跳至正文

IQ 由基因决定吗

目录

本篇问的不是「智力是什么」，而是「两件统计器械被读成了什么」。 g（心理测量的一阶公因子）与 h²（遗传率）都有过硬的技术定义与几十年的数据地基；但公共叙事里，它们一个被读成「脑子里可定位的聪明实体」，一个被读成「你命里注定的百分比」——再被接力读成「干预无用」「组间差异天定」。对称地，另一极把它们读成「IQ 只测应试能力」「g 是因子分析的幻觉」「双生子研究全部破产」「整个领域是优生学伪装」。本篇逐层称量：哪些是定理级/数据级事实，哪些是合法推论，哪一跳开始是升格，哪一跳是虚无化。去重分界：基因决定论的哲学面（记账视角 vs 本体实在）让位自私基因篇，本篇只借结论不重铸；测验偏见的测量学技术细节与复制危机方法论只点到不展开（让位复制危机篇）；「智能」的机器侧（AI eval、LLM 能力）不展开。组间差异的最终因果构成，本篇不裁——它是科学上至今未填的空位，不是任何一篇综述能下的判决；本篇只裁决「什么证据存在、什么推论合法、什么推论非法」。高风险声明：本篇涉及智力、遗传与群体差异等敏感主题，不评估任何个人或群体的价值，不作教育、临床、移民或社会政策建议；全部工作只是对科学主张与其推论结构的压力测试。

〇 母裁决：七层硬度光谱

层 主张 硬度 处置
① 不同认知测验成绩两两正相关（全正相关矩阵），一阶公因子 g 是统计事实；主流测验信度 FSIQ .86–.95+；11 岁与 80 岁同测验相关 r=.66 定理级/数据级·最硬地板 收下
② IQ 有真实跨域预测效度：学业 r≈.50、工作绩效校正后 .54、儿童期 IQ 高 1SD 对应死亡风险低 24%；双生子/分开抚养信号真：MZ/DZ 平均相关 .86/.60，分开养 MZ .68–.78，成年 h² 渐近约 .80 数据级·条件硬 收下，带边界
③ 弗林效应真：20 世纪测验分数每十年约涨 3 分，上升/拐点/反转全部可由家庭内变异复现；教育每多一年 IQ +1~5 分且终身持久；收养追平环境同伴；低剂量铅低段更陡 数据级·硬 收下——高 h² 与大环境效应并存（Dickens-Flynn 乘子）
④ 分子层真而未合拢：GWAS 数千位点、PGS 解释 4–16%；SNP h² 约 20–30% 对双生子 50–80%（missing heritability）；within-family 估计缩水约一半 数据级·条件硬·未结案 挂起，列边界
⑤ h² 被读成「个体能力的基因命定度」「高 h²=干预无用」「组内 h² 可外推组间」 概念误读·软·主菜 逐条打回（Visscher 2008 五条误读逐一在案）
⑥ 命定链「h²=.8→智力八成天定→干预无用→组间天定→价值排序」 vs 虚无链「IQ 只测应试→g 是幻觉→双生子全破产→领域即优生学」 升格/虚无化·上红线 双向逐跳打
⑦ 优生挪用史与 Pioneer Fund 资助谱系 史实真·与数据命题分开审计 披露，不作遗传谬误的弹药

母裁决一句话：g 与 h² 都是真的统计器械，也都被两名叫「命定」与「虚无」的司机开过了护栏——遗传率回答的只是「在这群人、这个年代、这个环境分布里，个体差异与基因型差异相关几成」；它对你一个人是什么、对另一群人差多少、对未来能改多少，三个问题都保持沉默。命定派替它抢答了全部三个，虚无派则没听见它确实回答了第一个。

一 唯象地基：g 是统计事实，测验是硬器械

先立守真锚。命定派与虚无派都从这一层偷东西：命定派偷走「g 真」冒充「g 的本体已被找到」，虚无派偷走「g 有批评者」冒充「数据本身不存在」。

g 的统计起源一手。Spearman 1904 年《American Journal of Psychology》长文提出双因素论：一切智力活动共享一个根本功能（g），各测验另有特异成分（s）——「all branches of intellectual activity have in common one fundamental function (or group of functions), whereas the remaining or specific elements of the activity seem in every case to be wholly different from that in all the others」（Classics in the History of Psychology 节选版，主笔亲核）；他称之为 「Universal Unity of the Intellective Function」，并断言各分科成绩「may be regarded as so many independently obtained estimates of the one great common Intellective Function」。术语勘误：1904 年原文没有「positive manifold」一词——该通行标签是后起的，Neisser et al. 1996 把它归于 Spearman 1927（APA 报告 §II：「all such correlations are typically positive and form what is called a ‘positive manifold.’ Spearman (1927) showed that…」）；撰稿时应表述为「1904 年观察到全正相关现象，后世称 positive manifold」。

跨世纪复现。Carroll 1993 年重分析 「more than 460 different datasets that included nearly all the more important and classic factor-analytic studies of human cognitive abilities since the time of Spearman」（Schneider & McGrew 2012 权威概述章），得三阶层结构（III 层=g；II 层八个宽能力 Gf/Gc/Gy/Gv/Ga/Gr/Gs/Gt；I 层约 70 个窄能力），后整合为 CHC 模型；学界评价为 「definitive foundation for current theory」。但同一份概述章自带本体地位警示（Box 4.1）：「Spearman did not consider g to be an ability, or even a thing… factors derived from factor analysis should not be reified prematurely」——连提出者本人都不把 g 当「物」。APA 报告同样克制：「Such a summary merely acknowledges that performance levels on different tests are correlated; it is consistent with, but does not prove, the hypothesis that a common factor such as g underlies those correlations」。

信度。Wechsler 总量表（FSIQ）内部一致性与 test-retest 信度 「average about .95 or higher」.pdf)（Hogan 2018 教科书转述 WAIS-IV 手册数据）；长期稳定性一手：Watkins 2021（WISC-V，225 名临床儿童、平均 2.6 年间隔）FSIQ 稳定系数 r=0.86（95%CI 0.82–0.89），分测验可低至 .50；跨测验元分析（Schuerger & Witt 1989，经 Watkins 2021 转述）间隔 0–10 月平均 .89、更长间隔 .80。数字边界：总量表极稳、分测验平庸——虚无派引用低数字时拿的是分测验，命定派引用高数字时拿的是总量表。

终身稳定性（同一测验跨 79 年）。Lothian/Aberdeen 出生队列用同一份 Moray House 测验：11→70 岁 r=.67（LBC1936）、11→77 岁 r=.63（ABC1921）、11→80 岁 r=.66（LBC1921，衰减校正后 .73，N=550）——Deary et al. 2004, JPSP（主笔亲核正文）；11→90 岁 r=.54（N=106，95%CI .37–.67）——Deary, Pattie & Starr 2013, Psychological Science。儿童期分数与近 80 年后的自己相关 .5–.7，这不是「应试技巧」能跨 80 年搬运的。

