Temporal and spectral EEG dynamics can be indicators of stealth placement

Stealth placement marketing, where consumers are unaware that they are being marketed to, attempts to reduce the audiences’ resistance to traditional persuasive advertising. It is a form of advertising that involves targeted exposure of brands or products incorporated in other works, usually with or without explicit reference to the brands or products. Brand placement can be presented in different visual and auditory forms in video programs. The present study proposed that different ‘representations’ (i.e., representable or non-representable) and ‘sounds’ (i.e., speech or musical sound) of brand placement can affect the viewers’ perception of the brand. Event-related potential results indicated significant differences in P1, N1, P2, N270, and P3. Further, event-related spectral perturbation results indicated significant differences in theta, alpha, beta, and gamma (30–100 Hz), in the right parietal, right occipital area, and limbic lobe. ‘Non-representable’ or ‘speech sound’ brand placement induced significant temporal and spectral EEG dynamics in viewers.


Discussion
The experimental hypotheses of the present study were centred on the effects of representable or non-representable product placement, and of speech sound or musical sound brand placements on brain EEG activity. The attention and awareness of non-representable and speech sound brand placements of the viewers were analysed based on the behavioural, ERP, and ERSP data.
Analyses of the behavioural data indicated that the representation and sound of brand placement had a main effect on the viewers' awareness (see Section 3.1). The viewers' awareness for NREBPL was higher than for REBPL. Thus, blatant placement led to better viewer recognition than subtle placement 54 . The viewers' awareness for SSBPL was higher than that for MSBPL. Compared with the recognition of program content using visual information, the use of auditory information under the program was easier 55 , implying that using auditory information alone to determine the video content will allow viewers to avoid the more complicated steps of visual processing 23 .
The present study demonstrated that the representation and sound of brand placement had a main effect on the viewers' preference. The viewers' preference was higher for REBPL than for NREBPL, and higher for MSBPL than for SSBPL. Thus, performance-integrated and creative placements are more likely to achieve higher viewer preference and more-positive attitudes [55][56][57] . Repeated, blatant brand placements will reduce the original impact of the program and lead to negative attitudes in viewers 57 . Background music is a clue to the persuasiveness of advertisements 58 . Using contemporary songs or music can also enhance the viewers' preference for advertisements 59 .
Moreover, the present findings that the representation and sound of brand placement did not have an interaction or main effect on the viewers' purchasing desire. Thus, although brand placement in films may enhance brand awareness, it does not seem to have a substantial impact on the viewers' purchasing desire 60,61 .
The present study demonstrated that more representable brand placement had greater effects on viewers' temporal and spectral EEG dynamics. Figure 3A-E indicate significant differences between REBPL and NREBPL for P1, N1, P2, N270, P3, and LPC (see Section 3.2). Our results showed that the amplitude of P1 was greater for REBPL than for NREBPL. P1 is evoked at the early stages of visuospatial information processing 10,49,62,63 , while REBPL involves the appearance of brand logo information in real space. P1 also reflects the attentional demands of task processing, such as using brightness or colour to judge the stimulus 10,16 . In contrast, NREBPL uses graphic processing of brand information, such as accentuating and highlighting the brand logo. The N1 amplitude of NREBPL was greater than that of REBPL. N1 is considered the signal for the allocation of attentional resources 14 and is related to the identification of objects or faces 9 . P2 is related to emotional evaluation 51 ; emotionally stimulating words and images evoke