Convolutional networks can model the functional modulation of MEG responses during reading DOI Creative Commons
Marijn van Vliet,

Oona Rinkinen,

Takao Shimizu

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2022, Volume and Issue: unknown

Published: Feb. 10, 2022

Abstract Neuroimaging studies have provided a wealth of information about when and where changes in brain activity might be expected during reading. We sought to better understand the computational steps that give rise such task-related modulations neural by using convolutional network model macro-scale computations necessary perform single-word recognition. presented with stimuli had been shown human volunteers an earlier magnetoencephalography (MEG) experiment evaluated whether same experimental effects could observed both model. In direct comparison between MEG recordings, accurately predicted amplitude three evoked response components commonly contrast traditional models reading, our directly operates on pixel values image containing text. This allowed us simulate whole gamut processing from detection segmentation letter shapes word-form identification, deep learning architecture facilitating inclusion large vocabulary 10k Finnish words. Interestingly, key achieving desired behavior was use noisy activation function for units as well obey word frequency statistics repeating training. conclude techniques revolutionized object recognition can also create reading straightforwardly compared neuroimaging data, which will greatly facilitate testing refining theories language brain.

Language: Английский

GPT-3 reveals selective insensitivity to global vs. local linguistic context in speech produced by treatment-naive patients with positive thought disorder DOI

Victoria Sharpe,

Michael Mackinley,

Samer Nour Eddine

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: July 9, 2024

Abstract Background Early psychopathologists proposed that certain features of positive thought disorder, the disorganized language output produced by some people with schizophrenia, suggest an insensitivity to global, relative local, discourse context. This idea has received support from carefully controlled psycholinguistic studies in comprehension. In production, researchers have so far remained reliant on subjective qualitative rating scales assess and understand speech disorganization. Now, however, recent advances large models mean it is possible quantify sensitivity global local context objectively probing lexical probability (the predictability a word given its preceding context) during natural production. Methods For each 60 first-episode psychosis patients 35 healthy, demographically-matched controls, we extracted probabilities GPT-3 based contexts ranged very local— single word: P(Wn | Wn-1)—to global— up 50 words: P(Wn|Wn-50, Wn-49, …, Wn-1). Results We show, for first time, characterized disproportionate versus linguistic Critically, this global-versus-local selectively predicted clinical ratings above beyond overall symptom severity. There was no evidence relationship negative disorder (impoverishment). Conclusions provide automated, interpretable measure can potentially be used disorganization schizophrenia. Our findings directly link phenomenology neurocognitive constructs are grounded theory neurobiology.

Language: Английский

Citations

3
Understanding words in context: A naturalistic EEG study of children’s lexical processing DOI Creative Commons
Tatyana Levari, Jesse Snedeker

Journal of Memory and Language, Journal Year: 2024, Volume and Issue: 137, P. 104512 - 104512

Published: March 8, 2024

When listening to speech, adults rely on context anticipate upcoming words. Evidence for this comes from studies demonstrating that the N400, an event-related potential (ERP) indexes ease of lexical-semantic processing, is influenced by predictability a word in context. We know far less about role children's speech comprehension. The present study explored lexical processing and 5-10-year-old children as they listened story. ERPs time-locked onset every were recorded. Each content was coded frequency, semantic association, predictability. In both adults, N400s reflect predictability, even when controlling frequency association. These findings suggest use top-down constraints words stories.

Language: Английский

Citations

2
Can prediction error explain predictability effects on the N1 during picture-word verification? DOI Creative Commons
Jack E. Taylor, Guillaume A. Rousselet, Sara C. Sereno

et al.

Imaging Neuroscience, Journal Year: 2024, Volume and Issue: 2, P. 1 - 24

Published: March 22, 2024

Abstract Do early effects of predictability in visual word recognition reflect prediction error? Electrophysiological research investigating processing has demonstrated the N1, or first negative component event-related potential (ERP). However, findings regarding magnitude and interactions with lexical variables have been inconsistent. Moreover, past studies typically used categorical designs relatively small samples relied on by-participant analyses. Nevertheless, reports generally shown that predicted words elicit less negative-going (i.e., lower amplitude) N1s, a pattern consistent simple predictive coding account. In our preregistered study, we tested this account via interaction between certainty. A picture-word verification paradigm was implemented which pictures were followed by tightly matched picture-congruent picture-incongruent written nouns. The target (picture-congruent) nouns manipulated continuously based norms association picture its name. ERPs from 68 participants revealed opposite to expected under framework.