预测效度（跨域、硬终点）。APA 报告汇总：学业成绩 r≈.50（「account for only about 25% of the overall variance」）、教育年限 ≈.55、工作绩效 r 在 .30–.50 之间、Hunter 1983 校正后均值 .54（「accounts for some 29% of the variance in job performance」）（主笔亲核镜像全文）。死亡率硬终点：Whalley & Deary 2001（BMJ，苏格兰 1932 年全员调查）：11 岁低 15 分对应 65 年后存活相对概率 0.79（95%CI 0.75–0.84）；Calvin et al. 2011 元分析（16 研究、1,107,022 人、22,453 例死亡）：认知分数高 1SD 对应死亡风险低 24%（95%CI 23–25），调整儿童期 SES 几乎不改变关联，调整成人 SES/教育分别衰减 34%/54%（任务书曾误记为「Calvin et al. 2011 BMJ」，实刊于 International Journal of Epidemiology 40(3):626–44，随文纠正）；Gottfredson & Deary 2004：苏格兰近全人口研究，儿童期 IQ 预测成人发病率与死亡率，控制社会经济变量后仍显著（APS 摘要页主笔亲核）。

反方锚一并收。Boring 名句出处考据：「Intelligence as a measurable capacity must at the start be defined as the capacity to do well in an intelligence test」——Boring 1923 年 6 月 6 日《New Republic》pp.35–37（Cambridge Handbook of Intelligence 书目核实；卷号记法有 35/36 两说）；注意语境：这是 Lippmann–Terman 论战中对操作定义的概括，常被脱离语境引用（俄勒冈大学学位论文专门指出「Boring’s comment has been the subject of controversy」）。现代批评一手：Richardson & Norgate 2015：「construct validity for IQ tests has always been difficult to establish… considerable caution needs to be exercised in citing such correlations for test validation purposes」（质疑 .5 级相关的校正程序与原始数据质量）。这一批评真实存在且有理有据，但它的锋芒对准的是「校正后 .54」这类数字的乐观上限，不是「预测效度为零」——未经校正的 .30–.50 原始相关不是校正程序造出来的。

第一层裁决：全正相关矩阵与 g 是定理级统计事实（1904 年观察到、460+ 数据集复现）；主流测验 FSIQ 信度 .86–.95+；11 岁与 80 岁相关 .66；学业/工作/死亡率的跨域预测真实存在。守真锚立住了：这不是一个「什么都测不到」的器械。 但同层三处边界也立住了：g 的提出者本人拒绝把它实体化、APA 明说相关结构「consistent with, but does not prove」共同因子假说、分测验长期稳定可低至 .5。第一层给后文的可用量是：器械真、读数稳、效度有上限——仅此而已。

二 叙事考古：从巴黎甄别表到优生挪用链

名号学是本篇必须做的功课：「IQ」这件器械的发明语境、挪用史与撤回史，每一拍都有原文。这一层同样是双向的——挪用史是真的，但挪用史的流行版本本身也有升格。

Binet 原意（1905 一手，四句全经主笔亲核）。量表受巴黎公共教育部委托，用途是甄别需要特殊教育安置的儿童：「Our purpose is to be able to measure the intellectual capacity of a child who is brought to us in order to know whether he is normal or retarded. We should therefore, study his condition at the time and that only. We have nothing to do either with his past history or with his future…」（Classics 全文，Kite 1916 英译）。量表终点即平均水平，结构上不为正常儿童排名：「a series of tests of increasing difficulty, starting from the lowest intellectual level that can be observed, and ending with that of average normal intelligence」。对量化分数的谨慎（承重句）：「This scale properly speaking does not permit the measure of the intelligence, because intellectual qualities are not superposable, and therefore cannot be measured as linear surfaces are measured, but are on the contrary, a classification, a hierarchy among diverse intelligences; and for the necessities of practice this classification is equivalent to a measure.」（自带脚注 1）。智力的核心是判断：「This faculty is judgment, otherwise called good sense, practical sense, initiative, the faculty of adapting one’s self to circumstances.」。1909 年《Les idées modernes sur les enfants》更直接反对固定量观：「A few modern philosophers seem to lend their moral approval to these deplorable verdicts when they assert that an individual’s intelligence is a fixed quantity, a quantity that cannot be increased. We must protest and react against this brutal pessimism.」（英文转录措辞各源略异，以法文原版为准；HAL 学位论文核证）。Binet 的量表是安置工具不是命定仪器——这一条双方都要吞：命定派失去了「发明人背书」，虚无派失去了「发明目的就是排名」的指控。

美国优生挪用链（带双向审计）。1908 年 Goddard 把 Binet-Simon 引入美国（Embryo Project Encyclopedia, ASU）；1912 年《The Kallikak Family》宣称同一祖先两条世系构成 「a natural experiment of remarkable value… That we are dealing with a problem of true heredity, no one can doubt」（第 2 章末；该书 12 个版本、译成德文，后有照片篡改指控 Fancher 1987 / Elks & O’Brien 2005）。1917 年 Ellis Island 研究：178 名三等舱乘客预筛样本，Kamin 转述的「83% of the Jews, 80% of the Hungarians, 79% of the Italians, and 87% of the Russians were ‘feeble-minded’」数字属实（Snyderman & Herrnstein 1983 核证出自 Goddard 1917 Table II, p.252，且指出 Kamin 引注互相颠倒）——但双向审计的另一面：Goddard 自己写明研究 「makes no attempt to determine the percentage of feeble-minded among immigrants in general or even of the special groups named」（p.244，S&H 带页码转引），并倾向环境解释：「We may argue that it is far more probable that their condition is due to environment than it is due to heredity…」（p.270，主笔在 S&H 1983 PDF 中亲手核到此句）。Terman 1916 年修订 Stanford-Binet 并引入 IQ 概念，第 1 章公开预言测验将 「bring tens of thousands of these high-grade defectives under the surveillance and protection of society… curtailing the reproduction of feeble-mindedness」（Classics 全文）；1922 年他有族群等级言论（「The immigrants who have recently come to us in such large numbers from Southern and Southeastern Europe are distinctly inferior mentally to the Nordic and Alpine strains…」，Terman 1922, p.660，S&H 带页码转引），但 1924 年公开退却：「Whether these ‘chronic’ traits reflect primarily the influence of endowment or of environment is a question to which no certain answer can at present be given」（Terman 1924, p.102，主笔在 S&H PDF 亲核）。Yerkes 一战陆军项目 1917–1919 共测验 1,726,966 名新兵（亚利桑那大学法学刊），Yerkes 本人 1921 年官方报告同样保留：「At best we can but leave for future decision the question as to whether the differences represent a real difference of intelligence or an artifact of the method of examination.」（p.704，S&H 转引）。

