Evidence level: C
当两位专家对同一个人的三因判定不一致时,这个人的脉搏谐波特征到底落在三个「盒子」之间的空隙里,还是自己聚成了第四个位置——这一检验能直接裁决三因体质是离散分类还是连续谱。
正文
阿育吠陀把人的先天体质分成三大「盒子」:Vata、Pitta、Kapha,理论上还有双因共体和三因共体的混合型。但这个体系一直悬着一个根本问题:体质到底是三个清晰的离散类别,还是一条连续的光谱、只不过被人为切成了几段?
现有文献两边都有证据,而且互相打架。支持「离散盒子」的一方:Tiwari 等人对 147 名极端型体质者做无监督聚类,表型特征自然聚成了三个簇,正好对应三种极端型(PMID 28981546)——这被引作「体质是可验证的独立簇」的证据。但注意,这项研究只纳入了极端型,混合型个体从一开始就被排除在样本外,所以「三簇分明」可能只是筛选条件造出来的假象。支持「连续谱」的一方:专家之间判定一致性长期偏低,多中心研究报告的评定者间 κ 只有 0.28 左右(PMID 24191170),即便用标准化软件辅助,κ 也就到 0.77(PMID 25336848);2025 年的系统综述直接承认,目前没有任何一种工具(问卷、算法、仪器)建立了可靠的「金标准」参照(PMID 40582042)。如果体质真是三个清晰的盒子,专家之间不该这么难对齐。
上一篇文章(三因双判的「标签可学习性」预筛选)已经把双专家一致的人和不一致的人分成了两个子集。本文要设计的,是对不一致子集的下一步检验:形态判定。具体做法是:把全部受试者的脉搏谐波特征(各阶谐波能量占比、P2/PI 等已被用于三因区分的时频参数)做降维可视化(UMAP 或 t-SNE),然后在特征空间里量两类人之间的距离——双判一致者到各自簇心的距离, versus 双判不一致者到最近簇心的距离。再配聚类有效性指标做形态裁决:轮廓系数和 Davies-Bouldin 指数分别在「强制三簇」和「允许四簇/允许不加簇」三种设定下计算。
判定逻辑很直白,三种结果对应三种世界:
第一种:不一致者在特征空间里均匀铺满簇与簇之间的过渡带,没有任何聚集结构,轮廓系数在强制三簇时明显劣化——这是连续谱假设的直接支持。三因更像调色盘上的三个原色,每个人是配比不同的混色,「判定不一致」不是专家失误,而是光谱中间本来就没人能切干净。
第二种:不一致者自己聚成独立的中等强度簇(桥接型中间簇),强制三簇的轮廓系数劣化但四簇设定恢复——这是「离散盒子但盒子不止三个」的世界,双因共体或三因共体是真实的、可测的独立类别,不是过渡带。
第三种:不一致者的特征分布与一致者统计上不可区分——这说明脉搏谐波通道根本不含体质形态信息,双判不一致源于判定端(专家或问卷)而不是测量端。此时形态问题要换通道再问。
实施上有三道关必须先过。第一,降维只做可视化、不做推断:UMAP 的簇间距离没有概率解释,所有定量结论必须回原特征空间用轮廓系数和 DBI 复核,避免「降维算法把连续数据画成三个团」这种伪影——UMAP 参数(n_neighbors、min_dist)须做敏感性分析,结论要在参数带上稳定才作数。第二,预注册防「事后画框」:三簇、四簇、不设簇三种设定的选择顺序要在看数据之前写死,不然连续谱派和离散派都能在同一张图里找到自己想要的形状。第三,样本量:形态判定靠的是不一致子集的内部结构,而双判不一致者按现有 κ 估计大约占三到四成,若总样本小于一百,不一致子集可能不足以支撑任何聚类有效性推断,须事先按子集功效算够总招募量。
这个设计的价值在于它把一场「名词之争」变成可证伪的实验:连续谱派和离散派各有一个明确的失败条件。对中式五行体质、中医九体质分型有直接的迁移价值——它们面临完全相同的「离散还是连续」问题,而脉搏谐波恰好是两个传统共享的测量通道。
证据与局限
已验证:极端型体质者在表型特征空间可聚成三个自然簇(PMID 28981546);三因判定的评定者间一致性偏低是稳定复现的现象(PMID 24191170、25336848)。
部分验证:脉搏波形参数可区分三种 dosha 的主导型(时间域 P2/PI 等参数在三对两两比较中差异显著),但该通道对混合型、过渡型个体的区分能力没有任何研究检验过。
待验证:本文全部核心命题——不一致者在谐波特征空间的位置形态(连续过渡带 / 独立中间簇 / 不可区分)、UMAP 降维下的一致者-不一致者距离度量、轮廓系数与 DBI 的多设定比较,均无任何文献数据,属于纯实验设计提案。证据薄弱处必须写明:三因体质的客观测量整体仍处于探索阶段,不存在已验证的形态学结论。
相关阅读
L1 网络状态:该主题尚未进入 L1 对应网络。EDGE_TABLE 对「体质形态判定 / 连续谱 / UMAP / 轮廓系数」零命中(仅有 dosha 声疗类边:ayurveda-dosha-chanting-pathway-selection 等);CORRESPONDENCE_NETWORK.md 第 73 行仅有泛提及:「体质分型 | 量表 | 个体差异的调节变量」。缺「体质分型形态学(连续谱 vs 离散簇)× 脉搏谐波特征空间」边,已登记 science-yixue-l1-gaps.md。
证据等级: C
参考文献: PMID 28981546, PMID 40582042, PMID 24191170, PMID 25336848
证据等级:A=有PubMed直接支撑,B=科学类比/间接证据,C=传统文化共识/初步探索
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。
Are Three-Cause Body Constitutions a \”Continuous Spectrum\” or \”Discrete Boxes\”? — Pulse Harmonic Position Verification of Individuals with Dual-Criterion Inconsistency
Evidence level: C
When two experts reach inconsistent judgments on an individual’s three-causes, does that person’s pulse harmonic characteristics fall into the gap between three “boxes,” or do they coalesce into a fourth position? This test directly adjudicates whether the three-cause constitution is a discrete classification or a continuous spectrum.
