Progress in Neurobiology

By definition, episodic memory is a conscious phenomenon. Memory traces reactivated by the hippocampus and reinstated in the sensory cortices need to enter conscious awareness for them to be re-experienced and overtly recalled. However, it remains unclear whether such reactivation in-and-of-itself ensures that memories will be overtly recalled. To investigate this, magnetoencephalography (MEG) recordings were analysed from thirty-one participants (18 female, 13 male) completing a video-word pair associates memory task. When combining linear classifiers and spectral analyses, sensory cortical reactivation could be observed without overt recall occurring, suggesting reactivation does not guarantee overt recall. Instead, overt recall was additively predicted by (i) an increase in reactivated representations rhythmically fluctuating within the alpha band, and (ii) a decrease in total sensory neocortical alpha power. These results are consistent with accounts which propose that reactivation benefits from desynchronising the network to provide representational space for stimulus-specific information, and/or amplifying stimulus-specific information above residual noise. Altogether, these results suggest that representational reactivation can occur without overt recall, and suggest a role for alpha oscillations in projecting internally-generated representations into conscious awareness.

We studied the effects of prioritization in a two-step retrocuing task in which participants hold two items in working memory, and the item not cued by the first cue cannot be dropped because it may be prioritized by the second cue. In Experiment 1, using a dense sampling procedure, we observed that recall performance oscillated at 15 Hz in the prioritization task, in comparison to 20 Hz in a matched task employing a neutral cue. Furthermore, the prioritized item was shielded from bias exerted by the uncued item, as well as from items from the previous trial. In Experiment 2, we recorded the EEG while participants performed variants of the two tasks. The prioritization cue uniquely triggered a phase reset at 15 Hz and an increase in oscillatory peaks at this frequency. Burst analysis ruled out bursting as a possible underlying factor. Time-resolved representational similarity analysis (RSA) revealed that the prioritization cue triggered representational transformations that were larger for the uncued item. The shielding effects of prioritization may arise from the transformation of the not-prioritized item into an "unprioritized" state that is implemented and maintained by a mechanism that cycles at 15 Hz.

The hippocampus (HPC), prefrontal cortex (PFC), and thalamic nuclei, such as reuniens (Re), form a reciprocally connected circuit that plays a critical role in processing hippocampus-dependent memory. Accumulating evidence suggests that this triangular modular circuit is crucial for performing cognitive tasks that require context-dependent memory, which belong to a class of behavioral tasks difficult for animals to learn. Experiments are gradually revealing what behavioral information these brain regions represent, but how the triangular circuit gives rise to the observed divisions of labor remains unknown. It is also unclear whether the triangular modular circuit brings any advantage in solving such tasks. Here, we addressed these questions by constructing a prefrontal-thalamo-hippocampal circuit model comprising interconnected long-short-term memory (LSTM) units and training it on contextual memory-dependent spatial navigation tasks. Our model revealed the critical roles of the distinct brain modules. The HPC module encoded spatial information, whereas the PFC module represented the spatiotemporal task structure in a context-dependent manner. The Re module integrated task-relevant information to facilitate learning in the PFC and HPC modules, dynamically harmonizing these modules. The thalamic coordination of the other modules enhanced the systems robustness in learning to navigate complex environments. This division of labor between the HPC, PFC, and Re modules was not specified a priori but emerged through learning, showing an interesting coincidence with the task-related activities of the prefrontal-thalamo-hippocampal circuit. Our results demonstrate that the multi-modular network structure is crucial for robust processing of context-dependent memory.

Beta-band activity (13-30 Hz) is a hallmark of human movement, yet a unifying account of its functional role remains unresolved. Although typically described as a sustained oscillation, beta activity is increasingly recognised to consist of transient bursts. More recently, beta bursts have been shown to exhibit heterogeneous waveforms. Here, we ask whether variability in burst shape corresponds to separable computational roles during motor adaptation. Using high-density MEG, we recorded neural activity while participants performed a visuomotor rotation task under either implicit (sensorimotor adaptation) or explicit (strategic re-aiming) learning conditions. Conventional metrics, beta power and burst rate, showed context-dependent modulation during preparation but provided limited insight into trial-by-trial behaviour. In contrast, sorting bursts according to their waveforms revealed a repertoire of burst types with dissociable temporal dynamics and context-dependent modulation. Crucially, during post-movement evaluation, distinct burst subtypes showed opposing and temporally specific relationships with behavioural error: one subtype decreased with increasing error, whereas others increased. Together, these findings indicate that beta activity comprises separable transient events with specific computational roles, and that accounting for waveform diversity is essential for understanding how cortical beta supports adaptive behaviour.

