Working memory is the tool that allows us to navigate an ever-changing world, as we assess what information to ignore, what information to retain for mere seconds, and what information to process as lasting, long-term memories. It is the process that is at work at this very moment, as you—the reader—move from sentence to sentence: each one, I would hope, building on the previous. Or perhaps your mind is elsewhere—your eyes scanning these lines in rhythm yet your attention lost—then, alas, I have failed. But working memory is still working—you have just established a stronger neural coalition with another brain region, and not that of the visual stream from the text before your eyes. If that is the case, I beg your return!
Working memory has been studied on all levels of neuroscientific analysis, from behavioral tests to the cellular mechanisms of neurotransmitter release and receptor function at the synapse. The idea that working memory is contained within one region of the brain is most likely a false assumption— however, the prefrontal cortex (PFC), the region just behind the forehead, is most often correlated with working memory, as well as a host of other executive functions that are uniquely human, such as complex planning, decision making, and the censorship of one’s impulses. Mark D’Esposito, researcher at the University of California in Berkeley, provides the following definition: “Working memory refers to the temporary retention of information that was just experience but no longer exists in the external environment, or was just retrieved from long-term memory” (D’Esposito, 2007).
Working memory is a particularly difficult cognitive process to study because of its rapidly transient nature. It is the moment after sensory information has entered the brain, but before consolidation into short-term or long-term memory occurs—it is the moment after you receive directions over the phone and where you begin to recite those steps to yourself in hopes of finding your destination.
Presumably the refinement of this process and the ability to instantly call upon the vast stores of memory in the brain to solve problems presented by working memory has been very evolutionarily adaptive for humans: perhaps this explains why working memory is thought to be correlated with the activity of the PFC, an area that has increased in size so rapidly since we parted evolutionary paths with the apes. As the human brain has tripled in size over the five million years of evolution from our ape ancestors, the PFC has increased in size sixfold. This type of rapid increase in cortical surface area over our evolutionary history is not seen anywhere else in the brain.
The prefrontal cortex first came into focus as a potential seat of working memory when single cell recordings found persistent, sustained levels of neuronal firing in that region during tests that require a monkey to retain information over a short period of time in order to perform goal-directed behaviors (Fuster and Alexander, 1971; Kubota and Niki, 1971). These tasks often require a remembered response to a previously administered stimulus cue, such as an eye saccade to the remembered location of a previously displayed flash of light. The single cell recordings from these early studies revealed that neurons in the PFC seemed to show fast-spiking behavior that was sustained for the period of time that the stimulus location was retained by the monkey.
In humans, these findings have been supported by functional magnetic resonance imaging (fMRI, which measures levels of bloodflow throughout the brain) techniques that show pronounced PFC activity during delay tasks (Curtis and D’Esposito, 2003). While these results lead us toward the PFC, figure 1b clearly shows that there are additional regions that yield pronounced fMRI readings during the delayed response task at hand. Other experiments have shown that these coalitions of activation during the retention phase fluctuate depending on whether the task is retrospective (requiring the recall of past sensory experience) or prospective (anticipating future action).
The prefrontal cortex can be divided into three distinct regions that have been shown to correlate with the processing of different types of information. While the distinctions are based largely on superficial fMRI data, the orbitofrontal and ventromedial areas seem to be most relevant to reward-based decision making. The dorsolateral areas seems to be critical in making decisions that call for the consideration of multiple sources of information, and the anterior and ventral cingulated cortex appear to activate most saliently when dealing with conflicting sensory data that requires integration (Krawczyk, 2002).
Like most of the brain, neurons in the PFC are activated by the predominant excitatory neurotransmitter, glutamate, and inhibited by GABA. While these transmitters are the most widespread and are necessary for PFC activity, other modulatory neurotransmitters have been shown to play a crucial role in working memory. Catecholamines—specifically dopamine and norepinephrine—play direct roles in working memory in humans and animals.
