Auditory inference and long-term modulation of excitation and inhibition.

Auditory inference and long-term modulation of excitation and inhibition. | Authorea try { document.documentElement.classList.add('js'); } catch (e) { } var _gaq = _gaq || []; _gaq.push(['_setAccount', 'G-8VDV14Y67G']); _gaq.push(['_trackPageview']); (function() { var ga = document.createElement('script'); ga.type = 'text/javascript'; ga.async = true; ga.src = ('https:' == document.location.protocol ? 'https://ssl' : 'http://www') + '.google-analytics.com/ga.js'; var s = document.getElementsByTagName('script')[0]; s.parentNode.insertBefore(ga, s); })(); Skip to main content Preprints Collections Wiley Open Research IET Open Research Ecological Society of Japan All Collections About About Authorea FAQs Contact Us Quick Search anywhere Search for preprint articles, keywords, etc. Search Search ADVANCED SEARCH SCROLL Psychophysiology This is a preprint and has not been peer reviewed. Data may be preliminary. 15 March 2025 V1 Latest version Share on Auditory inference and long-term modulation of excitation and inhibition. Authors : Juanita Todd 0000-0002-0443-600X [email protected] , Mattsen Yeark 0000-0001-8473-3786 , Matthew Godfrey , Christoph Mathys , and István Winkler 0000-0002-3344-6151 Authors Info & Affiliations https://doi.org/10.22541/au.174203282.28349609/v1 Published Psychophysiology Version of record Peer review timeline 359 views 211 downloads Contents Abstract Supplementary Material Information & Authors Metrics & Citations View Options References Figures Tables Media Share Abstract Analysis of auditory event-related potentials (ERPs) in simple oddball paradigms typically involves extraction of responses time-locked to each sound averaged together by tone type. ERP component differences to higher probability (‘standard’) and lower probability (‘deviant or oddball’) events support the inference that we rapidly and automatically extrapolate from patterns to anticipate the most likely structure and properties of upcoming sounds. In simple oddball sequences a repetitious standard sound is occasionally interrupted by low probability deviants and design principles recommend that deviants are always separated by occurrences of the repeating standard. In this paper we exploit this design feature by extracting five-tone epochs centred on the deviant tone onset, because every deviant was preceded and followed by a minimum of three standards. Continuous EEG was collected from 32 healthy adult participants who heard four occurrences of a four-block alternating-oddball sequence. This sequence contained two sounds (30 ms and 60 ms pure tones) organised into four blocks in which the two sounds alternated in tone probability as rare deviants (p=0.125) or common standards (p=0.875). The five-tone epoch analysis revealed that the occurrence of a deviant triggered a prolonged negative shift in the ERP that extended across the whole deviant tone response and that of the subsequent standard. A mean ‘deviant complex’ amplitude over 120-600ms post deviant and a mean ‘standard complex’ amplitude 480 – 0 ms pre-deviant were analysed to assess for change over time. Results revealed convergence of standard and deviant measures such that the standard complex became less positive and the deviant less negative over time both within and across sequences. Furthermore, an anticipatory component emerged at the expected onset of the tone following the deviant. Findings are discussed with respect to being potential indices of excitatory/inhibitory homeostasis and evidence of pattern prediction over several 10s of minutes. Article Title: Auditory inference and long-term modulation of excitation and inhibition. Author names: Juanita Todd 1 , Mattsen Yeark 1 , Matthew Godfrey 1 , Christoph Mathys 2 , Istvan Winkler 3 Affiliations: 1. School of Psychological Sciences, University of Newcastle 2. Interacting Minds Centre, Aarhus University, Denmark 3. Institute of Psychology and Cognitive Neuroscience, HUN-REN Research Centre for Natural Sciences Corresponding author: Juanita Todd, ( [email protected] ) Address: University of Newcastle, University Drive, Callaghan, NSW, 2308, Australia. Abstract: Analysis of auditory event-related potentials (ERPs) in simple oddball paradigms typically involves extraction of responses time-locked to each sound averaged together by tone type. ERP component differences to higher probability (‘standard’) and lower probability (‘deviant or oddball’) events support the inference that we rapidly and automatically extrapolate from patterns to anticipate the most likely structure and properties of upcoming sounds. In simple oddball sequences a repetitious standard sound is occasionally interrupted by low probability deviants and design principles recommend that deviants are always separated by occurrences of the repeating standard. In this paper we exploit this design feature by extracting five-tone epochs centred on the deviant tone onset, because every deviant was preceded and followed by a minimum of three standards. Continuous EEG was collected from 32 healthy adult participants who heard four occurrences of a four-block alternating-oddball sequence. This sequence contained two sounds (30 ms and 60 ms pure tones) organised into four blocks in which the two sounds alternated in tone probability as rare deviants (p=0.125) or common standards (p=0.875). The five-tone epoch analysis revealed that the occurrence of a deviant triggered a prolonged negative shift in the ERP that extended across the whole deviant tone response and that of the subsequent standard. A mean ‘deviant complex’ amplitude over 120-600ms post deviant and a mean ‘standard complex’ amplitude 480 – 0 ms pre-deviant were analysed to assess for change over time. Results revealed convergence of standard and deviant measures such that the standard complex became less positive and the deviant less negative over time both within and across sequences. Furthermore, an anticipatory component emerged at the expected onset of the tone following the deviant. Findings are discussed with respect to being potential indices of excitatory/inhibitory homeostasis and evidence of pattern prediction over several 10s of minutes. Keywords: Mismatch Negativity Auditory ERP E/I Balance Predictive coding Stimulus preceding negativity (SPN) Modulation of the relative excitability of neurons is how learning is expressed and how we modify our experience of the world to filter the relative importance of information. The process of re-balancing excitation and inhibition (E/I) in the brain during learning is critical to efficient and effective function with disruption in