Learn how current researchers are developing insights about the components of working memory, and what this might mean for our understanding of how we think and learn.

“Four-Nine-Seven, Oh-Two-Five-Four. Got it?”

“Sure. See you later!”

As soon as I hung up the phone, I realized that I had no paper or pen. As I rifled my room for writing implements, I wondered: “How am I remembering these numbers?”

This simple process of remembering things for a short period of time happens every day of our lives; it is fundamental to our experience of the world. Memory over a short period of time, called ‘working memory’, has generated much interest recently both because of its importance to many higher brain functions and the evolution of powerful techniques to study brain processes, such as PET and fMRI. Based on these techniques, scientists can pursue exciting questions about the neural underpinnings of working memory.

What is Working Memory?
A great deal of evidence indicates that working memory is an entirely different process from long term memory. For example, the famous neurological patient, H.M. who has not formed an explicit long term memory since the day of an operation to remove his hippocampi in 1954, has intact working memory. If you or I were to meet him, we could interact with him and sustain a normal conversation about Eisenhower or that new gadget television until a door slammed or something distracted his attention; at that point, we would have to begin again. Long term memory of the sort that H.M. is missing is operative over long periods such as hours, days, or even a lifetime. It has been clinically and experimentally well studied, and has been shown to involve brain regions such as the hippocampus. It is thought to be mediated by changes in cell functioning, such as long term potentiation (LTP).

In sharp contrast, working memory seems to be something profoundly different. Scientists, particularly psychologists and cognitive scientists, have long been curious about working memory because of its involvement in all cognitive processes. Early psychological work in the 1950’s and 1960’s led to the hypothesis of ‘short term memory’; a process of limited capacity and only operative over a few seconds. The concept of ‘working memory’ is an extension of this idea, with the added idea that short term memory is woven together with higher cognitive processes, such as learning, reasoning, and comprehension.

Unlike long term memory, which has a large clinical body of research, working memory has only recently become the focus of intense clinical study. It is often assayed in intelligence or cognitive examinations using span tests, in which patients are asked to repeat a set of digits in reverse order (if I read “8-9-3-2-1-9”, you would say “9-1-2-3-9-8”) or alphabetize a group of words that had been read aloud. Studies of patients with various frontal lobe lesions do not show a systematic deficit in storage. These studies indicate that working memory is not one process; rather, it is made up of several separable processes.

A Psychological Perspective
Alan Baddeley, in his landmark book Working Memory, captures three decades of psychological work on working memory systems. Many working memory experiments simply consist of stimuli that are to be remembers for a few seconds. A typical task might ask you to remember a few letters, numbers, or features of an object. Typically, there is a brief delay, after which the subject is ‘probed’, or asked what he or she remembers. From extensive studies like these, Baddeley proposed a model of working memory that involved three distinct subsystems. The best described is the ‘phonological loop’, a system that draws upon speech resources. For example, if I wanted to remember a set of numbers, I might catch myself whispering to myself — it turns out that speech systems are an integral part of working memory. The second component is the visuospatial sketchpad, a parallel system akin to an artist’s sketchbook for stimuli that cannot be verbalized, such as spatial information. The third main unit is the central executive, a system responsible for supervisory attentional control and cognitive processing. This last system, though poorly defined, is most alluring because it represents the very stuff of thought.

Where is Working Memory in the Brain?
The rich psychological research, the simplicity and fundamental nature of working memory systems, and the adaptability of working memory experiments make it ripe for new brain imaging technologies. Both PET and fMRI capitalize on properties of cerebral blood flow to make inferences about underlying neural activity. Founded upon Baddeley’s model of working memory, investigators have begun to explore neural correlates of working memory. Several neuroimaging studies provide evidence for a distinct neurological basis for a phonological loop, as well as separate processes for storage of items and retrieval. During the storage phase of verbal working memory tasks, activity is found in Broca’s area (involved in speech production) in addition to supplementary and premotor areas (involved in movement) in frontal cortex, and is strongly consistent with activity in areas involved in preparation of speech from other neuroimaging studies. In addition, different networks are involved in retrieval as compared with storage in the left lateralized frontal cortex.

