Working Memory

The neural basis of working memory  

A complex cognitive process such as working memory is likely to be mediated by a distributed network of distinct brain regions (D’Esposito & Postle, 2015). However, initial findings from both animal and human studies identified the prefrontal cortex (PFC) as a critical node in the network supporting working memory. The very first fMRI study performed in our laboratory (D’Esposito et. al, 1995), and one of the first published on this function using fMRI, established a clear link between the PFC function and working memory. This study provided a foundation for research in our lab that was aimed at understanding this brain-behavior relationship by identifying the precise mechanisms in which representations held in working memory are maintained over short periods of time. Studies from awake behaving monkeys using recordings from single neurons within the lateral PFC had consistently found persistent, sustained levels of neuronal firing during the retention interval in tasks that require a monkey to retain information over a brief period of time. This sustained activity was thought to provide a bridge between the stimulus cue (e.g. the location of a flash of light) and its contingent response (e.g. a later delayed saccade to the remembered location). Using fMRI in healthy human volunteers, our initial fMRI studies were able to demonstrate that persistent PFC activity was also observed during the retention period of delay tasks (Zarahn et. al, 1997, 1999). Subsequently, the development of improved fMRI methods developed in our lab (D’Esposito et al, 1999; Postle et. al, 2000) allowed us to further demonstrate that the magnitude of PFC activity during the retention interval correlated positively with working memory accuracy (Curtis et al., 2004). This relationship suggested that the fidelity of the actively maintained representation is reflected in the delay period activity. Thus, the existence of persistent neural activity during retention intervals of delay tasks is a powerful empirical finding that lends strong support to the hypothesis that such activity represents a neural mechanism for the active maintenance or storage of task-relevant representations.

Functional organization of the prefrontal cortex

While there was strong support that lateral PFC was critical for working memory maintenance processes, it was unclear whether functional subdivisions within PFC existed. Pat Goldman-Rakic and colleagues put forth a proposal that different PFC regions were critical for the active maintenance of different types of information . Based on monkey electrophysiological and lesion studies, it was proposed that persistent activity within ventrolateral PFC reflected the temporary maintenance of non-spatial codes (such as an object’s color and shape), whereas dorsolateral PFC activity reflected the maintenance of spatial codes (such as the location of an object in space). Another possible axis along which human lateral PFC may be organized is according to the type of operations performed upon information being actively maintained rather than the type of information being maintained. For example, Michael Petrides proposed that ventrolateral PFC is the site where information is initially received from posterior cortical association areas and where active comparisons of maintained information are made. In contrast, dorsolateral PFC is recruited only when monitoring and manipulation of this information is required.

In numerous published studies, we tested these models of PFC organization (e.g., D’Esposito, Rypma & Postle, 2000, D'Esposito & Postle, 2002). Based on this work, we determined that PFC organization is likely a hybrid of these two models. That is, distinct PFC subregions can support the temporary storage of different types of information as well as implement different types of working memory processes (e.g., maintenance vs. manipulation). Thus, we put forth a more evolved formulation of this model that proposed that the rostro-caudal axis of the frontal lobes is characterized by a functional gradient whereby more anterior regions of the frontal lobe engage more complex or abstract control processes and more posterior regions engage in control lower in the action hierarchy and closer to the motor response (Badre & D’Esposito, 2007). Moreover, beyond these regional PFC differences in abstraction, we have further hypothesized that the frontal lobes are organized hierarchically (Badre & D'Esposito, 2009). Within such an architecture, there should be a dominance relationship whereby higher, more anterior regions influence processing in lower, more posterior regions more than vice versa. We have obtained evidence from studying patients with focal frontal lesions that such a hierarchy exists (Badre et al., 2009). That is, it was demonstrated that rostral PFC lesions affect lower-order processing but disruption of caudal PFC regions does not affect higher-order processing.

How does the PFC interact with other brain regions to support working memory?