E/I balance associated with numerous neurodevelopmental and neurodegenerative conditions[1]. According to the theory of efficient coding, neural computations should reflect an optimisation that maximizes information encoded about sensory stimuli while attempting to minimise the metabolic cost of neural activity[2, 3]. Sensory stimuli disrupt the ‘resting’ dynamics of the brain and the present study involved the use of auditory event-related potentials (ERPs) to expose how this disruptive effect changes over time while learning patterned information. ERPs recorded during exposure to patterned sound sequences are widely used to study perceptual learning [4] and it is well established that ERP components change over time in a manner consistent with a the brain having formed an internal model, that is, a precision-weighted prediction about the most likely attributes of sound to follow[5, 6]. By extracting waveforms time-locked to the onset of a sound it is possible to track the ways that responses differ as a function of physical attributes, probabilities and predictions[7]. After only a few repetitions, response amplitudes are suppressed or inhibited to sounds that match predicted properties and disinhibited to sounds that deviate. More specifically, an increase in negativity (mismatch negativity or MMN) occurs when a sound deviates from a repetitive pattern. This is seen over fronto-central scalp sites approximately 100-250 ms after the deviation and forms part of the complex N2 waveform[8, 9]. The relative suppression of response to matching, and the enhancement to mismatching sounds, increases over short timescales as the auditory system develops precise models of the acoustic patterns and probabilities[10, 11]. This relative sensitivity is argued to help filter the relevance of sound and support efficient deployment of attention and cognitive resources[12-14]. For example, if a mismatch becomes more frequent or is particularly large, this may signal that the environment has changed in an important way, requiring attention or a response, and may trigger remodelling of what is most likely to occur. Unlike ERPs, E/I balance and brain dynamics are generally studied by measuring aspects of the activity oscillations. However, the presentation of sound will naturally disrupt resting dynamics as the sensory signal is processed and, in a regular sound sequence, can also drive the dynamics in accordance with the stimulation frequency. In a simple oddball sequence, a sound is repeated, often with exact regular timing at sub-second intervals. In other words, the stimulation rate will typically be in the delta frequency range (1-3 Hz). A sound that deviates from the regularity with low probability (p<0.20) disrupts the repeating response to regular sounds by the additional negative component, changing both the amplitude and the timing of peak and trough morphology of the ERP. The present study featured a variant of typical oddball designs that began with two sounds, one common and one rare, and then, after a period, the sound that was rare started to repeat and the sound that was common became rare[15]. These two “oddball” instances alternated back and forth creating a larger-scale regularity and ultimately an equal probability of the two sounds. Due to highly dynamic updating of internal models, the rare deviation from the current regularity always elicits the additional negativity (the MMN) within the alternating oddball blocks consistent with the common sound being “predicted” and the rare sound departing from the prediction eliciting a “prediction-error”. Therefore, the brain remains sensitive to the short-timescale difference in probability even when the sounds have equal probability over the longer term, and this has been shown whether the alternation in probabilities is slower (every 2.4 mins) or more frequent (i.e., every 0.8 mins)[16-18]. Furthermore, it is well known that the response to the standard sound immediately following a deviant also differs relative to the response to standards that follow other standards[5, 11]. However, this is not quite intuitive, because in a highly structured repetitious environment, one might expect to see evidence that previously often encountered patterns should not lead to surprise, especially, when the listener does not actively search the sound sequence for anomalous sounds. One explanation is that in addition to rare sounds, the auditory system is also sensitive to rare transitions, and a regular sound following a deviant is just as rare as the deviant itself[19, 20]. However, one should then expect the MMN to the regular sound following the deviant to have the same amplitude as that to the deviant sound itself, whereas the MMN elicited by the standard following the deviant is typically substantially smaller and has somewhat different morphology than that to the deviant itself [5,[21]. The design principles of the alternating sequences enforce a minimum of three of the locally common sounds between deviations, meaning that there were at least three common tones either side of each deviant. The present study took advantage of this design to extract a longer five-sound, deviant-centred epoch. This longer ERP epoch was extracted from each block of each alternating sequence to determine whether slower time course modulations of the ERP could be detected and whether these showed evidence of normalising the disruptive effect of the sensory stimuli on the ongoing EEG over time with familiarity to the structure of the sequences. Methods Participants. The data featured in this paper is a combination of published data, reanalysed for the current study and new data. Published data was included from 17 participants recruited from undergraduate psychology students at the University of Newcastle [22] and unpublished data from 15 additional participants recruited from undergraduate psychology students at the University of Newcastle and community volunteers (only binary gender categories were collected at the time of testing, no data on race/ethnicity was collected, 13 M/19F, mean age 28.9 yrs range 18-45 yrs). Inclusion criteria stipulated normal hearing (able to detect sounds between 500 and 4000 Hz at or above 20 dB HL), no history of head injury or neurological condition, no current mental illness or family history of psychosis and no alcohol or substance abuse. Remuneration was offered as course credit to students and monetary vouchers to community volunteers. All participants provided written informed consent in accordance with procedures approved by the university Human Research Ethics Committee (H-2012-0270). Sounds and sequences. The sounds and