larger P2 amplitudes 51,52 . When perceiving emotionally stimulating music or images, the P2 amplitude of positive stimulations is larger than that of negative stimulations 13 . Rare targets can also evoke a larger P2 64 . In the present study, the P2 amplitude evoked by REBPL was larger than that evoked by NREBPL. This might have been because, compared with NREBPL, where the brand is placed in a fixed position in the frame for the entire duration, REBPL involves the natural and random appearance of the brand, which evokes a greater emotional reaction. Moreover, the present findings demonstrated that the N270 amplitude in the frontal area for NREBPL was larger than that for REBPL. The presence of conflict or spatial differences evoke the N270 component 65,66 . N270 has also been used to detect the congruence of brand image and extension products 67 , particularly for frontal lobe activity 53 . In the NREBPL, the viewer may be unable to match the graphic brand logo information occurring throughout the video with the script content of the actual program, leading to a greater sense of conflict. The present study demonstrated that the P3 amplitude in the left frontal and parietal areas for REBPL was larger than that for NREBPL. Visual search tasks will evoke P3 in the frontal lobe 68 , which is an indicator of attentional control mechanisms. Clear selection of attentional targets will evoke a larger P3 48 , while novel and unpredictable stimuli also evokes P3 in the frontal and temporal lobes 69 . Moreover, P3 is an indicator of emotions 16,48 . Further, LPC is a signal of high-level categorization and decisions 70 , with incongruity evoking larger LPC amplitude 71,72 . Our study demonstrated that the LPC amplitudes of the left frontal and temporal areas for REBPL were larger than those for NREBPL. The susceptibility of viewers to the boredom of repeated exposure to visual stimuli will lead to delayed LPC in the frontal lobe 73 .
The present study demonstrated that brand placement accompanied by music was more likely to influence brain activity in the viewers. Figure 3G-K show significant differences between SSBPL and MSBPL in the P1, N1, P2, N270, and P3 amplitudes (see Section 3.2). The P1 amplitude of MSBPL was larger than that of SSBPL. P1 is an indicator for screening and inhibition of interfering and irrelevant auditory stimuli 74 . A reduction in the P1 amplitude implies a lower demand for auditory processing. SSBPL involves the reading of brand information, such that viewers can focus on processing the auditory information alone. In contrast, MSBPL involves the processing of both music and brand information, which increases the demands of inhibiting auditory interference. . Independent component analysis was performed on the electroencephalography signals of all the participants to obtain the statistically significant scalp map and equivalent dipole for clustering, which resulted in 12 brain areas. The figures show the scalp map and equivalent dipole of the 12 brain areas. *Indicates that the ERSP analysis revealed significant differences in these brain regions for the two factors (different representations and sounds of brand placement; p < 0.05), which include the right parietal area (BA 2), right occipital area (BA 17), and limbic lobe (BA 30). Parametric testing with FDR correction was performed for significance testing (p < 0.05) to further compare the differences for the varying representations and sounds at different frequencies ( (A-C) Brain regions and ERSP(s) results are shown, with significant differences between REBPL and NREBPL. The two factors indicate significant differences in the spectral powers of theta, alpha, beta, and gamma for Brodmann areas BA 2, BA 17, and BA 30 (p < 0.05). (A) In BA 2, REBPL shows a higher spectral power for 7-13 Hz (alpha) than NREBPL (p < 0.05), and NREBPL shows a higher spectral power for 80-100 Hz (high gamma) than REBPL (p < 0.05). (B) In BA 17, NREBPL shows a higher spectral power for 3-6 Hz (theta) than REBPL (p < 0.05); REBPL shows a higher spectral power for 9-20 Hz (alpha, beta) than NREBPL (p < 0.05); REBPL shows a higher spectral power for 20-30 Hz (beta) than NREBPL (p < 0.05); REBPL shows a