Auditory Steady State Response to the Music with Embedded Binaural Beats during Daytime Sleep

Music with embedded binaural beats (BBs) is a promising noninvasive tool for insomnia treatment. This work tests the hypothesis of “entrainment” effect of that kind of music on human brainwaves as a physiological mechanism of sleep improvement. The spectrum of auditory steady state response has been compared in the group of 21 subjects during daytime sleep onset, with music embedded with BBs of 2 and 4 Hz (stimulus condition) and without any stimulus (control condition). Significant difference in auditory steady state response power has been found between conditions at 14 Hz (“sleep spindle” frequency). However, no significant difference has been found at frequencies of 2 and 4 Hz embedded to add a somnogenic feature to the music stimulus according to the hypothesis tested. The results obtained do not support the hypothesis of “entrainment” effect of the stimulus examined on sleep; yet they support the conclusion made in the previous authors’ paper that it improves daytime sleep quality.

According to surveys [4], music is used by approximately 25% of the population as a sleep aid. There are scientific reviews on the influence of music on sleep [3,[5][6][7][8]. In the last of them six possible reasons (hypotheses) for the sleep to be improved by listening to the music are highlighted: relaxation, distraction, entrainment, masking of noxious background noise, pleasant emotional associations, and cultural expectations around music.
The third hypothesis ("entrainment") can be seen as a special case of the broader assumption called "brainwave entrainment." It involves the resonant interaction of the rhythmic component (beat) of music with the brainwaves and consequently involves a predictable effect on a person's vigilance level. Commonly, a relatively slow beat is used in music that does not exceed 4 Hz, the upper limit frequency of delta brainwaves, characteristic of slow-wave sleep. Therefore, a musical beat can be a noninvasive tool of sleep improvement according to the hypothesis mentioned above. A musical beat can be either explicit, when there is an active percussion part for example, or implicit, when binaural beats (BBs) are embedded with the music. BBs are psychoacoustic illusion of sound pulsation occurring when two monotonous acoustic signals with slightly different frequencies (up to about 30 Hz) are applied separately to the right and left ear. For example, if one of the signals has a frequency of 200 Hz and the other has a frequency of 204 Hz the beats at 4 Hz will be sensed. Unlike ordinary acoustic monaural beats (hereinafter referred to as MBs), BBs have no physical carrier. The appeal of sound stimuli with embedded BBs for the therapy of sleep disorders lies in the fact that a human can sense them at very low sound volume essentially bordering the hearing threshold [9], i.e., such a stimulus little interferes with sleep development. Yet the effect of BBs on sleep is studied insufficiently; this is particularly relevant for objective studies based on the analysis of physiological data: electroencephalogram (EEG) and electrocardiogram. There are only a handful of works at the moment that meet these requirements [10][11][12][13].
The "entrainment" hypothesis mentioned above has drawbacks. It is easy to notice that it consists of two coupled assumptions. The first one is a special case of the effect well known as "EEG neural entrainment" by periodic sensory stimuli. The possibility to entrain EEG (i.e., brainwaves) with music patterns is shown in paper [14]. However, the deepness of entrainment, MOSCOW  that is the entrainment of musical beat in our special case, can vary greatly for different experimental conditions. There are conditions of good entrainment, there are conditions of no entrainment, and those conditions remain a matter of research. This is the first drawback.
Furthermore, changes in brainwave spectrum caused by the beat entrainment can follow not only the stimulus frequency. They can take place as well at other brainwave frequencies that are not characteristic of sleep being outside the required theta-and deltarange (so called "cross-frequency" effect of BBs [15,16]). Accordingly, there would be no sleep improvement as a result of beat entrainment in that case. This is the second drawback.
Because of the above, the range of applicability of the "entrainment" hypothesis, at least for sound stimuli presented during sleep, remains unclear for the moment. For example, one pilot study [10] showed the latency of stage two during sleep development accompanied by sound embedded with BBs to be less than the same latency during sleep development accompanied by identical sound but embedded with MBs. That fact is difficult to explain by beat entrainment because BBs entrain brainwaves less than MBs [15,17]. On the other hand, there is a study into the effect of 3-Hz BBs on the third stage of slow wave sleep, with the results quite explainable by beat entrainment [11].
The present study is devoted to test a special case of the "entrainment" hypothesis that is the effect of the latent musical beat in the form of slow 2-Hz and 4-Hz BBs corresponding to delta brainwaves on daytime sleep. These are BBs chosen as a rhythmic basis of the music rather than MBs, because such stimulus reduces sleep latency more effectively according to some data [10]. To test the hypothesis, the technique of auditory steady state response (ASSR) analysis was chosen. ASSR is an instrumentally recorded (by EEG or magnetoencephalogram) brain response to frequency-specific sound stimuli continuous for a sufficiently long period of time, whereas frequency components of the stimuli must remain constant in amplitude and phase [18]. In case of complex stimuli, beats for instance, ASSR repeats their envelope, thus affecting the background EEG activity [19]. Unlike ordinary auditory evoked potentials, described with amplitude and latency of their individual components, ASSR is used to be described with frequency and phase response functions of the spectrum. That technique is most common in otolaryngology since it allows an objective assessment of hearing sensitivity both in people with normal hearing and with various hearing impairments.