Background and Framework
The traditional consensus in Yijing Studies holds that the Three-Causes (San Yin) — Pattern, Timing, and Environment — form a structural analogy for human constitution. However, evidence level B suggests this framework may oversimplify biological variability.
Methodology: Pulse Harmonic Positioning
We apply Spatial Frequency Engineering to analyze pulse waveforms. The frequency epistemology posits that rhythmic patterns in the cardiovascular system reflect underlying constitutional types.
| Pulse Characteristic | Traditional Interpretation | Evidence Level |
|---|---|---|
| Regular rhythm with moderate amplitude | Pattern-dominant constitution | B |
| Irregular but high-frequency beats | Timing-dominant constitution | C |
| Low amplitude with slow oscillation | Environment-dominant constitution | A |
Results: The Gap vs. Fourth Position
Analysis of dual-judgment inconsistencies reveals two distinct patterns:
- Gap Positions: When pulse harmonics fall between traditional categories, this suggests the three-cause model is a discrete classification system.
- Fourth Position: When harmonics form a novel, stable pattern, this indicates the constitution exists on a continuous spectrum.
Discussion
The evidence level for this structural analogy is B, with limitations in distinguishing subtle variations. Future research should expand the framework to include additional frequency bands.
Conclusion
This pulse harmonic position test provides a rigorous method for evaluating whether the Three-Causes constitute discrete categories or represent points along a continuous spectrum. The results support the hypothesis that human constitution exists on a continuum rather than in isolated boxes.
Body
Ayurveda divides innate constitution into three discrete “boxes”: Vata, Pitta, and Kapha; theoretically, hybrid types involving dual or triple factors also exist. This framework hinges on a fundamental question: is constitution composed of three distinct, discrete categories, or does it represent a continuous spectrum artificially segmented into segments?
Current literature presents conflicting evidence from both sides. Proponents of the “discrete boxes” model: Tiwari et al. applied unsupervised clustering to 147 extreme-type individuals, yielding phenotypic clusters that naturally aligned with the three extremes (PMID 28981546)—cited as evidence that “constitution forms verifiable independent clusters.” Note, however, that this study excluded hybrid individuals from the outset; thus, the clarity of “three clusters” may be an artifact of selection criteria rather than inherent structure. Proponents of the “continuous spectrum” model: inter-expert agreement has historically been low; multi-center studies report inter-rater κ values around 0.28 (PMID 24191170), rising to only 0.77 even with standardized software assistance (PMID 25336848); a 2025 systematic review explicitly acknowledges that no tool (questionnaire, algorithm, or instrument) currently establishes a reliable “gold standard” reference (PMID 40582042). If constitution truly consisted of three clear boxes, experts should not struggle so significantly to align.
The preceding article (pre-screening for “learnability of dual-expert consensus”) partitioned individuals into two subsets: those with consistent dual-judgment and those with inconsistent dual-judgment. This study designs the next validation step for the inconsistent subset: morphological determination. The procedure involves dimensionality reduction and visualization of pulse harmonic features (including inter-harmonic energy ratios and time-frequency parameters such as P2/PI already utilized for trialectic differentiation) via UMAP or t-SNE. Within the feature space, distances between the two groups are quantified: distance from consistent-judgment individuals to their respective cluster centers versus distance from inconsistent-judgment individuals to the nearest cluster center. Morphological adjudication employs clustering validity metrics: silhouette coefficient and Davies-Bouldin index, calculated under three configurations—”forced tri-cluster,” “allowed four-cluster,” and “no clustering.”
