Beta Spectral Power during Passive Listening in Preschool Children with Specific Language Impairment Open Access

Children with specific language impairment (SLI) show delays in language comprehension and production that cannot be ascribed to hearing loss, intellectual disabilities, or neurological deficits [1]. Children with SLI have difficulties with phonological awareness [2], discrimination [3], word learning and vocabulary [4‒6], attention, and emotional competence [7, 8]. SLI is usually accompanied by behavioral, gross, and fine motor problems, as well as reading and writing [9]. Despite the absence of neurological deficits, neuroimaging studies have reported that children with SLI exhibit delays, immaturity, or difficulties in brain function [10] such as atypical lateralization, with a lack of left lateralization [11]; atypical auditory processing [3, 12, 13] with a deficit in rapid auditory processing and selective attention [14], when compared to typical development (TD) peers.

The literature on electroencephalography (EEG) has extensively studied the role of beta rhythm in children’s language and auditory processing because of its multiregional and asymmetric activation during childhood [15‒17]. Beta waves, with a range of 13–30 Hz, are mostly located in the frontal regions, during TD [18], and in central areas, named frontal beta and Rolandic beta [19]. Furthermore, beta can be divided into two ranges: low beta (13–20 Hz), involved in focus and attention, and high beta (21–30 Hz), related to anxiety and hyperactivity disorder [20, 21]. Besides that, beta activation is linked with motor skills [22, 23].

During brain growth and maturation of EEG activities, spectral power (SP) fluctuates [24], which is reflected in the SP decrease in delta and theta, as well as with the increase in alpha and beta waves, until they begin to resemble an adult brain [16, 25]. Thus, during childhood, brain waves tend to be transitioned and mixed [26‒28] (e.g., in children aged 3–5 years old, alpha and theta rhythms dominate over posterior regions [29]). The amplitude of theta frequencies decreases with age and increases in beta wave [28]. Low beta with a range of 13–23 Hz is associated with language development in children and adults [30, 31].

The most acknowledged regions involved in language processing are frontal regions, involved in language, cognition, and behavior [32, 33], and temporal regions involved in language, memory, and hearing [34]. Hence, together these two regions are in a junction with intelligence [35]. Parietal regions are involved in somatosensory integration [36]. Despite that, specific parts of these regions are involved in more explicit language functions, i.e., inferior frontal gyrus, more specifically, the Broca area in language production [37], prefrontal regions in focus and memory [35], superior temporal gyrus, i.e., Wernicke area involved in language comprehension [35], inferior parietal lobe connected with some parts of language processing, as well [38].

Several developmental studies showed that language processing and neural topography and functions in children’s brain maturation are debatable. Skeide et al. [39] documented that syntax and semantics during language processing are linked with left superior temporal gyrus in children aged 4 years old; furthermore, left fronto-temporal activation is linked in children aged 5 years, which is adult-like brain behavior [40, 41]. Pre-schoolers’ auditory listening activates the ventral pathway, which includes the superior temporal gyrus and Broca area [37], whereas older children activate the dorsal pathway [39]. Other studies [31, 42] reported, in children aged from four to six, bilaterally inferior, medial, and right superior frontal beta desynchronization, which is close to F3, F4, F7, and F8 electrode placement [43]. Perone et al. [28] found a positive correlation between age and attention task in posterior beta activation. Vanvooren et al. [17] suggested that auditory processing is universal, but throughout age, it becomes more specific and asymmetric, especially for the beta rhythm. According to this author, diffuse desynchronization of low beta activity has been found during language processing in 5 years old, when compared to adults. In addition, the same results have been found bilaterally in the inferior temporal regions, which refer to T3 and T4 positioned electrodes. In the same study, beta synchronization was also bilateral, in frontal and temporal regions, but more right lateralized. Kropotov [19] documented negative correlations of low beta power with the F3 and P3 electrodes and middle beta with the Fp2 electrode during the task with auditory duration detection. High beta was in negative correlation for the dichotomy listening task at the F4 electrode. Studies including low beta frequencies, with a range of 13–25 Hz [22], 13–20 Hz [22], 15–25 Hz [44], and listening tasks revealed a decrease in parieto-occipital regions; increase in high beta in fronto central regions [45]; as well as decrease in left frontal and temporal regions during auditory processing of unknown stimuli [22].