Neuroimaging data often provides information about the localization of function where a particular cognitive function, such as a component of working memory, is ascribed to a particular brain region. However, the implementation of discrete cognitive functions is almost surely distributed across many nodes in a network. Importantly, fMRI has the unique ability to simultaneously image multiple regions of the brain. Thus, fMRI data can characterize interactions between the nodes in neural networks, such as the network that supports working memory. In our lab, we have developed and validated several multivariate statistical techniques for fMRI data in order to test network hypotheses (e.g. Kayser et al., 2009; Rissman et al., 2004; Sun et al., 2005). For example, we have published several studies that explored the functional connectivity between the PFC and other brain regions during working memory (Gazzaley et al., 2004; Fiebach et al., 2006). Consistently, we have found that the PFC interacts with a network of brain regions during the retention interval of a working memory task. These data provide support for the notion that a plausible mechanism for active maintenance is the coupling of abstracted, higher-order information in the PFC and stimuli-specific sensory information in the posterior association cortex through reverberant activity between these areas. In this manner, areas of multimodal association cortex such as PFC and parietal cortex, are in a position to integrate representations through connectivity to unimodal association cortex (e.g. temporal and occipital cortex). Another study revealed the importance of these functional interactions by showing that when an individual is distracted while trying to remember task-relevant information, functional connectivity in the working memory network is disrupted (Yoon et al., 2005). In summary, goal-directed behavior, which is both intentional and flexible, requires the temporary maintenance of a broad range of perceptual, mnemonic, and motor representations through coordinated network interactions.

Top-down control: how we remember what is relevant and ignore what is not

As we and others have demonstrated, in addition to recent sensory information, the PFC maintains the highest level of representations such as rules, goals, and intentions. Executive control can stem from the active maintenance of these PFC representations. Control processes (also called top-down processes) are those that guide behavior based on internal states such as knowledge from previous experience, expectations, and goals. Without such control, behavior is guided by bottom-up processes or those that are determined by the nature of sensory input. Humans with a faulty cognitive control system, such as those with PFC damage are ‘stimulus-driven’, responding in a habitual, almost reflexive manner, to events in their environment. Thus, the PFC may exert “control” in that the information it represents can bias unimodal association cortex in order to keep neural representations of behaviorally relevant sensory information activated when they are no longer present in the external environment (Miller & D'Esposito 2005). Thus, neural activity throughout the brain that is generated by input from the outside world, may be differentially enhanced or suppressed, presumably from top-down signals emanating from integrative brain regions such as PFC, based on the context of the situation. Thus, in this formulation, the “processing” component of working memory is that the “control” of actively maintained representations within primary and unimodal association cortex stems from the representational power of multimodal association cortex such as the PFC. If the PFC, for example, stores the rules and goals, then activation of such PFC representations will be necessary when behavior must be guided by internal states or intentions.

We have directly studied the neural mechanisms underlying top-down modulation by investigating the processes involved when individuals are instructed to remember relevant and ignore irrelevant information (Ranganath et al., 2004a; Gazzaley et al. 2005). For example, we have studied healthy young individuals while they performed a delay task during which on each trial they observed sequences of two faces and two natural scenes presented in a randomized order. However, during different time periods they were given different instructions informing them how to process the stimuli: 1) Remember Faces and Ignore Scenes, 2) Remember Scenes and Ignore Faces, or 3) Passively View faces and scenes without attempting to remember them. This task was performed with both fMRI and event related potential (ERP) recording. This allowed us to capitalize on the high spatial resolution of fMRI and the high temporal resolution of ERP. Our fMRI and ERP data revealed top-down modulation of both activity magnitude and processing speed in visual association cortical areas that process faces and scenes. For example, during the encoding period of the delay task, activity in the face area was enhanced, and occurred earlier, when faces had to be remembered as compared to when they were passively viewed. Likewise, face area activity was suppressed, and occurred later, when faces had to be ignored (with scenes now being retained instead across the delay interval) compared to a condition when they were passively viewed. Thus, there appears to be at least two types of top-down signals, one that serves to enhance task-relevant information, and the other that serves to suppress task-irrelevant information. It is well documented that the nervous system utilizes interleaved inhibitory and excitatory mechanisms throughout the neuroaxis (e.g., spinal reflexes, cerebellar outputs and basal ganglia movement control networks). Thus, it may not be surprising that enhancement and suppression mechanisms may exist to control memory function. By generating contrast via both enhancements and suppressions of activity magnitude and processing speed, top-down signals bias the likelihood of successful representation of relevant information in a competitive system.

