Evidence level: B
外气疗效到底来自”这个人修为高不高”(特质),还是来自”这一场他状态好不好”(状态)?传统说法是”功夫在功外”——修为是底子;但现代心理治疗研究却证明,同一位治疗师在不同会谈中状态波动本身就携带预测力。用嵌套多层模型把”特质”和”状态”同时放进一个模型,就能拆开这个困扰外气研究二十年的黑箱。
正文
2008 年那篇 112 例膝骨关节炎外气随机对照试验留下了一个悬案:两位治疗师手法相同、方案相同,疗效却显著不同(PMID 18654733)。二十年来,”治疗师个体差异”一直是个黑箱标签——研究者知道有的治疗师”发功”更灵,却说不清这份灵到底是他这个人”本来就好”(特质),还是他”那一场状态好”(状态)。这两者的差别不是文字游戏:如果主要是特质,那选人、培训就够了;如果主要是状态,那外气研究必须学会测量单次会谈内的实时生理,否则永远测不准疗效。
为什么必须用嵌套多层模型:组间与组内效应是两回事
这个问题的统计本质,是把”治疗师之间的差异”(between-person)和”治疗师自身不同会谈之间的波动”(within-person)拆开。早在 2011 年,Curran 与 Bauer 的经典综述就系统论证了:纵向数据最被低估的价值,就是能把时变协变量对结局的效应分解成组间与组内两部分,而多层模型(MLM)是标准工具(PMID 19575624)。2015 年 Wang 与 Maxwell 进一步厘清了三种中心化方式的差别——不中心化、总均值中心化、个体均值中心化(person-mean centering)——明确指出:只有用”个体均值中心化”把会谈层变量减去治疗师自身均值,才能得到纯粹的”组内(状态)效应”,不会被”组间(特质)差异”污染(PMID 25822206)。2023 年 Neubauer 等的多层结构方程框架还提示,组内效应本身还可能反过来塑造组间差异,说明两层信息不能互相替代(PMID 35446047)。
这套方法学就是”状态 vs 特质”检验的解剖刀:把每个治疗师的会谈内状态指标做个体均值中心化,得到纯粹的状态效应;把每个治疗师的状态指标均值(连同修为、经验、FIS 类表现分)放到治疗师层,得到特质效应;两者同模型估计,再比较加入状态层变量后 Δ方差/ΔAIC 是否显著——这就是”控制特质后,状态是否仍有独立增量”的严格检验。
特质层的证据:治疗师差异真实存在,但”修为类”特质只解释一小部分
治疗师效应的大小本身已有多项大规模实证。2017 年一项分析 1.1 万余名患者、85 名治疗师的研究显示,治疗师效应解释了脱落率方差的 12.6% 和恶化率方差的 10.1%(PMID 27424844);同年另一项英国常规数据研究在控制病例构成后,治疗师效应约占总结局方差的 5.8%(PMID 27424106)。但低强度干预的 1376 名患者、38 名治疗师研究却只测到 0%-1.3% 的治疗师 ICC(PMID 25207881)——治疗师效应不是常数,这本身就提示”特质之外还有会变的东西”。
更关键的是,”特质类”指标到底能解释多少。2025 年 Schwartz 团队对 99 名受训治疗师、1031 名门诊患者的分析显示:人际技能(TRIB 量表)显著预测结局,但仅额外解释 1.3% 的方差,且初始损害越重、技能效应越强(PMID 39874731)。也就是说,即使是经过严格行为评估的”特质”,也只能解释治疗师效应的一小块——其余大部分变异仍悬而未决,留给”状态”等会谈层因素。
状态层的证据:会谈内状态确实与当次会谈质量挂钩
会谈内状态有没有信息?2010 年开发的”治疗性临在”量表(TPI)给出了第一个线索:患者报告的治疗师临在感(一种会谈内状态体验)能显著预测当次会谈后的联盟与改变(PMID 20845229)。2018 年涵盖 82 个独立样本、6138 名患者的共情元分析也显示,治疗师的共情——一种可在会谈间波动的人际状态——与结局呈中等强度相关(r = .28,PMID 30335453)。这些证据来自心理治疗,但指向同一个结论:会谈内可变的”状态”质量不是噪音,它有预测力。
外气场景呢?直接证据来自 2005 年两项随机对照/交叉研究:外气治疗能显著降低心率、提高 HRV、降低 LF/HF(PMID 16048810),非线性分析还发现 Poincaré 图指数与近似熵升高(PMID 16173532)。这些是”会谈内状态指标在动”的铁证——但注意,它们测的是受试者,不是治疗师;治疗师自己的”发功状态”至今没有被逐会谈地测过。
一个可检验的实验设计
