5D Library › 默认模式网络、熵脑与神经振荡 · English

第六章 夹带:40 Hz 与双耳节拍

节律能从外部施加吗?

第二章说明了耦合振子可以步调一致。有节律的刺激,例如闪烁的光或脉动的声音,就是一种来自外部的耦合:原则上,它能把神经群体拉向自己的频率。大脑反应会跟随有节律的刺激,这一点已经确立;问题在于,这种效应能扩展到最先接收刺激的感觉区之外多远,以及它是否改变了任何重要的东西。

40 赫兹

证据

在被改造为会出现类阿尔茨海默病理改变的小鼠中,麻省理工学院蔡立慧(Li-Huei Tsai)团队发现,以 40 Hz 驱动 gamma 活动,先用光遗传学方法,后用 40 Hz 闪烁光的非侵入方法,能减少视皮层中的 β-淀粉样蛋白,并改变小胶质细胞的形态与活动 (Iaccarino et al. 2016)。该团队后来的研究把 40 Hz 光与声音结合起来。

证据

另一个独立团队报告,在两种小鼠模型中,40 Hz 闪光没有在视皮层之外夹带出固有的 gamma 振荡,一小时或七天的刺激也没有显著改变淀粉样蛋白负荷或小胶质细胞形态 (Soula et al. 2023)。

证据

在人体中,一项可行性研究发现 40 Hz 声光刺激是安全的,并能夹带皮层和一些更深部位的活动;一项纳入 15 名轻度阿尔茨海默病痴呆患者、每天刺激、持续三个月的小型随机预试验,报告刺激组脑萎缩较少,连接和记忆测试表现也有一些差异,作者称这些是探索性结果,需要更大规模的试验 (Chan et al. 2022)。

总的来说,40 Hz 研究是一条严肃的研究路线:动物结果互相矛盾,人体结果早期且规模小,更大规模的人体试验已经开始。它还不是一种治疗;面向健康人出售的「40 Hz 健脑」设备或音频,并不建立在这些证据之上。

双耳节拍

用耳机给左耳播放 200 Hz、给右耳播放 210 Hz 的纯音,许多人会听到一个 10 Hz 的缓慢起伏。与两个音在空气中混合时听到的普通拍音不同,这个起伏在任何一只耳朵的声音里都不存在。它是大脑产生的,产生于比较两耳声音时间差的脑干核团,主要是上橄榄复合体,听觉通路更高的部位也记录到相应反应 (Oster 1973; Ingendoh et al. 2023)。

推导

左耳 200 Hz、右耳 210 Hz。任何一只耳朵的信号在 10 Hz 处都没有能量;两个音在同一只耳朵里混合后的包络则有。戴耳机时,以 10 Hz 循环的是两耳之间的相位差。

import math

fs, n = 4000, 4000                       # one second at 4 kHz
fl, fr = 200.0, 210.0                    # tone to the left ear, tone to the right ear
t = [i / fs for i in range(n)]
left = [math.sin(2 * math.pi * fl * u) for u in t]
right = [math.sin(2 * math.pi * fr * u) for u in t]

def power_at(x, f):
    """Squared magnitude of the discrete Fourier component at f Hz."""
    re = sum(v * math.cos(2 * math.pi * f * i / fs) for i, v in enumerate(x))
    im = sum(v * math.sin(2 * math.pi * f * i / fs) for i, v in enumerate(x))
    return (re * re + im * im) / len(x) ** 2

mixed = [a + b for a, b in zip(left, right)]           # both tones in one ear
envelope = [v * v for v in mixed]                      # a crude rectifier
print(f"10 Hz component, left ear alone:       {power_at(left, 10):.4f}")
print(f"10 Hz component, right ear alone:      {power_at(right, 10):.4f}")
print(f"10 Hz component, envelope of the mix:  {power_at(envelope, 10):.4f}")
# With headphones the two tones never mix in the air: the 10 Hz 'beat' exists
# only as the interaural phase difference, which the brainstem compares.
for ms in (0, 25, 50, 75, 100):
    ipd = (2 * math.pi * (fr - fl) * ms / 1000) % (2 * math.pi)
    print(f"  t = {ms:3d} ms: right ear leads by {math.degrees(ipd):5.0f}°")
# 10 Hz component, left ear alone:       0.0000
# 10 Hz component, right ear alone:      0.0000
# 10 Hz component, envelope of the mix:  0.2500
#   t =   0 ms: right ear leads by     0°
#   t =  25 ms: right ear leads by    90°
#   t =  50 ms: right ear leads by   180°
#   t =  75 ms: right ear leads by   270°
#   t = 100 ms: right ear leads by     0°
上:戴耳机时,两耳之间的相位差每秒十次从 0° 循环到 360°。下:如果两个音在同一只耳朵里混合,声音本身每秒十次时强时弱。