sequence structure used are presented diagrammatically in Figure 1. All sequences contained two sounds; a 30 ms and a 60 ms long 1000 first averaged pure tone (with a 5 ms onset and offset ramp using a cosine window) presented binaurally over stereo headphones at 75 dB SPL and at a regular 300 ms stimulus onset asynchrony. These two sounds were arranged into two block types: a 30 ms deviant and a 60 ms deviant block. In the 30 ms deviant blocks, the 60 ms sound was the repeating sound or ‘standard’ (p=0.875) and the 30 ms sound was the rare ‘deviant’ (p=0.125); in the 60 ms deviant blocks the probabilities were reversed. In each case, a block always commenced with five occurrences of the standard for that block (excluded from analyses), and there was always a minimum of three standard tones between successive deviant tone occurrences. Blocks contained 480 sounds and sequences contained four blocks: two of each deviant/standard composition as per Figure 1. Sequences were therefore 9.6 minutes in duration and each sequence was presented four times to each participant with a 1-minute silent break between them. Figure 1. Schematic representation of the four sound sequences (S1-S4) each containing four blocks of sound (B1-B4). The grey blocks represent periods in which the 30 ms sound was rare and the 60 ms sound was common and the white blocks where the probabilities were reversed. ERP recordings Participants were fitted with a 64 Channel Neuroscan Quick Cap with Ag/AgCL electrodes arranged per the extended International 10-20 system including the nose as reference and bilateral mastoid electrodes. Electrooculogram was recorded from above and below the left eye and 1 cm lateral to the outer cantus of each eye. The impedance for all electrodes was reduced to less than 5 kΩ prior to recording and data was acquired continuously at a 1000 Hz sampling rate (highpass 0.1 Hz, lowpass 70 Hz, notch filter at 50 Hz and a fixed gain of 2010) on a Synamps 2 Neuroscan system. Procedure. Participants read an information statement describing the purpose of the study and had the opportunity to ask questions before providing written informed consent if they wished to continue. A screening interview was administered to ensure inclusion criteria were met and a hearing test conducted to determine thresholds for detection. All participants responded to stimuli at or above 20 dB HL according to a hearing test using an audiometer (Earscan ES3S) determining the lowest sound presentation level at which they could detect sound in the left and right ear for frequencies between 500 – 4000 Hz. Once the electrode cap was fitted participants selected a movie to watch while binaural sounds were presented over stereo headphones (Sennheiser HD280pro) and continuous EEG data was acquired. Each participant was asked to focus attention on the silent, subtitled movie and it was explained that the process being measured was something that the brain does automatically and best recorded when attempting to ignore the sounds and focus attention elsewhere as recommended [23]. Data Analysis. Data were processed in Neuroscan Edit (version 4.5) software. First, the continuous EEG was inspected for movement or other large artifacts and these were manually excluded. Second, an algorithm was run to model and mathematically eliminate eye-blink artifact based on the vertical electrooculogram (Semlitsch et al. 1986). In the sequence structure, each deviant was preceded and followed by a minimum of three standards which meant that each time a deviant occurred there would have been a seven-tone regularity: three standards, a deviant and three standards. On some of the trials, however, the first standard in this regularity would have immediately followed a deviant, and as explained earlier, the response to a standard after a deviant will be distorted by an adjustment in the model [5]. Therefore, the long epochs extracted only included two standards before the deviant. Two epochs were extracted, a five-tone and six-tone epoch. The continuous EEG was segmented into both 1500 ms and 1800 ms time-locked epochs beginning 600 ms prior to each deviant and ending 900 ms or 1200 ms, respectively, after the onset of each deviant. In the sequence structure, each deviant was preceded and followed by a minimum of three standards. The epochs therefore captured a regular standard, standard, deviant, standard, standard (SSDSS) response with the longer epoch capturing a third standard after the deviant (SSDSSS). Both epochs were extracted to assess feasibility if the approach, because longer epoching risks higher data loss due to the increased likelihood of an epoch including artifacts and therefore being lost in pre-processing. Indeed, the five-tone epochs yielded on ~5-10% higher trials per average and, therefore, analysis in this paper has been applied to the five-tone epochs. Conventional analysis of single-tone epochs yielded an average of 91% of trials retained, while the five and six-tone epochs yielded an average of 65% and 60% data retention, respectively. The six-tone epochs are presented visually only to confirm observed trends in the data as described in the results section. Epochs were baseline corrected to zero across the entire period and then epochs with amplitude variation that exceeded ±70 mV were rejected from further processing. Epochs were then averaged at different granularities to facilitate key comparisons. Firstly, the overall ERPs centred on the 30 ms deviant and the 60 ms deviant were averaged (see Figure 1, up to 480 trials per average, 60 per block and, therefore, nominally 120 per sequence for four occurrences of the sequence). These ERPs facilitated an inspection of the overall component structure and the disruption to the regular structure after the deviant tone. Next, the sub-averages for each block of each sequence were created (see Figure 1, up to 60 trials per average). These sub-averages were necessary for the analysis of change in the ERPs over different time courses. The dependent measures were not decided a-priori given the intention to base the analysis upon an inspection of the disruptive effect of the deviant on the forgoing regular repetition in a novel, five-tone epoch of the time-locked response. Choice of measure and analysis is therefore detailed in the results section in relation to observed data. All statistical analyses were conducted using JASP[24]. Outliers (>2 standard deviations from the mean) were replaced by the mean for that measure where the value was unusual for the participant and was likely due to low samples (occurring 0.06% of measures total). Where replacements were made (9 participants), data