higher spectral power for 30-60 Hz (low gamma) than NREBPL (p < 0.05); and NREBPL shows a higher spectral power for 60-100 Hz (high gamma) than REBPL (p < 0.05). (C) In BA 30, NREBPL shows a higher spectral power for 4-6 Hz (theta) than REBPL (p < 0.05); REBPL shows a higher spectral power for 9-20 Hz (alpha, beta) than NREBPL (p < 0.05); REBPL shows a higher spectral power for 30-40 Hz (low gamma) than NREBPL (p < 0.05); and NREBPL shows a higher spectral power for 60-100 Hz (high gamma) than REBPL (p < 0.05). (D-F) Brain regions and ERSP(s) results are shown, with significant differences between SSBPL and MSBPL. The two factors indicate significant differences in the spectral powers of alpha, beta, and gamma for Brodmann areas BA 2, BA 17, and BA 30 (p < 0.05). (D) In BA 2, MSBPL shows a higher spectral power for 15-15 Hz (beta) than SSBPL (p < 0.05), and MSBPL shows a higher spectral power for 30-40 Hz (low gamma) than SSBPL (p < 0.05). (E) In BA 17, MSBPL shows a higher spectral power for 10-20 Hz (alpha, beta) than SSBPL (p < 0.05), and MSBPL shows a higher spectral power for 60-100 Hz (high gamma) than SSBPL (p < 0.05). (F) In BA 30, MSBPL shows a higher spectral power for 7-30 Hz (alpha, beta) than SSBPL (p < 0.05), and MSBPL shows a higher spectral power for 30-100 Hz (low gamma, high gamma) than SSBPL (p < 0.05). The present study demonstrated that the N1 amplitude of SSBPL was larger than that of MSBPL. N1 is an indicator of pitch transitions in the early auditory cortical response 75 . The amplitude of MSBPL is larger under auditory stimulation alone than under joint visual and auditory stimulation 12 . The P2 amplitude of MSBPL was larger than that of SSBPL. Bilateral P2 enhancement reflects the acoustic discrimination of pure tones 76 . Conflicts between auditory speech and visual stimuli will evoke N270 77 . Our results indicated that the N270 amplitude of SSBPL was larger than that of MSBPL. Viewers may have felt a sense of abruptness towards brand information that was irrelevant to the content of the singing program and the program host's presentation. P3 has been used to detect the brain's perception of rhythm 78,79 . Moreover, P3 is evoked when capturing rare targets and refocusing attention 77,80 . In the present study, the P3 amplitude of MSBPL was larger than that of SSBPL. For MSBPL, the viewers may have had to divert their attention away from the rhythm of the songs to capture brand information, which in turn shifted and refocused their attention between the two types of information.
The ERSP results indicated that different representations and sounds of brand placement could lead to differences in temporal and spectral EEG dynamics in BA 2, BA 17, and BA 30 (see Section 3.3). Table 2 outlines the statistical results and provides a detailed description.
A more representable brand placement had a greater impact on the viewers' brain EEG activity in the right parietal (BA 2) area. Table 2 shows that the different representations of brand placement led to differences in the alpha band of BA 2; REBPL was significantly higher than NREBPL. The natural occurrence of REBPL in real space led to significantly higher spectral power than NREBPL, which involves embedding the graphic logo into the frame after film production and is unrelated to the video script. BA 2, in the parietal region, is responsible for the integration of visual, auditory, and somatosensory information 81,82 , while the parietal alpha band reflects the attentional demand of objects 83 . The PPC plays a role in the integration and processing of spatial discrimination 1,84-87 . Patients with right PPC damage will make errors in tasks involving discrimination of spatial depth 84 , which indicates that the right PPC fulfils the function of spatial discrimination. Activation of the alpha band in the frontoparietal and parietooccipital areas has been used to detect spatial discrimination functions 88,89 . Moreover, high gamma (>60 Hz) in BA 2 was significantly higher for NREBPL than for REBPL. Saccades towards the target location will activate the parietal high-gamma band, while alpha will simultaneously be reduced to suppress the memorised stimulus locations during saccades 90 .
