Blog

  • 一个特征量管五种楼形:尾流宽度能否统一预测涡脱频率的裁决

    Evidence level: B

    上一轮证明了”St 数可以被回归”,这一轮要问更难的一问:这个回归的自变量,到底该用盖楼时量得到的尺寸,还是用只有风洞里才看得见的尾流宽度?

    正文

    第一层:这个想法不是我们提的,是 Roshko 1948 年提的。 涡脱频率的常规写法是 f = St·U/D,D 是迎风宽度。但 Roshko 发现,只要把 D 换成”尾流里两条自由流线的间距”d′、把 U 换成分离点处的流速 Uₛ,得到的 St* = f·d′/Uₛ 在所有测过的柱体上都落在 0.16 附近——平板、楔形、圆柱都不例外(Roshko, On the Wake and Drag of Bluff Bodies, NACA TN 3168, 1948)。这就是”统一特征量”的最初形态:不是形状决定频率,是尾流的宽度决定频率;形状只是通过改变尾流宽度间接起作用。

    第二层:这条线索后来被两次加固。 Ponta & Aref 2004 年从涡街几何的稳定性出发,把 St-Re 关系从经验拟合推成了有理论骨架的式子(PMID 15447193);Roushan & Wu 2005 年给出它的结构化解释——关系的形式来自涡形成机制本身,与具体形状无关(PMID 15783649)。到 Kim & Wu 2015 年的肥皂膜实验,事情被推到最干净的一步:多种形状的障碍物生成层流涡街,Re→∞ 时 St(∞)=1/A≈0.21 与形状完全无关,形状差异全部被压缩进单个参数 B(PMID 26565336)。形状没有消失,它只是退到了一维。

    第三层:但有一个反证必须摆在这里。 2025 年 PNAS 那项跨 28 个物种的巡航飞行研究,结论对”St 是普适常数”是冷水:St 与飞行条件强相关,它的普适性只在”偏好条件”内成立;真正跨物种一致的是另一个量——广义涡形成时间,用涡量注入时长比上涡量增长率来定义,与飞行条件无关(PMID 40880533)。这说明 St 本身可能是个坏因变量:它是”几何 × 工况”的混合产物,拿它当预测目标,等于逼模型去学一个本来就不该统一的量。我们真正该预测的,也许是”无量纲涡形成长度”,而 St 只是它的一个投影。

    第四层:再附长度这个候选量,有直接的干预证据。 2026 年 Sci Rep 的 cross-slit 圆柱参数扫描做了件少见的事:它不改工况、只改几何(缝宽比 0.10-0.25、扫掠角),然后同时报三个数——再附长度下降 40-52.5%、阻力下降 26-32.5%、St 数在 Re=100 时下降 93.1%(PMID 42581059)。三个量同向大幅移动,说明再附长度与 St 之间不是松散相关,而是强耦合。但它同时暴露了一个麻烦:改动一个几何参数,所有候选特征量一起动,你无法从这类数据里判断谁是因、谁是果。 这正是本课题必须用”留一截面族”而不是”留一样本”来裁决的原因。

    第五层:建模这一侧,两条路都有前例,而且指向相反。 一条是黑箱:SuperWing 数据集用 4,239 个参数化机翼、28,856 个 RANS 流场训练 Transformer,在留出样本上阻力误差 2.5 个阻力单位,并且对 DLR-F6、NASA CRM 这类完全没见过的复杂机翼表现出零样本外推能力(PMID 42538339)——它证明”参数化几何 + 足够大的数据 + 现代网络”确实能跨外形外推,但代价是四千级样本。另一条是可解释:微塑料碎片沉降速度那项研究里,作者从受力分析推了一个新的不规则形状因子 ISF,结果比 Corey 形状因子准,也比机器学习方法准(PMID 39419206)。一个推出来的无量纲量,赢了一个训出来的网络。 这两条路线的对照,决定了本课题的建模顺序:先找特征量,再用黑箱做上界参照,而不是反过来。

    第六层:所以裁决设计要卡住一个零假设。 五类截面(方/圆/锥/切角/扭转)× 关键形态参数做统一 CFD 与风洞参数扫描,每个算例同时输出 St、尾流有效宽度 d′、分离角、再附长度 Lr/D、基压系数;然后并列三个模型:A 只用盖楼量得到的原始几何参数(宽深比、锥率、切角比、扭转角);B 只用统一特征量(d′/D、Lr/D 等);C 黑箱网络。核心零假设是:B 相对 A 没有增量。 检验方式是留一截面族交叉验证——用四类训练、预测剩下那一类,看相对误差是否守在 5-10% 以内。若 B 打不过 A,”统一特征量”就只是个换了名字的形状哑变量,统一是假的;若 B 显著优于 A 且能跨族外推,才算找到了那个比”截面类型”更深的变量。

    第七层:三种结局,对应三种形煞观。 ① B 胜且外推误差 <5-10%:存在一个可量、可算、可跨形使用的几何量,形煞评级可以脱离逐案风洞,进入”量三个尺寸算一个数”的阶段——但要注意,这时预测的是频率不是吉凶,中间还隔着”频率是否落在器官共振带”这一层。② B 只在部分截面族成立(扭转、切角最可能掉队):形煞的细分流派可能确实踩到了物理边界,但边界在”依赖方式”上而不在”吉凶等级”上。③ B 全面失败:形煞必须逐案验证,”一形一断”获得物理背书,而”峦头一诀、百形皆应”只能当经验口诀、不能当公式。

    回到易学。 形峦派内部其实一直有两句话在打架:一句是”峦头一诀,百形皆应”,讲的是万形同理;一句是”一形一断,不可类推”,讲的是就形论形。风水师靠肉眼在两句之间来回,靠的是手感。这篇研究做的就是把这句手感翻译成一个可判定的问题:到底存不存在一个”万形同理”的量? 存在,那两句里前一句对;不存在,后一句对。古人没能力做这件事,是因为他们看不见尾流;今天我们看得见,就不该再靠手感决定用哪一句。

    证据与局限

    已验证: 常规 St 数可用尾流宽度与分离点流速归一为通用 St*≈0.16,且在平板/楔/圆柱等尖缘或圆形截面上成立(Roshko 1948, NACA TN 3168);St-Re 统一关系 St=1/(A+B/Re) 对多种形状障碍物成立,Re→∞ 极限 St≈0.21 与形状无关、形状差异集中在参数 B(PMID 26565336,PMID 15783649,PMID 15447193);数值方法的边界处理与阻塞比会系统性改变所测 St,必须在协议层校正(PMID 14682839);St 的跨物种普适性受工况限制,广义涡形成时间才是跨条件不变量(PMID 40880533);几何微改(缝宽比/扫掠角)可同时大幅改变再附长度、阻力与 St(PMID 42581059);参数化几何 + 大规模数据可实现对未见外形的零样本气动外推(PMID 42538339);从受力分析推导的无量纲形状因子在预测不规则体沉降速度上优于机器学习(PMID 39419206)。

    部分验证: 各候选特征量(尾流有效宽度、分离角、再附长度、基压系数)与 St 的耦合关系在单个截面族内有充分证据,但”哪一个可跨五类截面统一使用”从未被同一批算例、同一套协议横向比较过;”用尾流宽度而非迎风宽度归一后 St 变常数”这一结论在高 Re、三维、强湍流来流条件下的适用性仍有争议。

    待验证: ① 是否存在一个无量纲几何特征量,可跨方/圆/锥/切角/扭转五类截面把 St 的外推误差压到 5-10% 以内;② 若 St 本身不是好的因变量,改用广义涡形成长度后跨截面统一性是否提高;③ 跨截面泛化失败时,形煞分类是否对应 St 依赖方式的物理分界;④ 用盖楼可量的三个尺寸替代风洞特征量,精度损失有多大(工程可用性边界)。

    相关阅读

    证据等级: B
    参考文献: PMID 26565336, PMID 15783649, PMID 15447193, PMID 14682839, PMID 40880533, PMID 42581059, PMID 42538339, PMID 39419206


    免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

    Five Feature Quantities for Predicting the Strouhal Number of Vortex Shedding from Wake Width

    Evidence level: B

    The previous round demonstrated that “St can be regressed.” This round asks the harder question: for this regression, should the independent variable be the dimensions measured during building design, or the wake width visible only in wind tunnels?

    Main Text

    Layer 1: This idea was not proposed by us; it originated with Roshko in 1948. The conventional formulation of the Strouhal frequency is $f = St \cdot U/D$, where $D$ is the windward width. However, Roshko found that replacing $D$ with “the spacing between two free streamlines in the wake,” denoted as $d’$, and replacing $U$ with the velocity at the separation point $U_s$, yields a modified Strouhal number $St^ = f \cdot d’/U_s$ that clusters near 0.16 across all tested bluff bodies—flat plates, wedges, and cylinders are no exception (Roshko, On the Wake and Drag of Bluff Bodies, NACA TN 3168, 1948). This represents the initial form of a “unified feature quantity”: not shape determines frequency, but wake width determines frequency; shape acts indirectly by altering wake width.*

    Layer 2: This line of inquiry was subsequently reinforced twice. Ponta & Aref (2004) derived a theoretically grounded equation for the $St-Re$ relationship starting from the geometric stability of vortex streets (PMID 15447193); Roushan & Wu (2005) provided a structural explanation— the form of the relationship stems from the vortex formation mechanism itself and is independent of specific shapes (PMID 15783649). By Kim & Wu’s 2015 soap film experiments, the matter was pushed to its cleanest step: obstacles of various shapes generated laminar vortex streets, and as $Re \to \infty$, $St(\infty) = 1/A \approx 0.21$ was independent of shape, with all shape differences compressed into a single parameter $B$ (PMID 26565336). Shape did not disappear; it merely receded to one dimension.

    Layer 3: However, a counter-evidence must be presented here. The 2025 PNAS study on cruising flight across 28 species concluded that “St is a universal constant” is a cold water: $St$ is strongly correlated with flight conditions, and its universality holds only within “preferred conditions”; what remains consistent across species is another quantity—the generalized vortex formation time, defined as the ratio of vorticity injection duration to vorticity growth rate, which is independent of flight conditions (PMID 40880533). This suggests St itself may be a poor dependent variable: it is a mixed product of “geometry × operating conditions,” and using it as a prediction target forces the model to learn a quantity that should not be unified. What we should truly predict is perhaps the “dimensionless vortex growth length,” with $St$ merely being one projection of it.

    Layer 4: Regarding this candidate quantity, there is direct intervention evidence. The 2026 Sci Rep cross-slit cylinder parameter scan did something rare: it kept operating conditions constant while only altering geometry (slit width ratio 0.10-0.25, sweep angle), then reported three metrics simultaneously—reattachment length decreased by 40-52.5%, drag decreased by 26-32.5%, and $St$ decreased by 93.1% at $Re=100$ (PMID 42581059). The three quantities moved significantly in the same direction, indicating a strong coupling between reattachment length and $St$, rather than loose correlation. However, it simultaneously exposed a complication: altering one geometric parameter causes all candidate feature quantities to move together, making it impossible to determine cause versus effect from such data. This is precisely why this study must use a “leave-one-section-family” approach rather than “leave-one-sample” to adjudicate.

    Layer 5: On the modeling side, both paths have precedents pointing in opposite directions. One is a black box: the SuperWing dataset trained a Transformer on 4,239 parameterized wings and 28,856 RANS flow fields; on held-out samples, it achieved a drag error of 2.5 drag units and demonstrated zero-shot extrapolation capability for complex wings unseen in DLR-F6, NASA CRM, etc. (PMID 42538339)—proving that “parameterized geometry + sufficient data + modern networks” can indeed extrapolate across forms, at the cost of thousands of samples. The other is interpretable: in a study on microplastic fragment settling velocity, authors derived a new irregular shape factor ISF from force analysis, which proved more accurate than the Corey shape factor and also more accurate than machine learning methods (PMID 39419206). A derived dimensionless quantity wins over a trained network. This comparison determines the modeling sequence for this project: first identify feature quantities, then use black boxes as upper bound references, rather than the reverse.

    Layer 6: Therefore, the adjudication design must hinge on a null hypothesis. Five section types (square/circular/conical/truncated/torsioned) × key morphological parameters undergo unified CFD and wind tunnel parameter scans, with each case simultaneously outputting $St$, effective wake width $d’$, separation angle, reattachment length $Lr/D$, and base pressure coefficient. Then three models are presented in parallel: A uses only raw geometric parameters measured during building design (width-depth ratio, taper ratio, truncation ratio, torsion angle); B uses only unified feature quantities ($d’/D$, $Lr/D$, etc.); C is a black box network. The core null hypothesis is: B provides no incremental gain relative to A. The testing method is leave-one-section-family cross-validation—train on four families, predict the remaining one, and check if relative error stays within 5-10%. If B fails against A, “unified feature quantities” are merely shape dummy variables with renamed labels, and unification is false; if B significantly outperforms A and can extrapolate across families, then a variable deeper than “section type” has been found.

    Layer 7: Three outcomes correspond to three views of Form Sha (shape煞). ① B wins with extrapolation error <5-10%: a quantifiable, calculable, cross-shape geometric quantity exists, allowing Form Sha ratings to detach from case-specific wind tunnels and enter a stage of “calculating one number from three dimensions”—but note, at this stage, the prediction is frequency, not auspiciousness, with an intermediate layer of “whether frequency falls within organ resonance bands.” ② B holds only for partial section families (torsion and truncation most likely to fail): the sub-schools of Form Sha may indeed have hit physical boundaries, but the boundary lies in “dependency mode” rather than “auspiciousness level.” ③ B fails comprehensively: Form Sha must be verified case-by-case, “one shape one judgment” gains physical endorsement, while “one principle for all shapes” can only serve as an empirical mnemonic, not a formula.

    Returning to Yijing Studies. Within the Form Sha school, two sentences have always been at odds: one is “one principle for all shapes,” referring to universal similarity; the other is “one shape one judgment, no analogy,” referring to judging based strictly on form. Feng Shui practitioners oscillate between these two via eye contact, relying on tactile intuition. This research translates that intuition into a determinable question: Does a “universal similarity” quantity actually exist? If it exists, the first sentence is correct; if not, the second is correct. Ancient scholars lacked the capability for this task because they could not see the wake; today we can see it, so we should not rely on intuition to decide which sentence to use.

    Evidence and Limitations

    Verified: Conventional $St$ numbers can be normalized into a universal $St^* \approx 0.16$ using wake width and separation point velocity, holding true for sharp-edged or circular cross-sections like flat plates/wedges/cylinders (Roshko 1948, NACA TN 3168); the unified $St-Re$ relationship $St=1/(A+B/Re)$ holds for various shape obstacles, where the $Re \to \infty$ limit $St \approx 0.21$ is shape-independent and shape differences concentrate in parameter $B$ (PMID 26565336, PMID 15783649, PMID 15447193); boundary handling and blockage ratios in numerical methods systematically alter measured $St$, requiring correction at the protocol level (PMID 14682839); cross-species universality of $St$ is constrained by operating conditions, with generalized vortex formation time being the invariant across conditions (PMID 40880533); geometric micro-modifications (slit width ratio/sweep angle) can simultaneously and significantly alter reattachment length, drag, and $St$ (PMID 42581059); parameterized geometry + large-scale data enables zero-shot aerodynamic extrapolation for unseen forms (PMID 42538339); derived dimensionless shape factors outperform machine learning in predicting settling velocities of irregular bodies (PMID 39419206).

    Partially Verified: Coupling relationships between candidate feature quantities (effective wake width, separation angle, reattachment length, base pressure coefficient) and $St$ have sufficient evidence within single section families, but “which one can be uniformly applied across five section types” has never been compared laterally using the same set of cases and protocols; the applicability of the conclusion “normalizing $St$ with wake width rather than windward width yields a constant” remains debated under high $Re$, three-dimensional, and strong turbulence inflow conditions.

    To Be Verified: ① Does a dimensionless geometric feature quantity exist that can extrapolate $St$ error across square/circular/conical/truncated/torsioned five section types to within 5-10%? ② If $St$ itself is not a good dependent variable, does switching to generalized vortex growth length improve cross-section unification? ③ When cross-section generalization fails, does Form Sha classification correspond to physical boundaries in $St$ dependency modes? ④ What is the precision loss when replacing wind tunnel feature quantities with three dimensions measurable during building design (engineering usability boundary).

    Related Reading

    Evidence Level: B
    References: PMID 26565336, PMID 15783649, PMID 15447193, PMID 14682839, PMID 40880533, PMID 42581059, PMID 42538339, PMID 39419206


    Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.

  • 没感觉到,身体也记账了吗:把「感知」当中介之前

    Evidence level: B

    「无形之煞」这四个字藏着两个不同的主张——一个是”低于你能感觉到的量”,一个是”绕过你的感觉直接起作用”。前者测得到,后者要证明一条因果链,而这条链目前没有被证明。

    正文

    风水讲煞气伤人”于无形”,随后的推论通常是:你感觉不到,不代表没影响。这句话在逻辑上完全站得住——感觉不到确实不等于没发生。但它被往下推一步就成了另一个主张:因为绕过感觉,所以更危险。这一步跳过了因果结构的检验。现代方法学里,这个结构有专门的名字:中介。

    第一,要检验的是链,不是点。

    中介模型写成 D→M→Y:暴露 D 改变中介 M,M 再改变结局 Y。要成立,两个箭头都得有,而且 M 必须站在通路上,不只是跟着一起动。如果 M 只是伴随现象(暴露同时引起 M 和 Y,M 与 Y 无因果关系),那”绕过感知”这个说法就没有对应物——通路根本不经过感知,谈不上绕过。判断这一点只能靠中介分析,不能靠”阈值两侧斜率不一样”来代替(PMID 29283590)。

    第二,那条阈值线本身是模糊的。

    心理物理的阈值不是一条物理刻度,是检出概率到 50% 的那个点,本身带误差。触觉检测阈值的重测 ICC 约 0.7,而频率辨别阈值只有约 0.3(PMID 32300187)——同一个人隔几天测,同一个”阈值”能差出一截。用一条带误差的线把人分成阈上/阈下两组,两组的标签里都掺着错分。而错分对”阈值两侧比较”这种设计是致命的:它不削平差别,它把差别抹匀。

    第三,阈下有生理反应,这一条是真的。

    必须先承认正面证据。让受试者坐在平台上做 0.03–0.2 Hz、0.5–5 mG 的极慢正弦运动,幅度低到没有任何运动感,皮肤交感神经活动(SSNA)的前庭调制仍然显著高于心搏调制,同时心搏调制与皮肤血流双双低于基线(PMID 32965160)。另一条线:把电前庭刺激设在正弦运动阈值的 50% 或皮肤感觉阈值的 80%,仍能改变平衡受挑战时的躯干侧向速度(PMID 32275736)。阈下不是”没反应”,是”没报告”。

    第四,但”有反应”不等于”反应经过感知”。

    这是全篇最关键的一刀。上面两项研究证明的是 D→Y 存在,它们没有、也无法证明 M 在中间。逻辑上二者独立:通路可以完全并行——一套走脑干-交感反射,另一套走皮层报告,互不经过。真正能判断的是”改动 M 而 D 不变,Y 是否跟着变”。现成的实验正好做了这件事:把”次声会致病”的警告给一组人看,高期望组在真次声和假次声下症状都升高(PMID 23477573);反过来,给已经报告症状的人讲清 nocebo 机制,症状与情绪回落到基线,而暴露条件没变(PMID 25981871)。M(期望与感知)动了,Y 动了,D 没动——这恰恰说明 Y 的变动可以不经过 D,也就说明 D→Y 那条链上,M 不是必经关口。

    第五,最强的一条反面证据是直接否证。

    37 名对噪声敏感的健康成人,72 小时、1.6–20 Hz、约 90 dB 峰值(模拟风机次声签名),双盲三臂交叉、洗脱超过 10 天,在隔声睡眠实验室里住。主结局入睡后觉醒时间(WASO)与假暴露无差异:−1.36 分钟(95% CI −6.60~3.88,p=0.60);而作为阳性对照的交通噪声把 WASO 拉长了 6.07 分钟。主观症状、心血管、神经行为表现,次声一项都没变差(PMID 36946580)。剂量给到这份上还没有效应,”阈下更危险”就失去了落点。

    第六,方法上还有两个坑。

    一是断点回归要求断点位置已知且无测量误差,阈值一有误差,断点在原假设下就不可识别,斜率”连续”是默认结果而非发现。二是间接效应的估计本身极不稳:当中介与结局方向相反(抑制)或效应微弱时,比例中介这个数会跑出 [0,1] 甚至改变符号,报告一个”中介了 43%”是没有意义的。这两条决定了:要做这个检验,必须预注册断点位置与中介模型,而不是事后找。

    第七,判决实验长什么样。

    三臂:阈上、阈下(同一暴露积分)、零暴露;中介变量必须用客观检出的 d′(信号检测论下的辨别力),不能用”我感觉到了吗”的自评;结局同时采 SSNA/HRV 与症状量表;用反事实框架分解自然间接效应。判据清楚:阈下臂 D→Y 显著而 D→M 不显著,则感知不是中介,通路并行;若 M 完全中介,则阈下臂本不该有效应——一旦有效应,中介假设就自相矛盾。两条路都指向同一个结论:感知是不是中介,是一个可以被数据判掉的问题,不该停在措辞上。

    易学落点。 风水说”无形”,古人观察到的现象是真的——有些屋子让人不适,而人说不清哪儿不对。但”无形”在古人那里是一个词,在现代要拆成两句:低于阈值的量,和绕过意识的路。第一句有实测支撑:阈下前庭刺激确实调制交感输出。第二句没有支撑,而且被 72 小时 90 dB 的双盲实验正面顶了回来。所以准确的说法不是”无形所以更凶”,而是”无形,所以更要指名道姓地说清走哪条路”——是低频声压、是热辐射不对称、是气流直吹、还是期望本身。说清通道,那几条老经验反而更硬;说不清,就只是把一个没测的量换了个更吓人的名字。

    证据与局限

    已验证
    – 阈下前庭刺激可显著调制皮肤交感神经活动并引起皮肤血管收缩,受试者无运动感(PMID 32965160)。
    – 阈下强度电前庭刺激可改变平衡受挑战条件下的姿势控制(PMID 32275736)。
    – 72 小时、约 90 dB 峰值的阈下次声暴露,双盲交叉条件下未改变睡眠、症状、心血管与神经行为指标(PMID 36946580)。
    – 症状报告可被期望操纵,且在假暴露下同样出现(PMID 23477573);给出 nocebo 解释后症状回落(PMID 25981871)。

    部分验证
    – 感知作为中介的必要性:现有研究支持”阈下有生理反应”与”症状可被期望驱动”,但尚无一项研究在同一批受试者身上同时测 D、M、Y 并做正式中介分解。本篇给出的是逻辑与间接证据,不是直接测量。
    – 心理物理阈值的稳定性:触觉检测阈值 ICC≈0.7 属可接受但不高(PMID 32300187),前庭与次声的阈值重测信度缺少同等级证据。

    待验证
    – 阈下与阈上暴露在同一暴露积分下的剂量-反应斜率是否连续,尚未见预注册的断点检验。
    – 客观检出 d′ 与自评”感觉到”在低频振动下的分离程度,缺少同一实验内的并列测量。
    – 期望操纵能否在改变症状的同时不改变 SSNA,这一”分离实验”尚未有人做——它才是判定并行通路的决定性证据。

    相关阅读

    证据等级: B
    参考文献: PMID 32965160, PMID 32275736, PMID 36946580, PMID 23477573, PMID 25981871, PMID 32300187, PMID 29283590


    免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

    Not Feeling It, Is the Body Keeping Accounts? Before Treating Perception as an Intermediary

    Evidence level: B

    The phrase “invisible sha” carries two distinct claims: one is “below the amount you can feel,” and the other is “acting directly by bypassing your sensation.” The former is measurable; the latter requires proving a causal chain, which has not yet been demonstrated.

