Assessing the Impact of Selective Attention on the Cortical Tracking of the Speech Envelope in the Delta and Theta Frequency Bands and How Musical Training Does (Not) Affect It

Assessing the Impact of Selective Attention on the Cortical Tracking of the Speech Envelope in the Delta and Theta Frequency Bands and How Musical Training Does (Not) Affect It | bioRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-M677548'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search New Results Assessing the Impact of Selective Attention on the Cortical Tracking of the Speech Envelope in the Delta and Theta Frequency Bands and How Musical Training Does (Not) Affect It Alina Schüller , Annika Mücke , Jasmin Riegel , Tobias Reichenbach doi: https://doi.org/10.1101/2024.08.01.606154 Alina Schüller a Department Artificial Intelligence in Biomedical Engineering, Friedrich-Alexander-University Erlangen-Nürnberg , Erlangen, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Annika Mücke a Department Artificial Intelligence in Biomedical Engineering, Friedrich-Alexander-University Erlangen-Nürnberg , Erlangen, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Jasmin Riegel a Department Artificial Intelligence in Biomedical Engineering, Friedrich-Alexander-University Erlangen-Nürnberg , Erlangen, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Tobias Reichenbach a Department Artificial Intelligence in Biomedical Engineering, Friedrich-Alexander-University Erlangen-Nürnberg , Erlangen, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site For correspondence: tobias.j.reichenbach{at}fau.de Abstract Full Text Info/History Metrics Preview PDF Abstract Oral communication regularly takes place amidst background noise, requiring the ability to selectively attend to a target speech stream. Musical training has been shown to be beneficial for this task. Regarding the underlying neural mechanisms, recent studies showed that the speech envelope is tracked by neural activity in the auditory cortex, which plays a role in the neural processing of speech, including speech in noise. The neural tracking occurs predominantly in two frequency bands, the delta and the theta band. However, much regarding the specifics of these neural responses, as well as their modulation through musical training, still remain unclear. Here, we investigated the delta- and theta-band cortical tracking of the speech envelope of attended and ignored speech using magnetoencephalography (MEG) recordings. We thereby assessed both musicians and non-musicians to explore potential differences between these groups. The cortical speech tracking was quantified through source-reconstructing the MEG data and subsequently relating the speech envelope in a certain frequency band to the MEG data using linear models. We thereby found the theta-band tracking to be dominated by early responses with comparable magnitudes for attended and ignored speech, whereas the delta band tracking exhibited both earlier and later responses that were modulated by selective attention. Almost no significant differences emerged in the neural responses between musicians and non-musicians. Our findings show that only the speech tracking in the delta but not in the theta band contributes to selective attention, but that this mechanism is essentially unaffected by musical training. Introduction Auditory selective attention refers to the ability of the human brain to segregate spatiotemporally overlapping speech streams into distinct auditory objects and to selectively attend one of them [ 1 ]. However, this ability requires significant cognitive resources and can be impeded by several factors, such as hearing impairment [ 2 , 3 ], speech-related learning impairment [ 2 ] or age-related decline in the ability to attend to a target speaker [ 4 , 5 ]. In contrast, musical training may prevent or even counteract difficulties in speech-in-noise perception [ 2 , 3 , 5 , 6 , 7 ]. Possible explanations for this hypothesis include that musical training can enhance brain plasticity [ 6 ] and functional connectivity [ 6 , 8 ], increase the auditory working memory [ 9 , 7 , 10 ], and thus improve auditory fitness [ 2 ], especially when the training started at a young age [ 2 , 11 ]. However, despite the importance of speech-in-noise comprehension for human oral communication and social life, the underlying neural processes, as well as the neural mechanisms leading to declined or improved abilities, are not yet fully understood. Recent investigations have employed non-invasive neuroimaging through electroencephalogra-phy (EEG) or magnetoencephalography (MEG) while participants listen to naturalistic speech in noise. Through subsequent statistical modeling, these recordings allow quantifying ongoing neural responses to repetitive, rhythmic aspects of speech stimuli, often referred to as neural speech tracking [ 1 , 12 , 13 , 14 ]. Perhaps the most robust neural tracking emerges in response to the speech envelope, a signal in the low-frequency range, between 1-15 Hz, that traces the amplitude fluctuations of a speech signal [ 13 , 15 ]. The neural tracking of the speech envelope does not simply reflect the bottom-up processing of the acoustic input, but is also influenced by higher cognitive processes, in particular by selective attention, which enhances the tracking [ 16 , 1 , 13 ]. Moreover, modulating the neural tracking through transcranial alternating current stimulation can impact and even enhance speech-in-noise comprehension [ 17 , 18 , 19 , 20 , 21 ]. Different functions have been attributed to cortical tracking in the theta band (4 - 8 Hz) and delta band (1 - 4 Hz) [ 22 ]. Theta-band tracking probably reflects the parsing of lower-level speech components, such as syllables and phonemes [ 22 , 23 , 24 , 25 ], and reflects the acoustic clarity [ 26 ]. Delta-band tracking is associated with neural responses to words, reflects higher-level linguistic processing and can inform on speech comprehension [ 15 , 26 , 25 ]. Theta-band tracking has indeed been found to be more sensitive to stationary background noise, whilst delta band tracking seems to be robust as long as the speech stimulus is still intelligible [ 15 , 27 ]. However, the specific contributions of delta- and theta-band tracking to selective attention have not yet been fully clarified. The neural specialization that may allow musicians to achieve better speech-in-noise comprehension has been in-vestigated through non-invasive neuroimaging as well, mostly focusing on the evaluation of short auditory stim-uli [ 28 , 9 , 29 , 30 ]. These investigations found, in particular, enhanced subcortical responses of people with musical training [ 31 , 32 , 29 ], as well as a larger right-hemispheric recruitment of neural resources in the auditory cortex [ 33 ]. However, there has only been very little work investigating the impact of musical training on cortical speech tracking. As a notable exception, Puschmann et al. (2019) found that in musically-trained individuals, attentional modulation of the cortical speech tracking was less pronounced, suggesting that the ignored stream is represented stronger in cortical activity [ 34 ]. However, they did not investigate delta and theta band cortical tracking separately. Here, we sought to differentiate the modulation of the neural tracking in the delta and in the theta band through selective attention, as well as how these individual responses may be shaped by musicianship. We hypothesized that attentional modulation only acts on the higher-level processing associated to the delta band, but not on the lower-level processing in the theta band, since the attentional effects have been observed comparatively late, at delays of 140 ms and later [ 35 ]. We further hypothesized that the attentional modulation of the delta-band tracking is more pronounced in musicians, contributing to an enhanced behavioral performance in speech-in-noise listening. Materials and Methods Experimental Design We used an MEG dataset from 52 participants (26 female, 26 male) aged 24.1 ± 3.1 years that was acquired in the scope of two of our previous studies [ 36 , 37 ]. All participants were right-handed and native German speakers. They had no history of neurological disease or hearing impairment. The study was approved by the ethics board of the University Hospital Erlangen (registration no. 22-361-S). Participants listened to two German audiobooks simultaneously, both of which were narrated by male speakers ( Figure 1 ). Two audiobooks were used alternatingly as attended audiobooks , which the participants were instructed to focus on. The first attended audiobook was ‘Frau Ella’ written by Florian Beckerhoff and read by Peter Jordan ( Speaker 1 ). The second attended audiobook was ‘Den Hund überleben’ written by Stefan Hornbach and read by Pascal Houdus ( Speaker 2 ). Simultaneously, an incoherent excerpt from an