Biology subjects

Auksztulewicz, R.

Publications and source records attributed to Auksztulewicz, R..

At least 19 recordsLinked to original sources

bioRxiv2026

Learned Audiovisual Predictability Modulates Depth-Resolved Hippocampal-Cortical Functional Coupling

Perception reflects the dynamic integration of prior expectations, uncertainty, and incoming sensory evidence, but the neural circuitry supporting these computations remains debated. Current theories implicate coordinated interactions between the hippocampus and neocortex, potentially at the level of hippocampal subfields and cortical layers. Evidence in humans has come primarily from the visual domain, while less is known about how these interactions support learned auditory predictions. Using submillimetre 7T fMRI, we investigated depth-resolved functional coupling between hippocampal subfields and temporal and frontal cortical regions during audiovisual associative prediction. Participants learned deterministic associations between visual cues and auditory syllables, allowing us to compare auditory processing in predictable and unpredictable contexts. Predictability-related connectivity changes showed a spatially structured pattern involving DG/CA3-pSTG and CA1-inferior frontal regions, with significant edge-wise effects resolved across cortical depths. These findings provide fine-grained evidence that learned auditory predictability modulates hippocampal-cortical functional interactions across temporo-frontal systems.

neuroscience↗

bioRxiv2026

Predictive perception via simultaneous learning and inference

Perception has been proposed to involve an inferential process that combines noisy sensory evidence with prior expectations to estimate the latent states of the environment. Exact Bayesian inference is intractable in the continuous, high-dimensional state spaces of the natural world. Variational schemes address this by restricting the form of the posterior, whereas sampling schemes represent the posterior with a finite set of samples. In both cases the approximation concerns the posterior rather than the state space over which it is defined. Here we place the approximation elsewhere. Rather than restricting the form of the posterior, we restrict the state space, modeling environmental states as discrete and thereby making Bayesian filtering exact. Beliefs are then unconstrained in shape and what has to be learned can be reduced to a transition matrix updated online by a local rule. The cost is that the states, rather than the distribution, must be specified in advance. The Simultaneous Learning and Inference Model (SLIM) combines this exact filtering with a gated Hebbian rule that learns transitions from inferred rather than observed states. In simulations SLIM recovers environmental dynamics under sensory noise and adapts when those dynamics change, with error growing only once the sensor becomes uninformative. A hierarchical instantiation reproduces local and global prediction error effects. Applied to two auditory decision-making experiments under noise, SLIM reproduces behavioral signatures of expectation and supplies trial-level measures of expectation and surprise. Exact inference over a small discrete state space with a single local Hebbian rule is therefore sufficient to account for these phenomena, without an explicit optimization objective and without a parametric approximation to the posterior.

neuroscience↗

bioRxiv2026

Predictive coding networks capture human neural representations missing in supervised DNNs

Neuroscientific learning theories propose that the brain acquires knowledge by constructing internal world models. Supervised learning, the dominant approach in deep neural networks (DNN), relies on external category labels, making it difficult to reconcile with biological learning. There is an increasing trend towards more biologically valid approaches, such as predictive (minimize future surprise) or contrastive (minimize response to expected, maximize to unexpected inputs) objectives, but these approaches typically rely on pretrained DNN models that vary widely in architecture, size, and hyperparameters, making direct comparisons difficult. Here, we isolate the effect of learning by comparing small, identical networks trained with predictive, contrastive, and supervised learning objectives, as well as local (layer-restricted) vs. global (full backpropagation) learning. We show that brain representations after statistical learning are better modeled by a predictive local target than a supervised or contrastive target, and that during learning, the brain attenuates category-specific representations while retaining predictive ones. We additionally show that predictive objectives explain brain variance that standard supervised DNNs do not, and that this variance is tied to predictive rather than stimulus or mismatch processing. These findings demonstrate a controlled approach for testing the algorithmic basis of learning and identify prediction as a core learning mechanism.

neuroscience↗

bioRxiv2026

A probabilistic atlas of the human thalamic reticular nucleus derived from 7T MRI

