Journal of Cognitive Neuroscience

Auditory working memory is often conceived of as a unitary capacity, with memory for different auditory materials (syllables, pitches, rhythms) thought to rely on similar neural mechanisms. One spontaneous behavior observed in working memory studies is chunking. For example, individuals often recount digit sequences in groups, or chunks, of 3 to 4 digits, and this chunking improves performance. Chunking may also operate in musical rhythm, with beats acting as chunk boundaries for tones in rhythmic sequences. Similar to chunking, beat-based structure in rhythms also improves performance. Thus, beat processing may rely on the same mechanisms that underlie chunking of verbal material. The current fMRI study examined whether beat perception is a type of chunking, measuring brain responses to chunked and unchunked letter sequences relative to beat-based and nonbeat-based rhythmic sequences. Participants completed a sequence discrimination task, and comparisons between stimulus encoding, maintenance, and discrimination were made for both rhythmic and verbal sequences. Overall, rhythm and verbal working memory networks overlapped substantially. When comparing rhythmic and verbal conditions, rhythms activated basal ganglia, supplementary motor area, and anterior insula, compared to letter strings, during encoding and discrimination. Letter strings compared to rhythms activated bilateral auditory cortex during encoding, and parietal cortex, precuneus, and middle frontal gyri during discrimination. Importantly, there was a significant interaction in the basal ganglia during encoding: activation for beat-based rhythms was greater than for nonbeat-based rhythms, but verbal chunked and unchunked conditions did not differ. The significant interaction indicates that beat perception is not simply a case of chunking, suggesting a dissociation between beat processing and grouping mechanisms that warrants further exploration.

Attention can be deployed in anticipation of visual stimuli based on features such as their color or direction of motion. This anticipatory feature-based attention involves top-down neural control signals from the frontoparietal network that bias visual cortex to enhance the processing of attended information and suppress distraction. So, for example, anticipatory attention control can enable effective selection based on stimulus color while ignoring distracting information about stimulus motion. But as well, anticipatory attention can be focused more narrowly, for example, to select specific colors or motion directions that define task-relevant events and objects. One important question that remains open is whether anticipatory attention control first biases broad feature dimensions such as color versus motion before biasing the specific feature attributes (e.g., blue vs. green). To investigate this, we recorded EEG activity during a task where participants were cued to either attend to a color (blue or green) or a motion direction (up or down) on a trial-by-trial basis. Applying multivariate decoding approaches to the EEG alpha band (8-12 Hz) activity during the attention control period (cue-target interval), we observed significant decoding for both the attended dimensions (color vs. motion) and specific feature attributes (blue vs. green; up vs. down). Importantly, the temporal onset of the dimension-level biasing (color vs. motion) preceded that of the attribute-level biasing (e.g., blue vs. green). These findings demonstrate that the top-down control of feature-based attention proceeds in a hierarchical fashion, first biasing the broad feature dimension, and then narrowing to the specific feature attribute. Significance StatementDuring voluntary feature-based attention, electrophysiological and neuroimaging studies have highlighted the role of anticipatory (top-down) biasing of the sensory cortex in enhancing the selection of attended stimulus attributes, but little is known about how this is achieved. In particular, it is not clear whether attending to an attribute such as a color (blue vs. green) or motion direction (up vs. down) first biases all neural structures coding that dimension (color/motion) before biasing the specific attribute, or if the top-down signals directly bias only the attended attribute. Using EEG and multivariate decoding, we report that top-down attention control follows a hierarchical organization: first, the broader attended feature dimension is biased, which is followed by the biasing of the specific feature attribute.