    Main Text

    Feng Shui posits that sha qi harms people “invisibly,” followed by the inference: not feeling it does not mean no effect. Logically, this holds water—absence of sensation does not equal absence of occurrence. However, taking this step further leads to another claim: because it bypasses sensation, it is more dangerous. This leap skips the verification of causal structure. In modern methodology, this structure has a specific name: mediation.

    First, one must test the chain, not the point.

    The mediation model is written as $D \rightarrow M \rightarrow Y$: exposure $D$ changes mediator $M$, which then changes outcome $Y$. For this to hold, both arrows must exist, and $M$ must lie on the pathway, not merely move in concert. If $M$ is merely a concomitant phenomenon (where exposure simultaneously causes $M$ and $Y$, with no causal link between $M$ and $Y$), then the claim of “bypassing perception” has no counterpart—the pathway does not go through perception, so bypassing is moot. Determining this requires mediation analysis, not substituting it with “different slopes on either side of the threshold” (PMID 29283590).

    Second, the threshold line itself is ambiguous.

    Psychophysical thresholds are not physical scales but the point where detection probability reaches 50%, inherently carrying error. The retest ICC for tactile detection thresholds is approximately 0.7, while frequency discrimination thresholds are only about 0.3 (PMID 32300187)—the same person measured days apart could yield different “thresholds.” Categorizing people into above/below threshold groups using a line with error introduces misclassification into both labels. For designs comparing slopes on either side of the threshold, misclassification is fatal: it does not flatten differences; it smears them out.

    Third, there are physiological responses below the threshold, and this is true.

    One must first acknowledge positive evidence. Subjects seated on a platform undergoing extremely slow sinusoidal motion at 0.03–0.2 Hz and 0.5–5 mG, with amplitudes low enough to produce no sense of motion, still show significant vestibular modulation of skin sympathetic nerve activity (SSNA) compared to cardiac modulation, while both cardiac modulation and skin blood flow fall below baseline (PMID 32965160). Another line of evidence: setting electrical vestibular stimulation at 50% of the sinusoidal motion threshold or 80% of the skin sensation threshold still alters trunk lateral velocity during balance challenges (PMID 32275736). Below-threshold is not “no response”; it is “no report.”

    Fourth, but “having a response” does not mean “the response passes through perception.”

    This is the most critical cut of the entire piece. The two studies above prove $D \rightarrow Y$ exists but neither can nor do prove $M$ lies in between. Logically, the two are independent: pathways can run fully in parallel—one via brainstem-sympathetic reflexes, the other via cortical reporting—without intersecting. The only way to judge is whether $Y$ changes when $M$ changes while $D$ remains constant. Existing experiments have precisely done this: showing a warning about “ultrasound causing illness” increased symptoms in a high-expectation group under both real and fake ultrasound (PMID 23477573); conversely, explaining nocebo mechanisms to those already reporting symptoms caused symptoms and emotional states to revert to baseline while exposure conditions remained unchanged (PMID 25981871). $M$ (expectation and perception) moved, $Y$ moved, but $D$ did not—this precisely indicates $Y$’s change can occur without $D$, meaning $M$ is not a necessary gate on the $D \rightarrow Y$ chain.

    Fifth, the strongest counter-evidence is direct refutation.

    Thirty-seven healthy adults sensitive to noise lived in a soundproof sleep laboratory for 72 hours, exposed to 1.6–20 Hz at approximately 90 dB peak (simulating infrasound signatures from fans), in a double-blind three-arm crossover design with washout exceeding 10 days. The primary outcome, wake after sleep onset (WASO), showed no difference between real and sham exposure: −1.36 minutes (95% CI −6.60~3.88, p=0.60); as a positive control, traffic noise extended WASO by 6.07 minutes. Subjective symptoms, cardiovascular, and neurobehavioral measures showed no worsening for infrasound (PMID 36946580). With dosing this high, there is still no effect; “more dangerous below threshold” loses its footing.

    Sixth, there are two pitfalls methodologically.

    First, difference-in-differences regression requires a known breakpoint with no measurement error; if the threshold has error, the breakpoint is unidentified under the null hypothesis, and “continuity” becomes the default result rather than a finding. Second, estimation of indirect effects is inherently unstable: when the mediator and outcome have opposing effects (suppression) or weak effects, the proportion mediated can fall outside [0,1] or flip signs; reporting “mediated 43%” is meaningless. These two factors dictate that such a test must be preregistered regarding the breakpoint and mediation model, not discovered post-hoc.

    Seventh, what a decisive experiment looks like.

    Three arms: above threshold, below threshold (same exposure integral), and zero exposure. The mediator variable must use objectively detected d’ (discriminability under signal detection theory), not self-report of “did I feel it?”; outcomes should simultaneously collect SSNA/HRV and symptom scales; use a counterfactual framework to decompose natural indirect effects. The criterion is clear: if $D \rightarrow Y$ is significant in the below-threshold arm but $D \rightarrow M$ is not, perception is not a mediator and pathways run in parallel; if $M$ fully mediates, the below-threshold arm should not have an effect—once it does, the mediation assumption becomes self-contradictory. Both paths point to the same conclusion: whether perception is a mediator is a question that data can adjudicate, not one that should stall on wording.

    Yijing Studies Implications

    Feng Shui speaks of “invisible,” and the phenomena observed by ancients are real—some rooms cause discomfort while people cannot pinpoint exactly what is wrong. However, “invisible” in ancient texts is a single word; in modern terms, it must be split into two sentences: quantity below the threshold, and a path bypassing consciousness. The first has empirical support: subthreshold vestibular stimulation does indeed modulate sympathetic output. The second lacks support and is directly contradicted by the 72-hour, 90 dB double-blind experiment mentioned above. Therefore, the accurate statement is not “invisible therefore more dangerous,” but rather “invisible, therefore we must explicitly name the pathway”—is it low-frequency sound pressure, asymmetric thermal radiation, direct airflow, or expectation itself? Clarifying the channel makes old experiences harder to dismiss; leaving it unclear merely renames an unmeasured quantity with a more frightening label.

    Evidence and Limitations

    Verified
    – Subthreshold vestibular stimulation significantly modulates skin sympathetic nerve activity and causes skin vasoconstriction without motion sensation (PMID 32965160).
    – Subthreshold intensity electrical vestibular stimulation alters postural control under balance challenge conditions (PMID 32275736).
    – 72-hour subthreshold infrasound exposure at ~90 dB peak under double-blind crossover conditions did not alter sleep, symptoms, cardiovascular, or neurobehavioral indicators (PMID 36946580).
    – Symptom reports can be manipulated by expectation and occur under sham exposure (PMID 23477573); symptoms revert upon providing nocebo explanations (PMID 25981871).

    Partially Verified
    – Necessity of perception as a mediator: existing research supports “subthreshold physiological responses” and “symptoms driven by expectation,” but no study simultaneously measures $D$, $M$, and $Y$ in the same subjects to perform formal mediation decomposition. This paper provides logical and indirect evidence, not direct measurement.
    – Stability of psychophysical thresholds: tactile detection threshold ICC≈0.7 is acceptable but not high (PMID 32300187); retest reliability for vestibular and infrasound thresholds lacks evidence at the same level.

    To Be Verified
    – Whether dose-response slopes are continuous under same exposure integrals for subthreshold vs. above-threshold exposures; no preregistered breakpoint tests exist yet.
    – The degree of separation between objectively detected $d’$ and self-report “feeling” under low-frequency vibration; lack of concurrent measurement within the same experiment.
    – Whether expectation manipulation can alter symptoms without altering SSNA; this “disentangling experiment” has not been performed—it is the decisive evidence for judging parallel pathways.

    Related Reading

    Evidence Level: B
    References: PMID 32965160, PMID 32275736, PMID 36946580, PMID 23477573, PMID 25981871, PMID 32300187, PMID 29283590


    Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.

  • 煞是一样的,灾不一样:把「体质差异」放进交互项之前

    Evidence level: B

    同一间屋子、同一个位置,有人难受有人没事——这个观察可能是真的,但”谁敏感”这件事能被测到什么程度,决定了它能不能被当成结论。

    正文

    风水与命理里有一句几乎无法反驳的话:煞是一样的,灾不一样。同一栋楼、同一个朝向、同一间卧室,有人睡得香,有人整夜翻。传统给的解释是个体化分层——命卦、八字、体质。现代流行病学给这类说法的名字叫效应修饰:暴露的效应量随第三个变量变化。名字对得上,但”对得上”和”测得出来”是两个问题。本篇只处理后者。

    第一,交互项是一个乘积,不是加强版的主效应。

    模型写成 E[Y]=β₀+β₁D+β₂S+β₃(D×S),β₃ 检验的是两条斜率之差,而每条斜率自己已经带着误差。这意味着 β₃ 的标准误几乎总是大于 β₁ 或 β₂,即使真存在效应修饰。样本量上的代价是硬的:检出交互通常要数倍于检出主效应的样本,而且不同算法在同一情景下算出来的数还能差出很多(PMID 10206617)。这不是保守,是算术。

    第二,分层用的那把尺,得先问它准不准。

    MSSQ(晕动易感性问卷)是这条线最常用的分层工具。作者称其重测信度可假定优于 0.8,但预测实验室晕动耐受的有效性只有 r≈0.45(PMID 10052582)——解释约两成方差。用一把只解释两成方差的尺去分组,组内混着的错分,比组间那点差别还大。

    第三,错分打在交互项上的方向,和打在主效应上不一样。

    非差异性错分通常把乘法交互的估计拉向零(看起来”没有修饰”),但对加法交互,它可能把估计推向远离零的方向(PMID 10613335)。同一个数据集,两种尺度,两个相反的偏倚方向。这就是为什么”我们没发现敏感度修饰效应”这句话本身信息量很低——它可能只是尺子钝。

    第四,二分还是连续,会改变结论本身。

    把 MSSQ 切成”易感/非易感”最省事,也最伤:切点任意、组内异质、高分区样本天然稀疏。所谓”两层斜率不同”,其实是”组均值差”的再包装。更稳的做法是让敏感度留在连续尺度上直接估 D×S 的斜率——但这又绕回第一层的功效问题。两条路都不轻松,选哪条得写清。

    第五,问卷测的可能不是前庭。

    MSSQ 与过去 12 个月非运动性恶心呕吐的相关是 r=0.3,其中偏头痛是最大贡献者(PMID 10052582)。这说明它测的更像是”最终共同呕吐通路的兴奋性”,一个比前庭宽得多的东西。另一条路给出更窄的答案:前庭感知阈值在个体内高度一致、个体间差异显著,作者直接称其为可作生物标志的先天特质(PMID 38658516);在驾驶模拟数据上,两个个体参数就能把晕动反应拟合改善 1.7 倍(PMID 40589464)。能用两个数描述的个体差异,不必用一张回忆问卷去代理。

    第六,症状这一端也不老实。

    sopite 综合征(嗜睡、倦怠、易怒)在定义上就不特异,2014 年的修订版专门强调它”症状非特异、与个体健康状态和运动刺激的存在都有关”(PMID 24919391)。而完全不可感知的低频正弦运动确实能调制皮肤交感神经活动、降低皮肤血流(PMID 32965160)——生理端有信号,主观端却是一团模糊词。拿模糊词做结局,交互项只会更模糊。

    第七,判决。

    要做这件事,五条硬要求:① 样本量按交互算,不按主效应算;② 敏感度留在连续尺度,预注册切点只进敏感性分析;③ 同时报乘法与加法两个尺度(含 RERI),并写明哪个是主判据(PMID 38562994);④ 分层工具优先用实测阈值而非回忆问卷;⑤ 无重复验证前,结论只能写”探索性”。第五层还留了个更狠的问题:如果阈值真是稳定特质,它该被当作协变量校正掉,而不是当作修饰因子去切人群——这两条路对同一份数据会给出不同结论,而选哪条本身没有数据能替你决定。

    易学落点。

    “煞同而灾异”这个观察,古人是看对的。但古人从”不一样”直接推到”我知道你为什么不一样”,中间少了一步:先证明你分得开。命卦、八字、体质,本质上都是在给个体差异起名字;起名字不是测量。现代方法学最不客气的一句话是:想分层,得先有一把解释得动方差的尺,再有一份算得动交互的样本。两样都没有的时候,”某人命里怕这个煞”不是结论,是一个还没被检验的假设——而且是一个功效上很难检验的假设。

    这不是把古法否掉,是把它从一个可以随口说的判断,还原成一个要花很大代价才能说的判断。

    证据与局限

    已验证:交互项检验的功效代价及不同算法间的分歧(PMID 10206617);MSSQ 的预测有效性 r≈0.45 及其与偏头痛等非运动性恶心的相关(PMID 10052582);非差异性错分对乘法/加法交互的相反偏倚方向(PMID 10613335);前庭感知阈值的个体内高一致性(PMID 38658516);两个个体参数可将晕动反应拟合改善 1.7 倍(PMID 40589464);不可感知低频运动对皮肤交感神经活动的调制(PMID 32965160)。

    部分验证:把 MSSQ 分数当作稳定的”体质”代理——信度支持,效度只部分支持;sopite 症状的非特异性(PMID 24919391)意味着主观结局的测量误差方向不明,结论只能弱化。

    待验证:4-16 Hz 建筑振动这条暴露上,”敏感度×暴露”的交互斜率能否检出、需要多大样本;实测前庭阈值与问卷分数在同一人群中的一致性;把阈值当协变量校正与当修饰因子分层,两种建模对结论的实际翻转幅度。这三项都还没有人做,本篇的判决是设计建议,不是实测结果。

    相关阅读

    证据等级: B
    参考文献: PMID 10206617, PMID 10052582, PMID 10613335, PMID 38658516, PMID 40589464, PMID 24919391, PMID 32965160, PMID 38562994


    免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

    Same Cause, Different Effects: Including ‘Physiological Differences’ as an Interaction Term

    Evidence level: B

    In the same room, at the same location, one person feels ill while another remains unaffected—this observation may be true, but the degree to which “who is sensitive” can be measured determines whether it can stand as a conclusion.

    Main Text

    In Feng Shui and Yijing Studies, there is an almost irrefutable saying: Same Sha, Different Calamity. In the same building, same orientation, same bedroom, one sleeps soundly while another tosses and turns throughout the night. Traditional explanations attribute this to individualized stratification—Life Hexagrams, Bazi (Four Pillars), and constitution. Modern epidemiology names this phenomenon effect modification: the effect of exposure varies according to a third variable. The terminology aligns, but “aligning” and “measuring it out” are two distinct problems. This piece addresses only the latter.

    First, an interaction term is a product, not an amplified main effect.

    The model is written as $E[Y]=\beta_0+\beta_1D+\beta_2S+\beta_3(D\times S)$. $\beta_3$ tests the difference between two slopes, while each slope already carries its own error. This means the standard error of $\beta_3$ is almost always greater than that of $\beta_1$ or $\beta_2$, even if true effect modification exists. The cost in sample size is rigid: detecting an interaction typically requires several times the sample size needed to detect a main effect, and different algorithms can yield vastly different results in the same scenario (PMID 10206617). This is not conservatism; it is arithmetic.

    Second, the ruler used for stratification must first be questioned for its accuracy.

    The MSSQ (Motion Sickness Susceptibility Questionnaire) is the most common stratification tool in this line. Authors claim its test-retest reliability can be assumed to exceed 0.8, but its predictive validity for laboratory motion sickness tolerance is only $r\approx0.45$ (PMID 10052582)—accounting for roughly 20% of the variance. Using a ruler that explains only 20% of the variance to group subjects results in misclassification within groups that exceeds the differences between groups.

    Third, the direction of misclassification applied to the interaction term differs from that applied to the main effect.

    Non-differential misclassification typically pulls the estimate of a multiplicative interaction toward zero (appearing “no modification”), but for an additive interaction, it may push the estimate away from zero (PMID 10613335). The same dataset, two scales, two opposing bias directions. This is why the statement “we did not find a sensitivity modification effect” carries very low information—it may simply mean the ruler is dull.

    Fourth, binary versus continuous scales alter the conclusion itself.

    Cutting the MSSQ into “susceptible/non-susceptible” is the most convenient yet most damaging approach: arbitrary cut-points, heterogeneity within groups, and naturally sparse samples in high-score regions. The claim of “two different slopes” is merely a re-packaging of “group mean differences.” A more stable approach is to keep sensitivity on a continuous scale to directly estimate the $D\times S$ slope—but this circles back to the power issue raised in the first layer. Neither path is easy; whichever is chosen must be clearly stated.

    Fifth, the questionnaire may not measure the vestibular system.

    The correlation between MSSQ and non-exertional nausea/vomiting over the past 12 months is $r=0.3$, with migraine being the largest contributor (PMID 10052582). This suggests it measures more like “excitability of the final common vomiting pathway,” something far broader than the vestibular system. Another path offers a narrower answer: vestibular perception thresholds show high consistency within individuals and significant differences between individuals; authors directly refer to it as an innate trait suitable as a biological marker (PMID 38658516). In driving simulation data, two individual parameters can improve the fit of motion sickness responses by 1.7 times (PMID 40589464). Individual differences describable by two numbers need not be proxied by a single recall questionnaire.

    Sixth, the symptom side is not straightforward either.

    Sopite syndrome (sleepiness, fatigue, irritability) is inherently non-specific by definition; the 2014 revision specifically emphasizes its “non-specific symptoms, related to individual health status and the presence of motion stimulation” (PMID 24919391). Meanwhile, imperceptible low-frequency sinusoidal motion can indeed modulate cutaneous sympathetic nerve activity and reduce skin blood flow (PMID 32965160)—there is a signal on the physiological end, but the subjective end remains a blur of vague terms. Using vague terms as an outcome only makes the interaction term more ambiguous.

    Seventh, the verdict.

    To conduct this work, five hard requirements apply: ① Sample size calculated for interaction, not main effects; ② Keep sensitivity on a continuous scale, with pre-registered cut-points entering only sensitivity analyses; ③ Report both multiplicative and additive scales (including RERI), and explicitly state which is the primary criterion (PMID 38562994); ④ Prioritize measured thresholds over recall questionnaires for stratification tools; ⑤ Before repeated validation, conclusions can only be written as “exploratory.” Layer five also leaves a harsher question: if the threshold is truly a stable trait, should it be treated as a covariate to adjust for, or as a modifier factor to stratify the population? These two paths yield different conclusions on the same data, and no data exists to decide which path to take.

    Yijing Studies Landing Point.

    The observation that “Sha is the same but calamity differs” is correct in the eyes of ancients. However, ancients jumped directly from “different” to “I know why you are different,” skipping a crucial step: first prove you can distinguish them. Life Hexagrams, Bazi, and constitution are essentially just naming individual differences; naming is not measuring. The most unyielding statement from modern methodology is: to stratify, one must first have a ruler capable of explaining variance and a sample size capable of calculating interactions. When neither exists, “someone is afraid of this Sha in their fate” is not a conclusion but an untested hypothesis—one that is difficult to test regarding statistical power.

    This does not negate ancient methods but restores them from casual judgments to conclusions requiring significant cost.

    Evidence and Limitations

    Verified: The power cost of interaction testing and divergence between different algorithms (PMID 10206617); MSSQ predictive validity $r\approx0.45$ and its correlation with non-exertional nausea/migraine (PMID 10052582); opposite bias directions of non-differential misclassification on multiplicative/additive interactions (PMID 10613335); high within-individual consistency of vestibular perception thresholds (PMID 38658516); two individual parameters improving motion sickness response fit by 1.7 times (PMID 40589464); modulation of cutaneous sympathetic nerve activity by imperceptible low-frequency motion (PMID 32965160).

    Partially Verified: Treating MSSQ scores as a stable proxy for “constitution”—reliability supported, validity only partially supported; the non-specificity of sopite symptoms (PMID 24919391) implies unclear measurement error direction for subjective outcomes, weakening conclusions.

    To Be Verified: Whether the “sensitivity × exposure” interaction slope can be detected on the 4–16 Hz building vibration exposure and what sample size is required; consistency between measured vestibular thresholds and questionnaire scores in the same population; the actual magnitude of conclusion reversal when treating thresholds as covariates versus stratifying as modifier factors. None of these have been done yet; the verdict in this piece is a design recommendation, not an empirical result.

    Related Reading

    Evidence Level: B
    References: PMID 10206617, PMID 10052582, PMID 10613335, PMID 38658516, PMID 40589464, PMID 24919391, PMID 32965160, PMID 38562994


    Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.

