimately related because, if consciousness is not causal but an epiphenomenon, it is hard to see how there could be such a thing as free will. Until the late 20th century, these topics were almost taboo for most neuroscientists, who preferred to work on problems that were directly approachable or related to human mental and neurologic health. However, now that neuroscientists think they have a basic understanding of the principles of brain operation, the debate has ensued about whether the brain is just a “machine made of meat” whose operational principles are generally understood and whose behavior should be as predictable as that of any machine.

Deterministic arguments that deny the existence of causal consciousness and free will are generally motivated at least partly by a reaction against the Western dualistic ideas of the existence of a soul that inhabits and controls a mechanistic body. In Western religious traditions, the soul is the possessor of free will, while the body is an organic machine in humans that is similar to that in animals, except they lack a soul. If one does not believe in a soul, what is left is a machine. The feeling of free will has been postulated to be an evolutionary trick that motivates us to feel good about some decisions (completely determined by the machine) that promote survival value and bad about decisions that do not promote survival.

Those who argue against consciousness being causal also tend to view the principles of the operation of the brain to be grounded in Newtonian, pre-quantum physics, which views the universe as being filled with particles and forces whose future unfolding proceeds entirely according to a set of physical laws. The state of any physical system should be, at least in theory, entirely predictable from knowledge of its current state and physical laws that govern its evolution from the current state. In this scenario, there can be no free will or any causal role for consciousness. However, ironically, modern quantum physics does not actually view the universe this way, and as discussed earlier, some quantum physicists assign consciousness to a primary role in the creation of reality.

Another objection to deterministic theories of brain function that exclude free will and causal consciousness is that they are based on ideas of linear causation that are appropriate for describing simple systems but not complex dynamic systems characterized by extensive feedback and self-modification from experience. Much of the nervous system consists of loops in which firing of cell A drives cell B, which drives cell C, which feeds back to cell A. Of course, if these connections were all excitatory, this system would blow up, as in epilepsy. However, operating under normal homeostatic control, the locus of causality in this loop is difficult to assign. Feedback systems generated by completely deterministic rules can also operate in a chaotic manner such that tiny differences in the initial conditions can produce large differences in subsequent behavior.

Some neuroscientists advocate a middle ground in which the brain is regarded as a meat machine operating according to physical laws, but something like free will exists as an emergent phenomenon enabled by consciousness. Consciousness has causal efficacy via its enabling of decision making using categorization, episodic memory, and the ability to calculate future outcomes explicitly, which are all enabled by language. The brain constructs, through experience and previous decisions, its own architecture for using information to predict the future and make decisions based on that information. Conscious introspection is a significant part of this architecture.

The essence of science is the ability to make accurate predictions from data. No one doubts that biology depends on chemistry, and chemistry depends on physics. However, in the search for the DNA mutation that causes some neurologic disorder, neuroscientists typically do not look for the answer to physics, or even chemistry. DNA is part of an information processing system whose operation is best understood in terms of information coding rather than its chemistry (although the mutation may have been a result of chemistry, such as a teratogen). Brains that possess consciousness act on information itself. You choose the lumber size for the support posts for your deck based on a calculation of loads and tensile strengths. You do not know the result of the calculation before you do it. Perhaps your action was determined from the beginning of the universe at the big bang, but believing this does not seem to be of any use in guiding decisions you actually make.

The informed decision point of view of consciousness being causal does not depend on postulating the existence of a nonphysical soul. The human agent who acts is a biologic mechanism that obeys the laws of chemistry and physics. Choice is certainly constrained by the biologic hardware that executes decisions because drives and motivations are built into the hardware by biology and personal and species evolution. However, choices also depend on information stored in the brain and its use for the evaluation of likely outcomes. Failing to use information to consciously make decisions results in poorer decisions. Failing to understand the role of conscious introspection in human behavior reduces the scientific ability to predict human in it.

Choice is not the exclusive domain of humans. A dog can be trained to do and not do certain things. Dogs appear to experience conflict in situations where their basic instincts and their training conflict. They hesitate in those situations before making choices. The difference is that when a dog chooses, it is not aware of the explicit outcomes of the choice or the factors that went into the choice, other than an association between one possible behavior and reward or punishment, whereas humans are aware of episodic history and can foresee the future.

The idea of free will as the conscious introspective processing of information is admittedly a functional and operational definition of it. The issue for most neuroscientists is not whether there exists an unconstrained freedom to make any choice whatsoever but, rather, whether it is of predictive power to consider the chooser’s conscious calculations in understanding what choice is made (behavior). One could imagine making 100 identical robots with a variety of goals embedded in their programming but whose programming was modifiable by experience. If these robots were turned out to roam the world, one could imagine at some particular time some particular choice would be made by half of them, and the opposite choice by the other half, because of the differences in their history. If they had language modules, we would find it more instructive, in terms of predicting their future behavior, to ask them why they made the choice they did, than to record their entire experience to reconstruct the changes in the operation of their neural net programming. Presumably, the robots would use the results of their decisions to make better future decisions based on their “belief” in their own consciousness and free will. Free will may be an illusion, but it is one that can have considerable explanatory and predictive power.

The moral principles related to values, conscience, and societal norms are called ethics. Belief in the brain as a deterministic machine has implications for ethics because, obviously, if there is no free will it is difficult to see how to make people responsible for their choices. With or without free will, if a human is demonstrably a threat, such as by committing a number of murders, there will be little controversy about society acting to remove the threat via imprisonment, although the harshness of the punishment may depend on the assessment of the mental state and mental competence of the killer. Killings may be attributed to a free-willed decision to kill for some gain (during a robbery), or the killer may be a psychopath unable to control killing. The psychopathy may have been inborn (genetics), the result of acquired brain dysfunction (a tumor in the amygdala), or psychologically acquired (parental abuse).

Difficult ethical questions arise related to neuroscience and physiologic psychology in the case of risky behavior. It is not illegal, per se, to be intoxicated. Fifty years ago, arguments were sometimes made in court that someone should be less harshly punished for a traffic accident because they “had one too many” and were not really responsible for their subsequent actions. Now, the opposite view is the norm. Driving while intoxicated is itself a crime punishable by loss of one’s driving license (removal) and, in some cases, incarceration (punishment).

Do lower grade risky behaviors that are habitual reduce the ability to exercise free will and justify legal action? Does playing violent video games lead to violent behavior? Is smoking marijuana the gateway to more dangerous drugs? Does watching pornography lead to the commission of rape? There are neuroscientific components to all of these questions concerning whether any habit leads to any subsequent unacceptable behavior, and if so, in which people and under what other circumstances. Should bad habits that are associated with subsequent unacceptable behavior be legally sanctioned? Choices people make are caused not only by the hardware of their brain, but also by the hardware configuration modified by education, experience, knowledge of law and societal norms, and anticipated future consequences. Although the brain is a machine that operates on principles of chemistry and physics, the existence of ethics depends on the assumption that choices can be and should be made at the level of conscious knowledge. This knowledge includes the ethical norms of the society about making choices. Most neuroscientists would say that the phrase “free will” is a construct that is useful for explaining and predicting behavior. How free one’s will is in any particular situation depends on parameters that are at least partly amenable to neuroscientific investigation.

SUMMARY

• Brain activity is highly correlated with consciousness, but philosophers and religious traditions disagree about the causal relationship between the two.

• Intelligence appears to be generally correlated with the ratio of brain size to body weight.

• Although conscious capabilities appear to reside mostly in the neocortex, the integrity and operation of subcortical structures such as the reticular formation are necessary to sustain consciousness.

• A common evolutionary scenario thought to underlie the rise of consciousness is the triune theory in which nonmammalian vertebrates evolved the brainstem “reptilian complex,” mammals added the limbic system and rudimentary neocortex, and primates evolved a significantly larger neocortex with particularly enlarged frontal lobes.

• Current research suggests that previous labeling of large neocortical areas such as the “association cortex” is misleading because much of this area is now known to be sensory or motor.

• The closest approximation to the role of association cortex in the brain is the prefrontal area of the frontal lobe.

• The DLPFC is crucial for the instantiation of working memory, a necessary component of consciousness.

• Ventromedial prefrontal cortex underlies representation and learning of social and emotionally salient contingencies whose valence and intensity produce “gut feelings” rather than rational verbal narratives.

• Given that Homo sapiens is the only species on earth thought to have true consciousness and the only species to have true language, most neuroscientists believe these facilities are strongly linked.

• The 2 brain hemispheres differ significantly in their capabilities, with the left being predominant in most individuals for language and the right being superior at spatial processing.

• Different stages of consciousness, or its lack, range from deep sleep to hypervigilance, controlled by brainstem mechanisms.

• Some specific neurologic and psychiatric disorders can be related to dysfunction in specific brain areas.

• Theories for the existence of consciousness range from brain size to synchronous neural firing to quantum mechanical wave functions, with none being universally accepted.

• By their nature, different theories for the origin of consciousness also differ regarding the existence and causality of free will.

OBJECTIVES

OVERVIEW OF MEMORY & MEMORY SYSTEMS

One of the essential cognitive functions, memory involves the mental processes by which information is encoded, stored, and retrieved. Memory is thought to involve physical changes in the brain, called engrams or memory traces. In cognitive neuroscience, memory is defined not only by encoding and storing of the engram, but also through expression, which refers to the retrieval of stored information that leads to modification of behavior or conscious thought. Memory allows us to accomplish everyday physical and living activities and is critical for other key cognitive processes including language, planning, reasoning, and problem solving; it is also a fundamental component of our personality and consciousness.

Memory is the basis for the amassment of knowledge, the importance of which is exemplified by Francis Bacon’s quote “Knowledge itself is power.” Diminished learning and memory occur in normal aging and are common features of intellectual disability, traumatic brain injury, schizophrenia, Parkinson disease, and Alzheimer disease. Based on its central role in human behavior and disruption in many brain disorders, memory has been studied extensively in the fields of psychology, neuroscience, and clinical medicine. In this chapter, the various memory systems will be elaborated. Complementing this chapter is Chapter 10, “Synaptic Plasticity,” in which synaptic plasticity mechanisms implicated in encoding of memory are described.

In the broadest sense, memory can be categorized into collective memory and individual memory. Collective memory is a shared pool of knowledge by members of a social group, usually requiring symbolic representation such as language, and often studied in the discipline of sociology. The modern concept of human individual memory (hereafter memory) involves characteristic features. First, memory encompasses at least 4 components: acquisition, consolidation, storage, and retrieval, also called recall. Consolidation involves processes that stabilize a memory trace after its initial acquisition; learning consists of acquisition and consolidation and is also called encoding. Learning is a graded phenomenon, and most forms of memory can be either short lasting or long lasting. Second, memory involves multiple memory systems that depend on the specific type of information processed. Consequently, memory can be classified according to 2 features: the time course of learning and storage and the nature of the information learned and stored.

Focusing first on the time course, human memory involves informational processing systems with distinct temporal components. The idea that there are temporal forms of memory was expressed by William James and others in 1890, who distinguished between primary and secondary memory. The “modal” or “multistore” model of memory, published in 1968 by Atkinson and Shiffrin, proposes a structure of memory composed of sensory memory (also called a sensory register) and a dual memory store that includes short-term memory and long-term memory (Figure 20–1).

Sensory memory involves initial sensing and processing of chemical and physical stimuli from the outside world, has a relatively high capacity, and is very short lived, on the order of a few hundreds of milliseconds to seconds. Short-term memory is defined as a short-term store of pertinent or goal-directed information; it depends on attention, the cognitive process that involves selectively concentrating on 1 aspect of the environment while ignoring other aspects. Short-term memory has a limited capacity and a lifetime of a few tens of seconds to minutes. In 1974, Baddeley and Hitch proposed a revision to the short-term memory model, transitioning to the concept of working memory. Working memory encompasses short-term memory and serves as both an encoding and retrieval processor, with an active maintenance of information in short-term memory storage and processes to manipulate information in the store. Long-term memory is the permanent store where information is retained for extended periods, from months to years to decades.

Focusing next on the nature of the information encoded, long-term memory is typically divided into declarative and nondeclarative types, also known as explicit and implicit memory, respectively, with several subtypes in each category (Table 20–1). Declarative memory includes semantic memory and episodic memory. Semantic memory includes memory of facts and information that is encoded with specific meaning. Episodic memory refers to experience and events that are encoded along spatial and temporal dimensions. Declarative memory usually involves conscious encoding and conscious recall and is also called explicit memory because it consists of information that is explicitly stored and retrieved. Nondeclarative memory includes memory for procedures, tasks, and skills, as well as emotional and conditioned responses. Nondeclarative memory is also called implicit memory and can be consciously or unconsciously encoded but is usually unconsciously recalled. These different memory systems can operate independently and in parallel, which allows conscious and unconscious memory systems to operate simultaneously and maximize memory throughput of the brain.

Different functional areas in the brain are involved in encoding and storing different types of information (Figure 20–2). Sensory memory involves primary, secondary, and association sensory cortices for each modality. Short-term memory and working memory are thought to involve predominantly the prefrontal cortex (PFC), but also include functions of the parietal and medial temporal lobes (MTLs) and basal ganglia. For long-term memory encoding, the majority of neuroanatomic regions that are associated with specific memory systems have been identified using clinical approaches. The hippocampus is involved in declarative and spatial learning, whereas the amygdala is involved in emotional memory. The basal ganglia and cerebellum are involved in procedural learning. Much less is known about the neuroanatomy of memory storage. In general, memories are thought to be stored in distributed neocortical circuits.

SENSORY MEMORY

Defined as the first component of the modal model, sensory memory involves the acquisition of sensory information obtained by sensory receptors and processed within the sensory cortices in the brain. Sensory memory, also called the sensory register, provides a moment-to-moment snapshot of an individual’s overall sensory experience. Sensory information is stored for only a few hundreds of milliseconds to seconds, allowing a brief but high-resolution impression of the sensation, just long enough for pertinent information to be conveyed to short-term memory. Sensory memory is specific for each of the senses, of which humans have 5: vision, hearing, smell, touch, and taste; the sensory stimuli are called sensory modalities.

Common features have been identified for all forms of sensory memory studied. First, information stored in sensory memory is modality specific. The visual sensory store is called iconic memory; the auditory store is echoic memory; and the somatosensory store is haptic memory. Olfactory memory and gustatory memory have also been identified but are not as well characterized. Second, the formation of a sensory trace is only weakly dependent on attention to the stimulus, and the duration and capacity of sensory memory are outside of cognitive control. Third, the sensory stores are very brief in duration, on the order of about 0.2 to 2 seconds, depending on the modality. Fourth, the sensory stores have a relatively large capacity, facilitating details that provide high resolution.

The sensory memory engram or trace is thought to involve the transient activation of sensory circuits underlying that modality, which occur within the primary, secondary, and association sensory cortices for that modality. A subset of information can be transferred from the rapidly decaying sensory memory into short-term memory via the process of attention. This filters sensory stimuli such that only relevant information at any given time will be transferred into short-term memory. Thus, the function of sensory memory is to provide a detailed representation of our sensory experience, from which specific relevant or goal-oriented information is extracted into short-term memory and processed.

SHORT-TERM & WORKING MEMORY

Short-term memory maintains current representations of goal-relevant information in a temporary store. In the modal model, movement of information from sensory memory into short-term memory requires attention, which can be either bottom up, stimulus driven, or top down, goal oriented. Short-term memory has a limited capacity of about 4 to 7 pieces, depending on the nature of the information stored, and a limited duration, on the order of tens of seconds to minutes. The capacity and duration of short-term memory can be enhanced. Chunking, the process of organizing material into meaningful groups, can increase the capacity of short-term memory. Rehearsal, the ongoing repetition of the information held in short-term storage, can extend the duration. Thus, in its original conception, short-term memory is the capacity for holding, but not manipulating, limited amounts of information in the mind in an accessible store for short periods of time.

The term working memory was coined by Miller, Galanter, and Pribram in the 1960s in the context of the mind operating like a computer, but it was not until the 1970s and 1980s that Baddley and Hitch proposed a model that defines working memory in its current structure (Figure 20–3). In their scheme, information in the short-term memory store can be dynamically organized and manipulated; addition of this functional role is a key feature of working memory. Working memory handles information that enters not only from sensory memory but also from recently recalled long-term stores. Hence, working memory is bidirectional, serving as both an input and output device, for the holding and processing of new and already stored information, for the purpose of cognitive tasks.

Another feature of working memory is that it contains at least 2 storage subsystems—one for visual and spatial information, called the visuospatial sketchpad, and another for verbal information, called the phonological loop. These subsystems are under the control of a component called the central executive. The central executive enables working memory systems to selectively attend to some stimuli and ignore others and coordinate cognitive processes especially when more than 1 task is simultaneously performed. In 2000, Baddeley extended the model by adding another component, the episodic buffer, which holds representations that integrate phonologic, visual, and spatial information, and possibly information not covered by those subsystems, such as semantic and musical information. The episodic buffer also provides the link between working memory and long-term memory.

Many cognitive scientists and psychologists today use the concept of working memory to replace or include the previous scheme of short-term memory, emphasizing the functional role of working memory in manipulating and using information in the store, compared with the storage role of short-term memory. For simplicity, hereafter the term short-term working memory will be used to refer to both the temporary storage and manipulation of information in the store. Short-term working memory has received substantial attention over the past few decades. Impairment of short-term working memory is observed in several brain disorders, including attention-deficit/hyperactivity disorder, Parkinson disease, and Alzheimer disease. In normal cognition, short-term working memory, together with attention, plays a major role in the processes of thinking. Short-term working memory facilitates planning, comprehension, reasoning, decision making, and problem solving. Consequently, measures of short-term working memory capacity are strongly related to performance in other complex cognitive tasks, and in fact, working memory is a better predictor of academic success than is intelligence quotient (IQ). Finally, short-term working memory is fluid and can be enhanced by mental training exercises.

There is consensus among neuroscientists that the PFC plays a substantial role in short-term working memory (Figure 20–4). Research on monkeys in the 1930s by Jacobsen and Fulton showed that lesions to the PFC impaired spatial short-term working memory performance. In the 1970s, Fuster and colleagues recorded activation of PFC neurons in monkeys while they performed short-term working memory tasks. Functional imaging studies over the past 2 decades confirmed activation of the PFC in humans during short-term working memory tasks. Functional magnetic resonance imaging and transcranial magnetic lesion studies have revealed potential additional roles for the parietal lobe, MTL, and basal ganglia in short-term working memory. This is not entirely surprising given the roles of the parietal lobe in attention and computation of sensory information and the roles of the MTL and basal ganglia in encoding long-term memories.