The extent to which attention and memory processes rely on shared as opposed to distinct mechanisms is critical for understanding the role of both processes in cognition. As working memory sits at the intersection of external, perceptual input and long-term internal storage, it provides the ideal testbed for investigating overlaps between attention and memory. We hypothesize that memory brain states, whole-brain activity patterns that support long-term memory encoding and retrieval, map onto the external/internal axis of attention. Specifically, we hypothesize that external attention, focusing on sensory information, recruits the encoding state and internal attention, focusing on stored information, recruits the retrieval state. To test this hypothesis, we conducted a scalp electroencephalography study in which participants engaged in a working memory paradigm with and without maintenance demands. We used an independently validated cross-study multivariate pattern classifier to measure memory brain state engagement during change and target detection tasks. We find that the encoding state is recruited for stimulus presentation during both tasks, whereas the retrieval state is selectively recruited during the delay of the change detection task. Together, these results suggest that memory states map onto the external/internal axis of attention to support working memory, long-term memory, and cognition more broadly.

Attentional states are highly dynamic and variable, fluctuating from moment to moment and showing stark differences across contexts. To what extent does functional brain reorganization capture variability in attentional states? In the present study we utilize a time-resolved measure of functional MRI connectivity to examine and compare the extent to which univariate activity and functional networks reflect second-to-second sustained attentional fluctuations. Sustained attention was measured objectively, using auditory and visual tasks, and subjectively while participants watched and listened to narratives. Results revealed that objective measures of sustained attention to images and sounds involved common patterns of neural activity and functional interactions. Additionally, networks related to sustained attentional performance during controlled tasks also predicted fluctuations in subjective attentional engagement while participants watched movies and listened to a podcast. Generalization between experimental and everyday task contexts highlights the robustness of time-resolved functional networks for capturing dynamic fluctuations in sustained attentional states.

Visual working memory (WM) must maintain relevant information, despite the constant influx of both relevant and irrelevant information. Attentional control mechanisms help determine which of this new information gets access to our capacity-limited WM system. Previous work has treated attentional control as a monolithic process-either distractors capture attention or they are suppressed. Here, we provide evidence that attentional capture may instead be broken down into at least two distinct sub-component processes: 1) spatial capture, which refers to when spatial attention shifts towards the location of irrelevant stimuli, and 2) item-based capture, which refers to when item-based WM representations of irrelevant stimuli are formed. To dissociate these two sub-component processes of attentional capture, we utilized a series of EEG components that track WM maintenance (contralateral delay activity), suppression (distractor positivity), item individuation (N2pc), and spatial attention (lateralized alpha power). We show that relevant interrupters trigger both spatial and item-based capture, which means that they undermine WM maintenance more. Irrelevant interrupters, however, only trigger spatial capture from which ongoing WM representations can recover more easily. This fractionation of attentional capture into distinct sub-component processes provides a framework by which the fate of ongoing WM processes after interruption can be explained.

Recent behavioral and neural research on reading shows that humans can extract syntactic structure from short sentences within a fraction of a second--faster than many estimates for recognizing the meaning of a single word. This challenges a core assumption of many language processing models: that combinatory operations depend on prior lexical access. Further, studies using parallel presentation of full sentences have revealed electrophysiological responses remarkably similar to those well established for single words. This raises the question of whether words, phrases, and sentences all move through the same processing stages, regardless of syntactic complexity. Using magnetoencephalography (MEG), we examined how single words, phrases, and sentences are processed when all visual information is available at once. Across all three levels, we observed highly similar waveform dynamics, with early responses reflecting bottom-up detection of form followed by activity in the left anterior and posterior temporal cortices and vmPFC consistent with combinatory processing. Of these regions, the left anterior temporal lobe (LATL) showed effects of bigram frequency suggestive of serial left-to-right dynamics. Together, these results support a Global-to-Serial Assembly (GLOSA) model in which the brain first detects the global form of the stimulus in a snapshot-like manner and then probes its combinatory properties through partially serial processes.

