A retinotopic code structures the interaction between perception and memory systems

Nature Neuroscience, Journal Year: 2024, Volume and Issue: 27(2), P. 339 - 347

Published: Jan. 2, 2024

Language: Английский

---

Transforming the Concept of Memory Reactivation

Trends in Neurosciences, Journal Year: 2020, Volume and Issue: 43(12), P. 939 - 950

Published: Oct. 8, 2020

A foundational finding in the field of memory is that content-sensitive patterns neural activity expressed during perceptual experiences are re-expressed when remembered, a phenomenon termed reactivation. However, reactivation obscures key differences how events and memories represented brain.Recent findings suggest systematic, spatial transformations from perception to retrieval. These occur within sensory cortex frontoparietal cortex.We consider why identify critical questions be addressed future research. Understanding ways which representations differ will critically inform theoretical accounts help clarify brain recreates past. Reactivation refers wherein experience at later time, putative marker memory. content has been observed across many cortical areas correlates with objective subjective expressions humans. because emphasizes similarities between memory-based representations, it represented. Here, we highlight recent evidence systematic (and where) brain. We argue best thought as spatially transformed versions representations. for When remembering an event past, often feels though re-experiencing event, bringing mind sensations, emotions, or thoughts characterized initial experience. This quality paralleled by important phenomenon: (see Glossary) evoked retrieved referred influential tool studying episodic Neural measures have shown predictive accuracy [1.Polyn S.M. et al.Category-specific precedes retrieval search.Science. 2005; 310: 1963-1966Crossref PubMed Scopus (416) Google Scholar, 2.Ritchey M. al.Neural similarity encoding related via hippocampal interactions.Cereb. Cortex. 2013; 23: 2818-2828Crossref 3.Gordon A.M. al.Cortical reinstatement mediates relationship content-specific subsequent recollection decisions.Cereb. 2014; 24: 3350-3364Crossref (76) Scholar], vividness [4.St-Laurent al.Distributed predict recollection.J. Cogn. Neurosci. 2015; 27: 2000-2018Crossref (30) Scholar,5.Kuhl B.A. Chun M.M. Successful elicits event-specific lateral parietal cortex.J. 34: 8051-8060Crossref (99) consequences [6.Kuhl reveals mechanisms updating memory.J. 2012; 32: 3453-3461Crossref (0) 7.Lee H. al.Decomposing remembering.Cereb. 2018; 29: 3305-3318Crossref (6) 8.Chanales A.J.H. al.Interference overlapping predicted states learning.Nat. Commun. 2019; 10: 5363Crossref Scholar]. Another notable aspect can measured contexts such rest [9.Tambini A. Davachi L. Persistence multivoxel into postencoding memory.Proc. Natl. Acad. Sci. U. S. 110: 19591-19596Crossref (92) 10.Schlichting M.L. Preston A.R. Memory supports upcoming learning content.Proc. 111: 15845-15850Crossref 11.Schapiro A.C. al.Human replay prioritizes weakly learned information predicts performance.Nat. 9: 3920Crossref (27) Scholar] sleep [12.Rudoy J.D. al.Strengthening individual reactivating them sleep.Science. 2009; 326: 1079Crossref (275) Scholar,13.Antony J.W. al.Cued influences skill 15: 1114-1116Crossref (151) behavioral recordings not feasible. Over past two decades, there tremendous advances both computational sophistication sensitivity methods used measure (Box 1), but these exploit same fundamental 'match' While this emphasis on appealing unquestionably important, here also provides incomplete view contents In particular, reactivation, definition, tests retrieval, necessarily fails capture memory.Box 1Methods Approaches Detecting ReactivationNoninvasive Neuroimaging MethodsMemory extensively studied using noninvasive imaging techniques [e.g., fMRI, electroencephalography (EEG), magnetoencephalography], simultaneously record population large parts Below, detail several popular analytic approaches measuring neuroimaging data. All test profiles activation Notably, applied varying levels specificity: stimulus category (e.g., faces vs scenes objects), specific exemplar apple butterfly), feature value color: red yellow).Univariate ActivationBrain regions identified mean level varies according content. demonstrated exhibiting similar (changes level) [19.Wheeler M.E. al.Memory's echo: vivid reactivates sensory-specific cortex.Proc. 2000; 97: 11125-11129Crossref Scholar].Pattern SimilarityNeural (activity voxels sensors) region. Patterns then region correlation pattern corresponding exceeding noncorresponding [2.Ritchey Scholar,84.Staresina B.P. al.Episodic medial temporal lobe.J. 18150-18156Crossref (106) Scholar].Decoding ModelsSupervised models trained learn mapping stimuli classes perception. tested classifier transfer (i.e., perception-trained model accurately labeling memory) Scholar,22.Kuhl al.Fidelity competition memories.Proc. 2011; 108: 5903-5908Crossref Scholar].Encoding ModelsModels specify mathematical relationships physical properties pixel values) fit data if parameters generalize retrieval) [85.Senden al.Reconstructing imagined letters early visual tight topographic correspondence mental imagery perception.Brain Struct. Funct. 