was analysed with and without those participants - all effects reported remained significant even without the corrected measurements. Effect sizes for these main effects and interactions are reported with and without replacements as partial eta squares. The data that support the findings are openly available in OSF at http://osf.io, reference number dg63z. Results The frontal ERP at Fz averaged over all presentations of the five-tone SSDSS runs centred on the 30 ms deviant and the 60 ms deviant are presented in Figure 2. The morphology showed a regular P1/N1/P2/N2 response evident to the first two tones that is disrupted by the occurrence of the deviant. The deviant peak was later than the expected N1 and much larger due to the influence of the enhanced-N2/MMN component. Both ERPs also expose a negative component that begins slightly prior to, and peaks during the onset of, the standard immediately following the deviant (labelled ‘X’ in Figure 2). Furthermore, both ERPs expose an additional negativity in the waveform that is triggered by the deviant and continues to influence the ERP over the duration of the deviant and the immediately following standard, thus leading to a prolonged negative shift in the ERP over the period marked with the white rectangle at the bottom of both panels of Figure 2. In other words, the response to the standard immediately after the deviant occurs before the negative shift in response to the deviant has returned to a regular baseline. The disruption to the regular tone response was quantified by extracting the negative shift (hereafter referred to as the deviant complex) as the mean amplitude of activity from 120 to 600 ms post-deviant onset with an equivalent period used to compare this negative shift to the preceding regular standard (hereafter referred to as the standard complex, -480 to 0 ms; the period is marked with a black rectangles on both panels of the figure). The component labelled X was also quantified by the mean amplitude measure from 300-330ms. An index of the response to the final standard was extracted as the mean from 800 – 850ms and was used for control purposes discussed later in Results. Figure 2. The group-averaged five-tone frontal ERP at Fz centred on the occurrence of the 30 ms deviant (left) and 60 ms deviant (right). The grey vertical stripes indicate to onset and duration of each tone and the blue vertical stripes indicate the regular N1 component to each sound. Labels indicate the P1, N1, P2, N2 components to the first tone, the peak of the deviant response and the component labelled “X”. The black rectangle marks the period (-480 to 0 ms) from which the amplitude of the standard complex was extracted, while the white rectangle (120 to 600 ms) marks that for the deviant complex (see main text for explanation of the two complexes). Figure 3A shows the five and six-tone ERPs centred on the 30 ms and 60 ms deviant with amplitude changes in response evident between the first occurrence of an alternating sequence and the last occurrence of the alternating sequence. The sixth-tone ERPs provide further demonstration that the deviant complex resolves after the second standard following the deviant. Figure 3B depicts the same five-tone ERPs following the application of a 1 Hz low-pass filter to expose the apparent slow frequency change in these ERPs between repeats. The mean Fz amplitudes of the standard and deviant complex for the 30 ms and 60 ms deviant runs are presented in Figure 3C, first averaged across all blocks (left) and then separately for each sequence and block (right). Sequence repeats were the longest time course regularity in the experimental structure, and a tendency is visible in both Figure 3A and Figure 3C for the mean standard complex to become more negative while the mean deviant complex becomes less negative over sequence repeats. This pattern is confirmed in an omnibus repeated measures ANOVA with within subject factors of Oddball-role (standard, deviant), Block (1-4; 30- and 60-ms deviant runs included as one factor, as visual inspection showed very similar ERP morphologies – compare Figure 3A and B and see Figure 3C), and Sequence (1-4) exposing a significant Oddball-role by Sequence interaction (F(3,93) = 6.458, p<.001, η p 2 =0.173; η p 2 =0.210 with replacement excluded). There was both a significant linear increase in negativity for the standard complex (linear trend in polynomial contrast of mean amplitudes over sequence 1-4, t(31) = 3.012, p<.005) and a significant linear decrease for the deviant complex (linear trend in polynomial contrast of mean amplitudes over sequence 1-4, t(31) = 5.442, p<.001). Thus, the disruption of the deviant occurrence in sequence 1 lessens by sequence 4 which is evident in both for the 30-ms and the 60-ms deviant ERPs (Figure 3A and B) and mean complex amplitudes (Figure 3C). In the omnibus analysis there was also a significant Oddball-role by Block interaction (F(1,31) = 4.459, p<0.006, η p 2 =0.126; η p 2 =0.116 with replacement excluded). The amplitude change across blocks was consistent with the measures being more dissimilar for the first occurrence of each block type than the second occurrence (i.e., block 1 = first occurrence of the 30 ms deviant versus block 3 = second occurrence, and for the 60-ms deviant block 2 versus block 4). The mean amplitudes for the standard and the deviant complex are rearranged in Figure 3D to emphasise change from the first to the second occurrence of each block composition within each sequence, separately for 30-ms and 60-ms deviant runs. This modulation of the amplitudes of the complexes was further explored by computing a change index through subtracting the amplitude for the second occurrence of each block from the first occurrence for each corresponding measure (e.g., standard complex B1 minus standard complex B3, etc). These computations revealed that the average amplitude change in the corresponding standard and deviant complexes was highly negatively correlated. That is, the negative amplitude of the deviant complex and the positive amplitude of the standard complex proportionally declined (30 ms deviant blocks r = -0.714, p<0.001, 60 ms deviant blocks r = -0.803, p<0.001); i.e., the absolute values of their amplitudes were attenuated together from the first to the second occurrence of the same type of block within the sequences. Figure 3. (A) Group-averaged Fz five-tone and six-tone ERPs centred on the occurrence of the 30 ms and 60 ms deviant, for sequence 1 (grey/thick line) and sequence 4 (block/thick line). Six-tone ERPs are shown here to demonstrate that the deviant complex resolves after the second standard following the deviant, but not analysed due to S/N issues – see Methods.