unrelated audiobook read by the other speaker, respectively, was presented as ignored audiobook . The first ignored audiobook was thus read by Speaker 2 , Pascal Houdus, and was titled ‘Looking for Hope’ by Colleen Hoover (translated into German by Katarina Ganslandt). The second ignored audiobook was ‘Darum’ by Daniel Glattauer and read by Speaker 1, Peter Jordan. All of the employed audiobooks were published by Hörbuch Hamburg and are available in stores. Speaker 1 read with a mean word frequency in the first attended audio of 3.6 Hz with an average syllable frequency of 5.8 Hz. In the second attended audio, Speaker 2 showed comparable speech characteristics with a mean word frequency of 3.7 Hz and a mean syllable frequency of 5.7 Hz. The audio data, as well as the MEG recordings, will be published upon submission. Download figure Open in new tab Figure 1: Overview of the experimental setup and the data processing pipeline. (a) Participants were presented with two audiobooks while attending one and ignoring the other. (b) Simultaneously, MEG was measured. (c) Based on the acquired data, source reconstruction with focus on the auditory cortex was performed. (d) From the source-level data, the frequency bands of interest were extracted using bandpass filtering. The corresponding audio features were extracted from the acoustic envelopes of the ignored and attended audio, respectively. (e) Lastly, temporal response functions (TRFs) were calculated by training a forward model to estimate the neural features from the audio input features. The resulting TRF magnitudes were compared between the attended and ignored condition, as well as between musicians and non-musicians. The experiment was split into ten trials. Each trial consisted of listening to one chapter from one of the attended audios. Simultaneously, a random duration-matched excerpt from the corresponding ignored audio stream was presented. Once the audiobook chapter ended, participants had to answer three single-choice listening comprehension questions about this chapter to ensure that the participants did indeed focus on the attended audio. After each trial, the attended audio, and thus the attended and ignored speakers, were switched. This was repeated until the first five chapters of both attended audios were presented. The chapters had varying lengths between 3–5 minutes, resulting in a total recording duration of 37 minutes. The audios were presented diotically at equal sound pressure levels of 67 db SPL(A). MEG data were acquired with the participants in a supine position and with their eyes focused on a fixation cross displayed on a screen above. Their head was placed in a 248-channel MEG system ( 4D Neuroimaging , San Diego, CA, USA). Before the start of the measurement, the head shape of each participant was captured using a digitizer ( Polhemus , Colchester, Vermont, Canada). Additionally, the head position relative to the MEG system was recorded using five integrated head-position indicator coils. MEG data was recorded with a sampling frequency of 1017.25 Hz. During online processing, an analog bandpass filter (1 - 200 Hz) and a noise reduction algorithm ( 4D Neuroimaging , San Diego, CA, USA) using 23 reference sensors were applied to eliminate environmental noise. The audio stimuli were presented to the participants using a customized setup that was validated and described in more detail in previous work [ 38 ]. In brief, it consisted of two flexible tubes that were connected to loudspeakers and led into the magnetically-shielded MEG chamber, connecting to earphones that the participants wore. Simultane-ously, the presented audio signal was fed as an additional input channel to the MEG data logger. This enabled the synchronization of MEG recordings and the auditory stimuli with a 1 ms precision. Assessment of musical training Participants were assigned to three groups based on their history of musical training, musicians (M), non-musicians (NM), and neutral participants (N). This categorization was performed using previously established criteria [ 37 ], namely starting age of playing an instrument, total years of undergoing musical training, and whether they were currently practising. To be classified as a musician, a participant had to start playing before the age of seven for a total period of ten or more years. Furthermore, they had to be regularly play an instrument at the time of the study. On the other hand, participants were classified as a non-musician if they did not undergo musical training any longer than three years in sum and only started aged older than seven. Participants who did not fit in either of these categories were assigned to a third, neutral group. Based on these criteria, we acquired data from 25 non-musicians (12 female, 13 male, aged 24.5 ± 3.4 years), 18 musicians (10 female, 8 male, aged 24.1 ± 3.1 years), and 9 neutral participants (4 female, 5 male, aged 24.2 ± 3.9 years). Analysis that disregarded musical training was conducted using the data from all 52 participants. For the comparison between musicians and non-musicians, participants from the neutral group were excluded, resulting in data from 43 participants. This number of participants is in line with previous studies that investigated speech processing with regards to the musicality of a subject [ 29 , 31 , 32 ]. Data Analysis Data processing and analysis were performed in Python using the libraries MNE [ 39 ], SciPy [ 40 ], and statsmodels [ 41 ]. MEG Data Preprocessing First, the acquired MEG recordings were cut to the intervals of interest where an acoustic stimulus was present. These pieces were then concatenated. The MEG data were further processed with a notch filter (50 Hz) to remove powerline interference. Afterwards, the frequency range of interest was extracted using a Butterworth bandpass filter (1 - 20 Hz, order n = 4). This filter was applied forward and backward to prevent introducing a filter shift. To increase computational efficiency, the filtered data was downsampled to 100 Hz. For source reconstruction, we used the average MRI brain template ‘fsaverage’ provided by the FreeSurfer software package [ 42 ]. The computational steps were performed using the Python package MNE [ 39 ]. The ‘fsaverage’ template was aligned to the participant-specific head positioning and head shape that was collected before each measurement. Next, a volume source space consisting of an equidistant grid of sources was created. As our analysis focused on auditory processes, sources were exclusively created in the auditory cortex and its surrounding region, including the middle temporal gyrus, the transverse temporal gyrus, the superior temporal gyrus and its banks, the supramarginal gyrus, and the insular lobe, in both the left and right hemisphere. For the subcortical segmentation and cortical parcellation of the brain regions within the volume, the ‘aparc+aseg’ template from FreeSurfer was employed. Sources were selected to have a distance of 5 mm between one another, resulting in 525 distinct source points. Using the head model and the source space, a forward solution was calculated estimating the magnetic field strength at each MEG channel produced by the sources. To account for the different types of tissue in the brain and their varying conductivities, the boundary element method (BEM) model provided by FreeSurfer for the ‘fsaverage’ template was employed. The resulting leadfield matrix characterizes the sensitivity of each MEG channel to the activity of each of the selected sources. A spatial filter was then applied to reconstruct the source activity from the sensor-level data and the estimated leadfields. Here, a linearly constrained minimum variance (LCMV) beamformer [ 43 ] was employed to estimate the activation of each source independently whilst deducting environmental noise measured in the empty MEG chamber. For analyzing neural tracking in the delta and theta band, the data was further filtered in the corresponding frequency range. To this end, we employed a forward-backward Butterworth bandpass filter (order n = 4) in the range of 1 - 4 Hz and 4 - 8 Hz, respectively. Speech envelope The audio files had a sampling frequency of 44.1 kHz. To retrieve the acoustic speech envelope, the analytic repre-sentation was obtained by applying a Hilbert transform. The magnitude of this analytic signal was then calculated, as it represents the instantaneous amplitude and hence the envelope of this signal. To analyze the neural tracking of the audio, we filtered the speech envelope in three different frequency bands. On the one hand, a broad range of frequencies ranging from 1 - 20 Hz was extracted, which is referred to as broadband feature in the following. On the other hand, two narrower frequency bands were analyzed as well, namely the delta band (1 - 4 Hz) and the theta band (4 - 8 Hz). The envelope was then filtered within the desired frequency