Beyond Neural Networks: The Role of Global Oscillations in Memory and Cognition

 

Ravinder Jerath^1* and Varsha Malani²

Received: March 20, 2026 | Published: April 10, 2026

Copyright^© 2026 Genesis Pub by Jerath R et al. This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 International License (CC BY 4.0). This license permits unrestricted use, distribution, and reproduction in any medium, provided the original author(s) and source are properly credited.

DOI: http://doi.org/10.52793/JNSR.2026.6(1)-S6

Abstract

For decades, the dominant paradigm in cognitive neuroscience has framed memory and cognition as emergent properties of discrete synaptic connections within neural networks. While this connectionist framework has yielded profound insights, it remains incomplete. A growing body of evidence suggests that large-scale, rhythmic electrical oscillations — spanning multiple brain regions and frequency bands — serve as a fundamental organizational substrate for cognition. These global oscillations coordinate the temporal dynamics of neural firing, gate synaptic plasticity, facilitate long-range communication, and support the binding of distributed representations into coherent mental states. This paper synthesizes current research on the role of theta, gamma, alpha, beta, and delta oscillations in episodic memory, working memory, attention, and higher-order cognition. We argue that a comprehensive theory of memory must move beyond static network topology to embrace the dynamic, rhythmic, and global nature of brain activity. We further discuss how oscillatory dysfunction underlies major cognitive disorders and how entrainment-based therapies may offer novel clinical interventions.

Keywords

Beyond Neural Networks; Global Oscillations.

Introduction

The human brain contains approximately 86 billion neurons, interconnected by an estimated 100 trillion synapses. The sheer complexity of this architecture has naturally inspired computational metaphors: the brain as a network, memory as weighted connections, cognition as information processing through hierarchical modules. The connectionist tradition, pioneered by McCulloch and Pitts (1943) and later formalized through Hebbian learning rules and artificial neural networks, has been extraordinarily productive. It has explained how pattern recognition, associative memory, and category learning can emerge from the statistical regularities of synaptic weights.

Yet several fundamental cognitive phenomena resist clean explanation within this framework. How does the brain bind together color, shape, motion, and identity into a unified percept of a single object? How does working memory maintain representations across time without continuous sensory input? How do episodic memories — rich, contextually specific, temporally ordered — differ from semantic memories encoded in overlapping networks? And how does the brain dynamically route information between regions depending on behavioral context?

A compelling answer to these questions lies not in the architecture of connections alone, but in the timing of neural activity. Rhythmic, synchronous oscillations — generated by the interplay of excitatory and inhibitory neural populations — provide a dynamic scaffold upon which cognition is built. These oscillations are not epiphenomenal noise; they are active computational mechanisms that organize when neurons fire, which neurons communicate with which, and how information is routed and bound across distributed cortical and subcortical systems.

This paper reviews the neuroscience of global brain oscillations with a focus on their contributions to memory and cognition. Section 2 provides a primer on oscillatory mechanisms and measurement. Section 3 examines specific frequency bands and their cognitive roles. Section 4 focuses on oscillations in episodic and working memory. Section 5 addresses attention and executive function. Section 6 discusses inter-regional coordination and the communication-through-coherence hypothesis. Section 7 examines oscillatory dysfunction in neurological and psychiatric disorders. Section 8 considers therapeutic implications. Section 9 offers a theoretical synthesis, and Section 10 concludes with future directions.

Neural Oscillations: Mechanisms and Measurement

The origins of brain rhythms

Neural oscillations arise from the intrinsic membrane properties of individual neurons and, more importantly, from the dynamics of recurrent excitatory-inhibitory (E-I) circuits. When a population of excitatory pyramidal neurons activates local inhibitory interneurons (particularly parvalbumin-positive basket cells), the resulting feedback inhibition creates rhythmic fluctuations in the local field potential (LFP). The frequency of oscillation is determined by the time constants of synaptic inhibition: fast GABA-A receptor-mediated inhibition generates high-frequency gamma oscillations (30–80 Hz), while slower GABA-B or metabotropic mechanisms contribute to lower-frequency rhythms.

Long-range oscillations emerge when multiple local circuits synchronize through axonal projections, gap junctions, or subcortical pacemakers. The thalamus plays a particularly important role as a hub that can entrain widespread cortical networks through thalamocortical loops. Subcortical structures including the hippocampus, basal ganglia, and brainstem nuclei contribute additional oscillatory inputs that modulate cortical dynamics.

Measurement techniques

The primary tool for studying brain oscillations in humans is electroencephalography (EEG), which records the summed postsynaptic potentials of millions of cortical neurons with millisecond temporal resolution. Magnetoencephalography (MEG) offers complementary spatial precision by detecting the magnetic fields associated with the same currents. Intracranial EEG (iEEG) and local field potential recordings in animal models provide direct access to oscillatory activity at the level of individual cortical areas and subcortical structures, with far superior spatial resolution.