The macaque lateral intraparietal area (LIP) is known for its role in visually guided saccades as well as in higher cognitive functions. However, its contribution to more basic visuomotor processes remains unclear. Here, we investigated whether neural activity in area LIP is also related to involuntary reflexive eye movements, specifically the fast phases of optokinetic nystagmus (OKN). To address this question, we compared spiking activity and local field potentials (LFPs) in area LIP of two male macaque monkeys during visually guided saccades and during kinematically similar OKN fast phases. Neurons exhibiting robust perisaccadic activation during voluntary saccades showed markedly reduced or no activity around the time of OKN fast phases and were not modulated by fast-phase amplitude or frequency. Using a Generalized Linear Model, we found that during OKN slow phases, area LIP reliably encoded gaze position and the direction of visual motion driving the reflexive eye movement. LFP analyses further revealed that beta-band power differed between voluntary saccades and OKN fast phases, whereas theta-band phase coherence increased following both types of fast eye movements, suggesting distinct local processing but shared post-movement network coordination. Our results reinforce the view of area LIP as a key area for integrating sensory and cognitive signals relevant for goal-directed action rather than a generic oculomotor controller.

The transformation of sensory information into goal-directed motor plans and actions is known to emerge from coordinated activity between parietal and motor areas. Within this network, anticipation plays a critical role, enabling the brain to predict upcoming sensory inputs and prepare appropriate actions before sensory information becomes available. However, it remains unclear whether fronto-parietal interactions during motor anticipation follow a serial hierarchical organization or reflect distributed and reciprocal processing. To address this open question, we trained two rhesus macaques to perform a visually-guided sequential reaching task, in which the predictability of target location increased within the sequence. Analysis of eye and hand movements revealed that the degree of movement anticipation increased with target predictability. The direction of the upcoming reach toward predictable targets could be decoded from preparatory neural activity prior to target onset in both dorsal premotor-primary motor cortex (PMd/M1) and parietal area 7A. Using feature-specific information transfer analysis, we found that information about the upcoming movement direction was transmitted between 7A and PMd/M1 through bidirectional, yet asymmetric, interactions. Contrary to classical hierarchical models predicting serial activation across parietal and motor areas, parietal-to-motor interactions did not occur earlier than motor-to-parietal interactions. Instead, our findings support a heterarchical and reciprocal fronto-parietal network in which anticipatory processes adjust the timing of preparatory activity to facilitate eye-hand coordination during reaches to predictable targets.

Sharp-wave ripples (SWRs) serve as a key network state in which the hippocampus replays spatiotemporal patterns that reflect animals experiences. However, how the hippocampal circuitry controls the generation, duration, and termination of these events remains unclear. To clarify how behaviorally salient locations are encoded into replay, we constructed a biologically plausible model of area CA3, including three classes of inhibitory neurons -- parvalbumin-positive basket cells, cholecystokinin-positive basket cells (CCKBC), and theta-OFF, ripple-ON cells (TORO). In our model, replay and SWRs emerge from TORO-mediated release of CCKBC inhibition onto pyramidal cells, as hypothesized by experiments. Furthermore, our model suggests that short-term depression at TORO-CCKBC synapses controls replay duration and identifies a mechanism through which replay overrepresents reward-associated locations. Our model predicts that replays terminating at these locations are shorter, which we confirm in experimental data. Overall, this study provides a mechanistic framework for SWRs and replay dynamics in the hippocampus.

Despite numerous investigations, a comprehensive electrophysiological characterization of iconic memory remains lacking. Through a partial report paradigm, we aimed to shed light on this topic by disentangling electrophysiological activity related to stimulus perception from that linked with the specific task. We collected EEG data from 26 participants while they performed a partial report task. They were shown circular arrays of six letters lasting 100 ms. After the stimulus, an acoustic cue instructed the participant to report on which side of the array. Differences between reporting conditions were primarily evident in the time window 850-1100 ms, characterized by a positive component predominantly over parieto-occipital electrodes ipsilateral to the reporting side. Through linear regression, we also found a positive relationship between P1 and participants accuracy, as well as negative relationships between P3, VCR, TIF, and accuracy. Our results provide an overview of the different processes involved in iconic memory, corroborating the distinction between a series of neural mechanisms responsible for encoding and maintaining the entire stimulus and higher-order processes in charge of selecting an information subset for conscious report. The TIF component, in particular, could act as a key filtering mechanism to prevent irrelevant information from being selected for further processing. Our results provide, for the first time, a thorough characterization of the electrophysiological dynamics behind iconic memory. Moreover, implications for the consciousness debate are discussed, particularly regarding the overflow argument and how our results could be read through its lens.