bands using a Butterworth bandpass filter (order n = 4) applied forward and backward to prevent filtering delays using the SciPy function sosfiltfilt [ 40 ]. The resulting features were then resampled to 100 Hz to match the sampling rate of the preprocessed, source-level MEG data. Temporal Response Functions (TRFs) To analyze the neural tracking of the extracted audio features, we determined the relation between them and the neural data. Using the extracted audio features as input, we calculated forward models to predict the corresponding neural responses to both the attended and ignored audio and for each frequency band. For this purpose, a linear model was trained that predicted the neural response for every point t in time at every neural source s based on a linear combination of acoustic feature values x t−τ covering the time interval [ t − τ max , t − τ min ]. For each time delay τ , a coefficient was estimated. τ min was chosen to be negative to verify the plausibility of the resulting model weights and to gain information about the noise level present in the TRF. As not all variation in the neural signal can be explained by the auditory input, a residual remains for every investigated time lag: The coefficients are referred to as the TRF of the source point s . The relative magnitudes of these coefficients characterize to which extent each delay contributes to the neural signal, and thus, how strong the neural response to the auditory input is after a certain time lag. Local maxima in the magnitudes of the TRFs indicate that neural responses are particularly pronounced at these time delays. However, unconstrained linear regression is not the optimal approach for speech features as an input signal, as it likely contains temporal correlations causing numerical instability when fitting the model [ 44 ]. Furthermore, due to the high amounts of input data and parameters, the model is prone to overfitting. To address these issues and thus retrieve more robust and reliable results, we employed ridge regression [ 45 , 46 , 47 , 36 ]. The TRF coefficients α (s) ∈ ℝ ( τ max − τ min )×1 were accordingly approximated as where X ∈ ℝ Δ t ×( τ max − τ min ) contains the speech envelope information over the entire stimulus length Δ t , y (s) ∈ ℝ T × 1 describes the neural recording at source s , λ is a predefined regularization parameter, and I refers to the identity matrix. A Python implementation of this TRF coefficient estimation method that was previously developed and utilized by our group [ 46 , 47 ] was used. To retrieve meaningful results, all speech features and neural features were scaled using z-score normalization. To select the regularization parameter, a subject-wise five-fold cross-validation testing 12 parameters ranging from 10 − 5 to 10 5 was applied. The respective optimal λ was calculated based on the prediction that yielded the highest correlation with the measured neural signal. Due to similar results across all participants, the mean optimal λ from the search space was used for all TRF coefficients per investigated frequency band to ensure comparability of the results. The resulting regularization parameters are listed in Table 1 . View this table: View inline View popup Download powerpoint Table 1: Overview of the parameters employed for computing the TRFs The time lag intervals were selected based on visual inspection of the TRF estimates and are depicted for all frequency bands in Table 1 . A step size of 10 ms between each time step was used, which corresponds to the sampling frequency of 100 Hz of the employed features. TRF Peak Extraction Distinct TRF coefficients were estimated for all participants and the two listening modes attended and ignored . Our main focus was to identify the most prominent responses and corresponding time lags and compare their intensities across different groups and conditions. Thus, to simplify the subsequent analysis, only the magnitudes of the individual TRFs were further regarded. The TRF magnitudes were min-max normalized at the population level. To determine the magnitude and average latency of the distinct peaks in the TRFs, the average TRF magnitudes were investigated at the population level. They were calculated by averaging the TRF magnitudes over all source points, all participants, and both presented speakers, for both attended and ignored listening modes. Based on these participants’ average responses, the most prominent peaks and their latencies were identified for each frequency band and listening mode. As the TRF peak latencies differed from one subject to another, we found that information loss occurred when individual peak magnitude values were retrieved by simply considering participants’ average peak latencies. Thus, we also extracted a set of individual TRF peak latencies and magnitudes from the source-averaged TRF magnitudes for each individual participant. To avoid distorting the results with very early or late outlier peaks that may appear in individual TRFs due to noise or artifacts, we defined a fixed search interval around each average peak per frequency band. For every peak type, all local maxima were identified and the respective maximum closest to the average was selected as the individual peak. If no local maximum was found in the search interval, the individual TRF was excluded from the analysis of the respective peak. Attentional Modulation of Cortical Responses To investigate the impact of attention on the cortical responses, TRF magnitudes were investigated on the single-subject level. For each participant, cortical responses to the attended and ignored audios were retrieved, using the envelope from the attended and ignored streams as input features, respectively. Thus, the two listening modes attended and ignored were compared to evaluate differences in cortical tracking of the respective speech envelopes. TRF magnitudes were obtained by averaging over all sources. From these subject-level TRFs, individual TRF peaks characterized by peak amplitudes and peak latencies were extracted as described above and compared between listening modes. To quantify the attentional modulation of the cortical responses for each participant and compare this measure be-tween the non-musician and musician groups, an attentional modulation index AI was calculated for each participant, speaker, frequency band, and TRF peak p as where describes the individual peak amplitude at peak p for the attended and for the ignored audio feature, respectively. Values of AI close to zero indicate similar response strengths to the ignored and attended stream and hence, no attentional modulation. In contrast, positive values of AI suggest the existence of attentional modulation by attenuation of the ignored stream or amplification of the attended stream, whereas negative values of AI imply the opposite effect, i.e., a stronger cortical representation of the ignored than the attended stream. Statistical Analysis All statistical tests were performed using the Python package statsmodels [ 41 ] and the stats -module from the Python package SciPy [ 40 ]. The significance level was set to p < 0.05 for all tests. Significance of Peaks in Population-Level TRFs The population-level TRFs for both the attended and the ignored listening mode were obtained by averaging over the magnitudes of the TRF coefficient from all participants and sources. The intervals in which these population-level TRF magnitudes were significantly different from noise were determined using a bootstrapping approach that compared the magnitudes to a distribution of noise models. Participant- and source-specific noise models were generated by fitting the ridge regression model on a reversed version of the auditory input features. Therefore, the noise models did not contain any meaningful relationship between the input feature and the predicted neural output. The resulting individual noise model weights were randomly shuffled across time lags, channels, and participants and then averaged in the same manner as the individual TRFs to generate the participants’ average noise over all regarded time lags. This procedure was repeated 10.000 times to obtain a noise model distribution. Using this distribution, empirical p -values were calculated for each TRF and time lag based on the amount of noise-level values with a lower magnitude than the investigated actual TRF magnitude. To account for multiple comparisons across all time lags, the resulting p -values were revised using the Bonferroni correction. Significance of Lateralization To determine how the spatial distribution of cortical responses to natural speech differs between earlier and later time delays and across the investigated frequency bands, source-specific average TRFs were calculated. This was achieved by averaging the source-level TRFs over all participants and attended audios. Only the attended audios were incorporated in this