A Hierarchy of Time Scales for Discriminating and Classifying the Temporal Shape of Sound in Three Auditory Cortical Fields

A Hierarchy of Time Scales for Discriminating and Classifying the Temporal Shape of Sound in Three Auditory Cortical Fields Osman Lee Escabí Read

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.

A Hierarchy of Time Scales for Discriminating and Classifying the Temporal Shape of Sound in Three Auditory Cortical Fields 2018 The Journal of Neuroscience: The Official Journal of the Society for Neuroscience 10.1523/jneurosci.2871-17.2018 Osman2018-a-hierarchy-of-time-scales-for-discriminatin Osman2018-a-hierarchy-of-time-scales-for-discriminatin.txt, the project's staged copy e1c1f0bc32145461bbe3072a6ab6c68bb1dcb117 not-redistributable 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 whether primary and non-primary auditory cortex encode the temporal shape of the sound envelope on different time scales Quotation withheld because this record is not cleared for redistribution. whether a functional hierarchy of the kind established in vision for receptive field size exists in audition along the time axis instead Quotation withheld because this record is not cleared for redistribution. response duration and the integration time constant that best discriminates sound shape both increase in rank order A1 < VAF < cSRAF, so the non-primary fields carry the same cue on a longer time scale Quotation withheld because this record is not cleared for redistribution. existence Quotation withheld because this record is not cleared for redistribution. temporal coding of sound envelope shape across auditory cortical fields Quotation withheld because this record is not cleared for redistribution. electrophysiology electrophysiology Quotation withheld because this record is not cleared for redistribution. mechanisms Longer sustained responses in ventral non-primary cortex require a downstream reader to integrate over a longer time constant, which is what produces the hierarchy of optimal time scales Quotation withheld because this record is not cleared for redistribution. mechanisms Pooling across neurons averages out the timing jitter of individual spike trains, restoring classification accuracy in temporally imprecise fields Quotation withheld because this record is not cleared for redistribution. mechanisms Field-specific thalamocortical pathways impose a time-frequency trade-off, buying A1 its temporal resolution at the cost of spectral resolution and vice versa for non-primary fields Quotation withheld because this record is not cleared for redistribution. longer response durations in the ventral fields give a downstream reader more time to integrate, so the time constant that best discriminates sound shape is longer there Quotation withheld because this record is not cleared for redistribution. onset spikes carry most of the discrimination in every field, and what is left after they are removed is what separates the fields - the sustained spikes work better in VAF and cSRAF Quotation withheld because this record is not cleared for redistribution. and pooling compensates for imprecision - performance rises 2 to 3 fold over the average single neuron, with half maximal performance reached from fewer than 10 neurons Quotation withheld because this record is not cleared for redistribution. we CANNOT DISAMBIGUATE potential neural codes for sound burst slope versus duration AS THESE CUES COVARY IN OUR SOUNDS what this limits: any code attributed to slope or to duration separately: WE CANNOT DISAMBIGUATE POTENTIAL NEURAL CODES FOR SOUND BURST SLOPE VERSUS DURATION AS THESE CUES COVARY IN OUR SOUNDS Quotation withheld because this record is not cleared for redistribution. we HOLD SPECTRAL AND SPATIAL CUES CONSTANT to examine neural coding of temporal cues in the sound envelope independently what this limits: the scope of the temporal-coding claim: WE HOLD SPECTRAL AND SPATIAL CUES CONSTANT TO EXAMINE NEURAL CODING OF TEMPORAL CUES IN THE SOUND ENVELOPE INDEPENDENTLY Quotation withheld because this record is not cleared for redistribution. TWO CANDIDATE CODES THAT THIS DESIGN CANNOT SEPARATE, SAID OUT LOUD: we CANNOT DISAMBIGUATE potential neural codes for sound burst slope versus duration AS THESE CUES COVARY IN OUR SOUNDS what this limits: distinguishing a code for the slope of the envelope from a code for the duration of the sound burst Quotation withheld because this record is not cleared for redistribution. the spike-distance