224: 1167-1183Crossref (11) Predicted decode remembered (encoding–decoding) [24.Naselaris T. al.A voxel-wise decodes images scenes.NeuroImage. 105: 215-228Crossref (101) Scholar,25.Bone M.B. al.Feature-specific memory.Nat. 2020; 11: 1945Crossref inverted reconstruct (inverted model) [46.Ester E.F. al.Parietal frontal encode stimulus-specific mnemonic working memory.Neuron. 87: 893-905Abstract Full Text PDF (148) Scholar].Invasive Physiological Optogenetic MethodsWhile current article focuses methods, benefit large-scale coverage brain, all aforementioned local potential single-cell electrophysiological invasively humans (when treated neurological diseases requiring electrode implants) [86.Schreiner Staudigl Electrophysiological signatures humans.Phil. Trans. R. Soc. B Biol. 37520190293Crossref (2) animal experiments. Other approaches, high-density recordings, more prevalent some restricted For example, single-unit datasets collected rodents essential characterizing replay, temporally structured firing sequences [62.Ólafsdóttir H.F. al.The role planning.Curr. 28: R37-R50Abstract (55) Only recently [87.Vaz A.P. al.Replay spiking human retrieval.Science. 367: 1131-1134Crossref techniques, applicable studies solely, revealed experimentally induced sufficient evoke behavior [88.Josselyn S.A. Tonegawa engrams: recalling imagining future.Science. 367eaaw4325Crossref (39) Noninvasive Methods yellow). Univariate Activation Brain Pattern Similarity Decoding Models Supervised Encoding Invasive One obvious way might terms amount they contain, presumably being weaker degraded version Empirically, manifest, reduced ability compared number clear violations idea: specifically, contain about perceptually experienced [14.Steel al.An interface scene posterior cerebral cortex.bioRxiv. (Published online May 26, 2020. https://doi.org/10.1101/2020.05.25.115147)PubMed 15.Long N.M. al.Hippocampal mismatch signals modulated strength predictions their outcomes.J. 2016; 36: 12677-12687Crossref (22) 16.Xiao X. al.Transformed retrieval.J. 2017; 37: 2986-2998Crossref (24) 17.Favila S.E. features amplified flexibly aligned top-down goals.J. 38: 7809-7821Crossref (3) The implication only whether those perception, systematically underestimate amount, strength, qualities opinion article, emerging originally use idea transformation emphasize information: (i) different and, (ii) localization (or predictable) opposed random noise signal degradation (Figure 1). possibility has, date, overshadowed dominant note cognitive theories long argued constructive act than reproductive [18.Schacter D.L. neuroscience memory.Annu. Rev. Psychol. 1998; 49: 289-318Crossref Thus, strong motivation moving beyond measurements index degree resemble toward characterize sections describe supporting proposal avenues understanding perception-to-memory (spatial otherwise) represents logical, timely, next step field. reported most frequently particular. Early active demonstrating coarse-level modality vision versus audition). More recently, established increasingly fine-grained areas. numerous ventral reflect face object) reactivated Scholar,20.Ranganath C. al.Inferior temporal, prefrontal, contributions maintenance associative 2004; 3917-3925Crossref (251) 21.O'Craven K.M. Kanwisher N. Mental places activates regions.J. 12: 1013-1023Crossref (638) 22.Kuhl Further, V1) low-level location, frequency, edges, orientation, [23.Kosslyn al.Topographical primary cortex.Nature. 1995; 378: 496-498Crossref 24.Naselaris 25.Bone 26.Bosch al.Reinstatement signaled hippocampus.J. 7493-7500Crossref (62) common thread among examples Yet, results subtle, found scene-selective responses located meta-analysis over 100 functional magnetic resonance (fMRI) recruits relatively aspects parahippocampal place area (PPA), whereas retrieving anterior PPA [27.Baldassano al.Two distinct scene-processing networks connecting memory.eNeuro. 3 (ENEURO.0178-16.2016)Crossref unpublished study, striking dissociation was replicated paradigm directly within-subjects design Similar (perception) (memory shifts other areas, including occipital Scholar,28.Silson E.H. posterior–anterior distinction construction 39: 705-717Crossref Resting state fMRI analyses ventral, lateral, connectivity consistent supported segregated Scholar,27.Baldassano 28.Silson 29.Baldassano al.Differential area.NeuroImage. 75: 228-237Crossref (84) Related work suggested posterior-to-anterior division extends scenes, objects, outside network [30.Lee S.H. al.Disentangling real-world objects.NeuroImage. 59: 4064-4073Crossref (89) Scholar,31.Bainbridge W.A. al.Distinct representational structure recall imagery.bioRxiv. April 21, https://doi.org/10.1101/842120)Google Collectively, reveal shift system. Specifically, preferentially cortex, depending currently available receptors recalled long-term Importantly, fact system do perceived fundamentally inconsistent Moreover, highly (as degradation) regions. As such, strongly manner Though consistently applications pattern-based analysis robust, [5.Kuhl Scholar,7.Lee 3305-3