analysis, as we wanted to analyze the spatial distribution of responses independent of possible top-down effects introduced by attentional modulation. As a result, the cortical responses to the attended speech envelope for each of the 525 source points in the area of the auditory cortex were retrieved. Snapshots from the source-level TRFs were extracted at each of the previously determined average response peak latencies. The momentary spatial distribution of intensities in the source space at these time lags was visualized through brain plots. To assess whether potential differences in these responses occurring between sources in the right and left hemispheres were statistically significant, a two-sided Mann-Whitney-U test was performed. This test was undertaken with the TRF magnitudes at the latencies of the population-level TRF peaks in all frequency bands. To correct for multiple comparisons for the different peaks within a frequency band, the Bonferroni correction was applied. To further quantify the extent of lateralization at the response peaks yielding significant differences between both hemispheres, a participants’ average lateralization index LI ( p ) for every peak p and frequency band was calculated as already applied similarly in previous literature [ 48 ]: where L ( p ) and R ( p ) describe the sum of peak magnitudes from sources in the left and right hemispheres, respectively. Thus, a negative LI ( p ) value indicates right-lateralized neural activity, whereas a positive LI ( p ) value implies a left-lateralized neural processing. Significance of Attentional Modulation The degree to which attended and ignored speech were tracked differently in the auditory cortex was assessed at the population level using different characteristics of the TRF peaks, such as peak amplitude, peak latency, and the attentional modulation index. When a certain TRF peak was not present in one listening mode for a particular participant, e.g., for the ignored audio, the respective peak was also excluded for the respective other mode, e.g., the attended audio, to allow a mode-wise comparison within participants. Statistical significance of differences in these metrics within the same group of participants, i.e., when comparing attended and ignored speech responses for all participants, were assessed using a two-sided pairwise t-test if the underlying data of the respective characteristic was normally distributed. Otherwise, a two-sided Wilcoxon signed-rank test was applied. The normality of the analyzed samples was determined using the Shapiro-Wilk test. When comparing characteristics between musicians and non-musicians, a two-sided unpaired t-test was utilized for normally distributed characteristics. Alternatively, for samples that did not follow a normal distribution, a two-sided Mann-Whitney-U test was utilized. For all tests within a frequency band, the Bonferroni correction was applied to compensate for multiple comparisons across the different TRF peaks. Relationship of Listening Comprehension and Cortical Response Strength During the study, participants answered single-choice listening comprehension questions regarding the attended audio stream. The percentage of correct answers per participant and attended audiobook was captured as a behavioral metric to quantify the performance of the individual participant. To relate the cortical response to this behavioral measure, Pearson’s correlation coefficients were calculated between the comprehension score and each TRF peak per feature and listening mode. The Bonferroni correction was used to adjust for multiple comparisons between different TRF peaks for all correlations within a frequency band. Furthermore, to evaluate the listening performance of musicians and non-musicians on a purely behavioral level, a two-sided unpaired t-test was performed to compare the comprehension scores of the two groups. Relationship of Musicality Features and Cortical Response Strength As detailed above, participants were asked about different aspects of their musical training: their starting age of playing an instrument, total years of undergoing musical training, amount of current training, and the total number of instruments played. (Note that only the first three aspects were used to classify the participants into the three categories.) Using a linear regression model, we assessed if the TRF peak characteristics related to these reported musicality aspects. A regression model was estimated for each peak using ordinary least squares. Due to the large number of estimates, this analysis was only performed for the delta and theta bands. All independent and dependent variables were scaled beforehand using z-score normalization. The musicality aspects served as independent variables, whereas the peak values were used as the dependent predicted features. Results Temporal and Spatial Distribution of Cortical Responses to Attended Speech As a first step, we assessed the temporal and spatial distribution of the neural speech tracking in the three employed frequency bands at the population level ( Figure 2 ). For the broadband responses, the average magnitude of the TRFs exhibited peaks at the time lags 110 ms and 240 ms ( Figure 2a ). In the delta band, peak neural responses emerged at the time lags 100 ms, 270 ms, and 540 ms. The cortical tracking in the theta band led to a main peak at 130 ms. The symmetrical sidelobes around this peak are caused by the narrow bandwidth of the speech feature and were therefore not regarded as independent peaks. Download figure Open in new tab Figure 2: Population-level magnitudes of the averaged temporal response functions (TRFs) of attended speech and their spatial distri-bution for the distinct frequency bands. Panel (a) shows the mean magnitudes of the TRFs across all source channels and participants for attended speech for each investigated frequency band. Based on these functions, the time lags at which cortical responses were maximal (dotted vertical lines) were extracted and assigned to one of the response types M100 TRF , M200 TRF , or M400 TRF . Panel (b) exhibits the distribution of TRF magnitudes at these maximum response times in the source volume ranging from low activation (yellow) to high activation (dark red). For brainplots showing significant differences in activations of right and left hemispheres, the respective lateralization indices (LI) are reported. The colored area corresponds to the investigated source region of interest In the following, responses peaking around 100 ms and 200 ms are referred to as M100 TRF and M200 TRF , respectively. Furthermore, the late TRF response peak observed in the delta band is referred to as M400 TRF in analogy to the N400 component in event-related potentials [ 49 ]. We thus found a M100 TRF in all three frequency bands, a M200 TRF in the broadband and the delta band, and a M400 TRF solely in the delta band. M100 TRF showed a highly significant right-lateralization in all three frequency bands, with lateralization indices LI = 0.26 ( p = 6e − 19) in the broadband, LI = 0.36 ( p = 1e − 40) in the delta band, and LI = 0.15 ( p = 2e − 11) in the theta band ( Figure 2b ). A similar degree of right lateralization was observed for M200 TRF with LI = 0.26 ( p = 1e − 31) in the broadband, and LI = 0.36 ( p = 5e − 46) in the delta band. In contrast, M400 TRF showed a bilateral distribution of activations ( LI = 0.01, p = 0.99). Attentional Modulation of Cortical Responses In the next analysis step, TRF responses to the ignored audios were compared to the attended responses at the population level ( Figure 3 ). For all three frequency bands, the TRF magnitudes were significantly higher than those of the noise model for the majority of the analyzed positive time lags. Download figure Open in new tab Figure 3: Population-level TRF magnitudes in the attended and ignored condition, as well as their relation to the comprehension scores. (a) The TRF magnitudes in the attended condition (black solid) are larger than those in the ignored condition (grey dashed) for the broadband and the delta-band response, but not in the theta band. The horizontal bars indicate the areas where the TRF magnitudes significantly differ from the noise models. Peaks with significant differences in amplitude between the attended and ignored condition are marked with asterisks indicating their significance level (*, 0.01 ≤ p < 0.05; **, 1e − 3 ≤ p < 0.01, ***, 1e − 4 ≤ p < 1e − 3, ****, p < 1e − 4). The search areas for the participant-wise TRF response peaks are shaded in grey. (b) The individual magnitudes at each peak were extracted per subject and related to the percentage of correct comprehension question answers through Pearson’s correlation coefficient r . A significant positive correlation between response magnitude and comprehension was found in the theta band, both for the attended and the ignored condition Analogous