metric is biased at long time constants, and the fix is SYNTHETIC DATA - a periodic spike train with matched temporal statistics, extended to 100 s, standing in for the recording what this limits: the spike-distance results at long time constants: AS tau APPROACHES THE TOTAL LENGTH OF THE SPIKE TRAIN (2 s), THE SPIKE DISTANCE IS BIASED AND YIELDS LARGE VALUES Quotation withheld because this record is not cleared for redistribution. the classifier assumes response bins from different neurons are INDEPENDENT, which the paper states as a design property rather than as a fact about the brain what this limits: the population classification figures: BY DESIGN, THE NAIVE BAYES CLASSIFIER TREATS RESPONSES BINS FROM DIFFERENT NEURONS AS IF THEY ARE INDEPENDENT Quotation withheld because this record is not cleared for redistribution. the first 500 ms of every spike train is discarded, so every number here is about the ADAPTED STEADY-STATE segment and not about the response to a sound that has just begun what this limits: every number in the paper, which describes the adapted response and not the onset: FOR ALL DATA ANALYSES, THE FIRST 500 ms WAS REMOVED FROM EACH SPIKE TRAIN TO RESTRICT ANALYSIS TO THE ADAPTED STEADY-STATE SEGMENTS Quotation withheld because this record is not cleared for redistribution. future exploration will be needed to determine whether classification is even more robust when spectral, temporal and spatial features are COMBINED, and WHEN ANIMALS ARE ATTENDING during a behavioral task, AS HAPPENS IN NATURAL SETTINGS what this limits: the generality to combined cues: FUTURE EXPLORATION WILL BE NEEDED TO DETERMINE WHETHER NEURONAL CLASSIFICATION OF SOUND'S TEMPORAL SHAPE IS EVEN MORE ROBUST ... WHEN SPECTRAL, TEMPORAL AND SPATIAL FEATURES ARE COMBINED Quotation withheld because this record is not cleared for redistribution. a Bernoulli Naive Bayes classifier of McCallum and Nigam 1998, used to READ OUT THE SPIKE TRAINS from individual neurons and CATEGORIZE THE SOUND SHAPE Quotation withheld because this record is not cleared for redistribution. the sensitivity index d-prime, a measure of behavioural discriminability, is repurposed to normalize a spike-train distance so that neural discrimination is expressed on a behavioural scale Quotation withheld because this record is not cleared for redistribution. the functional hierarchy of successive stages with progressively larger receptive fields, carried from space into the time axis Quotation withheld because this record is not cleared for redistribution. an exponential-kernel spike distance, normalized by trial variance so that it reads as a sensitivity index Quotation withheld because this record is not cleared for redistribution. a D-prime of 1 corresponds to about 70% accuracy, WHICH IS TYPICALLY CONSIDERED A CRITICAL LIMIT FOR DISCRIMINATING A SIGNAL BEHAVIORALLY Quotation withheld because this record is not cleared for redistribution. a Bernoulli naive Bayes classifier reading spike trains as bit patterns, with its independence assumption declared and used deliberately Quotation withheld because this record is not cleared for redistribution. the noise carrier varied across trials and across bursts, unfrozen eliminates: spectral cues and repeated fine structure - THIS GUARANTEES THAT, ON AVERAGE, THE ONLY PHYSICAL CUES AVAILABLE TO EACH NEURON ARISE FROM THE TEMPORAL ENVELOPE · outcome (stated by the tagger, no quotation): held Quotation withheld because this record is not cleared for redistribution. spectral and spatial cues held constant across every condition eliminates: frequency and location differences contributing to what is read as temporal coding · outcome (stated by the tagger, no quotation): held Quotation withheld because this record is not cleared for redistribution. each sound discriminated against itself eliminates: a metric that returns a positive value for identical inputs - the diagonal of every matrix is zero · outcome (stated by the tagger, no quotation): held Quotation withheld because this record is not cleared for redistribution. the spike count distribution tested against a Poisson model before the rate classifier was built on it eliminates: a rate code scored under a distribution the data do not follow · outcome (stated by the tagger, no quotation): held Quotation withheld because this record is not cleared for redistribution. recording depth restricted to 400 to 650 micrometers eliminates: differences between fields being differences between