Despite these findings, greater changes in beta rhythm are more linked with older children, whereas the presence of theta and alpha changes is more prominent in preschool children [46, 47]. For children’s beta rhythm, it is still unclear what to expect in various conditions [28].

Given the fact that beta rhythm is involved in language and auditory processing and has multiregional and asymmetrical activation during childhood, our study aimed to examine the low beta SP during auditory listening of words and nonwords in preschool-aged children with SLI and their TD peers. Low beta SP was analyzed from 19 electrodes and regions of interest (ROIs) through the topography and desynchronization of the beta rhythm in different listening tasks (words and nonwords), including the resting state (RS) as a baseline condition.

Our study included 100 participants, divided into two groups: children with SLI as an experimental group (E) (N = 50) and children with typical speech and language development (TD) as a control group (C) (N = 50) (Table 1). The same participants were used in our previous study [48], with alpha rhythm detection. Both groups were divided into two subgroups: (1) children aged from 4.0 to 4.11 years (E = 25, C = 25); (2) children aged from 5.0 to 5.11 years (E = 25, C = 25). All children were recruited from a local community (kindergarten and personal contact) on the territory of the City of Belgrade. After qualified speech and language pathologists conducted the speech-language, intelligence, and handedness assessment, children were divided into two groups: children diagnosed with SLI and TD children. This procedure was carried out until reaching a sufficient number of participants. They did not receive any speech or language therapy services before.

The inclusion criteria for the final sample were as follows: (a) all participants were Serbian language native speakers with (b) normal or corrected-to-normal vision, (c) normal hearing, without any neurological impairments, (d) no use of any medications that may affect EEG processing, (e) normal non-verbal intelligence (performance IQ of 85 or higher, as well as the existence of a large scatter between performance and verbal IQ coefficients for children with SLI), and (f) the SLI group differs by 1.25 standard deviations below average from the mean of the TD group in each of the language assessments measures [49]. All participants were boys. The reason for this is the significantly higher presence of speech-language pathology in boys than in girls [50, 51]. All participants were right handed according to the Edinburgh Inventory [52]. This study was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Committee of the Institute for Experimental Phonetics and Speech Pathology in Belgrade, Serbia (No. 1/21-3; date: January 18, 2021). Children’s parents/guardians provided written informed consent to participate in the study.

The speech and language assessment consisted of the following tests: (1) dictionary test for children aged 3–7 years [53], (2) the Peabody Picture Vocabulary Test [54], and (3) the Token Test [53, 55]. Intelligence assessment consisted of two tests: the Wechsler Intelligence Scale (WISC) [56] and the Brunet-Lezine Scale [57]. The mentioned tests were used in both groups before EEG recording to evaluate children’s cognitive abilities and to determine whether there was lower achievement on the verbal scale (high scatter between verbal and performance IQ), which is an important diagnostic criterion for SLI. Handedness measures were obtained with the Edinburgh Handedness Inventory (EHI) [58]. In Table 2, descriptive statistics for all speech-language assessments, for two age groups in children with TD and SLI, are presented. In children 4 years of age, there is statistically significant difference in achievement on all speech and language assessments in both groups (TD/SLI): Peabody (F = 0.605, t₍₄₈₎ = 8.077, p < 0.001); Token (F = 12.304, t_(33.384) = 13.119, p < 0.001); dictionary (F = 14.762, t_(30.446) = 8.853, p < 0.001). Likewise, in children 5 years of age in both groups (TD/SLI), there is statistically significant difference in achievement on all speech and language assessments: Peabody (F = 1.491, t₍₄₈₎ = 42.016, p < 0.001); Token (F = 30.865, t_(27.404) = 9.266, p < 0.001); dictionary (F = 14.104, t_(30.631) = 6.776, p < 0.001).

