ch 13

CHAPTER 13 Memory and Learning Neil V. Watson Simon Fraser University S. Marc Breedlove Michigan State University Trapped in the Eternal Now Every day is alone in itself, whatever enjoyment I’ve had, and whatever sorrow I’ve had…. Right now, I’m wondering, have I done or said anything amiss? You see, at this moment everything looks clear to me, but what happened just before? That’s what worries me. It’s like waking from a dream. I just don’t remember. —Henry Molaison (B. Milner, 1970, p. 37) Known as “Patient H.M.” in a classic series of research articles, Henry Molaison was probably the most famous research participant in the history of neuroscience. After a bicycle accident when Henry was an adolescent, he started to experience seizures, and by his late twenties they were out of control. Like the people with epilepsy we described at the end of Chapter 2, Henry decided to take the extreme measure of having a surgeon remove the brain sites where the seizures began. Because Henry’s seizures began in both temporal lobes, a neurosurgeon removed most of his medial temporal lobes in 1953. Henry’s surgery relieved his epilepsy, but at a terrible, unforeseen price: He couldn’t seem to form new memories (Scoville and Milner, 1957). For more than 50 years after the surgery, until his death in 2008, Henry could retain any new fact only briefly; as soon as he was distracted, the newly acquired information vanished. He didn’t know his age or the current date. For a while, he carried a note reminding himself that his father had died and his mother was in a retirement home. Henry knew something was wrong with him, because he had no memories from the years since his surgery, or even memories from earlier the same day, as the quote above indicates. Henry’s inability to form new memories meant that he couldn’t have a lasting relationship with anybody new. No matter what experiences he might share with someone he met, Henry would have to start the acquaintance anew the following day, because he would have no recollection of ever having met the person. In some ways, this dreadful loss of memory ended Henry’s journey as a human being—he could no longer grow in his experience of historical events, his friendships, or even a sense of his own life story. What was it about his surgery that created Henry’s predicament, and what does his experience teach us about learning and memory? All the distinctively human aspects of our behavior are learned: the languages we speak, how we dress, the foods we eat and how we eat them, our skills, and the ways we reach our goals. So much of our own individuality depends on learning and memory. We begin this chapter with a discussion of memory and the twentieth century research revealing that there are fundamentally different types of memory. Then we delve into what we know about how learning alters the structure of the brain in order to store memories, perhaps for a lifetime. 13.1 There Are Several Kinds of Learning and Memory The Road Ahead We begin our discussion about learning and memory by examining how they fail. Studying this material should allow you to: 13.1.1 Understand the two kinds of amnesia, one for memories before an event, and one for memories after an event. 13.1.2 Describe the two fundamentally different categories of memory. 13.1.3 Review the evidence that a particular circuit of three brain regions is crucial for forming certain types of memory. 13.1.4 Understand the two subtypes of declarative memories, the memories we can describe to other people. The terms learning, the process of acquiring new information, and memory, the ability to store and retrieve that information, are so often paired that it sometimes seems as if one necessarily implies the other. We cannot be sure that learning has occurred unless a memory can be elicited later. Many kinds of brain damage, caused by disease or accident, impair both learning and memory. We’ll start by looking at some brain damage cases that revealed different classes of learning and memory. For Patient H.M., the present vanished into oblivion Amnesia (Greek for “forgetfulness”) is a severe impairment of memory, usually as a result of accident or disease. Loss of memories that formed prior to an event (such as surgery or trauma)—called retrograde amnesia (from the Latin retro, “backward,” and gradi, “to go”)—is not uncommon. After an accident that damages the brain, people often have retrograde amnesia regarding events that happened a few hours or days before the accident, or even a year before. Despite dramatic depictions you may see onscreen, it is unlikely that longer-term (or “complete”) retrograde memory loss has ever occurred. Patient H.M.—Henry Molaison, whom we met at the start of the chapter—experienced a far more unusual symptom. In Henry’s case, most old memories remained intact, but he had difficulty recollecting any events that took place after his surgery. What’s more, he was unable to retain any new material for more than a brief period. The inability to form new memories after an event is called anterograde amnesia (the Latin antero means “forward”). Over the very short term, Henry’s memory was normal. If given a series of six or seven digits, he could immediately repeat the list without error. But when he was given a list of words to study and then tested on them after being distracted by another task, he could not repeat the list or even recall that there was a list. So Henry’s case provided clear evidence that short-term memory differs from longterm memory—a distinction, long recognized by psychologists on behavioral grounds (W. James, 1890), that we will discuss in more depth later in this chapter. He was unable to transfer short-term memory into long-term memory. Brenda Milner Earning her PhD in 1952 a few years before working with Henry Molaison, Dr. Milner is still an active scholar, even after her 105th birthday. View larger image Henry’s surgery removed the amygdala, most of the hippocampus, and surrounding cortex from both temporal lobes (FIGURE 13.1). The memory deficit seemed to be caused by loss of the medial temporal lobe, including the hippocampus, because people who had only the lateral temporal cortex removed had no memory impairment. Fascinatingly, despite his obvious memory problems, Henry improved greatly over days of practice on a mirror-tracing task (FIGURE 13.2A) (B. Milner, 1965). Each day, when asked if he remembered the test, Henry said no, yet his performance was better than at the start of the first day (FIGURE 13.2B). On the final day, he expressed surprise that the task was easier than he’d expected. So, was Henry’s memory loss limited to tasks that relied on verbal processing? Not quite. For example, people with amnesia like Henry’s can learn the skill of reading mirror-reversed text (FIGURE 13.3), which is a verbal task. FIGU R E 1 3 . 1 Brain Regions Crucial for Forming New Memories View larger image FIGU R E 1 3 . 2 Henry’s Performance on a Mirror-Tracing Task View larger image FIGU R E 1 3 . 