The present study demonstrated that brand placement accompanied by music had a greater impact on the viewers' brain EEG activity. Different sounds of brand placement led to differences in beta and low gamma (30-60 Hz) in BA 1 and was significantly higher for MSBPL than for SSBPL. BA 2 is the primary somatosensory cortex. Discrimination of auditory stimuli will also activate the somatosensory and motor cortices 26,91,92 . A beta increase was observed when a tone could be imagined to be beat 93 , while discrimination of speech pronunciation activated the left somatosensory cortex 26,94 . Several studies have demonstrated hemispheric differences in the analysis of speech and musical sounds by the brain 26,94 . The right PPC could be involved in the extraction of spatial, pitch, and temporal information from sound signals. It is also related to the right temporal function of retrieving music 95 . When discriminating non-speech sounds, the gamma activation of the right hemisphere is higher than that of the left hemisphere 96 . The frequency for speech discrimination will activate the temporal processing function of the left hemisphere 35,97,98 , while discriminating the melody and tone of music will activate the pitch processing function of the right hemisphere 99,100 . MSBPL involves music and songs composed of melodies and tones with pitch changes and has a regular rhythm. SSBPL is composed of speech sounds without pitch changes, melody, or tone, and does not have a regular rhythm.
The present study revealed that more representable brand placement had a greater impact on the viewers' brain EEG activity in the right occipital (BA 17) area. Different sounds related to brand placement led to differences  in the theta, alpha, beta, and low gamma (30-60 Hz) of BA 17, with those for REBPL being significantly higher than those for NREBPL. BA 17 is the primary visual cortex. Viewing three-dimensional images can better evoke a response in BA 17 than graphic images 101 . Occipital alpha changes can be considered a signal for the allocation of spatial attention 88,102 . In humans, beta activation in BA 17 indicates the processing of pictorial information 103 . Individuals with a greater ability to process pictorial information will evoke higher beta in BA 17 when memorizing pictorial images than those with poorer ability to process pictorial information. REBPL involves the appearance of brand logos in real space and the random presentation of the brand with relevant video content. When viewers view the video content, they will naturally notice the brand logo at the same time. NREBPL involves embedding the brand logo in the same position on the video frame after graphic processing. It is not related to the video content and is present throughout the video. The viewers are almost forced to view the brand logo when viewing the video. Compared with a stationary target, the spectral power of 35-51 Hz (low gamma) will increase when attending to moving targets 104 . In experiments on commercial advertisements 105 , when participants focus their attention on screens with movement or scene changes, there is a drastic decrease in their occipital alpha 22 . Alpha and theta perturbations in the visual cortex also participate in working memory 106,107 . Furthermore, theta and high gamma (>60 Hz) in BA 17 for NREBPL are significantly higher than those for REBPL. Occipital high gamma is related to early visual processing and object discrimination decision making 108,109 . Images that are easily discriminated elicit higher levels of high gamma than those that are not. When selectively attending to visual stimuli, 35-90 Hz (gamma) of the occipital visual cortex will increase 22 .
The present findings demonstrated that brand placement accompanied by music had a greater impact on the brain EEG activity of viewers. Different sounds of brand placement led to differences in alpha, beta, and high gamma in the BA 17, those evoked during MSBPL were significantly higher than those evoked during SSBPL. During MSBPL, brand logos are placed together with musical sounds, involving melodies and tones, whereas during SSBPL brand logos are placed with rhythmic speech that does not involve melodies and tones. Auditory stimulation suppressed the responses of the visual cortex BA 17 94,110 . Auditory stimulation can also alter the subjective (non-physical) perception of the primary visual cortex 111 . When listening to music, alpha, which is a signal for relaxation and introspection, is enhanced 112 . Stimuli with a regular and intense rhythm will evoke a larger beta, while a reduction in the alpha in the right hemisphere reflects linguistic rhythm processing 113 . The induction of beta and gamma can reflect the perceived intensity of tones 93,114 .