Language: Английский

---

Working memory representations in visual cortex mediate distraction effects

Nature Communications, Journal Year: 2021, Volume and Issue: 12(1)

Published: Aug. 5, 2021

Abstract Although the contents of working memory can be decoded from visual cortex activity, these representations may play a limited role if they are not robust to distraction. We used model-based fMRI estimate impact distracting tasks on in several field maps and frontoparietal association cortex. Here, we show distraction causes fidelity briefly dip when both memorandum distractor jointly encoded by population activities. Distraction induces small biases errors which predicted neural decoding early cortex, but other regions. disrupts representations, widespread redundancy with information is protect against catastrophic loss. In representation behavioral performance intertwined, solidifying its importance memory.

Language: Английский

---

Joint representation of working memory and uncertainty in human cortex

Neuron, Journal Year: 2021, Volume and Issue: 109(22), P. 3699 - 3712.e6

Published: Sept. 14, 2021

Language: Английский

---

Perception and memory have distinct spatial tuning properties in human visual cortex

Nature Communications, Journal Year: 2022, Volume and Issue: 13(1)

Published: Oct. 18, 2022

Abstract Reactivation of earlier perceptual activity is thought to underlie long-term memory recall. Despite evidence for this view, it unclear whether mnemonic exhibits the same tuning properties as feedforward activity. Here, we leverage population receptive field models parameterize fMRI in human visual cortex during spatial retrieval. Though retinotopic organization present both perception and memory, large systematic differences are also evident. Whereas there a three-fold decline precision from early late areas perception, pattern not observed This difference cannot be explained by reduced signal-to-noise or poor performance on trials. Instead, simulating top-down network model cortex, demonstrate that property well hierarchical structure system. Together, modeling empirical results suggest computational constraints imposed system architecture limit fidelity reactivation sensory cortex.

Language: Английский

---

Tasks and their role in visual neuroscience

Neuron, Journal Year: 2023, Volume and Issue: 111(11), P. 1697 - 1713

Published: April 10, 2023

Language: Английский

---

Neural population dynamics of human working memory

Current Biology, Journal Year: 2023, Volume and Issue: 33(17), P. 3775 - 3784.e4

Published: Aug. 17, 2023

Language: Английский

---

Transformative neural representations support long-term episodic memory

Science Advances, Journal Year: 2021, Volume and Issue: 7(41)

Published: Oct. 8, 2021

Memory is often conceived as a dynamic process that involves substantial transformations of mental representations. However, the neural mechanisms underlying these and their role in memory formation retrieval have only started to be elucidated. Combining intracranial EEG recordings with deep network models, we provide detailed picture representational from encoding short-term maintenance long-term underlie successful episodic memory. We observed during encoding. Critically, more pronounced semantic formats predicted better subsequent memory, this effect was mediated by consistent item-specific representations across events. The were further transformed right after stimulus offset, similar those than Our results suggest pass through multiple stages achieve recall.

Language: Английский

---

Visual working memories are abstractions of percepts

eLife, Journal Year: 2024, Volume and Issue: 13

Published: Feb. 2, 2024

During perception, decoding the orientation of gratings depends on complex interactions between grating, aperture edges, and topographic structure visual map. Here, we aimed to test how biases described during perception affect working memory (WM) decoding. For memoranda, used multiplied by radial angular modulators generate orthogonal for identical orientations. Therefore, if WM representations are simply maintained sensory representations, they would have similar biases. If abstractions features, be unbiased modulator no effect Neural patterns delay period activity while maintaining with one (e.g. radial) were interchangeable other angular) in parietal cortex, suggesting that insensitive perception. Then, visualized stimuli using models field map properties. Regardless biases, both modulated recoded into a single oriented line. These results provide strong evidence percepts, immune perceptual compel revisions theory.

Language: Английский

---

Centering cognitive neuroscience on task demands and generalization

Nature Neuroscience, Journal Year: 2024, Volume and Issue: 27(9), P. 1656 - 1667

Published: July 29, 2024

Language: Английский

---