Information that has been recently perceived or remembered can bias current processing. This has been viewed as both a corrupting (e.g., proactive interference in short-term memory) and stabilizing (e.g., serial dependence in perception) phenomenon. We hypothesize that this bias is a generally adaptive aspect of brain function that leads to occasionally maladaptive outcomes.

Recent experiments have shown that visual cognition blends current input with that from the recent past to guide ongoing decision making. This serial dependence appears to exploit the temporal autocorrelation normally present in visual scenes to promote perceptual stability. While this benefit has been assumed, evidence that serial dependence directly alters stimulus perception has been limited. In the present study, we parametrically vary the delay between stimulus and response in a spatial delayed response task to explore the trajectory of serial dependence from the moment of perception into post-perceptual visual working memory. We find that behavioral responses made immediately after viewing a stimulus show evidence of adaptation, but not attractive serial dependence. Only as the memory period lengthens is a blending of past and present information apparent in behavior, reaching its maximum with a delay of six seconds. These results dovetail with other recent findings to bolster the interpretation that serial dependence is a phenomenon of mnemonic rather than perceptual processes. However, even while this pattern of effects in group-averaged data has now been found consistently, we show that the relative strengths of adaptation and serial dependence vary widely across individuals. Finally, we demonstrate that when leading mathematical models of working memory are adjusted to account for these trial-history effects, their fit to behavioral data is substantially improved.

Recent experiments have shown that visual cognition blends current visual input with that from the recent past to guide ongoing decision making. This serial dependence is tuned to the similarity between consecutive stimuli and appears to exploit the temporal autocorrelation normally present in visual scenes to promote perceptual stability. While these benefits have been assumed, evidence that serial dependence directly alters stimulus perception has been limited. In the present study, we parametrically vary the delay between stimulus and response in a spatial delayed response task to explore the trajectory of serial dependence from the moment of perception into post-perceptual visual working memory. We find that behavioral responses made immediately after viewing a stimulus show evidence of adaptation, but not attractive serial dependence. Only as the memory period lengthens is a blending of past and present information apparent in behavior, reaching its maximum with a memory delay of six seconds. These results dovetail with other recent findings to bolster the interpretation that serial dependence is a phenomenon of mnemonic rather than perceptual processes. We also demonstrate that when leading mathematical models of visual working memory are adjusted to account for this trial-history effect, their fit to behavioral data is substantially improved.

Dopamine is required for working memory, but how it modulates the large-scale cortex is unknown. Here, we report that dopamine receptor density per neuron, measured by autoradiography, displays a macroscopic gradient along the macaque cortical hierarchy. This gradient is incorporated in a connectome-based large-scale cortex model endowed with multiple neuron types. The model captures an inverted U-shaped dependence of working memory on dopamine and spatial patterns of persistent activity observed in over 90 experimental studies. Moreover, we show that dopamine is crucial for filtering out irrelevant stimuli by enhancing inhibition from dendrite-targeting interneurons. Our model revealed that an activity-silent memory trace can be realized by facilitation of inter-areal connections and that adjusting cortical dopamine induces a switch from this internal memory state to distributed persistent activity. Our work represents a cross-level understanding from molecules and cell types to recurrent circuit dynamics underlying a core cognitive function distributed across the primate cortex.

Recent work has established that visual working memory is subject to serial dependence: current information in memory blends with that from the recent past as a function of their similarity. This tuned temporal smoothing likely promotes the stability of memory in the face of noise and occlusion. Serial dependence accumulates over several seconds in memory and deteriorates with increased separation between trials. While this phenomenon has been extensively characterized in behavior, its neural mechanism is unknown. In the present study, we investigate the circuit-level origins of serial dependence in a biophysical model of cortex. We explore two distinct kinds of mechanisms: stable persistent activity during the memory delay period and dynamic “activity-silent” synaptic plasticity. We find that networks endowed with both strong reverberation to support persistent activity and dynamic synapses can closely reproduce behavioral serial dependence. Specifically, elevated activity drives synaptic augmentation, which biases activity on the subsequent trial, giving rise to a spatiotemporally tuned shift in the population response. Our hybrid neural model is a theoretical advance beyond abstract mathematical characterizations, offers testable hypotheses for physiological research, and demonstrates the power of biological insights to provide a quantitative explanation of human behavior.

