Converging intracortical signatures of two separated processing timescales in human early auditory cortex

Converging intracortical signatures of two separated processing timescales in human early auditory cortex Baroni Morillon Trébuchon Liégeois-Chauvel Olasagasti Giraud

Tag layer of the DHSS Meta-Tagging Project. The article text is reproduced from the staged copy named in sourceDesc and is not altered; the tags are held separately in standOff so that overlapping and nested spans are preserved.

Converging intracortical signatures of two separated processing timescales in human early auditory cortex 2019 bioRxiv 10.1101/730002 baroni2019-converging-intracortical-signature baroni2019-converging-intracortical-signature.txt, the project's staged copy bc5b65d8f7ee481cbbdbd13ea82412e91e25853a open-access neuroscience

the paper's own words

the tag layer; a sentence written by the tagger about the paper, always anchored to a quotation but not itself a quotation

0 spans located by searching the stored text for the quote's exact words, allowing any whitespace between them existence Quotation withheld because this record is not cleared for redistribution. dual-timescale speech processing in human early auditory cortex Quotation withheld because this record is not cleared for redistribution. electrophysiology electrophysiology Quotation withheld because this record is not cleared for redistribution. mechanisms Synaptic interactions between excitatory and inhibitory populations operating at theta and gamma timescales, coupled by a negative feedback loop, generate the two-peaked (delta-theta and gamma) spectral profile of information content. Quotation withheld because this record is not cleared for redistribution. mechanisms Short-term synaptic depression of the bottom-up pathway depletes available neurotransmitter as the speech envelope arrives, producing large onset responses and attenuated responses to later syllables. Quotation withheld because this record is not cleared for redistribution. mechanisms The beta peak in early auditory cortex is inherited top-down from higher cortical areas rather than generated locally, which is why a single-area microcircuit model cannot reproduce it. Quotation withheld because this record is not cleared for redistribution. mechanisms Excessive self-excitation drives the network into strong baseline synchronisation and saturating oscillatory responses, so different speech segments elicit near-identical patterns and stimulus information is lost. Quotation withheld because this record is not cleared for redistribution. Successful decoding does not establish that the decoded activity plays a functional role in speech processing. what this limits: any functional claim built on the decoding: WE ACKNOWLEDGE THAT OUR DECODING RESULTS DO NOT UNAMBIGUOUSLY IMPLY A CORRESPONDING FUNCTIONAL ROLE IN SPEECH PROCESSING Quotation withheld because this record is not cleared for redistribution. The study is correlational; establishing causal contributions would require intervention, which was not done. what this limits: any causal reading: CAUSAL CONTRIBUTIONS ARE BEST EXPLORED WITH INTERVENTIONAL APPROACHES, WHICH ARE BEYOND THE SCOPE OF THIS WORK Quotation withheld because this record is not cleared for redistribution. The model fails to reproduce the baseline beta power peak and its speech-induced reduction. what this limits: the model's account of the spectral profile: A KEY FEATURE THAT OUR MODEL DOES NOT REPRODUCE IS THE PRESENCE OF A POWER PEAK IN THE BETA RANGE, WHICH IS REDUCED BY SPEECH STIMULATION Quotation withheld because this record is not cleared for redistribution. The model does not capture the full inter-trial variability of the iEEG data. what this limits: the model's account of trial-to-trial variability: OUR RATE MODEL ... IS NOT EXPECTED TO FULLY CAPTURE THE EXTENT AND COMPLEXITY OF THE INTER-TRIAL VARIABILITY IN THE iEEG DATASET Quotation withheld because this record is not cleared for redistribution. Where the study can record was decided by clinical need, not by the research question. what this limits: the spatial coverage of these recordings as a sample of auditory cortex: Quotation withheld Quotation withheld because this record is not cleared for redistribution. A known contaminant is acknowledged and then set aside by expectation rather than by measurement. what this limits: these signals as free of epileptiform contamination: the authors state isolated interictal discharges could have occurred and that their properties are EXPECTED to be unrelated to the variables of interest Quotation withheld because this record is not cleared for redistribution. The two significance thresholds are not estimated the same way: the stricter one required pooling null values across decoding types. about this quote: THE MATHS LAYER OF THE STAGED TEXT WAS