Evidence level: B
在一段恒暗记录里同时看到 24 小时和 29.5 天两个谱峰,只能说明”有两种节律”,不能说明”有两个钟”——独立振荡器和从属计数器会给出同一个谱图,这两者只在周期协变的方向上分开。
正文
上一篇《恒暗监测协议》讲的是怎么把这套实验做出来——多长、多暗、几个通道、几个人。这一篇只问一件事:记录拿到手之后,凭什么说那两个周期是两个独立的钟,而不是一个钟拖着另一个走?
同一个问题,动物里已经有两个相反的答案
海洋环节虫 Platynereis 给出的答案是”独立”:它前脑里既有昼夜钟也有近月钟,用 CK1δ/ε 抑制剂把昼夜钟的振荡打掉,近月钟照样按自己的节拍走(PMID 24075994)。变的是昼夜行为,不变的是月钟。
摇蚊 Clunio 给出的却是”从属”:它的自由运行半月周期会跟着光照周期长度线性变化;用持续光照把昼夜钟打散,半月节律也跟着消失(PMID 38764210)。它的”月钟”根本不是独立振荡器,而是在数昼夜周期——数够一定次数才发作一次。
同一个问题两种答案,说明”近月钟独立吗”不是靠定义能回答的,得靠实验结构回答。人类属于哪一支,目前完全没有数据(PMID 34050958、PMID 38336520)。
坑一:两个谱峰不等于两个钟
这是最要命的一条。把长记录做周期图,24 小时附近和 29.5 天附近各看到一个峰,很多人就此宣布”发现两个钟”。但两种结构都会给出这个结果:
– 两个独立振荡器:各跑各的周期,叠加成一串拍。
– 一个从属计数器:月信号靠数昼夜周期触发,昼夜周期一变,月周期跟着变。
两者谱图长得几乎一样。要分开,只能看周期协变的方向:人为改一个自变量(改明暗比例、强制拉长作息周期),再去看另一个周期动不动。Clunio 就是这么被识破的。
而恒暗隔离实验里恰恰没有第二个自变量可动。所以这类实验原理上就答不了”独立”,只能答”两者是否同时存在”。要答”独立”,需要”恒暗 + 一个可控的非 24 小时光照周期”的组合,即在月尺度上重做强制失同步(FD)——昼夜模板见 PMID 36376588。
坑二:时长不够时,29.5 天会被自己的近邻淹掉
周期图对长周期的检出有硬门槛:记录要覆盖好几个完整周期,谱峰才窄得住;否则峰宽化、能量泄漏到邻近频率,报出来的”29.5 天”可能只是噪声或昼夜周期的谐波——周期图误估周期有专门讨论(PMID 33406069),不等间隔数据上 Lomb-Scargle 的检出困难与其回应是经典争论(PMID 11665689、PMID 11594366),改进算法与小波各有代价(PMID 27482382、PMID 23382592)。
人类恒暗记录正好卡在这条线上。2026 年 Lombrives 洞穴 40 天 13 人的研究是目前最完整的一次(PMID 42009612):睡眠-觉醒周期从隔离前 24.1 小时漂到洞穴内 31.6 小时;核心体温周期在 19.5 至 86.4 小时之间大幅波动;出现”早期内部失同步、之后部分重新对齐”;群体同步性逐渐崩解。它证明人体多通道在无外部线索时会自发失去统一节拍。但 40 天只有约 1.35 个近月周期,任何 29.5 天成分的估计都不成立。
更麻烦的是混淆:自由运行下昼夜周期被拉长到 31.6 小时,正向 29.5 小时靠拢。若某个受试者的昼夜周期恰好漂到那个位置,”月钟谱峰”与”被拉长的昼夜谱峰”会在同一处重合——这种重合用谱图分不开。历史单例(PMID 4955822,Reinberg 1966 一名女性三个月地下隔离的直肠温度记录)也躲不开:n=1、时长短、结论依赖分析先验。
坑三:第一个长程数据集把”月节律=月经”这个前提拆了
2026 年 EBioMedicine 用可穿戴设备(n=623,最长 4 年)分析心率的多日节律(PMID 42628374):70% 的人检出显著多日节律,36% 检出两条以上;在 4-45 天这个带里,个体能聚成三类”多日时序型”——约 7 天、约 25-35 天、超过 35 天。月级节律在有月经的女性身上与月经同步,但在男性和绝经后女性身上同样出现。
这一条分量很重:若月级心率节律的唯一来源是卵巢激素,男性不该有。所以至少存在一个非生殖源的月级成分,这是”近月钟”这条线目前最有力的正向证据。
但这份数据还暴露了一件事:“月级”是一个很宽的频带。25 到 35 天都算,29.5 天只是其中一点。所谓”人际差异”里,多少是真正的周期差异,多少只是同一个钟在不同人身上表达了不同长度,靠横断面聚类分不开——月经周期长度的个体间与个体内变异本来就极大(PMID 37248288、PMID 32442161、PMID 32104920)。昼夜侧的同一问题早有定论:本征周期是个体特征、因人而异(PMID 9554574、PMID 17684566、PMID 22894720、PMID 23899423)。
落到易学上
《素问·六微旨大论》里有八个字常被读过去:”天气始于甲,地气始于子,子甲相合,命曰岁立。“天有天的六十日循环,地有地的六十日循环,两套各自计数、互不推进;”相合”不是一种状态,是一个事件——两者的起点碰到一起,那一年才成立。
《易·系辞》讲得更像算术:”乾之策二百一十有六,坤之策百四十有四,凡三百有六十,当期之日。”两个数凑出一个周期,靠的是耦合,不是谁等于谁。
古人这两句其实已经说透了:一个”岁”不是一条钟走出来的,是两套独立的计数器在特定相位上撞出来的。 恒暗隔离实验想做的,恰恰是把这两套计数器分开测。麻烦在于,我们现在的仪器能同时看到两个峰,却看不到那根”谁在牵谁”的线——这就是当前真正的空白。