Increasing working memory (WM) load incurs behavioral costs, and whether the neural constraints on behavioral costs are localized (i.e., emanating from the intraparietal sulcus) or distributed across cortex remains an active area of debate. In a pre-registered fMRI experiment, 12 humans (12 scanner-hours each) performed a visual WM task with varying memory load (0-4 items). We replicated a localized, load-dependent increase in univariate BOLD activity in parietal cortex. However, we also observed both systematic increases and decreases in univariate activity with load across the visual hierarchy. Importantly, multivariate activation patterns encoded WM load regardless of the direction of the univariate effect, arguing against a restricted locus of load signals in parietal cortex. Finally, we observed representational drift in activity patterns encoding memory load across scanning sessions. Our results suggest a distributed code for memory load that may be continually refined over time to support more efficient information storage.

The role of spatial attention in iconic memory - an early interface between visual perception and short-term memory - remains poorly understood. Across two experiments run in 2025, we investigated how endogenous spatial attention modulates iconic memory using a partial-report paradigm. In both exper-iments, pre-cues manipulated the allocation of spatial attention before stimulus onset. Stimulus arrays were then briefly presented to one hemifield and followed by a post-cue that probed iconic memory at varying delays after stimulus offset. In Experiment 1 (N = 47), valid attentional cues improved perfor-mance at short delays. Performance was modeled with an exponential decay function to dissociate effects on initial stimulus availability at short SOAs, the rate of iconic decay, and later transfer to working mem-ory. This analysis indicated that valid pre-cues increased initial stimulus availability relative to invalid pre-cues. In Experiment 2 (N = 66), a neutral pre-cue condition was added, and post-cues were pre-sented at three delays (0, 120, and 1240 ms). This revealed that performance differences were driven by attentional costs at invalidly cued locations, with no detectable benefits at validly cued locations relative to neutral cues. Together, these results show that spatial attention modulates the earliest measurable phase of iconic memory by shaping the initial sensory trace. The cost-dominated pattern suggests that attention primarily suppresses information at unattended locations rather than enhancing representa-tions at attended locations. This finding challenges the view of iconic memory as a pre-attentive sensory store and indicates that attentional selection operates earlier than previously assumed. Significance StatementThe influence of attention on iconic memory - a high-capacity, ultra-brief sensory store - remains highly debated. We demonstrate that endogenous spatial attention modulates iconic memory at its earliest measurable stage. Critically, this modulation reflects an attentional cost at unattended locations rather than a benefit at attended locations, suggesting that spatial attention acts through inhibitory suppression of irrelevant sensory input. These findings challenge models proposing that early sensory representation is categorically attention-free and establish a suppressive role for attention in shaping visual short-term memory at its earliest stage.

Working memory (WM) enables the temporary maintenance of information for guiding thought and action, but its limited capacity requires internal selective attention mechanisms to prioritize goal-relevant representations. Like external attention, internal selection can be shaped by both endogenous (goal-driven) and exogenous (stimulus-driven) factors. Classic work by Charles Folk and Roger Remington demonstrated that such factors interact competitively during external attention, producing stimulus-driven capture under certain conditions. Whether similar competitive dynamics govern internal selection remains unclear. Prior studies using retro-cue paradigms with extended cue-response delays have reported evidence for "retro-capture"--involuntary orienting toward task-irrelevant but cue-matching working memory content. Here, we combined EEG and eye-tracking to test whether retro-capture persists when the cue and response probe are simultaneous, eliminating the temporal window in which capture has previously been observed. Participants remembered two oriented bars and used either a pro-cue or an anti-cue to guide recall. Across behavioral, cortical, and oculomotor measures, we found no evidence for retro-capture. Instead, competition between endogenous and exogenous selection mechanisms uniformly delayed the selection of task-relevant WM content, with delays evident in response times, lateralized alpha-band activity, and gaze biases toward remembered locations. Our findings indicate that retro-capture is not an obligatory consequence of endogenous-exogenous competition in WM but may depend on the temporal separation between cue and action.