  • 空屋子量出来的 C-A 差,不是有人时那个数

    Evidence level: B

    上一轮裁决了 20 dB 是断点还是斜坡;这一轮退回更早一步——那个被减出来的差值本身,是不是同一个量。

    正文

    办公楼测频谱有个习惯:趁没人时测。安静、不打扰、可重复。习惯里藏着一个从没被量化的假设:空屋的 C-A 差,和有人时那个数,差多少。

    一、这不是偏差,是两个量

    C-A 差 = dBC − dBA。A 计权在 1 kHz 增益为零,到 125 Hz 已衰减 16 dB;C 计权几乎全程不衰减。这把尺量的就是低频占比:中高频被 A 计权留着,低频只有 C 记得住。

    人走之后的变化,恰好落在这条分界线上。人主要吸中高频——座椅、衣物、体表是多孔吸声体,对 500-2000 Hz 有效,对 50 Hz 以下近乎无效。空屋时中高频反射多、dBA 偏高,dBC 几乎不动。分子不动、分母变大,C-A 差被压低。 方向推得出来,量级没人测过。

    二、观众吸声这张表,低频站不住

    有人会说查表:人均吸声系数乘人数,一减就校正了。这张表在低频不能用。

    PMID 11757930 用四种方法算六个厅的满场混响,结果在 125 与 250 Hz 出现大分歧。归因两条:混响室与实际厅堂扩散度不同;人并排坐时对掠入射声的吸声机制根本不清楚。PMID 17004464 又加了条件:吸声系数只能来自尺寸、形状、吸声区位、混响时间都相似的空间;低混响下 Sabine 与 Eyring 系数比值可超 2 倍。

    拿混响室的人均吸声校准常规办公室,误差先落在低频——而那正是 C-A 差唯一在意的频段。

    三、人不是吸声体,人还是声源

    “两状态相减”最致命处在这。

    PMID 30669442 调查 1078 名员工:无关语音在开放区比共享办公室更显著地抬高烦恼、降低绩效、增加心理症状;人数超过 20 人时烦恼概率进一步上升;无声学处理的空间更糟;超过 70% 的人表示在灯光反馈提醒下愿降低说话音量。

    从空置走到满员,同一时刻发生两件方向相反的事:吸声增加让 dBA 降,发声增加让 dBA 升。 两项都随人数增长。压成一个校正系数,等于把两个机制倒进一个数里——这个数还随人数、工作性质、有无监测提醒而变。

    四、”空置”不是一个状态,是一个区间

    空置有多种。硬地面硬墙无软包椅是一种;铺地毯、满软包椅、纸箱资料堆积是另一种。

    量级参照:同样容积与满场混响要求下,椅子从中度软包换重度软包,厅容积需约增加 10%(PMID 17004464)。“椅子软不软”这个与人在不在无关的属性,影响量级和”人在不在”是同一档。

    所以校正系数不能是常数,只能是房间既有吸声量的函数。而这个量在低频恰好没有可靠测法。

    五、A 计权把污染放大到一比一

    人的吸声落在 500-2000 Hz,两个计权在此都接近零增益。”中高频被吸掉 3 dB”,在 dBA 上几乎就是 3 dB、在 dBC 上几乎为零——C-A 差直接跳 3 dB。

    3 dB 在烦恼判定上不是小数。PMID 33925579 用卧室实测得出,烦恼既随等效声级上升也随事件数上升,两者都要进模型;多源合成时带交互项的感知模型才预测得准(PMID 26197326、15376661)。3-5 dB 的测量偏移,足以让一栋楼从”不超标”翻到”超标”,或从”无煞”翻到”有煞”。

    六、两状态之差,是混杂的总和

    两个状态在时间上也分不开。满员时段就是工作日白天——也是 HVAC 高负荷、外窗启闭、设备全开的时段。而 HVAC 本身就是低频主源:16 或 31.5 Hz 倍频程超 65 dB 时,轻质办公结构可能起可感振动(PMID 22918142)。

    于是”空置 vs 满员”之差里同时装着:人的吸声、人的发声、HVAC 负荷、外窗状态、设备运行。四项全锁在”上班时间”这一个变量上。这叫完全混杂,不叫可校正偏差。PMID 32357581 的系统综述把暴露误分类列为烦恼研究主要偏倚之一——这条线不该自己再添一个。

    七、不要校正系数,要状态分层

    审判三条。一,报告必须写清测量状态:空置/半员/满员,人数、地材、椅面软包、有无声学处理;缺这四项的 C-A 差不可跨研究比较。二,C-A 差只在同状态内比较:楼与楼比,要么都在空置下测,要么都在满员下测;跨状态不是校正问题,是换了尺子。三,若必须跨状态,用物理量推:以人的等效吸声面积为中介,同房间同测点同一天测空与满两态,输出”每增 10 人 C-A 差降多少”的斜率及其房间间变异。

    判决实验一句话:同楼宇同测点同一天,上班前与上班中各三次,中间不换设备不换位置;另加”无人但有等效声源”第三态,把吸声项与发声项拆开算。

    风水有句”屋大人少,是为凶宅”,今天被当成心理感受,其实讲的是可测的东西:人和屋子之间的比例。

    物理判决给这句话补了两个条件。一,”人气”不是一个量而是两个方向相反的量——人在场既吸声又发声,吸的和发的还不是同一频段,所以既不能说人多了更吵,也不能说人多了更静,要看问哪一段。二,”人少”的对面不是”人多”,是”这屋子本来吸不吸声”:铺满地毯软椅的小房间,人多人少的差别可能还小于换不换椅子。

    所以这句话古人观察对了(空屋和满屋确实是两种屋子),但算成了一道加减题。现代测量说:这不是加减题,是分离题——先把两个机制分开,再谈吉凶。

    证据与局限

    已验证: 人(衣物与体表)主要吸收 500-2000 Hz、对 50 Hz 以下近乎无效,故空置测量会系统性压低 C-A 差;观众吸声的四种测定方法在 125/250 Hz 分歧显著,ISO 354 混响室法在低频最不可靠(PMID 11757930);吸声系数只能在同一尺寸/形状/吸声区位/混响时间的空间间借用,Sabine 与 Eyring 系数在低混响下比值可超 2 倍(PMID 17004464);无关语音的烦恼随办公人数上升、>20 人时加剧、>70% 愿在监测提醒下降音量(PMID 30669442);HVAC 在 16/31.5 Hz 超 65 dB 时轻质结构可能起振(PMID 22918142);烦恼同时受等效声级与事件数驱动,多源需交互项感知模型(PMID 33925579、26197326、15376661);暴露误分类是烦恼研究公认主要偏倚(PMID 32357581)。

    部分验证: “中高频抬升 500-2000 Hz 应为多少 dB”没有一个可直接引用的现场数字——现有材料只能给出方向(空置偏高)与量级参照(软包程度约 10% 容积当量),不能给出斜率;”占空比加权”这一设想的线性可加前提,与第三节的双向机制直接冲突,未经验证。

    待验证: 每增 10 人所对应的 C-A 差下降斜率及其房间间变异;”无人但有等效声源”第三态能否真正分离吸声项与发声项;同一楼宇内空/满两态的 HVAC 负荷与外窗状态能否被独立记录并作为协变量剥离。均属 P2 池子空白,无直接文献。

    相关阅读

    证据等级: B
    参考文献: PMID 11757930, PMID 17004464, PMID 30669442, PMID 22918142, PMID 33925579, PMID 26197326, PMID 15376661, PMID 32357581


    免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

    The C-A difference measured in an empty room is not the same as when someone is present.

    Evidence level: B

    The previous round of adjudication determined whether 20 dB represents a breakpoint or a slope; this round steps back further—specifically, is the difference itself, which is subtracted out, the same quantity.

    Main Text

    There is a habit in office building spectrum measurement: measure when no one is present. It is quiet, non-disruptive, and repeatable. Hidden within this habit is an assumption that has never been quantified: by how much does the C-A difference in an empty room differ from that value when occupied.

    I. This Is Not Deviation, But Two Quantities

    C-A difference = dBC − dBA. The A-weighting curve has zero gain at 1 kHz and attenuates by 16 dB at 125 Hz; the C-weighting curve remains nearly flat throughout. This scale measures low-frequency proportion: mid-high frequencies are retained by A-weighting, while only C-weighting remembers the low frequencies.

    The change occurring after people leave falls precisely on this dividing line. Humans primarily absorb mid-high frequencies—seats, clothing, and skin are porous absorbers effective for 500–2000 Hz but nearly ineffective below 50 Hz. In empty rooms, mid-high frequency reflection is high and dBA is elevated, while dBC remains almost unchanged. The numerator stays constant while the denominator increases, suppressing the C-A difference. The direction is deducible; the magnitude has not been measured.

    II. The Audience Absorption Table Does Not Hold at Low Frequencies

    One might argue to consult tables: multiply the average absorption coefficient by the number of people and subtract for correction. This table cannot be used at low frequencies.

    PMID 11757930 calculated full-room reverberation times for six halls using four methods, resulting in significant divergence at 125 and 250 Hz. Two attributions were identified: differences in diffusion between reverberation rooms and actual halls; the absorption mechanism for grazing incidence sound when people sit side-by-side remains fundamentally unclear. PMID 17004464 added further constraints: absorption coefficients could only be derived from spaces with similar dimensions, shapes, absorption locations, and reverberation times; under low reverberation conditions, the ratio of Sabine to Eyring coefficients could exceed 2.

    Applying the average absorption calibration from a reverberation room to a conventional office introduces errors first at low frequencies—precisely the band of interest for the C-A difference.

    III. Humans Are Not Just Absorbers, They Are Also Sound Sources

    The most critical flaw in “subtracting two states” lies here.

    PMID 30669442 surveyed 1,078 employees: irrelevant speech elevated annoyance and reduced performance more significantly in open areas than shared offices; the probability of annoyance increased further when the number of people exceeded 20; acoustically treated spaces performed worse; over 70% of respondents indicated they would lower their speaking volume under light feedback reminders.

    Moving from empty to fully occupied, two events occur simultaneously with opposite directions: increased absorption lowers dBA, while increased sound generation raises dBA. Both increase with headcount. Compressing them into a single correction coefficient equals stuffing two mechanisms into one number—a number that also varies with headcount, work nature, and the presence of monitoring reminders.

    IV. “Empty” Is Not a Single State, But a Range

    There are various definitions of “empty.” Hard floors and walls with no upholstered chairs constitute one type; carpeting, fully upholstered chairs, and stacked cardboard boxes constitute another.

    Magnitude reference: under the same volume and full-room reverberation requirements, switching from medium to heavy upholstery requires increasing hall volume by approximately 10% (PMID 17004464). The attribute of “whether chairs are soft or hard,” unrelated to occupancy, affects magnitude at the same order as “whether people are present.”

    Therefore, a correction coefficient cannot be constant; it must be a function of the room’s existing absorption. And this quantity lacks reliable measurement methods at low frequencies.

    V. A-Weighting Amplifies Pollution to a One-to-One Ratio

    Human absorption falls within 500–2000 Hz, where both weighting curves approach zero gain. “Mid-high frequencies absorbed by 3 dB” translates to nearly 3 dB in dBA and nearly zero in dBC—the C-A difference jumps directly by 3 dB.

    A 3 dB shift is not negligible in annoyance determination. PMID 33925579 found via bedroom measurements that annoyance increases with both equivalent sound level and event count; both must enter the model, and perception models with interaction terms for multi-source synthesis predict accurately (PMID 26197326, 15376661). A 3–5 dB measurement offset is sufficient to flip a building from “non-compliant” to “compliant,” or from “no Sha” to “with Sha.”

    VI. The Difference Between Two States Is a Sum of Confounding Factors

    The two states are also temporally inseparable. Full occupancy occurs during weekday daytime—also the period of high HVAC load, window operation, and full equipment operation. HVAC itself is a primary low-frequency source: when octave bands at 16 or 31.5 Hz exceed 65 dB, lightweight office structures may exhibit perceptible vibration (PMID 22918142).

    Thus, the difference between “empty vs. full” simultaneously contains: human absorption, human sound generation, HVAC load, window status, and equipment operation. All four are locked under the single variable of “working hours.” This is complete confounding, not correctable bias. PMID 32357581’s systematic review lists exposure misclassification as a major bias in annoyance studies—this line should not add another one.

    VII. Do Not Use Correction Coefficients; Use State Stratification

    Three rulings. First, reports must clearly specify measurement states: empty / partial occupancy / full occupancy, number of people, floor materials, chair upholstery, presence or absence of acoustic treatment; without these four items, C-A differences cannot be compared across studies. Second, compare C-A differences only within the same state: building-to-building comparisons must either all be measured under empty conditions or all under full occupancy; cross-state comparison is not a correction issue but changing the ruler. Third, if cross-state comparison is necessary, use physical quantities for inference: using human equivalent absorption area as a mediator, measure both empty and full states in the same room at the same day at the same points, outputting the slope of “how much C-A difference decreases per additional 10 people” along with inter-room variability.

    The experimental verdict in one sentence: Same building, same measurement point, same day, three times before work and three times during work, with no equipment or position changes in between; additionally add a third state of “no people but equivalent sound source” to separate the absorption term from the generation term.

    Feng Shui has a saying: “When the house is large but people are few, it is an凶宅 (haunted house).” Today this is treated as a psychological sensation, but actually refers to measurable things: the ratio between people and the room.

    Physics adds two conditions to this statement. First, “human energy” is not a single quantity but two quantities moving in opposite directions—people present both absorb and generate sound; the absorbed and generated sounds are not in the same frequency band, so one cannot say people make it louder nor quieter; it depends on which segment is asked. Second, the opposite of “few people” is not “many people,” but “whether this room inherently absorbs sound or not”: a small room filled with carpet and soft chairs may show less difference between many and few people than the difference between having or not having chairs.

    So the ancients observed correctly (empty rooms and full rooms are indeed two types of rooms), but calculated it as an addition/subtraction problem. Modern measurement says: this is not an addition/subtraction problem, but a separation problem—first separate the two mechanisms, then discuss auspiciousness or inauspiciousness.

    Evidence and Limitations

    Verified: Humans (clothing and skin) primarily absorb 500–2000 Hz and are nearly ineffective below 50 Hz, thus empty room measurements systematically suppress the C-A difference; four determination methods for audience absorption show significant divergence at 125/250 Hz, with ISO 354 reverberation room method being least reliable at low frequencies (PMID 11757930); absorption coefficients can only be borrowed across spaces with similar dimensions/shape/absorption location/reverberation time, and the ratio of Sabine to Eyring coefficients can exceed 2 under low reverberation (PMID 17004464); irrelevant speech annoyance increases with office headcount, worsens above 20 people, and >70% would lower volume under monitoring reminders (PMID 30669442); HVAC may induce vibration in lightweight structures at 16/31.5 Hz octave bands exceeding 65 dB (PMID 22918142); annoyance is driven simultaneously by equivalent sound level and event count, requiring interaction terms in multi-source perception models (PMID 33925579, 26197326, 15376661); exposure misclassification is a recognized major bias in annoyance studies (PMID 32357581).

    Partially Verified: There is no directly citable field number for “how many dB should mid-high frequencies between 500–2000 Hz increase”—existing materials can only provide direction (empty is higher) and magnitude reference (upholstery level approximates 10% volume equivalent), not a slope; the linear additivity premise of the “duty cycle weighting” hypothesis directly conflicts with the dual mechanism in Section III and remains unverified.

    To Be Verified: The slope of C-A difference decrease per additional 10 people and its inter-room variability; whether the third state of “no people but equivalent sound source” can truly separate the absorption term from the generation term; whether HVAC load and window status in empty vs. full states within the same building can be independently recorded and stripped as covariates. All belong to P2 pool gaps with no direct literature.

    Related Reading

    Evidence Level: B
    References: PMID 11757930, PMID 17004464, PMID 30669442, PMID 22918142, PMID 33925579, PMID 26197326, PMID 15376661, PMID 32357581


    Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.

  • 20 dB 是断点还是斜坡:低频「频谱失衡」判据的函数形式裁决

    Evidence level: B

    上一篇把 C-A 差推成候选预测指标,这一篇问一个更靠地基的问题:那个 20 dB 到底是一条线,还是一段坡。

    正文

    上一篇论证了 C-A 差(dBC 与 dBA 之差)可能比绝对声压级更能预测办公楼烦扰。但它脚下踩着一块没人验过的地基:20 dB 这个切点,究竟是物理断点,还是被反复抄写的经验整数?

    一、同一个现象,五个切点

    如果 20 dB 是物理断点,它不该是个可以商量的数。可工程界并不统一:DIN 45680 与 Alberta 监管口径取 20 dB,澳大利亚新南威尔士的工业噪声政策取 15 dB,Acoustics Australia 的综述列完全部出处后建议”至少 20 dB”,同时提到有人主张 25 dB;上海交大对住宅固定设施的研究给出 15 dB。同一物理量,切点跨度 10 dB。 这不像阈值的常态,更像判据的常态——20 dB 是一个被抄写的整数,不是被检验出的界线。

    二、两种形状,两种政策

    这是”函数形式”之争,而两个候选模型对世界的说法相反。阈值模型说:20 dB 以下无效应,越过才有——管线把楼压到 19 dB 就交差。连续梯度模型说:效应随差值平滑上升,每一分贝都算——最划算的减排恰恰在线以下。

    空气污染领域早打过这一仗。NMMAPS 在 20 个美国城市用对数线性、样条、阈值三模型对比,结论是无阈值、线性够用;法国 18 城(PMID 36602605)把四种函数形式放在同一批数据上,得到”超线性”——低浓度端更陡,中段趋平。那不是断点,是弯。 而这两种形状在同一张散点图上都能拟合得像模像样。

    三、分段回归的致命陷阱

    即便有人”检验过” 20 dB,也不等于有证据。断点在零假设下不可识别——它只在”确实存在断点”的备择假设里才有定义,常规 t 检验与 Wald 检验在此失效,必须用 Davies 检验或 Muggeo 伪分数检验校正零分布;而 segmented 类算法对初值极敏感,换个起点就收敛到另一个局部极小。没做这一步而报出的”阈值位于 X dB”,报的是一个没有零分布的 p 值。

    四、现有证据更像斜坡

    PMID 39147186 用生态瞬时评估加便携声级计追踪真实时段的烦恼,非线性拐点出现在实时 60 dB、回忆 70 dB,且随活动情境漂移——拐点位置不是声学常数,是活动类型的函数。PMID 36586852 研究飞机起飞低频声,把室外 C-A 差作第二个连续预测因子,解释方差直接到 98%,全程未引入任何断点。连续变量已经够用,切成二值只是丢信息。

    五、低频端本来就是陡的

    方向上也对不上。Møller 与 Pedersen 的综述(PMID 15273024)指出低频烦扰随声级上升得比中高频更快。若曲线在低频端是凸的,”低端平坦、越线才起效”的图像在方向上就是反的——低端不是平台,是陡坡。

    六、但”没测到”不等于”不存在”

    必须留退路。PMID 32018132 系统展示了代理暴露的测量误差如何结构性抹平低端非线性;NMMAPS 的专家委员会也专门警告:暴露测量误差可能掩盖真实阈值。所以本轮判决只有一条——20 dB 目前不可判定,而不是已被证伪。当已验证的因果阈值是越界,当已被推翻同样是越界。

    七、判决实验

    选 8-12 栋频谱形状差异大的办公楼,每栋 20-40 人;C-A 差实测分布须横跨 15-40 dB,每 5 dB 单元不少于 20 人——只在 20 dB 附近堆点的样本永远判不出来。同一批数据平行拟合四个模型:线性、先验 20 dB 分段、无先验断点分段(segmented + Davies)、限制性立方样条。结局用 ICBEN 连续评分,不二分。

    核心判据:Davies p>0.05 且样条在 20 dB 附近斜率变化不足 30% ⇒ 判连续梯度,20 dB 降为工程筛查线、不得作因果阈值;Davies p<0.05、估计断点落在 18-22 dB 且 bootstrap 区间宽小于 4 dB ⇒ 断点成立、可写进标准;断点漂到 30 dB 以外或区间宽过 10 dB ⇒ 判断点不可识别、只报连续剂量-反应;样条显示低端更陡 ⇒ 干预优先级翻转,先救低端而非先压超标楼。

    风水线上,”煞”本来就并存两套直觉:一套门槛式(”路冲不过三丈”),一套坡度式(”煞有轻重,日积月累”)。本文不站队,只把它变成能被数据裁决的统计问题——这是”化煞”从口诀走向剂量学的必经一步。

    证据与局限

    已验证: 20/15/25 dB 三个切点在同一工程领域并存,无统一实证来源;阈值模型与连续模型在污染流行病学中已被系统对撞过(PMID 36602605 四模型对比得超线性而非断点);噪声烦恼的非线性拐点随活动情境漂移(PMID 39147186);C-A 差作连续预测因子解释方差达 98%、无需断点(PMID 36586852);低频烦扰随声级上升比中高频更快(PMID 15273024)。

    部分验证: C-A 差对烦扰解释力”显著但很小”(Kjellberg 1997,经 PMID 37818147 转述)——提示其单独解释力有限,增量可能来自函数形式误设而非指标本身;断点方法学(Davies 检验、Muggeo 线性化、bootstrap 区间)已成熟,但尚无任何低频噪声研究正式套用。

    待验证: 20 dB 在多楼宇样本上是否对应真实断点;断点估计的置信区间宽度是否足以支撑一个监管数字;C-A 差与烦扰的关系究竟是超线性、阈值还是线性——目前无直接研究,属 P2 池子空白。

    相关阅读

    证据等级: B
    参考文献: PMID 36602605, PMID 39147186, PMID 36586852, PMID 15273024, PMID 32018132, PMID 37818147


    免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

    20 dB is a breakpoint or a slope: adjudicating the functional form of the criterion for low-frequency \”spectral imbalance\

    Evidence level: B

    The previous article promoted the C-A difference as a candidate predictive metric; this one asks a more foundational question: is that 20 dB figure a straight line or a segment of a slope.

    Main Text

    The previous article argued that the C-A difference (the gap between dBC and dBA) may predict office annoyance better than absolute sound pressure levels. However, it stands on unverified ground: Is the 20 dB cutoff a physical discontinuity or a repeatedly copied empirical integer?

    I. One Phenomenon, Five Cutoffs

    If 20 dB were a physical discontinuity, it should not be negotiable. Yet engineering standards diverge: DIN 45680 and Alberta regulations use 20 dB; New South Wales, Australia’s industrial noise policy uses 15 dB; Acoustics Australia’s review lists all sources and suggests “at least 20 dB,” while noting some advocate for 25 dB; Shanghai Jiao Tong University’s study on residential fixed installations yields 15 dB. The same physical quantity spans a 10 dB range for the cutoff. This resembles the norm for criteria rather than thresholds—20 dB is a copied integer, not an empirically verified boundary.

    II. Two Shapes, Two Policies

    This is a debate over “functional form,” where two candidate models offer contradictory accounts of reality. The threshold model states: no effect below 20 dB; effects only emerge after crossing the line—compliance is achieved if a building stays at 19 dB. The continuous gradient model states: effects rise smoothly with the difference; every decibel counts—the most cost-effective reduction occurs precisely below the line.

    This battle has been fought in air pollution. NMMAPS compared log-linear, spline, and threshold models across 20 US cities; the conclusion was no threshold, linear suffices. France’s study of 18 cities (PMID 36602605) fitted four functional forms to the same dataset, finding “super-linear”—steeper at low concentrations, flattening in the mid-range. That is not a cliff, but a curve. Yet both shapes fit the same scatterplot with equal plausibility.

    III. The Fatal Trap of Piecewise Regression

    Even if someone claims to have “verified” 20 dB, that does not equate to evidence. A discontinuity is unidentified under the null hypothesis—it only has definition under the alternative hypothesis that a discontinuity indeed exists; conventional t-tests and Wald tests fail here, requiring Davies testing or Muggeo pseudo-score testing to correct the null distribution; segmented algorithms are extremely sensitive to initial values, converging to different local minima with a changed starting point. Reporting a “threshold at X dB” without this step reports a p-value with no null distribution.

    IV. Existing Evidence Suggests a Slope

    PMID 39147186 used ecological instantaneous assessment with portable sound level meters to track annoyance during real-time periods; the non-linear inflection point appeared at 60 dB real-time and 70 dB recall, drifting with activity context—the inflection point is not an acoustic constant but a function of activity type. PMID 36586852 studied low-frequency noise from aircraft takeoffs, using outdoor C-A difference as a second continuous predictor, explaining 98% of variance without introducing any discontinuity. Continuous variables suffice; dichotomizing discards information.

    V. The Low-Frequency End Is Naturally Steep

    The direction also mismatches. Møller and Pedersen’s review (PMID 15273024) notes that low-frequency annoyance rises faster with sound level than mid-to-high frequencies. If the curve is convex at low frequencies, an image of “flat at the bottom, effective only after crossing” is reversed—the low end is not a plateau but a steep slope.