Analytical methods have become increasingly sophisticated. Time-frequency analyses (wavelet transforms, Morlet filtering, Hilbert transforms) allow researchers to track the amplitude, phase, and frequency of oscillations dynamically. Phase-amplitude coupling (PAC) — the nesting of high-frequency oscillations within specific phases of slower rhythms — has emerged as a particularly informative measure of cross-frequency coordination. Functional connectivity analyses assess the degree of phase synchrony or amplitude correlation between spatially separated regions.

Frequency bands and their functional significance

Brain oscillations are conventionally divided into frequency bands, each associated with distinct functional roles:

• Delta (0.5–4 Hz): Deep sleep, slow cortical potentials, long-range integration
• Theta (4–8 Hz): Hippocampal-neocortical communication, episodic encoding, spatial navigation
• Alpha (8–12 Hz): Inhibitory gating, attention, sensory suppression
• Beta (13–30 Hz): Sensorimotor processing, working memory maintenance, top-down control
• Gamma (30–80 Hz): Local feature binding, perceptual encoding, fast information transfer
• High-frequency oscillations (HFO, >80 Hz): Sharp-wave ripples in the hippocampus; memory consolidation

These bands are not independent; their interactions — particularly theta-gamma coupling — are central to theories of memory and cognition.

Frequency Bands and Cognitive Functions 

Theta oscillations: The hippocampal metronome

Theta oscillations (4–8 Hz) are perhaps the most extensively studied oscillations in the context of memory. First described in the rodent hippocampus [1]. hippocampal theta is robustly associated with active locomotion, spatial exploration, and learning. The medial septum serves as the primary pacemaker, projecting cholinergic and GABAergic fibers to hippocampal interneurons that enforce rhythmic inhibition.

In humans, scalp EEG studies consistently show frontal and temporal theta power increases during episodic encoding and retrieval tasks [2]. Intracranial recordings from neurosurgical patients confirm that hippocampal theta coherence with prefrontal and entorhinal cortex predicts successful memory formation [3,4]. The phase of theta modulates the efficacy of synaptic potentiation: long-term potentiation (LTP) is induced at the peak of theta, while long-term depression (LTD) occurs at the trough [5]. This phase-dependent plasticity creates a mechanism by which the timing of inputs — relative to the ongoing theta cycle — determines whether memories are strengthened or weakened.

Theta also plays a critical role in temporal coding. The "theta phase precession" phenomenon, first described [6], in hippocampal place cells, demonstrates that as a rat traverses a spatial environment, place cells fire progressively earlier within each theta cycle relative to the animal's position. This provides a within-cycle temporal code for spatial sequences that is independent of firing rate alone. Theoretical extensions of this principle suggest that theta phase coding may generalize to temporal sequences in episodic memory more broadly [7].

Gamma oscillations: Local computation and feature binding

Gamma oscillations (30–80 Hz) are generated locally by E-I circuits and are associated with active sensory and cognitive processing. The binding problem — how the brain integrates distributed features (color, shape, location) processed in separate cortical areas into a coherent unified percept — was proposed by [8], to be solved by synchronous gamma oscillations. Neurons representing different features of the same object would fire synchronously within the gamma cycle, while neurons representing different objects would fire at different phases or asynchronously.

This "binding by synchrony" hypothesis remains influential but contested. Evidence from both animal electrophysiology and human EEG/MEG shows that attentional modulation strongly influences gamma synchrony: attended stimuli elicit stronger gamma responses in visual cortex, and cross-area gamma coherence increases for task-relevant information [9] Gamma oscillations have also been linked to working memory: the number of items held in working memory appears to be encoded by distinct gamma cycles nested within each theta cycle [10], providing an elegant mechanism for multi-item maintenance.

Alpha oscillations: Gating by inhibition

Alpha oscillations (8–12 Hz) were the first brain rhythms to be systematically described [38], and were initially viewed as a signature of idle cortical states. Contemporary research has fundamentally revised this view. Alpha oscillations are now understood as an active inhibitory mechanism that gates the flow of sensory information and regulates cortical excitability.

The "pulsed inhibition" model proposes that alpha represents periodic bursts of inhibition, suppressing cortical processing at a rate of approximately 10 times per second [13]. Crucially, alpha power is not globally uniform: task-irrelevant cortical areas show increased alpha (suppression), while task-relevant areas show alpha desynchronization (increased excitability). This spatially specific alpha modulation implements a selective attention filter: information from attended spatial locations or stimulus features is processed during low-alpha states; while competing information is suppressed by high alpha in irrelevant areas.