Research by Goldman-Rakic and others in the 1990s demonstrated that networks of PFC excitatory pyramidal neurons continue to fire during, and are required for, short-term working memory. These and other studies led to a model stipulating that the encoding of short-term working memory involves increased and persistent firing of PFC neurons in recurrent networks. In the models, persistent activity is induced by sensory input (and presumably via input from long-term memory), which persists even after the input disappears, and different PFC subregions contribute to different components of the working memory system. More recently, researchers have hypothesized that working memory mechanisms involve activity-dependent (Hebbian) short-term synaptic plasticity (defined in Chapter 10, “Synaptic Plasticity”). Both short-term potentiation and short-term depression have been implicated in short-term working memory. In fact, potentiation has been proposed as a candidate mechanism that underlies the enhanced persistent firing model. Short-term potentiation involves enhanced neurotransmitter release by increasing presynaptic calcium ion (Ca²⁺) levels or duration and/or enhanced release mechanisms. Short-term depression is thought to involve depletion of neurotransmitter stores. These synaptic changes could function as short-term working memory traces. The memory trace would decay over time as the release activities and neurotransmitter levels are restored to their prestimulus levels. Other non-Hebbian synaptic and neuronal plasticity mechanisms may act in parallel with synaptic plasticity in short-term working memory mechanisms.

LONG-TERM MEMORY

Long-term memory is the second stage of the dual memory store model where large amounts of information can be stored for prolonged periods of time, from years to decades. In the model, long-term memory depends on short-term working memory for its input and is accessed by the process of recall, also known as retrieval, in which information is located and then brought back to short-term working memory (Figure 20–5). Long-term memory encompasses the concept of multiple memory systems, which are defined based on the types of information being learned and stored. The 2 main types of memory are declarative, also called explicit memory, and nondeclarative, also called implicit memory. Declarative memory encodes and stores semantic, episodic, and spatial information; is consciously learned and recalled; and depends specifically on the hippocampus for encoding. Nondeclarative memory includes procedural and motor memory and nonassociative and classical associative learning; is unconsciously learned; and can be unconsciously or consciously recalled. Each of the nondeclarative subsystems requires specific brain regions, such as the basal ganglia, cerebellum, or amygdala, for encoding of information in long-term memory.

It is well accepted that the formation of long-term memories progresses as a time-dependent process and that, as time advances, memories become stronger and less susceptible to interference. Before the memory is stabilized, long-term memory is subject to fading through the normal processes of forgetting. There are numerous hypotheses about forgetting, including mechanisms of passive and active forgetting. The 3 main passive forgetting models are (1) the loss of context cues that hinder retrieval, (2) interference from other similar memories that hinder retrieval, and (3) the natural decay of memory traces, called trace decay theory. Four forms of active forgetting have also been postulated. In interference-based forgetting, other competing information or activities accelerate the decay of memory traces. In motivated forgetting, cognitive mechanisms are voluntarily engaged to weaken a memory trace, often for unpleasant memories. In retrieval-induced forgetting, some aspects of a memory are recalled that suppress the recall of other aspects related to the recalled memory. In intrinsic forgetting, activity of forgetting cells and intrinsic biochemical and molecular pathways degrade the previously encoded memory traces. It seems likely that, depending on the region or type of information affected, different mechanisms may be employed for forgetting.

Repetition or rehearsal is often required to encode and preserve a long-term memory. For most declarative and procedural memory, items stored in short-term working memory move to long-term memory through repeated practice and use. Rehearsal can occur through behavior, reflection, or deliberate recall, called recapitulation, which are often dependent on the perceived importance, called valence, of the information. For some types of memory, information with extremely high valence, for example, fear, may transition rapidly into long-term memory with just 1 learning experience. Memory research has shown that learning is subject to interference. In contrast, once a memory has been stored in a long-term memory store, it no longer requires rehearsal and becomes immune to interference. However, after long-term memories are recalled and undergo reconsolidation, memories can become labile and malleable.

It is well established that human long-term memory can store a massive number of items, which has been well studied in visual memory, where subjects can recall details about thousands of images after viewing. Because it is thought that the engram involves biochemical and physical changes in synapses and neurons, it is expected there must be a physical limit to how many memories we can store. Although the size of the memory store has not been measured, it has been estimated. If each neuron participated in the storage of a single memory, the limit of the store would be 86 billion (nearly 10¹¹) pieces of information. Because 1 neuron can receive thousands of synaptic inputs and each neuron probably participates in multiple memories, this exponentially increases the brain’s memory storage capacity, with current estimates of the store at around a petabyte (a million gigabytes or 10¹⁵ units of information).

Learning involves acquisition and encoding of information, leading to the formation and stabilization of the memory trace, which for long-term memory involves processes called memory consolidation, of which there are 2 types, termed synaptic consolidation and systems consolidation. These types of memory consolidation occur with different but overlapping time courses, mechanisms, and brain areas. Synaptic consolidation represents the first phase, beginning within the first minutes after acquisition and lasting for several hours. During synaptic consolidation, the memory trace is encoded as changes in synapses that involve synaptic plasticity mechanisms, molecular changes that alter synaptic transmission, and morphology. Synaptic consolidation is thought to occur through modification of synapses in small groups of neurons in a specific brain region required for learning, for example the hippocampus for semantic information. Candidate mechanisms underlying synaptic consolidation are detailed in Chapter 10, “Synaptic Plasticity.” Some investigators now refer to this phase as synaptic and cellular consolidation because additional neuronal activities such as gene expression and global modulation of excitability may be involved.

During systems consolidation, which begins within hours but takes weeks to months to years to be accomplished, memory traces that were first encoded by synaptic consolidation are then distributed to other areas of the brain and are stored as stable alterations of neuronal circuits. One hypothesis is that the memory trace is gradually consolidated within a population of engram neurons in the region where the memory is stored. Recurrent activation, or “training,” of the engram neurons in the storage circuit by neurons in the learning circuit may be involved. In addition, changes in gene expression in the neurons of the consolidated memory circuit have also been demonstrated.

There is consensus that long-term memories are stored in widely distributed neuronal circuits throughout the cerebral cortex. Numerous intriguing ideas have been proposed about memory storage. Specific aspects of memories may be encoded within individual neural networks by specific patterns of synaptic connections. The memory of all the aspects of an object, fact, event, location, skill, or response likely involves several different groups of neurons in different parts of the brain. Memories may be stored as connections between groups of neurons that are primed to fire together in the same pattern that created the original experience. Each component of a memory may be stored in the brain area that initiated it. Long-term memories may be accessed independently, by visual, verbal, or other sensory clues. Memories may be encoded redundantly in various regions of the cortex, such that if 1 memory trace is lost, the duplicated engram or alternative route allows the memory to still be recalled. These compelling ideas await support by experimental data. Studies have demonstrated that sleep is an important factor in establishing long-term memories, because processes that occur during sleep play key roles in system consolidation.

The recall of memory refers to the process of retrieval of information from long-term memory stores. The 2-stage theory proposes that recall begins with a search and retrieval process and then progresses to a decision or recognition process where the correct information is chosen from what has been retrieved. Once a previously consolidated memory has been recalled, it can be actively reconsolidated. It has been proposed that reconsolidation serves to maintain, strengthen, and modify memories that had already been stored in long-term memory. During reconsolidation, memories become labile and malleable. Importantly, reconsolidation may provide the opportunity for significant memories to become updated, incorporating new information into the long-term memory store. Many important questions about long-term memory are currently under investigation in basic neuroscience research. For example, what are the mechanisms whereby previously stored memories remain stable but accessible and malleable. Another key question is: How does the brain maintain previously learned memories while it continues to learn new information that becomes incorporated into the networks of preexisting long-term memories?

DECLARATIVE/EXPLICIT MEMORY

One of the 2 main types of human long-term memory, declarative/explicit memory, is a memory system for factual knowledge, concepts, experiences, and spatial information. Four common features define declarative/explicit memory. It is consciously learned and consciously recalled (hence, the term explicit); after recall, it can be described or articulated (hence, the term declarative). Declarative/explicit memory is extremely flexible; numerous pieces of information can be associated under different situations. Anatomically, declarative/explicit memory requires the MTL, including the hippocampus and surrounding entorhinal, perirhinal, and parahippocampal cortices, for learning (Figure 20–6).

The 2 main subcategories of declarative memory are semantic memory and episodic memory. Semantic memory reflects factual information, meaning ideas and concepts; it includes vocabulary, verbal symbols, object knowledge and function, social customs, geography, understanding of math, and all the factual knowledge acquired over a lifetime. Episodic memory involves observational information attached to specific life experiences and events. These can be memories about what happens to a person directly or events that occur in situations around a person. Episodic memory involves the encoding and recall of events and experiences that include a location context and time component.

Two additional subcategories of declarative/explicit memory, termed autobiographical and spatial memory, have been defined. Autobiographical memory consists of episodic memory about personal experiences combined with semantic knowledge associated with those experiences and information about oneself. These include our personal experiences, such as likes, dislikes, and opinions that contribute to our sense of identity and form part of the “self-memory system” proposed by Conway and Pleydell-Pearce in 2000. Spatial memory is involved in recording and recalling information about one’s environment and spatial orientation. Spatial memory provides essential components for episodic memory but is also key for representation of location, spatial orientation, and goal-directed spatial navigation and thus has earned its own subcategory in declarative/explicit memory.

Evidence for the role of the MTL in long-term memory was first discovered in the 1950s from the study of patients who had undergone surgery for epilepsy or suffered damage to the MTL. The first, most famous, and most thoroughly studied patient was H.M. By age 27, H.M. had suffered for over 10 years from intractable temporal lobe epilepsy, caused by brain damage following a bicycle accident. Recurrent uncontrolled seizures were debilitating and rendered him unable to live a normal life. Experimental surgery was performed to reduce the seizures that involved a bilateral MTL resection, including removal of the hippocampal formation, amygdala, and parts of the multimodal association area of the temporal cortex. Although the surgery resulted in better control of his seizures, H.M. was left with a devastating memory deficit. He suffered from anterograde amnesia and was unable to form new semantic or episodic long-term memories. Consequently, H.M. lived minute to minute for the next 50 years of his life.

H.M.’s memory deficit was remarkably specific. His short-term working memory was largely intact. His prior memories were mostly intact for his lifetime up to several years before the surgery; he experienced some retrograde amnesia for information acquired in the several years before his operation. H.M. retained the command of language, including his vocabulary, indicating that semantic memory was preserved; his IQ remained the same. His ability to learn new procedural and motor skills and classical and operant conditioning appeared normal. What H.M. lacked after the surgery was the ability to transfer new information about people, places, and objects he had encountered or other daily experiences from short-term working memory into long-term memory. Other patients who have experienced bilateral lesions of the hippocampus or MTL, from surgery or as a result of disease, have shown comparable long-term memory deficits.

These clinical studies, together with a multitude of studies using animal models and human functional brain imaging, support the conclusion that the hippocampus and adjacent entorhinal cortex, the main input area from the cerebral cortex to the hippocampus, play essential roles in forming new declarative/explicit memories, and other structures in the MTL contribute important functions as well. In addition, these studies demonstrate that the MTL is not essential for short-term working memory, nondeclarative/implicit memory, or storage and retrieval of the majority of declarative/explicit memory (although a role for the MTL in recall of some autobiographical memory has been shown). After his death, H.M.’s name was revealed; the field of neuroscience owes a large debt of gratitude to Henry Molaison and other patients, whose participation in experimental studies catapulted the understanding of human long-term memory.

Regions in the MTL are also involved in perception and spatial representation, which in addition to providing input for memory, may perform additional cognitive functions. For example, the perirhinal cortex has been implicated in object recognition. Neurons have been identified in the rodent hippocampus, called “place cells,” that fire bursts of action potentials when the animal passes through a particular location in its environment. Other hippocampal neurons called “head direction cells” respond when the animal’s head points in a specific direction; “boundary cells” respond to the presence of an environmental boundary. Specific neurons in the entorhinal cortex act as “grid cells,” which respond when the animal traverses a set of small regions, and “speed cells,” which respond to the speed of the animal’s running. Together these MTL cells are thought to form a network that provides a neural representation of the animal’s specific location and movement in space, acting as a cognitive map and contributing to spatial memory as well.

What are the mechanisms involved in the formation of long-term declarative/explicit memory? As described in Section 4, it has been proposed that long-term memory involves consolidation. In a popular model, changes in synapses and neuronal gene expression within hippocampal neurons in the first few hours of learning underlie synaptic and cellular consolidation for declarative/explicit memory. These involve synaptic plasticity mechanisms, in particular long-term potentiation (LTP) and long-term depression (LTD). Described in Chapter 10, “Synaptic Plasticity,” LTP and LTD are activity-dependent persistent changes in synaptic transmission, which encompass signaling cascades that produce biochemical and morphologic changes in synapses, as well as signaling to the nucleus that leads to changes in neuronal gene expression. In systems consolidation, information is actively transferred as the hippocampus teaches the cortex more and more about the information it has acquired. The standard model of systems consolidation proposes that soon after learning, the same hippocampal and cortical neurons that encode memory are reactivated during retrieval; in the early stages, the hippocampus guides the reactivation of neocortical circuits during retrieval of memory. However, as time progresses, neocortical circuits are reactivated during memory retrieval and no longer require input from the hippocampus. Studies demonstrate that neocortical plasticity is required for memory consolidation, with many of the signaling cascades and genes identified in hippocampal LTP and LTD implicated in systems consolidation.

It is well documented that sleep plays an important role in formation of declarative/explicit memory. One model contends that, during slow wave sleep (SWS), reactivation of newly learned hippocampal memory representations transfers information to neocortical networks, where it is integrated into long-term representations through systems consolidation. Research suggests the sleep spindles, a sleep wave generated in the thalamus, play an additional role in boosting consolidation of declarative memories and/or incorporating new information into existing knowledge. Rapid eye movement (REM) sleep has also been implicated in declarative memory. Sleep may play an active role in producing specific brain waves or other processes that facilitate memory consolidation processes, or a passive role to protect from interference or provide ideal conditions for memory consolidation. In an active role model, specific brain waves or other processes that occur during SWS may contribute to the regulation of gene expression that controls proteins required for synaptic plasticity. It has been suggested that systems consolidation occurs during sleep, because this is a period when the necessary retrieval of memory required for consolidation occurs and would not interfere with daily cognitive functions.

NONDECLARATIVE/IMPLICIT MEMORY

Nondeclarative/implicit memory is the other major category of long-term human memory. Nondeclarative memory is reflected in automatic changes in behavior, tasks, responses, or performance that do not require conscious thought or conceptual recollection. It allows people to do things by rote without thinking about them. It allows us to perform many tasks effortlessly, but it is often difficult to explain exactly how we do them, hence the term nondeclarative. The 4 main subcategories of nondeclarative/implicit memory are (1) procedural, which includes motor and perceptual skills, tasks, and habit formation; (2) priming; (3) simple, classical and associative conditioning; and (4) nonassociative learning.

These subtypes of nondeclarative/implicit memory share several features in common. They are all types of unconscious memory that is evident in the performance of a task, behavior, or response. They are inflexible (ie, insensitive to surface perceptual changes). They are tightly connected to the original context under which the learning occurred. However, although similar in their characteristics, each of these subcategories involves different, but sometimes overlapping, brain regions for their functions. Similar to declarative/explicit memory, nondeclarative/implicit learning involves changes in the effectiveness of the synaptic pathways that underlie those memory systems, which is further described in Chapter 10, “Synaptic Plasticity.” Indeed, many of the signaling pathways and mechanisms involved in synaptic plasticity underlying memory were first discovered in nondeclarative/implicit systems.

One of the most well-recognized forms of nondeclarative/implicit memory is procedural memory, which enables people to perform certain tasks, skills, and habits automatically, without being consciously aware or remembering how to perform the actions (Figure 20–7). It is the “how to” knowledge and constitutes a large majority of nondeclarative/implicit memory. In everyday life, people rely on procedural memory for physical actions such as tying a shoe, driving a car, washing dishes, and turning on a smart phone, without conscious awareness or thinking. Even behaviors that we consider innate, such as walking, require components of procedural memory. Procedural memory also involves learning motor sequences that are inherent in more complex skill learning, such as playing a musical instrument or hitting a tennis ball. Many procedural memory tasks are unconsciously learned, but components can also involve conscious learning as well. Improving procedural skills involves repetition and practice. Studies of human clinical disorders provided the first clues as to the brain regions involved in procedural memory, which is now known to involve the basal ganglia and cerebellum. Parkinson disease pathology is largely limited to the basal ganglia, and patients with this disorder have deficits in the acquisition of new motor skills. Within the basal ganglia, the striatum contains the main neuronal cell nuclei linked to procedural memory. Damage to the cerebellum can affect learning of motor and procedural skills. The cerebellum functions in correcting movement and fine tuning the motor agility found in complex procedural skills such as painting, instrument playing, and accuracy sports.

Priming is a subcategory of nondeclarative/implicit memory concerned with perceptual identification of words, objects, and other stimuli. Priming produces an improved facility for detecting or processing a perceptual object based on recent experience; exposure to 1 stimulus using pictures, words, or music influences the response to a subsequent stimulus. A typical example of priming is when a person reads or hears a word, and for some period of time afterward, the person is more likely to use that word in conversation or in a word completion experiment. Priming not only improves the ability to identify stimuli but also alters judgments and preferences that involve the same stimuli. For example, priming leads to the illusion-of-truth effect, which suggests that subjects are more likely to believe or rate as true those statements that they have already heard, regardless of the truth of the statement. Studies on stroke patients indicate that priming involves the occipital and prefrontal cortices.

Associative memory is a subcategory of nondeclarative/implicit memory where an animal learns the predictive value of one stimulus for another. Essential for survival and generally reliable, associative memory is the neural correlate of cause and effect. Two types of associative learning, called classical conditioning and operant conditioning, were defined through extensive experimental studies used throughout the 20th century, which still form the core of associative memory research today. Classical conditioning involves associating 2 environmental stimuli, such that a response is conditioned to an unrelated stimulus, and involves unconscious reflexive responses and skeletal musculature. Operant conditioning involves associating a specific behavior with a reinforcing environmental event and involves the execution of a voluntary motor response.