    VI. “Not Measured” Does Not Equal “Does Not Exist”

    A retreat must be left open. PMID 32018132 systematically demonstrates how measurement error in proxy exposure structurally flattens low-end non-linearity; the NMMAPS expert committee also specifically warns that exposure measurement error may mask true thresholds. Thus, this round’s verdict has only one rule: 20 dB is currently indeterminate, not refuted. When a verified causal threshold is crossing the line, and when a debunked one is also crossing the line.

    VII. Judgment Experiment

    Select 8–12 office buildings with large spectral shape differences, each housing 20–40 people; the measured C-A difference distribution must span 15–40 dB, with at least 20 people per 5 dB unit—samples only piling up near 20 dB will never yield a verdict. Fit four models in parallel on the same data: linear, prior 20 dB piecewise, no-prior-discontinuity piecewise (segmented + Davies), restricted cubic spline. Outcomes use ICBEN continuous scoring, not binary.

    Core criteria: If Davies p > 0.05 AND spline slope change near 20 dB is <30% ⇒ judge continuous gradient, reduce 20 dB to an engineering screening line, do not treat as causal threshold; If Davies p < 0.05, estimated discontinuity falls within 18–22 dB AND bootstrap interval width is <4 dB ⇒ discontinuity confirmed, can be written into standards; If discontinuity drifts beyond 30 dB OR interval width exceeds 10 dB ⇒ judge discontinuity unidentified, report only continuous dose-response; If spline shows steeper low-end ⇒ intervention priority flips, save the low end before pressuring over-compliant buildings.

    In Feng Shui, “Sha” inherently coexists with two intuitions: one threshold-based (“road collision not exceeding three zhang”), one gradient-based (“Sha has weight, accumulating over time”). This article does not take sides but transforms it into a statistical problem adjudicated by data—this is the necessary step for “neutralizing Sha” moving from mnemonic rules to dosimetry.

    Evidence and Limitations

    Verified: The coexistence of 20/15/25 dB cutoffs within the same engineering domain lacks unified empirical sources; threshold models and continuous models have been systematically clashed in pollution epidemiology (PMID 36602605 found super-linear rather than discontinuity); non-linear inflection points in noise annoyance drift with activity context (PMID 39147186); C-A difference as a continuous predictor explains 98% variance without needing discontinuities (PMID 36586852); low-frequency annoyance rises faster with sound level than mid-to-high frequencies (PMID 15273024).

    Partially Verified: C-A difference’s explanatory power for annoyance is “significant but small” (Kjellberg 1997, recounted via PMID 37818147)—suggesting limited standalone explanatory power, with incremental gains possibly arising from misspecified functional forms rather than the metric itself; discontinuity methodology (Davies testing, Muggeo linearization, bootstrap intervals) is mature, yet no low-frequency noise study has formally applied it.

    To Be Verified: Whether 20 dB corresponds to a true discontinuity across multi-building samples; whether the confidence interval width of discontinuity estimates is sufficient to support a regulatory number; whether the C-A difference–annoyance relationship is super-linear, threshold-based, or linear—currently no direct research exists, representing a P2 pool gap.

    Related Reading

    Evidence Level: B
    References: PMID 36602605, PMID 39147186, PMID 36586852, PMID 15273024, PMID 32018132, PMID 37818147


    Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.

  • 换床位到底管不管用:先把”位置差多少”和”人差多少”分开算

    Evidence level: B

    这个设计藏了一个默认前提:住哪儿可以当作”暴露多少”的代理。在低频上这个前提要断两次——房间里根本没有那个”位置差”,而人也不是随机分到波腹上的。

    一、这间屋子先装不装得下”位置”

    10 Hz 波长 34.3 米,四分之一波长 8.6 米。驻波要在空间上有起伏,前提是某个方向放得下半个波长,即 17.15 米。普通卧室进深三到五米,在 4-16 Hz 这个常被叫”器官共振带”的频段里整间同相——不是”波腹这头、波节那头”,而是一起涨一起落。所以卧室里分”波腹组/波节组”,两组暴露在设计上就相等。

    能装下波的房间日常还真有:宿舍、大殿、厂房、走廊尽头的值班室。这条线不是没有对象,是对象不在住宅卧室里。

    二、结局这一端,一两夜测不出来

    一项 10,412 人、370 万夜的研究:估一个人的均值,2-7 夜够(总睡眠时间 7 夜、入睡时刻 3 夜、WASO 5 夜);估他的夜间波动要 41-65 夜;用 7-14 夜算变异性可靠性很差(PMID 41761769)。更早的实验室数据更细:QEEG 各频带功率和睡眠期 HRV 一夜就稳,睡眠时长、连续性、结构一夜不稳(PMID 22942507)。组间比较若建在”睡眠效率””觉醒次数”上,先得回答这个差是不是抽样噪声。

    三、暴露这一端同样:一次测量不等于长期暴露

    UK Biobank 算过 2,858 个有重复测量的连续变量:基线体检中位 ICC 0.64,饮食回顾只有 0.35(PMID 37336529)。ICC 是”一次测量能代表长期值的比例”,0.64 意味着效应估计朝零衰减约三分之一。空气污染那边同理,代理指标替代个人暴露带来结构性误差(PMID 32018132)。低频声压的跨日 ICC 至今没人报过。

    四、更麻烦的是,床位是人挑的

    9,783 名学生的数据显示,对居住环境的偏好既预测他实际暴露在什么环境里,又混杂了暴露-健康关联(PMID 30286433)。床位同理:怕吵的往里靠,怕闷的靠窗。位置不是随机的,它跟敏感度、健康基线、住房条件绑在一起。

    还有一层更实的:29 间卧室的单盲干预,只秘密调排风扇转速(低/中/高三档),腕表睡眠就有差别(PMID 37142023)。通风口与床位强相关——位置效应里混着风、温度、光。

    五、二分法会把剩下那点差别也切掉

    连续变量二分,损失信息与功效(PMID 9598427);整群设计里还会压低 ICC,两个效应方向相反(PMID 22733454)。混杂更糟:同一批模拟数据,年龄按连续变量进模型能控住混杂,分成组就出偏倚,且切点离中位数越远偏倚越大(PMID 17066378)。”波腹组/波节组”正是这种二分。

    六、剂量用绝对值还是相对值,结论会不同

    飞机噪声唤醒率研究给了个反直觉结果:用相对声级预测明显优于绝对室内声级,绝对声级在很多区间预测的唤醒率与零无显著差别(PMID 24180775)。低频更甚——室内 4-16 Hz 的绝对值里有一大块是全城共有的本底,真正跟”这个房间、这个位置”有关的是它比本底高出多少。

    七、还有:人到底在不在那个点上

    近高速路研究把”位置误差”和”时间-活动模式”一起算:地理编码误差,加上人在五个微环境各自待多久,共同决定指派误差(PMID 24010639)。一张床一晚 7 小时,房间 9 小时,剩下 15 小时在别处。

    这样设计才有意义

    换抽样框:进深超 17 米的空间才谈得上”位置”,报”相对本房间本底超出多少 dB”,不做波腹/波节二分。两端都先报可靠性:暴露侧报跨日 ICC,结局侧报 iSD 所需夜数(41-65 夜)或改用多夜聚合均值。最强设计不是”波腹组 vs 波节组”,是同一人换床位:位置、通风、温度、光照一起搬动,自身对照把”人”这个最大方差源消掉。

    易学落点

    风水”安床”是一门很具体的学问:床头不朝门、不压梁、不贴湿地。它讲的正是同一间屋子里,位置不同。

    物理判决是:在低频这一段,普通卧室里这个”位置差”根本不存在——不是没测到,是房间装不下波。真正随床位变的是通风、温度、光照、噪声源距离这些热学和中高频的量。“床位有讲究”在物理上站得住,但它站的是热环境和日常噪声的台,不是次声的台。

    这不是把古法否掉,是把它从”哪条通道都算”改成”说清走哪条通道”。说清之后那几条经验反而更硬:床头贴外墙=冷辐射加窗缝风,正对门=气流直吹加走廊噪声。

    证据与局限

    已验证:房间 4-16 Hz 的”位置差异”受波长限制,属声学几何硬约束;睡眠夜间变异与测量可靠性、连续变量二分的功效与偏倚损失、床位与通风/温度/光照的空间相关,均有直接文献支撑。

    部分验证:低频暴露的跨日 ICC、居住者按敏感度自选位置的程度、”相对声级优于绝对声级”在 4-16 Hz 是否成立——均为合理外推,缺直接实测。

    待验证:同一人换床位能否把位置效应从人与房间的方差里分离出来;大空间里是否存在足以支撑组间比较的暴露梯度。

    相关阅读

    证据等级: B
    参考文献: PMID 41761769, PMID 22942507, PMID 37336529, PMID 32018132, PMID 30286433, PMID 37142023, PMID 9598427, PMID 22733454, PMID 17066378, PMID 24180775, PMID 24010639


    免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

    Does Changing Beds Actually Work: Separating the Calculations of \”Position Difference\” and \”Personnel Difference\

    Evidence level: B

    This design relies on a default assumption: where one lives can serve as a proxy for “how much is exposed.” This premise must be rejected twice at low frequencies: first, there is no such “position difference” within a room; second, humans are not randomly distributed across antinodes.

    I. Does This Room Physically Accommodate “Position”?

    At 10 Hz, the wavelength is 34.3 meters; a quarter-wavelength is 8.6 meters. Standing waves require spatial variation, which presupposes that space accommodates at least half a wavelength (17.15 meters) in a specific direction. Typical bedroom depths range from 3 to 5 meters; within the 4–16 Hz band often termed the “organ resonance band,” the entire room oscillates in phase—it is not “antinodes here, nodes there,” but rather rising and falling together. Consequently, within a bedroom, “antinode groups” and “node groups” are equally exposed by design.

    Rooms that do accommodate waves in daily life include dormitories, large halls, factories, and duty rooms at the end of corridors. This line of inquiry is not without subjects; however, those subjects do not reside in residential bedrooms.

    II. Outcomes Cannot Be Determined in Just One or Two Nights

    A study involving 10,412 individuals and 3.7 million nights found that estimating an individual’s mean sleep requires only 2–7 nights (total sleep time over 7 nights, sleep onset over 3 nights, WASO over 5 nights); however, estimating his nightly variability requires 41–65 nights. Reliability based on 7–14 nights is poor (PMID 41761769). Earlier laboratory data provide finer detail: QEEG power across frequency bands and sleep-period HRV stabilize within one night, whereas sleep duration, continuity, and structure remain unstable overnight (PMID 22942507). If between-group comparisons are constructed based on “sleep efficiency” or “awakening counts,” one must first determine whether the difference is sampling noise.

    III. Exposure Similarly: One Measurement Does Not Equal Long-Term Exposure

    UK Biobank analyzed 2,858 continuous variables with repeated measurements: the baseline physical examination had a median ICC of 0.64, while dietary recall yielded only 0.35 (PMID 37336529). ICC represents the proportion of variance in a single measurement attributable to long-term values; an ICC of 0.64 implies that effect estimates attenuate toward zero by approximately one-third. Similar structural errors arise when using proxy indicators for personal exposure in air pollution studies (PMID 32018132). No one has yet reported the cross-day ICC for low-frequency sound pressure levels.

    IV. The More Complicated Issue Is That Beds Are Chosen by People

    Data from 9,783 students show that preferences for the living environment both predict actual environmental exposure and confound exposure-health associations (PMID 30286433). Beds follow the same logic: those fearing noise retreat inward, while those fearing stuffiness prefer windows. Position is not random; it correlates with sensitivity, health baselines, and housing conditions.

    There is also a more tangible layer: in a single-blind intervention across 29 bedrooms, only the fan speed was secretly adjusted (low/medium/high), yet wrist-worn sleep metrics differed significantly (PMID 37142023). Ventilation outlets correlate strongly with bed placement—position effects are confounded by airflow, temperature, and light.

    V. Binary Classification Would Also Cut Off Remaining Differences

    Binary classification of continuous variables loses information and statistical power (PMID 9598427); in cluster designs, it also suppresses ICC while reversing effect directions (PMID 22733454). Confounding worsens this: in the same simulated dataset, modeling age as a continuous variable controls for confounding, whereas grouping introduces bias, with bias increasing as the cutoff moves further from the median (PMID 17066378). The “antinode vs. node” grouping is precisely this type of binary split.

    VI. Whether Absolute or Relative Dose Determines Conclusions

    An airplane noise study on arousal rates provided a counterintuitive result: predicting arousal using relative sound levels performed significantly better than absolute indoor sound levels, as absolute levels showed no significant difference from zero across many intervals (PMID 24180775). Low frequencies are even more pronounced—within the absolute values of 4–16 Hz indoors, a large portion represents city-wide background noise. What truly relates to “this room, this position” is how much it exceeds that background.

    VII. Furthermore: Is the Person Actually at That Point?

    Studies near highways calculated “position error” alongside “time-activity patterns”: geographic coding errors, combined with the duration individuals spend in each of five microenvironments, jointly determine assignment error (PMID 24010639). One bed per night covers 7 hours; the room occupies 9 hours; the remaining 15 hours are spent elsewhere.

    Why This Design Matters

    Change the Sampling Frame: Only spaces deeper than 17 meters can meaningfully discuss “position.” Report “how many dB above the room’s own background,” avoiding binary antinode/node splits. Report reliability for both sides first: exposure side reports cross-day ICC; outcome side reports the number of nights required for iSD (41–65 nights) or adopt multi-night pooled means. The strongest design is not “antinode group vs. node group,” but swapping beds for the same person: moving position, ventilation, temperature, and light simultaneously eliminates “person” as the largest variance source via self-control.

    Yijing Application Points

    Feng Shui “bed placement” is a very specific discipline: the head of the bed should not face the door, rest against a beam, or be placed near damp areas. It addresses precisely differences in position within the same room.

    The physical verdict is that, at low frequencies, ordinary bedrooms do not possess this “position difference”—it is not undetected, but the room cannot accommodate the wave. What varies with bed placement are thermal and mid-high frequency quantities such as ventilation, temperature, lighting, and distance to noise sources. “Beds have讲究” (have significance) stands physically on the platform of thermal environments and daily noise, not infrasound.

    This does not negate ancient methods but clarifies which channel applies. Once clarified, these empirical rules become even stronger: head against an exterior wall equals cold radiation plus window gap drafts; facing the door equals direct airflow plus corridor noise.

    Evidence and Limitations

    Verified: The “position difference” in rooms at 4–16 Hz is constrained by wavelength, representing a hard acoustic geometric constraint; sleep-nightly variability, measurement reliability, loss of power and bias from binary classification of continuous variables, and spatial correlations between beds and ventilation/temperature/light all have direct literature support.

    Partially Verified: Cross-day ICC for low-frequency exposure, the degree to which residents self-select positions based on sensitivity, and whether “relative sound levels outperform absolute sound levels” holds at 4–16 Hz—all are reasonable extrapolations lacking direct measurement.

    To Be Verified: Whether swapping beds for the same person can separate position effects from person-room variance; whether sufficient exposure gradients exist in large spaces to support between-group comparisons.

    Related Reading

    Evidence Level: B
    References: PMID 41761769, PMID 22942507, PMID 37336529, PMID 32018132, PMID 30286433, PMID 37142023, PMID 9598427, PMID 22733454, PMID 17066378, PMID 24180775, PMID 24010639


    Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.

  • 仿真说这里最响:可那个「这里」到底有多大?

    Evidence level: B

    驻波热点不是地上一个圆圈,是一整片同相区。判仿真准不准之前,得先说清”位置”这个词在这个频率下还成不成立。

    正文

    第 486 号课题的设想是:给目标房间建声学模型,算出模态频率表和空间声压分布,再用传声器网格实测,比对波腹/波节位置,看吻合不吻合。

    这个设计里藏着两个默认:一是”位置”是明确的东西,二是”吻合”是可以算的。前一个在低频先塌,后一个在方法上塌。

    第一层:低频的”点位”不成立,网格越密越假

    声压场的空间分辨率由波长决定,一个相位翻转只需 λ/4。10 Hz 的波长是 34.3 米,λ/4 就是 8.6 米——5 米见方的房间里任意两点都小于这个数,全场基本同相。网格铺 20 厘米还是 2 米,测到的是同一个数,不是”更精确”。更麻烦的是场并不被动:同一纯音声源在混响室里换个位置,辐射声功率就显著不同(PMID 19640033)。”放大多少”从来不是只由房间决定的数。

    第二层:模态算术把这场实验画在了 17 米以外

    f₁ = c/2L。10 Hz 要有第一阶模态,最长边得 17.15 米;4 Hz 要 42.9 米。普通住宅短边 5 米对应 34.3 Hz,在次声带之上两倍。所以比对对象不是”任意房间”,而是长走廊、大厅、厂房、堂屋这类进深过 17 米的空间。在 3 米见方的卧室里做,测到的不是”模态分布”,是整间屋子一起呼吸的刚性压缩——那时全场同相,波腹命中率必然 100%,而 100% 什么也没验证。这不是否掉实验,是把对象收窄到能成立的那一段。

    第三层:边界阻抗是唯一的真未知,而它只能从场里反推

    低频吸声系数测不准是公开的老问题:混响室在 315 Hz 以下模态密度不足、扩散场失效,得另找算法(PMID 41711454 改用阻尼密度法);墙面阻抗在模态频率上只能靠特征值近似反推,论文自己写明”低频缺乏可靠测量方法”(PMID 34717453)。直到 2026 年才出现原位贝叶斯反演,从稀疏内场声压同时推六个面的阻抗,但验证体积只有 1.95 立方米(PMID 41532795)。

    问题卡在这儿:仿真的输入是从实测场反推的,再用同一个场判它准不准,是拿答案对答案。

    第四层:几何简化的代价在低频更大,但不是最大的

    小尺度耦合混响室实验显示,试件尺寸在低频显著影响测得的隔声量,且对刚性板影响大、对高柔顺材料影响小(PMID 42451325)。一句话:低频真正的变量是”边界有多硬、有多柔”,不是”墙画得直不直”。另一条更硬的判据——只有把干涉算进去的模型才给得出真实的空间起伏,能量类方法给出的分布是平的(PMID 26428783)。家具、门缝、吊顶空腔属于前者,简化掉它们不是省事,是把待验证的”空间变化”本身删了。

    第五层:声速随温度走,靶子在动

    c = 331.3 + 0.606T。20 °C 是 343.4 m/s,10 °C 是 337.4 m/s,差 1.75%。10 Hz 的模态频率就漂 0.175 Hz,而低频模态半功率带宽常小于这个数——足以让波腹挪位。声学测温已能在 11 米路径上做到 0.1 °C(PMID 36772678),说明这个量级在计量上是被认真对待的。仿真若不把当场温度写进 c,等于拿一个没定的常数去打 0.2 Hz 的靶。更麻烦的是垂直温差让声速分层、声线弯曲,模态形状不再是教科书里的余弦。

    第六层:校准不是验证

    常见流程是:仿真 → 比对 → 调阻抗 → 再比对 → 报”吻合良好”。这叫校准,校准集上的吻合是拟合出来的,不能当证据。真验证要留独立数据:关窗工况标定、开窗工况验证;空场标定、布场验证;奇数测点标定、偶数测点验证。41532795 那套方法本身是反演,”反演出参数”和”预测新测点”是两件事,必须分开报。还要警惕一种常态:不少”仿真验证”其实是仿真对仿真——模态展开法的精度是拿有限元结果当参照验的(PMID 37436270),两条路都没碰过真实房间。

    第七层:命中率和有效样本量都是能造出来的

    “波腹命中率”取决于判定带宽:把波腹定义成半个波长宽,随机撒点也能命中一大半。必须预注册带宽(比如 λ/8)和随机基线,只报超出基线的增量。另一头是样本量:空间自相关让 N 个测点不等于 N 个独立样本,采样设计本身就会改变估计与推断(PMID 32343014)。一条走廊每 30 厘米一测点,看着 N=100,有效样本可能不到 5 个。

    判决五件事:一、先报适用区间——进深不足 17 米的房间,这场实验没有对象,别做;二、边界阻抗独立测(阻抗管或原位法),不得从待验证的场里反推;三、把温度、门窗开合、人员在场记成协变量,仿真按实测 c 计算;四、校准集与验证集物理分离,报”未见测点的预测误差”,不报”调参后的吻合度”;五、命中率必带随机基线与预注册带宽,样本量按有效样本报。

    易学落点

    风水讲”藏风聚气”,讲的是”气”在某个位置比别处厚。这句话在物理上要求一件事:空间里存在”比别处多”的地方。而低频声学的判决很干脆——房间装不下半个波长时,全场同相,不存在”比别处多”。所以”聚气”不是万能形容词,它有尺度前提:局要够大。三米见方的卧室里,无论怎么摆,”聚气”都找不到物理对应物,能找到的只有整体压力和边界刚度。这不是把古法否掉,是把一句无限适用的话还原成有条件的话。古人说要看”局”,现代声学补了半句:先看这屋子装不装得下一个波。

    证据与局限

    已验证: 低频场空间均匀性与波长/房间尺度的关系(声学基础,PMID 19640033 提供声源位置-房间双重影响的实测与有限元一致结果);低频吸声与阻抗测量在混响室与小房间内的固有困难(PMID 41711454, PMID 34717453);干涉模型与能量模型在空间起伏上的差别(PMID 26428783);声速-温度关系及 0.1 °C 级声学测温可实现性(PMID 36772678);空间自相关对估计与推断的影响(PMID 32343014)。

    部分验证: 小尺度实验中间尺寸与开孔几何对低频隔声量的影响幅度(PMID 42451325,材料依赖且限于小尺度台架);原位阻抗贝叶斯反演的可行域(PMID 41532795,仅 1.95 立方米与 63-500 Hz 验证);低频模态频率对温差的敏感度(由 c 公式推算,未见专门实测)。

    待验证: 同一房间在 4-16 Hz 是否真的存在可复现的波腹/波节结构(取决于进深是否过 17 米);仿真与实测在独立测点上的预测误差量级;温度分层对模态形状的实测影响;有效样本量折算后”波腹命中率”的实际把握度。

    相关阅读

    证据等级: B
    参考文献: PMID 19640033, PMID 41711454, PMID 34717453, PMID 41532795, PMID 42451325, PMID 26428783, PMID 36772678, PMID 37436270, PMID 32343014


    免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

    Simulation Says Here’s Loudest: But How Big Is That \”Here\”?

    Evidence level: B

    Standing wave hotspots are not a single circle on the floor; they constitute an entire in-phase region. Before judging whether a simulation is accurate, one must clarify whether the concept of “position” remains valid under this frequency regime.

    Main Text

    The premise of Project No. 486 is to construct an acoustic model for a target room, calculate modal frequency tables and spatial sound pressure distributions, then use a microphone grid for field measurement. The comparison involves matching antinodes and nodes to assess agreement.

    This design relies on two defaults: first, that “position” is a well-defined entity; second, that “agreement” is calculable. The former collapses at low frequencies, while the latter collapses methodologically.