Brigham 的全力升格与公开撤回（撤回句主笔亲核）。Brigham 1923 年《A Study of American Intelligence》以陆军数据主张 Nordic/Alpine/Mediterranean 递减等级并主张限制移民（七点概括见 Snyderman & Herrnstein 1983, pp.987–988），Yerkes 作序背书（「presents not theories or opinions but facts… the menace of race deterioration」，Alicia Patterson 基金会《Inventing the SAT》转引）；1924 年 1 月对 National Republican Club 演讲称移民是 「carriers of feeble-mindedness」（Saretzky, ETS RM-82-4 转《纽约时报》1924-01-27 报道，PDF 中「importation of a low grade of people」主笔亲核存在）。1930 年，Brigham 在《Psychological Review》37(2):158–165 公开撤回：「…comparative studies of various national and racial groups may not be made with existing tests… one of the most pretentious of these comparative racial studies—the writer’s own—was without foundation.」（Saretzky 转引，「the writer’s own—was without foundation」主笔在 ETS 备忘录 PDF 亲核）；1938 年《纽约时报》更激进：「it is ridiculous to claim that any test score is related to the germ plasm, and that alone」（Saretzky 转引）。撤回限度（双向审计必载）：Saretzky 指出 Brigham 只说「existing tests」不能回答此问题，并未否认群体差异存在的可能性。

1924 移民法因果链：流行叙事 vs 修正派史学（双向审计核心案）。史实层：1924 年移民法确立以 1890 年人口普查为基准的 2% 国别配额、完全排除亚洲移民（美国国务院史官办公室）；Harry Laughlin 以「Expert Eugenics Agent」身份向众议院移民委员会作证。流行叙事（Gould 1981, p.232）称「国会辩论不断援引陆军测验数据」——修正派核查：Snyderman & Herrnstein 1983（《American Psychologist》38(9):986–995）系统核查国会记录后结论：「An examination of the historical record failed to uncover any support for either component of the claim. The testing community did not generally view its findings as favoring restrictive immigration policies like those in the 1924 Act, and Congress took virtually no notice of intelligence testing, as far as one can ascertain from the records and publications of the time.」（摘要，主笔亲核）；正文 p.993：法案 「in 32 sections makes no reference to intelligence tests, intelligence, feeblemindedness, or any other related term」（主笔亲核「makes no reference to intelligence tests」句）。优生运动真、Laughlin 证词真，但「测验数据直接驱动 1924 法案」的强版本未通过史料核查——且注意 S&H 自身的立场提示：Herrnstein 后来是《The Bell Curve》作者之一，该文属修正派史学，与 Kamin/Gould 叙事对立，引用时双方并列。

Jensen 1969：补偿教育失败论与组间假说（p.82 全段主笔亲核）。Jensen 1969《How Much Can We Boost IQ and Scholastic Achievement?》（Harvard Educational Review 39(1):1–123）开篇即 「Compensatory education has been tried and it apparently has failed.」（全文 PDF 第 56 行亲核）；HER 编者按摘要其论点为阶级与种族差异 [「cannot be accounted for by differences in environment but must be attributed partially to genetic [factors]」](https://gwern.net/doc/iq/1969-jensen.pdf)。p.82 关键段（主笔在全文 PDF 亲核整段）：「So all we are left with are various lines of evidence, no one of which is definitive alone, but which, viewed all together, make it a not unreasonable hypothesis that genetic factors are strongly implicated in the average Negro-white intelligence difference. The preponderance of the evidence is, in my opinion, less consistent with a strictly environmental hypothesis than with a genetic hypothesis, which, of course, does not exclude the influence of environment or its interaction with genetic factors.」——注意措辞形态：「not unreasonable hypothesis」「in my opinion」「does not exclude the influence of environment」。这是升格链的起点之一，但起点本身写的是「假说」而非「判决」。

The Bell Curve 与主流回应。Herrnstein & Murray 1994 对组间差异的正式措辞其实是不可知论：「If the reader is now convinced that either the genetic or environmental explanation has won out to the exclusion of the other, we have not done a sufficiently good job of presenting one side or the other. It seems highly likely to us that both genes and the environment have something to do with racial differences. What might the mix be? We are resolutely agnostic on that issue; as far as we can determine, the evidence does not yet justify an estimate. (p. 311)」（经 Springer 2024 章节、Daedalus 2004、Murray 2014 Q&A 等四源转录一致；未直核原书，按多源转录入账）。出版后，APA 科学事务委员会 1994 年 11 月一致表决设立 11 人 task force（Neisser 主席），任务是「a dispassionate survey of the state of the art: to make clear what has been scientifically established, what is presently in dispute, and what is still unknown」（报告前言）；1995-08-07 发布、1996 年 2 月刊于《American Psychologist》51(2):77–101——即本篇反复定锚的 Neisser 1996。同月另有 Gottfredson 起草、52 人联署的《Mainstream Science on Intelligence》（WSJ 1994-12-13 p.A18；官方重印 Intelligence 1997;24(1):13–23）（重印版主笔亲核；签名史：131 邀、100 复、48 拒签、52 联署——「48 declined」亲核）。

Gould vs Morton 三回合（虚无派自己的误计案）。Gould《The Mismeasure of Man》（1981/1996）把 Morton 头骨测量立为「科学被偏见操纵」的正典案例。Lewis et al. 2011（PLoS Biology）复测 Morton 已发表 670 颗头骨中的 308 颗（46%），结论：「Morton did not manipulate data to support his preconceptions, contra Gould… Ironically, Gould’s own analysis of Morton is likely the stronger example of a bias influencing results.」（摘要与「Our Verdict」节）；作者同时声明 Morton 的种族主义观点与 1849 年汇总表错误成立，且该文有正式勘误（PLoS Biol 2011;9(7)）。第三回合：Kaplan, Pigliucci & Banta 2015 反驳 Lewis et al.，「We take strong exception to Lewis et al.’s thesis that Morton was ‘right.’ We maintain that Gould was right to reject Morton’s analysis as inappropriate…」；另 Weisberg 2014《Remeasuring Man》（Evolution & Development 16:166–178）。三方未决，Lewis 结论并非终局——但它已足够证明：连「优生时代科学全是操纵」这个虚无派正典案例本身，都要打上未决标记。

资助谱系（披露而非弹药）。Pioneer Fund 1937 年由 Wickliffe Draper 创立（Lombardo 2002, Albany Law Review 65(3):743–830）；Jensen 三十年间获逾百万美元资助、Rushton 2002 年至 2012 年去世任主席（SPLC 档案页 Wayback 镜像；Rushton 本人 2002 年 Albany Law Review 文自述接任；AP 2021 讣闻三源互证）。资助关系必须披露、必须进入证据权重计算；但它不构成对数据命题的反证——那是遗传谬误。数据命题只能用数据与推论结构打。

场内红线文件（定锚）。ASHG 2018-10-19《ASHG Denounces Attempts to Link Genetics and Racial Supremacy》（AJHG 103(5):636）（Wayback PMC 镜像主笔六句亲核）：「Genetics demonstrates that humans cannot be divided into biologically distinct subcategories.」；「In this way, genetics exposes the concept of ‘racial purity’ as scientifically meaningless.」；「It follows that there can be no genetics-based support for claiming one group as superior to another… Any attempt to use genetics to rank populations demonstrates a fundamental misunderstanding of genetics.」。