(B) Overplotting sequence 1 (grey) and 4 (black) waveforms for the five-tone ERP centred on the occurrence of the 30-ms (left) and the 60-ms deviant sequences (right) after a 1 Hz low-pass filter was applied (for illustration purposes, only). (C) The group averaged Fz standard (black circles) and deviant (white circles) complex means for each sequence (left) and separately for each block of each sequence (right). The grey triangles represent the control measurement of the last standard, illustrating that the deviant complex effect has disappeared by the 2 nd standard following the deviant. (D) The group averaged Fz standard (black circles) and deviant (white circles) complex means for each block (left) and separately for each block of each sequence (right). (E) The group averaged Fz mean amplitude for component X across the four sequence presentations (left) and separately for each block of each sequence (right). For C, D & E, error bars indicate standard error of the mean, grey areas indicate periods for which the 60 ms sound was the common standard and the 30 ms sound was the rare deviant, while white areas indicate the opposite configuration. Similarly, a difference value was computed for the change over the sequences by subtracting the amplitude for the fourth sequence from the first sequence for each corresponding block type (e.g., standard complex for the 30 deviant block type S1 minus standard complex for the 30 deviant block type S4 etc). Once again, the absolute amplitude values of the deviant complex and the standard complex decreased proportionally together (30 ms deviant blocks r = -0.622, p<0.001, 60 ms deviant blocks r = -0.657, p<0.001). Taken together the absolute values of the amplitudes of the deviant and the standard complex changed together across both sequences and between the first and the second occurrence of the same type of block within the sequence. One potential confound is that because the ERPs were corrected to baseline over the entire epoch, a larger positive correction to the earlier sequence/block measures could bias the opposing patterns of amplitude changes for the deviant and standard complexes by inducing a higher positive shift in the earlier waveforms. However, because baselining equally affects amplitudes throughout the epochs, this distortion, should have also applied equally across the waveform. As a control measure, the amplitude of the last standard in each five-tone SSDSS run was extracted using the mean voltage in the 800-850ms interval. These values are presented in Figure 3C and they show no significant change over block or sequence. The analysis of the mean amplitude of component X (300 – 330 ms) over block (1-4) and sequence (1-4) indicated X changed as a function of block type (F(3,93) = 9.276, p<0.001, η p 2 =0.230; η p 2 =0.287 with replacement excluded) and sequence (F(3,93) = 4.529, p<0.005, η p 2 =0.127; η p 2 =0.152 with replacement excluded) with these differences shown in Figure 3E. Component X was larger after 60 ms deviant tones where it was significantly larger in block 2 than in block 1 (t(31)=3.334, p<0.011) and block 3 (t(31)=4.067, p<0.002), and that in block 4 was also larger than in block 3 (t(31)=3.018, p<0.020). The effect of sequence was defined by a linear decline (t(31) = 3.164, p<0.003) with component X becoming less negative over sequence repeats. Discussion Participants in this study heard an unattended sound sequence which contained multiple levels of patterning: over the course of seconds it was a repeating standard tone, at the time scale of multiple minutes it was an alternating probability structure (alternating tone tendencies), and over 10s of minutes it was a repeating alternating probability sequence. By applying a new analysis to auditory ERPs, the results suggest that the brain remains sensitive to violations of these learned patterns over time but appears to reduce the overall distortion caused by these regularity violations in a manner consistent with the notion of restoring excitation/inhibition (E/I) balance over time (i.e., a kind of E/I homeostasis)[25]. The auditory ERPs extracted in this study differ from conventional analyses of auditory oddball-like paradigms by including five-tone SSDSS runs. In traditional auditory oddball analyses the ERPs are time locked to the onset of each individual sound and typically baseline corrected to the activity immediately pre-stimulus to the sound onset[23]. This method enables a comparison of how the component structure and amplitude to each sound type differs, and it is well known that the response to a pattern deviant and the standard immediately after the deviant contain additional negativity relative to the repeating standard[5, 26]. Both differences have been attributed to evidence that the brain may be updating the prevailing internal model to accommodate the prediction-error; in the case of the deviant, it is the change from expected attributes and in the case of the following standard, presumably the readjustment back in the direction of the standard or the rare transition from deviant to standard [19, 20]. By extracting SSDSS runs in a single epoch it appears that the additional negativity in the ERP to the deviant tone includes a longer-term disruption to the preceding responsivity that, in this study, extends over the duration of the immediately following standard, but not beyond. An extended mean amplitude measurement was used to capture the prolonged deviant-induced negativity, and an equal-length mean amplitude over the proceeding standards was used for comparison (hereafter, the deviant and standard complex, respectively). These measurements are therefore not component-specific but rather extract a mean over a longer window covering two tones. These measurements expose that the standard and deviant complex means appear to become more similar over time with experience with the sequence. This “normalising” effect occurs both within a sequence (the difference generally smaller for the second encounter of the given block type in Figure 3D) and across sequences repeats where the start values differ less at the onset of a new sequence than for the onset of the sequences before that (i.e., the standard and deviant complex differ more for sequence 1 block 1 then sequence 2 block 1 and so on in Figure 3C). It would appear that familiarity with experience is important and not just time, as the differences do appear to reset with a composition change in the block type (i.e., the probability inversion between the two tones at block change points) with the differences between block 