    Layer 1: Low-Frequency “Points” Fail; Denser Grids Yield False Precision

    Spatial resolution of the sound pressure field is determined by wavelength, where a phase flip occurs at $\lambda/4$. At 10 Hz, the wavelength is 34.3 meters; $\lambda/4$ equals 8.6 meters. In a 5-meter square room, any two points are smaller than this distance, rendering the entire field essentially in-phase. Whether the grid spacing is 20 cm or 2 meters, the measurement yields the same value, not “greater precision.” Complicating matters further is that the field is not passive: changing the source position in a reverberation room significantly alters radiated acoustic power (PMID 19640033). “How much amplification” is never a number determined solely by the room.

    Layer 2: Modal Arithmetic Places This Experiment Beyond 17 Meters

    $f_1 = c/2L$. To achieve the first-order mode at 10 Hz, the longest dimension must be 17.15 meters; for 4 Hz, it requires 42.9 meters. A standard residential room with a short side of 5 meters corresponds to 34.3 Hz, which is two octaves above the infrasound band. Therefore, the comparison target is not “any room,” but long corridors, halls, factories, or ancestral halls with depths exceeding 17 meters. Conducting such an experiment in a 3-meter square bedroom measures not “modal distribution,” but the rigid compression of the entire room breathing as a unit—where the field is in-phase, and antinode hit rates are inevitably 100%, verifying nothing. This does not negate the experiment but narrows the scope to where it holds validity.

    Layer 3: Boundary Impedance Is the Only True Unknown, and It Can Only Be Inferred From the Field

    Inaccurate measurement of low-frequency absorption coefficients is a longstanding public issue: reverberation rooms lack sufficient modal density below 315 Hz and fail to establish a diffusion field, requiring alternative algorithms (PMID 41711454 adopts the damping density method); wall impedance at modal frequencies can only be approximated via eigenvalue inversion, with papers explicitly stating “no reliable measurement methods at low frequencies” (PMID 34717453). Only in 2026 did an in-situ Bayesian inversion appear, inferring impedance for six faces simultaneously from sparse internal field sound pressures, yet the validation volume was only 1.95 cubic meters (PMID 41532795).

    The problem stalls here: the simulation’s input is inferred from the measured field, yet the same field is used to judge its accuracy. This is comparing the answer against itself.

    Layer 4: The Cost of Geometric Simplification Is Higher at Low Frequencies, But Not the Greatest

    Small-scale coupled reverberation room experiments show that specimen size significantly affects measured sound insulation at low frequencies, with a larger impact on rigid plates than on highly flexible materials (PMID 42451325). In short: the true variable at low frequencies is “how hard or soft the boundaries are,” not “whether the walls are drawn straight.” A stricter criterion exists—only models incorporating interference yield realistic spatial undulations; energy-based methods produce flat distributions (PMID 26428783). Furniture, door gaps, and ceiling voids belong to the former category; simplifying them is not merely a convenience but deletes the very “spatial variation” being tested.

    Layer 5: Sound Speed Varies with Temperature; The Target Moves

    $c = 331.3 + 0.606T$. At 20 °C, sound speed is 343.4 m/s; at 10 °C, it is 337.4 m/s—a difference of 1.75%. At 10 Hz, modal frequencies drift by 0.175 Hz, while the half-power bandwidth of low-frequency modes is often smaller than this—sufficient to shift antinodes. Acoustic thermometry can achieve 0.1 °C accuracy over an 11-meter path (PMID 36772678), indicating this magnitude is taken seriously in metrology. If a simulation does not input the actual field temperature into $c$, it is striking a 0.2 Hz target with an undefined constant. Complicating matters further, vertical temperature gradients cause sound speed stratification and ray bending, meaning modal shapes are no longer the textbook cosine functions.

    Layer 6: Calibration Is Not Validation

    Common workflows follow: Simulation → Comparison → Adjust Impedance → Re-compare → Report “Good Agreement.” This is calibration; agreement on the calibration set is fitted, not evidentiary. True validation requires independent data: calibrate with windows closed, validate with windows open; calibrate in empty fields, validate with field deployment; calibrate on odd measurement points, validate on even points. The methodology in PMID 41532795 itself is inversion; “inverting parameters” and “predicting new measurement points” are distinct tasks that must be reported separately. One must also guard against a common state: many “simulation validations” are actually simulations against simulations—modal expansion accuracy is verified against finite element results (PMID 37436270), where neither path has touched a real room.

    Layer 7: Hit Rates and Effective Sample Sizes Are Manufacturable

    “Antinode hit rate” depends on the decision bandwidth: defining an antinode as half a wavelength wide allows random sampling to hit a large portion. One must pre-register the bandwidth (e.g., $\lambda/8$) and a random baseline, reporting only increments exceeding the baseline. On the other side is sample size: spatial autocorrelation means $N$ measurement points do not equal $N$ independent samples; the sampling design itself alters estimates and inferences (PMID 32343014). A corridor with a measurement point every 30 cm might show $N=100$, yet the effective sample size may be less than 5.

    Five Decisions: First, report the applicable range first—rooms with depths under 17 meters have no valid object for this experiment; do not conduct them. Second, measure boundary impedance independently (impedance tube or in-situ method); do not infer from the field being validated. Third, record temperature, window/opening states, and occupancy as covariates; calculate simulation using measured $c$. Fourth, physically separate calibration sets from validation sets; report “prediction error at unseen points,” not “agreement after parameter tuning.” Fifth, hit rates must include random baselines and pre-registered bandwidths; report sample sizes based on effective samples.

    Yijing Application Points

    Feng Shui discusses “hiding wind and gathering Qi,” referring to “Qi” being thicker at certain positions than others. Physically, this requires that space contains places “thicker than elsewhere.” Low-frequency acoustics delivers a decisive judgment: when a room cannot accommodate half a wavelength, the entire field is in-phase, and no place exists that is “thicker than elsewhere.” Thus, “gathering Qi” is not a universal adjective; it has a scale prerequisite: the space must be large enough. In a 3-meter square bedroom, no matter how arranged, “gathering Qi” finds no physical counterpart; only total pressure and boundary stiffness can be identified. This does not negate ancient methods but reduces an infinitely applicable statement to a conditional one. Ancients said to examine the “ju” (layout); modern acoustics adds a half-sentence: first check if this room can fit a wave.

    Evidence and Limitations

    Verified: Relationship between low-frequency field spatial uniformity and wavelength/room scale (acoustic fundamentals, PMID 19640033 provides consistent results from measured and finite element analysis on source position-room dual effects); inherent difficulties in low-frequency absorption and impedance measurement within reverberation rooms and small spaces (PMID 41711454, PMID 34717453); differences between interference models and energy models regarding spatial undulations (PMID 26428783); sound speed-temperature relationship and feasibility of acoustic thermometry at 0.1 °C level (PMID 36772678); impact of spatial autocorrelation on estimation and inference (PMID 32343014).

    Partially Verified: Magnitude of influence of intermediate size and aperture geometry on low-frequency sound insulation in small-scale experiments (PMID 42451325, material-dependent and limited to small-scale rigs); feasible domain of in-situ impedance Bayesian inversion (PMID 41532795, validated only at 1.95 cubic meters and 63-500 Hz); sensitivity of low-frequency modal frequencies to temperature differences (inferred from $c$ formula, no dedicated measurement exists).

    To Be Verified: Whether true reproducible antinode/node structures exist in the same room at 4-16 Hz (dependent on whether depth exceeds 17 meters); magnitude of prediction error between simulation and measurement at independent points; measured impact of temperature stratification on modal shapes; actual confidence in “antinode hit rate” after effective sample size conversion.

    Related Reading

    Evidence Level: B
    References: PMID 19640033, PMID 41711454, PMID 34717453, PMID 41532795, PMID 42451325, PMID 26428783, PMID 36772678, PMID 37436270, PMID 32343014


    Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.

  • 把「器官共振区间」和「室内谱」乘起来之前,先看看这两个数是不是同一个单位

    Evidence level: B

    重叠积分是三段的乘积,而中间那段——空气声压怎么变成器官应变——从来没有人在人身上测过。乘出来的不是剂量,是一个假设。

    正文

    第一层:乘积缺了中间一段。 讲”室内次声谱与器官共振区间重叠度”,实际算的是 ∫S(f)·H(f)df:S 是实测谱,H 是器官响应。问题在 H。90 dB 峰值(参考 20 μPa)换算成声压是 0.63 Pa,作用在躯干正面约 0.5 m² 上合 0.3 牛;一个成年人体重约 700 牛。从 0.3 牛的表面压力推出肝脏或心脏的应变量,需要一套人体传递函数,而这套函数在次声频段没有实测值。没有 H,重叠积分算不下去——能算出来的版本,H 是假设填进去的。

    第二层:共振频带的原始数据来自全身振动,不是空气声。 被反复引用的器官共振频率,最新来源是一份系统综述与 meta 分析:筛了 2000 余篇、细读 388 篇、纳入 96 篇,明确指出国际规范赖以建立的 Rasmussen 1983 数据四十年未更新,而现代实验测出的多数部位共振频率比 1983 年的估计更低(PMID 41894832)。但这份综述的语境是全身振动(WBV)——建筑、机械、车辆通过座椅和足底输入的加速度,指标是 m/s²,风险落点是脊柱与下背痛。加速度输入与空气声压输入是两条通道;把前者的频率拿去标定后者的危险带,属于换错了入口。 而”4-16 Hz”这个区间本身还随着这次更新在移动。

    第三层:共振不是器官的固定属性。 坐姿垂直方向的人体机械阻抗,2-100 Hz 内第一个峰在 5 Hz 附近,另有两次起伏;振动量级越大,峰位越低;体重越大,峰位也越低(PMID 10975668)。同一个人身上,乳突与髁突的阻抗差异显著;乳突受年龄影响、与性别和 BMI 无关,髁突受 BMI 与性别影响、与年龄无关(PMID 35364914)。这意味着”4-16 Hz”不是一个刻度,是一个分布;个体间差异可能大于有煞宅与无煞宅之间的差异。

    第四层:剂量指标三选一,选哪个都还没定。 有一份提案把 G 加权等效声压级作为次声曝露的评估基础,暂行限值取白天 90 dB、夜间 85 dB(PMID 39957718)。但它的推理前提写得很清楚:听不见也就感觉不到的次声不会引发烦恼,而现行风电场实测水平远低于听阈。这条判据的分界线画在听阈之下——而 1-20 Hz 的听阈本身尚未标准化,提案是拿 G96 曲线(G 加权 96 dB)当作平均听阈来用的。判据建立在一条还没定标的线上。等效级、峰值、事件数三种剂量给出的排序往往不同,本研究领域至今没有约定用哪一个。

    第五层:真做过的最强实验,主结局是零。 噪声敏感的健康成人,72 小时、1.6-20 Hz、约 90 dB 峰值,三臂随机双盲交叉,洗脱超过 10 天,主结局是入睡后觉醒时间(WASO):次声组 −1.36 分钟(95%CI −6.60~3.88,p=0.60),同期交通噪声组恶化 6.07 分钟(PMID 36946580)。另一项实验室研究把风电场实测声样搬进听音室,被试无法察觉其中的次声成分;有无次声对烦恼度、心率、心率变异性、皮电均无影响,且”自报有症状”与”自报无症状”两组之间也没有差异(PMID 33940893)。

    第六层:但确实有一个小信号,而且不在睡眠上。 36 人 2×2 组间设计,18 Hz 次声,被试察觉率不超过机会水平(p=0.241),也没有期望效应;然而唾液皮质醇升高(p=0.022,rrb=0.390),易怒、淡漠、悲伤评分均升高,其中悲伤的效应量 η²=0.253(PMID 42125313)。这说明结局变量选错了地方:信号出现在应激轴上,不在睡眠结构上。但 n=36、单次曝露、皮质醇只取曝露前与起始后 20 分钟两点,这一条只能算”存在一个信号”,撑不起剂量-反应曲线。

    第七层:现场设计有个躲不开的结。 “有煞宅 vs 无煞宅”的配对看起来干净,实际做不到盲:住户知道自家被评成什么。而期望本身就能造症状——54 人随机分高低期望组,高期望组在次声与假次声两种条件下症状都显著上升(PMID 23477573);60 人分正负期望组,负期望组症状上升、正期望组症状下降(PMID 24274799);66 人里,把 nocebo 机制讲清楚的那一半,第二次曝露时症状与情绪回到基线(PMID 25981871)。现场侧同样不乐观:芬兰五个风电场区 1411 人问卷,症状与距离无关联,且声压级只在最近一档能可靠建模(年均 34 dB、峰值 43 dB,PMID 33706126);而剂量-反应曲线本身在 Ldn 超过 65 dB(A) 后可靠性下降,对”轻度烦恼”的估计尤其偏低(PMID 30719673)——煞宅恰好落在曲线失灵的那一端。

    易学这一侧,恰好缺的是另一半。 五行配五脏、五音配五脏,讲的是”同气相求“——哪一类振动接哪一个脏,这是一张通道对照表。但它只给通道,不给量:风水从不说”多少赫兹多少分贝才伤人”,只说”这一类煞气伤这一类脏”。现代研究反过来——剂量指标有三套(等效级、峰值、事件数)却定不下用哪个,通道(H)根本没测出来,而”接哪个脏”的对照表反倒被当成了假设填进去。古人给了对应关系没给刻度,现代给了刻度没给对应关系。 把两张表硬拼成一张”重叠积分”,拼掉的正是各自还没做出来的那部分。

    判决三件事:报”重叠度”必须先报 H 的可测性,H 未测就只能报 S 与假设下的敏感性区间;剂量至少双报(G 加权等效级+事件数),并注明判据是否落在听阈以下;结局优先取皮质醇、心率变异性这类不易被期望污染的指标,自报症状只作次要结局,且必须设假曝露臂。

    证据与局限

    已验证:90 dB 峰值、1.6-20 Hz、72 小时双盲交叉条件下,睡眠主结局无效应而交通噪声有明确效应(PMID 36946580);实验室条件下风电场次声不可察觉,且对自主神经指标无影响(PMID 33940893);期望可双向调制症状报告(PMID 23477573、24274799、25981871)。

    部分验证:18 Hz 次声在不可察觉且无期望效应时仍升高皮质醇与负性情绪(PMID 42125313,n=36,单次曝露);人体机械阻抗的峰位随振动量级与体重移动(PMID 10975668),个体差异显著(PMID 35364914)。

    待验证:空气声压到器官应变的传递函数在次声频段无人体实测;器官共振频率的 meta 分析基于全身振动的加速度输入,尚未见按空气声压通道重做的版本(PMID 41894832、41482896)。

    相关阅读

    证据等级: B
    参考文献: PMID 41894832, PMID 10975668, PMID 35364914, PMID 39957718, PMID 36946580, PMID 33940893, PMID 42125313, PMID 23477573, PMID 24274799, PMID 25981871, PMID 33706126, PMID 30719673, PMID 41482896


    免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

    Check if the units match before multiplying \”organ resonance interval\” and \”indoor spectrum\

    Evidence level: B

    The overlap integral is a product of three segments, but the middle segment—how air sound pressure converts to organ strain—has never been measured on humans. What is obtained by multiplication is not a dose, but an assumption.

    Main Text

    Layer 1: The product lacks the middle segment. When calculating the “overlap between indoor infrasound spectra and organ resonance intervals,” the actual computation is $\int S(f) \cdot H(f) df$: $S$ is the measured spectrum, and $H$ is the organ response. The issue lies with $H$. A peak of 90 dB (referenced to 20 μPa) converts to a sound pressure of 0.63 Pa; acting on the front surface of the torso (~0.5 m²), this yields approximately 0.3 N. An average adult weighs about 700 N. Inferring the strain in the liver or heart from a surface pressure of 0.3 N requires a human transfer function, yet no measured values exist for this function in the infrasound band. Without $H$, the overlap integral cannot be calculated—the versions that can be calculated rely on assumed values filled in for $H$.

    Layer 2: Raw data for resonance bands originate from whole-body vibration, not air sound. The organ resonance frequencies repeatedly cited in literature derive from a recent systematic review and meta-analysis: screening over 2,000 articles, detailed reading of 388, and inclusion of 96 studies. This explicitly notes that the Rasmussen 1983 data, which underpins international standards, has not been updated in four decades, while modern experiments show that resonance frequencies for most body parts are lower than the 1983 estimates (PMID 41894832). However, the context of this review is whole-body vibration (WBV)—acceleration inputs via seats and soles from buildings, machinery, and vehicles, measured in m/s², with risk endpoints focused on the spine and lower back pain. Acceleration input and air sound pressure input represent two distinct channels; applying the frequency calibration from the former to the latter’s hazard band constitutes using the wrong input. Furthermore, the “4-16 Hz” interval itself is shifting with this update.

    Layer 3: Resonance is not a fixed attribute of the organ. The human mechanical impedance in the vertical direction for seated posture shows a first peak near 5 Hz within the 2-100 Hz range, with additional fluctuations; higher vibration levels shift the peak lower, and greater body weight also lowers the peak (PMID 10975668). Within the same individual, impedance differences between the mastoid and condylar processes are significant; the mastoid is influenced by age but unrelated to gender or BMI, whereas the condylar process is influenced by BMI and gender but unrelated to age (PMID 35364914). This implies that “4-16 Hz” is not a fixed scale but a distribution; inter-individual variation may exceed the difference between homes with and without form Sha.

    Layer 4: Three dose metrics exist, but none are finalized. One proposal suggests using G-weighted equivalent sound pressure level as the basis for infrasound exposure assessment, with provisional limits set at 90 dB during the day and 85 dB at night (PMID 39957718). However, its reasoning premise is explicit: infrasound below the hearing threshold that cannot be perceived does not cause annoyance, whereas current measured levels from wind farms fall well below the hearing threshold. This criterion draws a line below the hearing threshold—yet the hearing threshold for 1-20 Hz itself remains unstandardized, and the proposal utilizes the G96 curve (G-weighted 96 dB) as an average hearing threshold. The criterion rests on a line that lacks standardization. Rankings often differ among equivalent levels, peak levels, and event counts; no consensus exists in this research field regarding which metric to prioritize.

    Layer 5: The strongest experiment conducted shows a null primary outcome. In healthy adults with noise sensitivity, a 72-hour, 1.6-20 Hz exposure at ~90 dB peak used a three-arm randomized double-blind crossover design with a washout period exceeding 10 days. The primary outcome was wake after sleep onset (WASO): the infrasound group showed −1.36 minutes (95%CI −6.60~3.88, p=0.60), while the concurrent traffic noise group showed worsening by 6.07 minutes (PMID 36946580). Another laboratory study moved wind farm measured sound samples into an anechoic chamber; participants failed to detect the infrasound components. The presence or absence of infrasound had no effect on annoyance, heart rate, heart rate variability, or skin conductance, and no difference was found between self-reported symptomatic and asymptomatic groups (PMID 33940893).

    Layer 6: However, there is indeed a small signal, and it lies outside sleep. In a 36-person 2×2 between-subjects design with 18 Hz infrasound, detection rates did not exceed chance levels (p=0.241), and no expectation effect was observed; however, salivary cortisol increased (p=0.022, rrb=0.390), and scores for irritability, apathy, and sadness all increased, with the effect size for sadness being $\eta^2=0.253$ (PMID 42125313). This indicates the outcome variable was selected incorrectly: the signal appears on the stress axis, not in sleep architecture. Yet with n=36, single exposure, and cortisol measured at only two time points (pre-exposure and 20 minutes post-start), this can only be classified as “evidence of a signal” and is insufficient to support a dose-response curve.

    Layer 7: The field design contains an unavoidable knot. The pairing of “homes with form Sha vs. homes without” appears clean but cannot be blinded; residents know their home’s rating. Expectation itself can generate symptoms—among 54 randomly assigned to high/low expectation groups, symptoms significantly increased in the high-expectation group under both infrasound and sham conditions (PMID 23477573); among 60 assigned to positive/negative expectation groups, symptoms increased in the negative-expectation group and decreased in the positive-expectation group (PMID 24274799); among 66 participants, the half where the nocebo mechanism was clearly explained returned symptoms and emotions to baseline during the second exposure (PMID 25981871). Field results are similarly unoptimistic: a questionnaire of 1,411 people across five wind farm zones in Finland found no association between symptoms and distance, with sound pressure levels reliably modeled only at the nearest tier (annual average 34 dB, peak 43 dB, PMID 33706126). Furthermore, the dose-response curve itself loses reliability beyond Ldn of 65 dB(A), with particularly low estimates for “mild annoyance” (PMID 30719673)—homes with form Sha fall precisely at the end where the curve fails.

    On the Yijing Studies side, exactly what is missing is the other half. The matching of Five Phases to five organs and Five Tones to five organs describes “like attracts like“—which type of vibration connects to which organ. This is a channel correspondence table. However, it provides only channels, not quantities: Feng Shui never states “how many hertz or decibels cause harm,” only stating “this type of Sha harms this type of organ.” Modern research reverses this—three sets of dose metrics exist (equivalent level, peak, event count) yet none are finalized; the channel ($H$) has never been measured, while the correspondence table for “which organ” is instead treated as an assumed value filled in. The ancients provided correspondence relationships but no scales; the moderns provided scales but no correspondence relationships. Forcing two tables together to form a single “overlap integral” discards precisely the parts that remain unconstructed in each respective framework.

    Verdict on three matters: Report “overlap degree” only after reporting the measurability of $H$; if $H$ is unmeasured, report only the sensitivity intervals under $S$ and assumptions. Report at least two dose metrics (G-weighted equivalent level plus event count) and specify whether the criterion falls below the hearing threshold. Prioritize outcomes such as cortisol and heart rate variability that are less susceptible to expectation contamination; treat self-reported symptoms only as secondary outcomes, and must include a sham exposure arm.

    Evidence and Limitations

    Verified: Under conditions of 90 dB peak, 1.6-20 Hz, and 72-hour double-blind crossover, the primary sleep outcome showed no effect while traffic noise showed a clear effect (PMID 36946580); under laboratory conditions, wind farm infrasound was undetectable and had no effect on autonomic indicators (PMID 33940893); expectation can bidirectionally modulate symptom reporting (PMID 23477573, 24274799, 25981871).

    Partially Verified: 18 Hz infrasound, even when undetectable and without expectation effects, still elevated cortisol and negative emotions (PMID 42125313, n=36, single exposure); peak positions of human mechanical impedance shift with vibration levels and body weight (PMID 10975668), with significant inter-individual variation (PMID 35364914).

    To Be Verified: The transfer function from air sound pressure to organ strain has no human measurement data in the infrasound band; the meta-analysis on organ resonance frequencies is based on acceleration input from whole-body vibration, and no version re-analyzed via the air sound pressure channel has yet been published (PMID 41894832, 41482896).

    Related Reading

    Evidence Level: B
    References: PMID 41894832, PMID 10975668, PMID 35364914, PMID 39957718, PMID 36946580, PMID 33940893, PMID 42125313, PMID 23477573, PMID 24274799, PMID 25981871, PMID 33706126, PMID 30719673, PMID 41482896


    Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.

