A computational neuroethology perspective on body and expression perception

Trends in Cognitive Sciences, Journal Year: 2021, Volume and Issue: 25(9), P. 744 - 756

Published: June 16, 2021

Body and expression perception may be sustained by midlevel feature computations rather than body category-selective processes.Body coding in the brain organized statistics of posture movements natural language semantic categories.Midlevel features at stake biological exploit ethological characteristics organism–environment interactions.Midlevel processing on its own sustain rapid action preparation, not require or depend high-order category.Feelings can associated with precursors to conscious emotional states because they are an intermediate layer between unconscious processes fully formed states. Survival prompts organisms prepare adaptive behavior response environmental social threat. However, what specific appearance a conspecific that trigger such behaviors? For species, prime candidates for triggering defense systems visual face body. We propose novel approach studying ability gather survival-relevant information from seeing features. Specifically, we behaviorally relevant bodies expressions is coded levels brain. These relatively independent higher-order cognitive emotions. Instead, our embedded framework mobilizes computational models discovery. Human nonhuman primates experts gathering crucial survival movement perception. Social threat situations reactions them among most studied [1.Roelofs K. Freeze action: neurobiological mechanisms animal human freezing.Philos. Trans. R. Soc. B. 2017; 372: 20160206Crossref PubMed Scopus (5) Google Scholar,2.Terburg D. et al.The basolateral amygdala essential escape: rodent study.Cell. 2018; 175: 723-735Abstract Full Text PDF (47) Scholar]. it specifically about triggers behavior? In intuitive thinking question, typically retrieve one another salient characteristic as head orientation, position arms, overall velocity [3.Poyo Solanas M. role subjective expressions.Sci. Rep. 2020; 10: 6202Crossref (1) So far, very few studies have objectively measured which understanding expressive postures movements, even fewer looked their possible correlates. To understand why this needed, must frame context current research basis Most followed tracks adopting theoretical object category (see Glossary)-selective areas [4.Kanwisher N. Domain specificity perception.Nat. Neurosci. 2000; 3: 759-763Crossref (636) Scholar] top level hierarchical model [5.Bruce V. Young A. Understanding recognition.Br. J. Psychol. 1986; 77: 305-327Crossref Scholar,6.Haxby J.V. distributed neural system perception.Trends Cogn. Sci. 4: 223-233Abstract (3117) framework, emotion dependent successful sketch here different centered notion high-level representation gateway subsequent decoding, but (Figure 1). Midlevel classical low-level (e.g., edges, spatial frequency, motion direction) [7.Giese M.A. Poggio T. Neural recognition movements.Nat. Rev. 2003; 179-192Crossref (625) well intuitively notice believe act upon (i.e., categories emotions, actions, intentions) [8.Grill-Spector Weiner K.S. The functional architecture ventral temporal cortex categorization.Nat. 2014; 15: 536-548Crossref (307) Some examples derived analysis limb contraction Scholar,9.Poyo al.Computation-based brain.Cereb. Cortex. 30: 6376-6390Crossref (0) Scholar], hand distance [10.Zhan al.Subjective actions emotions involves interplay observation networks brain.BioRxiv. 2021; (Published online April 15, 2021. http://dx.doi.org/10.1101/2021.04.15.439961)PubMed Recent identified correlates agentic [11.Haxby al.Naturalistic stimuli reveal dominant representation.NeuroImage. 216: 116561Crossref (4) animacy [12.Thorat S. nature organization cortex.eLife. 2019; 8e47142Crossref sociality [13.Tarhan L. Konkle Sociality interaction envelope organize representations.Nat. Commun. 