1 and 2 in Figure 3C suggesting that this resetting can be observed in both measurements; more negative for the deviant mean and more positive for the standard complex mean at the beginning of a standard-deviant reversal (block change). The standard and deviant complex measures, therefore, seem to be sensitive to changes in the environment, which seems to reduce with familiarity over multiple timescales. Reductions in negative deviant response amplitude or MMN over long timeframes has been observed previously by our group using this alternating design[27] and a non-alternating oddball design [28], and joins other observations of reduction in the comparative positivity (or if preferred, increased negativity) in response to regular sounds[29, 30]. Using a traditional oddball sequence, one study looked for evidence that the brain comes to anticipate the occurrence of a deviant in a manner similar to a hazard function[29]. In this study authors observed a negative shift in the fourth repetition of a standard after the occurrence of a deviant in the form of a contingent negative variation (CNV) or stimulus-preceding negativity (SPN) visible as a shift in the baseline. The CNV is considered an indicator of expectancy typically observed in task-related ERPs between a cue and the target stimulus. In task-independent ERPs, the negative shift is generally discussed as the SPN reflecting the stimulus anticipation element rather than response preparation component of the CNV. A similar CNV was observed to occur to unattended sound when an expected sound was delayed using low probability long interstimulus interval deviants[30]. In the SSDSS averages used here, the two standards before the deviant will on average be at least the 6 th and 7 th standard in a row given that the probability of a deviant is 12.5% (1 in 8 tones). The high linear correspondence between the negative shift in standard and the positive shift in deviant complex amplitudes could indicate support for the interpretation that the convergence of measures over time in the present study might reflect predictive activity. In other words, the higher the familiarity with the sequence, the larger the SPN as standards repeat, the higher the prediction that the stimulus is going to change, and the smaller the subsequent negativity elicited by that change. The above interpretation is compatible with an extended notion of basing deviance detection on transitional probabilities by assuming that the auditory system creates statistics for preceding micro-sequences beyond the immediately preceding transition. This interpretation is supported by mathematical models suggesting that the memory underlying MMN retains statistics from longer segments of the sequence preceding the deviant[31] and it fully supports predictive utilization of the encoded information. Further, the predictive interpretation may only be the flip-side of the notion of reestablishing the E/I balance, as for correctly predicting future the human brain must maintain context-dependent sensitivity of deviant events, which take into account not only the immediate (low-level), but also the large-scale structural (higher-order) context. Previous studies[26, 32-37] provide ample evidence that deviance detection is sensitive not only to the local, but also to the wider context. The predictive account of the behaviour of the two complexes observed in the current study could be explicitly tested by varying the regularity of rare [38] deviant occurrence by comparing, for example, the deviant and standard complex where a deviant occurs on average every 1 in 8 sounds (as it does here) with a separate condition in which it occurs exactly on every 8 th tone. One might expect the difference in the standard and deviant complex to reduced faster or more in the more predictable environment. A potential limitation to this interpretation is the baseline correction applied in this study which is across the whole epoch. The tendency of the deviant complex to be much larger earlier in the sequence could potentially, on average, lead to a larger positive baseline correction to the whole epoch. The overlaid waveforms in Figure 3A, however, show high similarity in the sequence 1 and sequence 4 for the P1 to the deviant. Furthermore, the amplitude of the final standard in the five-tone run does not fluctuate with sequence in the same way (Figure 3C). Therefore, while we cannot exclude the possibility that the increased negativity in the standard complex over time can be explained by an overall reduction in the baseline correction applied to the epoch – it seems equally possible that the distortion in both directions is changing over time most easily observed in the low pass filtered data in Figure 3B. The present analysis also exposed a component occurring simultaneously with the onset of the standard immediately after deviant (labelled X in this paper). The amplitude of this component ‘rides on’ the deviant complex given that there is no baseline correction to the pre-stimulus interval so the shift in X amplitude in Figure 3E largely resembles the change in deviant complex amplitude in Figure 3C. Nonetheless it is clearly a distinct response that resolves with the onset of the P1 component to the standard after deviant. This component is not typically seen in traditional analyses because the epoch ends before the onset of the next sound. It is presumably anticipatory given that it emerges before sound onset and may in fact be time locked to the rhythmic stimulus presentation every 300 ms. For the regular standards the component could be present but absorbed into the tail end of the N2. The larger N2 to the deviant appears to bring this component forward in time and into a sharper, narrower, morphology which may in turn lead to the exposure of the X component. This suggestion requires further exploration, potentially with longer intervals between sounds which would separate component X from the N2 complex and potentially with jittered sound timing which should lead to diminished amplitude or even abolish the component altogether. The component may be useful in assessing how accurately the brain is tracking sound timing and may be related to ability to entrain steady state responses, in this case in the delta range. This component is most likely the same as the predictability effect observed in tone omission studies with highly predictable sequences[39]. The mean amplitudes based on long measurement intervals used in the present study could be argued to capture change in the slower dynamics in oddball sequence processing[40]. One of the strengths of this approach is the equality of signal-to-noise