据此提出嵌套多层模型检验:招募 ≥20 名治疗师(修为/经验分层),每名治疗师完成 ≥8 次外气会谈(对盲法受试者),全程记录治疗师 HRV 频域/RMSSD/呼吸相干/EEG 频段功率,会谈后受试者即时自评疗效与满意度。模型分三步:(1) 空模型估计治疗师层 ICC;(2) 加入特质层变量(修为年限、经验、FIS 类表现分);(3) 加入状态层变量(个体均值中心化的会谈内状态指标),用 Δ方差/ΔAIC 检验状态层的独立增量,并检验”状态×特质”交互项(修为高者的状态波动是否更影响疗效)。对照组加”模拟发功”条件,排除表演期望效应。
若状态层仍有显著增量,则”功夫在功外”获得现代操作化定义——修为是底子,但每一场”发功”的成败由当场的生理状态独立决定,治疗师可据此实时调节;若增量消失,则外气疗效基本由特质决定,研究重心应回到选人与培训。无论哪种结局,这个悬置二十年的黑箱都会被打开一条缝。
证据与局限
已验证: 组间/组内效应分解方法学成熟(PMID 19575624, PMID 25822206);治疗师效应真实存在且占 5%-13% 方差(PMID 27424106, PMID 27424844);特质类人际技能仅解释治疗师效应的一小部分(PMID 39874731);会谈内状态与联盟/结局相关(PMID 20845229, PMID 30335453);外气会谈内受试者状态指标确实在动(PMID 16048810, PMID 16173532)。
部分验证: “控制特质后状态仍有独立增量”在外气场景尚无直接数据,上述证据全部来自心理治疗与统计方法学,属跨领域类比;治疗师侧会谈内状态的逐会谈测量尚无先例。
待验证: 外气治疗师会谈内状态指标(个体均值中心化)在控制修为/经验/FIS 后是否仍显著预测单次结局;状态×特质交互的方向与强度;”模拟发功”对照下的期望/表演混杂是否改变增量结论;治疗师状态层增量的跨传统(气功/Reiki/治疗性触摸)一致性。
相关阅读
- 治疗师的”状态”决定疗效?会谈中 HRV 与呼吸相干预测外气单次结局
- 一次”发功”成败的仪表盘:会谈内状态轨迹为何比单点平均值更会预测外气疗效
- FIS 评分预测外气疗效的独立增量检验
- 研究池子:环境-生命频率接口
证据等级: B(组间/组内分解与治疗师效应有 PubMed 直接实证,但外气治疗师”状态 vs 特质”的头对头增量检验尚无直接数据,属跨领域类比 + 实验设计提案)
参考文献: PMID 19575624, PMID 25822206, PMID 27424106, PMID 27424844, PMID 39874731, PMID 20845229, PMID 16048810, PMID 16173532
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。
Is the power outside or inside the practice? Can the ‘entering a state’ during therapy sessions still independently predict single-session outcomes after controlling for therapi…
Evidence level: B
Does external qi efficacy stem from “how high the practitioner’s cultivation is” (trait) or “how good their state is in this session” (state)? Traditional wisdom posits “the work lies outside the practice”—cultivation is the foundation; however, modern psychotherapy research demonstrates that state fluctuations within a single practitioner across sessions carry predictive power. By employing nested multilevel models that simultaneously incorporate both “trait” and “state,” we can dissect the black box that has puzzled external qi research for two decades.