Moreover, the present study demonstrated that a more representable brand placement had a greater impact on the viewers' EEG activity. Different sounds related to brand placement led to differences in the alpha, beta, and low gamma (30-60 Hz) of BA 30, with those evoked during REBPL being significantly higher than those evoked during NREBPL. REBPL involves the presentation of the brand logo in a three-dimensional (3-D) space of actual scenes in the program. Thus, viewers will naturally notice the brand logo. NREBPL, however, involves the graphic processing of the brand logo and embedding it in the same position in the video frame. The logo is unrelated to the video content and is present throughout the video. This may cause viewers to perceive the incongruence of the brand logo, which is unrelated to the video content. BA 30, which is also known as the retrosplenial cortex, participates in contextual analysis 115 . Viewing a 3-D scene leads to greater activation of the BA 30 than viewing close-up scenes, individual objects, and images of human faces 116 . The retrosplenial cortex is the region responsible for analysing 3D geometric spatial information. This region is activated when viewing emotionally stimulating videos or performing episodic memory task, with the left side being related to general memory work, while the right side is related to spatial memory 117 . A decrease in alpha in the retrosplenial cortex not only reflects the demands of visual attention but is also related to coding and retrieval of spatial information 118 ., The beta of the retrosplenial cortex will decrease when performing obvious perception or action 119 . Furthermore, the theta and high gamma (>60 Hz) in BA 30 for NREBPL were significantly than those for REBPL. Theta adjacent to the parietooccipital area represents spatial retrieval and decision making 118 , whereas a decrease in theta adjacent to the retrosplenial cortex of the medial temporal lobe reflects spatial processing and memory formation 120 . During short-term spatial memory tasks incongruity between the actual and expected stimuli elicits the synchronicity of theta and high gamma   121 .Another finding from the current study was that brand placement accompanied by music had a greater impact on the viewers' brain EEG activity. Different sounds of brand placement led to differences in alpha, beta, low gamma, and high gamma in BA 30, with those evoked during MSBPL being significantly higher than those evoked during SSBPL. BA 30 is part of the posterior cingulate cortex and is adjacent to the primary visual cortex 122,123 . Gamma increase adjacent to the left temporal region represents the perception of melodies 124 . Listening and experiencing pleasure from music activates the retrosplenial cortex 125 and enhances the alpha wave in the brain 112 . Furthermore, listening to musical hallucinations activates beta in the retrosplenial cortex. When actual music has been present for a period, musical hallucinations will become a permanent landmark in the auditory scene, as though the retrosplenial cortex has a positioning effect in the 3D space 124 . Moreover, the gamma of the retrosplenial cortex will increase at rest 126 . MSBPL is a musical performance where a melody accompanies the brand logo information, whereas in SSBPL, brand information is not accompanied by a melody and is only a sound performance.
In summary, the present findings, which were enriched by EEG to record brain electrical activity from the scalp corresponded with the ERSP analysis of differences in the mean spontaneous EEG frequency spectrum under various stimuli conditions and with ERPs measures of averages of specifically time-locked brain responses evoked by different experimental stimuli or events.
The present findings verified the hypothesis that a more representable brand placement will have a greater impact on viewers' temporal and spectral EEG dynamics. The ERP analysis results demonstrated the significant differences in the P1, N1, P2, N270, P3, and LPC. REBPL induced greater brain responses to spatial processing, emotional stimulation, and attention to rare objects, whereas NREBPL evoked greater responses to stimulus discrimination, conflict processing, and susceptibility to repeated exposures. The ERSP analysis results indicated significant differences in theta, alpha, beta, and gamma (30-100 Hz), which occurred in the right parietal area, SCIeNTIfIC REpoRTs | (2018) 8:9117 | DOI:10.1038/s41598-018-27294-4 right occipital area, and limbic lobe. REBPL also activated the spatial discrimination function of the right parietal area (BA 2); spatial attentional demands, image processing, and tracking of moving targets in the primary visual cortex (BA 17); and the analysis, coding, and memory functions for contextual space in the retrosplenial area (BA 30). NREBPL activated the prosaccadic function of the right parietal area; object judgment and working memory function of the primary visual cortex; and clear perception and processing of incongruent stimuli in the retrosplenial cortex.