第二层裁决：挪用史每一拍都真——Binet 的安置工具被 Goddard/Terman/Brigham 接力升格成族群等级器械，与优生运动、Pioneer Fund 谱系纠缠百年；但流行叙事自己的升格也同样在案：Goddard 有环境限定被 Kamin 略去、Terman/Brigham/Goddard 都有公开退却、「测验驱动 1924 法案」未过 Snyderman-Herrnstein 的国会记录核查、连 Gould 的 Morton 正典案例都被反指控且三方未决。历史层给后文的可用量是：这个领域的名号与机构史必须披露——而披露的对象包括两边。挪用史是真的，但挪用史不能替唯象数据投票（第一节的 r=.66、信度 .9x、死亡率关联不因这段历史而消失）；唯象数据真，也不能洗白挪用史。

三 概念命门：遗传率是方差比，不是命定度

本篇的独占核心。命定链的每一跳都从「h² 是什么」的误读开始；虚无链则从「h² 被误读」滑向「h² 不存在」。先把定义钉死，再逐条审误读。

定义一手（主笔亲核）。Visscher, Hill & Wray 2008（Nature Reviews Genetics 9:255–266）：遗传率是「a ratio of variances, specifically as the proportion of total variance in a population for a particular measurement, taken at a particular time or age, that is attributable to variation in additive genetic or total genetic values」（正文 Definitions 节，课程镜像 PDF 与 Nature 版同版式，主笔 pdftotext 亲核）——针对特定群体、特定时间/年龄的一次测量的方差比。群体依赖性随之而来：「the heritability in one population does not, in theory, predict the heritability of the same trait in another population」（主笔亲核「particular population and says nothing about environments in other populations」句）。摘要承认这个概念自 Wright 与 Fisher 引入以来 「Despite continuous misunderstandings and controversies over its use and application」 仍是进化生物学与医学的关键参数（主笔亲核）。

五条误读逐一在案（Visscher 2008 Box 2，前两条主笔亲核）。误读一：「Heritability is the proportion of a phenotype that is passed on to the next generation」——「This is a commonly held misconception; phenotypes are not passed on but genes are.」（亲核）。误读二：「High heritability implies genetic determination」——高 h² 只说当前群体里观察到的变异主要由基因型变异解释，「However, it does not mean that the phenotype is determined once we know the genotype, because the environment can change or can be manipulated to alter the phenotype. For example, the well-documented secular rise in height in many human populations is not at odds with reported heritabilities of about 0.8, but is likely to reflect changes in the environment resulting from improved nutrition and medical care.」（亲核）。误读五（组间）：「heritability should not be used to make predictions about mean changes in the population over time or about differences between groups, because in each individual calculation the heritability is defined for a particular population and says nothing about environments in other populations」（亲核）——随文即荷兰身高判例：19 世纪中美国白男世界最高、比荷兰男高约 9 cm；20 世纪末荷兰男反超约 5 cm，「a trend that is likely to be environmental rather than genetic」（「about 9 cm」「5 cm taller」两处亲核）。APA 报告同判：「Heritability does not imply immutability.」（镜像全文亲核；Neisser 1997 答复文确认该句在期刊版 p.86——「’heritability does not imply immutability’ (p. 86)」，主笔亲核）。

「高 h²≠不可干预」判例三件套。近视：Dirani et al. 2006（GEM 双生子研究，345 MZ+267 DZ 对）：球镜等效遗传率男 88%/女 75%，眼轴长度 94%/92%（PubMed 摘要亲核）——近视是人群里遗传率最高的常见性状之一，而干预是配镜；Bishop 2015：「Yet, high heritability does not imply immutability… When, for instance, a child has the heritable myopia, we do not treat them as passive victims of their genetic destiny. Instead, they are given spectacles: an intervention that is outside the range of normal environmental experiences, but which is tailored to counteract the genetic effect.」。PKU：「Screening newborn infants for PKU, now required in all states, in combination with immediate initiation of the special low-protein diet, has virtually eliminated PKU as a cause of intellectual disability in the United States.」（NICHD 官方页）——苯丙酮尿症是单基因智力障碍（遗传率逼近 1.0），被饮食干预「几乎消灭」：遗传率与可干预性在这里以最纯粹的形式脱钩。身高：NCD-RisC 2016（eLife，1472 项人群研究、1860 万受测者）：「The tallest people over these 100 years are men born in the Netherlands in the last quarter of 20th century, whose average heights surpassed 182.5 cm」，引言自陈 「Although height is one of the most heritable human traits… cross-population differences are believed to be related to non-genetic, environmental factors.」。

组内≠组间：从 Lewontin 种子类比到 2024 年数学封顶。Lewontin 1970《Race and intelligence》（Bulletin of the Atomic Scientists 26(3):2–8）设计了两个互补思想实验（原刊扫描无文本层，引文经 OCR 转录＋Sesardic 2000 转引 1976 重印本 p.89＋Block 1995＋Schraiber 2024 四方互证）：其一，两个完全纯系玉米——组内无遗传变异、组内 h²=0，但组间平均差异全部遗传；其二，开放授粉品种分两把种子，一把全营养液、一把硝酸盐减半并缺锌——组内 h²=1.0，但组间差异全部环境。结论句：「the genetic basis of the difference between two populations bears no logical or empirical relation to the heritability within populations and cannot be inferred from it」（Lewontin 1970, p.7；此句经 Schraiber & Edge 2024 正文逐字转引，主笔在 EuropePMC 全文 XML 亲核）。Lewontin 1974 年在《American Journal of Human Genetics》26(3):400–411 的 annotation（PubMed 4827368；PMC1762622 有扫描全文、无机读文本层，正文逐字未能取得，入账标注）把它推广为方法论总批评；同期转述（Goldberger 1975）：「within-group heritability carries no implications for between-group heritability and, furthermore, that high heritability carries no implications for the effectiveness of environmental policies」。Block 1995（Cognition 56:99–128）教科书式复述（种子-田地版，全文 PDF）：「The application to race is obvious: heritability is high within whites. But high heritability within groups licenses no conclusion about differences between groups.」；其摘要判词：「Heritability is an uninteresting measure that only misleads us about race.」（主笔亲核）。2024 年数学封顶：Schraiber & Edge et al.（PNAS）对行为遗传学 h_B²、进化 P_ST/Q_ST、混合人群祖源方法三族逐一证明：「Perfect knowledge of within-group heritability provides no information about between-group heritability.」（导言，主笔亲核）；「Our results establish Lewontin’s interpretation generally.」（Discussion，亲核）；且文中记录 DeFries 曾称 Lewontin 此句「incorrect」，而作者证明 「Lewontin’s claim is correct in general and not only in the special cases he used for illustration.」（亲核）。