11: 3002Crossref concepts used validated perception, turn out reduce to, emerge from, computations. goal provide characterization whole-body naturalistic contexts. Notions maps [14.Graziano M.S. Ethological maps: paradigm shift motor cortex.Trends 2016; 20: 121-132Abstract (75) domains [15.Kaas J.H. al.Cortical ethologically behaviors primates.Am. Primatol. 2013; 75: 407-414Crossref already proved useful characterizing domain, has been proposed low- representations [16.Wurm M.F. Lingnau Decoding abstraction.J. 2015; 35: 7727-7735Crossref (80) Systematic discovery needs combined methods counter, hand, naïve observer bias and, other, dimensionality explosion unconstrained networks. Studies consistently described [17.Downing P.E. al.A cortical area selective body.Science. 2001; 293: 2470-2473Crossref (1281) Scholar,18.Peelen M.V. Downing Selectivity fusiform gyrus.J. Neurophysiol. 2005; 93: 603-608Crossref (420) patches monkey [19.Pinsk al.Representations faces parts macaque cortex: MRI study.PNAS. 102: 6996-7001Crossref images. humans, originally later two were reported: extrastriate (EBA) middle occipital gyrus/middle gyrus Scholar,20.van de Riet W.A. al.Specific common regions involved expressions.Soc. 2009; 101-120Crossref (83) (FBA) [18.Peelen Scholar,21.Hadjikhani Gelder Seeing fearful activates amygdala.Curr. Biol. 13: 2201-2205Abstract (208) respective roles EBA FBA still understood. It suggested more whereas biased towards images [22.Taylor J.C. al.Functional part areas.J. 2007; 98: 1626-1633Crossref (185) More recent evidence indicates also encode details pertaining shape, posture, [23.Downing Peelen occipitotemporal body-selective person perception.Cogn. 2011; 2: 186-203Crossref (112) lack clarity concerning functions related anatomical complexity. example, there substantial overlap complex (hMT+), makes difficult determine actual involvement former [24.Ross P.D. form dissociable pathways.Front. 5: 767Crossref (3) Scholar,25.Vangeneugden al.Distinct discriminations.J. 34: 574-585Crossref addition, single area, illustrated landmarks, field maps, stimulus comparisons [26.Weiner Grill-Spector Not area: using hMT+, parcellate limb-selective activations lateral cortex.NeuroImage. 56: 2183-2199Crossref Although clearly supports [21.Hadjikhani Scholar,27.Grèzes al.Perceiving fear dynamic expressions.NeuroImage. 959-967Crossref Scholar,28.Pichon al.Threat defensive responses independently attentional control.Cereb. 2012; 22: 274-285Crossref clear whether important recognition, depending emotion. shown modulates activity FBA, although no difference found other [29.Peelen al.Emotional modulation areas.Soc. Affect. 274-283Crossref Emotion-specific differences connectivity patterns 2C ). Interestingly, along these lines, fact seems sensitive sense point view interface perceptual [30.Zimmermann al.Is dorsal visuomotor stream?.Brain Struct. Funct. 223: 31-46Crossref addition areas, first magnetic resonance imaging (fMRI) showed 2) [27.Grèzes Scholar,31.de al.Fear fosters flight: mechanism contagion when perceiving expressed whole body.PNAS. 2004; 101: 16701-16706Crossref Scholar,32.Goldberg H. emotion–action link? Naturalistic preferentially activate stream.NeuroImage. 84: 254-264Crossref network shows increased threatening neutral Scholar,33.Pichon al.Two Comparing recognizing anger 47: 1873-1883Crossref (115) system, responsible plays role, especially case [31.de Scholar,34.Borgomaneri al.Early changes corticospinal excitability 14122Crossref (20) Scholar, 35.Hortensius al.When dominates mind: threat.Psychophysiology. 53: 1307-1316Crossref (14) 36.Meeren H.K.M. preferential right stream – MEG study.Sci. 6: 24831Crossref (16) 37.Borgomaneri al.Behavioral inhibition sensitivity enhances suppression watching expressions.Brain 222: 3267-3282Crossref do subcortical [9.Poyo Scholar,38.Utter A.A. Basso basal ganglia: overview circuits function.Neurosci. Biobehav. 