Main Text
The 2008 randomized controlled trial involving 1,120 patients with knee osteoarthritis and external qi left an unresolved case: two practitioners used identical techniques and protocols yet yielded significantly different outcomes (PMID 18654733). For two decades, “practitioner individual differences” has served as a black box label—researchers know some practitioners are “more effective at inducing qi,” yet they cannot clarify whether this efficacy stems from the person being inherently better (trait) or from having a better state in that specific session (state). The distinction between these two is not merely semantic: if the driver is primarily trait, then selection and training suffice; if it is primarily state, external qi research must learn to measure real-time physiology within single sessions, otherwise efficacy will remain unmeasurable.
Why Nested Multilevel Models Are Necessary: Between-Person and Within-Person Effects Are Distinct
The statistical essence of this issue involves separating “differences between practitioners” (between-person) from “fluctuations within a practitioner across different sessions” (within-person). As early as 2011, Curran and Bauer’s classic review systematically argued that the most undervalued asset of longitudinal data is its ability to decompose the effects of time-varying covariates on outcomes into between-person and within-person components, with multilevel modeling (MLM) serving as the standard tool (PMID 19575624). In 2015, Wang and Maxwell further clarified the differences among three centering methods—no centering, grand-mean centering, and person-mean centering—explicitly stating that only “person-mean centering” (subtracting the practitioner’s own mean from session-level variables) yields a pure “within-person (state) effect,” free from contamination by “between-person (trait) differences” (PMID 25822206). Neubauer et al.’s multilevel structural equation framework in 2023 also indicated that within-person effects may themselves shape between-person differences, suggesting the two layers of information cannot substitute for one another (PMID 35446047).
This methodology serves as the scalpel for testing “state versus trait”: by applying person-mean centering to each practitioner’s session-level state indicators, one obtains the pure state effect; by placing each practitioner’s mean state indicator (along with cultivation years, experience, and FIS-type performance scores) at the practitioner level, one obtains the trait effect; both are estimated simultaneously in the model. Comparing whether the addition of state-level variables yields a significant Δ variance/Δ AIC constitutes the rigorous test of “whether state retains independent incremental value after controlling for trait.”
Evidence at the Trait Level: Practitioner Differences Are Real, but “Cultivation-Type” Traits Explain Only a Small Portion
The magnitude of practitioner effects is already supported by multiple large-scale empirical studies. A 2017 analysis of over 11,000 patients and 85 practitioners showed that practitioner effects accounted for 12.6% of dropout variance and 10.1% of deterioration variance (PMID 27424844); a concurrent British routine data study controlling for case composition found practitioner effects accounted for approximately 5.8% of total outcome variance (PMID 27424106). However, a study involving 1,376 patients and 38 practitioners in low-intensity interventions detected only 0%-1.3% practitioner ICC (PMID 25207881)—since practitioner effects are not constant, this itself suggests “there is also something variable beyond trait.”
More critically, what proportion of variance can “trait-like” indicators actually explain? A 2025 analysis by Schwartz et al. of 99 trained practitioners and 1,031 outpatient patients showed that interpersonal skills (measured by the TRIB scale) significantly predicted outcomes but explained only an additional 1.3% of variance, with stronger effects observed among patients with heavier initial impairment (PMID 39874731). In other words, even rigorously assessed behavioral “traits” explain only a small fraction of practitioner effects; the majority of variance remains unresolved, leaving room for session-level factors such as “state.”
Evidence at the State Level: Intra-Session States Are Indeed Linked to Session Quality
Does intra-session state carry information? The “Therapeutic Presence Inventory” (TPI), developed in 2010, provided the first clue: patients’ reported sense of practitioner presence (an intra-session state experience) significantly predicted alliance and change following the session (PMID 20845229). A 2018 meta-analysis of empathy encompassing 82 independent samples and 6,138 patients also showed that practitioner empathy—a type of interpersonal state fluctuating between sessions—correlated moderately with outcomes (r = .28, PMID 30335453). These evidences originate from psychotherapy but point to the same conclusion: variable “state” quality within sessions is not noise; it possesses predictive power.