哲学层的对称警示（双方都有稻草人）。Sesardic 2000（Philosophy of Science 67(4):580–602）指控：Lewontin 及其追随者驳倒的只是 H1（「高组内 h² 蕴含非零组间 h²」）或 H2（「组内 h² 单独归纳确立组间 h²」），而严肃的遗传决定论实际持 H3（「组内 h² 加上其他经验信息，归纳支持非零组间成分」）——「While serious hereditarianism actually involves commitment to H3, Lewontin and the philosophers of science following in his footsteps have persistently criticized H1 (or occasionally H2), with the unfortunate result that they simply never managed to get in contact with the real hereditarian argument (which aims to support H3).」（争议站点镜像，书本身为正规期刊出版物）。文本证据：Jensen 本人 1973 年书 p.1 即写 「Although one cannot formally generalize from within-group heritability to between-groups heritability, the evidence from studies of within-group heritability does, in fact, impose severe constraints on some of the most popular environmental theories…」（扫描镜像）；Flynn 1980 独立支持此解读（经 Sesardic 转引）：高 h² 的相关性在于 「they force us to look for a plausible candidate for the role of X-factor」。本层裁决的双向落点：Lewontin-Block-Schraiber 判例确立的是「组内 h² 单独不能判决组间」——2024 年后这是数学定理级；Sesardic-Flynn 点出的也是真的——类比本身同样不能反向判决「组间必环境」，它只是拆掉了非法推论，没有代替任何人填写组间答案。空位留给第六节。

第三层裁决：h² 是针对特定群体、特定时间、特定环境分布的方差比——它从不度量「一个性状有多遗传」，更不度量「你这个人有几分天定」。五条误读（传给下一代／高 h²=基因决定／低 h²=无加性方差／h² 说明组间差异／高 h²=大效应基因）在 Visscher 2008 Box 2 逐一在案；近视配镜、PKU 饮食、荷兰身高三件判例把「高 h²=不可干预」物理否证；Lewontin→Block→Schraiber & Edge 2024 把「组内→组间」推论在数学上一般性封死。命定链的第一跳（h²=命定度）、第二跳（=不可干预）、第三跳（=组间判决书）在这一层全部打断。 同时本层也封了虚无派的逃路：误读被清算不等于参数不存在——方差比在其定义域内是合法、可估、被 Wright-Fisher 传统使用了近百年的量。

四 双生子与分开抚养：信号真，边界也真

估计量与地基数值（全部主笔亲核）。Falconer 公式：ĥ²=2(r_MZ−r_DZ)，ĉ²=2r_DZ−r_MZ（Verhulst et al. 式 4；通俗表述见 Bishop 2015）。汇总地基：「across many studies, the average MZ and DZ correlation was 0.86 and 0.60, respectively, based on 4,672 MZ and 5,546 DZ twin pairs」（Visscher 2008 Box 4 亲核）；IQ 双生子 h² 报告值 「remarkably consistent in the range of 0.5–0.8, across many age groups」（亲核）；同一 Box 4 给出反例带：Devlin, Daniels & Roeder 1997 母体效应模型下 narrow h²≈0.3、broad H²≈0.5（「modelling of maternal effects implied a narrow-sense heritability of only 0.3 and an estimate of broad-sense heritability of 0.5」亲核）——同一批数据，模型假设换一个旋钮，点估计掉一半，这本身就是对「h²=.8 是测量常数」读数的场内警告。APA 1996 的参数汇总（镜像全文亲核）：合并全部相关 「the heritability (h²) works out to about .50 and the between-family variance (c²) to about .25」；「In childhood h² and c² for IQ are of the order of .45 and .35; by late adolescence h² is around .75 and c² is quite low (zero in some studies).」；分开抚养 MZ 相关 「ranged from .68 to .78 in five studies involving adult samples from Europe and the U.S. (McGue et al., 1993)」。遗传率随年龄上升（Wilson Effect）：Bouchard 2013：「the heritability of IQ reaches an asymptote at about 0.80 at 18-20 years of age and continuing at that level well into adulthood… shared environmental influence decreases across age, approximating about 0.10 at 18-20 years of age」（PubMed 摘要亲核；任务书曾误记刊名为 Current Directions in Psychological Science，实刊 Twin Research and Human Genetics 16(5):923–930，随文纠正）。元分析总锚：Polderman et al. 2015（Nature Genetics）：17,804 个性状、2,748 篇、约 1456 万双生子对，「across all traits the reported heritability is 49%」，69% 性状符合纯加性模型（亲核）；Plomin & Deary 2015：「The heritability of intelligence increases from about 20% in infancy to perhaps 80% in later adulthood… Assortative mating is greater for intelligence (spouse correlations ~0.40)」（亲核）。

MISTRA 与其战争（双向并列）。Bouchard et al. 1990（Science 250:223–228）摘要（主笔亲核）：「about 70% of the variance in IQ was found to be associated with genetic variation」，且自加限定 「This evidence for the strong heritability of most psychological traits, sensibly construed, does not detract from the value or importance of parenting, education, and other propaedeutic interventions.」。样本审计（Joseph 2022 据原文 p.223/226）：1990 发表时实际 56 MZA+30 DZA 对，IQ 相关仅基于 42–48 对 MZA（WAIS 48／Raven 42／FPC 43），未报告任何 DZA 相关；2000 年终点样本 81 MZA+56 DZA。Joseph 2022（Human Development 66:48–65）摘要：「To arrive at the conclusion that ‘IQ is strongly affected by genetic factors,’ Bouchard and colleagues omitted their control group reared-apart dizygotic twin (DZA) IQ score correlations… I conclude that the MISTRA IQ study failed to discover genetic influences on IQ scores…」；[Bouchard 2023（Twin Research and Human Genetics 26:133–142）摘要回击：「(a) his criticisms of previous TRA studies depend on sources that were discredited prior to MISTRA… (d) his claim that MISTRA should be evaluated in the context of psychology’s replication crisis is refuted. The TRA studies are constructive replications… His endeavor is a concatenation of elision and erroneous statistical/scientific [practice]」](https://gwern.net/doc/genetics/heritable/adoption/2023-bouchard.pdf)。发表往来（Bouchard 2023 自述）：Human Development 编辑未邀回应、拒绝反驳稿，经申诉后要求软化措辞，Bouchard 撤稿转投 TRHG——双方并列，本报告不裁此战：它要进入权重计算的是「70% 背后的实际样本结构与未报 DZA」这一审计事实，与「DZA 非对照组隐瞒、系 Science 简报格式与样本小之故」这一辩护事实。

EEA 批评（虚无派最重的方法论炮弹，如实登记）。同环境假设：Joseph 2015《The Trouble with Twin Studies》ch.7：「the twin method’s controversial MZT-DZT twin ‘equal environment assumption’ is not supported by the evidence, and that MZT pairs grow up experiencing much more similar environments and treatment, and experience much greater levels of identity confusion and psychological attachment, than experienced by DZT pairs… genetic interpretations of all past, present, and future MZT–DZT twin method comparisons in the social and behavioral sciences must be rejected」（作者自撰章节摘要）。场内防线（三层）：其一，EEA violations 的方向与量级可被测量且已被多次测量，多数效度研究认为不足以推翻主结论；其二，分开抚养设计（MZA）绕过 EEA 的一部分；其三——最硬的一层——分子方法不经过双生子假设独立收敛（见第七节 GCTA/within-family）。rGE 补注：环境不是外生变量——Scarr & McCartney 1983 提出基因型→环境三型相关（passive/evocative/active）：「experience is directed by genotypes」（摘要亲核；PubMed 把原文「→」转写为「greater than/leads to」，引用时恢复原符号）——这既是 h² 估计的系统性抬高机制（环境被基因「染色」后计入基因项），也是 Dickens-Flynn 乘子的微观基础（第五节）。