MOJIBAKE AND HAS BEEN REPAIRED. This paper's extraction rendered '=' as U+00BC (the vulgar fraction one quarter) and '+' as U+00FE (thorn), so the quote read 'p ¼ 0.01' where the paper says 'p = 0.01'. It passed every check we own, because the staged text said the same wrong thing: the offsets were right and the recorded hash of the staged text matched. Repaired only where the character stands in an operator's place - spaced, with a number after it - because the same U+00BC is a REAL QUARTER in six other papers ('a ¼ inch condenser microphone'). Original in evidence as extracted, pre-repair file in an internal file. · what this limits: comparing results reported at p = 0.01 with those at p = 0.001 in this paper: the first threshold was estimated per classification and the second required pooling null values across decoding types Quotation withheld because this record is not cleared for redistribution. The paper's stricter significance threshold is estimable only by pooling null values across decoding types, so p = 0.001 and p = 0.01 in this paper are not two points on one scale but two differently-constructed tests. about this quote: THE MATHS LAYER OF THE STAGED TEXT WAS MOJIBAKE AND HAS BEEN REPAIRED. This paper's extraction rendered '=' as U+00BC (the vulgar fraction one quarter) and '+' as U+00FE (thorn), so the quote read 'p ¼ 0.01' where the paper says 'p = 0.01'. It passed every check we own, because the staged text said the same wrong thing: the offsets were right and the recorded hash of the staged text matched. Repaired only where the character stands in an operator's place - spaced, with a number after it - because the same U+00BC is a REAL QUARTER in six other papers ('a ¼ inch condenser microphone'). Original in evidence as extracted, pre-repair file in an internal file. Quotation withheld because this record is not cleared for redistribution. The fast-to-slow negative feedback loop is adopted as a circuit motif because control-theoretic arguments identify it as a building block giving both sensitivity and stability. Quotation withheld because this record is not cleared for redistribution. Regularized Least-Square Classifiers, explicitly described as a machine learning technique, are used to quantify how much stimulus information the neural signal carries. Quotation withheld because this record is not cleared for redistribution. stimuli delivered in a pseudo-randomised order at a fixed comfortable hearing level across both protocols eliminates: order effects and adaptation - a fixed sequence would confound stimulus identity with position in the run · outcome (stated by the tagger, no quotation): held Quotation withheld because this record is not cleared for redistribution. Electrode location Electrode location Quotation withheld because this record is not cleared for redistribution. Significance thresholds Significance thresholds about this quote: THE MATHS LAYER OF THE STAGED TEXT WAS MOJIBAKE AND HAS BEEN REPAIRED. This paper's extraction rendered '=' as U+00BC (the vulgar fraction one quarter) and '+' as U+00FE (thorn), so the quote read 'p ¼ 0.01' where the paper says 'p = 0.01'. It passed every check we own, because the staged text said the same wrong thing: the offsets were right and the recorded hash of the staged text matched. Repaired only where the character stands in an operator's place - spaced, with a number after it - because the same U+00BC is a REAL QUARTER in six other papers ('a ¼ inch condenser microphone'). Original in evidence as extracted, pre-repair file in an internal file. Quotation withheld because this record is not cleared for redistribution. interictal epileptiform discharges interictal epileptiform discharges Quotation withheld because this record is not cleared for redistribution. decoding decoding about this quote: THE MATHS LAYER OF THE STAGED TEXT WAS MOJIBAKE AND HAS BEEN REPAIRED. This paper's extraction rendered '=' as U+00BC (the vulgar fraction one quarter) and '+' as U+00FE (thorn), so the quote read 'p ¼ 0.01' where the paper says 'p = 0.01'. It passed every check we own, because the staged text said the same wrong thing: the offsets were right and the recorded hash of the staged text matched. Repaired only where the character stands in an operator's place - spaced, with a number after it - because the same U+00BC is a REAL QUARTER in six other papers ('a ¼ inch condenser microphone'). Original in evidence as extracted, pre-repair file in an internal file. Quotation withheld because this record is not cleared for redistribution. auditory cortex auditory cortex Quotation withheld because this record is not cleared for redistribution. dual timescale processing infrastructure dual timescale processing infrastructure Quotation withheld because this record is not cleared for redistribution. information content information content