3 Reading Mirror-Reversed Text View larger image The important distinction in Henry’s deficit is not between motor and verbal performances, but rather between two general categories of memory: 1. Declarative memory, or explicit memory, is what we usually think of as memory: facts and information acquired through learning. It is memory we are aware of accessing, which we can declare to others. This is the type of memory that was so profoundly impaired by Henry’s surgery. Tests of declarative memory take the form of requests for specific information that was learned previously. It is the type of memory we use to answer “what” questions—and thus is difficult to test in animals. 2. Nondeclarative memory, or implicit or procedural memory—that is, memory about perceptual or motor procedures—is shown by performance rather than by conscious recollection. Examples of procedural memory include learning the mirror-tracing task, at which Henry excelled, and the skill of mirror reading, or riding a bike (FIGURE 13.4). It is the type of memory we use for “how” problems and is often (but not always) nonverbal. FIGU R E 1 3 . 4 Two Main Kinds of Memory: Declarative and Nondeclarative View larger image A clever way to measure declarative memory in monkeys and other animals is the delayed non-matching-to-sample task (FIGURE 13.5), a test of object recognition that requires monkeys to declare what they remember by identifying which of two objects was not seen previously (Winters et al., 2010). Monkeys with damage to the medial temporal lobe, similar to H.M., are severely impaired on this task, as we’ll see next. FIGU R E 1 3 . 5 The Delayed Non-Matching-to-Sample Task View larger image RESEARCHERS AT WORK Which Brain Structures Are Important for Declarative Memory? To determine which brain regions are crucial for declarative memory, researchers selectively removed specific parts of the medial temporal lobes of monkeys to confirm that the amygdala —one of the structures removed in Henry’s surgery—is not crucial for performance on tests of declarative memory. However, removal of the adjacent hippocampus significantly impaired performance on these tests and, as shown in FIGURE 13.6, the deficit was even more pronounced when the hippocampal damage was paired with lesions of nearby cortical regions that communicate with the hippocampus: entorhinal, parahippocampal, and perirhinal cortices (Zola-Morgan et al., 1994). Humans similarly show larger impairments when both the hippocampus and surrounding cortex are damaged (Squire and Wixted, 2011). So Henry’s symptoms were probably caused by loss of the medial temporal lobe on both sides of the brain. FIGU R E 1 3 . 6 Memory Performance after Medial Temporal Lobe Lesions View larger image Damage to the medial diencephalon can also cause amnesia In 1960, a young man now known as Patient N.A. had a bizarre accident in which a miniature sword entered his nostril and injured his brain. Like Henry, N.A. showed profound anterograde amnesia after his accident (Squire and Moore, 1979), and he can give little information about events since then, although his memory for earlier events is near normal (Kaushall et al., 1981). Magnetic resonance imaging (MRI) of N.A. (FIGURE 13.7) shows damage to several limbic system structures in the medial diencephalon that have connections to the hippocampus: the dorsomedial thalamus and the mammillary bodies (so called because they are shaped like a pair of breasts—see Figure 13.1A). Like Henry Molaison, N.A. shows normal short-term memory and can gain new nondeclarative/procedural memories, but he is impaired in forming declarative long-term memories. The similarity in symptoms suggests that the medial temporal lobe damaged in Henry’s brain and these midline regions damaged in N.A. are parts of a larger memory system. FIGU R E 1 3 . 7 The Brain Damage in Patient N.A. View larger image That idea is reinforced by studies of people with Korsakoff’s syndrome (or Korsakoff’s amnesia), a degenerative disease in which damage is found in the mammillary bodies (FIGURE 13.8) and dorsomedial thalamus, but not in temporal lobe structures like the hippocampus. The mammillary bodies may serve as a processing system connecting the medial temporal lobes (which were removed from Henry Molaison) to the thalamus and, from there, to other cortical sites (Vann and Aggleton, 2004). People with Korsakoff’s syndrome often fail to recognize their memory problems, and may deny that anything is wrong with them. They often confabulate— that is, fill a gap in memory with a falsification that they seem to accept as true. Damage to the frontal cortex, also found in people with Korsakoff’s syndrome, probably causes the denial and confabulation that differentiates them from other people who have amnesia, such as Henry. FIGU R E 1 3 . 8 Brain Damage in People with Korsakoff’s Syndrome View larger image The main cause of Korsakoff’s syndrome is lack of the vitamin thiamine (Arts et al., 2017). Alcoholics who obtain most of their calories from alcohol and neglect their diet often have this deficiency. Treating them with thiamine can prevent further deterioration of memory functions but will not reverse the damage already done. These studies make it clear that a brain circuit that includes the hippocampus, the mammillary bodies, and the dorsomedial thalamus is needed to form new declarative memories. But these case studies also clearly show that established declarative memories, formed before brain damage, are not stored in these structures for the long term. If they were, they would have been lost when the structures were damaged. So where are memories stored? A leading candidate is the cerebral cortex, as we’ll see in Signs & Symptoms, next. How’s It Going? 1. What are the two main types of amnesia, and which deficit was more severe for Henry Molaison? 2. What are the two main types of memory, and which type was affected in Henry? 3. How did research with animals help pin down the brain regions required for forming new memories that we can declare to others? 4. Name the two brain regions, in addition to the medial temporal lobe, that are required to form new declarative memories. 5. What are the two main subtypes of declarative memory? FOOD FOR THOUGHT If forced to choose between losing all your past autobiographical memory, or being like Henry—unable to form new declarative memories of any sort—which would you choose? Why? SIGNS & SYMPTOMS Brain Damage Can Destroy Autobiographical Memories While Sparing General Memories One striking case study suggests that at least some declarative memories are stored in the cortex, and it also illustrates an important distinction between two subtypes of declarative memory. At age 30 Kent Cochrane, known to the world as Patient K.C., had a motorcycle accident that resulted in serious brain damage. He could no longer retrieve any personal memory of his past, although his general knowledge remained good. He conversed easily and played a good game of chess but could not remember where he learned to play chess or who taught him the game. Detailed autobiographical declarative memory of this sort is known as episodic memory; you show episodic memory when you recall a specific episode in your life or relate an event to a particular time and place. In contrast, semantic memory is generalized declarative memory, such as knowing the meaning of a word without knowing where or when you learned that word. If care