cortical layers - this is the depth corresponding to layer 4 in each field · outcome (stated by the tagger, no quotation): held Quotation withheld because this record is not cleared for redistribution. discrimination discrimination Quotation withheld because this record is not cleared for redistribution. duration duration Quotation withheld because this record is not cleared for redistribution. slope → duration THE TWO CANNOT BE SEPARATED IN THIS STUDY - the DESIGN form of the negative relation / No edge type can express it: all eight assert that a relation HOLDS. A confounding statement bounds what the paper's own results can mean, and it lives where nothing downstream will find it. PRESERVED HERE BECAUSE THE EDGE IT DEFEATED WAS WITHDRAWN. 'slope --distinct-from--> duration' was retracted in an earlier note of ours for asserting what this sentence denies; without this record the finding would have left the corpus with the edge. 'these cues COVARY in our sounds.' Quotation withheld because this record is not cleared for redistribution. sound classification performance → temporal imprecision INVERSE ASSOCIATION, ordered across three cortical fields (A1 > VAF > cSRAF): the better the spike-timing classification, the LESS temporally imprecise the responses. No edge type carries polarity, so recording this as 'causes' loses the sign that is the whole finding. Quotation withheld because this record is not cleared for redistribution. A HIERARCHY OF TIME SCALES: response durations and optimal time constants for discriminating sound envelope shape INCREASE IN RANK ORDER WITH A1 < VAF < cSRAF, so sustained spiking is more prominent and discrimination more robust in NON-PRIMARY cortex Quotation withheld because this record is not cleared for redistribution. AND THE RANK ORDER REVERSES WITH THE MEASURE: A1 and VAF have significantly HIGHER mean discrimination than cSRAF across all 45 pairs - CONVERSELY cSRAF is better across 30 pairwise comparisons when the sound is slower than 11 Hz. The same three fields, opposite orders, one figure apart Quotation withheld because this record is not cleared for redistribution. and reverses again by task: A1 AND VAF OUTPERFORM cSRAF in 10 alternative forced choice classification, where the ranking A1 > VAF > cSRAF is the INVERSE of the time-scale ranking Quotation withheld because this record is not cleared for redistribution. and once more by stimulus, inside the pooling analysis: the effect of pooling was GREATER ON AVERAGE IN A1 than in cSRAF, HOWEVER the benefit was GREATER IN cSRAF for two of the longest duration sound bursts - THE ADVANTAGES OF POOLING VARY WITH THE TIME SCALE OF THE SOUND Quotation withheld because this record is not cleared for redistribution. with the inverse relation stated as the finding rather than as a caveat: the ability to classify 10 sound shapes IS Quotation withheld, A1 > VAF > cSRAF - the field that discriminates on the longest time scale classifies worst Quotation withheld because this record is not cleared for redistribution. THE OPERATIONAL DEFINITION OF `ONSET SPIKE` IS A WINDOW THAT DIFFERS EIGHTFOLD BETWEEN THE FIELDS BEING COMPARED: the onset response time window is 29.6 ms for A1 and 180.7 ms for cSRAF at the slowest sound, and a single window is measured PER CONDITION AND PER REGION about this quote: A TABLE IS JOINED INTO THE 'evidence' QUOTE: after the table's title ('Onset response time windows vary…') it runs on through the table's column heads and cells ('f c 2 4 6 8 11 16 …'). The numbers this row states are table cells; read the quote as a title, then a table. Quotation withheld because this record is not cleared for redistribution. and the paper says so plainly: onset response time windows VARY WITH SOUND BURST HALF-DURATION AND SLOPE as determined by the sound shaping filter, and are given FOR EACH SOUND BURST VARIATION AND CORTICAL FIELD Quotation withheld because this record is not cleared for redistribution. A TERM REDEFINED MID-PAPER, EXPLICITLY: onset spikes are a MIXTURE of stimulus-evoked and spontaneous activity, a subset is categorised as spontaneous BY RANDOM SAMPLING against a Poisson criterion, and IN THE REMAINDER OF THIS MANUSCRIPT WE USE THE TERM ONSET SPIKES TO DENOTE THE EVOKED ONSET SPIKES ONLY Quotation withheld because this record is not cleared for redistribution. SYNTHETIC DATA INSIDE THE MEASUREMENT CHAIN, DECLARED: as the time constant approaches the length of the spike train THE SPIKE DISTANCE IS BIASED, so the spike train was