In memory, alpha power has been linked to the suppression of irrelevant memory traces during retrieval, protecting the target memory from interference. Posterior alpha reflects visual working memory load in an inverse fashion — as working memory capacity is exceeded, alpha suppression over visual cortex fails, and performance declines.

Alpha oscillations (8–12 Hz) were the first brain rhythms to be systematically described [38], and were initially viewed as a signature of idle cortical states. Contemporary research has fundamentally revised this view. Alpha oscillations are now understood as an active inhibitory mechanism that gates the flow of sensory information and regulates cortical excitability.

The "pulsed inhibition" model proposes that alpha represents periodic bursts of inhibition, suppressing cortical processing at a rate of approximately 10 times per second [13]. Crucially, alpha power is not globally uniform: task-irrelevant cortical areas show increased alpha (suppression), while task-relevant areas show alpha desynchronization (increased excitability). This spatially specific alpha modulation implements a selective attention filter: information from attended spatial locations or stimulus features is processed during low-alpha states; while competing information is suppressed by high alpha in irrelevant areas.

In memory, alpha power has been linked to the suppression of irrelevant memory traces during retrieval, protecting the target memory from interference. Posterior alpha reflects visual working memory load in an inverse fashion — as working memory capacity is exceeded, alpha suppression over visual cortex fails, and performance declines.

Beta oscillations: The status quo signal

Beta oscillations (13–30 Hz) have been associated with the maintenance of existing cognitive and motor states — a "status quo" rhythm that resists change. During working memory maintenance, sustained beta activity in prefrontal and parietal regions is thought to hold representations stable against distraction [16]. Beta oscillations in the cortico-basal ganglia-thalamic loop regulate the timing of motor program execution and cognitive set-shifting.

Long-range beta synchrony between frontal and parietal regions increases during tasks requiring top-down attentional control, suggesting a role in maintaining task-relevant processing templates. Disruption of beta activity, as seen in Parkinson's disease and schizophrenia, correlates with deficits in working memory and cognitive flexibility.

Delta oscillations and sharp-wave ripples

At the slowest end of the oscillatory spectrum, delta oscillations (0.5–4 Hz) during slow-wave sleep coordinate the coupling between hippocampal sharp-wave ripples (SWRs) and cortical sleep spindles (12–15 Hz). This three-way coupling — slow oscillation UP states → cortical spindles → hippocampal ripples — is thought to be the neural substrate of memory consolidation. During sharp-wave ripples, previously encoded hippocampal representations are "replayed" in a compressed and temporally reversed manner, allowing their transfer to neocortical long-term storage [15,16].

The sharp-wave ripple is one of the most compelling examples of an oscillatory mechanism with a specific memory function: its disruption in rodents selectively impairs spatial memory consolidation, while its artificial reactivation during sleep can enhance memory for specific experiences [17].

Oscillations in Memory Systems

Episodic memory encoding

Episodic memory — the capacity to encode, store, and retrieve personally experienced events with their spatial and temporal context — depends critically on the hippocampus and its oscillatory communication with the prefrontal cortex, entorhinal cortex, and parietal regions.

The hippocampal-entorhinal theta oscillation provides a temporal framework for encoding sequential information. The entorhinal cortex acts as the primary gateway for cortical inputs to the hippocampus, and its grid cells — neurons that fire in spatially periodic, hexagonal patterns — may provide a universal metric for encoding not just space but also conceptual and temporal relationships [18].

Grid cell activity is phase-locked to theta, suggesting that the encoding of spatial and temporal context is fundamentally an oscillatory computation.

Human studies employing depth electrode recordings have shown that successful episodic encoding is associated with: (1) increased hippocampal theta power, (2) increased theta coherence between hippocampus and prefrontal cortex, (3) increased gamma power nested at the theta peak, and (4) phase resetting of the theta oscillation at stimulus onset [19]. The subsequent memory effect (SME) — the difference in neural activity between items later remembered versus forgotten — is reliably indexed by these oscillatory measures, often more robustly than by firing rate alone.

Memory consolidation during sleep

The consolidation of episodic memories from a labile hippocampal trace into a stable neocortical representation occurs primarily during sleep, particularly slow-wave sleep (SWS). The "active systems consolidation" hypothesis [20]. proposes that memory-associated hippocampal patterns are reactivated during SWRs, and these reactivations are coordinated with cortical slow oscillations and thalamic spindles to drive synaptic changes in neocortical networks.