We employed the same procedure for EEG recordings and data pre-processing as reported in our previous study, with the detection of alpha SP during auditory listening [48]. EEG was recorded using a Nihon Kohden Corporation EEG 1200K Neurofax apparatus with an Electro-Cap and silver/silver chloride (Ag/AgCl) ring electrodes filled with electroconductive gel. Nineteen EEG channels were recorded (Fp1, Fp2, F3, F4, C3, C4, P3, P4, F7, F8, T3, T4, T5, T6, O1, O2, Fz, Cz, and Pz). The electrodes were positioned according to the 10/20 International System for Electrode Placement. The reference was set as (C3+C4)/2, which is the NK-9100K EEG system, with the ground electrode placed on the forehead. The impedance was maintained below 5 kΩ, with no more than 1 kΩ between electrodes. The lower filter was set at 0.53 Hz and the upper filter at 35 Hz. Electro-oculograms (EOGs) were recorded to detect eye blinks and horizontal or vertical eye movement. Heart-rate sensors were used for online artifact removal. The AC filter was set on. The sampling rate was 200 Hz.

During the experimental procedure, the participants were seated in a comfortable position in a sound, electrically shielded room, more precisely, in a square-shaped cube made of white nontransparent curtains to eliminate visual stimuli that may influence the experimental tasks. The experimental procedures, approximately, were performed at noon (12 am ± 2 h). An experienced researcher set up an EEG cap on the child’s head. A parent/legal guardian was present during electrode placement. When the technical requirements for recording were fulfilled, verbal instructions were provided to the children at the beginning of each EEG recording.

The first part of the experiment was a 2-min RS EEG recording (with the possibility of shorter periods of recording depending on children’s attention). The participant’s task was to keep their eyes open for one or 2 min to minimize their movements to minimize artifacts in the EEG trace. The 1-min RS was used as a baseline for comparisons with the auditory processing tasks.

The second part of the experimental procedure was the recording of the EEG signal during the listening task with two different stimuli: words and nonwords. Word listening (WL) involves listening to 10 different words, while nonword listening (NWL) involves listening to 10 nonwords. In total, there were 20 stimuli in the listening task for each participant, with random presentation of words and nonwords. The duration of the inter-stimulus interval was 1.5 s. All stimuli, in randomized order, were encoded and annotated using stimulus presentation software.

The duration of the experimental procedure was around 30 min. Before the experimental procedure, the participants’ parents were informed about the experiment.

Auditory stimuli from an existing stimulus database called the EEG protocol for auditory-verbal processing were used (otherwise used in the laboratory for cognitive neuroscience at Research and Development Institute “Life Activities Advancement Institute,” situated in Belgrade, Serbia). Stimuli in the EEG protocol consisted of words, nonwords, sentence questions, and narrative discourse (short familiar and unfamiliar stories). Words and nonwords were used in the study.

The most frequent feminine nouns in the Serbian language (words with the highest frequency of occurrence in the standard Serbian language) according to the Children’s Frequency Dictionary [59] were used as the final stimuli in WL. All the words were balanced in length (four sounds per two-syllable word and six sounds per three-syllable word) with consonant-vowel structures.

Nonwords are words that follow the phonological rules of Serbian language but do not have semantic meaning in Serbian. When forming a nonword, the frequency of sounds in the Serbian language, as well as the consonant-vocal-consonant-vocal structure, was considered. In addition, the sounds that appeared in the word lists were uniform, according to the frequency of occurrence in the list of nonwords. The terminal list for each task included five two-syllable and five three-syllable words, as well as five two-syllable and five three-syllable nonwords.

Stimuli (words and nonwords) were spoken by a professional male speaker who read the stimuli individually without variation in melody, rhythm, and emotional expression. All stimuli were recorded at a sampling frequency of 44.1 kHz with a 16-bit resolution. Stimuli were recorded in a sound-attenuated room using a Handy Recorder H4N (serial number 00217460, ZOOM Corporation, Japan) placed 20 cm from the speaker’s mouth. Recordings were saved as WAV files at a 44.1 kHz sampling frequency and 16-bit amplitude resolution. Their average duration was 500 ms (range: 485–525 ms). Respective recordings were used to generate the stimuli. The stimuli were presented binaurally to the participants using earphones with earplugs at a sound pressure level of 50 dB.

Segments with rough artifacts were visually inspected and removed from further analysis. Heart-rate artifacts were removed simultaneously during EEG recording using an implemented electrocardiogram (ECG) filter. The raw files were then converted to “eeg” format to be imported into the EEGLAB software on a MATLAB platform [60] for further analysis. All continuous data were filtered using a FIR band-pass filter with a pass band from 1.6 Hz to 30 Hz. The data were re-referenced to the average values. This means that re-referencing occurs in all channels, on average. Independent component analysis was performed to remove eye blinking and muscle activity artifacts from selected EEG segments.