to the attended TRFs, the ignored TRFs in the broadband condition exhibited both M100 TRF (time lag 110 ms) and M200 TRF (time lag 200 ms). The same responses were also visible in the ignored TRFs in the delta band, with M100 TRF and M200 TRF occurring with average delays of 80 ms and 250 ms, respectively. Additionally, weak M400 TRF response was extracted at 520 ms to allow a direct comparison with the attended responses. As for the attended TRFs, the ignored TRFs in the theta band showed a single main peak, M100 TRF . To quantify the differences between responses in the attended and ignored condition, individual time lags were extracted for each participant, listening scenario, and response type. The utilized search intervals for each frequency band are depicted in Table 2 . For the broadband feature, a M100 TRF was found in 94 % of all participant-level TRFs, whereas a M200 TRF was present in 91 % of broadband TRFs. In the delta band, M100 TRF appeared in 84 %, M200 TRF in 91 %, and M400 TRF in 90 % of all participant-level TRFs. In the theta band, a M100 TRF was identified for every subject. View this table: View inline View popup Download powerpoint Table 2: Search intervals for the peaks in the subject-level TRF magnitudes for the different frequency bands The amplitude of M100 TRF was significantly higher in the attended than in the ignored condition in the broadband ( p = 1e − 9) and the delta band ( p = 1e − 12). In the theta band, however, no significant differences between attended and ignored response amplitude were found. The same applies to M200 TRF in the broadband TRFs, where no considerable differences in attended and ignored response amplitudes were discernible. In contrast, in the delta band, both M200 TRF and M400 TRF magnitudes were significantly greater in the attended TRFs compared to the ignored TRFs (M200 TRF : p = 2e − 5; M400 TRF : p = 2e − 8). Furthermore, noticeable differences between attended and ignored responses were visible not only in the TRF mag-nitudes, but also in the time lags of the TRF peaks. Both M100 TRF and M200 TRF showed significant differences in delay, with the attended response occurring later than the ignored response in the broadband (M100 TRF : p = 8e − 3, M200 TRF : p = 2e − 11), the delta band (M100 TRF : p = 0.01, M200 TRF : p = 7e − 5), and also the theta band (M100 TRF : p = 5.0e − 4). No significant differences in response peak time lags were found for M400 TRF . These dif-ferences were also mostly reflected in the mean TRF peak time lags shown in Figure 3a , where the attended response peaks occurred with a delay of 10 ms to 40 ms compared to the ignored response peaks in all three frequency bands. Correlation Between Neural Responses and Participant Behavior To investigate whether the neural responses were related to behavior, we computed Pearson’s correlation coefficient between the subject-level TRF magnitudes and the subject’s percentage of correctly answered comprehension ques-tions ( Figure 3b ). For both the attended and the ignored listening mode, correlations were either close to zero, indicating that there was no relationship, or slightly positive ranging up to r = 0.25 (M100 TRF for attended audio input in the theta band). However, only the Pearson’s correlation coefficients estimated for the theta band were sig-nificant, revealing similar positive correlations for both ignored ( r = 0.23 , p = 0.04) and attended ( r = 0.25 , p = 0.02) modes. There was no significant difference in the comprehension scores between musicians and non-musicians ( p = 0.35). Cortical Responses in Musicians and Non-Musicians To assess the putative influence of musical training on the cortical speech tracking, we compared the neural responses for the group of musicians to that of non-musicians ( Figure 4 ). Similar to the population-level TRF magnitudes depicted in Figure 3a , for both groups attention-induced differences in the TRF magnitudes are visible in both the broadband and in the delta band. At the same time, no considerable differences between ignored and attended responses are discernible in the theta band. Download figure Open in new tab Figure 4: Magnitudes of the temporal response functions (TRFs) averaged across all sources for non-musicians (NM, blue) and musicians (M, red) in the attended condition (solid) and in the ignored case (dashed). The search areas for the subject-wise TRF peaks are shaded in grey. They were estimated based on the participants’ average signal and are thus identical for M and NM and ignored and attended responses. For all three frequency bands and both attended and ignored responses, a tendency towards higher M100 TRF peaks in non-musicians can be seen compared to the musicians. For M100 TRF peaks across all frequency bands and listening modes, the mean magnitude was smaller in musicians than in non-musicians. This difference was most pronounced for the ignored M100 TRF response in the delta band with a mean increase of 23.0% in non-musicians compared to musicians. M200 TRF is also slightly larger for non-musicians than musicians in both the attended (9.6%) and ignored (16.5%) broadband TRF and the attended delta-band TRF (7.4%). For the remaining TRF peaks, no noticeable differences emerged. A difference in latencies could be observed in the attended M200 TRF in the broadband response, as the musicians’ average peaked 40 ms after the non-musicians at 260 ms. A similar effect was discernible in the delta-band M200 TRF in the ignored condition, where the musicians’ peak was delayed by 20 ms. Peak latencies were otherwise similar in both groups. To assess the statistical significance of these differences, the subject-level TRF peaks and their time lags were compared between musicians and non-musicians. Furthermore, intra-group differences between ignored and attended response magnitudes characterized by the attentional modulation index AI were quantified. The results for all three frequency bands are depicted in Figure 5 . Download figure Open in new tab Figure 5: Comparison of subject-level TRF peak amplitudes (upper row), latencies (middle row) and attentional modulation indices (lower row) between musicians and non-musicians. For non-musicians, slight tendencies towards higher peak values in ignored M100 TRF and M200 TRF are recognizable, but only the ignored delta band M100 TRF is slightly significant ( p = 0.049). For musicians, a significantly longer latency occurs for the M200 TRF in the ignored delta band. In turn, musicians show marginally larger attentional modulation indices for these response types. Peaks with significant differences in amplitude between the attended and ignored condition are marked with asterisks indicating their significance level (*, 0.01 ≤ p < 0.05; **, 1e − 3 ≤ p < 0.01) A tendency of increased M100 TRF amplitude in non-musicians compared to musicians was still observable. However, these differences were only slightly significant for the ignored delta M100 TRF ( p = 0.049). All other disparities were not significant. The same applied to the latency of the M200 TRF responses, except for the latency of the M200 TRF in the delta band in the ignored condition, which was found to be significantly longer in musicians than non-musicians ( p = 0.03). In turn, attentional modulation was highest at M100 TRF , with musicians tending to show larger AI values than non-musicians in the broadband and the delta-band response. This difference was most pronounced in the delta band yielding an average attentional modulation index AI = 0.27 for musicians and AI = 0.20 for non-musicians. Additionally, in the broadband M200 TRF , musicians showed a stronger representation of the attended feature yielding AI = 0.1 on average, whereas non-musicians had a modulation value of AI = − 0.01, suggesting no modulation of attended and ignored cortical representations. However, none of these differences in the attentional modulation coefficient were statistically significant. Furthermore, in the theta band, the mean AI values of both groups were negligible ( AI = 0.01). Relationship of Musicality Features and Cortical Responses As a more fine-grained assessment of the influence of musical training on cortical speech tracking, we investigated whether there were relations between the subject’s different aspects of musical training and the TRF peak amplitudes. The estimated contributions of each of the aspects of musical training to predicting the neural responses are depicted in Figure 6 . No considerable contributions were found for the weekly practice feature. Feature weights for the number of instruments indicate a positive contribution of this property to the peak amplitudes throughout all frequency bands and listening scenarios, but did not reach statistical significance. Only two features, namely the starting age of learning an instrument and and the playing duration in years, were significant predictors. The