In the early 1980s, my lab modified the Brown-Peterson task to discover how long Henry’s short-term memory lasted. This experiment was part of a broad investigation aimed at dissecting the component processes of declarative memory—a system that supports long-term memory for events and facts. When Henry and four other amnesic patients performed the Brown-Peterson task, they were neck and neck with healthy control participants without memory deficits at the three-, six-, and nine-second delays. At fifteen and thirty seconds, however, because their short-term store was full to capacity, the patients’ scores fell well below those of the controls. In this experiment, normal people tapped into their long-term memory to retrieve information that needed to be held beyond about fifteen seconds, but people with amnesia could not.14 This study helped define the limits of short-term memory.
Our understanding of short-term memory has become more complex as we study how we use our memories in everyday life. As we take in information from the world, we engage numerous complex processes in the brain. If someone tries to multiply sixty-eight and seventy-three in her head, she is performing calculations, saving the results, combining numbers, and checking their accuracy. The task requires much more effort than just regurgitating items held in her short-term memory; it is mental labor. She draws on the abstract ideas of numbers and multiplication, applying that knowledge to the problem at hand. This sort of process is called working memory: an effortful extension of short-term memory, a mental workspace where cognitive operations take place.
How does working memory differ from short-term, immediate memory? Think of short-term memory as simple and working memory as complex. Working memory is short-term memory on overdrive. Both are temporary, but short-term, immediate memory is the ability to reproduce a small number of items after a short or no delay (like saying 3–6–9), whereas working memory requires storing small amounts of information and simultaneously working with that information to perform complex tasks (like multiplying 3 x 6 x 9 in your head). When we engage short-term memory, we simply repeat a limited amount of information, whereas when we access working memory we can monitor and manipulate that information any way we want. Working memory organizes whatever cognitive and neural processes are necessary to achieve a short-term goal—deciphering long sentences, solving problems, following the plotline of a movie, carrying on a conversation, keeping track of a baseball game play by play.
Although the concept of working memory was first introduced in 1960, it was not until the 1980s that articles about working memory began to appear in the neuropsychological literature. In 1962, however, Milner administered a problem-solving test to Henry that we later realized measured not only his problem-solving ability but also his working memory capacity. Milner placed four cards on a table, side by side, and told Henry, “Here are your key cards.” The first card had a red triangle, the second two green stars, the third three yellow crosses, and the fourth four blue circles. She told him to take a stack of 128 cards and place each card on the table in front of one of the key cards, wherever Henry thought it should go. After he placed a card down, she would say “Right” or “Wrong,” and he was supposed to use this information to make as many correct choices as possible. He started sorting the cards, and initially she said “Right” if he matched a card to a key card based on the color and “Wrong” if he matched it by shape or number. After ten correct choices, she changed the sorting category without telling him. Now she said “Right” if he sorted by shape. She then changed again, using number as the sorting principle. Henry completed the task well and made few errors. He was engaging his working memory: he had to keep his attention focused on the correct category while he placed the cards on the table, listened to Milner’s response, and decided on the next card placement accordingly. But despite his superior performance, he did not remember at the end of the test that he had been changing strategies—placing the cards first to match the color, next to match the shape, and then to match the number—in response to Milner’s cues.15
The card-sorting task shed additional light on what Henry could and could not do. His ability to stick with a category as long as Milner said “Right” and to shift to another category when she said “Wrong” attests to his capacity to pay attention throughout a long test, to discriminate among different shapes and colors, and to think and respond flexibly. All these computations occurred online without the need to tap into long-term memory. When the test was over and Henry tried to think back over the entire test and recollect all he had just done, he was at a loss. The critical long-term memory traces were nowhere to be found.
Around the same time, Milner had the opportunity to administer the card-sorting task to a patient in whom Penfield had removed the front third of both frontal lobes to alleviate epilepsy. This man sorted all 128 cards based only on shape, even when Milner repeatedly said “Wrong”—an extreme example of perseverative behavior, responding in the same way over and over again. This case and many others that Milner tested with removals of either the left or right frontal lobe showed convincingly that the flexibility of planning and thought required for the sorting task depended on normal frontal-lobe function. Based on these findings, we can state with confidence that Henry’s frontal-lobe capacities were excellent.16
In the 1990s, my lab dedicated a substantial effort in our studies with Henry to assessing his working memory. We expected that his working-memory capacities would be unaffected because he could monitor and manipulate items in his intact short-term stores. But the working memory tests sometimes posed a challenge for Henry in two respects. On one task, he had to respond quickly to keep up with the imposed pace of the test, and in some cases, he may not have had sufficient time to decide on and execute the correct response. On another test, the number of stimuli he had to monitor and manipulate exceeded the capacity of his immediate memory, and therefore, required engagement of long-term declarative memory, which he lacked.