was taken to space out the trials, Kent could acquire new semantic knowledge (Tulving et al., 1991). But even with this method, Kent could not acquire new episodic knowledge—he wouldn’t remember where he had learned that new material. Patient K.C. Brain damage from a severe motorcycle accident left Kent Cochrane (1951–2014) unable to retrieve episodic memories. View larger image Scans of Kent’s brain revealed extensive damage to the left frontoparietal and the right parieto-occipital cerebral cortex, as well as severe shrinkage of both right and left hippocampus and nearby cortex (Rosenbaum et al., 2005). As with Henry, the bilateral hippocampal damage probably accounts for Kent’s anterograde declarative amnesia. But that damage cannot account for Kent’s selective loss of nearly all his autobiographical memory, because other people with damage restricted to the medial temporal lobe, like H.M., retain autobiographical memories. Kent’s inability to recall any events in his life from many years before his accident may instead be a consequence of injuries to frontal and parietal cortex (Tulving, 1989). (Unlike dramatic portrayals of retrograde amnesia in fiction, Kent knew his name and recognized his family, although he couldn’t remember any particular past events with those people.) FIGURE 13.9 reviews the current view of the sequence of brain regions important for forming declarative memories. FIGU R E 1 3 . 9 Current Model of Declarative Memory Formation View larger image 13.2 Different Forms of Nondeclarative Memory Involve Different Brain Regions The Road Ahead Now let’s consider the different types of nondeclarative memory and the brain regions associated with each. Studying this material should allow you to: 13.2.1 List the different categories of nondeclarative memory. 13.2.2 Name the stages of memory formation and the vulnerability for losing information at each stage. 13.2.3 Describe a model of how we encode, consolidate, and retrieve memories. 13.2.4 Explain how retrieving a memory makes it vulnerable to distortion. 13.2.5 Describe a model of how long-term memories are stored in cortex. So far, we’ve seen that there are two different kinds of declarative memory: semantic and episodic. Likewise, there are several different types of nondeclarative memory, and we’ll see that different brain regions are involved in these different forms. Different types of nondeclarative memory serve varying functions Skill learning is the process of learning how to perform a challenging task through deliberate practice, over and over. Improving at the mirror-tracing task performed by Henry Molaison (see Figure 13.2) or learning to read mirror-reversed text (see Figure 13.3) are examples of skill learning. So too is the acquisition of everyday skills like learning to ride a bike or to juggle (well, okay, maybe juggling isn’t an “everyday” skill, but you get the idea). Henry demonstrated that the medial temporal lobe is not required to gain skills and retain them. Imaging studies have investigated learning and memory for different kinds of skills, including sensorimotor skills (e.g., mirror tracing), perceptual skills (e.g., reading mirror-reversed text), and cognitive skills (tasks involving planning and problem solving, common in puzzles like the Tower of Hanoi problem). All three kinds of skill learning are impaired in people with damage to the basal ganglia (see Figure 1.14A). Damage to other brain regions, especially the motor cortex and cerebellum, also affects aspects of some skills. Neuroimaging studies confirm that the basal ganglia, cerebellum, and motor cortex are important for sensorimotor skill learning (Makino et al., 2016; Spampinato and Celnik, 2018). Priming (or repetition priming) is a change in the way you process a stimulus, usually a word or a picture, because you’ve perceived it, or something similar, previously. For example, if a person is shown the word stamp in a list and later is asked to complete the word stem STA-, then they are more likely to reply “stamp” than, say, “start.” Priming does not require declarative memory of the stimulus—Henry Molaison and other people with amnesia have shown priming for words they don’t remember having seen. In contrast with skill learning, priming is not impaired by damage to the basal ganglia. In functional-imaging studies, perceptual priming (priming based on the visual form of words) is related to reduced activity in bilateral occipitotemporal cortex (Korzeniewska et al., 2020), while conceptual priming (priming based on meaning) is associated with reduced activation of the left frontal cortex (Matsumoto et al., 2021). So priming appears to be at least partly a function of the cortex. Other types of nondeclarative memories include learning that involves relations between events—for example, between two or more stimuli, between a stimulus and a response, or between a response and its consequence—and is called associative learning. In the best-studied form, classical conditioning, an initially neutral stimulus comes to predict an event. In famous experiments, Ivan Pavlov (1849–1936) found that a dog would learn to salivate when presented with an auditory or visual stimulus if the stimulus came to predict the presentation of food. So, repeatedly ringing a bell before putting meat powder in a dog’s mouth will eventually cause the dog to start salivating when it hears the bell alone. In this case the meat powder in the mouth is called the unconditioned stimulus (US), which already evokes an unconditioned response (UR; salivation in this example). The sound of the bell is called the conditioned stimulus (CS), and the learned response to the CS alone (salivation in response to the bell) is called the conditioned response (CR) (FIGURE 13.10). By the way, several sources on the web smugly declare that Pavlov never actually used a bell for a CS, but there’s plenty of evidence that he did (Tully, 2003). Some web “myths” are themselves myths! FIGU R E 1 3 . 1 0 Pavlovian (Classical) Conditioning View larger image Experimental evidence in lab animals shows that circuits in the cerebellum are crucial for simple eye-blink conditioning, in which a tone or other stimulus is associated with eye blinking in response to a puff of air. A PET study in humans confirmed this idea by showing a progressive increase in activity in the cerebellum during eye-blink conditioning (Logan and Grafton, 1995). People with hippocampal lesions can acquire the conditioned eye-blink response, but people with damage to the cerebellum on one side can acquire a conditioned eye-blink response only on the side where the cerebellum is intact (Papka et al., 1994). In instrumental conditioning (also called operant conditioning), an association is formed between the animal’s behavior and the consequence(s) of that behavior. An example of an apparatus designed to study instrumental conditioning is called the Skinner box, named for its originator, B. F. Skinner (FIGURE 13.11). In a common setup, the animal learns that performing a certain action (e.g., pressing a bar) is followed by a reward (such as a food pellet). Research in animals has not pinpointed the brain regions that are crucial for instrumental conditioning, perhaps because this type of learning taps so many different aspects of behavior. FIGU R E 1 3 . 