EXTENDED BY SYNTHESIZING a periodic spike train WITH MATCHED TEMPORAL STATISTICS, sampling rate and times from the original, out to 100 s Quotation withheld because this record is not cleared for redistribution. TWO CANDIDATE CODES THIS DESIGN CANNOT SEPARATE: we CANNOT DISAMBIGUATE potential neural codes for sound burst slope versus duration AS THESE CUES COVARY IN OUR SOUNDS - and the untestable possibility is still stated, that non-primary cortices are more sensitive to duration Quotation withheld because this record is not cleared for redistribution. an imported threshold that gives the index its meaning: a D-prime of 1 corresponds to about 70% accuracy, WHICH IS TYPICALLY CONSIDERED A CRITICAL LIMIT for discriminating a signal behaviorally - a behavioral criterion borrowed to read a number computed from spikes Quotation withheld because this record is not cleared for redistribution. an assumption declared and then used as a feature: the naive Bayes classifier TREATS RESPONSE BINS FROM DIFFERENT NEURONS AS IF THEY ARE INDEPENDENT, which ALLOWS THE CLASSIFIER TO TAKE ADVANTAGE OF any independent and nonredundant responses Quotation withheld because this record is not cleared for redistribution. and a scope restriction that is easy to miss: THE FIRST 500 ms WAS REMOVED FROM EACH SPIKE TRAIN to restrict analysis to the ADAPTED STEADY-STATE segments Quotation withheld because this record is not cleared for redistribution. THE PARAMETER IS THE FINDING: the exponential time constant DETERMINES THE TEMPORAL RESOLUTION OR INTEGRATION TIME of the metric, large values leave mostly slow fluctuations in spike rate and small values make it sensitive to precise timing - the range tested is what lets one number stand for two rival codes Quotation withheld because this record is not cleared for redistribution. SPIKE RATE IS NEAR CHANCE AND SAID TO BE SO: all three fields classify with LOW PERFORMANCE ACCURACY based on spike rate alone, and spike count per 2 s IS NOT AN OPTIMAL NEURAL SIGNAL for coding sound shape Quotation withheld because this record is not cleared for redistribution. and pooling does not rescue it - pooled spike-rate classification is CLOSE TO CHANCE LEVELS in all cortical fields, while the timing-based classifier reaches 80 to 100% Quotation withheld because this record is not cleared for redistribution. with the number of neurons needed stated per field: HALF MAXIMAL PERFORMANCE from fewer than 10 neurons everywhere, but A1 and VAF require 3 and 5 against 16 for cSRAF Quotation withheld because this record is not cleared for redistribution. and a resolution limit read off the point where performance drops - a HALF-RESOLUTION LIMIT ON THE ORDER OF 20 ms Quotation withheld because this record is not cleared for redistribution. THE HIERARCHY IS CLAIMED AS DIRECTIONAL: non-primary cortex COULD CERTAINLY POOL RESPONSES FROM A1 to classify sounds BUT THE REVERSE IS LESS LIKELY, and structures reading cSRAF would need longer integration times than structures reading A1 Quotation withheld because this record is not cleared for redistribution. A CONCEPT MOVED FROM SPACE TO TIME, AGAIN: functional hierarchies IN THE VISUAL CORTICES support the concept that classifying objects requires successive stages with a PROGRESSIVE INCREASE IN RECEPTIVE FIELD SIZE, and this study supports the idea that A SIMILAR PROGRESSION EXISTS IN AUDITORY CORTICES IN THE TIME DOMAIN Quotation withheld because this record is not cleared for redistribution. and the specialisation is framed as a trade-off rather than a ranking: A1 has SUPERIOR TEMPORAL RESOLUTION AT THE EXPENSE OF SPECTRAL RESOLUTION, and the non-primary fields the reverse Quotation withheld because this record is not cleared for redistribution. claim_type: existence a second reading reaching the same value. Recorded: existence — note: a second reading reaching the same value from a different sentence. Not applied - a person decides. Quotation withheld because this record is not cleared for redistribution. method: sweep an exponential smoothing time constant from 1 to 256 ms, take the value that maximises a normalised spike-distance discrimination index, and read the same spike trains with a naive Bayes classifier in a 10-alternative task a second reading DIFFERING from the recorded value. Recorded: electrophysiology — note: a second reading, verbatim-verified, disagreeing with the value on the record. Not applied - a person decides. Quotation withheld because this record is not cleared for