Statistical learning (SL) enables the extraction of regularities from sensory input, yet the neural dynamics supporting this process--particularly in the visual modality--remain incompletely understood. Sixty-seven adults were familiarized with a continuous stream of shape sequences containing statistical structure that defined shape triplets. We recorded EEG to familiar sequences (presented in isolation) and unfamiliar foils. Both early (N100) and late (N400) event related potential (ERP) components were significantly more negative for unfamiliar than familiar sequences, reflecting robust neural sensitivity to learned structure. Notably, these familiarity effects were evident in both high- and low-performing participants and were not predicted by overall behavioral sensitivity, suggesting that neural indices of learning can emerge independently of explicit recognition. Follow-up analyses incorporating trial-level accuracy revealed a striking crossover interaction: for sensitive participants, ERP familiarity effects were stronger on correct trials, whereas for insensitive participants, effects were larger on incorrect trials. These findings highlight a dissociation between neural and behavioral measures of statistical learning and underscore the value of ERPs in capturing latent learning processes that may elude conscious awareness.

Keeping track of multiple moving objects across dynamic real-world scenarios such as driving, team sports, or crowded social environments is a fundamental challenge for visual attention. We have previously demonstrated that as the number of tracked objects increases, the strength of attentional facilitation allocated to each individual object decreases, limiting tracking success. It is also well established that beyond the number of tracked objects, faster-moving objects and objects embedded amongst higher numbers of distractors are more difficult to track. Are these effects on tracking difficulty also mediated by less effective allocation of attention to tracked targets as in the case of tracking more targets? If so, one should expect the strength of attentional modulation to drop systematically with increasing speed and total number of moving stimuli. In two experiments (total n = 70), participants were instructed to track moving targets amongst identical distractors while we manipulated object speed (Experiment 1) and number (Experiment 2). As expected, tracking performance declined with both manipulations. However, steady-state visual evoked potentials (SSVEPs) recorded during successful tracking revealed that attentional enhancement of tracked targets compared with distractors did not drop with increasing speed or object number. In summary, bottom-up changes in the stimulus display and top-down attentional manipulations affect tracking performance in independent ways, with the balance between strength of attentional allocation and bottom-up demands of the task determining successful tracking. The allocation of attention itself seems to be determined exclusively by top-down goals rather than being reactive to bottom-up display characteristics. Open Practices StatementParticipant level data and analysis code for all experiments are available at (https://osf.io/ypgfs/) and Experiment 1 was preregistered (https://osf.io/pxh25/). Significance statementKeeping track of multiple moving objects is fundamental to navigating dynamic real-world scenarios. This ability is accomplished through multifocal attentional selection, which weakens as the number of tracked targets increases. This study asks whether other stimulus manipulations increase tracking difficulty by diluting attentional allocation. Using steady-state visual evoked potentials to measure selective attention during tracking, we demonstrate that both increases in speed and distractor number impair performance, however, they do not affect attentional enhancement of targets. This suggests that top-down attentional control operates independently from bottom-up demands.

The error-related negativity (ERN) and correct-related negativity (CRN) are event-related potentials (ERPs) that reflect performance monitoring following error and correct responses, respectively. Prior work demonstrates the ERN is sensitive to the motivational significance of errors, which increases under social observation. However, most studies testing how social observation impacts performance monitoring rely on trial-averaged ERPs, potentially obscuring meaningful fluctuations in ERN/CRN over time. Here, we had participants complete a Flanker task twice (social observation vs. alone) and employed mixed-effects modeling of single-trial ERPs to test if social observation impacts ERN/CRN trajectories over short (within blocks) or long (between blocks) timescales. We found that social observation selectively influenced ERN/CRN trajectories over short timescales: for blocks performed under social observation (but not alone), ERN magnitudes increased across trials and CRN magnitudes decreased. At longer timescales, ERN/CRN significantly decreased across all blocks, regardless of social observation and consistent with a vigilance decrement. To our knowledge, this is the first demonstration that social observation influences performance monitoring trajectories over short timescales. Results highlight the importance of analyzing ERN/CRN trajectories over relatively short timescales to fully characterize the impact of social observation on performance monitoring dynamics. These findings lay the groundwork for future investigation into whether social observation interacts with individual differences in motivation/affect to differentially impact performance monitoring dynamics.