The thalamic reticular nucleus (TRN) is a thin, inhibitory shell surrounding the thalamus. It regulates the thalamocortical information flow, and thereby plays a central role in attention, task switching, and the sleep-wake cycle. Despite its importance, the TRN remains poorly studied in the human brain. This is largely because its small size and deep anatomical location limit its visibility with conventional non-invasive neuroimaging techniques. Here, we assessed whether the human TRN can be reliably visualised and segmented in vivo using ultra-high field (UHF) magnetic resonance imaging (MRI) at 7 Tesla. High resolution (0.35 mm isotropic) partial-brain T2* and T1 scans were acquired from healthy individuals, followed by manual delineation of the TRN. These in vivo segmentations were compared with TRN estimates obtained from two high-quality postmortem datasets serving as an anatomical reference. In vivo segmentations of TRN volume and thickness closely matched measurements derived from the postmortem reference datasets, and quantitative comparisons showed high consistency in TRN shape and location across individuals while also capturing meaningful inter-individual variability. Using these segmentations, we constructed a publicly available probabilistic atlas of the human TRN. This atlas provides a new resource for incorporating TRN anatomy into functional, structural, and clinical neuroimaging studies. Our findings demonstrate that the human TRN can be robustly mapped in vivo at 7T and establish a foundation for future investigations into its structure and function.

neuroscience↗

bioRxiv2025

Reduced belief updating impairs adaptive step initiation in older adults

Ageing is associated with declines in mobility and balance that threaten independence and increase fall risk. Although these changes are often attributed to deterioration of muscular and sensory systems, growing evidence suggests that age-related alterations in cognitive control also contribute to postural instability and gait impairments. Here, we examined whether shifts in cognitive control strategy underlie impaired gait initiation in older adults by modelling how individuals update beliefs about upcoming actions. Older and younger adults performed a choice-stepping go/no-go task in which step likelihood was manipulated across trials. We measured anticipatory postural adjustments (APAs), the preparatory weight shifts preceding stepping and modelled age-related differences in APA onset using a hierarchical Bayesian framework. Older adults initiated slower APAs, particularly after repeated go trials. Computational modelling revealed reduced learning rates and heightened sensitivity to uncertainty in older adults, identifying altered belief updating as a cognitive mechanism shaping age-related changes in adaptive stepping.

neuroscience↗

bioRxiv2025

Neural correlates of implicit memory for reoccurring sound patterns

The human brain exhibits remarkable sensitivity to auditory patterns. Previous research has shown that beyond the detection of such patterns, humans form memories for them that are implicit, robust to interference, and persistent over time. Here, we examine the mechanisms underlying implicit memory formation for reoccurring regular tone patterns using UHF fMRI. Prior to scanning, implicit memory was induced through a standard behavioral task where participants detected regular patterns embedded in random tone sequences. Unbeknownst to participants, five regular patterns reoccurred sporadically (every 1.5 min) across trials. Consistent with previous reports, behavioral performance indicated a reaction time advantage to reoccurring patterns, confirming the formation of implicit memory traces. During passive listening in the scanner, regularity engaged the superior temporal gyrus, inferior frontal gyrus, insula, putamen, and hippocampus, a result we interpret as reflecting the higher perceptual precision of regular patterns. Reoccurrence interacted with regularity in the superior temporal gyrus and a frontal cluster, showing higher activity for reoccurring compared to novel regular patterns. The hippocampus showed sensitivity to reoccurrence in both regular and random patterns. Together, these results indicate that the neocortex may selectively amplify reoccurring inputs in line with their enhanced (memory driven) perceptual precision.

neuroscience↗

bioRxiv2025

Predictive Coding algorithms induce brain-like responses in Artificial Neural Networks

This study explores whether predictive coding (PC) inspired Deep Neural Networks can serve as biologically plausible neural network models of the brain. We compared two PC-inspired training objectives, a predictive and a contrastive approach, to a supervised baseline in a simple Recurrent Neural Network (RNN) architecture. We evaluated the models on key signatures of PC, including mismatch responses, formation of priors, and learning of semantic information. Our results show that the PC-inspired models, especially a locally trained predictive model, exhibited these PC-like behaviors better than a Supervised or an Untrained RNN. Further, we found that activity regularization evokes mismatch response-like effects across all models, suggesting it may serve as a proxy for the energy-saving principles of PC. Finally, we find that Gain Control (an important mechanism in the PC framework) can be implemented using weight regularization. Overall, our findings indicate that PC-inspired models are able to capture important computational principles of predictive processing in the brain, and can serve as a promising foundation for building biologically plausible artificial neural networks. This work contributes to our understanding of the relationship between artificial and biological neural networks, and highlights the potential of PC-inspired algorithms for advancing brain modelling as well as brain-inspired machine learning.