第四层裁决：双生子信号是真的、跨设计收敛的（亲子/收养/分开抚养同向），.86/.60 与「儿童 .45→成年 .75–.80」是几十年、上万对双生子的地基读数——「双生子研究什么都没测到」未立。但同层四条边界同样真：①同一批数据在母体效应模型下 narrow h² 可从 .5–.8 掉到 .3（Devlin 1997，登记在 Visscher Box 4）；②MISTRA 的「70%」背后是 42–48 对 MZA 且未报 DZA 相关，Joseph-Bouchard 2022/2023 战争未裁；③EEA 批评在场内有登记有防线，但不是可以假装不存在的异端；④rGE 意味着「基因项」里本来就混着被基因染色的环境——h² 是「基因+被基因吸引来的环境」的联合账目，这一点连场内都承认。拿这个带边界的信号去说「个体智力八成天定」，是第五节的弗林效应和第六节的组间判例要来处理的事。

五 弗林回旋镖：环境敏感性最硬的证据

如果 h²≈.8 读作「智力八成天定且环境动弹不得」，那么 20 世纪发生了一件不可能的事：全世界 IQ 分数暴涨了一个多标准差。这一节称量这件「不可能的事」——它同时是命定派的最大反证和虚无派的最大诱惑（「既然分数能整体漂移，测验是不是什么都没测」），两边都要挨这一镖。

唯象一手（摘要主笔亲核）。Flynn 1987（Psychological Bulletin 101(2):171–191）：「Data from 14 nations reveal IQ gains ranging from 5 to 25 points in a single generation. Some of the largest gains occur on culturally reduced tests and tests of fluid intelligence… The Dutch data proved the existence of unknown environmental factors so potent that they account for 15 of the 20 points gained.」（PsycNet 记录 Wayback 存档亲核；APA 同刊 101(3) 有 Correction：原文误把瑞士增涨幅度假说归于 U. Schallberger，已澄清）。注意摘要里的重量级表述是「单一代际 5–25 分」，而 Flynn 本人的假说是挑衅性的：「The hypothesis that best fits the results is that IQ tests do not measure intelligence but rather a correlate with a weak causal link to intelligence.」。「约每十年 3 分」的通行量级：Flynn 1987 正文为扫描件未能逐字核到，以两处亲核来源入账——Neisser 1996：「The average gain is about three IQ points per decade; more than a full standard deviation since, say, 1940.」（镜像亲核；总结节另给出「Mean IQs have increased more than 15 points–a full standard deviation–in the last fifty years」）；Trahan et al. 2014 元分析（Psychological Bulletin 140(5)）：现代（1972 后）Stanford-Binet 与 Wechsler 53 个比较，2.93 分/十年（95%CI 2.3–3.5），「was not consistent with the hypothesis that the Flynn effect is diminishing」（亲核）。

增益不在 g 上（虚无派拿不走整局的闸门）。Wicherts et al. 2004（Intelligence 32:509–537）对五个数据集（荷兰 WAIS 1967–1999、DAT 1984–1995、RAKIT 1982–1993，另重分析 Must et al. 与 Teasdale & Owen）做多组验证性因子分析：「The results of multigroup confirmatory factor analyses clearly indicate that measurement invariance with respect to cohorts is untenable.」（摘要亲核）；丹麦 BPP 结果正文判词：「Thus, we conclude that the Flynn effect found in this Danish comparison cannot be explained by an increase in latent intelligence (i.e., factor mean differences between cohorts).」（p.519 亲核）。含义双方各挨一半：分数的代际漂移不是潜变量 g 层面的整体抬升（虚无派不能拿它证明「g 跟着环境跑了」）；但代际之间测验的测量不变性不成立（命定派不能拿同一把尺子跨年代读「命定度」）。

上升与反转都是环境（家庭内设计）。Bratsberg & Rogeberg 2018（PNAS 115(26):6674–6678）：挪威征兵行政数据、出生队列 1962–1991（任务书曾误记为「1970–2009 出生队列」，摘要原文为 1962–1991，随文纠正），「the observed Flynn effect, its turning point, and subsequent decline can all be fully recovered from within-family variation. The analysis controls for all factors shared by siblings and finds no evidence for prominent causal hypotheses of the decline implicating genes and environmental factors that vary between, but not within, families.」（亲核）——上升、拐点、反转全部在家庭内复现，基因库变化假说与家庭间环境假说同时出局。

环境机制四锚（全亲核）。教育：Ritchie & Tucker-Drob 2018（Psychological Science 29(8)）：142 个效应量、42 数据集、>60 万人、三类准实验设计，「consistent evidence for beneficial effects of education on cognitive abilities of approximately 1 to 5 IQ points for an additional year of education… the effects persisted across the life span… Education appears to be the most consistent, robust, and durable method yet to be identified for raising intelligence.」（亲核）。收养：van IJzendoorn, Juffer & Poelhuis 2005（Psychological Bulletin 131(2)）：62 研究、17,767 名收养儿童，「Adopted children scored higher on IQ tests than their nonadopted siblings or peers who stayed behind… Adopted children did not differ from their nonadopted environmental peers or siblings in IQ」（亲核）。铅：Lanphear et al. 2005（EHP 113(7)）：7 国 1,333 名儿童汇总，血铅 2.4→30 μg/dL 对应 -6.9 分（95%CI 4.2–9.4）；分段 2.4→10 段 -3.9、10→20 段 -1.9、20→30 段 -1.1，「For a given increase in blood lead, the lead-associated intellectual decrement for children with a maximal blood lead level < 7.5 μg/dL was significantly greater」（亲核）——低段更陡：最穷的孩子为每一微克铅付出最贵。另有 EHP 2019 勘误（10.1289/EHP5685，内容未核）与一篇统计再评估（未一手核对，入账标注）。

悖论与其解法（本篇理论核心）。摆在桌面上的矛盾：h² 高（个体内差异主要由基因解释）× 代际均值巨变（环境敏感性）。Dickens & Flynn 2001（Psychological Review 108(2):346–369）给出形式模型（摘要主笔亲核）：「The authors present a formal model of the process determining IQ in which people’s IQs are affected by both environment and genes, but in which their environments are matched to their IQs. The authors show how such a model allows very large effects for environment, even incorporating the highest estimates of heritability.」（该文有 Erratum：Psychol Rev 2001;108(3):549，并有 2002 年两轮评论与回应）。机制（作者 Brookings 自撰通俗版，主笔亲核三句）：基因优势的持久性使其被匹配到更优环境——「people who have an advantage for a particular trait will become matched with superior environments for that trait; and that genes can derive a great advantage from this because genetic differences are persistent.」；社会乘子——「when something raises the average performance of society, that rise becomes a powerful cause in its own right, and raises the average performance further, and raises it further, until the original rise is greatly multiplied.」；遮蔽——「It shows that kinship studies hide or ‘mask’ the potency of environmental influences on IQ.」。一句话：亲缘研究里，环境的推力被「基因吸引环境」的匹配过程记进了基因账；而一个作用在全社会的环境变化（营养、教育、测验熟悉度）经乘子放大后，可以在 h²=.8 的世界里把均值推走一个标准差。「高遗传率与大环境效应悖论」不是悖论，是记账错觉。