former significantly contributed to the attended M200 TRF in the delta band ( p = 0.01), with an assigned feature weight of 0.54, whereas the latter exhibited a negative relation with the TRF peak amplitudes. This observation became significant for the ignored M100 TRF in the delta band, where the smallest model coefficient was estimated ( − 0.55, p = 0.04). Download figure Open in new tab Figure 6: Influence of the individual aspects of musical training on the cortical responses. For each response type, frequency band, and attention mode, a multiple linear regression model was fitted to pre-dict the response amplitude from the four aspects of musical train-ing (hours of current weekly practice, starting age of playing an in-strument, total duration of playing instruments, and the number of played instruments). The resulting weights of the linear model are assessed for statistical significance ( p -value displayed in the cor-responding cell). Two correlated aspects of musical training were found to be significant, the starting age and the duration of play-ing. Whereas the former is positively contributing to the attended M200 TRF response in the delta band (0.54), the latter is negatively affecting the ignored M100 TRF delta band response ( − 0.55). Discussion In this work, we utilized a large dataset of MEG recordings from 52 participants, each of whom listened to 37 minutes of continuous speech. We thereby found that the attentional modulation of the cortical speech tracking occurred in the delta band but not in the theta band. We further found that musical training had only a slight influence on the neural responses, with most aspects of the latter remaining unaffected by musicianship. Different Types of Cortical Responses Masked in Broadband Become Apparent in the Delta and Theta Band The speech envelope is a basic auditory feature that conveys various acoustic and linguistic information [ 14 , 13 ]. The low-frequency range, below 20 Hz, is of particular interest, as it contains the rates of words and syllables. Previous work has not only shown strong cortical envelope tracking in this frequency range [ 15 ] but also hypothesized that distinct functions can be attributed to delta- and theta-band tracking [ 25 ]. In particular, theta-band tracking is assumed to reflect lower-level syllable processing [ 50 , 26 , 23 ], whereas higher-level word processing is associated with delta-band responses [ 51 , 52 , 26 ]. Our results show noticeable differences between the cortical responses in the delta and theta bands, which remain obscured in the broadband due to the superposition of the individual components. Whereas in the theta band the most prominent response occurs around 100 ms, which is also visible in the delta band and broadband, additional response peaks are present in the delta band around 250 ms and 500 ms. Even though the latter is considerably weaker than the other response peaks, it is still clearly recognizable. In turn, in the broadband signal, the M200 TRF is discernible but has a much lower amplitude than in the delta band. No later response peak emerges in the broadband response. Low-Frequency Speech Envelope Tracking is Mainly Right-Lateralized To obtain a better understanding of the underlying mechanisms of the observed cortical responses, we also looked at their spatial distribution and lateralization. Except for the late M400 TRF , all cortical responses were significantly right-lateralized, with the highest activations centered in Heschl’s gyrus, i.e., the primary auditory cortex. This is in line with previous findings that also found a right-lateralization of low-frequency envelope tracking [ 53 ]. Furthermore, bottom-up processes on the sublexical level are dominated by the right hemisphere [ 54 , 55 ]. For top-down semantic processing starting from the word level, however, a lateralization shift towards the left hemisphere has been observed in previous studies, which might cause the lack of lateralization of the M400 TRF observed here [ 24 , 51 ]. Delta and Theta Band Responses are Modulated Differently by Selective Attention It is well-researched that in a competing speaker scenario, the envelopes of both speech streams are tracked individ-ually in the auditory cortex, with the cortical response to the ignored speaker being weaker than to the attended speaker [ 1 ]. We recover this behavior in the broadband for the M100 TRF but not for the M200 TRF . Because the broadband response contains both the delta and the theta band, the lack of attentional modulation of its M200 TRF peak is presumably due to the conflation of these two individual frequency bands. Therefore, we focused on the separate assessment of the cortical tracking in the delta and theta bands. In the delta band, attended speech leads to significantly stronger tracking than ignored speech, across a large range of delays and encompassing the M100 TRF , M200 TRF , and M400 TRF . In contrast, the theta band did not exhibit attentional modulation. However, we found a significant positive relationship between the theta band’s M100 TRF and the behavioral performance of the participants, which is in line with the crucial role of theta-band tracking in speech-in-noise intelligibility [ 22 ]. Because an M100 TRF emerges both in the delta- and in the theta-band tracking, but is only affected by attention for the delta-band response, this peak is likely not evoked by acoustic activity like an event-related potential. In fact, already the earliest evoked potentials in the cortex are modulated by attention [ 56 ], as opposed to the theta band’s M100 TRF . Therefore, we conclude from our results that delta- and theta-band tracking of the envelope are continuously executed in parallel, whereby theta tracking provides bottom-up information about the unfiltered auditory input, while in the delta band, background information is attenuated and further processed for linguistic information. A time delay of around 10 ms between the peaks in the responses to ignored and attended speech was reported in previous work [ 1 ], with the attended response peak occurring earlier than the ignored response peak. However, in our study, we observed the opposite effect: the response peaks of the ignored stream appeared significantly before the response peaks of the attended stream. This could indicate that the ignored stream is processed less thoroughly than the attended one, and therefore the response arises sooner. Few Significant Differences between Musicians and Non-Musicians When assessing the influence of musical training on cortical speech tracking, we only found minor significant effects on the M100 TRF and M200 TRF in the delta band. For the ignored speech signal, the delta band M100 TRF peak was slightly greater in non-musicians than in musicians. Furthermore, its magnitude was negatively impacted by the duration over which participants had played an instrument. Furthermore, the ignored M200 TRF peak was slightly delayed in musicians than in non-musicians. Additionally, its magnitude resulting from the attended speech signal had a positive correlation with the starting age of practicing an instrument. The other components of the neural speech tracking were not significantly affected by musicianship, although we observed a general trend of lower amplitudes for musicians as compared to non-musicians. To interpret these findings, we note that, in contrast to intra-subject modulation effects, where stronger tracking of a speech signal in background noise is indicative of better intelligibility, this does not necessarily apply when comparing cortical responses between participants. Previous studies investigating speech perception in the elderly in fact demonstrated that stronger cortical speech tracking correlated with lower intelligibility scores [ 57 , 58 ]. The strength of the neural tracking might thus not indicate improved comprehension, but rather increased listening effort. In line with these deliberations, Jantzen et al. (2014) observed that non-musicians showed increased cortical activity, especially for early responses [ 33 ]. This finding matches with the tendencies towards higher response magnitudes in non-musicians, which we observed here (despite a lack of statistical significance). At the same time, this might explain the slightly higher attentional modulation indices for early delta responses in musicians. This ‘inverse’ relationship between musicianship and the strength of the cortical tracking is also reflected in the relation between the aspects of musical training and the neural responses that we determined. The earlier the participants started playing an instrument, the weaker the delta-band responses were. This aspect of musical training hence seems to have a strong influence on speech-in-noise perception, which is consistent with the results of other studies [ 2 , 8 ]. We expect the M400 TRF to be most closely