Quotation withheld because this record is not cleared for redistribution. speech detection speech detection Quotation withheld because this record is not cleared for redistribution. speech discrimination speech discrimination Quotation withheld because this record is not cleared for redistribution. phase information phase information Quotation withheld because this record is not cleared for redistribution. power information power information Quotation withheld because this record is not cleared for redistribution. simple rate model simple rate model Quotation withheld because this record is not cleared for redistribution. negative feedback loop negative feedback loop Quotation withheld because this record is not cleared for redistribution. fast subnetwork fast subnetwork Quotation withheld because this record is not cleared for redistribution. slow subnetwork slow subnetwork Quotation withheld because this record is not cleared for redistribution. auditory perception auditory perception Quotation withheld because this record is not cleared for redistribution. scale-free process scale-free process Quotation withheld because this record is not cleared for redistribution. preferred timescales preferred timescales Quotation withheld because this record is not cleared for redistribution. delta-theta delta-theta Quotation withheld because this record is not cleared for redistribution. gamma gamma Quotation withheld because this record is not cleared for redistribution. fast subnetwork → slow subnetwork The model couples the fast and slow subnetworks through a negative feedback loop. Coupling-by-feedback is not is-a, part-of, same-as, distinct-from, contrasted-with, operationalised-as, causes, or renamed-across-fields. Quotation withheld because this record is not cleared for redistribution. significance threshold - AT p = 0.01: estimated separately for each classification considered about this quote: THE MATHS LAYER OF THE STAGED TEXT WAS MOJIBAKE AND HAS BEEN REPAIRED. This paper's extraction rendered '=' as U+00BC (the vulgar fraction one quarter) and '+' as U+00FE (thorn), so the quote read 'p ¼ 0.01' where the paper says 'p = 0.01'. It passed every check we own, because the staged text said the same wrong thing: the offsets were right and the recorded hash of the staged text matched. Repaired only where the character stands in an operator's place - spaced, with a number after it - because the same U+00BC is a REAL QUARTER in six other papers ('a ¼ inch condenser microphone'). Original in evidence as extracted, pre-repair file in an internal file. Quotation withheld because this record is not cleared for redistribution. significance threshold - AT p = 0.001: estimated from null values POOLED across decoding types, because the per-classification estimate was not good enough at that level about this quote: THE MATHS LAYER OF THE STAGED TEXT WAS MOJIBAKE AND HAS BEEN REPAIRED. This paper's extraction rendered '=' as U+00BC (the vulgar fraction one quarter) and '+' as U+00FE (thorn), so the quote read 'p ¼ 0.01' where the paper says 'p = 0.01'. It passed every check we own, because the staged text said the same wrong thing: the offsets were right and the recorded hash of the staged text matched. Repaired only where the character stands in an operator's place - spaced, with a number after it - because the same U+00BC is a REAL QUARTER in six other papers ('a ¼ inch condenser microphone'). Original in evidence as extracted, pre-repair file in an internal file. Quotation withheld because this record is not cleared for redistribution. model failure - treated as INFORMATIVE rather than as a defect: the features the model does not reproduce are said to highlight the limitations of the approach Quotation withheld because this record is not cleared for redistribution. model failure - treated as INFORMATIVE rather than as a defect: features the model fails to reproduce are said to highlight the limitations of the modelling approach, so a mismatch is read as a finding about the model Quotation withheld because this record is not cleared for redistribution. Dual timescale processing infrastructure - convergent auditory-cortex structure supported by spectral analysis and decoding Quotation withheld because this record is not cleared for redistribution. Information content - bimodal spectral decoding at delta-theta and gamma rather than intermediate beta frequencies Quotation withheld because this record is not cleared for redistribution. Speech detection - distinguishing speech from pre-stimulus neural activity Quotation withheld because this record is not cleared for redistribution. Speech discrimination - distinguishing neural patterns for different speech segments Quotation withheld because this record is not cleared for redistribution. Phase information - generally decodes better than power