What about the external qi context? Direct evidence comes from two randomized controlled/crossover studies in 2005: external qi treatment significantly reduced heart rate, increased HRV, and decreased LF/HF (PMID 16048810); nonlinear analysis further revealed elevated Poincaré plot indices and approximate entropy (PMID 16173532). These constitute ironclad evidence that “intra-session state indicators are moving”—however, note that they measured the subjects, not the practitioners; the practitioner’s own “induction state” has never been measured session-by-session to date.
A Testable Experimental Design
Based on this, we propose a nested multilevel model test: recruit ≥20 practitioners (stratified by cultivation/experience), each completing ≥8 external qi sessions with blind subjects, recording practitioner HRV frequency domains/RMSSD/respiratory coherence/EEG band power throughout, and having subjects rate efficacy and satisfaction immediately post-session. The model proceeds in three steps: (1) estimate practitioner-level ICC in an empty model; (2) add trait-level variables (cultivation years, experience, FIS-type performance scores); (3) add state-level variables (person-mean centered intra-session state indicators), testing the independent incremental value of the state layer via Δ variance/Δ AIC, and examining the “state × trait” interaction term (whether state fluctuations have a greater impact on efficacy among high-cultivation practitioners). A control group includes a “simulated induction” condition to exclude performance expectancy effects.
If the state layer retains significant incremental value, then “the work lies outside the practice” obtains a modern operational definition—cultivation is the foundation, but the success or failure of each “induction” is independently determined by physiological state in that moment, allowing practitioners to adjust in real-time; if the increment vanishes, external qi efficacy is largely determined by trait, and research focus should return to selection and training. Regardless of the outcome, this black box suspended for two decades will be cracked open.
Evidence and Limitations
Verified: Between-person/within-person effect decomposition methodology is mature (PMID 19575624, PMID 25822206); practitioner effects are real and account for 5%-13% of variance (PMID 27424106, PMID 27424844); trait-like interpersonal skills explain only a small fraction of practitioner effects (PMID 39874731); intra-session states correlate with alliance/outcomes (PMID 20845229, PMID 30335453); subject-level state indicators in external qi sessions are indeed dynamic (PMID 16048810, PMID 16173532).
Partially Verified: “Independent incremental value of state after controlling for trait” lacks direct data in the external qi context; all above evidence derives from psychotherapy and statistical methodology, constituting cross-domain analogy; session-by-session measurement of practitioner-side intra-session states has no precedent.
To Be Verified: Whether practitioner-level intra-session state indicators (person-mean centered) significantly predict single-session outcomes after controlling for cultivation/experience/FIS; the direction and strength of the state × trait interaction; whether expectancy/performance confounds alter incremental conclusions under “simulated induction” control; cross-traditional consistency of practitioner state-level increments (Qigong/Reiki/Therapeutic Touch).
Related Reading
- Does Practitioner “State” Determine Efficacy? Predicting External Qi Single-Session Outcomes via HRV and Respiratory Coherence in Sessions
- Dashboard for a Single “Induction” Success/Failure: Why Intra-Session State Trajectories Predict External Qi Efficacy Better Than Single-Point Means
- Testing Independent Incremental Prediction of External Qi Efficacy via FIS Scores
- Research Pool: Environment-Life Frequency Interface
Evidence Level: B (Between-person/within-person decomposition and practitioner effects have direct PubMed empirical support, but head-to-head incremental testing of “state versus trait” in external qi practitioners lacks direct data, constituting cross-domain analogy + experimental design proposal)
References: PMID 19575624, PMID 25822206, PMID 27424106, PMID 27424844, PMID 39874731, PMID 20845229, PMID 16048810, PMID 16173532
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