Afterward, a database of EEG segments for each task and trial is created. For the RS conditions, the data were segmented into 10-s epochs. The number of included epochs in further analysis was five per RS condition per child, which resulted in a total of 500 EEG epochs. For WL or NWL, the data marked in the EEG trace were segmented in 1-s epochs. Separately, WL and NWL had 10 trials per participant, resulting in 1000 EEG epochs per task. All data were saved in “set” file format. For statistical analysis, we used the participants’ averaged values of five trials of RS measurement and averaged values of 10 trials for each WL and NWL measurement.

The absolute beta SP was calculated for 19 electrodes. We used MATLAB script and EEGLAB software for spectral analysis and graphical presentation. The power spectral density estimate was calculated using Welch’s method in MATLAB (pwelch’s function). The absolute SPs for the RS, WL, and NWL tasks were determined for a beta rhythm ranging from 13 to 23 Hz (according to the Sharma et al. [31]). The results are shown as the mean average absolute SP in the beta range (µV2/Hz).

Descriptive statistics were calculated for speech/language assessment tests, intelligence assessment tests, and different age groups. Continuous variables are presented as means and 95% confidence intervals. Normality of the distribution was tested using graphical and mathematical models. The level of significance was set at p < 0.05. Statistical analysis was performed using SPSS 21 (IBM, Chicago, IL, USA) 2012 package. To explore the influence of different factors on beta processing, we used a generalized linear mixed model (GLMM) with the following model parameters:

• Type of measurement-repeated measures (task WL/NWL)

• Fixed effects (fixed factors and covariates): group, age, age group, task, IQm, Peabody, Token, dictionary, group × age group, group × task, age group × task, group × age group × task

• Target distribution and relationship (link function), distribution-gamma, link function identity

• Build option for the model: maximum iterations = 400, degrees of freedom variable across tests, tests of fixed effects and coefficients, robust estimation

This model was applied in all 19 electrodes, as well as ROI. ROI has been defined based on the functional and anatomical criteria for language processing: sum of electrodes’ beta SP, according to their locations, was averaged such as the left frontal region (LFR-Fp1, F3, F7), right frontal region (RFR-Fp2, F4, F8), left temporal region (LTR-T3, T5), right temporal region (RTR-T4, T6), left centroparietal region (LCPR-C3, P3), right centroparietal region (RCPR-C4, P4), and central region (CR-Fz, Cz, Pz). For the RS, we used the averaged beta SP values to compute all 19 electrodes.

For the statistically significant interaction, group × age group, we defined new variable groupage group and applied one-way ANOVA to examine if there is statistically significant difference between variable subgroups. The subgroups are defined as (1) TD group 4 years of age, (2) TD group 5 years of age, (3) SLI group 4 years of age, and (4) SLI group 5 years of age.

Differences between the average values of RS beta SP between age groups, as well as between the SLI and TD groups, were analyzed by univariate analysis, and Levene’s test of equality of error variances showed that univariate analysis can be performed (F(3, 96) = 1.10, p = 0.353). Univariate analysis of the RS-averaged beta SP showed that the group (SLI/TD) and age group (4 years of age/5 years of age) did not have a statistically significant impact as main factors or as an interaction term group × age group (see online suppl. Table 1; for all online suppl. material, see https://doi.org/10.1159/000539135).

The GLMM repeated measures revealed that a significant effect on beta rhythm processing during auditory listening had a group in the temporally positioned electrodes: T3 (p = 0.016), T6 (p = 0.002), additional in the LTR (p = 0.007) and RTR (p = 0.027). Furthermore, age groups have a statistically significant effect on beta processing in F4 (p = 0.044) and T6 (p = 0.001) electrodes, as well as in the RTR (p = 0.013) and RCPR (p = 0.030). The GLMM revealed a statistically significant differences in the interaction term groupage group in the Fp1 (p = 0.025), F8 (p = 0.010), T5 (p = 0.027), and Fz (p = 0.037) electrodes. No statistical significance was observed for other interactions. Task variable did not have statistically significant impact on model, neither as a fixed factor or as an interaction term. One-way ANOVA was performed to determine beta SP differences in groups (SLI/TD), different age groups (4 years of age/5 years of age), and group × age group interaction in those electrodes and ROI, which were statistically significant in the GLMM, presented in Table 3.