information, with detection-only exceptions Quotation withheld because this record is not cleared for redistribution. Power information - the comparison baseline that phase generally outperforms Quotation withheld because this record is not cleared for redistribution. Simple rate model - two different-timescale excitatory/inhibitory subnetworks connected by negative feedback Quotation withheld because this record is not cleared for redistribution. Negative feedback loop - reciprocal fast/slow subnetwork connectivity from Ge to Te and Te to Gi Quotation withheld because this record is not cleared for redistribution. Fast subnetwork - model subsystem with a natural frequency of 85.9 Hz Quotation withheld because this record is not cleared for redistribution. Slow subnetwork - model subsystem with a natural frequency of 3.8 Hz Quotation withheld because this record is not cleared for redistribution. Auditory perception - evidence suggests two separated frequency bands rather than scale-free dynamics Quotation withheld because this record is not cleared for redistribution. Scale-free process - the auditory-processing description rejected by the two-band evidence Quotation withheld because this record is not cleared for redistribution. Preferred timescales - ranges where sensory responses dominate and input information is encoded more effectively Quotation withheld because this record is not cleared for redistribution. Delta-theta - low-frequency peak of speech information decoding Quotation withheld because this record is not cleared for redistribution. Gamma - high-frequency peak of speech information decoding Quotation withheld because this record is not cleared for redistribution. contradicts → the scale-free account of ECoG/iEEG signals (Miller et al., 2009; He et al., 2010; He, 2014; Podvalny et al., 2015) Quotation withheld because this record is not cleared for redistribution. builds-on → Lakatos et al., 2005 Quotation withheld because this record is not cleared for redistribution. extends → Doelling et al. (2019) rate network model Quotation withheld because this record is not cleared for redistribution. replicates → prior reports of phase information outperforming power (Luo and Poeppel, 2007; Howard and Poeppel, 2010; Ng et al., 2013; Gross et al., 2013 and others) Quotation withheld because this record is not cleared for redistribution. speech detection —distinct-from→ speech discrimination speech detection speech discrimination the relation itself is separately evidenced · strength of the relation: 1 · note: The paper separately defines detection as speech-vs-baseline and discrimination as segment-vs-segment, using a 'not only ... but also' construction. Quotation withheld because this record is not cleared for redistribution. auditory perception —distinct-from→ scale-free process auditory perception scale-free process the relation itself is separately evidenced · strength of the relation: 0.9 · note: The paper says multiple pieces of evidence suggest the non-scale-free result, so weight 0.9 preserves that hedge. Quotation withheld because this record is not cleared for redistribution. Wilcoxon signed-rank test Wilcoxon signed-rank test Quotation withheld because this record is not cleared for redistribution. effect size effect size Quotation withheld because this record is not cleared for redistribution. t-test t-test Quotation withheld because this record is not cleared for redistribution. false discovery rate correction false discovery rate correction Quotation withheld because this record is not cleared for redistribution. permutation test permutation test Quotation withheld because this record is not cleared for redistribution. null distribution by label shuffling null distribution by label shuffling about this quote: THE MATHS LAYER OF THE STAGED TEXT WAS MOJIBAKE AND HAS BEEN REPAIRED. This paper's extraction rendered '=' as U+00BC (the vulgar fraction one quarter) and '+' as U+00FE (thorn), so the quote read 'p ¼ 0.01' where the paper says 'p = 0.01'. It passed every check we own, because the staged text said the same wrong thing: the offsets were right and the recorded hash of the staged text matched. Repaired only where the character stands in an operator's place - spaced, with a number after it - because the same U+00BC is a REAL QUARTER in six other papers ('a ¼ inch condenser microphone'). Original in evidence as extracted, pre-repair file in an internal file. Quotation withheld because this record is not cleared for redistribution. cross-validation cross-validation Quotation withheld because this record is not cleared for redistribution. cross-correlation cross-correlation Quotation withheld because this record is not cleared for redistribution. z-score standardization z-score standardization Quotation withheld because this record is not cleared for