SummaryDopamine is critical for working memory. However, its effects throughout the large-scale primate cortex are poorly understood. Here we report that dopamine receptor density per neuron, measured by receptor autoradiography in the macaque monkey cortex, displays a macroscopic gradient along the cortical hierarchy. We developed a connectome- and biophysically-based model for distributed working memory that incorporates multiple neuron types and a dopamine gradient. The model captures an inverted U-shaped dependence of working memory on dopamine. The spatial distribution of mnemonic persistent activity matches that observed in over 90 experimental studies. We show that dopamine filters out irrelevant stimuli by enhancing inhibition of pyramidal cell dendrites. The level of cortical dopamine can also determine whether memory encoding is through persistent activity or an internal synaptic state. Taken together, our work represents a cross-level understanding that links molecules, cell types, recurrent circuit dynamics and a core cognitive function distributed across the cortex.

In contrast to feedforward architecture commonly used in deep networks at the core of today's AI revolution, the biological cortex is endowed with an abundance of feedback projections. Feedback signaling is often difficult to differentially identify, and its computational roles remain poorly understood. Here, we investigated a cognitive phenomenon, called categorical perception (CP), that reveals the influences of high-level category learning on low-level feature-based perception, as a putative signature of top-down signaling. By examining behavioral data from a visual motion delayed matching experiment in non-human primates, we found that, after categorization training, motion directions closer to (respectively, away from) a category center became more (less) difficult to discriminate. This distance-dependent discrimination performance change along the dimension relevant to the learned categories provides direct evidence for the CP phenomenon.To explain this experimental finding, we developed a neural circuit model that incorporated key neurophysiological findings in visual categorization, working memory and decision making. Our model accounts for the behavioral data indicative of CP, pinpoints its circuit basis, suggests novel experimentally testable predictions and provides a functional explanation for its existence. Our work shows that delayed matching paradigms in non-human primates combined with biologicallybased modeling can serve as a promising model system for elucidating the neural mechanisms of CP, as a manifestation of top-down signaling in the cortex. 2 Significant StatementCategorical perception is a cognitive phenomenon revealing the influences of high-level category learning on low-level feature-based perception. However, its underlying neural mechanisms are largely unknown. Here, we found behavioral evidence for this phenomenon from a visual motion delayed matching experiment in non-human primates. We developed a neural circuit model that can account for this behavioral data, pinpoints its circuit basis, suggests novel experimentally testable predictions and provides a functional explanation for its existence. Our work shows that delayed matching paradigms in non-human primates combined with biologically-based modeling can serve as a promising model system for elucidating the neural mechanisms of categorical perception, as a manifestation of top-down signaling in the cortex.3

A growing body of evidence suggests that conscious perception of a sensory stimulus triggers an all-or-none activity across multiple cortical areas, a phenomenon called 'ignition'. In contrast, the same stimulus, when undetected, induces only transient activity. In this work, we report a large-scale model of the macaque cortex based on recently quantified structural connectome data. We use this model to simulate a detection task, and demonstrate how a dynamical bifurcation mechanism produces ignition-like events in the model network. Within this framework, the model predicts that feedforward excitatory transmission is primarily mediated by the fast AMPA receptors to ensure rapid signal propagation from sensory to associative areas. In contrast, a large fraction of the inter-areal feedback projections and local recurrent excitation depend on the slow NMDA receptors, to ensure ignition of distributed frontoparietal activity. Our model predicts, counterintuitively, that fast-responding sensory areas contain a higher ratio of NMDA to AMPA receptors compared to association cortical areas that show slow, sustained activity. We validate this prediction using in-vitro receptor autoradiography data. Finally, we show how this model can account for various behavioral and physiological effects linked to consciousness. Together, these findings clarify the neurophysiological mechanisms of conscious access in the primate cortex and support the concept that gradients of receptor densities along the cortical hierarchy contribute to distributed cognitive functions.