11 A Skinner Box View larger image Animal research confirms the various brain regions involved in different attributes of memory The caricature of the white-coated scientist watching rats run in mazes, a staple of cartoonists to this day, has its origins in the intensive memory research of the early twentieth century. The early work indicated that rats and other animals don’t just learn a series of turns but instead form a cognitive map (an understanding of the relative spatial organization of objects and information) in order to solve a maze (Tolman, 1949). Animals apparently learn at least some of these details of their spatial environment simply by moving through it (Tolman and Honzik, 1930). We now know that, in parallel with its role in other types of declarative memory, the hippocampus is crucial for spatial learning. The rat hippocampus contains many neurons that selectively encode spatial location (Morris and Derdikman, 2023). These place cells become active when the animal is in—or moving toward—a particular location (FIGURE 13.12A). If the animal is moved to a new environment, place cell activity indicates that the hippocampus remaps to the new locations (Kubie et al., 2020). Some rat neurons in the entorhinal cortex next to the hippocampus act like “grid cells,” likened to a latitude and longitude in a maze (Moser et al., 2017) (FIGURE 13.12B), which have been recorded in people too (Rolls, 2023). FIGU R E 1 3 . 1 2 Hippocampal Spatial Neurons View larger image Bird species that hide food in many locations have larger hippocampi than other birds have (Croston et al., 2015), indicating that natural selection favors enlargement of the hippocampus to enhance spatial learning. Brain regions involved in learning and memory: A summary FIGURE 13.13 updates and summarizes the classification of longterm memory that we’ve been discussing. Several major conclusions should be apparent by now, especially (1) that many regions of the brain are involved in learning and memory; (2) that different forms of memory rely on at least partly different brain mechanisms, which may include several different regions of the brain; and (3) that the same brain structure can be a part of the circuitry for several different forms of learning. Next we’ll discuss the stages by which memories, of any sort, can be preserved for a lifetime. FIGU R E 1 3 . 1 3 Subtypes of Declarative and Nondeclarative Memory View larger image How’s It Going? 1. Name three different types of nondeclarative memory, giving an example of each. What different parts of the brain have been implicated in each type? 2. What is a cognitive map? 3. What are hippocampal place cells, and why do they suggest a role for the hippocampus in spatial learning? Successive processes capture, store, and retrieve information in the brain The span of time over which a piece of information is retained in the brain varies. There are at least three different stages of memory. The briefest memories are called sensory buffers (for visual stimuli, they are sometimes called iconic memories); an example is the fleeting impression of a glimpsed scene that vanishes from memory seconds later. These brief memories are thought to be residual activity in sensory neurons. Somewhat longer than sensory buffers are short-term memories (STMs). If someone tells you a website name and you keep it in mind (perhaps through rehearsal) just until you type it into your browser, you are using STM. In the absence of rehearsal, STMs last only about 30 seconds. With rehearsal, you may be able to retain an STM until you turn to a new task a few minutes later, but when the STM is gone, it’s gone for good. Eventually, some memories become really long-lasting—the address of your childhood home, how to ride a bike, your first crush—and are called long-term memories (LTMs). A related concept is working memory, which refers to the ability to actively manipulate information in your STM, perhaps retrieving information from LTM, to solve a problem or otherwise make use of the information (Belletier et al., 2023). The concept of working memory has evolved (Cowan, 2022), but we will consider it to be a subset of STM where information is not just stored, but can be analyzed and manipulated by some “executive” part of our mind. As shown in FIGURE 13.14, the memory system consists of at least three processes: (1) encoding of raw information from sensory channels into STM, (2) consolidation of the volatile STM into more-durable LTM, and (3) eventual retrieval of the stored information from LTM for use in working memory. A problem at any stage can cause us to lose information. Although not depicted in the figure, this model suggests that the flow of information into and out of working memory is supervised by another part of the mind, an executive function, which we will discuss in more detail in Chapter 14. FIGU R E 1 3 . 1 4 The Stages of Memory View larger image Not all memories are created equal. We all know from firsthand experience that emotion can powerfully affect our memory for past events. For example, an emotionally arousing story is remembered significantly better than a closely matched but emotionally neutral story (Kim et al., 2021). But if people are treated with propranolol (a beta-adrenergic antagonist, or beta-blocker, that blocks the effects of epinephrine), this emotional enhancement of memory vanishes. It’s not that treated volunteers perceive the story as being any less emotional; in fact, they rate the emotional content of the stories just the same as untreated people do. Instead, the drug seems to directly interfere with the ability of adrenal stress hormones to act on the brain to enhance memory (Bolsoni and Zuardi, 2019). Long-term memory has vast capacity but is subject to distortion Henry Molaison’s case and the research it inspired have already told us several ways in which STM and LTM differ from one another. While the medial temporal lobe is not needed to encode sensory information into STM, or to retrieve that information from STM (Henry could repeat back to you a list of words or numbers), it is crucial for encoding that information in LTM. In terms of the model (see Figure 13.14), an intact hippocampus is required to consolidate declarative STMs into LTMs, indicating that the information is somehow transformed into a different format, one that may make it available for a lifetime. How much information can be stored in LTM? There must be a limit, but no one has been able to come up with a way to measure it. In one classic experiment, people viewed long sequences of color photos of various scenes; several days later, they were shown pairs of images— in each case a new image plus one from the previous session—and asked to identify the images seen previously. Astonishingly, participants performed with a high degree of accuracy for series of up to 10,000 different stimuli (Standing, 1973)! For all practical purposes, there seems to be no upper bound to LTM capacity (Brady et al., 2014). Pigeons have a similarly impressive visual memory (Vaughan and Greene, 1984). We take this capacity for granted and barely notice, for example, that