Though it has been proposed that the PFC provides a major source of the types of top-down signals that we have described, this hypothesis largely originates from suggestive findings rather than direct empirical evidence.  We have obtained direct evidence that the PFC is the site of these top-down signals by using two different approaches. First, by performing fMRI studies in healthy individuals after perturbation of prefrontal function with TMS, and second by scanning patients with focal PFC lesions. In these studies. (Miller et al., 2011; Lee & D'Esposito, 2013), we investigated the effect of disrupting prefrontal cortex function on the selectivity of category representations of faces or scenes in the temporal cortex. Different object categories, such as faces and scenes, are represented by spatially distributed yet overlapping areas in the extrastriate visual cortex and can be easily identified with fMRI. We hypothesized that there would beless selectivity to faces and scenes after disruption of prefrontal cortical function with TMS in healthy individuals or in patients with focal frontal lesions. This is precisely what we found. The face and scene areas were less selective to their corresponding category after TMS in healthy subjects, and in the patients with unilateral focal frontal lesions, the face and scene areas in the same hemisphere as the frontal lesion were less selective to their corresponding category than in the other hemisphere. Moreover, subjects, where there was the greatest reduction in category tuning following frontal TMS exhibited the greatest working memory deficit. Together, this causal evidence clearly supports the notion that the lateral prefrontal cortex is one source of top-down feedback signals that act via both gain and selectivity mechanisms.

The relationship between working memory and long-term memory

The dominant view in psychology and neuroscience has been that working memory and long-term memory processes are mediated by distinct memory systems. Consequently, researchers have investigated these forms of memory in isolation, focusing on the role of medial temporal regions (MTL) in long-term memory and PFC in working memory. Research from our laboratory has challenged this idea in several ways. First, we have demonstrated that brain regions involved in storing perceptual representations of different object categories (e.g., faces or houses) are active even when people are maintaining a mental image of an object in working memory (Druzgal et al., 2003, Ranganath et al., 2004a). These findings suggest that working memory relies on activation of stored long-term memory representations in posterior cortex, in contrast to the view that these forms of memory are supported by structurally distinct systems. This neural data is consistent with a cognitive model put forth by Cowan that proposes that the contents of working memory are not maintained within dedicated storage buffers, but rather are simply the subset of information that is within the focus of attention at a given time. Thus, working memory arises from hierarchically arranged faculties comprising long-term memory, the subset of working long-term-memory that is currently activated and the subset of activated memory that is the focus of attention. From a neuroscience perspective, it is counterintuitive that all temporarily stored information during goal-directed behavior requires specialized, dedicated buffers. Clearly, there could not be a sufficient number of independent buffers to accommodate the infinite types of information that need to be actively maintained to accommodate all potential or intended actions.

Second, we have shown that MTL regions—traditionally associated exclusively in long-term memory such as the hippocampus—in fact contribute to working memory maintenance (Ranganath & D’Esposito, 2004b). This study sparked significant controversy, but the findings were subsequently replicated in at least four other labs. Of course, we are well aware that neuroimaging studies are correlative and cannot demonstrate that the MTL is necessary for any transient form of memory. However, contrary to popular belief, studies of amnesic patients are consistent with the idea that the MTL are necessary for some forms of working memory maintenance. These assertions prompted new research by several labs (e.g. Neal Cohen and colleagues), each of whom demonstrated that under some circumstances, the MTL may be necessary for normal short-term retention.

Thus, working memory can be considered an emergent property of brain systems that have evolved to accomplish sensory-, representation-, and/or- action-related functions (D’Esposito, 2007). That is, there are no special-purpose working memory systems or modules. Rather, the sustained retention of information is a general property of the brain issuing from attentional mechanisms that are also important for many non-memory behaviors. This framework has implications that challenge currently prevailing views of working memory at both cognitive/theoretical and neurobiological levels of analysis. For example, at the theoretical level, it challenges the idea that working memory is supported by a neural system that has evolved to perform this special function (in the way that, e.g., the visual system is specialized for visual perception).

Key References:

Badre D, D’Esposito M. FMRI evidence for a hierarchical organization of the prefrontal cortex, Journal of Cognitive Neuroscience, 19:2082-2099, 2007.

Badre D, D’Esposito M.  Is the rostro-caudal axis of the frontal lobe hierarchical? Nature Reviews Neuroscience, 10:659-69, 2009.

Badre D, Hoffman J, Cooney JW, D’Esposito M.  Hierarchical cognitive control deficits following damage to the human frontal lobe, Nature Neuroscience, 12:515-22, 2009.