The present study verified the hypothesis that brand placement accompanied by music will have a greater impact on viewers' temporal and spectral EEG dynamics for the discrimination of preferences. The ERP analyses results indicated significant differences in P1, N1, P2, N270, and P3. MSBPL evoked greater brain responses to the suppression of irrelevant auditory information, discrimination of sound rhythms, and capture of rare targets. SSBPL led to greater processing of auditory stimulation alone, and conflicts in auditory and visual information. The ERSP analysis results indicated significant differences in alpha, beta, and gamma (30-100 Hz), which occurred in the right parietal area, right occipital area, and limbic lobe. MSBPL activated the beat imagination and non-speech discrimination functions of the right parietal area (BA 2); perception of tone intensity and regular rhythm in the right occipital region (BA 17); and perception of musical melodies and auditory positioning function of the retrosplenial cortex (BA 30).

Methods
Participants. Twenty-five participants (13 men and 12 women), with a mean age of 23.72 years, were recruited for the present study. The participants had university degrees or higher; were residents of Taipei, Taiwan; had corrected visual acuities of ≥0.8; and had no colour blindness, visual impairments, or a history of neurological or psychiatric disorders. The participants did not have a drug or alcohol addiction and were requested to avoid the intake of stimulants (e.g., coffee, alcohol etc.). The study was conducted in accordance with the Declaration of Helsinki 127 . The study was approved by the human trial institutional review board of Cathay General Hospital. All methods were performed in accordance with the approved guidelines. Informed consent was obtained from all participants prior to the experiments. Bring 'Em Back' . A total of 48 video clips, which contained features of the independent variables, were selected for the study. The duration of each clip, from which 105 experimental stimuli were extracted, was 25 s. REBPL and NREBLP (or SSBPL and MSBPL) had 105 stimuli. A 2 × 2 Latin square design was used to form the conditional framework for the experimental stimuli, based on the criteria that one factor will occur for stimuli conditions of each row or column 128,129 . The advantages of this design included the counterbalance of the same factor, small error, and high efficiency. Hence, it is frequently applied in quantitative marketing experiments [128][129][130] . The stimuli conditions of this experiment were as follows: (1) 'representable versus speech sound' brand placement, (2) 'non-representable versus speech sound' brand placement, (3) 'representable versus musical sound' brand placement, and (4) 'non-representable versus musical sound' brand placement (Fig. 1A). The resolution of the computer screen presenting the stimuli was 480 × 960 pixels. All programs had not been broadcasted in Taiwan. The presenters and the brands placed in the program had never been seen in the Taiwanese market. Hence, the participants were equally unfamiliar with all stimuli.

Stimuli.
Tasks and Procedures. During the experiment, noise, temperature, light, and other interferences were strictly controlled. The participants provided their responses alone in the EEG laboratory. The researcher observed the participant's condition and EEG recording from a screen outside the room. A Neuroscan EEG recording and analysis system (Scan 4.3.3 & Presentation), which included an electrode cap (Quik-Cap) and amplifier (SynAmps2), was used. The 32-channel EEG signals were recorded according to the International 10-20 system of electrode placement; experimental facilities referred to in previous studies 4, [131][132][133][134] were used for brain sources and signals recording.