knowledge of a language involves remembering at least 100,000 pieces of information. Most of us also store a huge assortment of information about faces, tunes, odors, skills, stories, and so on. The late Kim Peek (1951–2009) was a savant (from the French for “knowing”), a person with an unusually well-developed ability or skill. Born with several brain structural abnormalities, including an absence of the corpus callosum (Treffert and Christensen, 2005), Kim eventually memorized about 9000 books, each taking about an hour. He read the 656-page novel The Hunt for Red October in 75 minutes, and when asked, 4 months later, to name a minor character, not only did Kim know the name, but he cited the page number where the character appeared and quoted several passages on the page verbatim! Case studies of such individuals with exceptional memory indicate that without the usual process of pruning out unimportant memories, continual perfect recall can become uncontrollable, distracting, and exhausting (Luria, 1987; Parker et al., 2006). Despite the vast capacity of LTM, forgetting is a normal aspect of memory, helping to filter out unimportant information and freeing up needed cognitive resources (Kuhl et al., 2007). Interestingly, research indicates that the memory trace (the record laid down in memory by a learning experience, also known as an engram) doesn’t simply deteriorate from disuse and the passage of time; instead, memories tend to suffer interference from events before or after their formation. For example, the process of retrieving information from LTM causes the memories to become temporarily unstable and susceptible to disruption or alteration before undergoing reconsolidation and returning to stable status (Nader and Hardt, 2009). Thus we can create false memories when we use leading questions to have people retrieve memories. Asking “Did you see the broken headlight?” rather than “Was the headlight broken?” can incorporate the false detail as the memory is reconsolidated (Loftus, 2003). This possibility of planting false memories clouds the issue of “recovered memories” of childhood sexual or physical abuse, because controversial therapeutic methods such as hypnosis or guided imagery (in which the person is encouraged to imagine hypothetical abuse scenarios) can inadvertently plant false memories (Otgaar et al., 2019). A Prodigious Memory Kim Peek (1951–2009) was a savant who memorized about 9000 books. View larger image On the other hand, you can use the power of reconsolidation when studying—as long as you’re careful to check your facts. One of the best ways to improve learning is simply repeated retrieval (and thus, repeated reconsolidation) of the stored information with feedback to let you know what you got right or wrong (Karpicke and Roediger, 2008). In other words, test yourself repeatedly, as with the How’s It Going? questions in this book and/or flash cards. For your next exam, try making up some additional practice tests for yourself, or have a friend quiz you, instead of simply “cramming.” We still haven’t talked about the nitty-gritty of memory in the brain —what exactly changes in the brain when we learn? That’s our next topic. How’s It Going? 1. What are the stages of memory, and what do we call the processes by which information moves from one stage to the next? 2. What is the memory trace, and what are two explanations of why we sometimes cannot retrieve it? 3. Explain how reconsolidation makes us vulnerable to distorted memories. FOOD FOR THOUGHT A few rare people seem to remember every event in their lives, down to the weather and newspaper headlines for each day. What advantages and disadvantages might accompany such a prodigious autobiographical memory? 13.3 Memory Storage Requires Physical Changes in the Brain The Road Ahead In this part of the chapter, we will look at some of the ways in which new learning involves changes in synapses. Reading this material should enable you to: 13.3.1 List the possible ways in which changes in neural function and structure could encode memories. 13.3.2 Review evidence that exposure to an enriched environment can affect brain structure and affect future behavior. 13.3.3 Describe how a circuit involving the cerebellum mediates certain types of conditioning. In introducing the term synapse, Charles Sherrington (1897) speculated that synaptic alterations might be the basis of learning. Sherrington’s notion anticipated what remains one of the most intensive efforts in all of neuroscience, since most theories of learning focus on neuroplasticity (or neural plasticity), changes in the structure and function of synapses. Plastic changes at synapses can be physiological or structural Synaptic changes that may store information can be measured physiologically. The changes can be presynaptic, postsynaptic, or both. They can include changes in the amount of neurotransmitter released and/or changes in the number or sensitivity of the postsynaptic receptors, resulting in larger (or smaller) postsynaptic potentials. Inhibiting inactivation of the transmitter (by altering reuptake or enzymatic degradation) can produce a similar effect (FIGURE 13.15A). Synaptic activity can also be influenced by inputs from other neurons, causing extra depolarization or hyperpolarization of the axon terminals and therefore changes in the amount of neurotransmitter released (FIGURE 13.15B). FIGU R E 1 3 . 1 5 Synaptic Changes That May Store Memories View larger image Long-term memories may require changes in the nervous system so substantial that they can be directly observed (with the aid of a microscope, of course). After all, structural changes resulting from use are apparent in other parts of the body, as when exercise tones and shapes muscle. In a similar way, new synapses can form (or old synapses may die back) as a result of use (FIGURE 13.15C). Training can also lead to the reorganization of synaptic connections. For example, it can cause a more active pathway to take over sites formerly occupied by a less active competitor (FIGURE 13.15D). Varied experiences and learning cause the brain to change and grow The remarkable plasticity of the brain is easy to demonstrate. Classic studies found that simply living in a complex environment, with its many opportunities for new learning, produces pronounced biochemical and anatomical changes in the brains of rats (Renner and Rosenzweig, 1987). In studies of environmental enrichment, rats are randomly assigned to one of three housing conditions: 1. Impoverished condition (IC) Animals are housed individually in standard lab cages (FIGURE 13.16A). 2. Standard condition (SC) Animals are housed in small groups in standard lab cages (FIGURE 13.16B). 3. Enriched condition (EC) Animals are housed in large social groups in special cages containing various toys and other interesting features (FIGURE 13.16C). This condition provides enhanced opportunities for learning perceptual and motor skills, social learning, and so on. FIGU R E 1 3 . 