Classical conditioning was first studied in detail by Ivan Pavlov, through his seminal experiments with dog behavior. Subsequently, other experimental paradigms were developed, including conditioned fear response, classical eye blink response, and odor and taste aversion. Pavlovian classical conditioning involves a specific paradigm. Before conditioning occurs, a stimulus, called the unconditioned stimulus (US), produces a natural unlearned response, called the unconditioned response (UR). The US needs to be biologically potent and can be pleasant and even rewarding or unpleasant and even painful. The UR is an involuntary, automatic, or reflexive response. In Pavlov’s experiments, prior to conditioning, presentation of food was the US, and it produced salivation, the UR. The experiment involves another stimulus, called the neutral stimulus (NS), which has no effect on an animal before learning. The NS could be a person, object, place, noise, smell, or some other stimulus. For Pavlov, the NS was the sound of a metronome.

Conditioning begins when the NS is paired with the US. As soon as the pairing begins, the NS becomes what is called the conditioned stimulus (CS), and it becomes associated with the pleasant/unpleasant and rewarding/painful qualities of the US. After pairing is repeated, the animal exhibits a conditioned response (CR) to the CS when the CS is presented alone. The CR is usually similar to the UR, although there may be some differences. Consequently, with successful pairing, the CS is able to produce a CR that is reminiscent of the UR to the US. For Pavlov, the sound of the metronome elicited a salivation response in the dogs, before and even without the presentation of food.

In the conditioned fear response, an aversive stimulus, such as a foot shock, becomes associated with a specific environmental context or stimulus sound, resulting in the expression of a fear response, such as freezing. Fear conditioning involves the amygdala and, depending on the type of US, can also involve the hippocampus (for contextual fear conditioning) (Figure 20–8). In eye blink conditioning, before learning, a mild puff of air or shock to the cornea, the US, produces a reflexive eye blink response, the UR. During conditioning, the pairing of an auditory or visual stimulus (initially a NS, which becomes the CS) together with the US (air puff) leads to the CR, the eye blink. Studies have shown that eye blink conditioning involves the cerebellum. In olfactory conditioning, pairing of an odor with an aversive stimulus results in the animal’s aversion to that particular odor and involves the olfactory bulb. Conditioned taste aversion occurs when an animal associates the taste of a certain food with symptoms caused by a toxic, spoiled, or poisonous substance that produce nausea, sickness, or vomiting. The nucleus tractus solitarius in the brainstem has been implicated in conditioned taste aversion.

Many human behaviors can be modified through positive and negative reinforcement, so behavior can be controlled by predicted consequences. Operant conditioning is a type of learning in which the strength of a behavior is modified by the behavior’s consequences, such as reward or punishment. In operant conditioning, an individual makes an association between a particular behavior and a consequence, and the stimuli present when a behavior is rewarded or punished can also become associated. Throughout humanity, behaviors that lead to a successful outcome are often retained, whereas those that are unsuccessful or produce adverse effects are usually discarded. Moreover, operant conditioning has likely been used by parents in teaching children for tens of thousands of years.

Famously studied first by Edward Thorndike in the late 19th century and then B.F. Skinner in the mid-20th century, operant conditioning experiments can involve positive or negative reinforcement or punishment. Positive reinforcement occurs when a behavior or response is itself rewarding or the behavior is followed by another stimulus that is rewarding. Negative reinforcement occurs when a response or behavior is followed by the removal of an adverse stimulus. Positive punishment occurs when a behavior or response is followed by an aversive stimulus. Negative punishment occurs when a behavior or response is followed by the removal of a positive stimulus. The goal in both types of reinforcement is for the behavior to increase, whereas the goal in both types of punishment is for a behavior to decrease. Evidence suggests that both acetylcholine (produced by neurons in the basal nuclei) and dopamine (produced by neurons in the midbrain) are involved in operant conditioning.

Regarded as one of the simplest forms of implicit learning in animals, nonassociative learning is considered a fundamental form of learning that can be observed across all animal phyla and most sensory modalities. Nonassociative memory involves a change in the strength of a response to a single stimulus, resulting from repeated exposure to that stimulus, but does not require the formation of an association between one environmental stimulus and another. It automatically classifies the valence of a sensory input based on its temporal pattern without the need for concomitant or feedback signals. Nonassociative learning can be divided into habituation and sensitization. Habituation occurs when an animal learns, after repeated exposure, to ignore a stimulus that is neither beneficial nor harmful in its environment. In habituation, the strength or probability of a response, which is usually a reflex response, decreases when the stimulus and its response are repeated. Sensitization is the progressive increase of a response following repeated administrations of a stimulus. Well studied in invertebrates, nonassociative memory in humans is best characterized in adaptation in reflex responses involving sensory stimuli.

SUMMARY

• Memory involves information encoding, storage, and retrieval.

• Memory is organized according to temporal components called sensory memory, short-term/working memory, and long-term memory.

• Sensory memory involves sensing and initial processing of sensory stimuli.

• Short-term memory is defined as a short-term store of goal-directed information; it depends on attention and involves reverberating circuits in the prefrontal cortex.

• Working memory encompasses an active short-term memory store that can be accessed to process and manipulate information; in addition to the prefrontal cortex, it involves other areas of the frontal lobe and parietal cortex.

• Long-term memory is the permanent store where information is retained for extended periods, from years to decades; the long-term memory engram or trace involves biochemical and physical changes in the brain.

• Memory is also organized along informational components, called declarative/explicit and nondeclarative/implicit memory.

• Declarative/explicit memory includes semantic memory, which is memory of facts and information encoded with specific meaning, episodic memory of experiences and events, and spatial memory about one’s environment and spatial orientation.

• Nondeclarative/implicit memory includes memory for procedures, tasks, and skills, as well as emotional and conditioned responses.

• Encoding of long-term memory involves different brain regions: Declarative/explicit memories require the hippocampus and other temporal lobe structures, whereas nondeclarative/implicit memory requires the basal ganglia, cerebellum, amygdala, and/or reflex circuits.

• Encoding of long-term memory involves synaptic consolidation, which involves synaptic plasticity mechanisms and systems consolidation, where memory traces are distributed to other areas of the brain and stored as stable alterations of neuronal circuits.

• Sleep plays an important role in memory consolidation.

• Different memory systems operate independently and in parallel, which allows simultaneous and maximal memory throughput of the brain.

OBJECTIVES

CLASSIC MODEL

The classic model of the neural basis of language emerged from the work of 19th-century cognitive neurologists Paul Broca, Carl Wernicke, and Ludwig Lichtheim. Broca described patients with relatively preserved ability to understand speech but profound deficits in producing speech: slow, labored, telegraphic speech, often with speech errors or distortions of speech sounds and omission of function words (eg, articles and prepositions) and inflections (eg, –s for plurals, –ed for past tense). This disorder came to be known as Broca aphasia or expressive aphasia and to be associated with damage to the left inferior frontal gyrus (Broca area) and the underlying white matter. In contrast, Wernicke described patients with profound comprehension deficits but relatively fluent speech production. Although relatively rapid and well-articulated, the speech production in Wernicke aphasia (also called receptive aphasia) tends to have limited meaningful content due to use of a small set of high-frequency words and made-up words (also called “nonwords” or “abstruse neologisms”). Wernicke aphasia was thought to be associated with damage to Wernicke area, the posterior portion of the left superior temporal gyrus.

Lichtheim developed an integrated description of the language system and its instantiation in the cortical regions around the left hemisphere’s Sylvian fissure. This distinction between nonfluent expressive aphasia and fluent receptive aphasia and the role of peri-Sylvian regions in language remained largely unchanged into the late 20th century and remains deeply influential despite being challenged by 3 kinds of observations. First, many individuals with aphasia do not fit the classic aphasia subtypes; rather, they exhibit “mixed” or “unclassifiable” clusters of language deficits. Second, the symptoms are functional/behavioral and may be caused by a variety of very different underlying cognitive and neurologic deficits. For example, the telegraphic speech typical of Broca aphasia can be caused by a deficit in syntax (grammar) and sentence planning, a deficit in word retrieval and selection, or a deficit in articulatory motor planning and execution. Third, focal damage to the Broca area does not tend to produce Broca aphasia, and focal damage to the Wernicke area does not tend to produce Wernicke aphasia. Rather, those syndromes are caused by broader damage that includes the traditionally associated regions as well as other regions.

Partly motivated by these challenges, an alternative has emerged that focuses on the “primary systems”—the cognitive and neural systems that support language processing. These systems are as follows: (1) a distributed semantic system for representation of conceptual knowledge, with particularly important integrative hubs in the anterior temporal lobe and temporoparietal cortex (angular gyrus and posterior superior temporal gyrus) and related white matter tracts (particularly the inferior frontal-occipital fasciculus and the uncinate fasciculus); (2) a speech recognition system for recognizing and distinguishing speech sounds and spoken words, which builds on higher level auditory processing and primarily involves the superior temporal lobe; (3) a speech production system in which (i) segment-level or syllable-level articulatory planning and motor control rely on inferior parietal regions including somatosensory representations of the articulators in the inferior postcentral gyrus and extend into inferior precentral gyrus motor representations of the articulators; and (ii) sentence-level or utterance-level planning and execution (ie, grammatically structuring multiword utterances) rely on inferior and middle frontal gyri and underlying white matter (aslant tract); and (4) a ventral visual object recognition system for reading, particularly a region within the left fusiform gyrus known as the visual word form area.

MEANING

Semantic Representation

Language is ultimately about the communication of meaning, so semantic or conceptual representations are the keystone of the language system. When we encounter and interact with objects in the world, the primary perceptual and motor representations converge to more abstract conceptual representations of those objects. When we recall and think about object and action concepts, such as “hammer” and “kick,” we reactivate those perceptual, motor, and emotional experiences of the objects or events. For example, the representation of the concept “hammer” is based on the visual experiences of seeing hammers, the motor-action experiences of pounding nails, the auditory experiences of hearing hammering, and so on. This is called grounded or embodied semantics or perceptual symbol systems; that is, the conceptual representations are grounded in our bodily experiences. The neural consequence is that semantic knowledge draws on a distributed system that includes lateral occipital cortex representations of object shape, middle temporal/medial superior temporal representations of object motion, secondary auditory cortex representations of meaningful sounds, posterior temporal and inferior parietal representations of object-related action, and other perceptual, motor, and limbic regions.

Damage to individual components of this distributed system can cause specific semantic deficits associated with individual components. For example, atrophy of the auditory association cortex produces a specific deficit for processing words that refer to concepts with strong auditory elements (eg, thunder) compared to concepts with strong visual elements (eg, pyramid) or manipulation elements (eg, scissors). Damage to inferior temporal regions in the later stages of the ventral visual stream tends to produce deficits for concepts that are primarily distinguished by visual features, such as animals. In contrast, damage to inferior parietal cortex regions in the later stages of the dorsal visual stream tends to produce deficits for concepts that are primarily distinguished by action or manipulation features, such as tools.

This distributed system of modality-specific representations is integrated in 2 cross-modal “hubs”: 1 in the anterior temporal lobes (ATLs) and 1 in temporoparietal cortex (TPC). Figure 21–1 shows a schematic depiction of the distributed semantic system with hubs in the ATL and TPC. The ATL hub is particularly important for object identification and categorization. Frontotemporal lobar degeneration that particularly affects the ATL produces a deficit of semantic knowledge: the semantic variant of primary progressive aphasia (sPPA, also called semantic dementia). Individuals with sPPA have profound semantic deficits, while other aspects of perception and cognition are largely intact. In sPPA, errors in picture naming are almost exclusively more common or more typical category members (eagle → duck) or superordinate category labels (eagle → bird). Similar errors are made by patients with ATL damage due to focal epileptogenic lesions or stroke lesions (which typically extend beyond ATL). The semantic deficit in sPPA extends beyond just verbal tasks: In nonverbal tasks such as delayed copying and object decision, individuals with sPPA also make category typicality errors. For example, they may choose the elephant with small ears instead of large ears or draw a camel with no hump or a duck with 4 legs—all cases where the patient produced a response that is typical for the category (animals) but fails to reflect object-specific properties. That is, degeneration of the ATLs produces gradual blurring of object-specific within-category distinctions, while broader between-category distinctions are comparatively spared.

The TPC hub is particularly important for event or scenario representations, such as relationships between objects that are not members of the same category and do not share features but frequently co-occur together (also called thematic relations), for example, object–instrument relations (eg, dog – leash) or object–location relations (eg, zebra – savannah). TPC stroke damage tends to increase the production of thematically related semantic errors in picture naming (eg, cow → milk), and these relations are spared relative to perceptual features in patients with sPPA, which tends to affect the ATL hub more than the TPC hub.

Both cerebral hemispheres are involved in semantic processing, with a substantial degree of redundancy between the hemispheres, but also some hemispheric specialization. The profound semantic deficit in sPPA tends to arise from bilateral deterioration of the ATL. Unilateral damage to the semantic system tends to produce more limited semantic deficits, indicating that the spared contralesional hemisphere is able to largely compensate. Unilateral left ATL damage tends to affect verbal semantic tasks more strongly (eg, picture naming and word matching), whereas unilateral right ATL damage tends to affect visual semantic tasks more strongly (eg, picture matching). Presumably, this is due to the left ATL being more strongly connected to the (typically) left-lateralized language system.

White Matter

There is one notable exception to the pattern that unilateral damage produces modality-specific semantic deficits: Broad semantic deficits can arise after unilateral stroke damage to a “white matter bottleneck” medial to the insula and lateral to the basal ganglia. The inferior fronto-occipital fasciculus (IFOF), uncinate fasciculus, and anterior thalamic radiations converge in this region, so the diverse semantic deficits arise because a small amount of bottleneck damage can cause widespread dysfunction throughout the semantic system (key white matter tracts involved in language processing are shown in Figure 21–2). The IFOF is particularly important for linking visual processing in the occipital and ventral temporal cortex with the semantic hub in the ATL. The role of the uncinate fasciculus is somewhat more controversial, with alternative accounts focusing on its role in semantics of valence and social semantics (due to connecting the ATL with the orbitofrontal cortex) or in semantic control (due to connecting the ATL with the inferior frontal regions). There may also be a posterior white matter bottleneck in the vicinity of the TPC, but its anatomy and function are less well documented.

Semantic Access

Deficits such as sPPA and category-specific impairments (eg, impaired semantics of animals but not tools) suggest damage to the neural system that stores or represents semantic knowledge, such that part of that knowledge is lost. In these deficits, impaired performance is fairly consistent across repeated testing, different tasks, and different testing conditions. There is also a very different kind of semantic deficit in which semantic knowledge appears to be intact but access to that knowledge is ineffective, inefficient, or inconsistent. Patients with semantic access deficits (1) are sensitive to cueing (picture naming is facilitated by hearing the start of the target word), (2) perform better when test trials are separated by a long delay than by a moderate delay, (3) perform inconsistently on multiple tests of the same stimulus (ie, sometimes can name a picture and sometimes cannot name that same picture, usually with declining performance over repeated presentations), (4) perform worse in the presence of semantically related distractors, and (5) tend to have a correlation between their semantic deficit and executive function deficits. Each of these phenomena stands in stark contrast to the semantic deficit in sPPA, which exhibits none of them. Observation of these deficits—and their striking difference from the semantic deficit in sPPA—indicates that semantic cognition requires key components or mechanisms: storage of semantic knowledge and access to or control over that knowledge. However, there is not yet agreement about the computational basis of the access/control component or its neural basis, although the frontal cortex and the underlying white matter likely play an important role.

RECOGNITION & PRODUCTION OF SPEECH SOUNDS

The sound structure of language is called phonology, and the speech sound units are called phonemes. Phonemes are typically defined by “minimal pairs”: The difference in the initial sound between “tape” and “cape” produces a different word; therefore, the initial sounds in those 2 words are different phonemes. In contrast, the initial sound in “tape” will be acoustically different between “Christmas tape” and “Spanish tape,” but it will be heard as the same word in both cases, so that acoustic difference is not phonemic.

Speech Recognition: Ventral Stream

Recognition of phonemes is the major higher level auditory function of the auditory association cortex in the posterior portion of the superior temporal gyrus. Early spectrotemporal analysis of speech input is bilateral, with left hemisphere auditory cortex specialized for temporal analysis of rapid acoustic changes and right hemisphere auditory cortex specialized for fine-grained spectral frequency analysis. This division of labor is driven by a fundamental signal-processing limitation: a trade-off between temporal resolution and spectral resolution known as the Gabor limit. Asymmetric temporal sampling allows the auditory system to overcome this limitation with higher temporal resolution and lower spectral resolution in the left hemisphere and higher spectral resolution and lower temporal resolution in the right hemisphere.

One consequence of asymmetric temporal sampling is the relative dominance of the right hemisphere for auditory functions that are more reliant on spectral resolution than on temporal resolution. Such functions include music processing, recognition of emotion in speech, and speaker identification and recognition. Auditory agnosias—deficits of auditory recognition that are not attributable to hearing or cognitive deficits—tend follow this lateralization pattern. Impaired recognition of environmental sounds and music (amusia) generally results from right hemisphere damage, whereas deficits of speech sound recognition (verbal auditory agnosia, or word deafness) generally result from left hemisphere damage.

After initial spectrotemporal processing, subsequent stages of speech recognition are generally left-lateralized and rely on progressively more anterior portions of the superior temporal lobe. Recognition of phonemes and syllables engages the middle portion of the superior temporal gyrus; recognition of words and phrases engages the anterior portion, making contact with the semantic hub in ATL. This posterior-to-anterior progression in the superior temporal lobe is known as the ventral speech stream (Figure 21–3) and is the speech analog of the ventral visual processing stream for object recognition.

Speech Production: Dorsal Stream

Speech production requires very precise motor control of the articulatory apparatus—the lips, jaw, tongue, vocal folds, and velum (which controls whether air is allowed through the nasal passage, distinguishing sounds such as /m/ and /n/ from /b/ and /d/). This relies on core motor control systems of the basal ganglia, cerebellum, and motor cortex and can be disrupted by stroke and other disorders of the motor system such as Parkinson disease. Such motor-based disorders of speech production are called dysarthria. Stuttering is also thought to be due to dysfunction in the basal ganglia–thalamocortical motor circuits that impairs timing cues for the initiation of speech.

In addition to low-level motor control of the articulators, speech production requires articulatory action planning. The articulatory action planning system relies on a frontoparietal circuit that includes the precentral gyrus, insula, postcentral gyrus, and inferior parietal lobule. This system is known as the dorsal speech stream (see Figure 21–3) and is the speech analog of the dorsal visual processing stream for visually guided action. Damage to the dorsal speech stream tends to impair production of speech sounds with relatively spared lower level motor control of the articulatory apparatus and higher level aspects of language processing.