Despite decades of animal and human intracranial work highlighting the critical role theta oscillations (4-12 Hz) play in memory encoding and recall stage during navigation, the link between scalp recorded human theta oscillations and spatial encoding and recall is lacking. In the present study, we used right posterior theta (RPT) - a scalp-level theta signal believed to be generated in the medial temporal cortex - to examine spatial encoding and recall during virtual spatial navigation. In particular, we recorded EEG from 27 healthy subjects performing a novel virtual Linear Track Memory (LTM) task. During the encoding stage of the task, a reward cue was presented at one of five pillar locations along a linear track. During the recall stage, subjects were presented with images of the five pillars and five new pillars, and were asked to press a button when the rewarded target pillar location appeared. If correct, subjects received 5 cents for that trial. Memory performance was assessed using reaction time, d-prime (d), and response bias ({beta}), and RPT was measured following the onset of the reward cue at bilateral scalp electrodes P7 and P8. Consistent with previous work, RPT peaked approximately 170-300 ms over the right hemisphere (P8) after cue onset, which was significantly increased for reward cues during the encoding stage and for the target pillar during the recall stage. Importantly, general linear model regressions revealed that peak RPT power during the encoding stage significantly predicted higher d and {beta} scores during recall, supporting the relationship between RPT peak power and memory performance. Together, these findings support the proposal that RPT activity reflects the encoding of salient information for the purpose of spatial navigation and a promising candidate biomarker for memory-related functioning in health and disease (e.g., Alzheimers disease).

Selective attention allows us to prioritize or retrieve task-relevant features and items from working memory. However, previous work has largely relied on unisensory paradigms, leaving open the question of how attentional mechanisms act on audiovisual working memory representations. Here, using an EEG-based audiovisual delayed-match-to-sample task, we investigate whether attention to audiovisual working memory contents operates on the level of individual unisensory features or bound cross-modal objects. On each trial, participants memorized an audiovisual item. At test, they were randomly presented with either an auditory, visual, or an audiovisual probe stimulus and indicated whether the latter matched their working memory content. Compared with recalling the entire audiovisual object, recall of unisensory visual or auditory features resulted in poorer behavioral performance and elevated midfrontal theta power. Multivariate pattern analyses (MVPA) showed that task-irrelevant feature representations were retrieved in both auditory and visual probe trials, consistent with object-based retrieval. Moreover, these results shed light on the representational structure underlying cross-modal feature storage in working memory, suggesting that audiovisual features are stored as bound objects and that attentional selection of individual object features in one modality spreads to cross-modal object features in another modality. In task conditions where such incidental recall of task-irrelevant features is detrimental to task performance, this places greater demands on cognitive control mechanisms.

The profile of electrophysiological responses in the human hippocampus (HPC) during verbal memory processing has remained complex and unclear. Here, we studied 26 patients implanted with intracranial electrodes across 187 HPC sites (50% left, 2-18 per patient). During memory encoding and retrieval, a subset of HPC responsive sites demonstrated increased ripple events, along with elevated high-frequency (HFA >50 Hz), and low-frequency (LFA 1-8 Hz) activity. A nearly equal number of sites showed no changes in ripple rate but increased LFA power and a delayed response-locked decrease in HFA power. More importantly, both successful encoding as well as recognition of remembered words were strongly associated with the coordination of the timing of LFA and HFA increases across the two clusters of responsive HPC sites. Using direct cortical electrical stimulations, we confirmed overlapping, but partially distinct, cortical connections to the functionally distinct HPC clusters. Our findings suggest a mesoscale mosaic functional organization within the human HPC where adjacent sites with divergent electrophysiological responses may have specialized roles during verbal memory processing. More importantly, our findings suggest that successful human memory depends on the coordination of the timing of low and high frequency local fields generated across these functionally divergent neuronal population sites.

The key elements for fear extinction learning are unexpected omissions of expected aversive events, which are considered to be rewarding. Given its reception of reward information, we tested the hypothesis that the cerebellum contributes to reward-like prediction error processing driving extinction learning via its connections with the ventral tegmental area (VTA). Forty-three young and healthy participants performed a three-day fear conditioning paradigm in a 7T MR scanner. The cerebellum and VTA were active during unexpected omissions of aversive unconditioned stimuli in the initial extinction trials and in other learning phases, in line with the proposed role of prediction-error processing. Increased functional connectivity was observed between the cerebellum and VTA, indicating that they are functionally coupled during fear extinction learning. These results suggest that an interaction between the cerebellum and VTA should be incorporated into the existing model of the fear extinction network.