associated with semantic integration and contextualization of the speech signals. This response was remarkably similar between musicians and non-musicians. Together with the insignificance of most of the other aspects of the neural speech tracking that we examined, we conclude that cortical speech tracking is essentially not affected by musical training. The behavioral improvements in speech-in-noise perception observed in musicians appear to originate in other neural mechanisms, probably further downstream of the cortical speech tracking since the latter already involves early auditory responses. Conclusions In summary, our study showed that the attentional modulation of cortical speech tracking results from the delta band but not from the theta band. The theta band presumably reflects lower-level acoustic processing such as syllable parsing, which appears to be equally executed for the attended and the ignored speech stream. Neural activity in the delta band, in contrast, is responsible for higher-level linguistic processing, and our results show that attentional effects emerge at this stage. Musical training leaves both types of responses largely unchanged, even though we observed a tendency towards stronger cortical speech tracking in people with less musical training. Data availability The MEG data used in this study is available from the authors upon request. Author contributions Alina Schüller, Annika Mücke : Conceptualization, Interpretation of Results of Experiments, Formal analysis, Investigation, Writing original draft. Jasmin Riegel : Conceptualization, Data Acquisition, Reviewing and Editing of the paper. Tobias Reichenbach : Conceptualization, Interpretation of Results of Experiments, Supervision, Reviewing and Editing of the paper. Conflict of interest statement The authors declare no competing financial interests. Acknowledgments This project was supported by the German Federal Ministry of Education and Research (Cluster4Future SEMECO, project number 03ZU1210FB) and the German Science Foundation (DFG, project number 523344822). References [1]. ↵ N. Ding , J. Z. Simon , Emergence of neural encoding of auditory objects while listening to competing speakers , Proceedings of the National Academy of Sciences of the United States of America 109 ( 29 ) ( 2012 ) 11854 – 11859 . OpenUrl Abstract / FREE Full Text [2]. ↵ N. Kraus , B. Chandrasekaran , Music training for the development of auditory skills , Nature Reviews Neuro- science 11 ( 8 ) ( 2010 ) 599 – 605 . OpenUrl [3]. ↵ E. B. Coffey , N. B. Mogilever , R. J. Zatorre , Speech-in-noise perception in musicians: A review , Hearing Research 352 ( 2017 ) 49 – 69 . OpenUrl CrossRef PubMed [4]. ↵ P. E. Souza , K. T. Boike , K. Witherell , K. Tremblay , Prediction of speech recognition from audibility in older listeners with hearing loss: Effects of age, amplification, and background noise , Journal of the American Academy of Audiology 18 ( 1 ) ( 2007 ) 54 – 65 . OpenUrl CrossRef PubMed Web of Science [5]. ↵ A. Parbery-Clark , D. L. Strait , S. Anderson , E. Hittner , N. Kraus , Musical experience and the aging auditory system: Implications for cognitive abilities and hearing speech in noise , PLoS ONE 6 ( 5 ) ( 2011 ) e18082 . OpenUrl CrossRef PubMed [6]. ↵ Y. Du , R. J. Zatorre , Musical training sharpens and bonds ears and tongue to hear speech better , Proceedings of the National Academy of Sciences 114 ( 51 ) ( 2017 ) 13579 – 13584 . OpenUrl Abstract / FREE Full Text [7]. ↵ D. Strait , N. Kraus , Can you hear me now? Musical training shapes functional brain networks for selective auditory attention and hearing speech in noise , Frontiers in Psychology 2 ( 2011 ). [8]. ↵ S. Puschmann , M. Regev , S. Baillet , R. J. Zatorre , MEG intersubject phase locking of stimulus-driven activity during naturalistic speech listening correlates with musical training , Journal of Neuroscience 41 ( 12 ) ( 2021 ) 2713 – 2722 . OpenUrl Abstract / FREE Full Text [9]. ↵ A. Parbery-Clark , E. Skoe , N. Kraus , Musical experience limits the degradative effects of background noise on the neural processing of sound , Journal of Neuroscience 29 ( 45 ) ( 2009 ) 14100 – 14107 . OpenUrl Abstract / FREE Full Text [10]. ↵ K. K. Clayton , J. Swaminathan , A. Yazdanbakhsh , J. Zuk , A. D. Patel , G. K. Jr , Executive function, visual attention and the cocktail party problem in musicians and non-musicians , PLOS ONE 11 ( 7 ) ( 2016 ) e0157638 . OpenUrl CrossRef [11]. ↵ J. Zuk , O. Ozernov-Palchik , H. Kim , K. Lakshminarayanan , J. D. E. Gabrieli , P. Tallal , N. Gaab , Enhanced syllable discrimination thresholds in musicians , PLoS ONE 8 ( 12 ) ( 2013 ) e80546 . OpenUrl CrossRef PubMed [12]. ↵ Y.-P. Chen , F. Schmidt , A. Keitel , S. Rösch , A. Hauswald , N. Weisz , Speech intelligibility changes the temporal evolution of neural speech tracking , NeuroImage 268 ( 2023 ) 119894 . OpenUrl CrossRef [13]. ↵ C. Brodbeck , J. Z. Simon , Continuous speech processing , Current Opinion in Physiology 18 ( 2020 ) 25 – 31 . OpenUrl [14]. ↵ M. Gillis , J. Van Canneyt , T. Francart , J. Vanthornhout , Neural tracking as a diagnostic tool to assess the auditory pathway , Hearing Research 426 ( 2022 ) 108607 . OpenUrl [15]. ↵ N. Ding , J. Z. Simon , Adaptive temporal encoding leads to a background-insensitive cortical representation of speech , Journal of Neuroscience 33 ( 13 ) ( 2013 ) 5728 . OpenUrl Abstract / FREE Full Text [16]. ↵ J. A. O’Sullivan , A. J. Power , N. Mesgarani , S. Rajaram , J. J. Foxe , B. G. Shinn-Cunningham , M. Slaney , S. A. Shamma , E. C. Lalor , Attentional selection in a cocktail party environment can be decoded from single-trial EEG , Cerebral Cortex 25 ( 7 ) ( 2015 ) 1697 – 1706 . OpenUrl CrossRef PubMed [17]. ↵ A. Wilsch , T. Neuling , J. Obleser , C. S. Herrmann , Transcranial alternating current stimulation with speech envelopes modulates speech comprehension , NeuroImage 172 ( 2018 ) 766 – 774 . OpenUrl [18]. ↵ B. Zoefel , A. Archer-Boyd , M. H. Davis , Phase entrainment of brain oscillations causally modulates neural responses to intelligible speech , Current Biology 28 ( 3 ) ( 2018 ) 401 – 408 . OpenUrl CrossRef PubMed [19]. ↵ L. Riecke , E. Formisano , B. Sorger , D. Başkent , E. Gaudrain , Neural entrainment to speech modulates speech intelligibility , Current Biology 28 ( 2 ) ( 2018 ) 161 – 169 . OpenUrl CrossRef PubMed [20]. ↵ M. Keshavarzi , M. Kegler , S. Kadir , T. Reichenbach , Transcranial alternating current stimulation in the theta band but not in the delta band modulates the comprehension of naturalistic speech in noise , NeuroImage 210 ( 2020 ) 116557 . OpenUrl [21]. ↵ S. Kadir , C. Kaza , H. Weissbart , T. Reichenbach , Modulation of speech-in-noise comprehension through tran- scranial current stimulation with the phase-shifted speech envelope , IEEE Transactions on Neural Systems and Rehabilitation Engineering 28 ( 1 ) ( 2019 ) 23 – 31 . OpenUrl [22]. ↵ N. Ding , J. Z. Simon , Cortical entrainment to continuous speech: Functional roles and interpretations , Frontiers in Human Neuroscience 8 ( 2014 ). [23]. ↵ G. M. Di Liberto , J. A. O’Sullivan , E. C. Lalor , Low-frequency cortical entrainment to speech reflects phoneme- level processing , Current Biology 25 ( 19 ) ( 2015 ) 2457 – 2465 . OpenUrl CrossRef PubMed [24]. ↵ C. Brodbeck , A. Presacco , J. Z. Simon , Neural source dynamics of brain responses to continuous stimuli: Speech processing from acoustics to comprehension , NeuroImage 172 ( 2018a ) 162 – 174 . OpenUrl CrossRef PubMed [25]. ↵ G. Mai , W. S.-Y. Wang , Distinct roles of delta- and theta-band neural tracking for sharpening and predictive coding of multi-level speech features during spoken language processing , Human Brain Mapping 44 ( 17 ) ( 2023 ) 6149 – 6172 . OpenUrl [26]. ↵ O. Etard , T. Reichenbach , Neural speech tracking in the theta and in the delta frequency band differentially encode clarity and comprehension of speech in noise , Journal of Neuroscience 39 ( 29 ) ( 2019 ) 5750 – 5759 . OpenUrl Abstract / FREE Full Text [27]. ↵ T. Van Hirtum , B. Somers , E. Verschueren , B. Dieudonné , T. Francart , Delta-band neural envelope tracking predicts speech intelligibility in noise in preschoolers , Hearing Research 434 ( 2023 ) 108785 . OpenUrl CrossRef [28]. ↵ B. R. Zendel , C. Alain , Concurrent sound segregation is enhanced in musicians , Journal of Cognitive Neuroscience 21 ( 8 ) ( 2009 ) 1488 – 1498 . OpenUrl CrossRef PubMed Web of Science [29]. ↵ A. Parbery-Clark , S. Anderson , E. Hittner , N. Kraus , Musical experience strengthens the neural representation of sounds important for communication in middle-aged adults , Frontiers in Aging Neuroscience 4 ( 2012 ). [30]. ↵ B. R. Zendel , C.