Participants viewed a screen placed on a 74-cm-high desktop. The centre of the screen was placed within 10-20° of the participant's line of sight, at 60-70 cm. The experimental procedure is shown in Fig. 1B. Instructions were given before the experiment, requesting participants to relax and view the stimulus videos, before responding to the experimental questions using the keyboard. Behavioural questions were concerned with the participant's perception of brand placement. If the participants detected brand logo information placed in the video, they pressed the down key, with no limit to the number of key presses. A '+' fixation point was presented for 1,000 ms in the next screen, followed by the stimulus video for 25,000 ms. The other two behavioural questions were presented at the end of the video. The participants could control their own timings. This process constituted one cycle of the experimental procedure. Forty-eight stimulus videos were randomly presented in 48 trials. A rest period was allowed after every 12 videos, with the participant deciding the length of the rest period. The entire experimental process took approximately 35 min. The following three behavioural questions were presented: (1) Are you aware of the branded product in the video? (2) Do you prefer this form of brand placement? (3) Are you willing to buy the branded product when you see it in the video? The options for response included four levels, with 1 being the most positive attitude and 4 being the most negative attitude. The questions were designed to examine the consumer's 'awareness' , 'preference' , and 'purchasing desire' towards the brand placement. Behavioural Data Analyses. The behavioural data for each question were then analysed. ' Awareness' was calculated as the number of times the participant responded that they perceived brand logo information divided by the total number of stimuli for that stimulus condition., while 'preference' and 'purchasing decision' were calculated by summing up the participant's responses. A strongly positive attitude was awarded 4 points; somewhat positive attitude, 3 points; somewhat negative attitude, 2 points; and strongly negative attitude, 1 point. The two factors of brand placement (representation and sound) were used to form four stimulus conditions as follows: (1) representable and speech sound, (2) non-representable and speech sound, (3) representable and musical sound, and (4) non-representable and musical sound. A one-way analysis of variance 135 was performed to investigate their impact on the participant's behaviour and attitude.
Event-related Potential Analysis. Event-relative potentials (ERPs) are brain activity induced by multiple repeated stimulus events. ERP analyses helped determine temporal resolution, which reflected the information processing of the brain at different time epochs 10 . ERPs were distinguished from the recorded EEG signals using superimposition and averaging. ERP data were analysed using the Neuroscan software. Raw EEG data from the 25 participants were first filtered using a band-pass filter. The frequencies of the high-pass and low-pass filters were 0.1 and 30 Hz, respectively. Abnormal segments were deleted; DC signals were corrected to remove electro-oculography interference. EEG waveforms 200 ms before each stimulus and 1000 ms after each stimulus were sliced into epochs and subjected to baseline correction. The time epochs of P1 (90-150 ms) 8,48 , N1 (125-175 ms) [49][50][51] , P2 (150-250 ms) 13,51,52 , N270 (242-300 ms) 53 , P3 (300-500 ms) 16,54 , and LPC (500-900 ms) were observed. The epochs were then superimposed and averaged to obtain ERPs of brand placement under different conditions.
The overall exclusion rate of ERP epochs for the 25 participants and all stimulus conditions was 9.4%. The exclusion rates for individual stimulus conditions were as follows: 'representable versus speech sound' , 7.0%; 'non-representable versus speech sound' , 12.7%; 'representable versus musical sound' , 9.5%; and 'non-representable versus musical sound' , 11.6%. Each level of single independent variables has at least 30 epochs for each participant.
Electroencephalography is classically considered as being an excellent temporal resolution among the different brain imaging techniques 1 . Analysis of ERPs, using the average evoked response method, has now become predominant for human experiments over recent decades. The limitation of ERPs was time-locked and phase-locked EEG activity using simple averaging technique enabled to reflect a specific cognitive processing with time-related voltage fluctuations 136 . ERP averaging measures depended on time-locked, more specific response epoch and phase synchronization to brain activity of the event, not only the less phase resolution of EEG, but less overall measures of the time course regarding the underlying sources at scalp level 137,138 . Thus it can be enriched by EEG to record brain electrical activity from the scalp corresponded with the ERSP analysis of differences in the mean spontaneous EEG frequency spectrum under various stimuli conditions. Independent Component Analysis and Clustering. By using the algorithms for independent component analysis (ICA) in the EEGLab toolbar 139 , the present study controlled the noise generated during the experimental process and investigated the activation response of each brain region to different methods of brand placement. The underlying assumption of ICA was that the EEG signals recorded from the electrodes were composites originating from neural signals of different brain cortical regions 140,141 . Hence, ICA effectively removed the effects of participants' eye movements, noise, and electromyographic responses on EEG signals 142,143 . First, the frequency of raw EEG signals for all participants was reduced to 250 Hz, and then a 100-Hz low-pass filter and a 0.5-Hz high-pass filter were applied to reduce sampling and remove noise. EEG data were then segmented into epochs based on the 4 stimulus conditions. Data from 1,000 ms pre-stimulus to 4,000 ms post-stimulus were used for superimposition and averaging. Then, the filtered EEG signals were subjected to ICA to isolate 30 independent components (ICs).