第五层裁决：弗林效应是真的、量级是每十年约 3 分（14 国 5–25 分/代；Trahan 2.93 [2.3,3.5]）、方向上有升有停有反转（挪威家庭内复现全套曲线）、机制上有教育/收养/铅/营养四个独立锚——「环境动不了 IQ」在群体均值层面被定量否证，命定链第二跳再中一镖。但回旋镖对称地飞回虚无派：①增益不在 g 载荷上（Wicherts 2004）——代际漂移没有证明「g 随环境漂移」，潜变量层面发生了什么仍是开放的；②Flynn 本人「weak causal link」假说是他提出的待选假说而非被证结论，与第一节的终身稳定 r=.66、跨域预测效度同存一库；③个体层面的双生子信号与群体层面的均值漂移，在 Dickens-Flynn 模型里不但不矛盾，而且是同一台机器的两个读数。「h² 高」与「弗林效应大」同时为真，这才是本题的原始地貌；任何只能容纳其中一半的叙事都是升格。

六 组间差异证据室：什么已知，什么推论非法

最高规格纪律声明（本节适用）：本节只陈列三件事——①存在什么证据（双方最强论证并列）；②何种推论合法、何种非法（这是裁决）；③何处是诚实空位。本节不裁定任何组间差异的最终因果构成——那是科学上至今未填的空位；本节不评估任何群体的价值；所有数字只回答「研究测到了什么」，不回答「谁比谁好」。

主流定锚（APA 1996，全部主笔亲核）。测验偏差：「Considered as predictors of future performance, the tests do not seem to be biased against African Americans」（p.93，经 Rushton & Jensen 2005 p.241 逐字转引）。组内/组间因素分离：「the factors that create individual differences within a population may be unrelated to those that establish differences between populations」（p.95，经 Neisser 1997 答复文转述，亲核）。组间差异总判决（镜像全文亲核）：「Explanations based on factors of caste and culture may be appropriate, but so far have little direct empirical support. There is certainly no such support for a genetic interpretation. At present, no one knows what causes this differential.」——字样注记：Rushton & Jensen 2005 引作「no support」，Cambridge 书章与镜像作「no such support」，以镜像字样为准；页码各源标 93/94/97 不一。报告 §V 开篇另有：「There is not much direct evidence on this point, but what little there is fails to support the genetic hypothesis.」（亲核）。总结节：「Several culturally based explanations of the Black/White IQ differential have been proposed; some are plausible, but so far none has been conclusively supported. There is even less empirical support for a genetic interpretation. In short, no adequate explanation of the differential between the IQ means of Blacks and Whites is presently available.」（亲核）。

Eyferth 1961（德国占领军子女）。设计：二战后驻德美军士兵的私生子女，由德国白人母亲抚养。Nisbett 2005（同刊评论）：「Eyferth (1961) examined the IQs of several hundred German children fathered by Black GIs… The children of the Black GIs had an average IQ of 96.5. The children of the White GIs had an average IQ of 97. Because the (phenotypic) Black–White gap in the military was similar to that for the U.S. population, these data imply that the Black–White gap in the U.S. population as a whole is not genetic, even in part (Flynn, 1980, pp. 87–88).」（「96.5」「average IQ of 97」主笔亲核）；APA 1996 §V 的 Eyferth 段（镜像亲核）：「there was no mean difference between the test scores of those children」。反方三点保留（Rushton & Jensen 2005 引 Loehlin et al. 1975, pp.126–128，「French North Africans」亲核）：①孩子测试时年龄尚小（尚在 h² 随年龄上升的窗口之前）；②约 20–25% 的「黑人」父亲实为法属北非人；③黑人士兵经过入伍筛选。裁决形态：Eyferth 是「组间零遗传成分」假说的支持性证据而非判决——样本年龄、父系构成与筛选效应都是合法保留；但它同样使「组间大遗传成分」假说失去了一个本可得分的位置。

Minnesota 跨种族收养（数字以 Neisser 1997 亲核为准）。1976 年初测（约 7 岁）：Scarr & Weinberg 1976（American Psychologist 31(10):726–739），130 名黑人/混血儿童被优势白人家庭收养，摘要（OpenAlex 重建，含 OCR 伪迹，入账标注）：「The socially classified black adoptees… scored above the IQ and the school achievement mean of the white population.… The high IQ scores of the socially classified black adoptees indicate malleability for IQ under rearing conditions that are relevant to the tests and the schools.」。数字（Neisser 1997 答复文，主笔亲核）：黑人被收养者 106.3、白人 111.5、收养家庭亲生子女 116.7、「a full standard deviation above the expected IQ mean of Blacks in Minnesota」、双亲皆黑者 96.8；Scarr & Weinberg 由此结论 「the social environment plays a dominant role in determining the average IQ level of Black children」。1992 年随访（约 17 岁）：Weinberg, Scarr & Waldman 1992（Intelligence 16(1):117–135）：黑人被收养者均分降至 96.8、双亲皆黑者 89.4（Neisser 1997 亲核）；作者方摘要结论仍为 「Being reared in the culture of the tests and the culture of the schools benefits all children’s IQ scores and school achievements.」。双向保留：Neisser 1997 同时指出 「race and preadoptive experience were strongly confounded in the Minnesota study… the Minnesota data provide no clear evidence for the genetic hypothesis. But it is only fair to say that they do argue against certain versions of the environmental hypothesis (pending the necessary Flynn effect corrections).」（亲核）；且两次测试换过测验与常模，弗林效应校正完成前「raw figures like those above are relatively meaningless」（亲核）。后续交锋（Lynn 1994、Levin 1994 批评；Waldman, Weinberg & Scarr 1994 回应，Intelligence 19:13–44）未裁。Moore 1986：黑人儿童被黑人中产家庭收养均分 104、被白人中产家庭收养 117（Nisbett 2005 转述，「104」「117」亲核）（Moore 1986 摘要原文未能取得，psycnet 拦截，入账标注；设计 23+23、2 岁前收养、黑人女主试盲测，经教材转述）。

差距缩窄与群体均值的可移动性。Dickens & Flynn 2006（Psychological Science 17(10)）：9 个标准化样本、4 个主要测验，「Blacks gained 4 to 7 IQ points on non-Hispanic Whites between 1972 and 2002. Gains have been fairly uniform across the entire range of Black cognitive ability.」（亲核；任务书曾误记「1947–2002」，摘要原文为 1972–2002，随文纠正）。30 年缩窄 4–7 分——若差距是固定基因结构，这个速度不可解释；APA 1996 早已点出同一逻辑：「The Flynn effect shows that environmental factors can produce differences of at least this magnitude.」（总结节，亲核）。