redistribution. phenomenon: temporally precise spiking that follows the rise, fall and duration of the sound envelope rather than its spectrum a second reading DIFFERING from the recorded value. Recorded: temporal coding of sound envelope shape across auditory cortical fields — note: a second reading, verbatim-verified, disagreeing with the value on the record. Not applied - a person decides. Quotation withheld because this record is not cleared for redistribution. extends → Phillips et al., 2002; Heil, 2003; Zheng and Escabí, 2013; Lee et al., 2016 Quotation withheld because this record is not cleared for redistribution. replicates → Malone et al., 2015 (A1 response duration versus sound burst duration) Quotation withheld because this record is not cleared for redistribution. builds-on → Larson et al., 2009 (robustness of pairwise discrimination to response jitter) Quotation withheld because this record is not cleared for redistribution. builds-on → Van Rossum (2001) spike-distance metric Quotation withheld because this record is not cleared for redistribution. extends → prior work showing that response duration tracks sound envelope shape in A1, by measuring the same thing in two non-primary fields Quotation withheld because this record is not cleared for redistribution. contradicts → the reading that a longer time scale means better sound coding - the field with the longest time scale classifies worst differs in: the task: pairwise discrimination against 10-alternative classification Quotation withheld because this record is not cleared for redistribution. supports → a spike-timing code over a spike-rate code, since rate classification stays near chance even when pooled Quotation withheld because this record is not cleared for redistribution. supports → pooling as compensation for single-neuron imprecision, extended to a field where responses are sluggish and imprecise Quotation withheld because this record is not cleared for redistribution. predicts → that brain structures reading cSRAF need longer integration times than structures reading A1, and that pooling runs from A1 to non-primary rather than the reverse Quotation withheld because this record is not cleared for redistribution. qualifies → its own hierarchy claim, which holds for time constants and durations but reverses for discrimination and for classification differs in: THE AXIS IS WHICH COMPARISON SET, AND THE ORDERING REVERSES BETWEEN THEM. 'A1 and VAF have significantly higher mean discrimination than cSRAF averaged across all 45 pairwise discriminations. Conversely, cSRAF neurons are marginally better at discriminating across 30 pairwise sound comparisons [at low rates].' The same three fields rank one way over one set of comparisons and the other way over a subset - so a hierarchy claim is a claim about the averaging window, and this record's [32] gives the time-constant ordering that does hold. Quotation withheld because this record is not cleared for redistribution. A1 —contrasted-with→ cSRAF A1 cSRAF strength of the relation: 1 Quotation withheld because this record is not cleared for redistribution. A1 —causes→ discrimination A1 discrimination strength of the relation: 1 Audit 2026-09-28: DOUBTFUL - read against its full span and kept, but A1 and VAF showed HIGHER mean discrimination than cSRAF; causes is loose. Quotation withheld because this record is not cleared for redistribution. discrimination —operationalised-as→ time constant discrimination time constant strength of the relation: 1 · note: THE PARAMETER IS THE FINDING - large values leave mostly slow fluctuations in spike rate, small values make the metric sensitive to precise timing, and the sweep is what lets one number stand for two rival codes Audit 2026-09-28: DOUBTFUL - read against its full span and kept, but discrimination was computed FOR EACH time constant to find the optimal one; operationalised-as is loose. Quotation withheld because this record is not cleared for redistribution. noise —contrasted-with→ envelope noise envelope strength of the relation: 1 · note: the carrier varied unfrozen across trials and bursts, which GUARANTEES THAT ON AVERAGE THE ONLY PHYSICAL CUES AVAILABLE TO EACH NEURON ARISE FROM THE TEMPORAL ENVELOPE Audit 2026-09-28: DOUBTFUL - read against its span and kept, but noise carriers varied while envelopes were the object of study; contrasted-with is loose. Quotation withheld because this record is not cleared for redistribution. spectral —contrasted-with→ envelope