redistribution. median and percentiles median and percentiles Quotation withheld because this record is not cleared for redistribution. Pre-stimulus spectra often showed a delta shoulder and a beta bump; speech lowered beta power and raised delta-theta and gamma power. Quotation withheld because this record is not cleared for redistribution. Single-trial decoding of speech was better from phase than from power, and information content was bimodal across frequency — high at delta-theta and at gamma, low at the intervening beta band. Quotation withheld because this record is not cleared for redistribution. Phase's decoding advantage over power grew as the discrimination got harder, from an effect size of 0.79 for easy discrimination to 1.21 for hard discrimination. Quotation withheld because this record is not cleared for redistribution. Concatenating power and phase patterns gave no decoding gain over whichever of the two was already the more informative. Quotation withheld because this record is not cleared for redistribution. Power information for speech detection generalized across low and high frequencies but not to the alpha-beta range. Quotation withheld because this record is not cleared for redistribution. Phase-based decoders trained at one frequency failed when tested at a sufficiently different frequency, showing that speech information sits in frequency-specific phase patterns. Quotation withheld because this record is not cleared for redistribution. Unlike every iEEG subject, the model showed no significant beta power reduction with speech. Quotation withheld because this record is not cleared for redistribution. Both the model-iEEG and the iEEG-iEEG similarity profiles were bimodal: high at delta-theta and gamma, low at alpha-beta. Quotation withheld because this record is not cleared for redistribution. Recurrent excitatory connection strength in either subnetwork had the largest effect on decoding performance. Quotation withheld because this record is not cleared for redistribution. For the hardest discrimination, phase gave strongly significant decoding in 9 of 10 patients while power did so in only 4 of 10. Quotation withheld because this record is not cleared for redistribution. Regularized Least-Square Classification Regularized Least-Square Classification Quotation withheld because this record is not cleared for redistribution. A prime A prime Quotation withheld because this record is not cleared for redistribution. Phase Locking Value (PLV) Phase Locking Value (PLV) Quotation withheld because this record is not cleared for redistribution. speech detection speech detection Quotation withheld because this record is not cleared for redistribution. speech discrimination speech discrimination Quotation withheld because this record is not cleared for redistribution. prosody prosody Quotation withheld because this record is not cleared for redistribution. area under the ROC curve area under the ROC curve Quotation withheld because this record is not cleared for redistribution. spectral power spectral power Quotation withheld because this record is not cleared for redistribution. bias current bias current Quotation withheld because this record is not cleared for redistribution. strength of recurrent excitatory connections strength of recurrent excitatory connections Quotation withheld because this record is not cleared for redistribution. 10 — epilepsy patients from whom intracortical EEG was recorded Quotation withheld because this record is not cleared for redistribution. 3.8 Hz and 85.9 Hz — natural resonance frequencies of the slow and fast model subnetworks Quotation withheld because this record is not cleared for redistribution. 9/10 vs 4/10 — patients with strongly significant hard-discrimination decoding from phase versus from power Quotation withheld because this record is not cleared for redistribution. 7 and 22 Hz / 20 and 40 Hz — inter-individual spread of the two crossover frequencies between speech-induced activation and deactivation bands Quotation withheld because this record is not cleared for redistribution. 100 — simulated trials per phrase used for the model decoding analyses Quotation withheld because this record is not cleared for redistribution. 0.975 / 0.951 — median phase-decoding accuracy across gamma-subnetwork parameter variations, and its worst value at the most extreme parameters Quotation withheld because this record is not cleared for redistribution. 0.79 and 1.21 — Cohen's d for phase's advantage over power in easy and hard discrimination Quotation withheld because this record is not cleared for redistribution.

The full text of this article is not redistributable, so it is not included here. The evidence sentences are withheld with it; the tag layer and its grounding status remain. The article itself is at https://doi.org/10.1101/730002.