1 6 Experimental Environments to Test the Effects of Enrichment on Learning and Brain Measures View larger image In dozens of studies over several decades, a variety of changes in the brain were linked to environmental enrichment. For example, compared with IC animals: EC animals have a heavier, thicker cortex, especially in somatosensory and visual cortical areas (M. C. Diamond, 1967). EC animals show enhanced cholinergic activity throughout the cortex (Rosenzweig et al., 1961). EC animals have more dendritic branches on cortical neurons, and many more dendritic spines on those branches (FIGURE 13.17) (Greenough, 1976). EC animals have larger cortical synapses (M. C. Diamond et al., 1975), consistent with the storage of long-term memory in cortical areas through changes in synapses and circuits. EC animals have more neurons in the hippocampus because newly generated neurons (see Chapter 4) live longer (Kempermann et al., 1997). EC animals show enhanced recovery from brain damage (Will et al., 2004). FIGU R E 1 3 . 1 7 Measurement of Dendritic Branching View larger image These neural effects of experience, which were surprising when first reported for rats in the early 1960s, are now seen to occur widely in the animal kingdom—from flies to philosophers (Mohammed, 2001; Chan et al., 2018). But how can we study the physiology of learning when the mammalian cortex has many billions of neurons, organized in vast networks, and upwards of a billion synapses per cubic centimeter (Merchán-Pérez et al., 2009)? Researchers made progress by studying simple learning circuits, in various species, uncovering basic cellular principles of memory formation that may generalize to neurons throughout the brain. Invertebrate nervous systems show synaptic plasticity As we’ve discussed, neuroplasticity and the ability to learn are ancient adaptations found throughout the animal kingdom. At the neuronal level, even species that are only remotely related likely share the same basic cellular processes for information storage. One fruitful research strategy focuses on memory mechanisms in the very simple nervous systems of certain invertebrates. Invertebrate nervous systems have relatively few neurons (on the order of hundreds to tens of thousands). Because these neurons are arranged identically in different individuals, it is possible to construct detailed neural circuit diagrams for particular behaviors and study the same few identified neurons in multiple individuals. In these “simple” organisms, the search for memory mechanisms began with the simplest types of learning. Earlier we discussed one of the most basic forms of learning—associative learning about two stimuli, such as the case of a dog learning to associate the sound of a bell with food. Even simpler than associative learning are the types of learning that involve only one stimulus, called nonassociative learning. Perhaps the simplest form of nonassociative learning is habituation—a decrease in response to a stimulus as it is repeated. To be true habituation, the decreased response cannot be due to failure of the sensory system to detect the stimulus or due to an inability of the motor system to respond. Sitting in a café, you may stop noticing the door chime when someone enters. Your ears still detect the chime, and your body is perfectly capable of looking up to see what happened, but you’ve habituated to the sound. Scientists uncovered how the sea slug Aplysia learns to habituate to a stimulus (Kandel, 2009). If you squirt water at the slug’s siphon—a tube through which it draws water—the animal protectively retracts its delicate gill (FIGURE 13.18). But with repeated stimulation the animal retracts the gill less and less, as it learns that the stimulation represents no danger to the gill. Research demonstrated that this short-term habituation is caused by changes in the synapse between the sensory cell that detects the squirt of water and the motor neuron that retracts the gill. As less and less transmitter is released at this synapse, the gill withdrawal in response to the stimulation slowly fades (FIGURE 13.19A). FIGU R E 1 3 . 1 8 The Sea Slug Aplysia View larger image FIGU R E 1 3 . 1 9 Synaptic Plasticity Underlying Habituation in Aplysia View larger image The number and size of synapses can also vary with training in Aplysia. For example, if a slug is tested in the habituation paradigm over a series of days, each successive day the animal habituates faster than it did the day before. This phenomenon represents long-term habituation (as opposed to the short-term habituation that we just described), and in this case there is a reduction in the number of synapses between the sensory cell and the motor neuron (FIGURE 13.19B). This very simple organism taught us that learning can be accomplished either through a reduction in the strength of existing synapses or through a reduction in the number of synapses. A similar research program aimed at understanding simple learning in much more complicated species—mammals—revealed that learning could also increase the strength of synaptic connections, as we’ll see next. Classical conditioning relies on circuits in the mammalian cerebellum Success at studying the more complicated mammalian brain came when researchers probed simple associative learning: classical conditioning of the eye-blink reflex (R. F. Thompson and Steinmetz, 2009). When a puff of air is aimed at the cornea of a rabbit, the animal reflexively blinks. The neural circuit of the eye-blink reflex is relatively simple, involving cranial nerves and some interneurons that connect between cranial nerve nuclei in the brainstem (FIGURE 13.20A). Sensory fibers from the eye’s cornea detect the puff and send action potentials along axons of cranial nerve V (the trigeminal nerve) to its nucleus in the brainstem. From here, some interneurons’ axonal endings excite other cranial nerve motor nuclei (VI and VII), which in turn activate the muscles of the eyelids, causing the blink. FIGU R E 1 3 . 2 0 The Neural Circuit for Classical Conditioning of the Eye-Blink Reflex View larger image This eye-blink reflex can be classically conditioned. Over several trials, if an acoustic tone (CS) precedes the air puff (US) repeatedly, a simple conditioned response develops: the rabbit comes to blink (CR) when the tone is sounded (to review these terms and the basics of conditioning, see Figure 13.10). Early studies showed that destruction of the hippocampus and the rest of the medial temporal lobe has little effect on the conditioned eye-blink response in rabbits (Lockhart and Moore, 1975). Instead, researchers found that a cerebellar circuit is necessary for eye-blink conditioning (Poulos and Thompson, 2015). The trigeminal (cranial nerve V) pathway that carries information about the corneal stimulation (the US) to the cranial motor nuclei also sends axons to the brainstem. These brainstem neurons, in turn, send axons called climbing fibers to synapse on cerebellar neurons. The same cerebellar cells also receive information about the auditory CS by a pathway through auditory centers (FIGURE 13.20B). So information about the US and CS converges in the cerebellum. After conditioning, the occurrence of the CS—the tone—has an enhanced effect on the cerebellar neurons, so they now trigger eye blink even in the absence of an air puff (FIGURE 13.20C). Imaging studies confirm that the cerebellum is important for conditioning of the eyeblink reflex and other simple conditioning in humans (Timmann et al., 2010). How does the training change the strength of those synapses, so that now the tone triggers a blink? At least some of these plastic changes in the cerebellar neurons rely on a special synaptic mechanism that has been best studied in the hippocampus (Mao and Evinger, 2001), so we’ll turn our attention there to conclude the chapter. How’s It Going? 