2008; 32: 333-342Crossref (119) cerebellum [39.Sokolov al.Brain signaling absence language.PNAS. 117: 20868-20873Crossref A pathway pulvinar, superior colliculus, interacts support reflexes withdrawal, freezing, startle) [40.de Towards neurobiology language.Nat. 2006; 7: 242-249Crossref (460) Scholar,41.Dean P. al.Event emergency? Two mammalian colliculus.Trends 1989; 12: 137-147Abstract (397) does so [28.Pichon particular, supporting pivotal assignment affective value incoming stimuli, preparation behaviors, modulating attentional, perceptual, [42.Emery N.J. Amaral D.G. primate cognition.in: Lane R.D. Nadel Cognitive Neurocience Emotion. Oxford University Press, 2000: 156-191Google over decade, implicitly assumed constitute various attributes, same way Scholar,43.Shallice From Neuropsychology Mental Structure. Cambridge 1988Crossref 44.Kanwisher Yovel G. region specialized faces.Philos. 361: 2109-2128Crossref 45.Kanwisher module perception.J. 1997; 17: 4302-4311Crossref 46.Peelen 8: 636-648Crossref With gradual category-based models, encapsulated loosened. Certainly, available equally outside [32.Goldberg Scholar,40.de attributed needed all open question. Functional assume represent [47.Van Essen D.C. Maunsell Hierarchical streams 1983; 370-375Abstract (523) Scholar,48.Josephs E.L. Large-scale dissociations views objects, scenes, reachable-scale environments cortex.PNAS. 29354-29362Crossref stable, task detection, and/or attribute identification, passive viewing, explicit recognition) attributes emotion, gender) Scholar,49.Kanwisher quest FFA where led.J. 37: 1056-1061Crossref (42) growing showing factors significantly impact including (Box attention-related increases preferred during search tasks [50.Çukur al.Attention vision warps across brain.Nat. 16: 763-770Crossref (184) Scholar,51.Peelen al.Neural scene categorization cortex.Nature. 460: 94-97Crossref Scholar].Box 1A areasHierarchical each elaborates previous [48.Josephs task-related gender). some reported variability how represented indicating abstract, static, high-level, conceptual often assumed. facial known that, identical type influences active vs observation, orthogonal task) [116.Gur R.C. activation processing.NeuroImage. 2002; 651-662Crossref (256) 117.Habel U. al.Amygdala expressions: discrimination versus implicit processing.Neuropsychologia. 45: 2369-2377Crossref (110) 118.Hariri A.R. al.Neocortical stimuli.Biol. Psychiatry. 494-501Abstract (615) 119.Winston al.Common distinct direct incidental multiple emotions.NeuroImage. 84-97Crossref (297) There similar findings effects literature. emotion-naming compared color-naming Using multivariate pattern decoded high accuracy [120.Marrazzo dynamics visual, prefrontal parietal cortex.BioRxiv. July 2020. https://doi.org/10.1101/2020.07.14.202515)Google Similar strong task-driven electroencephalography (EEG) [121.Jessen Kotz S.A. vocal, facial, bodily 58: 665-674Crossref (89) Scholar].Substantial variations linked attention awareness. Along modulate processes, observed example differential saccades [122.Milders al.Detection modulated direction eye gaze.Emotion. 1456Crossref (38) hemispatial neglect patients, contralesional presentation overcomes deficits [123.Tamietto al.Once you feel it, see it: insula sensory-motor contribution awareness neglect.Cortex. 62: 56-72Crossref intact brain, consciously, opposed non-consciously, viewed major frontoparietal [61.Zhan al.Ventral pathways relate differently under continuous flash suppression.eNeuro. 5 (ENEURO.0285-17.2017)Crossref Scholar,124.de influence short review.Front. Integr. 54Crossref Scholar,125.Zhan tool angry awareness.PLoS One. 10e0139768Crossref without EBA, patients full bilateral lesion [126.Van den Stock al.Body patient primary lesions.Biol. e31-e33Abstract This category-specific receive input through V1-independent 2). shape sufficient drive responses, know partial feature-based image non-conscious percept. Substantial