neuroscience↗

bioRxiv2025

Predictive acoustical processing in human cortical layers

In our dynamic environments, predictive processing is vital for auditory perception and its associated behaviors. Predictive coding formalizes inferential processes by implementing them as information exchange across cortical layers and areas. With laminar-specific blood oxygenation level dependent we measured responses to a cascading oddball paradigm, to ground predictive auditory processes on the mesoscopic human cortical architecture. We show that the violation of predictions are potentially hierarchically organized and associated with responses in superficial layers of the planum polare and middle layers of the lateral temporal cortex. Moreover, we relate the updating of the brains internal model to changes in deep layers. Using a modeling approach, we derive putative changes in neural dynamics while accounting for draining effects. Our results support the role of temporal cortical architecture in the implementation of predictive coding and highlight the ability of laminar fMRI to investigate mesoscopic processes in a large extent of temporal areas.

neuroscience↗

bioRxiv2024

Dissociable dynamic effects of expectation during statistical learning.

The brain is thought to generate internal predictions, based on previous statistical regularities in the environment, to optimise behaviour. Predictive processing has been repeatedly demonstrated and seemingly explains expectation suppression (ES), or the attenuation of neural activity in response to expected stimuli. However, the mechanisms behind ES are unclear and various models of the mechanisms supporting ES have been suggested with conflicting evidence. Sharpening models propose that expectations suppress neurons that are not tuned to the expected stimulus, increasing the signal-to-noise ratio for expected stimuli. In contrast, dampening models posit that expectations suppress neurons that are tuned to the expected stimuli, increasing the relative response amplitude for unexpected stimuli. Previous studies have used decoding analyses to examine these effects, with increases in decoding accuracy interpreted in terms of sharpening and decreases related to dampening. The opposing process theory (OPT) has suggested that both processes may occur at different time points, namely that initial sharpening is followed by later dampening of the neural representations of the expected stimulus as learning progresses. In this study we aim to test this theory and shed light on the dynamics of expectation effects, both at single trial level and over time. Thirty-one participants completed a statistical learning task consisting of paired scene categories whereby a "leading" image from one category is quickly followed by a "trailing" image from a different category. Multivariate EEG analyses focussed on decoding stimulus information related to the trailing image category. Within-trial, decoding analyses showed that stimulus expectation increased decoding accuracy at early latencies and decreased decoding accuracy at later latencies, in line with OPT. However, across trials, stimulus expectation decreased decoding accuracy in initial trials and increased decoding accuracy in later trials. We theorise that these dissociable dynamics of expectation effects within and across trials can be explained in the context of hierarchical learning mechanisms. Our single trial results provide evidence for the OPT, while our results over time suggest that sharpening and dampening effects emerge at different stages of learning in the visual domain.