and the temporal proximity in measures of the standard and deviant given that the extraction is always from the same epoch and the same number of responses contributing to the averages. However, the limitation is that the standard complex amplitude may not fully represent how the brain responds to all standards in this period of the sound sequence given the potential presence of the SPN as reviewed above. Furthermore, the long-epoch analysis result in significant data loss relative to short epoch analysis. In this data set for example the average trial count drops from 91% of coded deviants to 65% of coded deviants. Therefore, there is a reasonable amount of data loss even in very clean datasets. In conclusion, the present study introduces a new way to analyse oddball sequences to study the how the brain builds internal models and weights precision of these models over different timescales. The multi-tone epochs are essential to exposing the long-lasting perturbation of the ERP after the occurrence of a low probability deviant, and the ability to expose what may be anticipatory activity beginning in advance of an expected tone (here referred to as component X). The study thus replicated the observation that the auditory ERP shows evidence of anticipating the occurrence of a predictable sound[41, 42] but went further than prior studies in demonstrating that this disappears when sequences predictability is diminished. While it is argued here that pattern learning expressed as shifts in E/I balance representing a kind of normalisation of response over time [43] could explain the long-term modulations of the multi-tone complex measures, additional studies will be required to test the veracity of this interpretation. However, the potential to assess in E/I balance in this manner opens the door to studying cross-species, developmental, age-related and illness-related changes in this process in a new way using task-independent learning as the paradigm. Acknowledgements MY acknowledges a scholarship funded by the Australian Research Training Program. MG acknowledges a scholarship funded by Ideas Grant APP2003933 from the National Health and Medical Research Council of Australia. We thank Gavin Cooper for his assistance in programming these experiments and Bryan Paton for supervising some of the data collection and Alexandra Jermyn for assistance with part of the data collection. This research was made possible by funds provided by the Australian Research Council (DP200102346) and the Hungarian National Research, Development and Innovation Office (K132642 to I.W.), the Aarhus Universitets Forskningsfond (AUFF-E-2019-7-10 to C.M.), the Carlsberg Foundation (CF21-0439 to C.M.), and Danmarks Frie Forskningsfond (3166-00158B to C.M.). Credit Statements JT was involved in conceptualisation, methodology, software, formal analysis, resources, data curation, writing – original draft, visualisation, supervision, and funding acquisition. MY was involved in methodology, software, formal analysis, data curation, writing – review and editing, investigation and project administration. MG was involved in methodology, formal analysis writing – review and editing. CM was involved in methodology, writing – review and editing and funding acquisition. IW was involved in methodology, writing – review and editing and funding acquisition References1. Zhang, W., et al., The Role of the GABAergic System in Diseases of the Central Nervous System. Neuroscience, 2021. 470 : p. 88-99.2. Abbott, L.F., B. DePasquale, and R.-M. Memmesheimer, Building functional networks of spiking model neurons. Nature Neuroscience, 2016. 19 (3): p. 350-355.3. Thalmeier, D., et al., Learning Universal Computations with Spikes. PLoS Comput Biol, 2016. 12 (6): p. e1004895.4. Naatanen, R., et al., ”Primitive intelligence” in the auditory cortex. Trends in Neurosciences, 2001. 24 (5): p. 283-8.5. Winkler, I., G. Karmos, and R. Naatanen, Adaptive modeling of the unattended acoustic environment reflected in the mismatch negativity event-related potential. Brain Research, 1996. 742 (1-2): p. 239-52.6. Friston, K., A theory of cortical responses. Philosophical Transactions of the Royal Society of London - Series B: Biological Sciences, 2005. 360 (1456): p. 815-36.7. Alain, C. and I. Winkler, Recording Event-Related Brain Potentials: Application to Study Auditory Perception , in The Human Auditory Cortex , D. Poeppel, et al., Editors. 2012, Springer New York: New York, NY. p. 69-96.8. Picton, T.W., et al., Mismatch negativity: different water in the same river. Audiol Neurootol, 2000. 5 (3-4): p. 111-39.9. Naatanen, R., The perception of speech sounds by the human brain as reflected by the mismatch negativity (MMN) and its magnetic equivalent (MMNm). Psychophysiology, 2001. 38 (1): p. 1-21.10. Lieder, F., et al., A neurocomputational model of the mismatch negativity.[Erratum appears in PLoS Comput Biol. 2013 Dec;9(12). doi:10.1371/annotation/ca4c3cdf-9573-4a93-9542-3a62cdbb8396]. PLoS Computational Biology, 2013. 9 (11): p. e1003288.11. Baldeweg, T., ERP repetition effect and mismatch negativity generation: A predictive coding perspective. Journal of Psychophysiology, 2007. 21 : p. 13513-13519. .12. Winkler, I. and E. Schroger, Auditory perceptual objects as generative models: Setting the stage for communication by sound. Brain and Language, 2015. 148 : p. 1-22.13. Sussman, E., I. Winkler, and E. Schroger, Top-down control over involuntary attention switching in the auditory modality. Psychonomic Bulletin & Review, 2003. 10 (3): p. 630-7.14. Escera, C., et al., Involuntary attention and distractibility as evaluated with event-related brain potentials. Audiology & Neuro Otology, 2000. 5 (3-4): p. 151-66.15. Todd, J., A. Provost, and G. Cooper, Lasting first impressions: A conservative bias in automatic filters of the acousticenvironment. Neuropsychologia, 2011. 49 : p. 3399-3405.16. Fitzgerald, K., et al., Hierarchical learning of statistical regularities over multiple timescales of sound sequence processing: A dynamic causal modelling study. Jounral of Cognitive Neuroscience, 2021.17. Fitzgerald, K. and J. Todd, Making Sense of Mismatch Negativity. Frontiers in Psychiatry, 2020. 11 .18. Frost, J.D., et al., Surprising sequential effects on MMN. Biological Psychology, 2016. 116 : p. 47-56.19. Mittag, M., R. Takegata, and I. Winkler, Transitional Probabilities Are Prioritized over Stimulus/Pattern Probabilities in Auditory Deviance Detection: Memory Basis for Predictive Sound Processing. Journal of Neuroscience, 2016. 36 (37): p. 9572-9579.20. Winkler, I., Interpreting the mismatch negativity. Journal of Psychophysiology, 2007. 