Understanding the neural process underlying the comprehension of visual images and sentences remains a major open challenge in cognitive neuroscience. We previously demonstrated with fMRI and DTI that comprehension of visual images and sentences describing human activities recruits a common semantic system. The current research tests the hypothesis that this common semantic system will display similar neural dynamics during processing in these two modalities. To investigate these neural dynamics we recorded EEG from naive subjects as they saw simple narratives made up of a first visual image depicting a human event, followed by a second that was either a sequentially coherent narrative follow-up, or not, of the first image. In separate blocks of trials the same protocol was presented using sentences. Analysis of the EEG signal revealed common neural dynamics for semantic processing across image and sentence modalities. Late positive ERPs were observed in response to sequential incoherence for sentences and images, consistent with previous studies that examined coherence in these two modalities separately. Analysis of oscillatory power revealed increased gamma-band activity for sequential coherence, again consistent with previous studies showing gamma increases for coherence and matching in sentence and image processing. Multivariate analysis demonstrated that training a classifier on data from one modality (images or sentences) allowed reliable decoding of the sequential coherence of data from trials in the untrained modality, providing further support for a common underlying semantic system for images and sentences. Processing sequential coherence of successive stimuli is associated with neural dynamics that are common to sentence and visual image modalities and that can be decoded across modalities. These results are discussed in the context of EEG signatures of narrative processing and meaning, and more general neural mechanisms for structure processing.

Human long-term memory for visual scenes is remarkably robust, yet the neural mechanisms supporting memory encoding and retrieval remain poorly understood when both processes must operate at the same time. For instance, this might happen when we encounter a familiar place while simultaneously forming new memories of this encounter. We investigated electrophysiological correlates of visual recognition memory using a continuous recognition task (CRT), in which participants judged a continuous stream of scene photographs as previously seen or new, such that encoding and retrieval occurred in parallel on every trial. To make recognition particularly demanding, stimuli were drawn from only four scene categories. Thirty-one participants performed the task while EEG was recorded, and we analyzed canonical ERP markers of retrieval (mid-frontal FN400, 300-550 ms; late parietal effect, LPE, 550-800 ms) and encoding (subsequent memory effect, SME) as a function of stimulus repetition and lag between consecutive presentations. FN400 showed robust old/new effects for both repetitions, whereas LPE differences emerged only at the second repetition. While FN400 amplitude was insensitive to lag, LPE amplitude decreased systematically with increasing lag, mirroring the behavioral pattern of declining accuracy and slower responses. A significant SME emerged selectively for images subsequently recognized on both repetitions, indicating that the SME in continuous recognition is specific for the most robustly encoded items and reflects the strength of encoding. Together, these findings show that canonical ERP markers of recognition memory are preserved even when encoding and retrieval operate concurrently, but their expression depends on how often and how recently an item has previously been encoded - parameters that can be flexibly manipulated within the CRT. This demonstrates that the CRT is sensitive to fine-grained temporal dynamics of memory formation and retrieval that could be missed under standard single-repetition designs.