neuroscience↗

bioRxiv2024

"What" and "when" predictions jointly modulate speech processing

Adaptive behavior rests on forming predictions based on previous statistical regularities encountered in the environment. Such regularities pertain not only to the contents of the stimuli ("what") but also their timing ("when"), and both interactively modulate sensory processing. In speech streams, predictions can be formed at multiple hierarchical levels, both in terms of contents (e.g. single syllables vs. words) and timing (e.g., faster vs. slower time scales). Whether and how these hierarchies map onto each other in terms of integrating "what" and "when" predictions remains unknown. Under one hypothesis neural hierarchies may link "what" and "when" predictions within sensory processing areas: with lower cortical regions mediating interactions for smaller units e.g., syllables, and higher cortical areas mediating interactions for larger units e.g., words. Alternatively, interactions between "what" and "when" predictions might rest on a generic, sensory-independent mechanism, mediated by common attention-related (e.g., frontoparietal) networks. To address those questions, we manipulated "what" and "when" predictions at two levels - single syllables and disyllabic pseudowords - while recording neural activity using magnetoencephalography (MEG) in healthy volunteers (N=22). We studied how syllable and/or word deviants are modulated by "when" predictability, both analyzing event-related fields and using source reconstruction and dynamic causal modeling to explain the observed effects in terms of the underlying effective connectivity. "When" predictions modulated "what" mismatch responses in a specific way with regards to speech hierarchy, such that mismatch responses to deviant words (vs. syllables) were amplified by temporal predictions at a slower (vs. faster) time scale. However, these modulations were source-localized to a shared network of cortical regions, including frontal and parietal sources. Effective connectivity analysis showed that, while mismatch responses to violations of "what" predictions modulated connectivity between regions, the integration of "what" and "when" predictions selectively modulated connectivity within regions, consistent with gain effects. These results suggest that the brain integrates "what" and "when" predictions that are congruent with respect to their hierarchical level, but this integration is mediated by a shared and distributed cortical network. This contrasts with recent studies indicating separable networks for different levels of hierarchical speech processing.

neuroscience↗

bioRxiv2023

Dissociable roles of the auditory midbrain and cortex in processing the statistical features of natural sound textures

It is well established that sound texture perception takes advantage of a hierarchy of time-averaged statistical features of acoustic stimuli, but much remains unclear about how these statistical features are processed in auditory subcortical and cortical regions. Here, we compared the neural representation of sound textures in the inferior colliculus (IC), and primary and non-primary auditory cortex (AC) of anesthetized rats with texture morph stimuli which gradually add statistical features of increasingly higher order. We generated texture morphs for a representative subset of 13 sound textures, chosen to span the three principal component dimensions of a corpus of 200 natural texture recordings. For each of the 13 texture types, six different exemplar morphs were synthesized using different random seeds. All exemplars of each texture type have the same long term statistics, but they differ in their acoustic waveforms. Meanwhile, different texture types differ in statistical features as well as waveforms. An analysis of transient and ongoing multi-unit responses to our stimulus set showed that the IC units were sensitive to every type of statistical feature, albeit to a varying extent, but only a small proportion of AC units were overtly sensitive to any statistical features. Differences in texture types explained more of the variance of IC neural responses than did differences in exemplars, indicating a degree of "texture-type tuning" in the IC, but this was, perhaps surprisingly, not the case for AC responses. We also evaluated the accuracy of texture type classification from single-trial population activity and found that IC responses became more informative as more summary statistics were included in the texture morphs, while for AC population responses classification performance remained constant, and consistently lower, than for the IC. These results argue against the idea that AC plays an important role in encoding statistical features of natural sound textures.

physiology↗

bioRxiv2023

Multimodal acoustic-electric trigeminal nerve stimulation modulates conscious perception

Multimodal stimulation has the potential to reverse pathological neural activity and alleviate symptoms in neuropsychiatric diseases. However, the reliability of this approach and the mechanisms through which it improves consciousness remain largely unknown. We investigated the effects of multimodal stimulation combining music stimulation with electrical trigeminal nerve stimulation in healthy human participants. We assessed conscious perception before and after acoustic-electric stimulation and investigated the mechanisms underlying the putative stimulation effects. Our results show that (1) acoustic-electric stimulation improves conscious tactile perception in healthy human participants without a concomitant change in auditory perception, (2) this improvement is caused by the interplay of the acoustic and electric stimulation rather than any of the unimodal stimulation alone, and (3) the effect of acoustic-electric stimulation on conscious perception correlates with inter-regional connection changes in a recurrent neural processing model. These findings provide evidence that multimodal acoustic-electric stimulation can promote conscious perception and offer insights into its underlying mechanisms.

neuroscience↗

bioRxiv2023

Distinct mechanisms underlie value-driven modulation of visual cortex by previously rewarded visual and auditory stimuli