Our study collected a total of 750 ICs (25 participants * 30 ICs) of EEG signals to implement the analysis of ICAs. All EEG signal data computed by the ICAs were characterised by Brodmann brain area used the K-means 139,144 to cluster 750 ICs into the following 12 brain areas: left frontal area, frontal midline area, right frontal area, left temporal area, central area, right temporal area, left parietal area, parietal area, right parietal area, left occipital area, occipital area, and right occipital area. The Talairach xyz coordinated with the regions centroids 145,146 were matched to the Brodmann areas 147 to investigate their brain functions. However, as shown in Fig. 4, the numbers in each cluster area tended to vary. The weight of ICs shown in the scalp map represented the regional locations of active sources in brain activity, enabling the examination of the source of brain activation in the participants 139 . Subsequently K-means clustering was also used in the study of dipole locations and power spectrum density 139,144 . Significant differences of frequency bands oscillations in the ERSP corresponded to the significant brain areas. ERSP analysis. In the present study, the ERSP was the spectral perturbation map of the brain, with time as the horizontal axis and frequency as the vertical axis 148 . The EEGLab software was used for ERSP analysis. Wavelet transform was performed to obtain 3-to 100-Hz band intervals to transform the signals of each trial into temporal and frequency data. The Infomax ICA algorithm 141,149 from the EEGLab toolbox 139 was employed to effectively find representations for image and audio for tasks such as feature compression and noise removal (e.g. eye movement and blinking, single-electrode noise, muscle activities) when processing EEG signals 142,143,150,151 . Accordingly, the filtered EEG signals with representations were classified into images of independent components corresponding to brain regions. ERSP maps were obtained after the data were normalised using the spectral intensity at baseline to observe intensity changes of the delta (1-3 Hz), theta (4-7 Hz), alpha (8-13 Hz), beta (13-30 Hz), low-gamma (31-60 Hz), and high-gamma (>60 Hz) frequency bands, as well as their corresponding neurophysiological and cognitive implications. Significant between-group differences were analysed using a T-test to examine EEG changes in different brain regions when participants viewed the 4 types of stimulus conditions. P-values were adjusted using the false discovery rate (FDR)-controlling multiple testing procedure (the fdr.m routine from the EEGLAB toolbox) 152 .
ERSP analysis of event-related dynamics of the EEG spectrum induced by the moment of the beginning of single stimuli trials, were not averaging phase and time-locked response epochs to the stimuli. Moreover, stimuli, such as auditory stimuli, which were not captured by ERP were omitted 138 since the responses would have elicited a wide range of frequency band oscillations in the ERSP. The ERSP revealed features of event-related brain dynamics that were not contained in the ERP average of the same response epochs 137 . The ERSP was an analysis of single response epochs, multi-channel source localization studies that was capable of obtaining event-related power spectrum density in terms of frequency oscillations from EEG data, which could be phase-coherent/incoherent across selected spatially adjacent or non-adjacent cortical areas, plotted in the plane of time (ms, the horizontal axis) by frequency (Hz, the vertical axis) 153 , corresponding to the temporal-spatial resolution. Efforts to observe and characterise phase-incoherent event-related brain dynamics had been an emerging research topic since 154,155 .