Depletion of catecholamines in rats has been shown to severely impair working memory in rats (Simon et al. 1979), whereas the application of dopamine into the PFC of monkeys performing working memory tasks has been shown to significantly increase spike activity (Sawaguchi 2001).
Yet the correlation of increased working memory ability on a behavioral level and higher levels of catecholamines is not exponential: an important body of research has established the presence of an inverted-U-shaped response curve that suggests working memory functions with an optimal range of dopamine D1 receptor stimulation in the PFC,
with insufficient and excessive stimulation showing dramatic drop-offs in working memory-dependent task performance (Williams and Goldman-Rakic, 1995). Results have suggested that dopamine is involved in the PFC not as a reward mechanism for successful actions but as a modulatory factor that aids, and perhaps causes—at the correct levels of release—the accurate recall of relevant elements of working memory.
To that effect, a 2005 study showed that distinct molecular processes are at work for memory retrieval lasting seconds versus memories recalled after several minutes (Runyan and Dash, 2005). In this experiment, slice preparations from rats were studied after tasks requiring quick, working memory, versus longer “short-term” memory retrieval, such as that seen in the experiment discussed above. The results show that there is a clear difference in PKA activity between these two temporal periods of information storage and retrieval: in working memory, PKA action was demonstrated to be detrimental, while in short-term memory, PKA function was necessary to avoid behavioral error. The explanation of these results set forth in the study suggests that in longer periods of memory retention the need to select between conflicting internal representations becomes necessary—a process that, on the cellular level, is mediated by PKA activity—while working memory relies more on transient, singular representations without as much conflicting information.
While these cellular studies offer insight into the dynamics of PFC circuits, our task is to connect these synaptic behaviors to the behavior of the organism, as far as working memory is concerned. In addition to the action of dopamine, NMDA receptor action seems to be a crucial component of local PFC circuitry. NMDA receptors are particularly intriguing in the study of working memory because of their unique properties that require depolarization to remove a magnesium block and allow ions to flow into the cell, furthering depolarization. This delayed-onset behavior has been suggested as a possible mechanism for PFC pyramidal cells to sustain firing after initial excitation, and thus as a possible cellular trace of the storage of working memory.
NMDA receptor activation has been repeatedly linked to the innervation of the dopaminergic system (Tsukada et al. 2005). One can begin to see how reverberating connections between sensory regions and the PFC could be the telltale sign of working memory, dependent on NMDA receptor-mediated dopaminergic modulation in pyramidal neurons in the PFC as well as inhibition of background noise to focus on the retention of specific sensory cues for their relay to appropriate, goal-directed outputs, or consolidation into longer-term memories by the hippocampus. A study published last month in the Journal of Neural Systems by researchers at Boston University explains how dopaminergic neurons may keep short-term memories alive in the PFC by firing sharply initially and then maintaining a lower voltage, just enough to prime the cell for a quick re-fire.
As we begin to sort out what neurotransmitters, receptors, and types of neurons are involved in the circuitry of the PFC and thus ostensibly working memory, we can begin to see how cellular events, such as the NMDA-mediated release of dopamine onto pyramidal neurons, can modulate specific synaptic dynamics and lead to the sustained activity that is the hallmark of working memory. Can we make the causative association between the firing of neurons in the PFC—the presumed integrator of various incoming streams—and the behavioral output of the organism? Are there distinct maps in the PFC for spatial orientation, color, or other sensory data that are overlaid in interesting ways, or are these selective response fields ever-shifting, dependent on the incoming stream from the sensory system(s) at play?
Moving from correlation to causation will be the next major challenge for research into the most transient forms of memory we posses. The research has important implications for our understanding of internal representations, as information is moved from sensory inputs to higher cortical regions. If we can narrow down what is necessary and sufficient to explain a given representation of an external stimulus in the briefest of retained intervals, then we could be moving closer to understanding at least a piece of subjective, first-person experience.