One such disorder is apraxia of speech, in which speech sounds can be distorted with abnormal rhythm, stress, or prosody, and patients struggle to initiate words or “grope” for the right arrangement of articulators to produce the correct speech sound. Patients may also make frank speech sound substitution errors, producing well-articulated nonwords that are similar to the target word (eg, ghost → “goath”). Although subtly different and possibly having somewhat different underlying causes, speech sound distortions and substitution errors tend to be fairly highly correlated and to be caused by very similar patterns of neurologic damage to the precentral and postcentral gyri, insula, and inferior parietal lobule.

The analogy between the dorsal visually guided action system and the dorsal speech production system also helps to understand the computational architecture of speech production. Limb actions are guided by visual perception of object location and size and by somatosensory feedback of limb position. However, the neural transmission and processing of these sensory feedback signals is too slow for efficient correction; by the time the feedback signal was processed and a correction signal sent, the position of the limb would have changed substantially and the correction signal would be wrong. To overcome this problem, motor control systems use an internal forward model, which makes online predictions regarding the state of the effector and the sensory feedback. This internal forward model allows the system to predict errors and send correction signals before the errors occur (and before sensory feedback is sent), so the correction signal can arrive in time for efficient, online error correction. The internal forward model is tuned by actual sensory feedback, so its predictions can (gradually) adjust to changes in the environment or effector properties. The computational forward model principles apply to both limb and articulatory motor control.

The targets of speech production are both motor and acoustic. That is, the target is an articulatory action sequence that produces a particular acoustic signal. The auditory system provides feedback about the outcome of the articulatory actions (like the visual system does for limb actions). The somatosensory system provides online feedback about the configuration of the articulators (as it does for limb actions). Internal forward models make predictions about the somatosensory and acoustic/auditory consequences of the articulatory motor actions, which are compared against the targets to allow online correction. Immediate error corrections are a simple example of the prediction-based correction mechanism in action; for example, utterances such as “v-horizontal,” where the production of “vertical” is corrected to “horizontal” within 150 milliseconds. This is far less than the approximately 400-millisecond minimum that would be required to recognize that an incorrect speech sound was produced and execute a new articulatory command. More generally, a substantial proportion of error corrections happen too fast for an external feedback system, indicating the involvement of an internal prediction-based error detection and correction system.

Because the formats of motor and sensory representations are different, these correction signals require sensory-motor transformation in order to make the forward model prediction and to compute the correction signal. The precise neural implementation of these sensory-motor transformations is still being debated, although the cerebellum and the planum temporale (or the parietotemporal boundary at the posterior end of the Sylvian fissure) have been proposed as promising candidates.

SENTENCE PROCESSING

Sentences unfold gradually over time but have a hierarchical or propositional underlying structure. That is, even simple sentences such as, “John loves Mary” or “Bill gave the book to Susan,” are expressions of relationships between constituents that are distributed throughout the sentence. Sentence comprehension requires assembling those constituents into the correct relational structure, and sentence production requires unfolding that structure into an appropriate sequence of words. Thus, temporal sequencing, working memory, planning, and cognitive control are key elements of sentence processing.

From the earliest days of neurolinguistics, the neural basis of syntax has been a central question. The syntax or grammar of a language is the set of rules or patters that govern how sentence constituents can be assembled. That is how we know that “The boy chased the dog” means roughly the same thing as “The dog was chased by the boy,” but that “The dog chased the boy” describes a very different event. Much of this research has focused on so-called agrammatism, a disorder thought to be specific to syntactic or grammatical processing. Despite these efforts, it has proven virtually impossible to disentangle syntactic processing from closely related cognitive functions such as sequence processing, working memory, and cognitive control. Therefore, although a neural basis for syntax remains elusive and controversial, there is substantial agreement about the neural systems that support sentence processing.

Sentence Comprehension

Sentence comprehension can be measured by asking participants to find the picture that matches a sentence such as, “The boy chased the dog.” If the distractor picture is a dog chasing a boy, then it can be correctly rejected only after correct comprehension of the full sentence. Comprehension of such “reversible” sentences can be compared against a “lexical” case where just recognizing the individual constituents is sufficient; for example, a distractor picture of a boy chasing a girl. Impaired comprehension of reversible sentences with spared lexical sentence comprehension is associated with damage to the inferior parietal cortex (angular gyrus and supramarginal gyrus), extending into the posterior superior temporal gyrus. The role of this region in sentence comprehension may be related to the event semantic hub in TPC; sentences describe events, so sentence comprehension is a form of event comprehension. It may also be related to the role of the parietal cortex in working memory because sentence comprehension requires active memory for word sequences.

Multiword Utterances

Following the general pattern that language production is more difficult than language comprehension, deficits of sentence production are more readily apparent than deficits of sentence comprehension. Agrammatism is typically defined in terms of impaired sentence production, such as omission of grammatical morphemes (eg, function words such as articles and prepositions and suffixes such as –s for plural and –ed for past tense) and making word order errors. In both progressive aphasia (agrammatic variant of primary progressive aphasia) and nonprogressive aphasia (stroke), these deficits tend to co-occur with other aspects of nonfluent language production such as slow, effortful speech, short utterances, and production of speech sound errors (discussed earlier). They are, perhaps surprisingly, not strongly associated with impaired sentence comprehension; many so-called agrammatic patients perform relatively well on sentence comprehension tasks that require syntactic analysis, which is one reason to doubt the existence of a central syntactic processing mechanism.

Effective sentence production requires articulatory control and word retrieval. As discussed in previous sections, speech production depends on a dorsal system of inferior parietal regions and frontal motor control regions, and the left ATL plays a particularly important role in semantically driven production of individual words. Broader speech rate measures (eg, words per minute) tend to also reflect integrity of the dorsal speech production system. When syntactic errors are isolated from other factors, the critical regions are substantially more anterior: the inferior and middle frontal gyri, possibly extending into the frontal aslant tract, a white matter tract that connects inferior frontal regions with the anterior cingulate and presupplementary motor area.

The neural systems supporting speech and sentence production bear a striking resemblance to the neural systems supporting limb action production and planning. Disorders of object-related actions (limb apraxia) are primarily associated with inferior parietal damage, extending into frontal motor regions, much like disorders of speech production. Higher level action planning and sequencing involves the frontal lobe in a hierarchical posterior-to-anterior progression: Posterior frontal cortex (ie, motor and premotor areas) is responsible for simple motor acts such as picking up a milk carton or pouring milk into coffee, more anterior regions represent larger action “chunks” or subgoals such as adding milk to coffee (which requires picking up the milk, pouring it, and putting it down), and even more anterior regions represent even larger action plans such as making a cup of coffee. Speech and sentence production can similarly be understood in terms of simple actions (articulating an individual speech sound or syllable), small action chunks (articulating a word or common word sequence), larger action chunks (planning and articulating a complete sentence), and overall action goals (telling a story). As with limb actions, there is a nested, hierarchical structure of goals and subgoals that requires maintenance and updating, and it is supported by a frontal-parietal cortical neural system.

READING

Basics & Mechanics

Spoken language appears to have emerged spontaneously a fairly long time ago and is learned by (almost) all children without formal instruction. In contrast, written language is a relatively recent invention (at least in evolutionary terms)—approximately 6000 years ago—and learning to read and write requires formal instruction. Thus, the reading system is based on some of the same neural systems as spoken language, with additional reliance on particular kinds of visual processing.

Writing systems differ substantially across languages, but they have some consistent elements: Most characters are drawn with 1 to 4 strokes, and character shapes tend to be as distinct as possible. At the initial input stage, reading is constrained by the anatomy of the retina and oculomotor dynamics. Text needs to be fixated (foveated) in order to extract enough high spatial frequency detail for character recognition. Text that appears in the parafovea provides partial information and allows a parafoveal preview of the next word. Eye movements during reading are typically short saccades, covering about 7 letters or 2 degrees, and moving from word to word. About 30% of words are skipped; these are typically short, very common words (eg, “the”) that can be processed without a dedicated fixation (ie, based on parafoveal preview and context). About 10% to 15% of saccades are regressions—refixations of an earlier word. Regressions are important for reading comprehension, especially for correcting initial comprehension failures: Comprehension accuracy decreases if regressions are prevented, especially for “garden path” sentences that have misleading structures.

Average fixation time during reading is approximately 250 milliseconds, but this varies widely. Fixation duration is primarily determined by linguistic and cognitive factors such as word familiarity and predictability, rather than visual factors such as word length. Even if a word disappears after 60 milliseconds, the fixation duration is largely unaffected: Visual perception of the word happens very quickly; the fixation time is the time required to comprehend the word, integrate it into the sentence representation, and plan the next saccade. Fixation times are widely used to study reading processes because they provide such a precise, natural, and minimally invasive measure of the cognitive demands of reading.

Reading a single word requires connecting 3 components of the language system: visual recognition of the characters (orthography), production (and recognition) of speech sounds (phonology), and word meaning (semantics). These 3 components are bidirectionally interconnected for the purpose of typical language tasks (eg, word recognition, word production, word reading) and are commonly called the “triangle model.” In alphabetic orthographies such as English, Spanish, and Hebrew, there is a systematic relationship between the visual characters and the sound of the word. In these cases, the orthography-phonology part of the triangle can directly decode (ie, read aloud) a word based on those grapheme-phoneme correspondences. This works well for words such as mint, lint, and hint, and for pseudo-words such as deest and beest. However, this system will have more difficulty with words that deviate from the pattern, such as pint. Decoding such “irregular” or “inconsistent” words is additionally supported by the semantic route of the triangle, and patients with semantic deficits (eg, sPPA) often have difficulty reading these kinds of words (discussed more later in the section on dyslexias). The triangle model also identifies the 3 main neural components of the reading system: inferior/basal temporal-occipital cortex for visual orthographic processing, peri-Sylvian regions for phonological processing, and anterior temporal cortex for semantic processing.

Ventral Stream & Visual Word Form Area

Like other kinds of object recognition, letter recognition is based on the ventral visual stream, including V1 and V2 edge and contour detectors (letter features), and ventral occipitotemporal representations of letter-like shapes and 2-letter combinations (bigrams). This pathway culminates with the visual word form area (VWFA), a region of the left fusiform gyrus near the lateral occipitotemporal sulcus, which is reliably activated more strongly when processing real words than nonword letter strings or mirror-reversed words. The VWFA response to real words increases as people learn to read in both their first language and in a second language. This is an experience-dependent process and not just developmental maturation. Adult illiterates who learn to read show increased VWFA activation specifically for real words. Among adult native English speakers, those who can also read Hebrew show greater VWFA response to Hebrew words than those who cannot read Hebrew (both groups show equal VWFA response to English words). VWFA lesions produce pure alexia, which is characterized by letter-by-letter reading, indicative of intact low-level visual processing (letter recognition) but impaired holistic word recognition.

Recall that reading is a recent invention (and widespread literacy is even more recent)—too recent for human genome evolution to have produced an innate neural word recognition module. VWFA is properly understood as an emergent property of the visual object recognition system, a system that is co-opted (or “recycled”) for reading and that itself influenced the form of writing systems. As a visual object recognition task, word recognition is paralleled by face recognition, which is similar in 3 key respects: (1) it is highly practiced, (2) it requires fine-grained visual discrimination of high spatial frequency information, and (3) it requires holistic object recognition. The VWFA is homologous to the fusiform face area (FFA), a portion of the right fusiform gyrus that responds selectively to faces in a way that is similar to the VWFA response to visual words. In young (preliterate) children, the FFA is bilateral; the right-lateralization of the FFA emerges as children learn to read, and this portion of the left fusiform gyrus becomes specialized for visual word recognition rather than face recognition (ie, becomes the VWFA), presumably because spoken language is left-lateralized.

Dyslexias

Acquired Dyslexia

Acquired dyslexia is a disorder of reading in previously literate individuals as a result of brain damage that spares (most) other perceptual and cognitive functions. Acquired dyslexia has 3 main forms. Pure alexia (also called alexia without agraphia or pure word blindness) is typically due to damage to the VWFA and is primarily characterized by slow reading aloud with abnormally large (linear) effects of number of letters. Damage to the white matter fibers that connect the VWFA to the posterior occipital cortex areas (ie, inferior longitudinal fasciculus) also plays an important role in chronic pure alexia, suggesting that VWFA is a contact point between visual processing and language processing. For patients with pure alexia, the reading deficit is paralleled by impaired performance for matching unfamiliar visual symbols (eg, Kanji), consistent with a deficit of holistic visual discrimination of high spatial frequency information.

Phonological dyslexia is a relatively selective deficit of reading (and writing/spelling) nonwords relative to real words, often involving lexicalization errors (eg, gat → “cat”). This pattern reflects increased reliance on semantics during reading due to impaired direct connections between orthography and phonology. Patients with phonological dyslexia often also exhibit phonological deficits in spoken language processing, suggesting a more general phonological processing deficit (although a few notable exceptions to this pattern have been documented). Phonological dyslexia is generally associated with damage to peri-Sylvian speech processing regions (discussed earlier), including the superior temporal lobe, angular and supramarginal gyri, inferior frontal cortex, frontal operculum, and insula.

Surface dyslexia is characterized by impaired reading (and writing/spelling) of irregular words, especially low-frequency irregular words such as pint, with comparatively spared reading of regular words (eg, mint, lint, hint) and nonwords (sint). Spared ability to sound out nonwords and regular words indicates that the system for mapping orthography to phonology remains intact. Intact grapheme-phoneme mapping is reflected by the errors for irregular words, which are typically regularizations (eg, reading pint to rhyme with mint). Surface dyslexia is strongly associated with semantic deficits in nonreading tasks, commonly arising in neurodegenerative diseases (especially sPPA) and rarely due to focal lesions (eg, stroke). Thus, the impairment appears to be specific to a semantically mediated system for reading irregular words, while leaving the direct orthography-to-phonology mapping system intact.

In addition to these main types of acquired dyslexia, reading deficits can also arise as a result of damage to related peripheral (ie, visual) systems. One example of a peripheral dyslexia is neglect dyslexia, in which patients fail to identify the initial portion of a letter string and which tends to occur in patients with left-sided spatial neglect. Deep dyslexia is an intriguing and complex syndrome in which patients produce semantic (eg, castle → “knight”) and visual (eg, skate → “scale”) errors in reading, tend to lexicalize nonwords (as in phonological dyslexia), and perform better when reading concrete nouns (eg, table, chicken) than abstract words or modifiers such as adjectives and adverbs (eg, envy, swiftly). Deep dyslexia may be a particularly severe form of phonological dyslexia, possibly with some additional semantic deficit (although some patients appear to have intact comprehension, arguing against a semantic deficit). Another possibility is that deep dyslexia is a consequence of (not very effective) reading using the right hemisphere. The ability of the right hemisphere to take over language functions in response to left hemisphere damage and whether such compensation is adaptive or maladaptive are active areas of current research.

Developmental Dyslexia

Developmental dyslexia is characterized by difficulties with word recognition and by poor spelling and decoding abilities, in the absence of frank deficits of other perceptual or cognitive abilities and despite the provision of effective classroom instruction. These difficulties typically result from a deficit in phonological processing, although secondary consequences may include problems in reading comprehension and reduced reading experience. The phonological deficits in developmental dyslexia are frequently framed in terms of phonological awareness, the ability to intentionally manipulate phoneme-level aspects of spoken language; for example, a phoneme deletion task such as saying the word “split” without the /p/ sound. Phonological awareness skills in children are early predictors of reading ability, and dyslexic deficits in phonological awareness tend to persist into adulthood. Interventions that strengthen phonological awareness skills are effective strategies for improving decoding in children with dyslexia. Speeded word production (eg, rapid naming of letters, numbers, or colors) and verbal short-term memory are also often impaired in developmental dyslexia.

Dyslexia is highly heritable, with a 30% to 50% chance of being passed from parent to child. Several genes in the DYX2 (dyslexia susceptibility-2) locus on chromosome 6p22 have been associated with dyslexia, including KIAA0319, TTRAP, and DCDC2. Magnetic resonance spectroscopy has revealed N-acetyl-aspartate, choline, and glutamate abnormalities in developmental dyslexia. Neuroanatomic studies have found reduced gray matter volume in the left planum temporale and bilateral posterior superior temporal lobe and inferior parietal lobe, as well as reduced white matter volume in temporoparietal regions, particularly the arcuate fasciculus (superior longitudinal fasciculus) and corona radiata.

Functional neuroimaging studies reveal that, relative to control participants, dyslexics exhibit less activation in posterior portions of the reading system, particularly the inferior occipitotemporal cortex (ie, VWFA), but also the posterior superior temporal gyrus and inferior parietal cortex (angular and supramarginal gyri). In contrast, dyslexics exhibit greater activation of frontal language areas, such as left inferior frontal and precentral gyri. These findings align with the broader set of results showing that reading development is associated an anterior-to-posterior shift: Early (novice) readers show more activity in frontal language areas; as reading skill and automaticity increase, frontal activation decreases and activation of the posterior temporoparietal areas increases. This shift appears to be delayed in dyslexia, either due to (possibly genetic) neuroanatomic abnormalities in those posterior regions or due to cognitive deficits in phonological processing that engages those areas.

CONCLUSION

The neural system supporting human language is centered on the left peri-Sylvian cortex. Building on higher level auditory processing, the superior temporal cortex is responsible for speech perception and spoken language comprehension. The posterior-to-anterior progression of speech sound representations to word and phrase representations is called the ventral speech stream. Inferior frontoparietal cortical regions are responsible for articulatory planning and execution, and this speech production system is closely related to neural systems involved in limb action planning and execution. This dorsal speech stream also follows a posterior-to-anterior progression from speech sound and syllable production (parietal and frontal motor/premotor regions) to sentence planning and production (inferior and middle frontal gyri). Semantic knowledge relies on a distributed system of modality-specific representations of perceptual, motor, and other experiences integrated in “hubs” in the ATLs and TPC. Reading is relatively recent development that requires explicit training and builds on existing systems for spoken language, semantic knowledge, and expert visual object recognition. The visual word form area in the left fusiform gyrus is a critical connection between the visual system and the language system that is particularly important for efficient word reading.

SUMMARY

• In the traditional model, the neural language system is primarily in the left hemisphere, and composed of a posterior superior temporal region (Wernicke area) and an inferior frontal region (Broca area). This model has been substantially extended and revised, largely due to contemporary neuroimaging methods.

• Semantic processing depends on anterior temporal and temporoparietal hubs that integrate information from a distributed system of cortical regions.

• Recognition of speech sounds and spoken words builds on bilateral cortical auditory processing and primarily involves the ventral speech stream in the left superior temporal lobe.

• Production of speech sounds and syllables begins with motor control systems in the cerebellum, basal ganglia, and motor cortex.