The neural correlates of spatial or object storage, in pursuit of the visuospatial sketchpad, is somewhat more tenuous. Neuroimaging studies yield that there are different areas activated in spatial or object memory tasks compared to those in verbal working memory tasks. Neuroimaging studies also suggest a difference in storage systems compared with retrieval systems in spatial or object working memory, indicating that there are again separate networks at work.

The Elusive Central Executive
The most fascinating line of inquiry confronts the idea of a ‘central executive’, a control system that mediates attention and regulation of processes occurring in working memory. The idea of a central executive was first postulated by Baddeley. Many investigators have seen evidence supporting the idea of a central executive; they have observed higher cognitive activity in an area in the prefrontal cortex, called DLPFC (Dorsolateral Prefrontal Cortex), during difficult tasks. This area shows activity during object working memory, and what are termed ‘executive processes’, such as planning, focusing attention on an object, switching between tasks, and ‘inhibition’ of short term storage (which are often tested using probes designed to distract subjects). One powerful design to study executive processes is to tax working memory systems to its capacity, or to present the subject with two tasks to perform simultaneously. As the reasoning goes, if you make working memory systems work hard, the central executive will intervene to manage the increased load. Examples of such difficult tasks include remembering a set of numbers while doing simple math or the famous Stroop task, where color names are presented in different colors (“red”, for example, might be presented in green text).

A few neuroimaging studies using these difficult tasks support the notion of a central executive control system. In one fMRI study conducted at the University of Pennsylvania, participants had to place objects in a category and decide whether two visual displays differed by rotation. In the dual task condition, frontal areas showed increased activity, including DLPFC and the anterior cingulate gyrus (an attentional area). Both areas are active in attention and inhibition tasks, and the anterior cingulate has been implicated in PET studies of the Stroop test. Despite these studies, the concept of a central executive still remains tantalizing and mysterious, and much further exploration remains to be done.

Other Evidence About Working Memory
In addition to neuroimaging studies, there is converging evidence from animal models and cellular studies. Typically, awake, behaving monkeys are studied with electrophysiology, and the interconnections of individual circuits can be mapped out. Many of these circuits, thought to be comprised of large pyramidal cells, are focused in monkey prefrontal cortex, analogous to that of humans. Careful studies of these neurons reveal exquisite patterns of neuronal connectivity, and several models of working memory are derived from this connectivity.

The pharmacology of working memory also proves fascinating. Dopamine, a largely inhibitory neurotransmitter with many functions, is thought to play a major role in working memory. The frontal cortex has many dopaminergic pathways, which may modulate the activity of the pyramidal cells in the frontal cortex. Again, although much evidence has been marshalled by scientists who study working memory in monkeys, the relationship to human working memory systems is unclear.

As more information becomes available about working memory, it will become possible to think clearly about diseases such as schizophrenia and Alzheimer’s disease, conditions that show clear deficits in working memory. Many scientists have studied the way that working memory interacts with these and other diseases; however, without knowing more about the structure of working memory, it is difficult to draw further conclusions about its specific interaction with neuropsychological disease. Another intriguing line of research involves tracking working memory systems in aging; as humans age, there is a clear decline in their working memory capacity. It is not yet clear what component of the systems of working memory is responsible for this decline.

Despite the current intensive inquiry into how we remember things on a short time scale, the components of working memory remain shrouded in mystery. Further research on the systems of working memory will result in greater understanding of this fundamental system that we use almost every moment of our lives, providing insights into the higher cognitive processes that it feeds.

Kumar Narayanan graduated from Stanford University in 2000 and is currently studying Neurobiology in the M.D./Ph.D. program at Yale University. His hometown is Seattle, Washington.

Baddeley A., Recent developments in working memory., Curr Opin Neurobiol. 1998 Apr;8(2):234-8. Review.

Smith EE, et al. Storage and executive processes in the frontal lobes., Science. 1999 Mar 12;283(5408):1657-61.