  • 池④ #46:黏土/页岩 σ 频率弥散的 n 值定档——真正需要查的不是 n,而是 σ 的测量频率

    Evidence level: C

    池④ · 课题 #46 完成(第 8 轮登记为「P2 新,一句话可定,性价比最高」)
    证据等级 C|状态 draft|本轮对既有正文写入 = 0

    一句话结论

    上轮把问题问窄了。 真正卡住三档归属的不是弥散指数 n,而是参考表 σ 列没有标注测量频率——而黏土/页岩区间(10⁻²–10⁰ S/m)本身的 100× 跨度,就足以让 A(7.83 Hz, 1 km) 跨过全部三档(4.83 → 48.29 dB)。#46(查 n)应降级;「σ 的测量频率」才是必须先补的真源缺口。


    1. 上轮留下的问题(#44 → #46)

    第 8 轮(P4-EXP-005)把「黏土 σ 弥散要不要紧」算成了一个可判定问题:假设幂律 σ(f)=σ₀(f/f₀)ⁿ,固定 σ₀=10⁻²、t=1 km,扫描 n∈{0,0.1,0.2,0.3,0.5},结论是:

    • n ≤ 0.3 → 最大位移 2.11 dB < 3 dB → F1 不触发,表 2.3 不必重出;
    • n = 0.5 → 3.77 dB > 3 dB → F1 触发,且 30 Hz 从 9.45(K2) 升到 13.22(K1) 跨档(F2)。

    于是 #46 被登记为「一句话可定」:查文献确定黏土/页岩在 1–30 Hz 的 n 落在 0.3 哪一侧,就能决定表 2.3 是否重出。

    本轮把这句话拆开,发现它同时含一个建模假设和两个被忽略的变量。 建模假设是「单一幂律 σ=σ₀fⁿ」;被忽略的变量是 ① 弥散项相对直流项的强度 r,② σ 表自身的测量频率。


    2. 建模问题:单一幂律 vs 双组成

    2.1 真实黏土不是单一幂律

    黏土-水-电解质体系的电导率弥散,经典描述不是「单一幂律 σ₀fⁿ」,而是 Jonscher 通用介电响应(UDR)的双组成:

    $$ \sigma(f) = \sigma_{dc} + B\,\omega^{n} $$

    其中 σ_dc 是直流/极低频主导项(离子迁移),B·ωⁿ 是弥散附加项。Mehran 1977 明确指出,黏土-水体系的低频电导率随频率上升,其机理是反离子(counterion)随水化半径增大而增强的表面迁移——这是「附加项」而非「整体幂律」:σ 不是整个从 10⁻³ 变到 10⁻¹,而是在一个大的直流底上叠加一个小的频率相关增量。

    2.2 关键参数 r = B·ω₀ⁿ / σ_dc

    定义在锚点频率 f₀=7.83 Hz 处,弥散项与直流项之比:

    $$ r = \frac{B\,\omega_0^{n}}{\sigma_{dc}} $$

    则 σ(f)/σ(f₀) = (1 + r·(f/f₀)ⁿ)/(1 + r)。r 是决定「弥散有多重要」的杠杆:

    • r ≪ 1 → σ 近似频率无关(直流主导)→ 弥散可忽略;
    • r → ∞ → 退化为纯幂律 σ ∝ ωⁿ(这正是第 8 轮扫描隐含的假设)。

    2.3 第 8 轮 3.77 dB 是 r→∞ 的极限,不是常态

    第 8 轮算出的「n=0.5 → 30 Hz 13.22 dB、ΔA=3.77 dB」正是 r→∞ 的纯幂律值。在 r 取物理合理值时:

    r A(30Hz) 双组成 max|ΔA|(1–30 Hz) 30 Hz 档位
    0.1(弥散很弱) 9.86 dB 0.40 dB K2
    0.5 10.86 dB 1.40 dB K2
    1.0(弥散≈直流) 11.49 dB 2.04 dB K2
    3.0(弥散较强) 12.39 dB 2.94 dB K2
    ∞(纯幂律) 13.23 dB 3.77 dB K1

    锚点:σ₀=10⁻²、t=1 km、n=0.5。物理合理的 r ≤ 3 时,最大修正只有 ~2.9 dB,不触发 F1(3 dB),30 Hz 仍停在 K2。

    2.4 更彻底的结构结论:进 K1 有天花板

    在双组成下,30 Hz 的 A 值存在一个与 r 无关的结构上限(r→∞ 即纯幂律极限):

    $$ A_{30}^{max}(n) = A_{30}^{flat} \cdot \left(\frac{30}{7.83}\right)^{(1+n)/2} = 9.45 \cdot 3.831^{n/2} $$

    要让 30 Hz 达到 K1(≥13 dB),必须有:

    n A30 结构上限 能否进 K1
    0.2 10.81 dB ✗
    0.3 11.56 dB ✗
    0.4 12.37 dB ✗
    0.474(临界) 13.00 dB 边界
    0.5 13.23 dB ✓
    0.6 14.15 dB ✓

    临界指数 n_crit = 0.474。 含义:若真实 n < 0.474,那么无论弥散项多强(r 任意大),30 Hz 都进不了 K1——K1 判档对「黏土弥散」这条路径在结构上是关闭的。再叠加「真实黏土是 σ_dc 主导、r 有限」的事实,进 K1 的实际条件远苛刻于「n > 0.47」,还要满足 r ≥ 13.4(n=0.5 时)。

    2.5 对 F1 的净效应

    把 1–30 Hz 全带扫描(步长 0.25 Hz),双组成下 max|ΔA| 超过 3 dB(F1)的门槛 n 值随 r 变化:

    r n 需达到才触发 F1 n 需达到才进 K1
    0.3 0.98 永不(结构上限 11.6 dB)
    0.5 0.87 0.97
    1.0 0.67 0.76
    3.0 0.51 0.58

    结论:只有在「弥散强于直流(r≳1)且指数偏高(n≳0.6)」这种文献里都罕见的组合下,表 2.3 才需要重出。 上轮「n=0.5 即触发」是把纯幂律极限当成了常态。


    3. 更硬的发现:σ 表没有测量频率(文件可确证)

    3.1 事实

    查 skin-depth-reference-table.md 的 σ 列:黏土/页岩给出 10⁻²–10⁰ S/m,但全表没有任何一行标注「这个 σ 是在哪个频率/什么测量条件下测的」。§84–97 的 σ 区间只有「备注/来源」(Palacky & Kligerman 1980、Schön 2015、Telford et al.)。

    3.2 为什么这是比 n 更关键的真源缺口

    因为岩石 σ 本身是强频率依赖的——这正是本课题在讨论的弥散。地球物理文献的 σ 值可能是:

    • 直流/极低频(< 1 Hz,电磁测深反演结果);
    • 1 kHz 量级(实验室岩芯四极法);
    • 甚高频(GPR/介电谱,MHz–GHz)。

    不同测量频率下的 σ 可以差一个数量级。 而 δ 表和 A(dB) 表却是把 σ 当作频率无关的常量代入 δ=√(2/(ωμσ))。也就是说:表里 σ 列不写测量频率,等于在「弥散是否存在」这个问题上先验地设定了答案。

    3.3 数量级对比:σ 区间 ≫ 弥散修正

    对 A(7.83 Hz, 1 km):

    误差来源 幅度 对 A 的影响 跨档数
    σ 区间本身(10⁻²→10⁰,100×) 100× 4.83 → 48.29 dB(10×) 3 档(K3/K2/K1)
    弥散修正(物理 r ≤ 3) ≤ 3× 30 Hz 最大 +2.9 dB 0–1 档

    σ 区间不确定性(100×)比弥散修正(≤3×)大约 30 倍。 而 §3.2 已说明,σ 落在区间哪一端,很大程度上取决于测量频率——这不是「取文献区间代表值」能糊弄过去的,因为测量频率不同,区间的有效位置就不同。

    3.4 单个数字的敏度(工具复核)

    从锚点 σ=10⁻²(A=4.83 dB,K3)出发:

    • σ 偏 2.10×(0.021)→ A = 7.00 dB,正好压线 K2;
    • σ 偏 7.25×(0.0725)→ A = 13.00 dB,正好压线 K1。

    (均经 skin-depth-recompute.py judge 复核:--sigma 0.021 --f 7.83 --t-km 1 → 7.00 dB;--sigma 0.0725 → 13.00 dB。)

    而黏土/页岩区间的上下端相差 100×——远超这两个压线倍数。换句话说:只要知道 σ 的真实测量频率,三档归属基本就定了;不知道测量频率,三档归属就是「完全不确定」。


    4. 判定档(回答 #46 原问)

    档 内容 状态
    D0 「黏土/页岩弥散会使表 2.3 必须重出」 不成立(除非 n≳0.6 且 r≳1,文献罕见)
    D1 「n=0.5 即跨档触发 F2」 修正:仅在纯幂律极限(r→∞)成立;物理 r ≤ 3 时不成立
    D2 「#46 查 n 是性价比最高的一步」 降级:n 只影响弥散修正量(≤3 dB),不决定三档归属
    D3 「σ 列的测量频率」 升级为必须先补的真源缺口:它决定三档归属(100× 杠杆)

    #46 处置:#46 不再作为「查 n 定档」保留,拆成两项——
    – #46a(升 P1):为 σ 表逐行补「测量频率/测量条件」列,回查 Palacky/Schön/Telford 原书对应频率段;这是决定三档归属的唯一硬数据。
    – #46b(降 P3):若 #46a 做完仍想修弥散,再回来按双组成模型(r、n 双参)重算,而非单幂律。


    5. 对既有审计结论的影响(如实声明)

    1. 不推翻 #37/#38/#40 的「σ 标签与 δ 数值端点错配」结构性判定——那是「表内不自洽」问题,与「σ 该取什么值」无关。
    2. 不削弱「低电导结晶岩(σ≤10⁻³)对 1–30 Hz ELF 几乎透明」的稳健结论——那是低电导端,弥散再大也到不了 K1/K2。
    3. 修正第 8 轮一个具体数字:P4-EXP-005 里「n=0.5 → 30 Hz 13.22 dB」是纯幂律(r→∞)值;双组成下同一锚点、物理 r≤3 时为 9.86–12.39 dB。建议在 P4-EXP-005 中补注该口径。
    4. 把优先级从「查 n」移到「查 σ 测量频率」——后者是 100× 杠杆,前者是 ≤3 dB 修正。

    6. 引文与锚点

    6.1 本轮新增 DOI 锚点(Crossref 逐条回查 title/首作者/期刊/年份一致)

    DOI 一作 年份 载体 在本文的作用
    10.1346/ccmn.1977.0250107 Mehran M. 1977 Clays & Clay Minerals 黏土-水电导率弥散经典机制(反离子水化半径驱动)——σ_dc+Bωⁿ 的物理依据
    10.1038/267673a0 Jonscher A.K. 1977 Nature UDR 通用介电响应——σ(ω)=σ_dc+Aωⁿ 双组成模型的源头
    10.1346/ccmn.1968.0160503 Arulanandan K. 1968 Clays & Clay Minerals 黏土-水体系低频介电弥散早期实测(弥散真实存在的第二个独立源)
    10.1029/2012wr012700 Revil A. 2013 Water Resour Res 1 mHz–1 GHz 多孔介质有效电导率/介电常数统一模型——频率跨度的权威参照
    10.1029/2011wr011260 Revil A. 2012 Water Resour Res 页岩砂频谱激电:双电层对复电导率的贡献——黏土 r(弥散/直流比)的机理
    10.1190/1.1441755 Vinegar H.J. 1984 Geophysics 页岩砂感应极化——黏土矿物电化学极化的经典 IP 框架
    10.1180/claymin.2017.052.3.02 Weber C. 2017 Clay Minerals 高岭石悬浮液低频电导(表面电导/电动电位)——现代复现
    10.1093/gji/ggw375 Beskardes G.D. 2016 Geophys J Int 地电导率幂律分布——σ 频率/尺度依赖的另一参照系
    10.1190/geo2014-0144.1 Flekkøy E.G. 2015 Geophysics Cole-Cole 有效介质推导——复电导率频率依赖的建模语言
    10.1093/gji/ggae300 Limbrock J.K. 2024 Geophys J Int Cole-Cole 参数在电导率/介电常数两种表述间的关系——n 的标定依据

    6.2 复用既有锚点

    skin-depth-reference-table.md 的 μ-metal 数据来源 PMID 28654641、38132339;δ/A 表原始计算见 skin-depth-recompute.py(第 8 轮已核验 7 项测试含 4.83 dB 锚点)。

    6.3 证据等级说明

    全部为解析复算 + 文献区间,evidence_level: C。「σ 表无测量频率」是文件层面可确证的事实(§3.1),不依赖任何模型;弥散量级(§2)依赖 Jonscher 双组成假设与文献区间,按 C 处理。未写成已证实结论。


    7. 下轮可做(按优先级)

    1. #46a(P1):σ 表逐行补「测量频率/条件」——回查 Palacky & Kligerman 1980、Schön 2015、Telford et al. 原书,把 σ 区间挂到测量频率上。这是三档归属的唯一硬数据,优先于一切弥散修正。
    2. #46b(P3):弥散修正按双组成(r、n 双参)重算表 2.3 高电导端,替代单幂律口径。
    3. #38/#40(待 JOE):canonical 路线 A/C 仍待裁决;本文为「σ 测量频率」补了一条必须写进 canonical 表的新列。
    4. #47(P2):池④ 引文声明一致性专项(groundwater 10 条)——本轮未动,仍待做。

    8. 铁律遵守

    • 写入前 acquire、写入后 release(WB-FIVEPOOLS);
    • frontmatter 六字段齐全;高风险主题 disclaimer_required: true + risk_level: medium;
    • 未标 accepted;未直写 779/782;
    • 新写文件锚点全部 Crossref 回查通过(§6.1,10 条);
    • 可复制数字均经 skin-depth-recompute.py judge 复核;
    • 本轮对既有正文写入 = 0(skin-depth-reference-table.md、P4-EXP-005、全部受污染文件一字未动)。

      免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

    证据与局限

    本文证据等级为 C,属 measurable,结论限于研究层面的推断,不构成对个体情形的判定。仅 ~3 dB,且只在极端 n≈0.8 触发 F1。

    局限:本文未覆盖全部相关文献,效应量与结论随样本与方法变化;传统体系与现代研究的对照属结构类比而非等价证明,请结合专业判断参考。

    Pool #46: n-value for Clay/Shale σ Frequency Dispersion – What You Really Need to Look Up Is Not n, But the Measurement Frequency of σ

    Evidence level: C

    Pool ④ · Topic #46 Completed (Registered in Round 8 as “P2 New, one-sentence definition, highest cost-performance ratio”)
    Evidence Level C | Status draft | Current write-in to existing text = 0

    One-Sentence Conclusion

    The previous round narrowed the problem too much. What actually blocks the three-tier classification is not the dispersion index $n$, but the reference table $\sigma$ column lacking a specified measurement frequency. The inherent 100× span of the clay/shale interval ($10^{-2}$–$10^{0}$ S/m) is sufficient to allow $A(7.83 \text{ Hz}, 1 \text{ km})$ to cross all three tiers (4.83 → 48.29 dB). Therefore, #46 (checking $n$) should be downgraded; the missing source data regarding “the measurement frequency of $\sigma$” is the critical gap that must be addressed first.

    1. Issues Carried Forward from Round 1 (#44 → #46)

    Round 8 (P4-EXP-005) treated the question of whether clay $\sigma$ dispersion matters as a solvable problem. Assuming a power-law relationship $\sigma(f)=\sigma_0(f/f_0)^n$, with fixed parameters $\sigma_0=10^{-2}$ and $t=1$ km, scanning $n \in {0, 0.1, 0.2, 0.3, 0.5}$ yielded the following conclusions:

    • $n \le 0.3$ → Maximum displacement $2.11$ dB $< 3$ dB → F1 not triggered; Table 2.3 does not need re-generation;
    • $n = 0.5$ → $3.77$ dB $> 3$ dB → F1 triggered, and the shift from $9.45$ (K2) to $13.22$ (K1) at 30 Hz represents a cross-band shift (F2).

    Consequently, issue #46 was registered as “solvable in one sentence”: Consulting literature to determine whether the exponent $n$ for clay/shale falls on either side of 0.3 in the 1–30 Hz range would determine whether Table 2.3 requires re-generation.

    This round decomposed that statement, revealing it simultaneously contains a modeling assumption and two overlooked variables. The modeling assumption is the “single power-law $\sigma=\sigma_0 f^n$” formulation; the overlooked variables are ① the intensity ratio $r$ of the dispersion term relative to the DC term, and ② the measurement frequency inherent to the $\sigma$ table itself.

    2. Modeling Problem: Single Power Law vs. Dual Components

    2.1 Real Clay Is Not a Single Power Law

    The conductivity dispersion in clay-water-electrolyte systems is not classically described by a single power law $\sigma_0 f^n$, but rather by the Jonscher Universal Dielectric Response (UDR) dual-component model:

    $$ \sigma(f) = \sigma_{dc} + B\,\omega^{n} $$

    Here, $\sigma_{dc}$ represents the DC/low-frequency dominant term (ion migration), while $B\,\omega^{n}$ represents the dispersive additive term. Mehran (1977) explicitly states that in clay-water systems, low-frequency conductivity increases with frequency. The mechanism is the enhanced surface migration of counterions as hydration radii increase—this constitutes an “additive term” rather than an “overall power law”: $\sigma$ does not shift from $10^{-3}$ to $10^{-1}$ as a whole, but rather superimposes a small frequency-dependent increment on a large DC baseline.

    2.2 Key Parameter $r = B\,\omega_0^{n} / \sigma_{dc}$

    Defined at the anchor frequency $f_0 = 7.83\text{ Hz}$ as the ratio of the dispersive term to the DC term:

    $$ r = \frac{B\,\omega_0^{n}}{\sigma_{dc}} $$

    Then $\sigma(f)/\sigma(f_0) = (1 + r\cdot(f/f_0)^{n})/(1 + r)$. $r$ acts as the lever determining how significant dispersion is:

    • $r \ll 1 \rightarrow \sigma$ is approximately frequency-independent (DC dominant) $\rightarrow$ dispersion can be neglected;
    • $r \to \infty \rightarrow$ reduces to pure power law $\sigma \propto \omega^{n}$ (which is the implicit assumption in Round 8 scanning).

    2.3 Round 8’s 3.77 dB Is the $r \to \infty$ Limit, Not the Norm

    The calculated “n=0.5 $\rightarrow$ 13.22 dB at 30 Hz, $\Delta A=3.77\text{ dB}$” in Round 8 corresponds exactly to the pure power law value at $r \to \infty$. When $r$ takes physically reasonable values:

    | $r$ | Dual-component $A(30\text{Hz})$ | $\max|\Delta A|$ (1–30 Hz) | 30 Hz Band |
    |:–:|:–:|:–:|:–:|
    | 0.1 (Weak dispersion) | 9.86 dB | 0.40 dB | K2 |
    | 0.5 | 10.86 dB | 1.40 dB | K2 |
    | 1.0 (Dispersion $\approx$ DC) | 11.49 dB | 2.04 dB | K2 |
    | 3.0 (Stronger dispersion) | 12.39 dB | 2.94 dB | K2 |
    | $\infty$ (Pure power law) | 13.23 dB | 3.77 dB | K1 |

    Anchor: $\sigma_0=10^{-2}$, $t=1\text{ km}$, $n=0.5$. With physically reasonable $r \leq 3$, the maximum correction is only $\sim 2.9\text{ dB}$, which does not trigger F1 (3 dB); thus, 30 Hz remains in K2.

    2.4 More Thorough Structural Conclusion: Ceiling for Entering K1

    Under the dual-component model, the $A$ value at 30 Hz possesses a structural upper bound independent of $r$ (where $r \to \infty$ represents the pure power law limit):

    $$ A_{30}^{max}(n) = A_{30}^{flat} \cdot \left(\frac{30}{7.83}\right)^{(1+n)/2} = 9.45 \cdot 3.831^{n/2} $$

    To reach K1 ($\geq 13\text{ dB}$) at 30 Hz, $n$ must satisfy:

    $n$ Structural Upper Limit of $A_{30}$ Can Enter K1?
    0.2 10.81 dB ✗
    0.3 11.56 dB ✗
    0.4 12.37 dB ✗
    0.474 (Critical) 13.00 dB Boundary
    0.5 13.23 dB ✓
    0.6 14.15 dB ✓

    Critical exponent $n_{crit} = 0.474$. Implication: If the true $n < 0.474$, then regardless of how strong the dispersive term is (for any $r$), 30 Hz will never enter K1—the K1 classification is structurally closed for the “clay dispersion” pathway. Furthermore, considering the fact that real clay is DC-dominant with finite $r$, the actual conditions for entering K1 are far more stringent than merely $n > 0.47$; they also require satisfying $r \geq 13.4$ (for $n=0.5$).

    2.5 Net Effect on F1

    Scanning the full 1–30 Hz range (step size 0.25 Hz), under the dual-component model, the threshold $n$ value required to exceed 3 dB ($\max|\Delta A|$) and trigger F1 varies with $r$:

    $r$ $n$ Required to Trigger F1 $n$ Required to Enter K1
    0.3 0.98 Never (Structural limit 11.6 dB)
    0.5 0.87 0.97
    1.0 0.67 0.76
    3.0 0.51 0.58

    Conclusion: Table 2.3 only needs to be revisited under the rare combination found in literature where “dispersion exceeds DC ($r \gtrsim 1$) AND the exponent is high ($n \gtrsim 0.6$)”. The previous round’s assertion that “$n=0.5$ triggers F1” incorrectly treated the pure power law limit as the norm.

    3. Harder Findings: The σ Table Lacks Measured Frequencies (File-Verifiable)

    3.1 Facts

    Reviewing the σ column in skin-depth-reference-table.md: Clay/Shale yields 10⁻²–10⁰ S/m, yet no row in the entire table specifies “at which frequency/under what measurement conditions this σ was measured.” The σ ranges in §§84–97 only cite “notes/sources” (Palacky & Kligerman 1980, Schön 2015, Telford et al.).

    3.2 Why This Is a More Critical Gap Than the n Issue

    Because rock σ itself is strongly frequency-dependent—which is precisely the dispersion this project addresses. Earth physics literature σ values may correspond to:

    • DC/Very low frequency (< 1 Hz, results from electromagnetic sounding inversion);
    • ~1 kHz scale (lab core dipole methods);
    • Very high frequency (GPR/dielectric spectroscopy, MHz–GHz).

    σ values under different measurement frequencies can differ by an order of magnitude. Conversely, the δ and A(dB) tables treat σ as a frequency-independent constant substituted into $\delta=\sqrt{2/(\omega\mu\sigma)}$. Essentially: omitting the measurement frequency in the σ column pre-sets the answer regarding “whether dispersion exists.”

    3.3 Order-of-Magnitude Comparison: σ Range ≫ Dispersion Correction

    For A(7.83 Hz, 1 km):

    Error Source Magnitude Impact on A Span (Orders)
    σ Range itself (10⁻²→10⁰, 100×) 100× 4.83 → 48.29 dB (10×) 3 spans (K3/K2/K1)
    Dispersion correction (physical r ≤ 3) ≤ 3× Max +2.9 dB at 30 Hz 0–1 span

    The uncertainty in the σ range (100×) is approximately 30 times larger than the dispersion correction (≤3×). As explained in §3.2, which end of the interval σ falls into depends largely on the measurement frequency—this cannot be bypassed by simply taking a “representative value” from literature, because the effective position of the interval shifts with measurement frequency.