证据与局限
已验证: 环节虫 Platynereis 的昼夜钟与近月钟可在振荡层面解耦——抑制 CK1δ/ε 打掉昼夜钟后近月钟仍保留(PMID 24075994);摇蚊 Clunio 的自由运行半月周期由昼夜周期计数决定,半月周期随 T-cycle 长度线性变化,昼夜钟被打散后半月节律随之消失(PMID 38764210);人类 40 天洞穴群体隔离中昼夜节律显著自由运行(24.1→31.6 h)、核心体温周期 19.5-86.4 h 高度可变、出现早期内部失同步与群体同步崩解(PMID 42009612);人类可穿戴心率数据(n=623,最长 4 年)中 70% 检出多日节律、36% 检出两条以上,月级节律在男性与绝经后女性中同样存在(PMID 42628374);昼夜本征周期是个体特征并存在个体间差异与可塑性(PMID 9554574、PMID 17684566、PMID 22894720、PMID 23899423、PMID 32619243、PMID 30926824);月经周期长度存在巨大个体间与个体内变异(PMID 37248288、PMID 32442161、PMID 32104920);FD 协议可解耦昼夜成分并估计本征周期(PMID 36376588);周期图对长周期存在误估与泄漏风险,Lomb-Scargle 与小波方法各有代价(PMID 33406069、PMID 11665689、PMID 11594366、PMID 27482382、PMID 23382592);已有的近月钟综述与存在性论证(PMID 28553258、PMID 32198116、PMID 34050958、PMID 38336520、PMID 27871090、PMID 35622889、PMID 30767890、PMID 42380750、PMID 34028873)。
部分验证: 人类近月自由运行的直接记录仅有一次三个月地下隔离的单例直肠温度记录(PMID 4955822)与一次 40 天洞穴隔离(PMID 42009612),前者 n=1、后者不足 1.5 个近月周期;”人类近月钟独立于昼夜钟”尚无任何扰动式检验;动物界的近月钟机制已被证明并非单一类型(独立型与计数型并存),人类归属未定。
待验证: 对昼夜钟施加可控扰动(非 24 小时光照周期、FD、光脉冲)后,近月成分的周期是否跟着改变(区分独立振荡器与从属计数器);个体昼夜本征周期 τ 与近月周期是否相关(计数机制的证伪预测);在恒暗记录中,29.5 天窗与 24 小时谐波、时长不足导致的谱峰宽化能否被稳定分离;哪一生理通道可作为近月钟的可靠读出。
相关阅读
证据等级: B
参考文献: PMID 24075994, PMID 38764210, PMID 42009612, PMID 42628374, PMID 4955822, PMID 36376588, PMID 33406069, PMID 11665689, PMID 11594366, PMID 27482382, PMID 23382592, PMID 9554574, PMID 17684566, PMID 22894720, PMID 23899423, PMID 32619243, PMID 30926824, PMID 37248288, PMID 32442161, PMID 32104920, PMID 34028873, PMID 42380750, PMID 34050958, PMID 38336520, PMID 28553258, PMID 32198116
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。
Two Peaks Do Not Equal Two Clocks: Are Circadian and Lunar Rhythms Independent in Constant Darkness?
Evidence level: B
Observing both a 24-hour and a 29.5-day spectral peak in a constant darkness record indicates only that “two rhythms exist,” not that “two clocks exist”—both independent oscillators and subordinate counters yield identical spectral outputs; these mechanisms are distinguished solely by the direction of period covariance.
Main Text
The previous article, “Constant Darkness Monitoring Protocol,” detailed how to conduct this experiment—duration, darkness levels, channel count, and participant numbers. This article addresses a single question: Upon obtaining the record, on what basis can we claim those two cycles represent two independent clocks rather than one clock dragging another?