We have learned much about the brain regions that support reading by measuring neuronal responses to single words, but we know little about how the brain processes multiple words simultaneously. This fMRI study fills that gap by varying the number of English words presented while holding the amount of visual stimulation constant. We adapted the "simultaneous suppression" paradigm, which has demonstrated that the response to multiple stimuli presented simultaneously is typically smaller than the sum of responses to the same stimuli presented sequentially. On each trial, participants viewed rapid sequences of three frames. Each frame contained two character strings, most of which were pseudo-letters with visual features matched to familiar letters. The experimental conditions differed in the number of English words in the sequence: zero words; one word; two words sequentially; or two words simultaneously. Behaviorally, participants performed worse at detecting two words presented simultaneously than sequentially. In the brain, the BOLD response increased linearly with the number of words presented in several reading-related regions of the left hemisphere: the inferior frontal sulcus, the intraparietal sulcus, the superior temporal sulcus, and text-selective occipito-temporal regions. But the mean response magnitudes did not differ between sequential and simultaneous presentation of two words. Nonetheless, the sensitivity of ventral temporal text-selective regions to two words lexical frequencies was attenuated by simultaneous presentation. This suggests that words were processed less deeply when they competed simultaneously. Our new paradigm, therefore, reveals processing capacity limits in the reading circuitry.

For almost half a century, target-distractor similarity has been known to induce different visual search modes. When a target is highly salient, it can pop out, suggesting parallel processing of all items irrespective of set size. By contrast, high similarity among items requires item-by-item comparison with an attentional template, a characteristic of serial search. Despite this long-standing distinction, little is known about the neural correlates of search modes, as typical differences in visual displays confound interpretation. Here, we contrasted the neural correlates of serial and parallel search under visually identical displays. Across distinct blocks, we biased 24 participants (21 female) toward parallel or serial search by varying target-distractor similarity, thereby directing attentional focus toward a single feature or conjunction. Embedded among inducer trials, test trials were visually identical across all blocks and afforded both parallel and serial search. Behavioral analyses confirmed successful induction of distinct search modes during test trials. EEG decoding reliably discriminated search modes and these neural patterns generalized across inducer and test trials. Attentional deployment toward the target differed across search modes, revealing topographical differences in target location representations. The strength of target-location representations correlated with response times, indicating that during parallel test trials participants switched search strategies when the target was not detected early. Moreover, target representations diverged between search modes: a temporally stable pattern emerged during serial search, suggesting reliance on working memory, whereas parallel search was characterized by more dynamic representations, likely reflecting prioritization of the relevant feature. These findings demonstrate that search history shapes search mode, giving rise to clearly distinct neural dynamics even under visually identical stimulation.

The rapid, continuous flow of spoken language places strong demands on attention, and it is thought that listeners meet these demands by predicting when important information will occur and allocating attention accordingly. However, to date there is little direct evidence for the involvement of preparatory attention during language processing. In this study, we investigate preparatory attention during spoken language comprehension by measuring alpha neural activity with EEG, a known measure of temporal attentional preparation. Alpha activity leading up to target words that were either focused or defocused by a preceding discourse question did not vary as a function of focus, challenging the assumption that attention is pre-allocated to the timing of focused words. On the other hand, we found that trial-by-trial fluctuations in alpha activity predicted both the depth of processing and the subsequent memory for new information. Specifically, pre-target alpha modulated a centro-parietal Dm subsequent memory effect for focused words, linking preparatory attention to memory encoding during comprehension. Together, these findings bridge psycholinguistic studies on information structure and cognitive neuroscience research on temporal attention, offering novel insights into the role of alpha activity in attentional dynamics during spoken language processing.