-D. Tremblay , S. Belleville , I. Peretz , The impact of musicianship on the cortical mechanisms related to separating speech from background noise , Journal of Cognitive Neuroscience 27 ( 5 ) ( 2015 ) 1044 – 1059 . OpenUrl CrossRef [31]. ↵ G. Musacchia , M. Sams , E. Skoe , N. Kraus , Musicians have enhanced subcortical auditory and audiovisual processing of speech and music , Proceedings of the National Academy of Sciences 104 ( 40 ) ( 2007 ) 15894 – 15898 . OpenUrl Abstract / FREE Full Text [32]. ↵ A. Parbery-Clark , A. Tierney , D. L. Strait , N. Kraus , Musicians have fine-tuned neural distinction of speech syllables , Neuroscience 219 ( 2012 ) 111 – 119 . OpenUrl CrossRef PubMed Web of Science [33]. ↵ M. Jantzen , B. Howe , K. Jantzen , Neurophysiological evidence that musical training influences the recruitment of right hemispheric homologues for speech perception , Frontiers in Psychology 5 ( 2014 ). [34]. ↵ S. Puschmann , S. Baillet , R. J. Zatorre , Musicians at the cocktail party: Neural substrates of musical training during selective listening in multispeaker situations , Cerebral Cortex 29 ( 8 ) ( 2019 ) 3253 – 3265 . OpenUrl CrossRef PubMed [35]. ↵ C. Brodbeck , A. Jiao , L. E. Hong , J. Z. Simon , Neural speech restoration at the cocktail party: Auditory cortex recovers masked speech of both attended and ignored speakers , PLoS Biology 18 ( 10 ) ( 2020 ) e3000883 . OpenUrl CrossRef PubMed [36]. ↵ A. Schüller , A. Schilling , P. Krauss , S. Rampp , T. Reichenbach , Attentional modulation of the cortical contri- bution to the frequency-following response evoked by continuous speech , Journal of Neuroscience 43 ( 44 ) ( 2023 ) 7429 – 7440 . OpenUrl Abstract / FREE Full Text [37]. ↵ J. Riegel , A. Schüller , T. Reichenbach , No influence of musical training on the cortical contribution to the speech-FFR and its modulation through selective attention , In review ( 2024 ). [38]. ↵ A. Schilling , R. Tomasello , M. R. Henningsen-Schomers , A. Zankl , K. Surendra , M. Haller , V. Karl , P. Uhrig , A. Maier , P. Krauss , Analysis of continuous neuronal activity evoked by natural speech with computational corpus linguistics methods, Language , Cognition and Neuroscience 36 ( 2 ) ( 2021 ) 167 – 186 . OpenUrl [39]. ↵ A. Gramfort , M. Luessi , E. Larson , D. A. Engemann , D. Strohmeier , C. Brodbeck , R. Goj , M. Jas , T. Brooks , L. Parkkonen , M. S. Hämäläinen , MEG and EEG data analysis with MNE-Python , Frontiers in Neuroscience 7 ( 267 ) ( 2013 ) 1 – 13 . OpenUrl [40]. ↵ P. Virtanen , R. Gommers , T. E. Oliphant , M. Haberland , T. Reddy , D. Cournapeau , E. Burovski , P. Peterson , W. Weckesser , J. Bright , S. J. van der Walt , M. Brett , J. Wilson , K. J. Millman , N. Mayorov , A. R. J. Nelson , E. Jones , R. Kern , E. Larson , C. J. Carey , İ. Polat , Y. Feng , E. W. Moore , J. VanderPlas , D. Laxalde , J. Perktold , R. Cimrman , I. Henriksen , E. A. Quintero , C. R. Harris , A. M. Archibald , A. H. Ribeiro , F. Pedregosa , P. van Mulbregt , SciPy 1.0 Contributors, Scipy 1.0: Fundamental algorithms for scientific computing in Python , Nature Methods 17 ( 2020 ) 261 – 272 . OpenUrl [41]. ↵ S. Seabold , J. Perktold , statsmodels: Econometric and statistical modeling with Python , in: 9th Python in Science Conference , 2010 . [42]. ↵ B. Fischl , FreeSurfer , NeuroImage 62 ( 2 ) ( 2012 ) 774 – 781 . OpenUrl CrossRef PubMed Web of Science [43]. ↵ B. Van Veen , W. Van Drongelen , M. Yuchtman , A. Suzuki , Localization of brain electrical activity via linearly constrained minimum variance spatial filtering , IEEE Transactions on Biomedical Engineering 44 ( 9 ) ( 1997 ) 867 – 880 . OpenUrl CrossRef PubMed Web of Science [44]. ↵ M. J. Crosse , G. M. Di Liberto , A. Bednar , E. C. Lalor , The multivariate temporal response function (mTRF) toolbox: A MATLAB toolbox for relating neural signals to continuous stimuli , Frontiers in Human Neuroscience 10 ( Nov. 2016 ). [45]. ↵ W. Biesmans , N. Das , T. Francart , A. Bertrand , Auditory-inspired speech envelope extraction methods for improved EEG-based auditory attention detection in a cocktail party scenario , IEEE Transactions on Neural Systems and Rehabilitation Engineering 25 ( 5 ) ( 2017 ) 402 – 412 . OpenUrl CrossRef [46]. ↵ O. Etard , M. Kegler , C. Braiman , A. E. Forte , T. Reichenbach , Decoding of selective attention to continuous speech from the human auditory brainstem response , NeuroImage 200 ( 2019 ) 1 – 11 . OpenUrl CrossRef PubMed [47]. ↵ M. Kegler , H. Weissbart , T. Reichenbach , The neural response at the fundamental frequency of speech is modulated by word-level acoustic and linguistic information , Frontiers in Neuroscience 16 ( Jul. 2022 ). [48]. ↵ M. L. Seghier , Laterality index in functional MRI: Methodological issues , Magnetic Resonance Imaging 26 ( 5 ) ( 2008 ) 594 . OpenUrl CrossRef PubMed Web of Science [49]. ↵ M. Kutas , K. D. Federmeier , Thirty years and counting: Finding meaning in the N400 component of the event-related brain potential (ERP) , Annual Review of Psychology 62 ( 2011 ) 621 – 647 . OpenUrl CrossRef PubMed Web of Science [50]. ↵ A. Hyafil , L. Fontolan , C. Kabdebon , B. Gutkin , A.-L. Giraud , Speech encoding by coupled cortical theta and gamma oscillations , eLife 4 ( 2015 ) e06213 . OpenUrl CrossRef PubMed [51]. ↵ C. Brodbeck , L. E. Hong , J. Z. Simon , Rapid transformation from auditory to linguistic representations of continuous speech , Current Biology 28 ( 24 ) ( 2018b ) 3976 – 3983 .e5. OpenUrl CrossRef PubMed [52]. ↵ M. P. Broderick , A. J. Anderson , G. M. Di Liberto , M. J. Crosse , E. C. Lalor , Electrophysiological correlates of semantic dissimilarity reflect the comprehension of natural, narrative speech , Current Biology 28 ( 5 ) ( 2018 ) 803 – 809 . OpenUrl CrossRef PubMed [53]. ↵ M. F. Assaneo , J. M. Rimmele , J. Orpella , P. Ripollés , R. de Diego-Balaguer , D. Poeppel , The lateralization of speech-brain coupling is differentially modulated by intrinsic auditory and top-down mechanisms , Frontiers in Integrative Neuroscience 13 ( 2019 ). [54]. ↵ C. Brodbeck , S. Bhattasali , A. A. Cruz Heredia , P. Resnik , J. Z. Simon , E. Lau , Parallel processing in speech perception with local and global representations of linguistic context , eLife 11 ( 2022 ) e72056 . OpenUrl CrossRef [55]. ↵ H. Luo , D. Poeppel , Phase patterns of neuronal responses reliably discriminate speech in human auditory cortex , Neuron 54 ( 6 ) ( 2007 ) 1001 – 1010 . OpenUrl CrossRef PubMed Web of Science [56]. ↵ V. Poghosyan , A. A. Ioannides , Attention modulates earliest responses in the primary auditory and visual cortices , Neuron 58 ( 5 ) ( 2008 ) 802 – 813 . OpenUrl CrossRef PubMed [57]. ↵ A. Presacco , J. Z. Simon , S. Anderson , Evidence of degraded representation of speech in noise, in the aging midbrain and cortex , Journal of Neurophysiology 116 ( 5 ) ( 2016 ) 2346 – 2355 . OpenUrl CrossRef PubMed [58]. ↵ I. M. D. Karunathilake , J. L. Dunlap , J. Perera , A. Presacco , L. Decruy , S. Anderson , S. E. Kuchinsky , J. Z. Simon , Effects of aging on cortical representations of continuous speech , Journal of Neurophysiology 129 ( 6 ) ( 2023 ) 1359 – 1377 . OpenUrl View the discussion thread. Back to top Previous Next Posted August 01, 2024. Download PDF Email Thank you for your interest in spreading the word about bioRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following Assessing the Impact of Selective Attention on the Cortical Tracking of the Speech Envelope in the Delta and Theta Frequency Bands and How Musical Training Does (Not) Affect It Message Subject (Your Name) has forwarded a page to you from bioRxiv Message Body (Your Name) thought you would like to see this page from the bioRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share Assessing the Impact of Selective Attention on the Cortical Tracking of the Speech Envelope in the Delta and Theta Frequency Bands and How Musical Training Does (Not) Affect It Alina Schüller , Annika Mücke , Jasmin Riegel , Tobias Reichenbach bioRxiv 2024.08.01.606154; doi: https://doi.org/10.1101/2024.08.01.606154 Share This Article: Copy Citation Tools Assessing the Impact of Selective Attention on the Cortical Tracking of the Speech Envelope in the Delta and Theta Frequency Bands and How Musical Training Does (Not) Affect It Alina Schüller , Annika Mücke , Jasmin Riegel , Tobias Reichenbach bioRxiv 2024.08.01.606154; doi: https://doi.org/10.1101/2024.08.01.606154 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Neuroscience Subject Areas All Articles Animal Behavior and Cognition (7644) Biochemistry (17728) Bioengineering (13917) Bioinformatics (42038) Biophysics (21489) Cancer Biology (18637) Cell Biology (25553) Clinical Trials (138) Developmental Biology (13401) Ecology (19941) Epidemiology (2067) Evolutionary Biology (24367) Genetics (15622) Genomics (22547) Immunology (17764) Microbiology (40475) Molecular Biology (17208) Neuroscience (88749) Paleontology (667) Pathology (2842) Pharmacology and Toxicology (4834) Physiology (7659) Plant Biology (15175) Scientific Communication and Education (2047) Synthetic Biology (4304) Systems Biology (9835) Zoology (2272)