Past reward associations may be signaled by stimuli from different sensory modalities, however it remains unclear how different types of reward-associated stimuli modulate perception. In this human fMRI study, we employed a paradigm involving a visual discrimination task, where a visual target was simultaneously presented with either an intra-(visual) or a cross-modal (auditory) cue that was previously associated with rewards. We hypothesized that depending on the sensory modality of the cues distinct neural pathways underlie the value-driven modulation of visual areas. Two steps of analyses were conducted: first, using a multivariate approach, we confirmed that previously reward-associated cues enhanced the target representation in the early visual areas. Then, using effective connectivity analysis, we tested three possible patterns of communication across the brain regions that could underlie the modulation of visual cortex: a direct pathway from the frontal valuation areas to the visual areas, a mediated pathway through the attention-related areas, and a mediated pathway that additionally involved distinct sensory association areas for auditory and visual rewards. We found evidence for the third model and demonstrate that reward-related information is communicated across the valuation and attention-related brain regions such as the intraparietal sulcus across for both visual and auditory cues. Additionally, the long-range communication of reward information also involved the superior temporal areas in case of auditory reward-associated stimuli. These results suggest that in the presence of previously rewarded stimuli from different sensory modalities, a combination of domain-general and domain-specific mechanisms are recruited across the brain to adjust visual processing.

neuroscience↗

bioRxiv2023

Hierarchical Deviant Processing in Auditory Cortex of Awake Mice

Detecting patterns, and noticing unexpected pattern changes, in the environment is a vital aspect of sensory processing. Adaptation and prediction error responses are two components of neural processing related to these tasks, and previous studies in the auditory system in rodents show that these two components are partially dissociable in terms of the topography and latency of neural responses to sensory deviants. However, many previous studies have focused on repetitions of single stimuli, such as pure tones, which have limited ecological validity. In this study, we tested whether the auditory cortical activity shows adaptation to repetition of more complex sound patterns (bisyllabic pairs). Specifically, we compared neural responses to violations of sequences based on single stimulus probability only, against responses to more complex violations based on stimulus order. We employed an auditory oddball paradigm and monitored the auditory cortex (ACtx) activity of awake mice (N=8) using wide-field calcium imaging. We found that cortical responses were sensitive both to single stimulus probabilities and to more global stimulus patterns, as mismatch signals were elicited following both substitution deviants and transposition deviants. Notably, A2 area elicited larger mismatch signaling to those deviants than primary ACtx (A1), which suggests a hierarchical gradient of prediction error signaling in the auditory cortex. Such a hierarchical gradient was observed for late but not early peaks of calcium transients to deviants, suggesting that the late part of the deviant response may reflect prediction error signaling in response to more complex sensory pattern violations.

neuroscience↗

bioRxiv2022

"What" and "When" Predictions Modulate Auditory Processing in a Contextually Specific Manner

Extracting regularities from ongoing stimulus streams to form predictions is crucial for adaptive behavior. Such regularities exist in terms of the content of the stimuli (i.e., "what" it is) and their timing (i.e., "when" it will occur), both of which are known to interactively modulate sensory processing. In real-world stimulus streams, regularities also occur contextually - e.g. predictions of individual notes vs. melodic contour in music. However, it is unknown whether the brain integrates predictions in a contextually congruent manner (e.g., if slower "when" predictions selectively interact with complex "what" predictions), and whether integrating predictions of simple vs. complex features rely on dissociable neural correlates. To address these questions, our study employed "what" and "when" violations at different levels - single tones (elements) vs. tone pairs (chunks) - within the same stimulus stream, while neural activity was recorded using electroencephalogram (EEG) in participants (N=20) performing a repetition detection task. Our results reveal that "what" and "when" predictions interactively modulated stimulus-evoked response amplitude in a contextually congruent manner, but that these modulations were shared between contexts in terms of the spatiotemporal distribution of EEG signals. Effective connectivity analysis using dynamic causal modeling showed that the integration of "what" and "when" prediction selectively increased connectivity at relatively late cortical processing stages, between the superior temporal gyrus and the fronto-parietal network. Taken together, these results suggest that the brain integrates different predictions with a high degree of contextual specificity, but in a shared and distributed cortical network. Significance statementPredictions of stimulus features, present in different statistically-regular contexts in the environment, are crucial to forming adaptive behavior. However, it is unknown if the brain integrates predictions selectively according to such contextual differences. By recording human electroencephalography during experimental manipulations of time-based and content-based predictions, we found that those predictions interactively modulated neural activity in a contextually congruent manner, such that local (vs. global) time-based predictions modulated content-based predictions of sequence elements (vs. chunks). These modulations were shared between contextual levels in terms of the spatiotemporal distribution of neural activity. This suggests that the brain integrates different predictions with a high degree of contextual specificity, but in a shared and distributed cortical network.