Previous research has linked endogenous alpha oscillations ([~]7-13 Hz) to temporal integration windows in visual perception, with higher individual alpha frequency (IAF) predicting improved temporal segregation. Here, we investigated whether alpha-rhythmic temporal integration is a factor in visual crowding and whether this relationship is mediated by spatial grouping mechanisms. 47 participants performed a Vernier discrimination task, in which we manipulated both the stimulus onset asynchrony (SOA) between flankers and targets, and the spatial configuration of the flankers. Specifically, flankers were arranged to either induce crowding or "uncrowding", through the manipulation of good-Gestalt properties. Our results show that crowding has a temporal integration period of around 170 ms but this varies substantially across individuals. Importantly, resting-state IAF predicted individual variance in temporal integration windows: individuals with faster endogenous alpha rhythms could begin to segregate targets from distractors at shorter SOAs. Crucially, this effect was specific for crowding-inducing flankers and disappeared when flankers led to uncrowding. These results suggest that top-down spatial grouping can overwrite the temporal integration constraint imposed by alpha oscillations, highlighting both the relevance of alpha for understanding limits on peripheral visual processing as well as the flexible and context-dependent role of alpha in temporal integration. SIGNIFICANCE STATEMENTThe frequency of alpha-band activity varies across individuals and has previously been linked to temporal integration windows for low-level stimulus properties (e.g., flashes). However, the relevance of individual alpha frequency (IAF) for everyday perception is less clear. We tested whether IAF predicts temporal integration in spatial crowding, a phenomenon that strongly limits perceptual abilities. We found that individuals with higher IAF were better able to spatially segregate targets from flankers at shorter flanker-target offsets, indicating that they spatially integrated the flankers over a smaller temporal window. Interestingly, this effect disappeared when crowding was alleviated by flanker configurations following good-Gestalt principles. Our results suggest that bottom-up mediated spatio-temporal constraints on perception are linked to alpha oscillations but can be overridden by top-down spatial cues.

Episodic recollection involves retrieving context information bound to specific events. However, autobiographical memory largely comprises recurrent, similar experiences that become integrated into joint representations. In the current study, we extracted a neural signature of temporal context from scalp electroencephalography (EEG) to investigate whether recalling a recurring event accompanies the reinstatement of one or multiple instances of its occurrence. We asked 52 young adults to study and recall lists of words that included both once-presented and repeated items. Participants recalled repeated items in association with neighboring list items from each occurrence, but with stronger clustering around the repetitions initial occurrence. Furthermore, multivariate spectral EEG analyses revealed that neural activity from just prior to the recall of these words resembled patterns of activity observed near the items first occurrence, but not its second. Together, these results suggest that the initial occurrence of an event carries stronger temporal context associations than later repetitions. Research Transparency StatementThe authors report no conflicts of interest with respect to the authorship or publication of this article. This research was conducted with support from National Institutes of Health grant MH55687. The current study was not preregistered. Data are available upon reasonable request and with proper approval from the University of Pennsylvania research and ethics entities. Requests should be directed to Adam Broitman. Key EEG and behavioral data analysis scripts are available for download at https://github.com/awb99cu/repFRcode.Studymaterialsareavailablefordownloadathttps://memory.psych.upenn.edu/PEERS.

Statistical learning--the ability to extract patterns from noisy continuous experiences--is fundamental to human cognition. Yet, how contextual factors shape this process remains poorly understood. Music is an important example of such contextual factors, because it is ubiquitous in human experience and provides a rich temporally-structured stimulus that can co-occur with other learning processes. Here we demonstrate that pairing music fundamentally enhances visual statistical learning, and this is correlated with systematic reorganization of large-scale brain networks. Using fMRI and a novel probabilistic sequence learning paradigm, we show that familiar melodies significantly improved participants ability to segment continuous visual streams into events and learn sequential relationships. Neuroimaging analyses revealed that the presence of music fundamentally altered the neural network organization that coordinates learning mechanisms: while sequence learning in silence engaged frontal-parietal networks associated with explicit pattern extraction, providing musical temporal structure as a context shifted learning toward MTL-vmPFC circuits recently implicated in schema-guided memory processing. Machine learning analyses confirmed these architectural differences, with the music condition achieving optimal neural prediction of behavioral performance through distributed connectivity patterns while control condition relied on concentrated processing. Our findings support a Cross-Modal Temporal Scaffolding Theory, demonstrating that structured temporal context signals from one modality (here, music) can create more efficient neural states for sequence processing in another through dual mechanisms: enhanced memory integration through schema-guided learning and reduced demands on explicit control resources. These results identify network-level principles for optimizing statistical learning, with broad implications for understanding how environmental context shapes human learning capacity.