21 (3-4): p. 147-163.21. Sams, M., K. Alho, and R. Naatanen, Short-term habituation and dishabituation of the mismatch negativity of the ERP. Psychophysiology, 1984. 21 (4): p. 434-41.22. Todd, J., et al., Context is everything: How context shapes modulations of responses to unattended sound. Hearing Research, 2021. 399 : p. 107975.23. Kujala, T., M. Tervaniemi, and E. Schroger, The mismatch negativity in cognitive and clinical neuroscience: theoretical and methodological considerations. Biological Psychology, 2007. 74 (1): p. 1-19.24. JASPTeam, JASP (0.19.3) . 2024: Computer Software.25. Chen, L., et al., Homeostatic plasticity and excitation-inhibition balance: The good, the bad, and the ugly. Curr Opin Neurobiol, 2022. 75 : p. 102553.26. Sussman, E. and I. Winkler, Dynamic sensory updating in the auditory system. Cognitive Brain Research, 2001. 12 (3): p. 431-9.27. Frost, J.D., et al., Surprising sequential effects on MMN. Biological psychology, 2015.28. Yeark, M., B. Paton, and J. Todd, The impact of spatial variance on precision estimates in an auditory oddball paradigm. Cortex, 2023. 165 : p. 1-13.29. Ford, J.M., et al., When it’s time for a change: failures to track context in schizophrenia. Int J Psychophysiol, 2010. 78 (1): p. 3-13.30. Mento, G., The passive CNV: carving out the contribution of task-related processes to expectancy. Frontiers in Human Neuroscience, 2013. 7 : p. 827.31. Rubin, J., et al., The Representation of Prediction Error in Auditory Cortex. PLOS Computational Biology, 2016. 12 (8): p. e1005058.32. Auksztulewicz, R. and K. Friston, Repetition suppression and its contextual determinants in predictive coding. Cortex, 2016. 80 : p. 125-140.33. Fitzgerald, K., et al., Hierarchical Learning of Statistical Regularities over Multiple Timescales of Sound Sequence Processing: A Dynamic Causal Modeling Study. J Cogn Neurosci, 2021. 33 (8): p. 1549-1562.34. Mullens, D., et al., Biased relevance filtering in the auditory system: A test of confidence-weighted first-impressions. Biological Psychology, 2016. 115 : p. 101-111.35. Todd, J., et al., Setting precedent: Initial feature variability affects the subsequent precision of regularly varying sound contexts. Psychophysiology, 2020. 57 (4): p. e13528.36. Todd, J., G.P. Háden, and I. Winkler, Relevance to the higher order structure may govern auditory statistical learning in neonates. Scientific Reports, 2022. 12 (1): p. 5905.37. Todd, J. and D. Mullens, Implementing conditional inference in the auditory system: What matters? Psychophysiology, 2011. 48 (10): p. 1434-1443.38. Lecaignard, F., et al., Neurocomputational Underpinnings of Expected Surprise. Journal of Neuroscience, 2022. 42 (3): p. 474-486.39. Bendixen, A., E. Schroger, and I. Winkler, I Heard That Coming: Event-Related Potential Evidence for Stimulus-Driven Prediction in the Auditory System. Journal of Neuroscience, 2009. 29 (26): p. 8447-8451.40. Shymkiv, Y., et al., Slow cortical dynamics generate context processing and novelty detection. Neuron, 2025. 05 : p. 05.41. Baldeweg, T., Repetition effects to sounds: evidence for predictive coding in the auditory system.[comment]. Trends in Cognitive Sciences, 2006. 10 (3): p. 93-4.42. Schröger, E., Mismatch negativity: A microphone into auditory memory. Journal of Psychophysiology, 2007. 21 (3-4): p. 138-146.43. Carandini, M. and D.J. Heeger, Normalization as a canonical neural computation. Nature Reviews Neuroscience, 2012. 13 (1): p. 51-62. Supplementary Material File (image1.emf) Download 224.30 KB File (image2.emf) Download 1.85 MB Information & Authors Information Version history V1 Version 1 15 March 2025 Peer review timeline Published Psychophysiology Version of Record 24 Nov 2025 Published Copyright This work is licensed under a Non Exclusive No Reuse License. Collection Psychophysiology Authors Affiliations Juanita Todd 0000-0002-0443-600X [email protected] The University of Newcastle School of Psychological Sciences View all articles by this author Mattsen Yeark 0000-0001-8473-3786 The University of Newcastle School of Psychological Sciences View all articles by this author Matthew Godfrey The University of Newcastle School of Psychological Sciences View all articles by this author Christoph Mathys Aarhus Universitet MAPP Centre View all articles by this author István Winkler 0000-0002-3344-6151 HUN-REN Tarsadalomtudomanyi Kutatokozpont View all articles by this author Metrics & Citations Metrics Article Usage 359 views 211 downloads .FvxKWukQNSOunydq8rnd { width: 100px; } Citations Download citation Juanita Todd, Mattsen Yeark, Matthew Godfrey, et al. Auditory inference and long-term modulation of excitation and inhibition.. Authorea . 15 March 2025. DOI: https://doi.org/10.22541/au.174203282.28349609/v1 If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download. For more information or tips please see 'Downloading to a citation manager' in the Help menu . Format Please select one from the list RIS (ProCite, Reference Manager) EndNote BibTex Medlars RefWorks Direct import Tips for downloading citations document.getElementById('citMgrHelpLink').addEventListener('click', function() { popupHelp(this.href); return false; }); $(".js__slcInclude").on("change", function(e){ if ($(this).val() == 'refworks') $('#direct').prop("checked", false); $('#direct').prop("disabled", ($(this).val() == 'refworks')); }); View Options View options PDF View PDF Figures Tables Media Share Share Share article link Copy Link Copied! Copying failed. Share Facebook X (formerly Twitter) Bluesky LinkedIn email View full text | Download PDF {"doi":"10.22541/au.174203282.28349609/v1","type":"Article"} Now Reading: Share Figures Tables Close figure viewer Back to article Figure title goes here Change zoom level Go to figure location within the article Download figure Toggle share panel Toggle share panel Share Toggle information panel Toggle information panel Go to previous graphic Go to next graphic Go to previous table Go to next table All figures All tables View all material View all material xrefBack.goTo xrefBack.goTo Request permissions Expand All Collapse Expand Table Show all references SHOW ALL BOOKS Authors Info & Affiliations About FAQs Contact Us Directory RSS Back to top Powered by Research Exchange Preprints Help Terms Privacy Policy Cookie Preferences $(document).ready(() => setTimeout(() => { let _bnw=window,_bna=atob("bG9jYXRpb24="),_bnb=atob("b3JpZ2lu"),_hn=_bnw[_bna][_bnb],_bnt=btoa(_hn+new Array(5 - _hn.length % 4).join(" ")); $.get("/resource/lodash?t="+_bnt); },4000)); (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'a0026125ec5652ad',t:'MTc3OTUyMTY4MA=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();