The temporal precision of this multi-oscillatory coupling is remarkable. Hippocampal SWRs preferentially occur during the UP states of cortical slow oscillations, when cortical excitability is maximal. Sleep spindles — waxing and waning bursts of thalamocortical activity at 12–15 Hz — are temporally nested within the slow oscillation UP state and occur just before or coincident with hippocampal ripples [21]. This precise timing appears essential: experimental desynchronization of ripples and spindles using transcranial stimulation impairs next-day recall, while their amplification enhances it.

Working memory

Working memory — the capacity to temporarily maintain and manipulate information — is classically associated with sustained neural firing in prefrontal cortex (PFC). However, purely persistent activity models face a challenge: they require unrealistically stable network states and do not naturally account for item individuation (how the brain keeps multiple items in working memory distinct rather than blended).

The theta-gamma code for working memory, proposed by Lisman and Idiart [10], and elaborated by Jensen and Lisman [11], offers an elegant solution. According to this model, each item in working memory is represented by a distinct gamma subcycle nested within the theta oscillation. A 7 Hz theta rhythm with 40 Hz gamma provides approximately 7 gamma cycles per theta cycle — strikingly consistent with the classical "magical number seven" limit of working memory capacity [12]. Each gamma cycle reactivates a different item in sequence, enabling the maintenance of multiple distinct items within a single theta cycle.

Empirical support for this model comes from intracranial recordings in humans and monkeys showing that working memory load modulates both theta power and the strength of theta-gamma coupling in PFC and hippocampus [22,13]. Moreover, the sequential reactivation of working memory items is phase-locked to the theta rhythm, with different items appearing at different theta phases [23].

Oscillations, Attention, and Executive Function

Top-down attentional control

Attention is not a single process but a family of mechanisms that include spatial orienting, feature-based selection, temporal preparation, and executive control. Each of these components has been associated with specific oscillatory dynamics.

Executive function and prefrontal oscillations

Supporting evidence comes from studies of visual attention in monkeys: gamma-band coherence between V4 and frontal eye fields increases for attended stimuli, while alpha-band coherence increases for unattended ones [24]. In humans, MEG studies show that directing spatial attention modulates alpha lateralization (suppression contralateral to the attended location) and gamma enhancement in visual cortex, consistent with the CTC.

Executive functions — planning, cognitive flexibility, inhibitory control, task-switching — rely heavily on PFC and its oscillatory interactions with other brain regions. Frontal theta power increases during tasks requiring conflict monitoring [25]. error processing, and the maintenance of task rules. Cross-frequency coupling between frontal theta and posterior gamma has been proposed as a mechanism by which PFC exerts top-down control over sensory cortex.

The basal ganglia, long recognized for their role in action selection and cognitive flexibility, are increasingly understood to contribute to oscillatory dynamics. The subthalamic nucleus (STN) plays a key role in generating beta oscillations that propagate through the cortico-striato-thalamo-cortical loop. Cognitive demands that require breaking existing motor or cognitive "sets" are associated with transient suppression of beta power (beta event-related desynchronization), consistent with the status quo role of beta described above.

Inter-Regional Coordination and Global Brain States

The global workspace and oscillatory broadcast

The global workspace theory of consciousness [26,27], proposes that conscious cognition involves the ignition of a widely distributed network — including prefrontal, parietal, and cingulate regions — that broadcasts information to specialized processors throughout the brain. This ignition is accompanied by a characteristic pattern of long-range synchrony: gamma-band coherence increases dramatically between frontal and posterior regions when a stimulus crosses the threshold of conscious awareness.

Non-conscious processing, by contrast, is characterized by local, feedforward gamma activity confined to sensory cortex, without the long-range synchrony that accompanies conscious access. This distinction — local oscillations for unconscious processing versus global, long-range synchrony for conscious awareness — provides a testable, mechanistic account of the neural correlates of consciousness [27].

Default mode network and low-frequency coherence

The default mode network (DMN) — comprising medial PFC, posterior cingulate, angular gyrus, and medial temporal lobe — is active during rest, mind-wandering, episodic recall, and future thinking. The internal coherence of the DMN is largely maintained by slow, infra-slow oscillations (<0.1 Hz) and their interaction with faster oscillatory dynamics. Disruption of DMN coherence is one of the earliest and most consistent neural biomarkers of Alzheimer's disease, preceding the appearance of amyloid plaques and cognitive symptoms by years.

The DMN's oscillatory relationship with task-positive networks (frontoparietal control network, dorsal attention network) is characterized by anticorrelation: DMN activity suppresses during externally directed cognition and resurges during internally directed cognition. This anti-correlated relationship is enforced by the mutual inhibition between these networks, mediated in part by alpha oscillations.