这对「脑波夹带」的说法很重要。单耳拍或扬声器播放的拍音,是声音真实的振幅调制,听觉系统会跟随它。双耳节拍则是细微的两耳相位差,需要由脑干计算出来;大脑对它的反应更弱。它是否会进一步把皮层节律拉向拍频,正是研究所检验的问题。

证据

一项汇总 22 项研究的元分析发现,双耳节拍对记忆、注意、焦虑和疼痛等结局总体有中等效应(Hedges’ g=0.45g = 0.45),在任务前开始聆听、聆听时间更长时效应更大 (Garcia-Argibay et al. 2019)。一篇系统综述汇总了 14 项测量脑振荡的研究,其中 5 项结果与夹带一致,8 项与之矛盾,1 项结果混合,作者认为证据尚无定论 (Ingendoh et al. 2023)。

推导

0.45 的效应量在实际中意味着什么,以及那篇脑电研究综述的统计。

from statistics import NormalDist

Phi = NormalDist().cdf

def superiority(g):
    """Probability that a random person from the stimulated group scores
    better than a random person from the control group (equal variances)."""
    return Phi(g / 2 ** 0.5)

def overlap(g):
    """Share of the two distributions that overlaps (Cohen's OVL)."""
    return 2 * Phi(-abs(g) / 2)

for label, g in (("small", 0.2), ("binaural-beat meta-analysis", 0.45), ("large", 0.8)):
    print(f"g = {g:.2f} ({label}): better than control {100 * superiority(g):.0f} % "
          f"of the time; distributions overlap {100 * overlap(g):.0f} %")

# EEG entrainment studies in one systematic review (see the text)
studies = {"consistent with entrainment": 5, "contradictory": 8, "mixed": 1}
total = sum(studies.values())
for k, v in studies.items():
    print(f"  {k:28s} {v:2d} of {total}")
# g = 0.20 (small): better than control 56 % of the time; distributions overlap 92 %
# g = 0.45 (binaural-beat meta-analysis): better than control 62 % of the time; distributions overlap 82 %
# g = 0.80 (large): better than control 71 % of the time; distributions overlap 69 %
#   consistent with entrainment   5 of 14
#   contradictory                 8 of 14
#   mixed                         1 of 14

这样大小的效应意味着:听双耳节拍的人,大约 62% 的时候会比不听的人表现好,而不是每次都好;两组分数的分布约有五分之四重叠。而且,能证明这种效应是通过夹带脑节律、而不是通过放松、期望或单纯安静地听声音起作用的证据很弱。这就是第八章的实验把双耳节拍与另一种安静聆听作比较、而不是与什么都不做作比较的原因。

假说 · 尚未验证

尚未确立。 双耳节拍报告的行为效应,由安静专注的聆听和期望解释,而不是由拍频处的夹带解释。该假说预测:在参与者不知道研究假设的情况下,双耳节拍在注意与情绪指标上,不优于同样吸引人的非拍音声音或同样时长的专注听息练习。如果在一项预注册试验中,双耳节拍以值得重视的幅度优于这类积极对照,假说就被推翻。

参考文献

Chan, D., H.-J. Suk, B. L. Jackson, et al. 2022. “Gamma Frequency Sensory Stimulation in Mild Probable Alzheimer’s Dementia Patients: Results of Feasibility and Pilot Studies.” PLoS ONE 17 (12): e0278412.
Garcia-Argibay, Miguel, Miguel A. Santed, and José M. Reales. 2019. “Efficacy of Binaural Auditory Beats in Cognition, Anxiety, and Pain Perception: A Meta-Analysis.” Psychological Research 83 (2): 357–72.
Iaccarino, H. F., A. C. Singer, A. J. Martorell, et al. 2016. “Gamma Frequency Entrainment Attenuates Amyloid Load and Modifies Microglia.” Nature 540 (7632): 230–35.
Ingendoh, R. M., E. S. Posny, and A. Heine. 2023. “Binaural Beats to Entrain the Brain? A Systematic Review of the Effects of Binaural Beat Stimulation on Brain Oscillatory Activity, and the Implications for Psychological Research and Intervention.” PLoS ONE 18.
Oster, Gerald. 1973. “Auditory Beats in the Brain.” Scientific American 229 (4): 94–102.
Soula, M., A. Martín-Ávila, Y. Zhang, et al. 2023. “Forty-Hertz Light Stimulation Does Not Entrain Native Gamma Oscillations in Alzheimer’s Disease Model Mice.” Nature Neuroscience 26 (4): 570–78.

Fan, Y. (2026). The Default Mode Network, the Entropic Brain and Neural Oscillations: 5-HT2A Receptor Kinetics, Pineal Calcite Microcrystals and the Computational Neuroscience of Meditative Attention [默认模式网络、熵脑与神经振荡] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23697320 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).