Language: Английский

---

Feature-space selection with banded ridge regression

NeuroImage, Journal Year: 2022, Volume and Issue: 264, P. 119728 - 119728

Published: Nov. 8, 2022

Encoding models provide a powerful framework to identify the information represented in brain recordings. In this framework, stimulus representation is expressed within feature space and used regularized linear regression predict activity. To account for potential complementarity of different spaces, joint model fit on multiple spaces simultaneously. adapt regularization strength each space, ridge extended banded regression, which optimizes hyperparameter per space. The present paper proposes method decompose over variance explained by model. It also describes how performs feature-space selection, effectively ignoring non-predictive redundant spaces. This selection leads better prediction accuracy interpretability. Banded then mathematically linked number other methods with similar mechanisms. Finally, several are proposed address computational challenge fitting regressions large numbers voxels All implementations released an open-source Python package called Himalaya.

Language: Английский

---

The Dorsal Visual Pathway Represents Object-Centered Spatial Relations for Object Recognition

Journal of Neuroscience, Journal Year: 2022, Volume and Issue: 42(23), P. 4693 - 4710

Published: May 4, 2022

Although there is mounting evidence that input from the dorsal visual pathway crucial for object processes in ventral pathway, specific functional contributions of cortex to these remain poorly understood. Here, we hypothesized computes spatial relations among an object9s parts, a process forming global shape percepts, and transmits this information support categorization. Using fMRI with human participants (females males), discovered regions intraparietal sulcus (IPS) were selectively involved computing object-centered part relations. These exhibited task-dependent effective connectivity cortex, distinct other regions, such as those representing allocentric relations, 3D shape, tools. In subsequent experiment, found multivariate response posterior (p)IPS, defined on basis part-relations, could be used decode category at levels comparable regions. Moreover, mediation analyses further suggested IPS may account representations pathway. Together, our results highlight recognition. We suggest source ability categorize objects shape. SIGNIFICANCE STATEMENT Humans novel rapidly effortlessly. Such categorization achieved by structure, is, parts. Yet, despite their importance, it unclear how are represented neurally. computed which typically implicated visuospatial processing. fMRI, identified selective cortex. can categorization, even mediate region thought findings shed light broader network brain

Language: Английский

---

Multiple cortical visual streams in humans

Cerebral Cortex, Journal Year: 2022, Volume and Issue: 33(7), P. 3319 - 3349

Published: July 14, 2022

The effective connectivity between 55 visual cortical regions and 360 was measured in 171 HCP participants using the HCP-MMP atlas, complemented with functional diffusion tractography. A Ventrolateral Visual "What" Stream for object face recognition projects hierarchically to inferior temporal cortex, which orbitofrontal cortex reward value emotion, hippocampal memory system. Ventromedial "Where" scene representations connects parahippocampal gyrus hippocampus. An Inferior STS (superior sulcus) Semantic receives from Stream, parietal PGi, ventromedial-prefrontal system language systems. Dorsal via V2 V3A MT+ Complex (including MT MST), connect intraparietal LIP, VIP MIP) involved motion actions space. It performs coordinate transforms idiothetic update of representations. Superior inputs STV, auditory A5, is activated by expression, vocalization, important social behaviour,

Language: Английский

---

Texture-like representation of objects in human visual cortex

Proceedings of the National Academy of Sciences, Journal Year: 2022, Volume and Issue: 119(17)

Published: April 19, 2022

Significance Humans are exquisitely sensitive to the spatial arrangement of visual features in objects and scenes, but not textures. Category-selective regions cortex widely believed underlie object perception, suggesting such should distinguish natural images from synthesized containing similar scrambled arrangements. Contrarily, we demonstrate that representations category-selective do discriminate feature-matched scrambles can different categories, a texture-like encoding. We find insensitivity feature Imagenet-trained deep convolutional neural networks. This suggests need reconceptualize role as representing basis set complex features, useful for myriad behaviors.