Traveling waves and spatial organization

A recent development in the field concerns traveling waves: oscillatory activity that propagates across cortical tissue as a wave rather than occurring synchronously everywhere. Far from being a curiosity, traveling waves appear to be a fundamental mode of cortical operation. Theta waves traveling along the hippocampal long axis organize the sequential processing of spatial and temporal information [28]. In neocortex, alpha waves traveling from occipital to frontal regions during top-down attention processing may implement predictive coding mechanisms.

The spatial direction and speed of traveling waves carry information: the direction of propagation distinguishes sensory-driven feedforward processing (posterior-to-anterior) from memory-driven feedback processing (anterior-to-posterior). This spatial organization adds an additional dimension to the temporal code provided by oscillatory phase.

Oscillatory Dysfunction in Cognitive Disorders

Alzheimer's disease

Alzheimer's disease (AD) is characterized by progressive episodic memory failure, followed by widespread cognitive decline. At the oscillatory level, AD brains show reduced hippocampal theta power and theta-gamma coupling, decreased gamma activity in cortical networks, disrupted slow-wave sleep oscillations and sleep spindles, and reduced inter-regional coherence in the DMN. Strikingly, gamma oscillatory disruption may occur decades before clinical symptom onset.

The observation by Iaccarino [29], that 40 Hz flickering light stimulation (entraining gamma oscillations) reduces amyloid and tau pathology in mouse models of AD has generated enormous clinical interest. A subsequent large human trial (the GENUS trial) used combined auditory and visual 40 Hz entrainment and found improvements in hippocampal volume, functional connectivity, and some [30,31]. The mechanistic basis may involve gamma-frequency coordination of glial (astrocyte and microglia) activity in clearing amyloid — a link between macroscale oscillations, cellular metabolism, and neuropathology.

Schizophrenia

Schizophrenia is associated with widespread oscillatory abnormalities, most prominently deficits in gamma power and synchrony. The auditory steady-state response (ASSR) — the brain's oscillatory response to rhythmic auditory stimulation — is robustly reduced at 40 Hz in schizophrenic patients, replicated across dozens of studies worldwide. This gamma deficit is linked to dysfunction of parvalbumin-positive interneurons, which are essential for generating and maintaining gamma oscillations.

Working memory deficits in schizophrenia are associated with abnormal PFC theta-gamma coupling and reduced alpha-band suppression of task-irrelevant information. The disorganized, fragmented thought characteristic of psychosis may reflect a failure of oscillatory coordination — the breakdown of the temporal scaffolding that normally binds distributed representations into coherent mental states.

Major depressive disorder (MDD) is associated with disrupted slow-wave sleep architecture, reduced sleep spindle density, and aberrant default mode network activity. Elevated frontal alpha asymmetry (greater left than right alpha power) has been proposed as a trait marker of depression vulnerability. Treatment-resistant depression responds to transcranial magnetic stimulation (TMS) targeting frontal oscillatory dynamics, with efficacy correlating with normalization of connectivity between the subgenual cingulate and DMN.

Therapeutic Implications: Oscillatory Entrainment

Principles of neural entrainment

The brain's oscillatory dynamics are not fixed; they can be entrained — phase-locked and modulated — by external rhythmic stimuli. This can be achieved through sensory stimulation (auditory, visual, or tactile rhythms), transcranial electrical stimulation (tACS, tDCS), transcranial magnetic stimulation (TMS), or direct intracranial stimulation. The principle of neural entrainment provides a non-invasive pathway for manipulating cognition and potentially treating neurological disease.

Transcranial alternating current stimulation (tACS) — the application of sinusoidal electrical current at specific frequencies — can entrain cortical oscillations at the stimulation frequency, strengthen inter-regional synchrony, and modulate cognitive performance. Theta-frequency tACS over temporal-parietal regions enhances associative memory encoding. Alpha-frequency tACS can shift the balance of attentional allocation. Gamma-frequency stimulation — both electrical and sensory — has shown promise in reducing AD pathology.

Closed-loop and personalized approaches

A significant limitation of open-loop entrainment approaches is their inability to account for the natural variability in individual oscillatory states. Closed-loop systems — which continuously monitor ongoing brain activity and deliver stimulation precisely timed to specific oscillatory phases — represent the next frontier. Phase-specific TMS delivered at the peak of individual theta cycles produces stronger memory enhancements than non-phase-locked stimulation [32]. Similarly, closed-loop auditory stimulation during sleep, timed to the UP state of slow oscillations, enhances the slow oscillation amplitude and next-day memory performance in both young and older adults [33].

Personalized neurostimulation protocols — calibrated to an individual's specific oscillatory profile, cognitive deficits, and genetic factors — represent a promising avenue for precision medicine approaches to cognitive disorders.