Two Opposing Answers Already Exist in Animals
The marine annelid Platynereis provides an answer of “independence”: it possesses both a circadian clock and a circalunar clock in its forebrain. Applying CK1δ/ε inhibitors to abolish the circadian oscillation leaves the circalunar clock running on its own beat (PMID 24075994). The circadian behavior changes, but the lunar clock remains unchanged.
The mosquito Clunio, however, provides an answer of “subordination”: its free-running semilunar period changes linearly with the length of the light-dark cycle; when the circadian clock is disrupted by continuous light, the semilunar rhythm disappears alongside it (PMID 38764210). Its “lunar clock” is not an independent oscillator but rather counts circadian cycles—triggering once after a certain number of cycles are completed.
Two answers to the same question indicate that whether the “circalunar clock is independent” cannot be answered by definition alone but requires experimental structural evidence. Currently, there is no data for humans (PMID 34050958, PMID 38336520).
Pitfall One: Two Spectral Peaks Do Not Equal Two Clocks
This is the most critical point. When performing periodograms on long records, observing peaks near 24 hours and 29.5 days often leads to the announcement of “discovery of two clocks.” However, both structural models produce this result:
– Two independent oscillators: Each runs its own period, superimposing into a beat pattern.
– One subordinate counter: The lunar signal triggers by counting circadian cycles; as the circadian period changes, the lunar period follows.
Their spectral plots look almost identical. To distinguish them, one must examine the direction of period covariance: artificially alter an independent variable (e.g., light-dark ratio or force-feeding a non-circadian period), then observe whether the other period shifts. Clunio was thus exposed this way.
In contrast, constant darkness isolation experiments lack a second movable independent variable. Therefore, such experiments cannot theoretically answer “independence” but can only address “whether both coexist.” To answer “independence,” a combination of “constant darkness + a controllable non-24-hour light cycle” is required, essentially re-doing forced desynchronization (FD) at the lunar scale—circadian templates see PMID 36376588.