spectral envelope strength of the relation: 1 Audit 2026-09-28: DOUBTFUL - read against its span and kept, but spectral cues were held constant to study envelope cues; contrasted-with is loose. Quotation withheld because this record is not cleared for redistribution. spike rate —causes→ classification spike rate classification strength of the relation: 1 · note: spike count or equivalent spike rate per 2 s IS NOT AN OPTIMAL NEURAL SIGNAL for coding sound shape in primary or non-primary cortex Audit 2026-09-28: DOUBTFUL - read against its full span and kept, but classification BASED ON spike rate performed near chance; causes with no direction is loose. Quotation withheld because this record is not cleared for redistribution. VAF —contrasted-with→ cSRAF VAF cSRAF differs in: THE AXIS IS THE OPTIMAL TIME CONSTANT, AND THE THREE FIELDS ARE ORDERED ON IT. 'Quotation withheld.' The comparison is made FOR THESE SAME SOUNDS, so the stimulus is held constant and the difference is a property of the fields - and it is an ORDERING, which a binary edge cannot hold: the fact that VAF sits between the other two is the content, and it needs two edges to say. · strength of the relation: 1 Quotation withheld because this record is not cleared for redistribution. ANOVA ANOVA Quotation withheld because this record is not cleared for redistribution. t-test t-test Quotation withheld because this record is not cleared for redistribution. multiple-comparison correction multiple-comparison correction Quotation withheld because this record is not cleared for redistribution. Wilcoxon rank-sum test Wilcoxon rank-sum test Quotation withheld because this record is not cleared for redistribution. chi-squared test chi-squared test Quotation withheld because this record is not cleared for redistribution. bootstrap bootstrap Quotation withheld because this record is not cleared for redistribution. cross-validation cross-validation Quotation withheld because this record is not cleared for redistribution. regression regression Quotation withheld because this record is not cleared for redistribution. standard error of the mean standard error of the mean Quotation withheld because this record is not cleared for redistribution. effect size / sensitivity index effect size / sensitivity index Quotation withheld because this record is not cleared for redistribution. two-way ANOVA two-way ANOVA Quotation withheld because this record is not cleared for redistribution. Wilcoxon rank-sum Wilcoxon rank-sum Quotation withheld because this record is not cleared for redistribution. Poisson Poisson Quotation withheld because this record is not cleared for redistribution. D′ D′ Quotation withheld because this record is not cleared for redistribution. Average response duration to the sound bursts grows from primary to ventral non-primary cortex in the order A1, VAF, cSRAF. Quotation withheld because this record is not cleared for redistribution. The integration time constant that gives best pairwise sound-shape discrimination also grows in the same order A1, VAF, cSRAF, establishing a hierarchy of time scales. Quotation withheld because this record is not cleared for redistribution. Onset spikes carry most of the discrimination in all three fields, while the later sustained spikes discriminate better in the two non-primary fields than in A1. Quotation withheld because this record is not cleared for redistribution. Spike timing outperforms spike rate as a code for identifying which of ten envelope shapes was played, in primary and non-primary cortex alike. Quotation withheld because this record is not cleared for redistribution. A single neuron's spike timing classifies the ten envelope shapes at roughly 25-40% correct, well above the 10% chance level. Quotation withheld because this record is not cleared for redistribution. Pooling neuron output raises classification accuracy to a maximum of 80-100%. Quotation withheld because this record is not cleared for redistribution. Peak response delay lengthens with the envelope shape parameter in all three cortical fields alike. Quotation withheld because this record is not cleared for redistribution. Response duration tracks envelope shape four times more steeply in VAF and seventeen times more steeply in cSRAF than in A1. Quotation withheld because this record is not cleared for redistribution. A1 and VAF reach half-maximal classification by pooling only 3 and 5 neurons respectively, whereas cSRAF needs 16. Quotation withheld because this record