Language: Английский

---

Modelling continual learning in humans with Hebbian context gating and exponentially decaying task signals

PLoS Computational Biology, Journal Year: 2023, Volume and Issue: 19(1), P. e1010808 - e1010808

Published: Jan. 19, 2023

Humans can learn several tasks in succession with minimal mutual interference but perform more poorly when trained on multiple at once. The opposite is true for standard deep neural networks. Here, we propose novel computational constraints artificial networks, inspired by earlier work gating the primate prefrontal cortex, that capture cost of interleaved training and allow network to two sequence without forgetting. We augment stochastic gradient descent algorithmic motifs, so-called "sluggish" task units a Hebbian step strengthens connections between hidden encode task-relevant information. found introduce switch-cost during training, which biases representations under towards joint representation ignores contextual cue, while promotes formation scheme from layer produces orthogonal are perfectly guarded against interference. Validating model previously published human behavioural data revealed it matches performance participants who had been blocked or curricula, these differences were driven misestimation category boundary.

Language: Английский

---

Drawing as a versatile cognitive tool

Nature Reviews Psychology, Journal Year: 2023, Volume and Issue: 2(9), P. 556 - 568

Published: July 17, 2023

Language: Английский

---

Dissecting neural computations in the human auditory pathway using deep neural networks for speech

Nature Neuroscience, Journal Year: 2023, Volume and Issue: 26(12), P. 2213 - 2225

Published: Oct. 30, 2023

Abstract The human auditory system extracts rich linguistic abstractions from speech signals. Traditional approaches to understanding this complex process have used linear feature-encoding models, with limited success. Artificial neural networks excel in recognition tasks and offer promising computational models of processing. We representations state-of-the-art deep network (DNN) investigate coding the nerve cortex. Representations hierarchical layers DNN correlated well activity throughout ascending system. Unsupervised performed at least as other purely supervised or fine-tuned models. Deeper were better higher-order cortex, computations aligned phonemic syllabic structures speech. Accordingly, trained on either English Mandarin predicted cortical responses native speakers each language. These results reveal convergence between model biological pathway, offering new for modeling

Language: Английский

---

Brain-optimized deep neural network models of human visual areas learn non-hierarchical representations

Nature Communications, Journal Year: 2023, Volume and Issue: 14(1)

Published: June 7, 2023

Deep neural networks (DNNs) optimized for visual tasks learn representations that align layer depth with the hierarchy of areas in primate brain. One interpretation this finding is hierarchical are necessary to accurately predict brain activity system. To test interpretation, we DNNs directly measured fMRI human V1-V4. We trained a single-branch DNN all four jointly, and multi-branch each area independently. Although it was possible representations, only did so. This result shows not V1-V4, encode brain-like may differ widely their architecture, ranging from strict serial hierarchies multiple independent branches.

Language: Английский

---

Many but not all deep neural network audio models capture brain responses and exhibit correspondence between model stages and brain regions

PLoS Biology, Journal Year: 2023, Volume and Issue: 21(12), P. e3002366 - e3002366

Published: Dec. 13, 2023

Models that predict brain responses to stimuli provide one measure of understanding a sensory system and have many potential applications in science engineering. Deep artificial neural networks emerged as the leading such predictive models visual but are less explored audition. Prior work provided examples audio-trained produced good predictions auditory cortical fMRI exhibited correspondence between model stages regions, left it unclear whether these results generalize other network and, thus, how further improve this domain. We evaluated model-brain for publicly available audio along with in-house trained on 4 different tasks. Most tested outpredicted standard spectromporal filter-bank cortex systematic correspondence: Middle best predicted primary cortex, while deep non-primary cortex. However, some state-of-the-art substantially worse predictions. recognize speech background noise better than quiet, potentially because hearing imposes constraints biological representations. The training task influenced prediction quality specific tuning properties, overall resulting from multiple generally support promise audition, though they also indicate current do not explain their entirety.

Language: Английский

---