For the electrodes and ROI where group has statistically significant impact on model, and where interaction term do not have statistically significant impact on model, one-way ANOVA captured that beta SP mean values were higher in all cases in SLI group (T3: M = 0.46, SD = 0.45; T6: M = 1.08, SD = 0.61; LTR: M = 0.66, SD = 0.35; RTR: M = 0.80, SD = 0.42), in comparison to TD group (T3: M = 0.35, SD = 0.28; T6: M = 0.92, SD = 0.47; LTR: M = 0.59, SD = 0.27; RTR: M = 0.70, SD = 0.29) (see Fig. 1, with graphs). For the electrodes and ROI where the age group had a statistically significant impact on the model, and where interaction term do not have statistically significant impact on model, one-way ANOVA captured that beta SP mean values were higher in the older group than in the younger group in all cases (younger group: F4: M = 0.79, SD = 0.46; T6: M = 0.94, SD = 0.59; RTR: M = 0.72, SD = 0.38; RCPR: M = 0.52, SD = 0.21; older group: F4: M = 0.97, SD = 0.57; T6: M = 1.06, SD = 0.50; RTR: M = 0.78, SD = 0.34; RCPR: M = 0.59, SD = 0.27) (see Fig. 2, with graphs). Furthermore, we conducted one-way ANOVA with Games-Howell post hoc tests to establish pairwise comparison across multiple subgroups (group × age group interaction term) of the new variable groupage group (see online suppl. Table 2).

The one-way ANOVA with Games-Howell post hoc test indicated that there is no statistically significant difference between 4 subgroups of groupage group variable for Fp1 electrode. For the T5 electrode, statistically significant differences were observed between subgroups 1 and 2, i.e., TD group 4 years of age and TD group 5 years of age (p = 0.019), and subgroups 3 and 4, i.e., SLI group 4 years of age and SLI group 5 years of age (p = 0.015). For the Fz electrode, statistically significant differences were observed between subgroups 1 and 3, i.e., TD group 4 years of age and SLI group 4 years of age (p = 0.020), and subgroups 3 and 4, i.e., SLI group 4 years of age and SLI group 5 years of age (p = 0.001).

For statistically significant electrodes (T5 and Fz), in both groups (TD/SLI), the beta SP was higher in older children. Furthermore, the difference between the TD and SLI groups occurred in the central frontal electrode (Fz) in the younger group, with the higher beta SP in TD children (see Fig. 3). Additionally, significant differences for the interaction term groupoage group were obtained between subgroups 2 and 3, i.e., TD group 5 years of age and SLI group 4 years of age, for F8 (p = 0.031), T5 (p = 0.001), and Fz electrode (p = 0.001), with higher beta SP in older TD group. Evident differences between the older TD group and the younger SLI group confirms the suitability of auditory processing in TD children and maturation of the auditory system.

The present study examined prospective differences in EEG beta rhythm SP during auditory listening to words and nonwords between children with SLI and their TD peers. Consequently, these tasks may be powerful tools for diagnosis, knowing that nonword production and perception is a strong diagnostic instrument for SLI [61, 62]. Our results revealed that the younger group of children differed from the older group in beta auditory processing. Additionally, children with language impairments seem to have a higher SP in the beta range than typically developed children.

Our findings pointed to strong differences in bilateral temporal activation of the beta rhythm between TD and SLI children, which is in relation to similar research [63, 64], supporting the presence of the Wernicke area, positioned in the left posterior superior temporal gyrus [63, 64], which is involved in complex language skills, such as semantic [65], and most importantly the manifestation of auditory language processing in temporal regions [66‒68]. The absence or decrease of beta rhythm SP in anticipated brain regions in the TD population, revealed in our study, and in other studies, has been explained as an unexpected presence of unknown stimuli [45, 69, 70] or preparation for motor activation [23]. Van Elketal [71] documented a decrease in prefrontal beta during passive listening because of the absence of required motor tasks. An increase in beta-rhythm SP and a decrease in alpha-rhythm SP are linked to greater attention engagement [72, 73], which was not the case in our study within the TD group. Namely, the obtained results may indicate the absence of noticing the requested auditory tasks, especially in the TD group.