The neural origin of bottom-up saliency for exogenous attention remains highly controversial. In this study, we investigated whether the earliest activity in the primary visual cortex (V1) encodes saliency signals defined by the eye-of-origin and feature-conjunction information. Electroencephalography (EEG) recordings from the human occipital cortex revealed early responses to eye-of-origin (E) and/or orientation (O) singletons, with larger response amplitudes to the double-feature (EO) singletons. The short onset latency (58-70 ms) and polarity reversal of the responses indicate a V1 origin. Importantly, the latency and amplitude of these responses predicted behavioral detection performance. Together, these findings suggest that the timing and amplitude of the earliest signals in V1 represent the saliency of combined feature contrasts for bottom-up attention. These signals unlikely originate from projections of other proposed source areas of saliency, due to the scarcity of necessary monocular neurons to process eye-of-origin information. HighlightsO_LIEye-of-origin information, invisible to the SC, elicits an early saliency signal in V1 within 50-100 ms. C_LIO_LICombined feature contrast enhances V1 saliency responses in a nonlinear fashion. C_LIO_LIThe latency and amplitude of V1 saliency responses predict behavioral detection performance. C_LI

Motor learning alters activities in multiple brain areas. While learning-induced activity change in these areas has been investigated, how the information flow changes in the network across those areas remains to be thoroughly examined. We analysed wide-field calcium imaging data spanning from the premotor cortex to the parietal cortex of marmosets while they learned a two-target forelimb-reaching task. We applied non-negative matrix factorization (NMF) to the activity data and extracted about 30 localized activity components. Encoding model analysis indicated that learning was associated with a decrease in activity components related to hand movements, and an increase in those related to external and reward signals. Causality analysis by embedding entropy (EE) revealed increases in causal links across activity components in different areas and stabilization of the network structure with behavioural improvements. These results indicate that motor learning entails both a redistribution of task-related activity and a reorganization of large-scale cortical network interactions.

Neurons couple to various degrees to the activity level of the local neighboring population whereby strongly coupled choristers and weakly coupled soloists have been identified as two extremes of a continuous spectrum. At the same time neuronal populations undergo coordinated ON and OFF cortical state activity fluctuations, which are locally modulated by attention. The population coupling of soloists and choristers suggests that soloists should show limited alignment with cortical state fluctuations, while choristers should exhibit profound alignment. To test this, we recorded neurons across cortical layers in macaque areas V1 and V4, while animals performed a feature based spatial attention task. As expected, we found a wide range of population coupling strength of neurons. In line with our prediction, coupling of choristers to cortical state changes (ON-OFF transitions) was generally stronger than that of soloists. The strength of population coupling of neurons was similar during spontaneous and stimulus driven activity. Allocation of attention to the receptive field reduced the population coupling strength. Attentional modulation of neurons was positively correlated with population coupling strength. While neurons on average retained their coupling strengths across conditions, some neurons change coupling strength condition dependent, thereby potentially enhancing the coding abilities of cortical circuits.

Human attention is inherently transient and limited in span to only a few moments without lapsing. The intrinsic dynamics of large-scale neurocognitive networks are thought to contribute to these lapses and result in the unavoidable fluctuations in attention that constrain its span. However, it remains unclear how the millisecond temporal dynamics of specific electrophysiological brain states contribute to the endogenous maintenance of attention or the onset of attentional lapses. In the present study, we investigated whether the strength and millisecond dynamics of brain electric microstates differentiate states of focus from inattention and contribute to the endogenous maintenance of attention over short and long timescales. We recorded 128-channel EEG while participants maintained their attention during the wait time delay of trials in the Sustained Attention to Cue Task (SACT) and segmented the EEG into a categorized time series of microstates based on data-driven clustering of topographic voltage patterns. The findings revealed that the prevalence and rate of occurrence of microstates C and E in the wait time delay of trials differentiated trials in which the target stimulus was correctly detected from incorrectly detected. These same microstates were also implicated in the maintenance of attention over short and long timescales, with their time-varying dynamics changing systematically during the wait time delay of trials and over the course of the task session. Together, these findings demonstrate the sensitivity of microstates to variation in attentional states and suggest that the millisecond dynamics of these brain states contribute to the maintenance of attention over time.