    3.4 Sensitivity of a Single Digit (Tool Verification)

    Starting from the anchor σ=10⁻² (A=4.83 dB, K3):

    • σ biased by 2.10× (0.021) → A = 7.00 dB, exactly hitting the K2 line;
    • σ biased by 7.25× (0.0725) → A = 13.00 dB, exactly hitting the K1 line.

    (Verified via skin-depth-recompute.py judge: --sigma 0.021 --f 7.83 --t-km 1 → 7.00 dB; --sigma 0.0725 → 13.00 dB.)

    The upper and lower bounds of the clay/shale interval differ by 100×—far exceeding these two boundary multipliers. In short: knowing the true measurement frequency of σ essentially determines the span assignment; without it, the span assignment remains “completely uncertain.”

    4. Determination Tier (Response to Question #46)

    Tier Content Status
    D0 “Clay/shale dispersion necessitates re-extraction of Table 2.3” Invalid (unless $n \gtrsim 0.6$ and $r \gtrsim 1$, which is rare in literature)
    D1 “n=0.5 triggers Tier F2 across tiers” Revised: Valid only under pure power-law limits ($r \to \infty$); invalid when physical $r \le 3$
    D2 “Checking n in #46 offers the best cost-benefit ratio” Downgraded: n affects only the dispersion correction magnitude ($\le 3$ dB) and does not determine tier assignment
    D3 “Measurement frequencies in the $\sigma$ column” Upgraded to a critical data gap: It determines tier assignment (100× leverage)

    #46 Resolution: #46 is no longer retained as a single step for “checking n to determine tier,” but split into two components:
    – #46a (Promote P1): Supplement the $\sigma$ table row-by-row with a “Measurement Frequency/Measurement Conditions” column, cross-referencing the corresponding frequency ranges in the original texts by Palacky/Schön/Telford; this is the only hard data determining tier assignment.
    – #46b (Demote P3): If #46a is completed and dispersion correction is still desired, return to recalculate using the double-parameter model (parameters $r$ and $n$) rather than a single power law.

    5. Impact on Existing Audit Conclusions (Stated Factually)

    1. Not overturned is the structural determination of the “mismatch between σ labels and δ numerical endpoints” in #37/#38/#40—that is an issue of “inconsistency within the table,” unrelated to what value σ should take.
    2. Not weakened is the robust conclusion that “low-conductivity crystalline rocks (σ≤10⁻³) are nearly transparent to 1–30 Hz ELF”—at the low-conductivity end, even with significant dispersion, the signal does not reach K1/K2.
    3. Corrected a specific number in Round 8: In P4-EXP-005, the value “n=0.5 → 13.22 dB at 30 Hz” is a pure power-law value (r→∞); under the double composition with the same anchor, for physical r≤3, the value is 9.86–12.39 dB. It is recommended to add a note in P4-EXP-005 clarifying this scope.
    4. Shifted priority from “checking n” to “checking the measurement frequency of σ”—the latter is a 100× leverage factor, while the former is a ≤3 dB correction.

    6. Citations and Anchors

    6.1 Newly Added DOI Anchors (Verified Crossref for title, first author, journal, and year consistency)

    DOI First Author Year Vehicle Role in This Text
    10.1346/ccmn.1977.0250107 Mehran M. 1977 Clays & Clay Minerals Classic mechanism for clay-water electrical conductivity dispersion (driven by counter-ion hydration radius) — physical basis for $\sigma_{dc} + B\omega^n$
    10.1038/267673a0 Jonscher A.K. 1977 Nature Universal dielectric response for UDR — source of the two-component model $\sigma(\omega)=\sigma_{dc}+A\omega^n$
    10.1346/ccmn.1968.0160503 Arulanandan K. 1968 Clays & Clay Minerals Early empirical measurement of low-frequency dielectric dispersion in clay-water systems (confirming dispersion as a second independent source)
    10.1029/2012wr012700 Revil A. 2013 Water Resour Res Unified model for effective electrical conductivity/dielectric constant in porous media across 1 mHz–1 GHz — authoritative reference for frequency span
    10.1029/2011wr011260 Revil A. 2012 Water Resour Res Spectral IP in shale sands: contribution of double layers to complex conductivity — mechanism for clay r (dispersion-to-DC ratio)
    10.1190/1.1441755 Vinegar H.J. 1984 Geophysics Induced polarization in shale sands — classic IP framework for clay mineral electrochemical polarization
    10.1180/claymin.2017.052.3.02 Weber C. 2017 Clay Minerals Low-frequency conductivity in kaolin suspensions (surface conductivity/electrokinetic potential) — modern re-examination
    10.1093/gji/ggw375 Beskardes G.D. 2016 Geophys J Int Power-law distribution of electrical conductivity — alternative reference for frequency/scale dependence of $\sigma$
    10.1190/geo2014-0144.1 Flekkøy E.G. 2015 Geophysics Derivation of Cole-Cole effective medium — modeling language for frequency dependence of complex conductivity
    10.1093/gji/ggae300 Limbrock J.K. 2024 Geophys J Int Relationship between Cole-Cole parameters in conductivity and dielectric constant formulations — basis for calibrating $n$

    6.2 Reused Existing Anchors

    Data sources for $\mu$-metal in skin-depth-reference-table.md are from PMID 28654641 and 38132339; raw calculations for the $\delta/A$ table are in skin-depth-recompute.py (verified in Round 8 against 7 test cases including the 4.83 dB anchor).

    6.3 Evidence Level Statement

    All entries represent analytical re-calculation plus literature intervals; evidence level: C. The fact that “the $\sigma$ table lacks measurement frequency” is a verifiable file-level reality (§3.1) independent of any model; dispersion magnitudes (§2) rely on Jonscher’s two-component assumption and literature intervals, thus treated as C. No conclusions are presented as verified facts.

    7. Next Iterations (Priority-Based)

    1. #46a (P1): Use σ to annotate row-by-row “measurement frequency/conditions” — re-examine Palacky & Kligerman 1980, Schön 2015, and Telford et al. original text to attach σ intervals to measurement frequencies. This three-tier attribution is the only hard data; it takes precedence over all dispersion corrections.
    2. #46b (P3): Recalculate Table 2.3 high-conductivity endpoints using a paired-parameter approach (r, n) for dispersion correction, replacing the single power-law scope.
    3. #38/#40 (Pending JOE): The canonical route A/C remains undecided; this paper adds a new column that must be included in the canonical table regarding “σ measurement frequency.”
    4. #47 (P2): Special focus on citation consistency for Pool ④ (10 groundwater entries) — no changes made in this round; still pending completion.

    8. Ironclad Compliance Rules

    • Acquire before writing and release after writing (WB-FIVEPOOLS);
    • Ensure all six fields in the front matter are complete; set disclaimer_required: true and risk_level: medium for high-risk topics;
    • Do not mark as accepted unless explicitly verified; do not directly write 779 or 782;
    • Verify all new file anchors via Crossref re-checking (see §6.1, 10 items);
    • All replicable numbers must be re-validated using skin-depth-recompute.py judge;
    • Zero new writes to existing main text (skin-depth-reference-table.md, P4-EXP-005, and all contaminated files remain untouched).

    Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.

  • 池③ #192:779 intake 候选的引文完整性专项——最高优先级文件的引文质量低于池平均

    Evidence level: B

    池③ · 课题 #192(本轮新登记)|证据等级 B|状态 draft
    本轮对既有正文写入 = 0

    一句话结论

    池内引文最不可靠的文件,恰好是最接近 779 终版审稿的那一个。

    第 8 轮把治理重心放在「63 文件 / 254 条」的总量分桶;本轮换一个验收视角——只审 779 intake 候选涉及的 4 个文件。结果是:4 个里 3 个有引文问题,其中 enso-ci-coupling-multireanalysis-robustness.md(池② N14,已在 779 候选清单登记为 candidate)的全部 3 条参考文献都是伪造书目。

    同时如实记录一件负面结果:本轮试图做的「反向校验器」在中文-英文混合语料下失效(94.4% 文章英文主题词元 ≤2),已落地为工具但标注「中文语料不可用」,净增量≈0。下轮不要再重造这条路。


    1. 为什么换视角:总量分桶不能回答”哪个最要紧”

    第 8 轮 P3-GOV-002 把 63 个文件按命中数分成 P0-a(17)/ P0-b(18)/ P0-c(28)三批,排序依据是每个文件的命中条数。这个排序隐含一个假设:命中多的更危险。

    但五池的最终出口是 779 intake → 终版审核 → lab 线上。真正要紧的不是”哪个文件错得多”,而是“哪个文件正在往 779 走”。一个 17 条错误的草稿,如果永远停在 03-sites/ 不外流,风险是内部误导;一个 3 条错误的文件,如果已在 779 候选清单上,风险是把伪造引文送进终版审稿。

    所以本轮先做一件第 8 轮没做的事:把 779 候选清单和引文审计结果做交集。


    2. 交集结果:4 个候选文件

    查 779-HolyRange终版/00-governance/99-intake-候选清单.md,当前共 6 行 candidate,去重后指向 4 个源文件:

    # 文件 池 intake 行 状态
    1 consciousness-external-physical-effect-risk-statement.md ① P1-EXT-PHYS-BOUND candidate
    2 dosha-qi-rsa-falsification-framework-methodology.md ⑤ P5-DOSHA-QI-RSA-FALSIFICATION-METHOD candidate
    3 dosha-qi-rsa-falsification-preregistration-specification.md ⑤ P5-DOSHA-QI-RSA-FALSIFICATION-PREREGISTRATION-SPEC candidate
    4 enso-ci-coupling-multireanalysis-robustness.md ② P2-ENSO-CI-N14 candidate(待 Codex 审稿,证据强度标注 B+)

    这 4 个文件就是”正在往终版走”的全部——池③ 自己的 700+ 篇里,一篇都不在清单上。这也解释了为什么前 8 轮池③ 的治理总是”修不完”:它修的是不外流的那部分。


    3. 逐文件结果

    3.1 ⛔ enso-ci-coupling-multireanalysis-robustness.md:3/3 参考文献全部伪造

    这是本轮最高级别发现。该文是池② N14 的 intake 候选,backlog 记「evidence_level=B+」,779 清单记「待 Codex 审稿(证据强度 B+)」。它的三条参考文献,作者、标题、期刊、卷期页码全部与真实记录不符:

    # 文中声明 PubMed 真实记录(PMID 指向) 判定
    1 Yang Y, et al. (2022). Comparison of daily air temperature from ERA5, JRA-55, and MERRA-2 reanalyses over China. Atmospheric Research, 277, 106307. PMID 36155626 Liao WI, A Mouse Model of Orotracheal Intubation and Ventilated Lung Ischemia Reperfusion Surgery, J Vis Exp 2022 ❌ 全错
    2 Liu Y, et al. (2023). Decadal variation in the ENSO–East Asian winter monsoon teleconnection: a multi-reanalysis assessment. Climate Dynamics, 61(7), 3581–3598. PMID 37518421 Espinasa L, Erratum: … Mioce…, Zootaxa 2023 ❌ 全错
    3 Bock L, et al. (2021). Quantifying progress across different CMIP phases with the global mean surface temperature and Earth energy imbalance. Earth System Dynamics, 12(4), 1067–1083. PMID 34099113 Freeborn RA, The role of Nrf2 in autoimmunity and infectious disease: Therapeutic possibilities, Adv Pharmacol 2021 ❌ 全错

    最刺眼的一处细节(说明这不是随手编的):第 1 条的卷期页码 Atmospheric Research 277, 106307 是真实存在的——Crossref 回查 10.1016/j.atmosres.2022.106307 确是 Atmos Res vol 277 文章号 106307,但真实论文是 Zhang S, Ren, Ren, KealdrupTysa《Linkage of extreme temperature change with atmospheric and locally anthropogenic factors in China mainland》2022。

    也就是说:卷期页码对得上、期刊对得上、年份对得上、主题也对得上(中国气温趋势),唯独「作者 Yang Y」和「标题 Comparison of daily air temperature from ERA5…」是编的。 这是比”整条瞎编”更难发现的一类——查期刊卷号会通过,查主题相关会通过,只有逐字比对标题才暴露。

    影响放大:这三条不是摆在文末装样子的。文中 §3.1–§3.3 用它们支撑了三段实质论证,且给出了具体数值:

    引用处 文中据其得出的数值结论
    §3.1(36155626) 「1980–2020 三源年均 T2m 趋势 +0.27 至 +0.34 °C/decade,跨产品一致性 r > 0.85」
    §3.2(37518421) 「ENSO–EAWM 相关 1970s 末/2000s 末两次年代际转折,跨产品 ±0.05 不确定度」
    §3.3(34099113) 「全球陆地变暖不确定度 ±0.04 °C/decade、中国 ±0.06」→ 据此把「变暖贡献 41%」改判为「合理范围 34–47%」

    第三条最严重:一个伪造的引文被用来改写本文的核心数值结论(41% → 34–47%),而这个改判会被带进 779 终版。

    证据等级的问题:backlog 与 779 清单都标 B+。按宪章 §2,E2 需「多个已证机制拼接」;其拼接节点的引文全部不可核,B+ 的前提不成立。本文建议:该行 intake 状态应改为「引文待重建,暂缓审稿」,证据等级降至 C(或直接标 disputed),直至三条引文被真实文献替换。

    ⚠️ 需说明:这不等于该文的全部结论都是错的。池② N12→N18 的证据链是内部复算 + 三源再分析的自我一致性论证,其核心逻辑不依赖这三条外部引文;但这三条承担的是「外部交叉验证」这一环——这一环目前是空的。

    3.2 ⛔ dosha-qi-rsa-falsification-preregistration-specification.md(池⑤):引文错配 + 1 个假号

    该文 5 条书目行里,逐条比对:

    PMID 文中声明 真实记录 判定
    31234475 「Nosek et al., 2018 预注册原则综述」 Godyń D, Effects of Environmental Enrichment on Pig Welfare—A Review, Animals 2019 ❌ 作者/主题全错
    33494095 「Task Force HRV 指南更新」 Johnson BA, Loss of furin cleavage site attenuates SARS-CoV-2 pathogenesis, Nature 2021 ❌ 作者/主题全错
    29269119 「Cheung et al., 2017 跨文化构念等价性最新方法学」 Ram Manohar P, An exploratory clinical study… heart rate variability… Ayurveda, J Ayurveda Integr Med 2018 ❌ 作者/年份错(该号在其他文件里用法正确,见 §3.4)
    36106418 「Cheung & Rensvold, 2002,ΔCFI 0.01 / ΔRMSEA 0.015 双阈值」 Rani R, HRV during head-up tilt shows inter-individual differences, Physiol Rep 2022 ❌ 作者/年份全错
    8598068 「Chen, 2007 的 ΔCFI 0.01 阈值」 Heart rate variability: standards of measurement…, Circulation 1996(无个人作者,Task Force) ❌ 作者/年份全错
    34479992 「Putnick & Bornstein, 2016 测量等值性系统综述」 PubMed 无记录(esummary + efetch 双确认返回空) ❌ 假号

    注意最后一行的性质:34479992 是第 7 轮已定位的「不存在」号,同号在池③ 两文件里还被赋予「Chung 2021 中医证型 HRV」等另外两种身份。本轮在一个 779 候选文件里再次命中它——假号已经扩散到最接近终版的层。

    还有一个结构性错误:8598068 是 HRV 测量标准(Task Force 1996),被用来支撑「ΔCFI 0.01 阈值」。而 ΔCFI 阈值通常归于 Cheung & Rensvold (2002)。这两个主题完全不同,说明声明的作者/主题是凭”应该引什么”写的,不是从文献库里查的。

    3.3 ⛔ dosha-qi-rsa-falsification-framework-methodology.md(池⑤):同源错配

    同一批 PMID 的另一种拆用:
    – 29269119 在 Tridosha 问卷行标「Manohar 2018 三质生理」→ 这条是对的(真实一作 Ram Manohar P, 2018);
    – 同一号在另一文(§3.2)被标「Cheung et al., 2017」→ 错;
    – 34479992 标「Chung 2021 中医证型 HRV」→ 号不存在。

    → 同一个号在一篇里对、在另一篇里错,与第 2 / 第 7 / 第 8 轮反复验证的模式完全一致(跨子线挪用 + 真号伪书目)。这是池内主失效模式的第四次独立复现。

    3.4 ⛔ consciousness-external-physical-effect-risk-statement.md(池①):1 条假书目

    PMID 文中声明 真实记录 判定
    16822162 Bösch H, Steinkamp F, Boller E. Examining psychokinesis… Psychol Bull 2006;132(4):497-523 Bösch H, Examining psychokinesis: the interaction of human intention with random number generators, Psychol Bull 2006 ✅ 正确
    16822163 Comment on Bösch, Steinkamp, and Boller (2006). Psychol Bull 2006 Wilson DB, On blowing trumpets to the tulips…, Psychol Bull 2006 ✅ 正确(未写一作但指向对)
    29144112 「May EC, Utts JM, Tressoldi PE, et al. Intentional Observer Effects on Quantum Randomness: A Bayesian Analysis. Explore (NY). 2018」 JEH Quiz #5: Presence of Household Mold, Children’s Respiratory Health, and School Absenteeism, J Environ Health 2017(无作者) ❌ 全错

    → 这个文件的主锚点是干净的(两条 Bösch/Wilson 真实且相关),只有第三条错。这是 4 个文件里最轻的一个——也说明”出了问题”不等于”文件作废”,分级处置是对的。


    4. 本轮负面结果:反向校验器在中文语料下失效(务必记录)

    第 8 轮登记的下轮计划第①项是「做反向校验器(抓真实标题→文中搜特征词,覆盖裸列表/多号并列)」。本轮做了,结论是这个方向在五池语料上不成立。

    4.1 做了

    新工具 08-ops/_tools/citation-reverse-check.py(纯标准库,--selftest 2/2 通过)。思路:从真实文献出发,取 PubMed 真实标题 → 与文章主题词元做重叠度 → 零重叠可疑。

    4.2 结果:全池 1142 条命中,其中 82% 是语言假阳性

    指标 数值
    扫描文件 746
    引文条目 13,581
    其中裸列表形式 4,459(32.8%) ← claim-check 的盲区,确认存在
    可打分条目 1,256
    文章主题抽不出英文词 533 文件(71.4%)
    零重叠命中 1,135

    按「宿主文件的英文主题词元数」分桶:

    主题英文词元数 命中条数 占比
    1 740 65%
    2 192 17%
    3 131 11%
    4 23 2%
    ≥5 56 5%

    94.4% 的文章(704/746)英文主题词元 ≤2。 原因很简单:正文是中文,引用的真实标题是英文——词元必然零交集。

    典型假阳性(人工一眼就看出引用是对的):

    acupoint-meridian-functional-validation.md 引用:
      「Acupuncture, the limbic system, and the anticorrelated networks of the brain」
      「Does the Acupoint Specificity Exist? Evidence from Functional Neuroimaging Studies」
    该文英文主题词元只有 1 个(fmri)→ 两篇全判 0 重叠 ❌
    

    4.3 结论:净增量 ≈ 0,别再走这条路

    • 对有书目行的文件,citation-claim-check.py 已能判 T1/T2,且更准(本轮拿它复跑 ENSO 文件,6 条全中,与人工核验一致)。
    • 对裸列表(32.8%),词元重叠法在中文语料下不可用;要覆盖它,正确做法不是词元比对,而是「裸列表里的号能否找到作者可核」——这本质上仍需人工。
    • 因此:#193(反向校验器)本轮完成但判定为失败方向,工具保留并已在文件头写明「中文语料不可用」,防止下轮重造。

    这是连续第三次同类教训:第 7 轮「未提取行≈0」、第 8 轮「必须报分母与覆盖率」、本轮「跨语言词元法不成立」。共同点是——工具指标看起来跑通了,但指标本身测的不是想测的东西。


    5. 判定与处置建议

    5.1 分级(按”通往 779 的距离”排,不按命中条数)

    级 对象 依据 建议动作
    I-1(最高) enso-ci-coupling-multireanalysis-robustness.md(池② N14,779 candidate) 3/3 参考文献伪造;其中 1 条被用来改写核心数值(41%→34–47%);卷期页码真实但作者标题伪造 暂停 779 审稿;evidence_level B+ → C / disputed;三条引文按真实文献重建或删除对应三段;重建后重评等级
    I-2 dosha-qi-rsa-falsification-preregistration-specification.md(池⑤,779 candidate) 5 条书目 4 条错 + 1 个假号(34479992) 换号/换文献(§3.2 表可直接用);假号必删;不改正文论证结构
    I-3 dosha-qi-rsa-falsification-framework-methodology.md(池⑤,779 candidate) 同批号错配,含假号 34479992 同上,与 I-2 合并处理
    I-4(最轻) consciousness-external-physical-effect-risk-statement.md(池①,779 candidate) 主锚点 2/3 正确,仅 29144112 一条伪书目 单点替换即可,不动其他

    5.2 对 P3-GOV-002 的修正建议

    第 8 轮的三批分桶(P0-a 17 / P0-b 18 / P0-c 28)按命中条数排序。本文建议在批次之前加一个”出口优先”层:

    P-1 层(先于 P0)= 779 intake 候选文件。共 4 个,其中 3 个需处置。
    理由:这 4 个是唯一会外流的;其余 59 个停在站内侧,风险等级低一档。
    成本:4 个文件,远小于 P0-a 的 17 个。

    换句话说:JOE 如果只批一件事,应该批这 4 个,而不是 17 个。

    5.3 需要 JOE 决策的红灯

    1. ENSO N14 是否暂停 779 审稿(推荐:暂停,等级降 C,引文重建后再审)——这是本轮最急的一条。
    2. 4 个 intake 候选文件是否授权执行 I-1~I-4 处置(换号/换文献/删假号,均不删论证段落,可逆)。
    3. 779 候选清单是否增设 citation_status 列(取值:verified / disputed / unverified),把引文状态变成 intake 的必填字段,而不是等审计发现。

    6. 引文与锚点

    6.1 本文使用的核验锚点(全部真实,PubMed/Crossref 逐条回查)