Language: Английский

Citations

1
Unpleasant words can affect the detection of morphosyntactic errors: An ERP study on individual differences DOI Creative Commons
Lucía Vieitez, Isabel Padrón, Marcos Díaz‐Lago

et al.

Psychophysiology, Journal Year: 2024, Volume and Issue: 61(12)

Published: July 31, 2024

Abstract In recent years, several ERP studies have investigated whether the early computation of agreement is permeable to emotional content words. Some reported interactive effects grammaticality and emotionality in left anterior negativity (LAN) component, while others failed replicate these results. Furthermore, novel findings suggest that grammatical processing can elicit different neural patterns across individuals. this study, we aim investigate interaction between restricted participants with a specific profile. Sixty‐one female native speakers Spanish performed an judgment task noun phrases composed determiner, noun, unpleasant or neutral adjective could agree disagree gender preceding noun. Our results support existence two brain profiles: negative positive dominance (individuals showing either larger LAN P600 amplitudes ungrammatical stimuli than ones, respectively). Interestingly, pattern groups diverged at points along time course. Thus, group showed as 200 ms, parallel autonomous LAN/N400 window. Instead, for was found around evidencing effect emerged only confirm role individual differences interplay grammar emotion level call inclusion perspective on syntactic processing.

Language: Английский

Citations

1
Convergent neural signatures of speech prediction error are a biological marker for spoken word recognition DOI Creative Commons
Ediz Sohoglu, Loes Beckers, Matthew H. Davis

et al.

Nature Communications, Journal Year: 2024, Volume and Issue: 15(1)

Published: Nov. 18, 2024

Abstract We use MEG and fMRI to determine how predictions are combined with speech input in superior temporal cortex. compare neural responses words which first syllables strongly or weakly predict second (e.g., “bingo”, “snigger” versus “tango”, “meagre”). further the same when mismatch during pseudoword perception “snigo” “meago”). Neural representations of suppressed by strong match sensory but show opposite effect mismatch. Computational simulations that this interaction is consistent prediction error not alternative (sharpened signal) computations. signatures observed 200 ms after syllable onset early auditory regions (bilateral Heschl’s gyrus STG). These findings demonstrate computations identification familiar spoken unfamiliar pseudowords.

Language: Английский

Citations

1
An implemented predictive coding model of lexico-semantic processing explains the dynamics of univariate and multivariate activity within the left ventromedial temporal lobe during reading comprehension DOI Creative Commons
Lin Wang,

Samer Nour Eddine,

Trevor Brothers

et al.

NeuroImage, Journal Year: 2024, Volume and Issue: unknown, P. 120977 - 120977

Published: Dec. 1, 2024

Language: Английский

Citations

0
Selectivity for “Non-Food” versus “Food” Nouns Is Increased in Healthy Adults in Response to Elevated Peripheral Blood Glucose Levels as Indicated by Event-Related Potentials (ERPs) DOI
V. A. Ivanov, О. В. Кручинина,

Yu. A. Chiligina

et al.

Journal of Evolutionary Biochemistry and Physiology, Journal Year: 2024, Volume and Issue: 60(6), P. 2369 - 2380

Published: Nov. 1, 2024

Language: Английский

Citations

0
Convolutional networks can model the functional modulation of MEG responses during reading DOI Creative Commons
Marijn van Vliet,

Oona Rinkinen,

Takao Shimizu

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2022, Volume and Issue: unknown

Published: Feb. 10, 2022

Abstract Neuroimaging studies have provided a wealth of information about when and where changes in brain activity might be expected during reading. We sought to better understand the computational steps that give rise such task-related modulations neural by using convolutional network model macro-scale computations necessary perform single-word recognition. presented with stimuli had been shown human volunteers an earlier magnetoencephalography (MEG) experiment evaluated whether same experimental effects could observed both model. In direct comparison between MEG recordings, accurately predicted amplitude three evoked response components commonly contrast traditional models reading, our directly operates on pixel values image containing text. This allowed us simulate whole gamut processing from detection segmentation letter shapes word-form identification, deep learning architecture facilitating inclusion large vocabulary 10k Finnish words. Interestingly, key achieving desired behavior was use noisy activation function for units as well obey word frequency statistics repeating training. conclude techniques revolutionized object recognition can also create reading straightforwardly compared neuroimaging data, which will greatly facilitate testing refining theories language brain.

Language: Английский

Citations

0