Efficient metabolic waste clearance, via the postulated glymphatic system, is essential for neural homeostasis. However, direct visualization of tissue-cerebrospinal fluid (CSF) exchange remains limited, leading to ongoing debate in the neuroscientific field. The present work revealed evidence of tissue-CSF water exchange in the live human cortex, by employing a novel MRI technique demonstrating the flux of water molecules across the perivascular interface. We observed robust water exchange inside the cortical ribbon, which was more prominent than white matter and deep brain tissue. We validated that the signal originates from CSF and is independent of cerebral perfusion. Water exchange between tissue and CSF declined with age. Furthermore, we demonstrated for the first time that tissue-CSF exchange was impaired in Alzheimers disease (AD), in particular in regions where the perivascular space is clogged by anti-amyloid immunotherapy.

Stroke survivors often experience balance impairments which may result in increased risk of falling. However, it is unclear whether and how a stroke changes cortical involvement for postural control. To clarify this issue, we assessed the effect of stroke on sway-based corticokinematic coherence (CKC), which is a measure of the coupling between cortical electrophysiological signals and postural sways. To that end, we recorded the center-of-pressure fluctuations and electroencephalographic cortical activity of 34 stroke survivors and 34 healthy participants performing balance tasks during which sensory information was manipulated, by either removal or alteration. We found significantly increased CKC derived from medio-lateral sway in stroke participants when standing on foam compared to healthy controls, suggesting an increase in cortical involvement to compensate for the decreased function of the hemiparetic side, even in highly functional stroke survivors. Moreover, a relationship was found between CKC and clinical scores (Berg Balance Scale and Fugl-Meyer). This suggests that CKC could be used as a biomarker to track progress beyond traditional functional recovery as measured by clinical scores.

Hierarchy in the brain emerges across spatial and temporal scales, enabling transformations from rapid sensory encoding to sustained cognitive control. Hierarchical gradients are well established in sensory systems. In contrast, the hierarchical organization of the primate motor cortex remains debated, partly due to its agranular architecture and the absence of clear laminar input-output projections, that obscures the distinction between feedforward and feedback pathways. In particular, the relative hierarchical position of the dorsal premotor cortex (PMd) and the primary motor cortex (M1) cannot be resolved from anatomy alone. To investigate their relative organization, we here adopted a multimodal approach using intrinsic timescales derived from both single-unit spiking activity (SUA) and local field potentials (LFPs) in macaques performing a delayed-match-to-sample reaching task. We found convergent evidence for inter-areal temporal hierarchy, with longer spiking timescales and smaller LFP aperiodic spectral exponents in M1. Across cortical depth, however, temporal organization depended on signal modality. LFP spectral exponents were significantly smaller in deep than superficial layers in both areas, and LFP-autocorrelation timescales were longer in deep layers in M1. In contrast, spiking activity did not show significant laminar differences in intrinsic timescales. Functionally, neurons with longer timescales exhibited more stable representations of the planned movement direction during motor preparation in PMd and slower temporal evolution of movement encoding during execution in both areas. In conclusion, multimodal temporal measures converge on the same hierarchical organization across these two motor areas, with M1 placed higher than PMd. Our study provides the first characterization of intrinsic spiking timescales across cortical layers in any cortical area and shows that laminar temporal organization depends on the neural signal analyzed. This divergence likely reflects their distinct physiological origins. Spikes capture neuronal output, whereas LFPs primarily reflect synaptic and dendritic population activity, potentially integrating differential contributions from apical and basal dendritic inputs.

Working memory (WM) provides a flexible but capacity-limited workspace for maintaining information over short intervals, whereas long-term memory (LTM) serves as a vast and enduring repository for preserving information over extended periods. Decades of research suggest that they are two distinct yet connected systems that together enable adaptive behavior. The link between WM and LTM may not be straightforward, however, as recent evidence has shown that activation-dependent competition among items in WM can weaken their representations in LTM. In the current study, we examined how dynamic competition among items for limited WM resources affects their retention in LTM. We induced competition between items by manipulating temporal expectations in a WM task with either a short (1 s) or a long (4 s) memory delay. Human participants (N = 20) initially prioritized items expected to be tested early, but shifted their priority to items expected to be tested later when the early test did not occur. Using electroencephalography (EEG) and multivariate pattern analysis (MVPA), we tracked the dynamic fluctuations in WM contents based on expected task relevance across the delay window. We linked these temporal profiles during WM with the long-term recognition performance of each item and found that forgetting was associated with a marked decrease in neural evidence for items deemed no longer relevant during the later delay period. These results demonstrate that WM representations fluctuate with temporal expectations and that the de-prioritization of items during WM maintenance is what drives their long-term forgetting.