• Phonological aspects of speech production rely on the dorsal speech stream inferior parietal regions: angular and supramarginal gyri, somatosensory representations of the articulators in the inferior postcentral gyrus, and inferior precentral gyrus motor representations of the articulators.

• Sentence planning and production (grammatically structuring multiword utterances) relies on inferior and middle frontal gyri.

• The primary white matter tracts involved in language processing are the arcuate fasciculus (superior longitudinal fasciculus), IFOF, uncinate fasciculus, and aslant tract.

• Reading builds on the spoken language system, adding a visual word recognition component primarily in the left fusiform gyrus to the phonological and semantic systems of spoken language.

OBJECTIVES

OVERVIEW

Emotions are the feelings generated by basic instincts and drives and by the assessment of progress or frustration of those drives. Emotions depend on the limbic system. The limbic system, in turn, interacts with the autonomic nervous system to control whether the body should be in the sympathetic fight/flight body state or oriented toward homeostasis. The limbic system also interacts with sensory organs that assess the state of the environment and with the neocortex, which modulates behavior based on contextual memories. A particularly important part of the brain related to emotion and limbic function is the amygdala–ventromedial frontal lobe system that is responsible for assessing emotionally salient risks, such as those associated with living in social groups.

DEFINING EMOTION

Basic instincts and drives make their presence felt by generating emotions that motivate the organism toward some action. Emotional drive is both generated by and produced from the state of the autonomic nervous system. Emotional drives are typically based on immediate survival needs such as thirst, hunger, reproductive activity, and social rank maintenance. Note that some of these are internally generated, such as thirst and hunger, whereas others are externally triggered, such as the appearance, sound, or smell of a potential mate or predator. External emotion-generating inputs are transmitted from cephalic sensory organs. The behaviors triggered by external sensory input often depend on learned associations from previous experience.

Emotions can be described as existing along dimensions such as approach/withdrawal and intensity. Fear and disgust are emotions that are associated with withdrawal behavior, whereas happiness promotes approach. Each type of emotion has an associated intensity that is also experienced. Emotions are associated with body states mediated by different relative levels of sympathetic versus parasympathetic neurotransmitters as well as other neurotransmitters and neuromodulators. According to the James-Lange theory of emotion, emotional responses are processed initially at limbic brain levels and the autonomic nervous system, with cognitive responses occurring later along a slower pathway involving higher brain areas. By this theory, the initial emotional reaction to sensory input should be similar in humans to that of primates and other mammals because we share similar, phylogenetically older limbic brain structures. The cognitive human overlay would be unique because of unique human consciousness and language. It is also clearly the case, however, that cognitive processes, such as imagination, can generate emotions in a top-down manner.

The Uses of Emotion

Emotions bring about behaviors that satisfy basic needs necessary for survival, such as eating and fleeing predation. At this level, emotions exist in virtually all mammals and probably all vertebrates. Seeking pleasure (food, mates) and avoiding pain (predation, heights) are associated with a similar constellation of behaviors in all vertebrates. In nonmammalian, so-called cold-blooded vertebrates, many of the triggers for emotion have evolved to be inborn, such as fear of snakes and spiders and attraction toward the opposite sex. These inborn emotions are likely to be universal across many vertebrates because they are tied to the basic operation of the autonomic nervous system and the limbic system that controls it.

In mammals, there is more emotionally driven generation of risks and rewards based on learned associations from previous experience. This is particularly true in mammalian social hierarchies where every single individual has a unique rank, sometimes across multiple rank dimensions, and all interactions between individuals must follow socially defined rank protocols. In many mammals, emotion is communicated by body posture, facial expressions, and vocalizations.

Emotional Expression

Social mammals communicate more complex emotional repertoires than nonsocial mammals or nonmammalian vertebrates. Darwin, in his book The Expression of the Emotions in Man and Animals, argued that emotional communication is selected for because it aids survival in social groups. Given the similarity in hierarchical structures in primate social groups, it is not surprising that some emotional facial expressions are nearly universal. Figure 22–1 shows 6 faces displaying what are thought to be the 6 universal human facially displayed emotions of fear, anger, sadness, happiness, disgust, and surprise, from the work of Ekman and colleagues.

ANATOMIC AREAS MEDIATING EMOTION

The limbic system in nonmammalian vertebrates is the top of the hierarchy of autonomic control. The limbic system processes and integrates sensory information from internal sensors of body state such as blood pressure, sugar level, and temperature with external sensory input from smell, vision, audition, taste, and skin senses. In its interaction with the ventromedial prefrontal cortex, the limbic system evaluates this sensory input in light of learned associations from experience. The cingulate cortex, which is mesocortex, is an executive area at the top of the limbic system. Two memory areas, the hippocampus and amygdala, are associated with the creation and use of memory by the limbic system. The hippocampus has both limbic and nonlimbic connections for creating and learning the context associated with an event to be remembered. The amygdala processes inputs from the brainstem and autonomic nervous system and interacts reciprocally with the ventromedial (also called orbitofrontal) cortex and the basal ganglia.

The Limbic System

The main limbic areas involved in mood and emotion are shown in Figure 22–2. Executive areas include the cingulate cortex above the corpus callosum and the ventromedial prefrontal cortex in the prefrontal lobe area below the dorsolateral prefrontal cortex. Several temporal lobe cortical areas also interact extensively with limbic areas. The amygdala is a memory area whose relationship to the ventromedial prefrontal cortex is similar to that of the hippocampus with the rest of neocortex. The hippocampus was originally included as an area of the limbic system by Maclean. Although there are connections between the hippocampus and the limbic system, the hippocampus is extensively connected to other, nonlimbic areas of neocortex.

Another major brain area associated with the limbic system is the insula, located medial to the temporal lobe. Some of the main connections between the limbic system and several cortical areas are shown in Figure 22–3. There are also connections between limbic structures and the dorsolateral prefrontal cortex, the seat of working memory.

The overall connection scheme of cortical and subcortical limbic connections is shown in a block diagram format in Figure 22–4. The bold lines indicate the original Papez circuit of the limbic system, which includes the projection of several anterior thalamic nuclei to the cingulate cortex. The input to these thalamic nuclei comes chiefly from the hippocampus via its output tract, the fornix, via the mammillary body. The thin lines in Figure 22–4 show limbic projections mostly discovered more recently that are not part of the original Papez circuit, although the amygdala has long been known to be limbic.

The Amygdala

The amygdala on each side of the brain sits just anterior to the hippocampus. It has 3 main divisions: the basolateral, central, and cortical nuclei, shown in Figure 22–5. The basolateral nucleus receives significant input from the hippocampal formation and projects to medial prefrontal, orbito- (ventromedial) prefrontal, and cingulate cortical areas. The central nuclei receive ascending input from the brainstem and project to the hypothalamus. The cortical nuclei receive inputs from the olfactory bulb and also project to the hippocampus.

The amygdala interacts with the ventromedial prefrontal cortex in a manner somewhat similar to the interaction between the hippocampus and dorsolateral prefrontal cortex. Much of the prefrontal cortex instantiates a working memory of the configuration of the environment associated with a situation about which learning needs to occur. Activated neurons in the prefrontal cortex project to and activate coincidence detectors in the hippocampus and amygdala where synaptic weight changes, typically associated with N-methyl-D-aspartate (NMDA) receptors, allow the reactivation of the prefrontal neurons that activated neurons in the hippocampus and amygdala by reciprocal connections.

How the amygdala encodes situation context for learning is illustrated schematically in Figure 22–6. Neurons in the amygdala receive inputs from different neurons in multiple cortical areas, the hippocampus, and the limbic system. For any given contextual situation, a few neurons in the amygdala “grid” will receive inputs from the arbitrary set of activated neurons associated with that situation. The activated synaptic weights will be increased. Because the connections are reciprocal, similar situations that activate those amygdala neurons will activate the situation-coding neurons that constituted the representation of that situation.

The hippocampus is associated with learning nearly any context, such as navigational routes. The amygdala–ventromedial prefrontal cortex system, on the other hand, is involved in learning about risk and reward situations, particularly situations that are followed by negative consequences. This system creates memories that can engender fear of things such as guns pointed at someone or aggressive facial expressions. Damage to the amygdala does not eliminate fear of intrinsically fearful things such as snakes, heights, or leopards. However, it reduces the appropriate fear reactions to fear that must be learned, such as the sudden appearance of brake lights on the car ahead.

Ventromedial Prefrontal Cortex

The ventromedial or orbitofrontal cortex is shown in the sagittal section on the right in Figure 22–3 (some treatments consider the orbitofrontal cortex to be restricted to only the most ventral part of the ventromedial cortex, with the latter comprising a larger area). Ventromedial prefrontal neocortex is reciprocally connected to the amygdala and is essential for the assessment of risk and the associated emotional response of fear. In addition to the amygdala, inputs to the ventromedial prefrontal cortex come from the dorsolateral thalamus, olfactory bulb, and temporal lobe. Outputs from the ventromedial prefrontal cortex go reciprocally to the amygdala, as well as to the hypothalamus, hippocampus, cingulate cortex, and temporal lobe.

Ventromedial prefrontal cortex damage results in an inability to inhibit inappropriate behavior, particularly learned, socially inappropriate behavior. Humans with ventromedial prefrontal cortex damage often cannot avoid risky behavior because of the lack of a gut feeling of fear associated with the risks. Thus, they are poor at avoiding high reward but high loss activities such as gambling. Humans with ventromedial prefrontal damage are also poor at detecting and avoiding behavior that is socially unacceptable, such as speech laced with pointless profanity.

Another hallmark of ventromedial or orbitofrontal damage is what is called “utilization behavior.” A patient with damage to this part of the brain may, when left in a room with paint cans and brushes, begin painting the wall. The sight of objects triggers their use, even when inappropriate, because the gut feeling normal people get about socially inappropriate behavior is missing. Extreme damage to ventromedial prefrontal cortex is associated with the inability to experience empathy, and some patients with this damage become sociopaths.

There are notable lateralization differences in the prefrontal cortex, including the ventromedial prefrontal cortex. Activity in the left prefrontal cortex is biased toward approach on an approach/withdrawal axis, while activity in the right prefrontal area is associated with withdrawal. Damage to the left prefrontal cortex is thought to tip the approach/withdrawal balance toward right-dominated withdrawal and depression. Damage to the right prefrontal cortex, in contrast, shifts the balance to the left and is associated with manic behavior. These laterality effects are stronger in the ventromedial than dorsolateral prefrontal cortex.

Although the limbic system is phylogenetically old, the ventromedial prefrontal cortex has expanded in primates and humans with the rest of the cortex to mediate more complex emotional processing. It has often been supposed that emotion is the opposite of reason. Emotions, however, are the product of a complex neural processing system that is experience and case based, rather than rule based. The fact that the output of this system is via a low-dimensional feeling matrix rather that a high-dimensional language-based cognitive matrix does not imply that the calculations performed by this system are primitive. The autonomic/limbic/ventromedial prefrontal system acts rapidly before complex conscious processing and awareness, as is necessary for survival. Some aspect of how the car in front of you is being driven causes you to back away without your being aware of doing so or being able to say why, but it saves your life.

PHYSIOLOGIC CHANGES ASSOCIATED WITH EMOTION

Emotions are the “language” by which the limbic and autonomic nervous systems communicate with the rest of the brain. The fact that this language of is low dimensionality compared to conscious, verbal language used by other parts of the brain does not mean that the calculations done by this system are primitive and low dimensional. The guidance given by emotion for appropriate behavior in complex social situations may be based on complex, unconsciously processed cues and nondeclarative memories, including body language, tone of voice, facial expressions, and previous experiences with these elements. Humans experience strong “gut feelings” when their behavior in social groups is deemed unacceptable and may even experience such feelings in anticipation of unacceptable behavior, such as fear of making a mistake in public speaking. We are constantly unconsciously processing facial expressions, body postures, and tones of voice of those around us. Even the contemplation of particular dangerous moves in a cerebral exercise such as playing chess can evoke such feelings.

The output of the emotion-processing system tends to be mostly describable along the approach/withdrawal and intensity axes, which can be partly mapped onto the autonomic nervous system sympathetic/parasympathetic polarities and intensities for each division.

High & Low Road Inputs

The American neuroscientist Joseph LeDoux has proposed a low and high road theory for sensory processing. In this scheme, information that is sent by cephalic sensors to the thalamus for transmission along the slow high road to the neocortex is also sent in parallel on the fast low road to the amygdala for more immediate action. Olfactory input, for example, which is dominated by emotional approach/withdrawal motivation, projects directly to the amygdala–ventromedial prefrontal system, which can trigger avoidance behavior before conscious awareness of the identity of the smell, which is based on slower high road cortical processing. The block diagram of olfactory bulb projections in Figure 22–7 shows the direct projection from the olfactory bulb to the amygdala versus the more complex olfactory projection system that involves prefrontal and frontal cortex.

A similar situation exists for other senses in which a high road projection to thalamus and neocortex is paralleled by a low road projection to the limbic system. The low road system activates both the autonomic nervous system and initiates behavior before conscious processing informs us of the identity of the stimulus. Danger in both physical and social situations activates the amygdala–ventromedial system.

Autonomic Responses

LeDoux’s term for the output of the amygdala–ventromedial system is somatic markers, what we usually call “gut feelings.” These have a strong autonomic component, often switching the body from homeostasis (parasympathetic) to fight/flight sympathetic activation. Sympathetic activation increases activity in the hypothalamus-pituitary-adrenal (HPA) axis, leading to internal changes such as increased heart rate, sweating, and pupil dilation. A general term for activators of the HPA axis is stress, as shown in Figure 22–8. An important transmitter in this system is cortisol. Cortisol levels increase during stress, which acts as a negative feedback signal to reduce adrenal activation. However, chronic stress associated with continuously elevated cortisol levels that is not effective in reducing sympathetic activation has negative effects on the body, such as impairment of immune function, heart disease, and impairment of learning and memory.

AFFECTIVE DISORDERS

Patients with damage to ventromedial prefrontal cortex lack normal aversion to risky behavior, such as gambling, but are cognitively aware that the behavior may be detrimental. This type of damage is also characterized by utilization behavior. In contrast, the hallmark of amygdala damage is an extreme loss of any fear.

Sociopathy

There is considerable overlap between psychopathy and sociopathy, with the consensus being that sociopathy is a weak form of psychopathy. A unifying scheme for defining psychopathy is that it is characterized by a lack of gut feelings of remorse or empathy. It may occur in individuals who are often otherwise above average in intelligence. Overlapping sociopathic traits include lack of emotional attachment and manipulativeness. Given the considerable evidence that damage to the amygdala–ventromedial system produces such traits, it seems highly likely that unknown damage to this system underlies most sociopathic behavior. Such damage may be organic, from, for example a tumor or head blow, or possibly secondary to environmental effects such as child abuse.

Autism

Autism is a spectrum disorder, probably originating from mutations at more than 10 gene loci and arising from multiple other causes. As a spectrum, it ranges in severity from extremely mild (the absent-minded college professor) to profound mental retardation. Unifying concepts for autism include Simon Baron-Cohen’s idea of an autistic lack of theory of mind, the ability to comprehend and react to another’s point of view and intentions. This is exhibited as antisocial behavior that, although once was interpreted as maliciousness, is now seen as an inability to comprehend that others have feelings that can be hurt. Autistic people are thought to be socially withdrawn and inept because of an inability to act in a socially appropriate manner, due to an underlying perceptual deficit.

Baron-Cohen has extended the theory of mind idea of autism to the idea of autism as an ultra-male brain. Females, on average, score much higher than males on tests of empathy, while males score higher on tests of systemizing, such as spatial manipulations and mechanical knowledge. An essential point of this idea is that both males and females exist on a spectrum of male versus female “brains,” with some females having more male-type brains, and some males with more female-type brains. His theory then is that autistics, whether male or female, have brains at the extreme male end, with very poor empathy, but sometimes savant abilities in calculation or art. Some data have suggested that prenatal exposure to high testosterone tends to produce male brain types and associated autistic traits in humans and laboratory animals. These effects may be mediated by underdevelopment of the ventromedial prefrontal system that underlies social intelligence. As such, autism would be viewed as a deficiency in emotional intelligence or perception, not a lack of emotional experience.

CONCLUSION

The autonomic and limbic systems evolved to control the balance between homeostasis and species-typical action behaviors such as fight, flight, and reproduction. Highly social mammals, such as primates, have extensive autonomic reaction to risks and dangers associated with social interactions and signals, many of which are learned. The amygdala–ventromedial system and the anterior cingulate cortex are high-resolution executives that reciprocally interact with limbic and lower structures to mediate complex social and emotional intelligence. Although the outputs of these systems are embedded in low-dimensional matrices such as gut feelings, their unconscious computations are high dimensional, complex, and necessary for existence in a complex social environment.

SUMMARY

• Emotions are feelings generated by drives and instincts and by assessment of progress toward those drives.

• Emotion depends on the limbic system, the middle component of the proposed evolutionary triune structure of the brain.

• Emotional reactions can be characterized along the dimensions of valence and intensity.

• According to the James-Lange theory of emotion, emotional reactions precede cognitive appraisal.

• In mammals, emotions are particularly evoked by social hierarchy situations.

• Approximately 6 basic human emotions are thought to be universal.

• Emotional expression involves activation of the limbic and autonomic nervous systems.

• The amygdala is crucial for emotional learning via its interaction with ventromedial prefrontal cortex.

• According to LeDoux’s “low road” theory, emotions are processed by a rapid path from the thalamus to other subcortical structures such as the amygdala, independently from a slower, but more accurate “high road” path involving the projection from thalamus to neocortex.

• Affective disorders are characterized by disturbances in emotional cognition and behavior.

OBJECTIVES

INTRODUCTION

The result of billions of years of evolution, the innate biological clock is a nearly ubiquitous feature of life on Earth. The conservation of its basic function—the maintenance of a stable relationship between the organism’s internal physiologic processes and the environmental light cycle—across such a wide swath of species speaks volumes about its importance. This timekeeping mechanism is not simply a response to changing light, but rather an innate clock that responds slowly and predictably to changes in the architecture of daily light cycles. Behaviorally, this clock allows organisms to predict daily changes in the environment and to react accordingly. Physiologically, the clock serves as a master regulator of many processes, providing a temporal pattern for internal organization and output. In humans and other mammals, the master circadian clock is located in the suprachiasmatic nuclei of the hypothalamus—a pair of densely packed nuclei receiving direct light input from specialized cells in the retina and other indirect timing and physiologic information from a slew of other areas in the nervous system and body. In this chapter, we will explore the basic properties of circadian and seasonal rhythms, the more complex molecular constituents of the clock, and finally, their impact on human health.