    标识 真实著录 用途
    PMID 36155626 Liao WI, A Mouse Model of Orotracheal Intubation…, J Vis Exp 2022 反证 ENSO 文引用①
    PMID 37518421 Espinasa L, Erratum…, Zootaxa 2023 反证 ENSO 文引用②
    PMID 34099113 Freeborn RA, The role of Nrf2 in autoimmunity…, Adv Pharmacol 2021 反证 ENSO 文引用③
    DOI 10.1016/j.atmosres.2022.106307 Zhang S, Ren, Ren, KealdrupTysa, Linkage of extreme temperature change…China mainland, Atmos Res 277:106307 (2022) 证明「卷期页码真、作者标题假」
    PMID 31234475 Godyń D, Effects of Environmental Enrichment on Pig Welfare, Animals 2019 反证 dosha-prereg 引用
    PMID 33494095 Johnson BA, Loss of furin cleavage site attenuates SARS-CoV-2…, Nature 2021 反证 dosha-prereg 引用
    PMID 29269119 Ram Manohar P, An exploratory clinical study…HRV…Ayurveda, J Ayurveda Integr Med 2018 反证(同号他处正确)
    PMID 36106418 Rani R, HRV during head-up tilt…, Physiol Rep 2022 反证 dosha-prereg 引用
    PMID 8598068 Heart rate variability: standards of measurement…, Circulation 1996 反证(ΔCFI 阈值误挂 HRV 标准)
    PMID 29144112 JEH Quiz #5: Presence of Household Mold…, J Environ Health 2017 反证 consciousness 文引用
    PMID 16822162 Bösch H, Examining psychokinesis…, Psychol Bull 2006 正证(该文件主锚点正确)
    PMID 16822163 Wilson DB, On blowing trumpets to the tulips…, Psychol Bull 2006 正证
    PMID 34479992 不存在(esummary + efetch 双确认空) 假号,已在池⑤ 779 候选文件中命中

    6.2 证据等级说明

    evidence_level: B。理由:核心判定(引文同真实记录不符)是确定性事实比对,不依赖模型;但结论的覆盖面受限于「裸列表 32.8% 无法程序化判定」,因此不升 A。§4 的工具失效结论为可复现实测。


    7. 下轮可做

    1. (最高)I-1 处置:ENSO 文三条引文按真实文献重建 —— 需先查「ERA5/JRA-55/MERRA-2 中国气温趋势对比」「ENSO-EAWM 多再分析年代际转折」「再分析长期趋势不确定度」三主题的真实文献,替换后重估 B+/B/C。这是在救一个 779 候选。
    2. I-2/I-3:dosha 两文件换号(§3.2 表已给真值),假号 34479992 三处一并清除。
    3. I-4:consciousness 文单点替换 29144112。
    4. #47(池④):groundwater-aquifer-elf-sleep-quality.md 10 条专项——同属”按出口排序”的下一批。
    5. 不要再做:词元重叠类反向检测器(§4 已证否)。

    8. 铁律遵守

    • 写入前 acquire、写入后 release(WB-FIVEPOOLS);
    • frontmatter 六字段齐全;高风险主题 disclaimer_required: true + risk_level: medium;
    • 未标 accepted;未直写 779/782(本文未修改 779 候选清单,只提建议);
    • 新写文件的核验锚点全部 PubMed/Crossref 回查通过(§6.1,13 条,含 1 条不存在号的确认);
    • 本轮对既有正文写入 = 0(4 个 intake 候选文件、skin-depth-reference-table.md、全部受污染文件一字未动)。

      免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

    证据与局限

    本文证据等级为 B,属 measurable,结论限于研究层面的推断,不构成对个体情形的判定。不可靠的文件,恰好是最接近 779 的那一个。

    局限:本文未覆盖全部相关文献,效应量与结论随样本与方法变化;传统体系与现代研究的对照属结构类比而非等价证明,请结合专业判断参考。

    Pool 3 #192: 779 Intake Candidate Citation Completeness Special – Citation Quality of Highest Priority Files is Below Pool Average

    Evidence level: B

    Pool ③ · Topic #192 (Newly Registered in This Round) | Evidence Level B | Status draft
    Writing into existing text in this round = 0

    One-Sentence Conclusion

    The most unreliable cited document is paradoxically the one closest to the 779 final review draft.

    In Round 8, governance focused on aggregating totals across the “63 documents / 254 entries” bucket; this round adopts a different validation perspective, auditing only the 4 documents involved in the 779 intake candidates. The result is that 3 of the 4 contain citation issues, with enso-ci-coupling-multireanalysis-robustness.md (Pool ② N14, registered as a candidate in the 779 intake list) having all 3 references fabricated.

    Additionally, we record a negative outcome: the proposed “reverse validator” failed under mixed Chinese-English corpora (94.4% of articles have ≤2 English topic tags), has been deployed but flagged as “Chinese corpus unavailable,” yielding a net gain ≈0. Do not retrace this path in the next round.

    1. Why Shift Perspective: Total Counts and Binning Cannot Address “Which Is Most Critical”

    Round 8, P3-GOV-002, categorized the 63 files into three groups—P0-a (17), P0-b (18), and P0-c (28)—based on the number of hits per file. This sorting implicitly assumes that higher hit counts indicate greater risk.

    However, the ultimate exit point for the Five Pools is 779 intake → final review → lab online. The critical factor is not “which file contains the most errors,” but rather “which file is currently moving toward 779.” A draft with 17 errors poses an internal risk of misleading stakeholders if it remains confined to 03-sites/ and does not leak. Conversely, a file with only 3 errors poses the risk of introducing fabricated citations into the final review process if it has already entered the 779 candidate list.

    Therefore, this round prioritizes an action not taken in Round 8: intersecting the 779 candidate list with the citation audit results.

    2. Intersection Results: 4 Candidate Files

    Referencing 779-HolyRange-final/00-governance/99-intake-candidate-list.md, there are currently 6 lines of candidates; after deduplication, they map to 4 source files:

    # File Pool Intake Line Status
    1 consciousness-external-physical-effect-risk-statement.md ① P1-EXT-PHYS-BOUND candidate
    2 dosha-qi-rsa-falsification-framework-methodology.md ⑤ P5-DOSHA-QI-RSA-FALSIFICATION-METHOD candidate
    3 dosha-qi-rsa-falsification-preregistration-specification.md ⑤ P5-DOSHA-QI-RSA-FALSIFICATION-PREREGISTRATION-SPEC candidate
    4 enso-ci-coupling-multireanalysis-robustness.md ② P2-ENSO-CI-N14 candidate (Pending Codex review; evidence level rated B+)

    These 4 files constitute the entirety currently advancing toward the final version. Within Pool ③’s own 700+ articles, none appear on this list. This explains why Pool ③’s governance has consistently been “unending” in the first eight rounds: it addresses the portion that does not flow out.

    3. File-by-File Findings

    3.1 ⛔ enso-ci-coupling-multireanalysis-robustness.md: All 3 References Fabricated

    This represents the highest-level finding for this round. The manuscript is a candidate intake for Pool ② N14, marked in the backlog with evidence_level=B+, and listed in the 779 checklist as “Pending Codex Review (Evidence Strength B+)”. All three of its references are fabricated: author, title, journal, volume, issue, and page numbers all mismatch real records:

    # Claimed Citation PubMed Real Record (PMID) Verdict
    1 Yang Y, et al. (2022). Comparison of daily air temperature from ERA5, JRA-55, and MERRA-2 reanalyses over China. Atmospheric Research, 277, 106307. PMID 36155626 Liao WI, A Mouse Model of Orotracheal Intubation and Ventilated Lung Ischemia Reperfusion Surgery, J Vis Exp 2022 ❌ Completely Incorrect
    2 Liu Y, et al. (2023). Decadal variation in the ENSO–East Asian winter monsoon teleconnection: a multi-reanalysis assessment. Climate Dynamics, 61(7), 3581–3598. PMID 37518421 Espinasa L, Erratum: … Mioce…, Zootaxa 2023 ❌ Completely Incorrect
    3 Bock L, et al. (2021). Quantifying progress across different CMIP phases with the global mean surface temperature and Earth energy imbalance. Earth System Dynamics, 12(4), 1067–1083. PMID 34099113 Freeborn RA, The role of Nrf2 in autoimmunity and infectious disease: Therapeutic possibilities, Adv Pharmacol 2021 ❌ Completely Incorrect

    Most glaring detail (indicating this is not random fabrication): The volume, issue, and page number for Reference 1, Atmospheric Research 277, 106307, is real. Crossref verification of 10.1016/j.atmosres.2022.106307 confirms it is indeed Volume 277, Article 106307 in Atmos Res. However, the actual paper is by Zhang S, Ren, Ren, KealdrupTysa, titled “Linkage of extreme temperature change with atmospheric and locally anthropogenic factors in China mainland” (2022).

    In other words: The volume, issue, page number, journal name, and year all match; the topic also aligns (China temperature trends). Only the author “Yang Y” and the title “Comparison of daily air temperature from ERA5…” are fabricated. This is a more subtle error than “random fabrication”—journal volume checks pass, topic relevance passes, but only a word-for-word title comparison exposes it.

    Amplified Impact: These three references are not merely decorative at the end of the text. They are used in §§3.1–3.3 to support substantive arguments with specific numerical values:

    Citation Location Numerical Conclusion Derived from Citation
    §3.1 (36155626) “1980–2020 multi-source annual T2m trends +0.27 to +0.34 °C/decade, cross-product consistency r > 0.85”
    §3.2 (37518421) “ENSO–EAWM correlation two decadal shifts at late 1970s/late 2000s, cross-product ±0.05 uncertainty”
    §3.3 (34099113) “Global land warming uncertainty ±0.04 °C/decade, China ±0.06” → Used to revise “41% warming contribution” judgment to “reasonable range 34–47%”

    Reference 3 is most severe: A fabricated citation was used to rewrite the manuscript’s core numerical conclusion (41% → 34–47%), and this revised judgment will be carried into the final version of document 779.

    Evidence Level Issue: Both the backlog and the 779 checklist mark this as B+. Per Charter §2, E2 requires “multiple verified mechanisms assembled”; the premise of B+ fails because all citation nodes are unverifiable. Recommendation: Change intake status to “References Pending Reconstruction, Hold Review” and downgrade evidence level to C (or mark as disputed) until the three citations are replaced with real literature.

    ⚠️ Clarification: This does not mean all conclusions in the manuscript are incorrect. The evidence chain from Pool ② N12→N18 relies on internal recalculation and multi-source reanalysis for self-consistency; its core logic does not depend on these three external citations. However, these three citations bear the responsibility for the “external cross-validation” link—this link is currently empty.

    3.2 ⛔ dosha-qi-rsa-falsification-preregistration-specification.md (Pool ⑤): Citation Mismatch + 1 Fake PMID

    In the five bibliography lines of this manuscript, a line-by-line comparison reveals:

    PMID Claimed Citation Real Record Verdict
    31234475 “Nosek et al., 2018 preregistration principles review“ Godyń D, Effects of Environmental Enrichment on Pig Welfare—A Review, Animals 2019 ❌ Author/Topic Completely Incorrect
    33494095 “Task Force HRV Guidelines Update“ Johnson BA, Loss of furin cleavage site attenuates SARS-CoV-2 pathogenesis, Nature 2021 ❌ Author/Topic Completely Incorrect
    29269119 “Cheung et al., 2017 cross-cultural construct equivalence latest methodology” Ram Manohar P, An exploratory clinical study… heart rate variability… Ayurveda, J Ayurveda Integr Med 2018 ❌ Author/Year Incorrect (This PMID is correctly used elsewhere, see §3.4)
    36106418 “Cheung & Rensvold, 2002, ΔCFI 0.01 / ΔRMSEA 0.015 dual thresholds” Rani R, HRV during head-up tilt shows inter-individual differences, Physiol Rep 2022 ❌ Author/Year Completely Incorrect
    8598068 “Chen, 2007 ΔCFI 0.01 threshold” Heart rate variability: standards of measurement…, Circulation 1996 (Task Force, no individual author) ❌ Author/Year Completely Incorrect
    34479992 “Putnick & Bornstein, 2016 measurement equivalence systematic review” No Record in PubMed (confirmed empty via esummary + efetch) ❌ Fake PMID

    Note on the nature of the last entry: 34479992 is a “non-existent” PMID identified in Round 7. The same PMID was assigned two other identities (“Chung 2021 TCM Syndrome HRV”) in two files in Pool ③. This manuscript hits it again in a 779 candidate file—the fake PMID has spread to the layer closest to the final version.

    Another structural error: 8598068 is the HRV measurement standard (Task Force 1996), yet it is used to support the “ΔCFI 0.01 threshold.” The ΔCFI threshold is typically attributed to Cheung & Rensvold (2002). These are entirely different topics, indicating the cited author/topic was written based on “what should be cited” rather than retrieved from a literature database.

    3.3 ⛔ dosha-qi-rsa-falsification-framework-methodology.md (Pool ⑤): Homogeneous Mismatch

    Another deployment of the same batch of PMIDs:
    – 29269119 is marked as “Manohar 2018 Three-Dosha Physiology” in the Tridosha questionnaire line → Correct (First author Ram Manohar P, 2018);
    – The same PMID is marked as “Cheung et al., 2017” in another text (§3.2) → Incorrect;
    – 34479992 is marked as “Chung 2021 TCM Syndrome HRV” → PMID does not exist.

    → The same PMID is correct in one paper but incorrect in another, consistent with the pattern repeatedly verified in Rounds 2, 7, and 8 (cross-subline misappropriation + real PMID fake bibliography). This is the fourth independent reproduction of the pool’s primary failure mode.

    3.4 ⛔ consciousness-external-physical-effect-risk-statement.md (Pool ①): 1 Fake Bibliography Entry

    PMID Claimed Citation Real Record Verdict
    16822162 Bösch H, Steinkamp F, Boller E. Examining psychokinesis… Psychol Bull 2006;132(4):497-523 Bösch H, Examining psychokinesis: the interaction of human intention with random number generators, Psychol Bull 2006 ✅ Correct
    16822163 Comment on Bösch, Steinkamp, and Boller (2006). Psychol Bull 2006 Wilson DB, On blowing trumpets to the tulips…, Psychol Bull 2006 ✅ Correct (not first author but correct reference)
    29144112 “May EC, Utts JM, Tressoldi PE, et al. Intentional Observer Effects on Quantum Randomness: A Bayesian Analysis. Explore (NY). 2018“ JEH Quiz #5: Presence of Household Mold, Children’s Respiratory Health, and School Absenteeism, J Environ Health 2017 (No authors listed) ❌ Completely Incorrect

    → This file’s primary anchor is clean (two Bösch/Wilson entries are real and relevant); only the third entry is wrong. This is the lightest of the four files—also indicating that “having problems” does not mean “file invalid,” and graded disposition is appropriate.

    4. Negative Result in This Round: Reverse Validator Fails Under Chinese Corpus (Must Record)

    The first item in the plan for the next round registered in Round 8 was to “build a reverse validator (verify real titles by searching for characteristic terms in the text, covering bare lists and multiple numbers side-by-side)”. We executed this in the current round. Conclusion: This direction is not viable for the Wu Chi corpus.

    4.1 Execution

    New tool 08-ops/_tools/citation-reverse-check.py (pure standard library, --selftest passed 2/2). Approach: Start from real literature, take PubMed real titles → calculate overlap with article MeSH terms → flag as suspicious if zero overlap.

    4.2 Results: 1,142 Hits in Full Pool, 82% Language False Positives

    Metric Value
    Scanned files 746
    Citation entries 13,581
    Bare list format 4,459 (32.8%) ← Blind spot for claim-check; confirmed existence
    Scoreable entries 1,256
    No English terms extracted from article MeSH 533 files (71.4%)
    Zero-overlap hits 1,135

    Bucketed by “Number of English MeSH terms in host file”:

    English MeSH Count Hits Proportion
    1 740 65%
    2 192 17%
    3 131 11%
    4 23 2%
    ≥5 56 5%

    94.4% of articles (704/746) have ≤2 English MeSH terms. The reason is straightforward: the main text is in Chinese, while the cited real titles are in English → MeSH terms inevitably have zero intersection.

    Typical False Positives (obvious to human inspection that the citation is correct):

    acupoint-meridian-functional-validation.md Citation:
      「Acupuncture, the limbic system, and the anticorrelated networks of the brain」
      「Does the Acupoint Specificity Exist? Evidence from Functional Neuroimaging Studies」
    The article's English MeSH terms total only 1 (fmri) → Both papers flagged as 0 overlap ❌
    

    4.3 Conclusion: Net Increment ≈ 0, Do Not Pursue This Path Again

    • For files with bibliography lines, citation-claim-check.py can already judge T1/T2, and is more accurate (re-ran on ENSO files in this round; all 6 matches correct, consistent with manual verification).
    • For bare lists (32.8%), the MeSH overlap method is inapplicable under Chinese corpus conditions; to cover them, the correct approach is not MeSH comparison but “whether numbers in bare lists can locate authors for verification” — this essentially still requires human effort.
    • Therefore: #193 (Reverse Validator) completed this round but judged as a failed direction; the tool is retained and explicitly marked in file headers as “inapplicable for Chinese corpus” to prevent re-creation in the next round.

    This is the third consecutive lesson of this type: Round 7 “unextracted lines ≈ 0”, Round 8 “must report denominator and coverage”, this round “cross-language MeSH method fails”. The common point is — the tool metrics appear to run smoothly, but the metrics themselves do not measure what we intend to measure.

    5. Determination and Disposal Recommendations

    5.1 Grading (Ranked by “Distance to 779,” not hit count)

    Level Target Basis Recommended Action
    I-1 (Highest) enso-ci-coupling-multireanalysis-robustness.md (Pool ② N14, 779 candidate) 3/3 references fabricated; one used to rewrite core values (41%→34–47%); volume/page numbers real but author titles fabricated Suspend 779 review for ENSO N14; downgrade evidence_level B+ → C / disputed; reconstruct or delete the three corresponding citations based on real literature; re-evaluate level after reconstruction
    I-2 dosha-qi-rsa-falsification-preregistration-specification.md (Pool ⑤, 779 candidate) 5 bibliographic entries: 4 incorrect + 1 fake ID (34479992) Change ID or replace literature (§3.2 table can be used directly); delete fake ID; do not alter the main argument structure
    I-3 dosha-qi-rsa-falsification-framework-methodology.md (Pool ⑤, 779 candidate) Same batch ID mismatch, includes fake ID 34479992 Same as I-2; process together with I-2
    I-4 (Lightest) consciousness-external-physical-effect-risk-statement.md (Pool ①, 779 candidate) Primary anchor correct on 2/3; only one fake bibliography (29144112) Single-point replacement sufficient; do not touch others

    5.2 Correction Recommendations for P3-GOV-002

    The three batch bucketing in Round 8 (P0-a 17 / P0-b 18 / P0-c 28) is currently sorted by hit count. This paper recommends adding an “Exit Priority” layer before the batches:

    P-1 Layer (Preceding P0) = 779 intake candidate files. Total 4, of which 3 require disposal.
    Rationale: These 4 are the only ones at risk of external leakage; the remaining 59 remain within-site with a lower risk tier.
    Cost: Only 4 files, far fewer than the 17 in P0-a.

    In other words: If JOE reviews only one item, it should be these 4, not the 17.

    5.3 Red Flags Requiring JOE Decision

    1. Should ENSO N14 suspend 779 review? (Recommended: Suspend, downgrade to C, re-review after citation reconstruction) — This is the most urgent item for this round.
    2. Should authorization be granted to execute I-1~I-4 disposal for the 4 intake candidates? (Change IDs/literature/delete fake IDs; do not delete argument paragraphs; all actions are reversible).
    3. Should the 779 candidate list add a citation_status column? (Values: verified / disputed / unverified), making citation status a mandatory field for intake rather than waiting for audit discovery.

    6. Citations and Anchors

    6.1 Verification Anchors Used in This Text (All Verified via PubMed/Crossref)

    Identifier Verified Citation Purpose
    PMID 36155626 Liao WI, A Mouse Model of Orotracheal Intubation…, J Vis Exp 2022 Refutes ENSO citation①
    PMID 37518421 Espinasa L, Erratum…, Zootaxa 2023 Refutes ENSO citation②
    PMID 34099113 Freeborn RA, The role of Nrf2 in autoimmunity…, Adv Pharmacol 2021 Refutes ENSO citation③
    DOI 10.1016/j.atmosres.2022.106307 Zhang S, Ren, Ren, KealdrupTysa, Linkage of extreme temperature change…China mainland, Atmos Res 277:106307 (2022) Confirms “volume/page accuracy but author/title discrepancy”
    PMID 31234475 Godyń D, Effects of Environmental Enrichment on Pig Welfare, Animals 2019 Refutes dosha-prereg citation
    PMID 33494095 Johnson BA, Loss of furin cleavage site attenuates SARS-CoV-2…, Nature 2021 Refutes dosha-prereg citation
    PMID 29269119 Ram Manohar P, An exploratory clinical study…HRV…Ayurveda, J Ayurveda Integr Med 2018 Refutes (correctly cited elsewhere under same number)
    PMID 36106418 Rani R, HRV during head-up tilt…, Physiol Rep 2022 Refutes dosha-prereg citation
    PMID 8598068 Heart rate variability: standards of measurement…, Circulation 1996 Refutes (misapplied ΔCFI threshold to HRV standards)
    PMID 29144112 JEH Quiz #5: Presence of Household Mold…, J Environ Health 2017 Refutes consciousness citation
    PMID 16822162 Bösch H, Examining psychokinesis…, Psychol Bull 2006 Confirms (primary anchor correct)
    PMID 16822163 Wilson DB, On blowing trumpets to the tulips…, Psychol Bull 2006 Confirms
    PMID 34479992 Does not exist (confirmed empty via esummary + efetch) False identifier; matched in pool⑤ candidate file #779

    6.2 Evidence Level Explanation

    evidence_level: B. Rationale: The core determination (citation mismatch against actual records) relies on deterministic factual comparison independent of models; however, the scope of the conclusion is constrained by the “bare list” where 32.8% cannot be programmatically adjudicated, preventing an upgrade to A. The tool failure conclusion in §4 is reproducible via direct measurement.

    7. Next Round Actions

    1. I-1 (Highest Priority): Reconstruct the three citations in the ENSO text based on authentic literature. This requires verifying real sources regarding “ERA5/JRA-55/MERRA-2 China Temperature Trend Comparison,” “ENSO-EAWM Multi-Reanalysis Decadal Shifts,” and “Uncertainty in Reanalysis Long-Term Trends.” Replace the erroneous references and re-evaluate the rating from B+ to B or C. This addresses a critical candidate for the 779 classification.
    2. I-2/I-3: Replace the identifiers in the two dosha files (§3.2 table provides correct values). Simultaneously remove the three instances of the false identifier 34479992.
    3. I-4: Replace the single point in the consciousness text with 29144112.
    4. #47 (Pool ④): Execute the ten specialized tasks for groundwater-aquifer-elf-sleep-quality.md. These fall under the next batch of “sort by export.”
    5. Do Not Proceed: Do not run word-embedding overlap reverse detectors (§4 already refutes their validity).

    8. Ironclad Compliance Rules

    • Acquire before writing and release after writing (WB-FIVEPOOLS);
    • Ensure all six fields in frontmatter are complete; set disclaimer_required: true and risk_level: medium for high-risk topics;
    • No entries marked as accepted; do not directly write 779 or 782 (this text did not modify the 779 candidate list, only offered suggestions);
    • Verification anchors for all newly written files have passed PubMed/Crossref re-checks (§6.1, 13 items, including confirmation of one non-existent ID);
    • Zero writes to existing body text in this round (four intake candidate files, skin-depth-reference-table.md, and all contaminated files remained untouched).

    Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.