Pharmacological modulation

Multiple pharmacological agents modulate brain oscillations. Benzodiazepines, which potentiate GABA-A receptors, suppress theta and gamma while enhancing beta and spindle activity — consistent with their amnestic effects. Acetylcholinesterase inhibitors (used in AD treatment) enhance hippocampal theta by increasing cholinergic tone in the medial septum-hippocampal axis. Ketamine, an NMDA receptor antagonist, paradoxically increases gamma power while disrupting alpha and theta dynamics — effects that may underlie both its psychotomimetic and rapid antidepressant properties.

The discovery that specific oscillatory patterns can be targeted pharmacologically opens the possibility of "oscillotherapy" — drugs designed not to target specific receptors but to normalize specific frequency bands and inter-regional synchrony profiles associated with cognitive deficits.

Theoretical Synthesis: Toward a Dynamic Systems Account of Cognition

The limits of the static network metaphor

The connectionist framework — in which cognitive functions are properties of the static topology of synaptic weights — faces fundamental limitations. It cannot easily explain why the same neural circuits produce radically different cognitive states depending on behavioral context, arousal, pharmacological state, or time of day. It struggles to account for the rapid, dynamic routing of information between brain regions that attention and executive control require. And it provides no natural mechanism for the temporal sequencing that underlies episodic memory, language, and planning.

The oscillatory framework complements connectionism by adding a temporal dimension. Cognitive states are not simply patterns of activation across a fixed network; they are patterns of activity organized in time by rhythmic oscillations. The same synaptic connections can support radically different computational outcomes depending on the oscillatory state of the network — an insight with profound implications for both normal cognition and neurological disease.

Hierarchical temporal integration

A particularly powerful theoretical framework integrates oscillations into a hierarchical account of temporal processing [34,35]. Different cortical regions are tuned to different timescales of input: primary auditory cortex integrates information over tens of milliseconds (in the gamma range), higher auditory areas integrate over hundreds of milliseconds (theta range), and frontal and default mode regions integrate over seconds to minutes (delta range). This cortical hierarchy of temporal receptive windows maps directly onto the oscillatory hierarchy of frequency bands.

The coupling between these hierarchical oscillations — with faster rhythms nested within slower ones — implements a nested temporal coding scheme in which the meaning of a gamma-encoded feature depends on the theta cycle in which it occurs, which in turn depends on the delta phase, and so on. This hierarchical coupling provides a mechanism for context-sensitive processing: the same sensory input can be interpreted differently depending on the current state of slow oscillations that encode temporal and behavioral context.

Predictive processing and oscillations

The predictive processing framework [36,37], proposes that the brain constantly generates top-down predictions about sensory input and updates these predictions based on prediction errors propagating up the cortical hierarchy. Oscillations map naturally onto this framework: alpha and beta oscillations carry top-down predictions from higher to lower cortical areas, while gamma oscillations carry bottom-up prediction errors. Phase synchrony between cortical levels is modulated by the match between prediction and input — when predictions are accurate, gamma is suppressed; when predictions fail, gamma surges, triggering updates in the predictive model.

This integration of oscillatory neuroscience with predictive processing provides a unified account of perception, attention, memory, and learning that goes beyond both static network models and purely oscillatory descriptions. Memory, in this framework, is not simply stored in synaptic weights but is actively reconstructed at retrieval through the top-down oscillatory propagation of stored predictions.

The role of glial cells and the extracellular matrix

A largely underappreciated dimension of brain oscillations involves non-neuronal contributions. Astrocytes — which outnumber neurons in many brain regions — are electrically coupled through gap junctions into large syncytia that can sustain slow extracellular potential fluctuations. These glial-mediated potentials interact with neuronal oscillations through modulation of extracellular potassium and glutamate concentrations. Recent evidence suggests that astrocytic calcium waves occur at infra-slow timescales (<0.1 Hz) and may entrain neuronal oscillations across millimeters of cortex.

Perineuronal nets (PNNs) — dense extracellular matrix structures that ensheath parvalbumin interneurons — modulate the speed and precision of inhibitory interneuron firing, and thus the frequency of gamma oscillations. Disruption of PNNs in adolescence extends the critical period of cortical plasticity; in adulthood, PNNs stabilize oscillatory dynamics and consolidate memory circuits. The interaction between the extracellular matrix and oscillatory dynamics represents a frontier where cellular neuroscience and systems neuroscience converge.

Future Directions and Conclusions

Future directions

Several major questions remain open and are likely to drive the field forward over the coming decade.

The causal role of oscillations: Most evidence for oscillatory contributions to memory and cognition is correlational. While optogenetic studies in rodents can establish causality by driving specific interneuron populations at defined frequencies, equivalent causal demonstrations in humans require the development of more precise, targeted neurostimulation methods. Closed-loop TMS and intracranial stimulation paradigms will be critical.