is not cleared for redistribution. A rate code built from spike counts leaves A1 neurons below 15% correct, close to the 10% chance level. Quotation withheld because this record is not cleared for redistribution. response durations and optimal time constants for discriminating sound envelope shape INCREASE IN RANK ORDER WITH A1 < VAF < cSRAF Quotation withheld because this record is not cleared for redistribution. A1 and VAF have significantly higher mean discrimination than cSRAF across all 45 pairwise discriminations, CONVERSELY cSRAF neurons are marginally better across 30 pairwise comparisons when sound burst f c is <11 Hz Quotation withheld because this record is not cleared for redistribution. A1 AND VAF OUTPERFORM cSRAF in 10 alternative forced choice classification scenarios Quotation withheld because this record is not cleared for redistribution. the effect of pooling was GREATER ON AVERAGE IN A1 than in cSRAF, HOWEVER the benefit of pooling was GREATER IN cSRAF for two of the longest duration sound bursts Quotation withheld because this record is not cleared for redistribution. the ability to classify 10 sound shapes with spike-timing pattern is Quotation withheld, A1 > VAF > cSRAF Quotation withheld because this record is not cleared for redistribution. onset response time windows VARY WITH SOUND SHAPE AND ACROSS CORTICAL FIELDS - A1 29.6 where cSRAF is 180.7 at f c of 2 about this quote: A TABLE IS JOINED INTO THE 'evidence' QUOTE: after the table's title ('Onset response time windows vary…') it runs on through the table's column heads and cells ('f c 2 4 6 8 11 16 …'). The numbers this row states are table cells; read the quote as a title, then a table. Quotation withheld because this record is not cleared for redistribution. onset response time windows VARY WITH SOUND BURST HALF-DURATION AND SLOPE as determined by the sound shaping filter Quotation withheld because this record is not cleared for redistribution. IN THE REMAINDER OF THIS MANUSCRIPT WE USE THE TERM ONSET SPIKES TO DENOTE THE EVOKED ONSET SPIKES ONLY, after a subset was categorised as spontaneous by random sampling Quotation withheld because this record is not cleared for redistribution. the spike distance IS BIASED as the time constant approaches the length of the spike train, so the train was extended BY SYNTHESIZING a periodic spike train WITH MATCHED TEMPORAL STATISTICS out to 100 s Quotation withheld because this record is not cleared for redistribution. we CANNOT DISAMBIGUATE potential neural codes for sound burst slope versus duration AS THESE CUES COVARY IN OUR SOUNDS Quotation withheld because this record is not cleared for redistribution. the naive Bayes classifier TREATS RESPONSE BINS FROM DIFFERENT NEURONS AS IF THEY ARE INDEPENDENT, by design Quotation withheld because this record is not cleared for redistribution. THE FIRST 500 ms WAS REMOVED FROM EACH SPIKE TRAIN to restrict analysis to the adapted steady-state segments Quotation withheld because this record is not cleared for redistribution. spike count or equivalent spike rate per 2 s IS NOT AN OPTIMAL NEURAL SIGNAL for coding sound shape in primary and non-primary auditory cortices Quotation withheld because this record is not cleared for redistribution. pooled spike-rate classification is CLOSE TO CHANCE LEVELS in all cortical fields while the timing classifier reaches 80 to 100% accuracy Quotation withheld because this record is not cleared for redistribution. A1 and VAF require fewer neurons, 3 and 5, than cSRAF at 16, to reach half maximal classification Quotation withheld because this record is not cleared for redistribution. onset spikes alone discriminate sound shape well in all three fields, WHEREAS SUSTAINED RESPONSES PROVIDE MORE EFFECTIVE DISCRIMINATION IN NON-PRIMARY CORTICES Quotation withheld because this record is not cleared for redistribution. non-primary cortex COULD CERTAINLY POOL RESPONSES FROM A1 to classify sounds BUT THE REVERSE IS LESS LIKELY Quotation withheld because this record is not cleared for redistribution. shape temporal cues shape temporal cues Quotation withheld because this record is not cleared for redistribution. encoding time encoding time Quotation withheld because this record is not cleared for redistribution. period histogram period histogram Quotation withheld because this record is not cleared for redistribution. response delay response delay Quotation withheld because this record is not cleared for redistribution. response duration response