Our study documented right-lateralized activation in the frontal, temporal, and centroparietal regions in junctions with age groups. Our results revealed that the younger group of children differed from the older group in beta auditory processing. Accordingly, studies with adults and fMRI have shown that right temporoparietal activations occur during sentence comprehension [74] and passive speech listening [75, 76]. Bilateral frontal and temporoparietal activation has also been documented during tasks involving passive speech listening [77, 78]. Higher beta SP during auditory listening in the older group, also revealed in our study, confirms the fact that beta SP increases during brain maturation [26, 28, 79]. Additionally, beta activation linked with age in the SLI group as well as in the group with TD children in temporal regions indicates the presence of phonological decoding [48], and on the other hand, concerning Wernicke-Broca’s articulation loop with a secondary association to parietal regions involved in speech sound listening [80]. Furthermore, centroparietal activation suggests the existence of mental lexicon [81], and its responsibility in semantic and phonological decoding, as well [82, 83].

Consistently, the increased beta activity observed mostly in right-lateralized regions within TD and SLI different age groups and related to auditory listening in SLI children exhibits a lack of brain activation [84] (especially in the left-lateralized regions), or the presence of compensatory mechanisms [11]. In TD children, even though speech perception is left lateralized [85], there is evidence of homology between right- and left-lateralized brain activation during speech perception in children up to 6 years of age [86]. In children with SLI, there is evidence of auditory or language processing supported by the right hemisphere [87].

Beta activity in specific brain regions triggered by two auditory measurements revealed differences in different groups and age groups, indicating that brain maturation is still in progress [88]. Vanvooren [17] reported reduced neural synchronization for the beta band in preschool children, tentatively in 5-year-old children during auditory processing, which could be explained, as mentioned above, by brain immaturity [88] or immaturity for phoneme awareness, which is typical for that age [2, 65]. Moreover, essential differences in auditory maturation seem to become apparent from the age of 5 years [89].

As mentioned before, ROI is based on the functional and anatomical criteria for language processing. For example, language processing involves multifarious brain regions, such as frontal [32, 33], temporal [34], and parietal region [36]. Despite that, according to Karunanayaka et al. [80] the specific parts of these three regions are involved in the articulation loop from speech decoding to production. Therefore, we were interested in focusing not only on ROI but also on a specific area of ROI by analyzing the results of each electrode location independently.

Our results indicate that certain observed ROIs demonstrate a statistically significant impact on beta processing. However, when each electrode is analyzed separately, this significance is not observed in every electrode. More precisely, in the group interactions, LTR and separately, T3 and T5 have statistical significance, which confirms the fact of strong involvement of this ROI in auditory processing [35]. Opposite to the left-lateralized findings, RTR and T6, but not T4, have significant impact on beta processing, which points to the T6 as a strongly involved electrode location in language processing. Namely, T6, i.e., localized posterior temporal, could be in junction with semantic processing [77]. Statistically significant impact on beta processing obtained for RCPR, but not separately in C4 or P4 electrode, indicates that this whole brain activity is in junction with memory integration of different stimuli to long-term mental lexicon [90]. From this point, electrode information may be beneficial for specific language processing tasks, and the ROI information, for simpler language processing tasks.

Our results have the opposite direction compared to other findings. More specifically, other studies reported that the beta rhythm has a high presence in frontal regions, that is, prominent beta rhythm [18, 91]. Strong engagement of the frontal regions in different auditory [67] or attention tasks [20] has been documented. Likewise, studies have reported strong left-lateralized speech perception in children and adults [78, 92], which was not the case in our study. However, keeping in mind that the younger group of TD children showed higher values of beta SP compared with the younger group of SLI children in the Fz electrode is powerful evidence of beta synchronization during different tasks, especially in the frontal regions [19].

Our results revealed that the younger group of children differed from the older group in beta auditory processing. In addition, children with language impairments seem to have a higher SP in the beta range than typically developed children.