EVOLUTIONARY ORIGINS

At the formation of the solar system, the planetary nursery in the accretion disk had a natural motion, orbiting our nascent Sun and serving as the birthplace of dozens of planetoids colliding with each other like a game of cosmic pinball. These planetoids had a natural rotation parallel with the orbit of the accretion disk, but these planetary rotations tended to vary wildly as their axes wobbled in space or they slowed until they were tidally locked with their star. For Earth, the planet’s rotation was cemented and stabilized approximately 4.4 billion years ago when a Mars-sized planetoid named Theia crashed into it and formed our moon. The Moon stabilized Earth’s axis, slightly off perpendicular to the planet’s orbit around the Sun. The Moon also gave us tides once the planet had cooled enough from the impact that liquid water oceans could form. It is through geologic evidence of these powerful tides—when first formed, the Moon was 6 to 10 times closer than present—that we know that the early Earth spun at the absolutely blistering pace of 4 hours per day. Over time, as the radius of the Moon’s orbit drifted outward to its current position, the drag of the lunar tide slowed Earth’s rotation to the 24 hours we now know. Because circadian clocks can be found in life-forms as ancient as cyanobacteria and in most living things today, including humans, one cannot help but appreciate that they must have given early life an advantage. In fact, a TimeTree of Life analysis would indicate that the most recent common ancestor of cyanobacteria and humans lived some 4.1 billion years ago (with estimates as early as approximately 4.3 billion years ago), when we know that the Earth’s rotation was still much accelerated. Along with the extraordinary discovery of 4.1 billion-year-old organic carbon in 2015, this evidence suggests that the earliest life on Earth not only had to deal with its harsh and downright inhospitable environs, but also had to be able to synch their daily cycle with a period of only several hours. Yet, genes used by animals to control parts of their clocks are exaptations of older, more ancient genes that can be found in bacteria and are being used for such processes as DNA repair.

As such, the day–night cycle is a critical environmental factor shaping the evolution of life on Earth. In Earth’s tumultuous history, extraterrestrial impacts such as the Moon-forming event, along with geologic upheaval and vulcanism, were not uncommon. Although early life evolved in the water, these factors wreaked havoc on the ability of the Sun’s light to reach the oceans, at times darkening the planet for many years. In at least 3 extreme instances in the planet’s history, the entire globe was covered with a sheet of white ice that was miles thick in some places—the so-called “Snowball Earth.” The last of these deep ice ages lasted millions of years 800 to 600 million years ago, bouncing most of the Sun’s light back into space and trapping early life in a dark ocean. Yet we find in the present that the endogenous circadian clock is an almost ubiquitous feature of life on Earth, suggesting that it would have offered some increased fitness, even then. How did circadian rhythms become the rule rather than the exception? It is thought that the ability to anticipate day/night transitions and coordinate internal metabolic processes was advantageous for the earliest life on the planet. The evolution of photosynthesis led to large amounts of oxygen in the atmosphere and an ozone to filter damaging solar radiation. Before that time, it is thought that organisms used their circadian clocks to judge surface safety by sinking to deeper water during the day and surface feeding at night—the so-called “escape from light” hypothesis proposed by Colin Pittendrigh, who was a founding father of the chronobiology field. By tracking light–dark transitions, later photosynthetic organisms could capture the Sun’s light and fix carbon on a regular schedule. Organisms that fed on these photosynthesizers shared the ancestral characteristic and used their own clocks to time their feeding behavior. Regardless of the predator and prey relationship, an endogenous circadian rhythm allowed an organism to maintain a fairly consistent relationship (or phase angle) with the external day–night cycle, despite changes in the primary time cues (or zeitgebers). Circadian clocks could be entrained, rather than purely driven or synchronized, allowing the organismal clock to generally keep time despite irregularities in weather and seasons and/or the absence of light, such as due to volcanic ash, asteroid impact, or even Snowball Earth. This stable phase angle provided a mechanism for adaptive flexibility to the ever-changing geophysical day: the ability to predict time of day persistently, even in periods of bad weather, geologic upheaval, or severe climate change. Thus, entrainment offered some fitness that allowed feeding or photosynthesis to occur on a generally regular schedule. This last point explains why completely divergent species such as cyanobacteria and humans evolved the ability to produce circadian rhythms, despite the fact that their last common ancestor lived some 4.1 billion years ago when the length of the day was far shorter than 24 hours. Either circadian rhythms were so important to early evolving life that they persist in most species today, or circadian rhythms have become so important that convergent evolution has made them a general property of life on Earth.

Seminal work by Carl Johnson and colleagues elegantly demonstrated in cyanobacteria that circadian clocks do offer competitive fitness when the endogenous rhythm is resonant with the external environment. In mammals, where such direct competition experiments are difficult, Pat DeCoursey made a fortuitous observation in captive ground squirrels that having a functional circadian clock did indeed decrease mortality rates: While the squirrels could not get out of a pen, a predatory weasel somehow got in and disproportionately picked off clock-lesioned squirrels while at the same time providing invaluable evidence for the importance of a behavioral circadian rhythm in mammals. This incident inspired DeCoursey and her team to monitor the behavior of wild-type and clock-lesioned ground squirrels using Global Positioning System tags in the wild, further bolstering the initial results. In addition to compromised survival due to predator–prey relationships, the circadian clock can also offer fitness due to overall health of the organism. Recent work in rodent species shows that certain genetic mutations can change the speed of the clock. For example, mice with a mutation that causes a faster clock with a shorter period (ie, 22-hour cycle) die from heart and kidney failure at earlier ages when housed in conditions that provide a standard, 24-hour, light:dark (LD) 12:12 photoperiod. Remarkably, these mutant mice appear to age and live normally when housed in a 22-hour T-cycle with an LD 11:11 photoperiod.

Thus, the mammalian biological circadian clock, located in the hypothalamus of the brain (see later section on the suprachiasmatic nucleus), is an indispensable component of life in the wild. It would follow, then, that understanding how mutations in circadian genes disrupt the function of behavioral and physiologic rhythmicity is important not only on the merits of scientific interest, but also for its implications to human health.

PROPERTIES OF CIRCADIAN RHYTHMS

By definition, circadian rhythms are the physiologic processes whereby an organism regulates its activity and coordinates other physiologic processes on a self-sustained, daily cycle. The word circadian comes from the Latin words circa, which means “about,” and diem, which means “day.” Behavioral rhythms govern several characteristics in animals: feeding/fasting, drinking, waking/sleeping, attention, cognitive performance, torpor, and body temperature, among others. Physiologic circadian rhythms can be found in hormone release, immune system function, and even cell division. Presently, a circadian rhythm has 3 fundamental properties: (1) the rhythm is self-sustained and approximately 24 hours in length; (2) the rhythm is temperature compensated; and (3) the rhythm is entrained by external factors such as light.

The first property of circadian clocks was initially observed by a French scientist named Jean-Jacques d’Ortous de Mairan, who was fascinated with the daily leaf movements of the heliotropic plant, Mimosa pudica, and so he examined whether his observations were simply a direct response to the sunlight or something more ingrained in the plant. After placing the plant in a dark cabinet, he observed that daily leaf movement persisted, thus illustrating the first characteristic of a circadian clock: persistence of the rhythm in constant conditions (generally light and temperature) at a near 24-hour period. In rodent models, this endogenous period is most often assessed by behavioral monitoring—such as activity recorded by wheel running (Figure 23–1, left), infrared motion detection (Figure 23–1, right), laser beam break, or water drinking—in constant darkness. Mice, for example, are the nocturnal species that exhibit activity during the dark of night. If placed in a monitored cage in constant darkness, mice will generally exhibit a free-running period (FRP or τ) of slightly less than 24 hours and continue to run only during what they perceive as “night.” The same form of activity monitoring (known as actigraphy) may be used to assess behavioral rhythms of humans, but the FRP of diurnal humans tends to run slightly longer than 24 hours. Even in humans, wakefulness occurs during what they perceive as “day” in the absence of photic or social cues in constant routine conditions in which people are kept awake with a constant posture in continual dim ambient light, with hourly meals. The imperfect, non–24-hour clock timing found across the animal kingdom has a purpose: The closer an animal’s FRP is to 24 hours, the more difficult it is to entrain to the 24-hour, light–dark cycle.

The second property of circadian rhythms is rather remarkable: Unlike other basic biochemical reactions that increase their reactivity as the temperature increases—usually 2- to 3-fold for every 10°C increase—biological clocks are seemingly unaffected by temperature. In other words, the clock continues to run at a normal periodicity regardless of ambient environmental temperature. How the clock manages to accomplish this feat is rather a mystery, but it may have something to do with redundancy built into the molecular feedback loop (see later section titled “Molecular Architecture of the Circadian Clock”). Regardless, as discussed on the topic of clock evolution, the clock must be able to maintain its innate circadian rhythm at or near its native period and be able to process incoming signals from the environment. That latter property is the third and final characteristic of a circadian rhythm.

The light–dark cycle created by the Earth’s rotation is a natural, robust, and reliable oscillation. In the modern world, each day is exactly 24 hours. In a seminal series of papers on mammalian rhythms by Colin Pittendrigh and Serge Daan, the fourth paper suggests that conservation of the phase relationship between the master oscillator (Earth’s rotation—Φ) and its slave oscillator (the circadian biological clock—φ) is the essence of entrainment. This relationship is called phase angle (ψ) and is arguably the most important aspect of a circadian rhythm. A zeitgeber (German for “time giver”) is an external cue that sets or aligns the internal biological clock to the external 24-hour light–dark cycle. The most ancient and influential of all zeitgebers is light, which readily adjusts the phase of the biological clock such that the internal circadian rhythm is synchronized to the environmental stimulus. In humans, entrainment manifests as a behavioral period equal to the period of the 24-hour entraining stimulus with a stable phase relationship (ie, phase angle) between the circadian rhythm and the environment. There are 2 proposed models of entrainment: the continuous (parametric) model and the discrete (nonparametric) model. Both models have been championed by founders of the field, and both have merits and drawbacks. In the continuous model, Jurgen Aschoff suggested that the intensity of light proportionally changed the speed of the biological clock in a phase-dependent manner, thus squeezing or stretching the internal circadian clock period to fit into the actual environmental day. In contrast, Colin Pittendrigh suggested that a discrete model could explain entrainment of the circadian clock by simply shifting the phase (Δφ) or timing of each cycle by a specific amount. This model used a phase response curve (PRC) to explain phase-dependent shifts to discrete pulses of light that, alone or in combination, added up to the period of the environmental day (Figure 23–2). Thus, in a 24-hour light–dark cycle, the internal circadian period (τ) is equal to the external 24-hour cycle of light and darkness (T).

In Pittendrigh’s discrete model, a zeitgeber adjusts clock phase by an amount of time that corrects for the difference between the period of the master oscillator and the internal circadian clock. Light phase shifts a particular rhythm (eg, hormone levels, neuronal activity, behavioral activity levels) by an amount that depends on when the light is presented with respect to the animal’s subjective time, as defined by the animal’s internal circadian clock. With regard to light, the largest phase delays occur when presented during the early subjective night, and the largest phase advances occur when light is presented during the late subjective night, just before dawn. The functional result of a phase delay is that the rhythm peaks later than it would have if there had been no stimulus given. When a phase advance occurs, the rhythm peaks earlier than it would have if there had been no light stimulus given. The relationship between when a light pulse is given during the animal’s subjective day or night and the resulting phase shift can be plotted using a PRC, in which the time of the light presentation (with respect to the internal circadian day) is on the x-axis and the size of the phase shift is plotted on the y-axis (see Figure 23–2).

The ability to entrain to the geophysical day with a constant phase angle is of paramount importance, especially when considering that the amount of time between dawn and dusk changes with the seasons. These changes in photoperiod, or the amount of daytime relative to nighttime over a 24-hour period, may perturb the phase relationship between the master and slave oscillators. Further work on mammalian entrainment by Pittendrigh and Daan sought to address how phase angle is conserved across changes in photoperiod due to season (see later section titled “Seasonality”). They proposed that the circadian clock consists of 2 coupled oscillators—a morning and an evening oscillator. They suggested that each oscillator is entrained to a light–dark transition (either dawn or dusk) and that the relationship between these oscillators accounts for photoperiodic encoding. For diurnal species such as humans, where the majority of activity is relegated to the daytime, dawn is referred to as zeitgeber time 0, or ZT 0, regardless of the local time at which dawn occurs. Likewise, the onset of activity in diurnal animals is referred to as circadian time 0, or CT 0, when the individuals are left to free-run in constant conditions. Rather confusingly, dusk is referred to as ZT 12 regardless of what the local time is for primarily nocturnal species, whose activity is generally isolated to nighttime. The onset of activity for nocturnal species is therefore referred to as CT 12 when the animal is left to free-run in constant conditions (constant darkness).

These distinctions between how we refer to the passage of time on a circadian scale relative to the local time are difficult but important to understand. For example, consider 2 nocturnal mice with different internal circadian periods and therefore different onsets of activity relative to the local time. If a person in Bangor, Maine, were to check on the behavior of these 2 mice at local sunset on the first day of summer (ie, 8:25 pm Eastern Daylight Time [EDT] or 20:25 EDT), that would correspond to ZT 15:35 for that person, but the internal time for the 2 mice could be very different, depending on their native period and how long they have been in the dark (eg, CT 3 or CT 16). In other words, it’s complicated. The further away from the equator one goes, the bigger is the photoperiodic disparity between seasons: In the continental United States, its northernmost city of Lynden, Washington, has a summer photoperiod of LD 16.2:7.8 but only goes through a period of astronomical twilight rather than “night,” whereas the southernmost incorporated place in the United States, Key West, Florida, has a summer photoperiod of only LD 13.7:10.3 with full night and day. It is difficult enough to keep track of the nuances of seasonally changing and latitude-dependent photoperiods to even consider how the circadian clock actually processes the information. How seasonality and photoperiodism are encoded in mammals will be discussed in the later section titled “Seasonality,” along with the rules outlined by Pittendrigh and others concerning how entrainment and the conservation of phase angle persist at a tissue or cellular level.

SUPRACHIASMATIC NUCLEUS: THE CENTRAL MAMMALIAN PACEMAKER

In the latter half of the 20th century, the nascent field of chronobiology picked up speed on the mammalian front with the publication of classic and brilliant works by the field’s founding fathers. Colin Pittendrigh and Serge Daan published a masterpiece on the properties of rodent behavioral rhythms. Jurgen Aschoff furthered our understanding of human rhythms and the effects of light on the speed of the clock, postulating what Pittendrigh would later call Aschoff’s Rules, and Franz Halberg was finding rhythms in all types of organisms. What eluded these investigators, however, was the specific location of the mammalian biological clock. In 1967, Curt Richter published a study suggesting that ablation of the hypothalamus led to circadian behavioral arrhythmicity. Exactly who should receive credit for the subsequent discovery of the central pacemaker is still hotly debated, but in 1972, 2 groups independently identified the suprachiasmatic nucleus (SCN) of the hypothalamus as the biological clock in mammals. Robert Moore and Nicholas Lenn identified the retinohypothalamic tract projecting from the retina—long thought to be the primary sensor of circadian light in mammals—to the SCN. Simultaneously, Friedrich Stephan and Irving Zucker followed Richter’s lead and concluded that the SCN was the pacemaker by electrolytic lesioning of specific nuclei in the hypothalamus. Only lesioning of the SCN produced behavioral arrhythmicity. Progress in the field quickened as the SCN was found to exhibit rhythmicity of neuronal firing rates in vivo and in vitro. The SCN was firmly established as the master circadian pacemaker in mammals by transplant studies that rescued behavioral rhythmicity in SCN-lesioned animals and even conveyed upon the transplant recipient the donor’s circadian period.

The SCN is medially located in the anterior hypothalamus, dorsal to the optic chiasm and inferolateral to the third ventricle in mice (Figure 23–3A). Each nucleus is approximately 200-µm wide at its widest point on the coronal plane, 200 to 250 µm dorsoventrally, and approximately 500 to 600 µm from the rostral to caudal extremes. It is densely packed with about 10,000 small (approximately 10-µm wide) neurons, most of which are γ-aminobutyric acid (GABA)-ergic. Due to their density, these SCN neurons are readily identifiable in a coronal mouse brain section as translucent bulbs above a darker optic chiasm (Figure 23–3B). These neurons increase their spike frequency during the circadian day and upregulate nighttime firing rate in response to phase-shifting light pulses.

SCN cells have several characteristics that make them unique from other cells. One distinguishing feature is that SCN cells receive photic information directly from the retina, allowing them to remain entrained to the 24-hour light–dark cycle. The SCN sends efferent projections to other hypothalamic nuclei, ultimately influencing pineal gland function and melatonin release. Second, SCN cells have network properties that allow them to synchronize their activity with one another due to neuronal firing, chemical synapses, and gap junctions. Third, these small compact neurons are capable of regeneratively firing action potentials (eg, pacemaking). The frequency of these electrical events exhibits a 24-hour rhythm at the population and single-cell level, even when neurons are maintained in a low-density culture that minimizes synaptic communication.

The SCN Is Innervated by the Retina & Is Characterized by Anatomically Localized Neuropeptides

As mentioned earlier, the mammalian circadian clock is entrained to the external environment primarily by light through phase delays in the early night and phase advances in the late night. This phase-shifting process begins with photic activation of conventional photoreceptors and specialized melanopsin ganglion cells within the retina. These intrinsically photosensitive ganglion cells then signal photic information directly to the SCN via glutamatergic innervation by the retinohypothalamic tract. Glutamate and pituitary adenylate cyclase–activating peptide (PACAP) released from retinal terminals activate postsynaptic N-methyl-D-aspartate (NMDA) receptors (as well as VPAC₁, VPAC₂, and PAC₁ receptors), resulting in an influx of calcium, membrane depolarization, and an increase in spontaneous action potential firing. Early work suggested that these retinorecipient neurons are located in the ventrolateral core of the SCN (see Figure 23–3B). However, more recent studies suggest that projections from melanopsin ganglion cells actually innervate the entire nucleus more extensively than previously thought. There are slight species-specific differences in the pattern of innervation, but most rodent models follow a general layout. Studies on light-induced expression of the immediate-early gene c-Fos suggest that many retinorecipient cells express vasoactive intestinal polypeptide (VIP) and gastrin-releasing peptide (GRP). A conserved anatomic pattern of neuropeptide expression exists in the SCN from multiple rodent models. In addition to the retinorecipient core, the SCN is also characterized by a dorsomedial shell that produces arginine vasopressin, met-enkephalin, and angiotensin II. Core neurons innervate not only each other, but also the shell of the same nucleus. There are also internuclear connections between the 2 cores and the 2 shells.