1. What are some of the ways that learning could alter synaptic structure or function? 2. Describe the effects of different environments on the brains of rats. 3. How does the circuitry in Aplysia change in the course of short-term and long-term habituation? 4. Which brain region is crucial for classical conditioning in mammals, and how does it play its role? FOOD FOR THOUGHT Do the extensively documented effects of environmental enrichment on the development of brain structure and function in animals have any relevance for social policy decisions regarding humans? 13.4 Synaptic Plasticity Can Be Measured in Simple Hippocampal Circuits The Road Ahead In the final part of the chapter, we will look at the biochemical signals that alter the strength of synapses that underlie memory. Reading this material should enable you to: 13.4.1 Explain the properties of a particular type of glutamate receptor that support long-term changes in synaptic strength. 13.4.2 Critically evaluate the possibility that such longterm changes in synaptic strength play a role in memory formation. Modern ideas about synaptic plasticity have their origins in the theories of Donald Hebb, who proposed that when a presynaptic and a postsynaptic neuron were repeatedly activated together, the synaptic connection between them would become stronger and more stable (the phrase “Cells that fire together wire together” captures the basic idea). These Hebbian synapses could then act together to store memory traces (Hebb, 1949). This idea was eventually confirmed in the 1970s when researchers discovered an impressive form of neuroplasticity in the hippocampus, which appeared to be Hebbian synapses (Bliss and Lømo, 1973; Schwartzkroin and Wester, 1975). In the classic experimental setup, electrodes are placed within the hippocampus, positioned so that the researchers can stimulate a group of presynaptic axons and record the electrical response of a group of postsynaptic neurons. Normal, low-level activation of the presynaptic cells produces stable and predictable excitatory postsynaptic potentials (EPSPs) (see Chapter 2), as expected. But when a brief high-frequency burst of electrical stimuli, called a tetanus, is applied to the presynaptic neurons, causing them to produce a high rate of action potentials that drive the postsynaptic cells to fire repeatedly, the response of the postsynaptic neurons changes. Now the postsynaptic cells produce much larger EPSPs; in other words, the synapses appear to have become stronger, more effective. This stable and long-lasting enhancement of synaptic transmission is termed long-term potentiation (LTP; potentiation means “strengthening”; FIGURE 13.21). FIGU R E 1 3 . 2 1 Long-Term Potentiation Occurs in the Hippocampus View larger image We now know that LTP can be generated in conscious and freely behaving animals, in anesthetized animals, and even in isolated slices of brain. LTP is also evident in a variety of invertebrate and vertebrate species. Once induced by a tetanus, LTP can last for weeks or more. So, at least superficially, LTP appears to have the hallmarks of a cellular mechanism of memory: a long-lasting change in synaptic strength. This hint at a cellular origin prompted research into the molecular and physiological mechanisms underlying LTP. NMDA receptors and AMPA receptors collaborate in LTP The region called the hippocampal formation consists of two interlocking C-shaped structures: the hippocampus itself and the dentate gyrus. At least three different pathways in the hippocampal formation display LTP, and it is seen in other brain regions too (Bliss, 2022). The most studied form of LTP occurs at synapses that use the excitatory neurotransmitter glutamate, and it is critically dependent on a glutamate receptor subtype called the NMDA receptor (after its selective ligand, N-methyl-d-aspartate). Treatment with drugs that selectively block NMDA receptors completely prevents new LTP in this region, but it does not affect synaptic changes that have already been established. As you might expect, these postsynaptic NMDA receptors—working in conjunction with other glutamate receptors called AMPA receptors—have some unique characteristics, which are responsible for LTP. During normal, low-level activity, the release of glutamate at the synapse activates only the AMPA receptors. So the EPSP is mediated entirely by these AMPA receptors. The NMDA receptors cannot respond to the glutamate, because magnesium ions (Mg ) block the NMDA receptor’s calcium ion (Ca ) channel (FIGURE 13.22A); thus, few Ca ions can enter the neuron. The situation changes, however, if larger quantities of glutamate are released—say, in response to a barrage of action potentials caused by a tetanus. That stronger stimulation of the AMPA receptors depolarizes the postsynaptic membrane so much that the Mg plug is ejected from 2+ 2+ 2+ 2+ the NMDA receptor’s channel (FIGURE 13.22B). Now the NMDA receptors are also able to respond to glutamate, admitting large amounts of Ca into the postsynaptic neuron. Thus, NMDA receptors are fully active only when “gated” by a combination of strong depolarization (via AMPA receptors) and the ligand (glutamate). FIGU R E 1 3 . 2 2 Roles of the NMDA and AMPA Receptors in the Induction of LTP in the CA1 Region CaMKII, calcium/calmodulin-dependent protein kinase II; CREB, cAMP responsive element–binding protein; Glu, glutamate; PKC, protein kinase C; TK, tyrosine kinase. View larger image The large influx of Ca at NMDA receptors activates a variety of intracellular enzymes that affect AMPA receptors in several important ways (FIGURE 13.22C) (Kessels and Malinow, 2009). First, the enzymes cause existing nearby AMPA receptors to move to the active synapse, and they modify the AMPA receptors to increase their conductance of Na and K ions (Sanderson et al., 2008). In addition, more AMPA receptors are produced and inserted into the postsynaptic membrane. Thus, after the tetanus there are more 2+ 2+ + + AMPA receptors, and those receptors are more effective, so the synaptic response to glutamate is strengthened (see Figure 13.22B). There are presynaptic changes in LTP too. When the postsynaptic cell is strongly stimulated and its NMDA receptors become active and admit Ca , an intracellular process causes the postsynaptic cell to release a retrograde transmitter—probably a diffusible gas (a gasotransmitter; see Chapter 3)—that travels back across the synapse and alters the functioning of the presynaptic neuron (see Figure 13.22B). The retrograde transmitter induces the presynaptic terminal to release more glutamate than previously, thereby strengthening the synapse some more. So, LTP involves active changes on both sides of the synapse. So far, we’ve talked about how activity can make existing Hebbian synapses stronger. However, evidence suggests that the same mechanisms can affect whether new synapses are formed and old synapses retracted. In these systems it appears that when several presynaptic neurons fire at the same time, they “gang up” on the postsynaptic cell, depolarizing it enough that the NMDA receptors are activated to strengthen those connections. Conversely, any presynaptic neurons that tend to fire out of synchrony with the other inputs are not likely to depolarize the postsynaptic neurons enough to activate NMDA receptors. Eventually, the strengthened inputs seem to sprout new, additional connections, while the weakened synapses fade away (FIGURE 13.23). 