For the test with time constraints, the N-back test, the stimuli were patches of color (e.g., red, green, blue), presented one at a time on a computer screen at a rate of one color every two seconds. We asked Henry to press one button whenever the current color matched the color that had appeared immediately before, and a different button when it did not match. After he completed a number of such trials, the task became more challenging because we instructed him to press the button whenever he saw a color that matched the one that occurred two before it (i.e., with one intervening stimulus) and a different button when the color two before did not match.
On the N-back test with colors, Henry’s performance was intact, attesting to the integrity of three key cognitive processes—information maintenance (Henry had to hold the colors online as they appeared), information updating (he had to constantly update the color he was holding online), and response inhibition (he had to inhibit the tendency to always press the non-target button because matches occurred less frequently than non-matches). When the stimuli were colors, the two-second time limit did not handicap him.
We later administered two similar N-back tests using six spatial locations and six meaningless shapes instead of colors. With these new stimuli, Henry’s performance was no better than if he had been guessing. He had great difficulty maintaining the spatial locations and meaningless shapes in his working memory perhaps because he could not quickly attach verbal labels to them and respond correctly during the two-second window. Faced with test stimuli that were difficult to verbalize and the need to respond quickly, Henry was unable to succeed on this task.
On another test, self-ordered choosing, we measured Henry’s ability to plan and keep track of a sequence of his own responses. He saw six designs displayed in a grid on a computer screen with three in the top row and three in the bottom row. We asked him to choose one design. Then he saw a new screen with the same six designs, each presented in a different position. This time he had to choose a different design. On the four subsequent trials, he again saw the six designs, each time in a different location, and
again had to choose one he had not selected before. Henry completed this procedure three times in succession, and his performance was comparable to that of a control participant. In subsequent testing, however, when we increased the number of designs to eight and then twelve, Henry made more errors than the control. Because his errors tended to occur in the later trials, he could have been getting interference from constantly monitoring his experience of the early trials. Moreover, the demand of keeping track of eight and then twelve items, likely exhausted the limits of his immediate memory, and he did not have long-term declarative memory to fall back on.
Henry’s limited performance on our working memory tasks does not lead to the conclusion that working memory depends on an intact hippocampus. He was thwarted by the tests that required him to respond rapidly or to engage his long-term memory. When the capacity of immediate memory is insufficient to hold the number or complexity of stimulus items online, declarative memory and the medial temporal-lobe circuits that support it mediate successful performance. When these circuits are dysfunctional as in Henry and other amnesic patients, they are likely to fail on challenging working memory tests. In 2012, memory researchers at the University of California, San Diego, reviewed ninety articles in the neuroscience literature on this topic and arrived at the same conclusion—when the requirements of a task are greater than the capacity of working memory can accommodate, performance is supported by long-term, declarative memory.17
Research on working memory has become a vast area of investigation, with more than twenty-seven thousand scientific papers on this topic. Ongoing studies in thousands of laboratories continue to dissect working memory processes and circuits to establish brain-behavior correlations in animals and humans. Because working memory relies on multiple cognitive processes—attention, impulse control, storing, monitoring, ordering, and information manipulation—it recruits multiple brain circuits in parallel. Consequently, working memory is highly vulnerable to neurological conditions, so we see working memory deficits in patients with attention-deficit/hyperactivity disorder, autism, Alzheimer disease, Parkinson disease, HIV, and strokes, and even normal, healthy aging. Individuals in these groups sometimes have trouble engaging working memory to achieve their goals because this challenging mental effort requires the brain to be fit; even subtle abnormalties can cause performance to suffer.
The notion of working memory emerged not from neuroscience labs but rather from the field of applied mathematics. Norbert Wiener, widely considered the most brilliant American-born mathematician of his generation, proposed in 1948 that the brain is like a computing machine. With this insight he established the discipline of cybernetics, the study of control processes in humans and machines.18
The metaphor of the human brain as an information processor had a far-reaching effect on the field of neuroscience. It influenced George A. Miller, Eugene Galanter, and Karl H. Pribram, three mathematically oriented thinkers at Stanford’s Center for Advanced Study in the Behavioral Sciences, to unite the disciplines of cybernetics and psychology in their 1960 book, Plans and the Structure of Behavior. In this pioneering work, they argued that behavior must be directed by an overarching “Plan.” They made a radical proposal that the brain could be compared to a computer, and the mind to a computer program, or Plan.19
Plans and the Structure of Behavior introduced the term working memory, a concept that quickly became an active area of research in cognitive science and cognitive neuroscience. This vast area of research goes beyond memory formation by addressing our ability to achieve complex goals. Specific goal-directed plans, analogous to computer programs, are stored somewhere and retrieved while they are being executed. “The special place may be on a sheet of paper,” they wrote. “Or (who knows?) it may be somewhere in the frontal lobes of the brain. Without committing ourselves to any specific machinery, therefore, we should like to speak of the memory we use for the execution of our Plans as a kind of quick-access, ‘working memory.’ As Milner’s card-sorting study with Henry demonstrated, not only were the authors spot-on in their characterization of working memory, they were also correct in their guess that it resides in the frontal lobes of the brain. We now know that the prefrontal cortex is critical for holding multiple thoughts in mind to construct and carry out a plan, just as Henry did when he performed the card-sorting task.