VIP Plays a Key Role in Circadian Entrainment & Behavior

Prepro-VIP is the precursor from which VIP and peptide histidine isoleucine are derived, and both are structurally similar to PACAP. VIP is a secreted peptide in the glucagon family that is expressed rhythmically in the core cells of the SCN, with a peak during the night in 12:12 LD conditions, but not in constant lighting conditions (constant darkness [DD] and constant light [LL]). All 3 share binding affinity for 3 receptor types in the brain—VPAC₁, VPAC₂, and PAC₁—but VIP has a higher specificity for the VPAC receptor subtypes. VPAC₂ (gene name Vipr2) is a G-protein–coupled receptor that is highly expressed in the mammalian SCN, and its mRNA shows a biphasic pattern of expression in LD and DD. Signaling through this receptor activates adenylyl cyclase to increase the concentration of cyclic adenosine monophosphate (cAMP), and it has been shown that rhythms in firing activity in the SCN are responsive to the phase-dependent phase-shifting effects of VIP and VPAC₂ agonists via protein kinase A and mitogen-activated protein kinase (MAPK) in vitro. Exposure to constant light has been shown to significantly depress VIP concentrations in the rat SCN in a light dose–dependent manner. This result is supported by in vitro data suggesting that NMDA phase-delays neuronal firing activity in rats and causes a drop in the VIP content of core cells. These same cells respond to pulses of light during the dark period by upregulating transcription of the core circadian clock genes (Per1 and Per2; see later section titled “Molecular Architecture of the Circadian Clock”), producing a phase shift. Previous studies have shown that VIP is sufficient to phase-advance and alter electrical activity when applied alone at CT 20 to 24 in vivo and moderately sufficient to phase-delay when applied alone at CT 12 to 14 in hamsters. Altogether, these results strongly suggest a key role for VIP in photic entrainment and for VPAC₂ as the essential VIP receptor in the mammalian circadian clock.

In addition to VIP, GRP and GABA also contribute to SCN coupling. For example, mice lacking VIP receptors (Vipr2^–/–) and VIP (VIP^–/–) have arrhythmic SCN electrical activity, along with arrhythmic or severely disrupted wheel running activity. Furthermore, daily application of a VPAC₂ agonist restores molecular rhythmicity and synchrony of firing rate rhythms to VIP^–/– SCN neurons, indicating that VIP is mediating SCN coupling and behavioral rhythms. When GRP is applied to SCN cultures from Vipr2^–/– mice, SCN synchronization is also restored, suggesting that other neuropeptides are important in this network. Finally, daily GABA application to 2 dissociated neurons from the same culture results in resynchronization of firing rate, indicating that GABA is sufficient for SCN coupling. However, in the intact SCN network, GABA_A receptor activation slows resynchronization after SCN organotypic cultures are forced into a state in which 2 populations in a single nucleus are uncoupled. Ironically, decoupling induced by VPAC₂ antagonists can be reversed by blockade of GABA_A receptors, suggesting that GABA signaling also mediates VIP-induced synchronization.

The Mammalian Circadian Clock Can Be Set by Nonphotic Stimuli

In addition to light, nonphotic zeitgebers or stimuli are also able to entrain the circadian clock. Such nonphotic stimuli include arousal-producing stimuli and exercise such as sleep deprivation, novel wheel running activity, saline injection, and intermittent shaking of cages. When these stimuli are presented during the subjective day, rodents will advance the onset of their activity on following days. The presence of a nonphotic stimulus is conveyed to the SCN through the following 2 main pathways: (1) a geniculohypothalamic tract (GHT) and (2) the median raphe nuclei pathway. The GHT originates from the thalamic intergeniculate leaflet (IGL) and uses neuropeptide Y (NPY), GABA, and endorphins as neurotransmitters, whereas the dorsal and median raphe nuclei pathway, which projects to the IGL, is primarily serotonergic and projects to the SCN either directly (median raphe) or indirectly (dorsal raphe) via the IGL.

The large phase shifts that occur during the subjective day when animals are presented with nonphotic stimuli are mediated, at least in part, by NPY release into the SCN from the IGL. During the subjective day, NPY itself can produce large phase advances in vivo when injected into the SCN region or applied in vitro directly to SCN slices. Furthermore, confining hamsters to a small box could not block the phase advances that were caused by NPY injections into the SCN during the subjective day, demonstrating that activity or exercise is not needed to induce a phase advance when the SCN is injected with NPY. In a second experiment within the same study, hamsters were also exposed to a novel wheel for 3 hours, with the experimental group being injected with NPY antiserum into the SCN and the control group being injected with normal rabbit serum into the SCN. Hamsters injected with normal serum exhibited phase advances similar to those of the unoperated hamsters, suggesting that NPY conveys nonphotic information to the SCN, whereas the hamsters injected with the antiserum to NPY shifted their activity by <15 minutes.

Early studies showed that VIPergic neurons are under the influence of serotonergic innervation and that treatment with the serotonin-depleting reagents para-chlorophenylalanine or 5,6-dihydroxytryptamine causes a marked decrease in VIP immunoreactivity—an effect that can be reversed with a 5-hydroxytryptamine (5-HT) receptor 1B agonist. This depletion initiates the rhythmic property of VIP mRNA expression in DD and suggests that serotonin works antagonistically to light. The increase of available serotonin by treatment with monoamine oxidase (MAO) inhibitors or by knocking out the MAOA gene increases VIP expression in rats and mice, respectively. 5-HT_1B receptors are localized to the retinohypothalamic tract, suggesting an indirect effect on VIP expression by modulation of glutamate release. Similarly, 5-HT_1A or 5-HT₇ receptors are located on the core neurons of the SCN and allow serotonin to directly modulate VIP release and neuronal activity. Because serotonin plays a role in nonphotic entrainment and modulates the SCN directly, mice with VIP signaling deficiency—and thus lacking robust responses to photic stimuli—may have enhanced responses to nonphotic stimuli.

The Efferent Projections of the Mammalian Circadian Clock Include Both Paracrine & Synaptic Signals

As mentioned earlier, the SCN is both necessary and sufficient for driving 24-hour rhythms in physiology and behavior, including rhythms in sleep–wake activity, energy homeostasis, heart rate/blood pressure, body temperature, and hormone release. Thus, it is not surprising that the SCN projects to a wide array of brain regions and impacts the entire brain through both direct and indirect neural pathways. A classic study showed that blocking the SCN output signal (ie, sodium-dependent action potentials) with tetrodotoxin (TTX) will dissociate the “hands” of the clock, while the “gears” keep on turning. Specifically, hamsters received chronic delivery of TTX to the SCN for several weeks while housed in running wheels in DD. During the TTX delivery, 24-hour locomotor rhythms were eliminated; however, they were reinstated after TTX was discontinued. Remarkably, the phase of the reinstated locomotor rhythm was predicted by the pre-TTX phase, indicating that the internal clock had been running all along, even though the outward manifestation of the clock was disrupted.

Numerous brain regions receive direct synaptic connections from the SCN, including the median preoptic area, the bed nucleus of the stria terminalis, the IGL, the paraventricular nucleus of the hypothalamus, the dorsal medial nucleus of the hypothalamus, and the arcuate nucleus. A primary direct efferent pathway from the SCN is the subparaventricular zone, which is also a putative “switch” for diurnality. Indirect pathways to the pineal gland and pituitary are critical for 24-hour rhythmic release of hormones such as melatonin and corticosterones, respectively. The pathway from the paraventricular nucleus is important for 24-hour control of the autonomic nervous system. Altogether, this system produces 24-hour rhythms in nutrients (eg, glucose, fatty acids) and hormones (eg, insulin, glucagon, leptin, adiponectin, ghrelin).

Hormones, regulation of body temperature, and feeding behavior have all been implicated as possible mediators of the efferent resetting signal from the SCN to downstream neural and peripheral targets. Humoral output of the SCN may also be a mediator of peripheral tissue synchronization. A classic study lesioned the SCN in hamsters and then restored locomotor rhythms by transplanting fetal SCN “micropunches” into the ventricles of arrhythmic animals. Furthermore, to determine whether restoration of rhythmicity was dependent upon neural outgrowth or humoral signaling, the SCN tissue was contained within a semi-permeable capsule before transplantation, and the same result was observed. However, it is important to note that not all circadian rhythms are restored by SCN fetal transplants; specifically, rhythmic release of glucocorticoids, melatonin, and luteinizing hormone fails to recover. Taken together, these results indicate that both humoral and neuronal signals are important for synchronization of peripheral clocks. Several candidate releasable factors proposed are transforming growth factor-α, prokineticin 2 (PK2), and cardiotrophin-like cytokine. For example, PK2 expression peaks during the subjective day in the SCN, and injection of PK2 to the SCN region suppresses locomotor activity in rats at night but results in an increase in activity during the day.

CIRCADIAN MULTIOSCILLATOR SYSTEM

So far, we have introduced the circadian system as relying on a single, primary master oscillator. However, research over the past 2 decades has revealed a much more complex system in which individual cells and tissues, both neural and peripheral, have the ability to generate 24-hour rhythms in isolation from the SCN. Rhythmicity of these secondary oscillators frequently damps out after several cycles due to reduction of both cellular clock gene expression amplitude and tissue-level amplitude, due to the individual cellular oscillators drifting out of phase with one another. Thus, the SCN provide an important daily resetting signal to the secondary pacemakers in the brain as well as in tissues throughout the body such as the lung, liver, heart, and spleen. The hierarchical structure of the circadian system is also evident during resynchronization following a large shift in the light–dark cycle—simulating jet lag. Using a bioluminescent reporter of clock gene activity (see later section titled “Molecular Architecture of the Circadian Clock”), researchers have been able to track ex vivo rhythms of various tissue clocks in real time. Explants of numerous peripheral and neural tissues have been cultured following a 6-hour phase advance of the light–dark cycle at 1, 3, and 6 days following the shift. Several studies have now shown that the SCN rather quickly reentrains to the new light–dark cycle, whereas other tissues take much longer. Locomotor behavior generally takes 1 day for each hour of jet lag to adjust, so a 6-hour phase shift would take at least 6 days for full reentrainment. These ex vivo clock gene reporter studies show that for some tissues, such as the liver, 6 days is not sufficient for reentrainment.

The first autonomous neural oscillator outside of the SCN to be discovered was the retina. The retinal clock is critical for adaptation to changes in environmental lighting, including dark adaptation (via melatonin) and light adaptation (via dopamine). Not surprisingly, these neurotransmitters can also directly regulate retinal circadian phase. Since then, other extra-SCN brain regions have been identified as having daily oscillations in clock gene expression. These include nuclei in the thalamus, hypothalamus, amygdala, olfactory bulb, cerebellum, hippocampus, and many others (Figure 23–4). Outside of the SCN, the olfactory bulb has been shown to maintain the most robust rhythms ex vivo in clock gene expression reported using bioluminescence. In situ hybridization studies of the rat brain have reported rhythmic clock gene expression in numerous regions of the forebrain (prefrontal cortex, rostral agranular insula, paraventricular nucleus, amygdala, and hippocampus). The majority of the extra-SCN oscillators have circadian phases that are delayed by 4 to 12 hours from the SCN.

Among the forebrain regions mentioned earlier, the hippocampal clock has been a primary focus of research in the past few decades because of documented 24-hour rhythmicity in neurogenesis, long-term potentiation, and signaling cascades that are important for memory formation (eg, MAPK and cAMP). Rhythms in these processes contribute to day–night differences in several hippocampal-dependent tasks including acquisition, recall, and extinction. For example, spatial working memory and novel location memory are augmented during the awake phase (night for nocturnal rodents), and these day–night differences persist when mice are housed under DD (and, thus, are under circadian clock control). The molecular clock in the forebrain appears to be necessary for these rhythms in memory since knocking out the positive limb of the molecular clock (see next section) results in poor memory at both times of day. The local clock in the hippocampus is entrained by the light–dark cycle (via the SCN) as well by other external factors. For example, food availability is critical for resetting the phase of the hippocampal clock, and synchronization of the feeding–fasting cycle with the light–dark cycle is necessary for memory and long-term potentiation. Mice that are fed during the day (rest phase) show a shifted rhythm of clock gene expression in the hippocampus as well as poor performance in novel object recognition and reduced long-term potentiation. Like the hippocampus, it is likely that the local circadian clocks in other extra-SCN oscillators also play important roles in neural function and behavior.

MOLECULAR ARCHITECTURE OF THE CIRCADIAN CLOCK

The Transcription/Translation Feedback Loop

Several clock genes take part in an interlocking transcriptional and translational feedback loop to activate and repress each other in a manner that results in an approximate 24-hour rhythm of gene expression (Figure 23–5). The key circadian genes in mammals include the activators Bmal1 and Clock (or also Npas2), which activate the E-box promoter for the Period genes Per1 and Per2 and the repressor Cryptochrome genes Cry1 and Cry2. In mammals, Clock is constitutively expressed, and Bmal1 peaks in expression during the subjective circadian night, or “lights off.” The transcription factor RORa serves to activate Bmal1 transcription and, in addition, competes with Rev-Erbα for binding to the promoter of Bmal1. CLOCK and BMAL1 dimerize in the cytoplasm after translation, are phosphorylated by casein kinase (CK)-2α, and then translocate back into the nucleus to bind to E-boxes that upregulate Per, Cry, and Rev-Erbα expression, forming the “positive arm” of the negative feedback loop.

After translation, either PER1 and CRY1 or PER2 and CRY2 dimerize and then are phosphorylated by CK-1ε/δ. This serves to either degrade the proteins or to repress their own transcription in the “negative arm” of the feedback loop by translocating back into the nucleus and inhibiting their own promoter’s activation by BMAL1/CLOCK. Concurrently, Rev-Erbα inhibits Bmal1 transcription by competing with RORa for promoter binding in a second negative feedback loop. Because the Per genes are highly transcribed during the circadian day, they are a useful measure of circadian period ex vivo (Figure 23–6).

Altogether, both positive and negative arms of the feedback loop result in rhythmic expression, with PER and CRY expression levels being highest during the late day and lowest during the middle of the night within the SCN. In particular, posttranslational modification has been shown to play a major role in determining period and phase. Phosphorylation is the main posttranslational modification being considered. In addition to CK2-α, CK1-ε, and CK1-δ, other kinases such as CK2-β and glycogen synthase kinase 3 (GSK3)-β have been shown to phosphorylate BMAL1, CRY1, CRY2, PER1, and PER2, which either allows the transcription factors to enter the nucleus or to be degraded by the proteasome.

Nearly every cell in the body (and many brain cells) rhythmically expresses these core circadian clock genes. The molecular clock drives transcription of a plethora of other clock-controlled genes (43% of all protein-coding genes in humans); however, which genes are rhythmically transcribed depends on the cell or tissue type. One important resource for exploring and comparing clock-controlled genes among rhythmic tissue is CircaDB (http://circadb.hogeneschlab.org). This website reports >3000 cycling genes from cell lines as well as peripheral and neural (brainstem, cerebellum, hypothalamus, and SCN) tissues. The relationship between the central clock in the SCN and the local molecular clocks in peripheral tissues is somewhat complex. One way to study this relationship is to disrupt the molecular clock throughout the body and then restore the clock only in the brain (restoring SCN output). Several years ago, the Hogenesch laboratory followed this approach and found that restoration of the molecular clock in the brains (only) of ClockΔ19-mutant mice rescued behavioral arrhythmicity and clock gene rhythmicity in the liver. These results suggest that SCN output is sufficient to drive the local clock in peripheral organs. However, clock-controlled genes specific to the liver were reduced in number and had lower amplitude rhythms. Thus, the local clock in peripheral tissues is important for functions specific to that cell type. This approach has yet to be applied to SCN–extra-SCN relationships. The results of such a study would reveal the importance of the SCN signal versus the local molecular clock in secondary oscillators in the brain.

SEASONALITY

After the Moon-forming impact, the new Earth was left spinning on an axis slightly off perpendicular to the planet’s orbit around the Sun. While this is not thought to be unusual, planets that lack a proportionally large moon either proceed with a floating axis or will right their axis to perpendicular (with respect to its orbit around a star) with time. The gravitational influence of the nascent Moon served to stabilize this off-kilter axis, creating reliable seasons. The amount of seasonal daily light, or photoperiod, varies from the most light on the first day of summer to the least light on the first day of winter. The range of light per day depends on latitude, with the most extreme latitudes (the poles) receiving the most extreme photoperiods. As life flourished on the planet, strategies for thriving with seasonal changes in photoperiod developed and are maintained today. The ability to predict oncoming hardship in the winter or abundance in the summer is thought to increase fitness in plants and animals alike. For plants such as trees, the shorter daytime indicates a decrease of higher energy short-wavelength light and the coming colder temperatures. The trees lose their foliage at this time of year and go relatively dormant. As daytime length increases, photoperiod signals the oncoming warmth and a transition to photosynthesis, and an increase in foliage and gamete production occurs.

In animals, change in photoperiod indicates a necessity for behavioral change. For nocturnal animals, summer photoperiod means that all activity has to be squeezed into a shorter night, and with the abundance of growing plant life, it is the prime time of year to breed, and thus, some male animals experience gonadal growth while females become receptive to mating. As winter approaches, the photoperiod indicates a time for hibernation and/or pelage change—lighter color and thicker coat to prepare for winter cold.

Not all animals experience these physiologic changes, but those that do are said to be photoperiodic. Laboratory mice (Mus musculus) are thought not to be photoperiodic, but they do experience a behavioral reorganization called seasonality following changes in photoperiod. In their seminal paper on entrainment, Pittendrigh and Daan demonstrated that rodents compress their main activity bout (α) in long photoperiods. In addition, this compression of activity length results in an aftereffect, or transient reorganization of the circadian rhythm that is evident in constant conditions. Specifically, the compression or shortening of the primary activity bout that is evident in long photoperiods correlates with a decreased FRP in mice—a faster clock.

The reorganization of the clock and how this accounts for behavioral changes had remained a mystery until recently, with the development of in vivo electrophysiologic and ex vivo circadian gene reporter techniques. In situ hybridization experiments showed that gene expression in response to a light pulse is different in animals taken from short (LD 8:16) and long (LD 16:8) photoperiods, specifically the immediate early gene c-Fos and pineal N-acetyltransferase, the enzyme used to produce the neurohormone melatonin. In fact, the tissue-level architecture of in vivo electrical activity of mouse SCN is dramatically different between photoperiods, initially suggesting that seasonality is encoded in the SCN by changes in the phase distribution of individual neuronal clocks.