Oscillations in artificial intelligence: Standard artificial neural networks lack oscillatory dynamics. Incorporating oscillatory mechanisms — particularly theta-gamma coupling and alpha-mediated gating — into AI architectures may yield systems with more human-like working memory, context sensitivity, and continual learning capabilities. Reservoir computing and liquid state machine models, which exploit the temporal dynamics of recurrent networks, represent early steps in this direction.

Individual differences and biomarkers: Individual oscillatory profiles vary substantially, and this variability is partly heritable. Understanding how genetic and environmental factors shape oscillatory dynamics will enable the development of cognitive biomarkers — oscillatory signatures that predict cognitive trajectories, disease risk, and treatment response.

Developmental oscillations: The postnatal development of cognitive capacities is accompanied by dramatic changes in oscillatory dynamics. Gamma activity matures late, consistent with the prolonged development of parvalbumin interneurons. Theta-gamma coupling strengthens across childhood and adolescence in parallel with improvements in episodic memory capacity. Understanding how oscillatory development constrains and enables cognitive development may shed light on neurodevelopmental disorders.

Cross-species translation: While rodent models have provided the majority of mechanistic insights into hippocampal oscillations, there are important differences between rodent and human oscillatory profiles. Bridging this gap — through careful comparative electrophysiology and the development of primate models — will be essential for translating mechanistic insights into clinical therapies.

Conclusions

The study of global brain oscillations has fundamentally enriched our understanding of how memory and cognition arise from biological neural systems. Far from being irrelevant background noise, oscillations are active computational mechanisms that organize the timing of neural activity, implement attentional selection, bind distributed representations, scaffold the sequential structure of episodic memory, and coordinate the dialogue between hippocampus and neocortex during memory consolidation.

The framework presented here — in which cognition is a dynamic, rhythmic, globally coordinated phenomenon — complements rather than replaces the connectionist tradition. Synaptic connections provide the anatomical substrate; oscillations provide the dynamic organizing principle that breathes life into that substrate. A full theory of memory and cognition must encompass both the static architecture of neural networks and the temporal dynamics of oscillatory synchrony.

As measurement technologies improve, as causal manipulation methods become more precise, and as theoretical frameworks integrate oscillatory dynamics with synaptic plasticity and predictive processing, the field is poised to achieve a genuinely comprehensive account of how the human brain transforms experience into memory, and memory into thought.