Pitfall Two: Insufficient Duration Causes the 29.5-Day Peak to Be Swamped by Neighbors
Periodograms have hard thresholds for detecting long cycles: records must cover several complete cycles for peaks to narrow; otherwise, peak broadening occurs with energy leakage into neighboring frequencies. The reported “29.5 days” may merely be noise or a harmonic of the circadian cycle—periodogram misestimation of long cycles has been discussed (PMID 33406069). Detection difficulties with non-uniformly spaced data and responses via Lomb-Scargle represent a classic debate (PMID 11665689, PMID 11594366), with improved algorithms and wavelets each carrying costs (PMID 27482382, PMID 23382592).
Human constant darkness records sit precisely on this line. The 2026 Lombrives Cave study (40 days, n=13) remains the most complete to date (PMID 42009612): sleep-wake cycles drifted from 24.1 hours pre-isolation to 31.6 hours inside the cave; core body temperature cycles fluctuated wildly between 19.5 and 86.4 hours; “early internal desynchronization followed by partial realignment” occurred; group synchrony gradually collapsed. It proves that human multi-channel systems spontaneously lose unified pacing without external cues. However, 40 days covers only ~1.35 circalunar cycles; any estimation of a 29.5-day component is invalid.
Compounding this is confusion: under free-running conditions, the circadian period lengthens to 31.6 hours, approaching the 29.5-hour mark. If a subject’s circadian period drifts exactly to that position, the “lunar clock spectral peak” and the “lengthened circadian spectral peak” will coincide at the same location—this coincidence cannot be separated by periodograms. Historical single cases (PMID 4955822, Reinberg 1966, a female subject’s rectal temperature record during three months of underground isolation) also evade this issue: n=1, short duration, conclusions rely heavily on analytical priors.
Pitfall Three: The First Long-Term Dataset Decouples the Premise “Lunar Rhythm = Menstrual Cycle”
The 2026 EBioMedicine study using wearable devices (n=623, up to 4 years) analyzed multi-day heart rate rhythms (PMID 42628374): 70% showed significant multi-day rhythms, and 36% showed two or more; within the 4-45 day band, individuals clustered into three types of “multi-day temporal patterns”—approximately 7 days, approximately 25-35 days, and over 35 days. Lunar-level rhythms synchronize with menstruation in women with periods, but also appear in men and postmenopausal women.
This carries significant weight: if the sole source of lunar-level heart rate rhythms were ovarian hormones, men should not exhibit them. Thus, at least one non-reproductive lunar component exists, representing the strongest positive evidence for the “circalunar clock” line so far.
However, this data also reveals another issue: “Lunar-level” is a broad frequency band. Both 25 and 35 days count; 29.5 days is merely one point within it. Within “inter-individual differences,” how much represents true period variation versus the same clock expressing different lengths across individuals cannot be separated via cross-sectional clustering—inter-individual and intra-individual variability in menstrual cycle length is already massive (PMID 37248288, PMID 32442161, PMID 321064920). The same issue on the circadian side has already been settled: intrinsic periods are individual characteristics, varying between persons (PMID 9554574, PMID 17684566, PMID 22894720, PMID 23899423).
Returning to Yijing Studies
In the Su Wen · Six Micros Great Theory, eight characters are frequently read: “Weather begins at Jia, Earthly Qi begins at Zi; Zi and Jia combine; this is called the Year Establishment.” Heaven has its own sixty-day cycle, Earth has its own sixty-day cycle; both count independently without advancing each other; “combination” is not a state but an event—when their starting points meet, the year is established.
The I Ching · Xici speaks more like arithmetic: “The strategies of Qian total 216, those of Kun total 144; together they make 360, corresponding to the days of a year.” Two numbers combine to form a cycle through coupling, not equality.