It turns out that how seasonality is encoded in the SCN is a bit more complicated, but in an interesting way: How the SCN encodes season is not just dependent on the photoperiod that the organism inhabits presently, but also on what photoperiod they had been exposed to early in perinatal development. Several different labs demonstrated that for animals reared in an equinox photoperiod (LD 12:12), the phase relationship of neurons relative to each other is the major factor behind how the SCN encodes season and, therefore, how an animal subsequently behaves. In other words, if one were to record the circadian rhythm from a single neuron using something like the Period gene reporter in Figure 23–6B, one would find that the rhythms from many neurons in the SCN would line up (synchronize) in phase if they were recorded from an animal that had just experienced a short, winter photoperiod (LD 8:16) before the experiment. Contrarily, one would find that the rhythms from neurons derived from an animal that had just experienced a long, summer photoperiod (LD 16:8) would be much more asynchronous, peaking across the circadian “daytime.” Reasonably, SCNs harvested from equinox photoperiod demonstrate an intermediate phenotype. Therefore, in equinox-developed animals, season is encoded by neuronal circadian phase relationships. But this is only part of the story.

How the SCN encodes season, it turns out, depends on what photoperiod the animal experienced around birth. Although the phase distribution of neurons within the SCN is due to proximal photoperiod, the resultant tissue-level integration in response to seasonal change is profoundly different based on one’s season of development. Summer-developed mice exhibit remarkably stable molecular clock rhythms regarding the phase at which SCN neurons peak in Per1 expression, very close to dusk (ψ). Winter-developed mouse SCN, in contrast, varies phase angle widely, depending on the proximal season, with Per1 expression peaking more than an hour after dusk in animals harvested from LD 8:16 or peaking 2 hours before dusk in animals harvested from LD 16:8. Equinox-developed animals display an intermediate phenotype. The ethologic implications alone are stunning: If the peak of Per1 expression correlated with time of activity onset, imagine how this result might impact Dr. DeCoursey’s study involving ground squirrels discussed earlier! Furthermore, the tissue-level dynamics are not just a result of changing neuronal phase distributions, as mentioned in the previous paragraph, but also altered neuronal rhythms themselves, with the shape of the distribution changing drastically by broadening or narrowing between summer and winter photoperiods, but only if the animal was developed in a short, winter photoperiod. In addition, neuronal period decreases in long photoperiod, but only if the animal was developed in a long photoperiod. Taken together, it is fairly clear that seasonal encoding in the SCN occurs not only at the tissue level, but also at the neuronal level, and because the environment has such a profound and lasting effect on the animal’s physiology, seasonal encoding may be an example of an epigenetic modification.

These fascinating results on what should otherwise be a mundane topic are underscored by further work on the behavioral ramifications of seasonal birth, seasonality, and photoperiodism. A series of studies on seasonally developed hamsters and mice have demonstrated correlates of anxiety and depression dependent not only on the proximal photoperiod (ie, a tendency to be sadder in short photoperiod), but also on developmental photoperiod (ie, long photoperiod–developed mice are generally more anxious, and short photoperiod–developed mice are generally more “depressed”). With that in mind, one is forced to wonder what effect seasonality and seasonal birth have on humans. Homo sapiens are, in nature, seasonal breeders, with most births happening in the summer to early fall. In an industrialized society, we lose that natural behavior because we can provide for infants year-round, so an understanding of the seasonal effects of light is critical in explaining aberrations in human health.

Changes in seasonal photoperiod have been associated with mood disorders and mental disease in humans. Seasonal affective disorder, for example, is a mood disorder that affects between 0.4% and 2.7% of the US population every year and is treated with bright light therapy or selective serotonin reuptake inhibitors. A 5% to 8% winter-spring excess of births has been reported for both schizophrenia and mania/bipolar disorder. There is a spring and summer excess of births for autism and a winter-spring excess of births for neurosis. Winter-born humans have been reported to show both stronger morning preference than the other seasons on the Morningness-Eveningness Questionnaire and significantly lower (sadder) Global Seasonality scores than individuals born in the summer. Many of these disorders revolve around the serotonergic system, and for good reason: A recent study in mice reported that photoperiod at birth programs the intrinsic electrical and receptive properties of neurons in the dorsal raphe nuclei—the midbrain region that is responsible for serotonergic signaling to higher order areas of the brain. These studies suggest that birth season—or perhaps the seasonality of late gestation—imprints the organization of the brain systems that control cognition, emotion, and/or circadian rhythmicity.

In all, seasonality is an important aspect of circadian rhythmicity on a planet with changing light cycles. Its impact on human health is profound, but it is important to reiterate 1 particular aspect of seasonal physiology, and that is on sleep properties. In humans, summer sleep is very different from winter sleep. Summer sleep is usually consolidated into a single bout and compressed into a shorter night than other times of year. Winter sleep, on the other hand, is wholly different: It is decompressed, spread over a much longer night, and usually occurs in 2 or more bouts. This compression and decompression of sleep, which correlates with duration of melatonin release, can contribute to the aforementioned disorders of human health, especially in those who may be prone to the disorders because of seasonal birth.

HUMAN SLEEP

Sleep Architecture

An in-depth discussion of sleep is covered in Chapter 34, but a brief description is included here to describe the interrelationship between sleep and circadian systems. During waking, electroencephalogram (EEG) recordings show 2 different types of brainwave activity: alpha activity (at 8 to 12 Hz) and beta activity (at 13 to 30 Hz). Alpha activity occurs when a person is resting quietly and not engaging in any strenuous activity. Beta activity has a lower amplitude because many different neural circuits are being activated at once.

Sleep has been divided into 2 main types: rapid eye movement (REM) and non–rapid eye movement (NREM). The original scoring guidelines include 4 stages of NREM sleep and 1 stage of REM sleep. The new system developed by the American Academy of Sleep Medicine identifies 3 stages of NREM sleep (NREM stages 1, 2, and 3) and 1 REM sleep stage. Once an individual falls asleep, he or she will progress through stages 1 to 3 of NREM sleep (with stage 3 being the deepest level of sleep) and then ascend back up through the stages until the REM sleep stage is reached. One cycles lasts approximately 70 to 90 minutes. A person experiences approximately 4 or 5 of these cycles per night (Figure 23–7).

Stage 1 represents the transition from wakefulness to sleep and consists of theta activity (3.5 to 7.5 Hz), indicating neuronal firing is becoming more synchronized. At this point, a person may experience hypnic jerks, which are sleep twitches caused by muscle contractions followed immediately by relaxation. This may also be experienced as a falling sensation. Upon entering stage 2 sleep, EEG activity is characterized by sleep spindles and K complexes (Chapter 34). A sleep spindle is a burst of waves (12 to 14 Hz) that lasts about half a second. K complexes are sharp waves that are associated with inhibition of neuronal firing. Stage 3 EEG activity is then characterized by synchronized, high-amplitude, delta waves (<3.5 Hz). Heart rate, breathing rate, and brain activity also decrease at this stage, which is also known as slow-wave sleep.

After stage 3, EEG activity during REM sleep is characterized by irregular, low-voltage, fast waves that mimic those during stage 1 sleep. Using this criterion, REM sleep is very light. Eyes also begin to move rapidly back and forth while eyelids remain closed. At this point, postural muscles become very relaxed. This is referred to as sleep paralysis. Using this criterion, REM sleep is deep. This is why REM sleep is sometimes known as paradoxical sleep. Heart rate, blood pressure, and breathing also tend to vary more in REM sleep than in stages 2 to 3. People who are woken up out of REM sleep also tend to report that they had been dreaming, whereas someone woken up out of slow-wave sleep would deny that they had been dreaming.

Two-Process Model of Sleep

Sleep is regulated by 2 different processes: the circadian clock (process C) and the sleep homeostatic process (process S) (Figure 23–8). Homeostatic sleep builds up exponentially during time awake and declines exponentially during sleep. At the same time, the circadian clock drives a rhythm in attention and arousal that is independent of the duration of prior wakefulness. The distance between the 2 processes adds up to account for total sleep pressure. During a 24-hour period of sleep deprivation, many people feel a “second wind” on the following morning as process C kicks back in, overcoming the continual increase in process S.

The 2-process model of sleep can also help explain dips in performance often experienced as an afternoon slump after lunch or as an early morning fog. Process C drives a 24-hour rhythm in performance that peaks in the evening for most people. Process S results in peak performance in the morning after recovering from the initial wake-up period (approximately 1 to 4 hours), referred to as sleep inertia. After that, performance driven by process S slowly declines throughout the day, but overall performance results from a combination of both process C and process S. Thus, performance is fairly high during the early part of the day due to process S and fairly high during the evening hours due to process C. That leaves the afternoon as a time when process S is declining but process C has yet to reach a full peak, causing a reduction in performance at this time of day. Of course, during the overnight hours, performance is low due to reductions in both processes. As a result, jobs that require workers to be awake and alert in the middle of the night often struggle with balancing productivity, performance, and worker safety.

Functions of Sleep

For millennia, people and scientists have wondered about the purpose of sleep—even Aristotle recognized its importance for preservation of life because sleep allows the body’s organs to rest and recover. A related hypothesis states that the purpose of sleep is to conserve energy. During sleep, energy is conserved due to decreased body temperature. As a result, animals increase sleep duration when food is in short supply, thus reducing energy expenditure. Sleep improves memory. In humans, task learning improves following a night of sleep. REM sleep specifically aids in consolidation of nondeclarative (procedural) memory, whereas NREM sleep aids in consolidation of declarative memories. Similarly, natural immune response is modified by partial sleep deprivation. After just 1 night of sleep deprivation, the activity of natural killer cells, a type of white blood cell, is reduced by 28%. Sleep loss can also result in a reduction of circulating immune complexes, secondary antibody responses, and antigen uptake. Finally, recent research has highlighted the importance of sleep for proper function of the glymphatic system, in which glial cells regulate the flow of cerebrospinal fluid by shrinking and swelling. The results showed that the flow of fluid slows and the space between cells decreases during waking and increases during sleep by as much as 60%! This mechanism of sleep is very important for removing toxins from the brain, including β-amyloid, a protein implicated in Alzheimer disease.

CLINICAL CORRELATES

As described earlier, circadian rhythms impact a variety of physiologic functions. Therefore, it is no surprise that circadian disruption can impair memory, reduce mental and physical reaction times, and exacerbate or increase the risk of human disease such as depression, insomnia, diabetes, obesity, immune dysfunction and cancer. In the laboratory, mice are often used to model the impact of circadian disruption on human disease since they exhibit many of the same circadian characteristics as humans, including 24-hour rhythms in temperature, melatonin, cortisol, attention, and activity. Like mice, the SCN serves as the central circadian pacemaker and is entrained by photic and nonphotic input. One notable difference is temporal niche, such that humans are diurnal and active in the daytime, whereas mice are nocturnal and active at night. Despite this major behavioral difference, clock genes are also expressed in a circadian rhythm in human tissues including white blood cells and buccal cells. As with all genes, circadian genes are prone to heritable mutations over time. These conserved variants have been identified in many of the circadian genes, with significant phenotypic consequences that have driven the field of circadian geneticists to create mouse models of the genetic change and further our understanding of circadian protein roles, function, and dysfunction.

Much effort has been made to identify genetic causes for circadian sleep disorders, which have been defined by the International Classification of Sleep Disorders (ICSD) as “a persistent or recurrent pattern of sleep disturbance due primarily to alterations in the circadian timekeeping system or a misalignment between the endogenous circadian rhythm and exogenous factors that affect the timing or duration of sleep.” Six disorders have been recognized as true circadian rhythm sleep disorder by the ICSD; these include advanced sleep phase syndrome (or familial advanced sleep phase syndrome [FASPS]), delayed sleep phase syndrome (DSPS), non–24-hour sleep–wake disorder (N-24; also known as free-running disorder), irregular sleep–wake rhythm/phase (ISWR), jet lag, and shift work.

To date, FASPS has been the only circadian disorder found to display true Mendelian inheritance. This disorder is characterized by an approximate 4-hour advanced phase angle of sleep onset every day. These people wake up in the very early morning (1 to 4 am) and fall asleep very early as well (6 to 7 pm), even in the presence of social pressure to stay awake. Although this disorder is likely to have many genetic sources, 2 such sources have been identified as either of 2 complementary mutations: a S662G mutation in the hPer2 phosphorylation site or a T44A mutation in the phosphate transfer domain of CK1-δ.

The etiology of DSPS has not been as straightforward. DSPS is the most common of the CRSDs and is characterized by extremely late bedtimes (2 to 6 am) and midday wake times (10 am to 2 pm). Although there is 1 reported case of inherited DSPS where the genetic defect is unknown, a length polymorphism located in exon 18 of the Per3 gene has shown the most promising association with the phenotype. A single nucleotide polymorphism in the Clock gene (3111 T>C) has also been a promising genetic lead, with people carrying the C/C genotype exhibiting strong evening preference, although not DSPS per se. Like FASPS, however, there are likely many genetic causes for the phenotype that have yet to be identified. Numerous clinical trials have attempted to treat delayed circadian phase with light therapy. Even delayed sleep phase associated with mental illness such as attention-deficit/hyperactivity disorder (ADHD) was successfully treated in 1 pilot study using early morning light therapy. Moreover, the reduction in ADHD symptoms was correlated with the size of the phase advance induced by light therapy.

N-24 is common in blind people and is a result of a lack of perception of photic entraining stimuli, which then leads to the person free-running through the geophysical days. N-24, therefore, does not necessarily have a primary genetic component. ISWR is characterized by the absence of circadian sleep–wake, and its cause is unknown. Jet lag is the unpleasant physiologic and psychological aftereffects of traveling across multiple time zones, and its severity is proportional to the number of time zones crossed, direction of travel (west to east is considered more severe for humans), sleep deprivation, presence of zeitgebers at the destination, and individual tolerance. While the cause of jetlag is known to be the misalignment of circadian phase with the geophysical phase at the destination (or disruption/misalignment of multiple internal endogenous oscillators), much work has been done to alleviate the symptoms and elucidate their root causes. To date, no major genetic studies have been undertaken in humans regarding N-24, ISWR, or jet lag.

The final recognized circadian rhythm sleep disorder, shift work disorder, is caused by circumstances similar to jet lag, in that misalignment of circadian phase with the zeitgebers (light, social pressure, stimulants) during the work shift causes unpleasant aftereffects or sleepiness, especially in people on a rotating schedule. Studies in individuals involved in shift work (estimated at 1 in 5 American workers) have discovered how genetic polymorphisms in circadian genes associate with shift work tolerance, sleep strategy, and other CRSDs. For example, multilocus models of polymorphisms in several clock genes are associated with alcohol or caffeine consumption and sleepiness of hospital shift work nurses. One contributing factor to sleepiness and adaptation to shift work is the decision of when to sleep on days off. For example, many shift workers switch back to sleeping at night on days off. Different sleep strategies are used by hospital shift work nurses, who frequently work several 12-hour night shifts in a row followed by several days off. The most maladaptive strategies are ones in which nurses take long, daily naps on or stay awake for >24 hours on at least 1 day per week. Nurses using these strategies reported greater sleep disturbance, an earlier chronotype, and more cardiovascular problems.

Characteristics of human mental health display an intriguing association with circadian rhythms. The symptoms of jet lag and shift work briefly described earlier offer evidence that disruption of circadian phase angle can lead not only to side effects, but also the possibility of more serious conditions. Patients with mood disorders often exhibit altered circadian phase in body temperature, cortisol, and even melatonin. It has been suggested that the connection between mood disorders and circadian rhythms may involve the literal connection between the serotonergic raphe nuclei of the brainstem and the SCN, both directly and indirectly via the IGL of the thalamus. Numerous 5-HT receptor subtypes are expressed in the SCN, and the interplay between the circadian system and the serotonergic system is thought to be extensive at both the molecular and tissue levels. In combination with the extant data on the role of clock genes in normal and abnormal human circadian rhythmicity, these data suggest not only that circadian genes could play a potentially huge role in mental health, but also that human circadian neural organization is an appropriate starting point for research into novel treatments for mental disorders. For example, some medications may have improved efficacy or reduced side effects when taken at a particular time of day. This approach is often referred to as chronotherapy.

CONCLUSION

Circadian rhythms are 24-hour oscillations in behavior and physiology that arose out of necessity to survive in a world with regularly occurring light and dark periods. In mammals, these rhythms are controlled by a primary clock located in the SCN of the hypothalamus that orchestrates secondary clocks throughout the rest of the brain and body. The SCN is necessary and sufficient for driving 24-hour rhythms in sleep/wake, locomotor activity, feeding/fasting, hormone release, heart rate/blood pressure, and so on. SCN neurons and nearly every cell in the body can generate 24-hour rhythms in transcription due to a molecular clock composed of interlocking feedback loops. However, SCN neurons have an additional feature of being able to couple into a network that allows animals to shift their clock phase to match the environmental cycle of light and dark, food availability, and even the presence of predators. This network also allows adaptation to changes in the length of the light period as the seasons change across an annual cycle. The core features of an internal clock—persistence in constant conditions, entrainment to the environment, and temperature compensation—are evident in nearly all life on Earth, including humans. The circadian clock is important for human health, and disruption of circadian rhythms (as in jet lag and shift work) can exacerbate and/or increase the risk of disease. Translation of the findings from animal models and human laboratory studies may lead to novel chronotherapeutic treatments for disease.

SUMMARY

• Circadian rhythms are 24-hour oscillations in behavior and physiology that arose out of necessity to survive in a world with regularly occurring light and dark periods.

• Mammalian circadian rhythms are controlled by a primary clock located in SCN of the hypothalamus that orchestrates secondary clocks throughout the rest of the brain and body.

• Nearly every cell in the body can generate 24-hour rhythms in transcription due to a molecular clock composed of interlocking feedback loops.

• The SCN is necessary and sufficient for driving 24-hour rhythms in, for example, sleep/wake, locomotor activity, feeding/fasting, hormone release, heart rate, and blood pressure, and SCN neurons form a coupled network that allows animals to shift their clock phase to match the environmental cycle of light and dark, food availability, and even the presence of predators.

• The circadian network allows adaptation to changes in the length of the light period as the seasons change across an annual cycle.

• The core features of an internal clock—persistence in constant conditions, entrainment to the environment, and temperature compensation—are evident in nearly all life on Earth, including humans.

• The circadian clock is important for human health, and disruption of circadian rhythms (as in jet lag and shift work) can exacerbate and/or increase the risk of disease.

• Translation of the findings from animal models and human laboratory studies may lead to novel chronotherapeutic treatments for disease.