Other studies suggest that SLI could be a significant cause of “low-level impairment” in auditory perception [93‒95] with the possibility of atypical asymmetry in children with SLI [96] and bilateral and diffuse speech perception compared with TD children [93]. Interestingly, a review study reported that investigations of auditory processing in children with SLI is scarce and have contradictory outcomes [97]. These findings may be related to the fact that beta auditory processing in preschool children with SLI is present in multivariate regions such as the frontal, temporal, and parietal regions.

Our general judgment follows that stronger beta activation that occurs in children with SLI in the frontal, temporal, and parietal regions might indicate the presence of unknown stimuli with a semantically unknown concept; lack of activation of activate awareness, mental lexicon, and articulation loop to reach a semantic background of the word or nonword. In addition, higher beta in SLI children indicates a potential need for motor imitation or verbal repetition of stimuli for improved understanding of the stimuli. The absence of beta rhythm activation in anticipated brain regions in TD children may suggest brain immaturity, planning for motor activation, and an unexciting presence of the stimulus, as well as the nature of the task (task with passive listening without requiring production). Unusual results obtained from ROI and separately, from electrode locations, could suggest that during brain development especially for language improvement, it is still uncertain to predict the pattern of speech and language processing. In addition, these findings could be beneficial for further research and to better understand the neural mechanisms underlying language processes.

In our research, we used an Electro-Cap with 19 electrodes, while recent studies have attempted to use an EEG system with 64 electrodes integrated with event-related measures (ERP) [29, 51, 98, 99]. In addition, EEG as a “window of the mind” is less considerable than modern imaging techniques, such as PET and fMRI [100]. Our study has certain limitations. First, the small sample size may have influenced our overall evaluation and conclusion. Furthermore, we used an EEG cap with 19 electrodes. We based our analysis on the fact that EEG can serve as an objective method of predicting the achievement of rehabilitation [101], the application of neurorehabilitation (neurofeedback) [102], especially for beta rhythm [103], and in the brain-computer interface for people with various neurological disorders [104, 105]. In this regard, we believe that the findings obtained in this study, even with a 19-electrode EEG cap, could be reliable and provide guidance for further research, especially concerning typically and atypically developed children. Furthermore, the small sample size may have affected our general judgment and conclusion.

This finding suggests that beta activation caused by WL and NWL in TD and SLI children appears to be a very complex activity in different brain regions, indicating the presence of early changes in functional brain connectivity during electrogenesis. Considering that all cognitive processes are expressed in our daily lives, analysis of brain waves and their changes may provide better insight into the dynamic mechanisms, which are the basics of cognitive processes. Generally, it is possible to observe the effects of cognitive involvement during auditory perception in children, such as maturation, attention engagement, memory capacity, and emotional maturity. Accordingly, future research should focus on spatial brain activation and interregional interactions during speech perception, as well as production. Knowing which neurophysiological markers are involved in specific cognitive functions is an essential first step in therapy and rehabilitation of children with SLI.

This project was conducted in cooperation with the Faculty of Medical Sciences of the University of Kragujevac.

This study protocol was reviewed and approved by the Ethics Committee of the Institute for Experimental Phonetics and Speech Pathology in Belgrade, Serbia (No. 1/21-3; date: January 18, 2021). Written informed consent was obtained from parent/legal guardians to participate in the study.

The authors have no conflicts of interest to declare.

This work was partially supported by the Ministry of Education, Science, and Technological Development of the Republic of Serbia under the project “influence of psychophysiological, sociological, and cultural factors on speech and language in the child population.”

S.F., N.S., LJ.J., and M.M. contributed to the formal analysis; S.F., N.S., and R.B took part in data collection and software usage; S.F. and M.S. analyzed the data; S.F. designed the original draft; S.M. and A.G. cooperated in visualization of the paper; and M.S. contributed to final approval and supervision. All authors have read and approved the submission of the manuscript.

Raw data were produced at the Laboratory of Cognitive Neuroscience Department, Research and Development Institute “Life Activities Advancement Institute,” Belgrade, Serbia, and at the Department of Speech, Language, and Hearing Sciences, Institute for Experimental Phonetics and Speech Pathology, Belgrade, Serbia. The data that support the findings of this study are not publicly available due to privacy reasons of research participants but are available from Saška Fatić as the corresponding author on request.