Curtis CC, Rao VY, D’Esposito M. Maintenance of spatial and motor codes during oculomotor delayed response tasks, Journal of Neuroscience, 24:3944-52, 2004.

D’Esposito M, Detre AJ, Alsop DC, Shin RK, Atlas S, Grossman M.  The neural basis of the central executive system of working memory, Nature, 378:279-281, 1995.

D’Esposito M, Zarahn E, Aguirre G. Event-related fMRI: implications for cognitive psychology, Psychological Bulletin, 125:155-64, 1999

D’Esposito M, Postle BR, Rypma B. Prefrontal cortical contributions to working memory: Evidence from event-related fMRI studies, Experimental Brain Research, 133:3-11, 2000.

.D’Esposito M, Postle BR. The organization of working memory function in lateral prefrontal cortex: evidence from event-related functional MRI. In: DT Stuss & RT Knight (Eds.) Principles of Frontal Lobe Function. Oxford University Press, New York, 2002.

D'Esposito M. From cognitive to neural models of working memory, Phil. Trans. Biol. Sc., 362: 761-72, 2007.

D’Esposito M, Postle BR. The cognitive neuroscience of working memory, Annual Reviews of Psychology, 66:115-42, 2015.

Druzgal TJ, D’Esposito M.  Dissecting contributions of prefrontal cortex and fusiform face area to face working memory, Journal of Cognitive Neuroscience, 15:771-84, 2003.

Fiebach CJ, Jesse Rissman J, D’Esposito M. Modulation of the inferotemporal cortex activation during verbal working memory maintenance, Neuron, 51:251-61, 2006.

Gazzaley A, Rissman J, D’Esposito M. Functional connectivity during working memory maintenance, Cognitive, Affective and Behavioral Neuroscience, 4: 580-99, 2004.

Gazzaley A, Cooney J, McEvoy K, Knight RT, D’Esposito M. Top-down modulation of visual processing: converging fMRI and ERP evidence, Journal of Cognitive Neuroscience, 17:507-17, 2005.

Kayser AS, Sun FT, D’Esposito M, A comparison of Granger causality with other multivariate techniques in fMRI analysis of the motor system, Human Brain Mapping, 30:3475-94, 2009.

Lee TG, D’Esposito M. The dynamic nature of top-down signals originating from prefrontal cortex: A combined fMRI-TMS study, Journal of Neuroscience, 32:15458-66, 2012.

Miller BT, D’Esposito M. Searching for the “top” in top-down control?” Neuron, 48:535-8, 2005.

Miller BT, Fegen D, Vytlacil J, Pradhan S, D’Esposito M. The prefrontal cortex modulates category selectivity in human extrastriate cortex, Journal of Cognitive Neuroscience, 23:1-10, 2011.

Postle BR, Zarahn E, D’Esposito M. Using event-related fMRI to assess delay-period activity during performance of spatial and nonspatial working memory tasks, Brain Research Protocols, 5:57-66, 2000.

Ranganath C, Degutis J, D’Esposito M. Category-specific modulation of inferior temporal activity during working memory maintenance, Cognitive Brain Research, 20:37-35, 2004a.

Ranganath C, Cohen MX, Dam C, D’Esposito M. Inferior temporal, prefrontal, and hippocampal contributions to visual working memory maintenance and associative memory retrieval, Journal of Neuroscience, 24:3917-25, 2004b

Rissman J, Gazzaley A, D’Esposito M. Measuring functional connectivity during distinct stages of cognitive task, NeuroImage, 23:752-63, 2004.

Sun FT, Miller LM, D’Esposito M. Measuring temporal dynamics of functional networks using phase spectrum of fMRI data. NeuroImage, 28:227-37, 2005.

Yoon JH, Curtis CE, D’Esposito M. Differential effects of distraction during working memory on delay-period activity in the prefrontal cortex and the visual association cortex, NeuroImage, 29:1117-26, 2005

Zarahn E, Aguirre G, D’Esposito M.  A trial-based experimental design for functional MRI, NeuroImage, 6:122-138, 1997.

Zarahn E, Aguirre G, D’Esposito M.  Temporal isolation of the neural correlates of spatial mnemonic processing with functional MRI, Cognitive Brain Research, 7:255-268, 1999.