Faculty » Weiwei Zhang

The research program in my lab integrates behavioral, computational, cognitive neuroscience, and translational approaches to address fundamental questions of theoretical importance about perception, attention, and memory in healthy subjects and in various clinical populations.  Most of our basic science experiments focus on the nature of visual memory representations and the encoding and forgetting processes involved in both working memory and long-term memory. The lab employs state of the art techniques, including 64-channel and 128-channel Biosemi electroencephalography (EEG) systems and an Eyelink 1000 eye tracking system.

Capacity & Resolution in Visual Working Memory

Although the human brain has a vast store of long-term memories (some memories for a lifetime), it also maintains temporary, disposable, scratch-pad memories (working memory) that last only a few seconds and are essential for performing tasks such as adding two numbers or comparing the attractiveness of two faces.  Working memory plays a key role in virtually all real-time cognitive tasks, and variations in working memory capacity are central to both normal and pathological variations in cognitive abilities.  Miller’s seminar work (7±2) led to a flurry of research on working memory as being one of the major limiting factors in human cognition. However, no previous research has convincingly addressed the most fundamental question about visual working memory, namely, whether there is a small set of discrete representations (like a fixed number of photos that can be taken from a roll of film) or a potentially infinite set of representations that vary in resolution depending on the number of objects being represented (like a large number of low-resolution digital photographs vs. a small number of high-resolution digital photographs stored on a flash memory card). We have recently developed a new set of experimental and analytical methods that finally allow this issue to be settled. We have demonstrated that human observers, when presented with more than a few simple objects, store a subset of objects with good precision and retained no information about the others (Zhang & Luck, 2008, Nature). In addition, observers could not increase the number of representations by decreasing the quality of the representations in working memory, even though they could make such a trade-off in iconic memory (Zhang & Luck, 2011, Psych Sci). This line of research combines two fundamental approaches to memory research [see Koriat et al., 2000, Annual Review of Psychology]: a quantity-oriented approach emphasizing the storage capacity of memory as well as an accuracy-orientated approach focusing on the level of correspondence between a memory representation and the physical stimuli being represented. 

Beyond item-based representations

This mixture model effectively describes capacity limitations, but its application may be better suited to study explicit and discrete item-based representations (e.g., isolated objects, lists of words, etc.) than summary representations of visual scenes (e.g., “ensemble” or “gist”). We have recently started to apply some theoretical constructs and quantitative methods (e.g., Dual Process Signal Detection model) from long-term memory research to working memory in order to separate out and quantitatively assess global gist and individual item representations using analyses of receiver operating characteristic (ROC) curves (Zhang & Yonelinas, 2011, Psychonomics). We found manipulations of configural encoding selectively affected familiarity (reflected in ROC curvature), whereas manipulations of item information selectively affected recollection (reflected in the y-intercept). This line of research not only provides a quantitative assessment of global gist, but also attributes capacity-limited and capacity-unlimited aspects of working memory to discrete item-based recollection and continuous gist-based familiarity, respectively.

Processes in Visual Working Memory

General Douglas MacArthur famously remarked that “Old soldiers never die; they just fade away” (General Douglas MacArthur, April 19, 1951). For decades, researchers have concluded that working memories, like old soldiers, simply fade away as they age, becoming progressively less precise as they are retained for longer periods of time.  However, we anticipated some sudden termination of working memories (hence “sudden death”) would occur because it is a natural consequence of the maintenance of information in an active, feedback-driven system.  This is analogous to a computer system that gradually overheats with no obvious performance degradation and then suddenly shuts down. In line with this prediction, we recently found that visual working memories suffered no significant loss of strength or precision over a short period of time. After about four seconds, some memories suddenly terminated in a probabilistic manner, disappearing without a trace (Zhang & Luck, 2019, Psych Sci). An all-or-none working memory forgetting may sound counter-intuitive and inconsistent with phenomenological awareness that visual analog world fades away gradually, like old film photos. However, the digital human mind may make us more flexible and adaptive in everyday life in that it avoids overloading catastrophe and makes room in limited-capacity working memory so that new information can be dynamically and promptly encoded and processed. Interestingly, the process of transforming fragile perceptual representations into durable working memory representations is also a discrete process in which the representation is lost unless it has reached a particular level of activation in working memory (Zhang & Luck, 2008, Nature).

Feature-based Attention

For decades most theories of attention (and attention researchers) have assumed that attention operates by focusing on spatial locations, with nonspatial features such as color being used solely to guide attention to potentially relevant locations. The most compelling and direct evidence that spatial attention has a special status derives from ERP studies, in which spatial attention influences the amplitude of the sensory evoked response, beginning with the P1 wave (< 100 ms poststimulus).  Such modulations of P1 amplitude have become the “gold standard” for demonstrating that attention modulates feedforward sensory activity in the visual system.  No early P1 modulations have been previously observed for nonspatial attention, which has always been found to operate after 100 ms. Moreover, the effects of nonspatial attention have been found only for stimuli presented at attended locations, with no effect for stimuli presented at ignored locations.  This has led most researchers to conclude that nonspatial attention operates after, and is contingent upon, spatial attention. However, we have recently provided the first unambiguous demonstration that feature-based attention influences the initial feedforward wave of sensory processing in the extrastriate cortex, just like spatial attention. We found feature-based attention can influence feedforward sensory processing as early as spatial attention can (within 100 ms of stimulus onset), but only under conditions of strong competition between attended and unattended feature values (Zhang & Luck, 2009, Nature Neuroscience).

Translational & Applied Research

Working memory for simple visual features is tightly linked to a broad set of issues in psychology, ranging from oculomotor control to fluid intelligence and psychopathology. For example, the capacity of visual working memory is strongly associated with individual differences in intelligence, and it is disrupted in conditions such as schizophrenia (Gold et al., 2010, Archives of General Psychiatry). In contrast, medial temporal lobe (MTL) patients with focal lesions in hippocampus exhibit selective impairments in the precision but not the capacity of visual working memory. Some of our ongoing research makes it possible to provide quantitative evaluations of the working memory deficits in various psychiatric and neurological conditions, including schizophrenia, ADHD, depression, dementia, and cholinergic deficits. Furthermore, working memory is important for education and learning because of the various functional roles of working memory in higher cognition, such as the tight coupling between working memory capacity and fluid intelligence. In addition, the precision of working memory representation is strongly correlated with creativity. Translational and applied research like these has been and will continue to be integrated with the basic science research in the lab.

Selected Publications

Zhang, W., & Luck, S. J. (2011). The Number and Quality of Representations in Working Memory. Psychological Science, 22(11), 1434 - 1441.

Gold, J. M., Hahn, B., Zhang, W., Robinson, B. M., Kappenman, E. S., Beck, V. M., et al. (2010). Reduced capacity but spared precision and maintenance of working memory representations in schizophrenia. Archives of General Psychiatry. 67(6), 570-577.

Zhang, W., & Luck, S. J. (2009). Sudden Death and Gradual Decay in Visual Working Memory. Psychological Science, 20(4), 423-428.

Zhang, W., & Luck, S. J. (2009). Feature-Based Attention Modulates Feedforward Visual Processing. Nature Neuroscience, 12(1):24-25.

Zhang, W., & Luck, S. J. (2008). Discrete Fixed-Resolution Representations in Visual Working Memory. Nature, 453(7192), 233-235.