Over the next decade, the study of working memory exploded, as psychologists and neuroscientists sought to dissect the underlying cognitive and neural processes. In 1968, psychologists Richard Atkinson and Richard Shiffrin described a detailed model for human memory in “Human Memory: A Proposed System and Its Control Processes,” still one of the most-cited writings in the literature on human memory. They divided memory into three stages: the sensory register, the short-term store, and the long-term store. The sensory register is the first entry point for incoming information from the senses. The information resides there for less than a second and then decays. In their single-process model, the short-term store is working memory; it receives input from the sensory register as well as from long-term stored memories. Information flows along a continuum, from the short-term store to the long-term store, which is a relatively permanent silo.20
Although the Atkinson-Shiffrin single-process model was influential, it did not fully account for the mechanism of long-term memory formation. If the model were correct, then Henry should not have had amnesia because, with the passage of time, information present at the short-term stage would have flown automatically into the long-term stage. Clearly, this did not happen—Henry’s brain could not translate information from short-term processing mechanisms to long-term processing mechanisms. Still, Atkinson and Shiffrin’s model is noteworthy for defining the short-term store as the individual’s working memory where control processes come into play. Control processes vary from individual to individual. We decide what to pay attention to; rehearse information to keep it in the short-term store; and create mnemonics, such as Every Good Boy Deserves Fudge, which music students invoke to recall the notes on the lines of the treble clef—EGBDF. The Atkinson-Shiffrin model started the quest to understand the strategies that influence the processing of information held in working memory.
In 1974, psychologist Alan Baddeley and his colleague Graham J. Hitch posited that working memory is not a unitary system. They proposed that it consisted of three subsystems: a central executive that calls the shots, and two slave systems that do the hard work—one devoted to visual information and one to language. This model generated an explosion of experiments that tried to identify the mechanisms operating within each subsystem, how the transient processes interact with long-term memory, and the brain areas that are recruited during the performance of working memory tasks. Scientists have studied working memory in healthy people of all ages, twins, bilinguals, menopausal women, the congenitally blind, smokers, insomniacs, people under stress, and people with numerous neurological and psychiatric disorders. These experiments have had wide-ranging impact, with implications for education, evaluating treatments, training programs and their utility, and evaluation of psychiatric disorders.21
In recent years, however, Baddeley’s model, which relies on dedicated holding areas—one for vision, one for hearing, and one for unique events—has been superseded by the concept of a more dynamic system. Current thinking considers working memory processes as interacting with long-term memory stores. By this view, information in working memory is kept alive by active processing in multiple areas within the temporal, parietal, and occipital lobes, in the same specialized brain areas where the initial perception of that information occurred. So, the circuits that are called into play when we first hear a name, see a face, or enjoy a landscape are the same ones that are active when we later remember that name, face, or landscape. The network that is recruited by working memory processes at a particular time depends on the content of our working memory and what we are trying to accomplish.
Twenty-first-century models of working memory emphasiz
e the interactions between short- and long-term memory. Cognitive neuroscientists Bradley Postle, Mark D’Esposito, and John Jonides have each championed the view that working memory integrates specific information from different time periods—the sights, sounds, smells, tastes, and skin and body sensations that just entered the brain, and the contents of long-term memory that are relevant to these inputs. For example, if we try to multiply thirty-six by thirty-six in our heads, a task that engages working memory, we need to access our stored knowledge about numbers and multiplication to do the calculation. These researchers see working memory as an emergent phenomenon that arises from cooperation among many brain areas. As a result, the human brain is capable of multitasking and dealing with different kinds of information concurrently, switching from one task to another with great flexibility.22
Imagine a woman in a restaurant who is listening to a waiter list the day’s specials. She keeps the list of dishes active in her working memory while simultaneously evaluating each dish, based on knowledge stored in her long-term memory. After shuffling these options in her head, she decides not swordfish, because of its mercury content; not fried chicken, because it has too much fat; but the vegetarian pasta sounds similar to a dish I liked before. She orders the pasta. Although she made this decision quickly, it came about through cooperation across networks of brain areas, enabling her to monitor and manipulate different kinds of information.
Permanent Present Tense Page 9