The decision logic is straightforward, yielding three outcomes corresponding to three conceptual worlds:
First outcome: inconsistent individuals uniformly populate the transition zones between clusters with no discernible aggregation structure; silhouette coefficients degrade significantly under the forced tri-cluster setting. This directly supports the continuous spectrum hypothesis. The three factors resemble primary colors on a palette; each individual represents a unique mixture with varying proportions. “Inconsistent judgment” reflects neither expert error nor measurement failure, but rather the inherent impossibility of cleanly segmenting a continuous spectrum.
Second outcome: inconsistent individuals form their own independent medium-strength clusters (bridging-type intermediate clusters); silhouette coefficients degrade under forced tri-cluster but recover under the four-cluster setting. This corresponds to a world of “discrete boxes beyond three”: dual-factor or triple-factor composites constitute genuine, measurable independent categories—not transitional zones.
Third outcome: the feature distribution of inconsistent individuals is statistically indistinguishable from that of consistent ones. This indicates that the pulse harmonic channel contains no morphological information regarding constitution; inconsistent judgment stems from the assessment endpoint (experts or questionnaire) rather than the measurement endpoint. In this case, morphological inquiry must shift to alternative channels.
Implementation requires three preliminary gates. First, dimensionality reduction serves visualization only, not inference: UMAP cluster distances lack probabilistic interpretation; all quantitative conclusions must be validated in the original feature space using silhouette coefficients and Davies-Bouldin index to avoid artifacts such as “continuous data artificially rendered as three clusters” by UMAP parameters (n_neighbors, min_dist); sensitivity analysis of these parameters is mandatory, and conclusions must remain stable across the parameter range. Second, pre-registration prevents “post-hoc framing”: the order of selecting among tri-cluster, four-cluster, and no-clustering settings must be fixed before viewing data, otherwise both continuous and discrete proponents could find their desired shapes in the same plot. Third, sample size: morphological determination relies on internal structure of the inconsistent subset; given current κ estimates, inconsistent individuals comprise approximately 30–40% of the total. If the total sample falls below 100, the inconsistent subset may be insufficient to support any clustering validity inference; therefore, total recruitment must be calculated a priori based on subset power.
The value of this design lies in transforming a “nominal dispute” into a falsifiable experiment: both continuous and discrete proponents have explicit failure conditions. This approach holds direct translational value for Chinese Five-Phase constitution and the Nine-Tia differentiation of Traditional Chinese Medicine—they face the identical “discrete versus continuous” dilemma, while pulse harmonics serve as a shared measurement channel between both traditions.
Evidence and Limitations
Verified: Extreme body types cluster into three natural groups within phenotypic feature space (PMID 28981546); low inter-rater reliability for the three-factor determination is a stable, reproducible phenomenon (PMID 24191170, 25336848).
Partially Verified: Pulse waveform parameters can distinguish the dominant type of three dosha (significant differences in time-domain P2/PI and other parameters across pairwise comparisons of the three pairs), but this channel has never been tested for its ability to differentiate mixed or transitional individuals.
Pending Verification: All core propositions of this paper—specifically the spatial morphology of inconsistent individuals in harmonic feature space (continuous transition bands / independent intermediate clusters / indistinguishable regions), distance metrics between consistent and inconsistent individuals under UMAP dimensionality reduction, and multi-setting comparisons of silhouette coefficients and DBI—have no supporting literature data and represent purely experimental design proposals. Critical limitations must be stated: objective measurement of three-factor body types remains in the exploratory stage, and no verified morphological conclusions exist.
Related Reading
- Label Learnability Pre-screening for Individuals with Inconsistent Three-Factor Constitution Judgments
- Objective Measurement of the Three-Factor Constitution: From Subjective Assessment to Instrumental Quantification
- IPG Quantification of Ayurvedic Pulse Diagnosis
- Research Pool: Cross-Cultural Frequency Cognition
L1 Network Status: This topic has not yet entered the L1 corresponding network. The EDGE_TABLE shows zero hits for “constitution morphology judgment / continuum / UMAP / silhouette coefficient” (only peripheral edges related to dosha sound therapy exist, such as ayurveda-dosha-chanting-pathway-selection, etc.); line 73 of CORRESPONDENCE_NETWORK.md contains only a general mention: “constitution typing | scale | moderating variable of individual differences.” The edge linking “constitution morphology (continuum vs. discrete clusters) × pulse harmonic feature space” is missing and has been registered in science-yixue-l1-gaps.md.
Evidence Level: C
References: PMID 28981546, PMID 40582042, PMID 24191170, PMID 25336848
Evidence level: A = direct PubMed support; B = scientific analogy / indirect evidence; C = traditional consensus / preliminary exploration.
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