These two ancient sentences actually reveal the truth: A “year” is not traced by a single clock but emerges from two independent counters colliding at a specific phase. Constant darkness isolation experiments aim precisely to separate and measure these two counters. The difficulty lies in our current instruments seeing both peaks simultaneously but failing to see the thread of “who pulls whom”—this is the true current gap.
Evidence and Limitations
Verified: In Platynereis, circadian and circalunar clocks can be decoupled at the oscillatory level—inhibiting CK1δ/ε abolishes the circadian clock while preserving the circalunar one (PMID 24075994); in Clunio, the free-running semilunar period is determined by counting circadian cycles, varying linearly with T-cycle length, and disappearing when the circadian clock is disrupted (PMID 38764210); in human 40-day cave group isolation, circadian rhythms showed significant free-running (24.1→31.6 h), core body temperature cycles varied highly between 19.5-86.4 h, early internal desynchronization and group synchrony collapse occurred (PMID 42009612); wearable heart rate data in humans (n=623, up to 4 years) showed multi-day rhythms detected in 70%, two or more in 36%, and lunar-level rhythms exist in both men and postmenopausal women (PMID 42628374); intrinsic circadian periods are individual characteristics with inter-individual differences and plasticity (PMID 9554574, PMID 17684566, PMID 22894720, PMID 23899423, PMID 32619243, PMID 30926824); menstrual cycle length exhibits massive inter-individual and intra-individual variability (PMID 37248288, PMID 32442161, PMID 321064920); FD protocols can decouple circadian components and estimate intrinsic periods (PMID 36376588); periodograms carry risks of misestimation and leakage for long cycles, with Lomb-Scargle and wavelet methods each having costs (PMID 33406069, PMID 11665689, PMID 11594366, PMID 27482382, PMID 23382592); existing reviews and existence arguments for circalunar clocks (PMID 28553258, PMID 32198116, PMID 34050958, PMID 38336520, PMID 27871090, PMID 35622889, PMID 30767890, PMID 42380750, PMID 34028873).
Partially Verified: Direct records of human free-running circalunar rhythms exist only once: a three-month underground isolation single-case rectal temperature record (PMID 4955822) and a 40-day cave isolation study (PMID 42009612); the former has n=1, the latter covers less than 1.5 circalunar cycles; “human circalunar clock independent of circadian clock” has yet to undergo any perturbation-based testing; mechanisms of circalunar clocks in animals are proven not to be a single type (both independent and counting types coexist), with human classification undecided.
To Be Verified: Whether the period of the circalunar component changes after applying controllable perturbations to the circadian clock (non-24-hour light cycles, FD, light pulses), distinguishing independent oscillators from subordinate counters; whether there is a correlation between individual intrinsic circadian period τ and the circalunar period (falsification prediction of counting mechanisms); whether spectral peak broadening caused by insufficient duration in constant darkness records can be stably separated from 29.5-day peaks and 24-hour harmonics; which physiological channel can serve as a reliable readout for the circalunar clock.
Related Reading
- Constant Darkness Monitoring Protocol: Does the Human Body Hide a Free-Running 29.5-Day Circalunar Clock?
- After Turning Off the Ovary, Does the Endometrium Still Walk Its Own Clock?
- Finding a 29.5-Day Clock in Human Blood
- Environment-Life Frequency Interface
Evidence Level: B
References: PMID 24075994, PMID 38764210, PMID 42009612, PMID 42628374, PMID 4955822, PMID 36376588, PMID 33406069, PMID 11665689, PMID 11594366, PMID 27482382, PMID 23382592, PMID 9554574, PMID 17684566, PMID 22894720, PMID 23899423, PMID 32619243, PMID 30926824, PMID 37248288, PMID 32442161, PMID 321064920, PMID 34028873, PMID 42380750, PMID 34050958, PMID 38336520, PMID 28553258, PMID 32198116
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.
Leave a Reply