References

1. Jung R, Kornmüller AE. (1938) Eine Methodik der Ableitung lokalisierter Potentialschwankungen aus subcorticalen Hirngebieten. Archiv für Psychiatrie und Nervenkrankheiten. 109(1):1-30.
2. Klimesch W. (1999) EEG alpha and theta oscillations reflect cognitive and memory performance: A review and analysis. Brain Res Rev. 29(2–3):169-95.
3. Fell J, Klaver P, Lehnertz K, Grunwald T, Schaller C, et al. (2001) Human memory formation is accompanied by rhinal–hippocampal coupling and decoupling. Nature Neurosci. 4(12):1259-264.
4. Lega BC, Jacobs J, Kahana M. (2012) Human hippocampal theta oscillations and the formation of episodic memories. Hippocampus. 22(4):748-61.
5. Huerta PT, Lisman JE. (1995) Bidirectional synaptic plasticity induced by a single burst during cholinergic theta oscillation in CA1 in vitro. Neuron. 15(5):1053-63.
6. O'Keefe J, Recce ML (1993) Phase relationship between hippocampal place units and the EEG theta rhythm. Hippocampus. 3(3):317-30.
7. Buzsáki G, Moser EI. (2013) Memory, navigation and theta rhythm in the hippocampal-entorhinal system. Nature Neuroscience. 16(2):130-38.
8. Singer W, Gray CM. (1995) Visual feature integration and the temporal correlation hypothesis. Ann Rev Neurosci. 18:555-86.
9. Fries P. (2015) Rhythms for cognition: Communication through coherence. Neuron. 88(1):220-35.
10. Lisman JE, Idiart MA. (1995) Storage of 7 ± 2 short-term memories in oscillatory subcycles. Sci. 267(5203):1512-15.
11. Jensen O, Lisman JE. (1998) An oscillatory short-term memory buffer model can account for data on the Sternberg task. J Neurosci. 18(24):10688-699.
12. Miller GA. (1956) The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychol Rev. 63(2):81-97.
13. Jensen O, Mazaheri A. (2010) Shaping functional architecture by oscillatory alpha activity: Gating by inhibition. Fronti Human Neurosci. 4:186.
14. Lundqvist M, Rose J, Herman P, Brincat SL, Buschman TJ, et al. (2016) Gamma and beta bursts underlie working memory. Neuron. 90(1):152-64.
15. Buzsáki G. (1989) Two-stage model of memory trace formation: A role for "noisy" brain states. Neuroscience. 31(3):551-70.
16. Diekelmann S, Born J. (2010) The memory function of sleep. Nature Reviews Neurosci. 11(2):114-26.
17. de Lavilléon G, Lacroix MM, Rondi-Reig L, Benchenane K. (2015) Explicit memory creation during sleep demonstrates a causal role of place cells in navigation. Nature Neurosci. 18(4):493-95.
18. Moser EI, Kropff E, Moser MB. (2008) Place cells, grid cells, and the brain's spatial representation system. Ann Rev Neurosci. 31:69-89.
19. Hanslmayr S, Staresina BP, Bowman H. (2016) Oscillations and episodic memory: Addressing the synchronization/desynchronization conundrum. Trends in Neurosci. 39(1):16-25.
20. Born J, Wilhelm I. (2012) System consolidation of memory during sleep. Psychol Res. 76(2):192-203.
21. Staresina BP, Bergmann TO, Bonnefond M, van der Meij R, Jensen O, et al. (2015) Hierarchical nesting of slow oscillations, spindles and ripples in the human hippocampus during sleep. Nature Neurosci. 18(11):1679-86.
22. Axmacher N, Henseler MM, Jensen O, Weinreich I, Elger CE, et al. (2010) Cross-frequency coupling supports multi-item working memory in the human hippocampus. Pro National Acad Sci. 107(7):3228-33.
23. Siegel M, Warden MR, Miller EK. (2009) Phase-dependent neuronal coding of objects in short-term memory. Proceed National Acad Sci. 106(50):21341-346.
24. Bosman CA, Schoffelen JM, Brunet N, Oostenveld R, Bastos AM, et al. (2012) Attentional stimulus selection through selective synchronization between monkey visual areas. Neuron. 75(5):875-88.
25. Cavanagh JF, Frank MJ, Klein TJ, Allen JJB. (2010) Frontal theta links prediction errors to behavioral adaptation in reinforcement learning. NeuroImage. 49(4):3198-209.
26. Baars BJ. (1988) A Cognitive Theory of Consciousness. Cambridge University Press.
27. Dehaene S, Changeux JP, Naccache L, Sackur J, Sergent C. (2006) Conscious, preconscious, and subliminal processing: A testable taxonomy. Trends in Cognitive Sci. 10(5):204-11.
28. Lubenov EV, Siapas AG. (2009) Hippocampal theta oscillations are travelling waves. Nature. 459(7246):534-39.
29. Iaccarino HF, Singer AC, Martorell AJ, Rudenko A, Gao F, et al. (2016) Gamma frequency entrainment attenuates amyloid load and modifies microglia. Nature. 540(7632):230-35.
30. Chan D, Suk HJ, Jackson BL, Milman NP, et al. (2021) Gamma frequency sensory stimulation in probable mild Alzheimer's dementia patients: Results of a preliminary clinical trial. PLOS ONE. 16(11):e0255155.
31. Murdock MH, Yang CY, Sun N, Phelan D, Bharioke A, et al. (2024) Multisensory gamma stimulation promotes glymphatic clearance of amyloid. Nature. 627(8003):374-81.
32. Helfrich RF, Mander BA, Jagust WJ, Knight RT, Walker MP. (2018) Old brains come uncoupled in sleep: Slow wave-spindle coupling during non-REM sleep and its role in memory consolidation. Curr Biol. 28(1)78-88.
33. Ngo HV, Martinetz T, Born J, Mölle M. (2013). Auditory closed-loop stimulation of the sleep slow oscillation enhances memory. Neuron. 78(3):545-53.
34. Hasson U, Yang E, Vallines I, Heeger DJ, Rubin N. (2008) A hierarchy of temporal receptive windows in human cortex. J Neurosci. 28(10):2539-50.
35. Giraud AL, Poeppel D. (2012) Cortical oscillations and speech processing: Emerging computational principles and operations. Nature Neurosci. 15(4):511-17.
36. Clark A. (2013) Whatever next? Predictive brains, situated agents, and the future of cognitive science. Behavioral Brain Sci. 36(3):181-204.
37. Friston K. (2010) The free-energy principle: A unified brain theory? Nature Reviews Neurosci. 11(2):127-38.
38. Berger H. (1929) Über das Elektrenkephalogramm des Menschen. Archiv für Psychiatrie und Nervenkrankheiten. 87(1):527-70.
39. Fries P. (2005) A mechanism for cognitive dynamics: Neuronal communication through neuronal coherence. Trends in Cognitive Sci. 9(10):474-80.