2+ FIGU R E 1 3 . 2 3 At Hebbian Synapses, Neurons That Fire Together Wire Together View larger image Scientists are intrigued by LTP because this momentary burst of neural activity, the tetanus, can change synaptic strength for a long time. It’s easy to imagine how another momentary burst of neural activity, in this case triggered by a learning experience, could change synaptic strength, and that change in synaptic strength might be a memory trace. But is this just a case of an overactive imagination, or is LTP truly involved in learning? Is LTP a mechanism of memory formation? Even the simplest learning involves circuits of multiple neurons and many synapses, and more-complex declarative and procedural memory traces must involve vast networks of neurons, so we are unlikely to conclude that LTP is the only mechanism of learning. However, evidence from several research perspectives implicates LTP in at least some forms of memory: 1. Correlational observations The time course of LTP bears strong similarity to the time course of memory formation. 2. Somatic intervention experiments In general, pharmacological treatments that interfere with LTP also tend to impair learning. So, for example, NMDA receptor blockade interferes with performance in the Morris water maze (a test of spatial memory) and other types of memory tests (R. G. Morris et al., 1989). Knockout mice that lack functional NMDA receptors only in the CA1 region of the hippocampus appear normal in many respects, but their hippocampi are incapable of LTP and their declarative memory is impaired (Rampon et al., 2000). Conversely, mice engineered to overexpress NMDA receptors in the hippocampus have enhanced LTP and better-than-normal long-term memory (Y. P. Tang et al., 2001). 3. Behavioral intervention experiments In principle, the most convincing evidence for a link between LTP and learning would be “behavioral LTP”: a demonstration that training an animal in a memory task induces LTP somewhere in the brain. Such research is difficult because of uncertainty about exactly where to put the recording electrodes in order to detect any induced LTP. Nevertheless, several examples of successful behavioral LTP have been reported (Whitlock et al., 2006). Taken together, these findings support the idea that LTP is a kind of synaptic plasticity that underlies (or is very similar to) certain forms of learning and memory. One theory is that an experience, such as seeing a tiger, activates some form of LTP in the hippocampal formation, which uses axons projecting to the outermost layers of cortex, to serve as a “memory switch.” Over time, this hippocampal input to layer I cortex modifies synapses there to encode the memory, perhaps forever (FIGURE 13.24). If the experience is emotionally intense, activating the amygdala (Chapter 11; see Figure 11.13C), then amygdala input to cortex may enhance that encoding of the memory, making it last longer and/or be more vivid. If this theory is correct, then the cause of Henry Molaison’s tragic amnesia may have been the loss of the hippocampal formation and therefore the chance to store new memories in his cortex. Henry’s memories before the surgery, already encoded in his cortex, were preserved. FIGU R E 1 3 . 2 4 A Theory of LTM Storage in Layer I of Cortex View larger image It’s strange to think that microscopic changes in synapses could be so crucial for living a full human life. Eventually Henry stopped being shocked when he saw his gray-haired, wrinkled reflection. But it’s not clear whether he understood—for long—that his parents had passed away. As a final act of generosity to a field of science that he helped launch, Henry arranged to donate his brain for further study after he died (to read more on this topic, search for “project HM” at The Brain Observatory), making a series of more than 2000 brain sections available to researchers. To the end, although Henry could remember so little of his entire adult life, he was courteous and concerned about other people. Henry remembered the surgeon he had met several times before his operation: “He did medical research on people…. What he learned about me helped others too, and I’m glad about that” (Corkin, 2002, p. 158). Henry never knew how famous he was or how much his dreadful condition taught us about learning and memory; despite being deprived of one of the most important characteristics of a human being, he held fast to his humanity. How’s It Going? 1. Describe how LTP is measured in the hippocampus. 2. What happens to AMPA receptors and NMDA receptors during LTP? 3. What evidence suggests that LTP may underlie some forms of learning and memory? FOOD FOR THOUGHT If it becomes possible to develop drugs that specifically enhance LTP and therefore memory formation, it should improve memory. What limits, if any, should there be in allowing people to take such a drug? RECOMMENDED READING Baddeley, A. D., Eysenck, M., and Anderson, M. C. (2020). Memory (3rd ed.). London, UK: Routledge Press. Gluck, M. A., Mercado, E., and Myers, C. E. (2020). Learning and Memory: From Brain to Behavior (4th ed.). New York, NY: Worth. Kandel, E. R. (2018). The Disordered Mind: What Unusual Brains Tell Us about Ourselves. New York, NY: Farrar, Straus, & Giroux. Rudy, J. W. (2021). The Neurobiology of Learning and Memory (3rd ed.). Sunderland, MA: Oxford University Press/Sinauer. Slotnick, S. D. (2017). Cognitive Neuroscience of Memory. Cambridge, UK: Cambridge University Press. VISUAL SUMMARY You should be able to relate each summary to the adjacent illustration, including structures and processes. The online version of this Visual Summary includes links to figures, animations, and activities that will help you consolidate the material. Visual Summary Chapter 13 View larger image LIST OF KEY TERMS Amnesia AMPA receptors anterograde amnesia associative learning basal ganglia cerebellum classical conditioning cognitive map confabulate consolidation Declarative memory delayed non-matching-to-sample task dentate gyrus dorsomedial thalamus encoding Enriched condition (EC) episodic memory glutamate habituation Hebbian synapses hippocampus Impoverished condition (IC) instrumental conditioning Korsakoff’s syndrome learning long-term memories (LTMs) long-term potentiation (LTP mammillary bodies memory memory trace neuroplasticity NMDA receptor Nondeclarative memory Patient H.M. Patient K.C. Patient N.A. place cells Priming reconsolidation retrieval retrograde amnesia retrograde transmitter semantic memory sensory buffers short-term memories (STMs) Skill learning Standard condition (SC) tetanus