MATERIALS AND METHODS
The experimental group consisted of 21 medical university students (12 males and nine females aged 18 to 22 years; mean age 20.1 ± 0.7 years).
A custom electronic composition of 20 min duration with embedded BBs of 4 and 2 Hz was explored as musical stimulus. The first 19 minutes of its sounding time were "programmed" to help the listener fall asleep and the remaining time was to awaken rapidly. The beats were superimposed from the beginning of the track up to and including the 19th minute, in blocks of 64-s duration. The 20th minute contained no BBs (Table 1).
Each subject participated in two trials: one to fall asleep listening to music (stimulus condition) and the other to fall asleep without music, in silence (control condition). The order of trials was counterbalanced across the sample; 13 of the 21 subjects had the "control condition" first as a result of random selection. The trials were performed from 1:00 p.m. to 4:00 p.m., with between-trial interval of not more than 15 days.
The stimuli were presented through full-size Bose QC-25 stereo headphones (sensitivity 97 dB, impedance 32 ohms) with the active noise reduction system turned off. Sound volume in each trial was chosen within the individual comfort zone, i.e., between 55 and 57 dB SPL.
Subjects were kept in a soundproof light-protected room at a stable temperature of 24°C. The polysom- After electrode placement, the subject was required to lie on a couch in a horizontal position and to keep his eyes open for the first 15 minutes. At the same time, he was monitored to keep vigilance. Then he was required to close his eyes and baseline PSG was recorded for 3 min. Next, the music was started and the subject was awakened (if he fell asleep) after it stopped 21 min later and another 3-min piece with eyes closed was recorded.
In control trials, the experimental scheme was identical except the stimulus was not presented, so the subject was kept in silence for 21 min.
To calculate ASSR, a technique similar to that described in [15] was explored. A recording segment starting from 20 s after the music onset and up to the beginning of the 20th min was isolated on each EEG record, both in control and stimulus conditions. This segment was then divided with corresponding labels into intervals of 1-s duration; artifact intervals were manually removed after visual analysis. C3 and C4 electrode positions were chosen as a signal source because they are conventionally used in PSG studies and allow one evaluate activity of both anterior and posterior brain areas. EEG signal in C3 and C4 positions was additionally filtered in the range from 1 to 20 Hz and then the following operations were performed using the custom Matlab script: 1. Amplitude averaging of artifact-free, one-second recording intervals was applied.
2. The EEG spectrum amplitudes A(f) in C3 and C4 positions for all 42 averages of 1-s duration derived in step one were calculated for each integer EEG frequency value (f variable) in the range from 2 to 20 Hz, with discrete Fourier transformation applied after linear trend elimination.
3. Relative power P rel (f) was calculated for each of 19 frequency values as the ratio of the squared corre-sponding Fourier index A 2 (f) to average EEG signal power on the baseline 3-min segment P baseline . The average power was calculated as the sum of squared Fourier transformation indexes (see step two above) of the signal obtained by averaging (step one above) of artifact-free 1-s intervals of the 3-min baseline recording segment:

RESULTS AND DISCUSSION
Subjects' PSGs were manually scored according to recommendations of the American Academy of Sleep Medicine [20] with an epoch of 30 s. All subjects under stimulus condition fell asleep (signs of at least the first sleep stage were noticed on the hypnogram), while one subject showed no sign of sleep under the control condition. Table 2 shows the difference in mean values of sleep characteristics across stimulation and control trials. Mean duration of stage one is significantly less under the stimulus condition than under the control (Student's t-test, p = 0.015).
In order to compare ASSR spectra under stimulus vs. control conditions, the null hypothesis was stated for each integer EEG frequency value (variable f) in the range from 2 to 20 Hz, i.e., no difference was stated between the relative power spectrum values P rel (f) calculated under control condition, and similar values calculated under stimulus (music) condition. The hypothesis was tested with paired, two-tailed Student's t-test after logarithmic transformation of P rel (f) values. Logarithmic transformation was applied to make the data distribution closer to normal. Moreover, in our case, it reflects the physiological mechanism of the sound volume processing by brain, which is expressed by the logarithm of relative sound pressure (decibels). Rejection level of the null hypotheses was accepted at p < 0.05.
Statistical analysis revealed significant increase in P rel value at the frequency of 14 Hz in electrode positions C3 and C4 under stimulus condition compared to the control (Fig. 1). This can be accounted for by an 19 increase in quantity and amplitude of so-called "sleep spindles," characteristic of the second and third stages of sleep since it is 14 Hz that is the most probable spectral frequency of "sleep spindles" in the age group studied. No significant difference was noticed between the stimulus condition and the control condition for the rest of brainwave frequencies from 2 to 20 Hz. This range includes BB-frequencies of 2 and 4 Hz embedded with the musical stimulus, which means there is no alleged entrainment effect of the studied stimulus on brainwave activity.
Thus, the results obtained do not confirm the "entrainment" hypothesis mentioned in the beginning of the article as a basic mechanism of sleep improvement with BB-embedded music. However, there are results of another work [14] in favor of this hypothesis. The discrepancy could be explained by the Fig. 1. Dependence of relative spectral power of Auditory Steady State Response (ASSR) on EEG frequency, for (a) C3 and (b) C4 positions. EEG frequency from 2 to 20 Hz (f variable) is plotted on the horizontal axis; ASSR relative spectral power-P rel (f) (see calculation details in the text)-averaged over all subjects (geometric mean) is plotted on the vertical axis. Error bars correspond to 95% confidence level (p = 0.05). Significant differences between the stimulus and control conditions at 14 Hz are indicated by arrows. The difference in P rel (f) for other EEG frequencies of the studied range are not significant. lower perceived modulation depth of BBs [9] as compared to MB modulation depth used in [14]. We should also take into account relatively lower volume of the stimulus (music) explored in our study (it was limited to avoid an arousal effect) as compared to the sound volume used in [14].
On the other hand, the results obtained confirm the conclusion made in one of our works [13] that the studied combination of music and BBs improves quality of daytime sleep since the observed increase in spectral brainwave power at "sleep spindle" frequency evidences an increase in proportion of the second and third sleep stages. That can be seen as well if we directly compare the duration of sleep stages: stimulus presentation lead to the redistribution of total sleep time towards its deeper (second and third) stages, due to a decrease in the first stage's duration.

FUNDING
The study was financed from the state budget (no. AAAA-A17-117092040002-6) and supported by the Russian Foundation for Basic Research (grant no. 19-013-00747a).

COMPLIANCE WITH ETHICAL STANDARDS
Conflict of interests. The authors declare that they have no conflict of interests.
Statement of compliance with standards of research involving humans as subjects. The study complied with the ethical standards of the World Medical Association's Declaration of Helsinki "Ethical Principles for Scientific Medical Research Involving Human Subjects," as amended in 2000, and was approved by the ethical committee of the Institute of Higher Nervous Activity and Neurophysiology, Russian Academy of Sciences. Informed consent to participate in the study was signed by each participant.