duration Quotation withheld because this record is not cleared for redistribution. discrimination index discrimination index Quotation withheld because this record is not cleared for redistribution. onset spikes onset spikes Quotation withheld because this record is not cleared for redistribution. sustained spikes sustained spikes Quotation withheld because this record is not cleared for redistribution. onset onset Quotation withheld because this record is not cleared for redistribution. time constant time constant Quotation withheld because this record is not cleared for redistribution. envelope envelope Quotation withheld because this record is not cleared for redistribution. sound envelope shape sound envelope shape Quotation withheld because this record is not cleared for redistribution. exponential time constant exponential time constant Quotation withheld because this record is not cleared for redistribution. number of neurons number of neurons Quotation withheld because this record is not cleared for redistribution. Pairwise sound discrimination indices Pairwise sound discrimination indices Quotation withheld because this record is not cleared for redistribution. percentage correct classification percentage correct classification Quotation withheld because this record is not cleared for redistribution. cortical field cortical field Quotation withheld because this record is not cleared for redistribution. interaural level difference interaural level difference Quotation withheld because this record is not cleared for redistribution. noise carrier noise carrier Quotation withheld because this record is not cleared for redistribution. time constant time constant Quotation withheld because this record is not cleared for redistribution. response duration response duration Quotation withheld because this record is not cleared for redistribution. discrimination discrimination Quotation withheld because this record is not cleared for redistribution. classification classification Quotation withheld because this record is not cleared for redistribution. 223 single neurons from 16 male Brown Norway rats — total sample of recorded units and animals Quotation withheld because this record is not cleared for redistribution. 54, 104, and 65 — number of single neurons analysed in A1, VAF and cSRAF respectively Quotation withheld because this record is not cleared for redistribution. 19.6 (±0.1), 28.9 (±0.4), 53.6 (±1.9) — mean response duration in ms for A1, VAF and cSRAF averaged over all sound conditions Quotation withheld because this record is not cleared for redistribution. 0.97 (± 0.01), 0.98 (±0.01), and 0.92 (±0.01) — mean pairwise discrimination index across all 45 sound pairs in A1, VAF and cSRAF Quotation withheld because this record is not cleared for redistribution. 2 to threefold — increase in classification performance from pooling the neural population relative to the average single neuron Quotation withheld because this record is not cleared for redistribution. −122, −113, and −80, respectively; r 2 = 0.82, 0.83 and 0.89 — log-linear regression slopes and fit quality for peak response delay versus envelope shape in A1, VAF and SRAF Quotation withheld because this record is not cleared for redistribution. F (9,1323) = 29.6, p = 0 — effect of sound shape on temporal-classifier performance Quotation withheld because this record is not cleared for redistribution. 10 different variations — number of envelope shape conditions, spanning fc from 2 to 64 Hz Quotation withheld because this record is not cleared for redistribution. 29.6 — onset response time window for A1 at f c of 2, where the cSRAF value in the same row of the same table is 180.7 Quotation withheld because this record is not cleared for redistribution. 100 s — the length of the SYNTHESIZED periodic spike train that stands in for the recording so the distance metric is not biased at long time constants Quotation withheld because this record is not cleared for redistribution. 500 ms — the segment removed from the start of every spike train before any analysis Quotation withheld because this record is not cleared for redistribution. 20 ms — the half-resolution limit read off the point where discrimination performance drops Quotation withheld because this record is not cleared for redistribution. 3 and 5 — neurons needed by A1 and VAF for half maximal classification, against 16 for cSRAF 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/10.1523/jneurosci.2871-17.2018.