neuroscience↗

bioRxiv2022

Omission responses in field potentials but not spikes in rat auditory cortex

Non-invasive recordings of gross neural activity in humans often show responses to omitted stimuli in steady trains of identical stimuli. This has been taken as evidence for the neural coding of prediction or prediction error. However, evidence for such omission responses from invasive recordings of cellular-scale responses in animal models is scarce. Here, we sought to characterise omission responses using extracellular recordings in the auditory cortex of anaesthetised rats. We profiled omission responses across local field potentials (LFP), analogue multiunit activity (AMUA), and single/multi-unit spiking activity, using stimuli that were fixed-rate trains of acoustic noise bursts where 5% of bursts were randomly omitted. Significant omission responses were observed in LFP and AMUA signals, but not in spiking activity. These omission responses had a lower amplitude and longer latency than burst-evoked sensory responses, and omission response amplitude increased as a function of the number of preceding bursts. Contrary to theories of neural entrainment, rhythmic stimulus presentation did not increase low-frequency phase-locking of neural activity specific to the stimulus presentation rate. Together, our findings show that omission responses are observed in LFP and AMUA signals, with laminar specificity, but are not observed in spiking activity, and do not show evidence for low-frequency phase locking. This has implications for models of cortical processing that require many neurons to encode prediction error in their spike output, and may have some consistency with representation of error in dendrites electrotonically distant from the soma.

neuroscience↗

bioRxiv2021

Simultaneous Mnemonic and Predictive Representations in the Auditory Cortex

Recent studies have shown that stimulus history can be decoded via the use of broadband sensory impulses to reactivate mnemonic representations. It has also been shown that predictive mechanisms in the auditory system demonstrate similar tonotopic organization of neural activity as that elicited by the perceived stimuli. However, it remains unclear if the mnemonic and predictive information can be decoded from cortical activity simultaneously and from overlapping neural populations. Here, we recorded neural activity using electrocorticography (ECoG) in the auditory cortex of anesthetized rats while exposed to repeated stimulus sequences, where events within the sequence were occasionally replaced with a broadband noise burst or omitted entirely. We show that both stimulus history and predicted stimuli can be decoded from neural responses to broadband impulse at overlapping latencies but linked to largely independent neural populations. We also demonstrate that predictive representations are learned over the course of stimulation at two distinct time scales, reflected in two dissociable time windows of neural activity. These results establish a valuable tool for investigating the neural mechanisms of passive sequence learning, memory encoding, and prediction mechanisms within a single paradigm, and provide novel evidence for learning predictive representations even under anaesthesia.

neuroscience↗

bioRxiv2021

The precedence effect in spatial hearing emerges only late in the auditory pathway

BackgroundTo localize sound sources accurately in a reverberant environment, human binaural hearing strongly favors analyzing the initial wave front of sounds. Behavioral studies of this "precedence effect" have so far largely been confined to human subjects, limiting the scope of complementary physiological approaches. Similarly, physiological studies have mostly looked at neural responses in the inferior colliculus, or used modeling of cochlear mechanics in an attempt to identify likely underlying mechanisms. Studies capable of providing a direct comparison of neural coding and behavioral measures of sound localization under the precedence effect are lacking. ResultsWe adapted a "temporal weighting function" paradigm for use in laboratory rats. The animals learned to lateralize click trains in which each click in the train had a different interaural time difference. Computing the "perceptual weight" of each click in the train revealed a strong onset bias, very similar to that reported for humans. Follow-on electrocorticographic recording experiments revealed that onset weighting of ITDs is a robust feature of the cortical population response, but interestingly it often fails to manifest at individual cortical recording sites. ConclusionWhile previous studies suggested that the precedence effect may be caused by cochlear mechanics or inhibitory circuitry in the brainstem and midbrain, our results indicate that the precedence effect is not fully developed at the level of individual recording sites in auditory cortex, but robust and consistent precedence effects are observable at the level of cortical population responses. This indicates that the precedence effect is significantly "higher order" than has hitherto been assumed.

neuroscience↗