刻板印象威胁（效应真、解释边界也真）。Steele & Aronson 1995（JPSP 69(5)）：「Blacks underperformed in relation to Whites in the ability-diagnostic condition but not in the nondiagnostic condition (with Scholastic Aptitude Tests controlled).」（摘要亲核）。解释边界：Sackett, Hardison & Cullen 2004（American Psychologist 59(1)）：「this research is widely misinterpreted in both popular and scholarly publications as showing that eliminating stereotype threat eliminates the African American-White difference in test performance. In fact, scores were statistically adjusted for differences in students’ prior SAT performance… The authors caution against interpreting the Steele and Aronson experiment as evidence that stereotype threat is the primary cause of African American-White differences in test performance.」（亲核）；交锋链 Steele & Aronson 2004 同期回应（Am Psychol 59(1):47–48）及 2005 年三篇评论在案。环境变量的效应是真实的实验效应，「消除威胁即消除差距」是公共叙事的过度收割——环境派自己的升格链也在案。

反方最强论证（如实陈列）。Rushton & Jensen 2005（Psychology, Public Policy, and Law 11(2):235–294）摘要（主笔亲核）：「The culture-only (0% genetic–100% environmental) and the hereditarian (50% genetic–50% environmental) models of the causes of mean Black–White differences in cognitive ability are compared and contrasted across 10 categories of evidence… The new evidence reviewed here points to some genetic component in Black–White differences in mean IQ.」（十类证据：全球分布、g、遗传率、脑大小、跨种族收养、混血、回归、生活史、人类起源、环境变量）。注意反方自己的方法论声明（正文 Section 2，p.239，亲核）：「The cause of individual differences within groups has no necessary implication for the cause of the average difference between groups. A high heritability within one group does not mean that the average difference between it and another group is due to genetic differences, even if the heritability is high in both groups. However, within-groups evidence does imply the plausibility of the between-groups differences being due to the same factors, genetic or environmental.」——连 Rushton & Jensen 都在字面上承认组内 h² 不能直接外推组间；他们的论证形态是 Sesardic 所说的 H3（组内证据＋十类旁证的「概率提升」），而非 H1/H2。Jensen 的「default hypothesis」（1999 年 PSYCOLOQUY Precis 作者自述）：「the causes of the phenotypic differences between contemporary populations of recent African and European descent arise from the same genetic and environmental factors, and in approximately the same magnitudes, that account for individual differences within each population.」——这是待检验假说的自我定位，不是已证结论。

分子层的硬约束（PGS 移植性）。Martin et al. 2019（Nature Genetics 51:584–591）：「those available today are several times more accurate in individuals of European ancestry than other ancestries. This disparity is an inescapable consequence of Eurocentric biases in genome-wide association studies… clinical uses of PRS today would systematically afford greater improvement for European-descent populations.」（亲核；该文有 2021 年 Publisher Correction，仅图注面板互换，不影响结论）。Duncan et al. 2019（Nature Communications 10:3328）：2008–2017 年 PGS 研究 67% 纯欧裔、19% 纯东亚裔、仅 3.8% 涉非/西/原住民队列；欧裔衍生 PGS 在非欧裔样本预测表现显著更低（t=-5.97, df=24, p=3.7×10⁻⁶）（亲核；任务书所给 DOI 尾号 11112-x 有误，实为 11112-0，随文纠正）。EA4 FAQ 3.4/3.5（SSGAC 官方）：欧裔样本 PGI 解释约 13% 方差，而「the polygenic index predicts only about 2% of the variance in individuals with African genetic ancestries」；「polygenic indexes of individuals from different genetic ancestry groups cannot be meaningfully compared.」（FAQ PDF 主笔亲核「cannot be meaningfully compared」句及上下文）。含义：连「同一祖源组内」的多基因评分都还在个位数到十几个百分点的方差解释水平，跨祖源更是断崖——用今天的 PGS 给组间差异定遗传成分，器械本身还不够格。

本节裁决。推论合法性清单：合法——组内 h² 真实存在、组间均值差真实存在（在特定测量时点）、Eyferth/收养/缩窄/刻板威胁构成环境侧的证据集合、R&J 的十类旁证构成遗传侧的假说集合、双方都可以继续设计判决性研究。非法——「组内 h² 高 → 组间差异有遗传成分」（Lewontin-Block-Schraiber 数学封死，连 R&J 字面承认）；「组间均值差 → 任一组的优劣排序」（ASHG 2018：用遗传学给群体排序「demonstrates a fundamental misunderstanding of genetics」）；「环境证据集合 → 组间差异已证 0% 遗传」（Eyferth/Minnesota 都有合法保留，环境侧同样没有判决书）。诚实空位——组间差异的因果构成未知：APA 1996「no adequate explanation… presently available」至今未被推翻；Gottfredson 1994 第 22 条 「There is no definitive answer to why IQ bell curves differ across racial-ethnic groups… Most experts believe that environment is important in pushing the bell curves apart, but that genetics could be involved.」（亲核）与本库判读一致：空位是空位，两边都不许拿空位当判决书用。

七 分子新战场：GWAS、missing heritability 与家庭内革命

双生子的信号（第四节）必须接受分子层的独立复核。2010 年后 GWAS 给出了不含双生子假设的第二套账目——它确认了信号，也改写了数字，并且正在改写「遗传效应」这个词的含义。

GWAS 地基（全亲核）。Davies et al. 2018（Nature Communications 9:2098）：CHARGE+COGENT+UK Biobank 共 300,486 人，148 个全基因组显著独立位点、709 个基因；「Using polygenic scores, up to 4.3% of variance in general cognitive function is predicted in independent samples.」（亲核；该文有 2019 年 Erratum）。Sniekers et al. 2017（Nature Genetics 49:1107–1112）：78,308 人元分析；「Despite intelligence having substantial heritability (0.54)…」；儿童（0.45）与成年（0.80）双生子 h² 差异巨大，但「we show substantial genetic correlation (rg = 0.89, LD score regression P = 5.4 × 10-29)」；正文 SNP h²=0.20（SE=0.01），「likely to be biased downwards」（亲核）。教育侧：Lee et al. 2018（EA3，Nature Genetics 50:1112–1121）：约 110 万人、1,271 个独立显著 SNP；多性状联合 PGS 解释教育年限 11–13%、认知表现 7–10% 方差（亲核）。Okbay et al. 2022（EA4，Nature Genetics 54:437–449）：约 300 万人、3,952 个近独立显著 SNP；PGI 解释 EA 方差 12–16%；「Direct effects (i.e., controlling for parental PGIs) explain roughly half the PGI’s magnitude of association with EA and other phenotypes.」（亲核）。

missing heritability 与内部张力。GCTA 原型：Yang et al. 2010（Nature Genetics 42:565–569）：身高全部 SNP 联合解释 45% 方差，「most of the heritability is not missing but has not previously been detected because the individual effects are too small to pass stringent significance tests」（亲核）。认知性状的 SNP h² 带：Davies 2018 引言：「Twin studies find that general cognitive function has a heritability of more than 50%… SNP-based estimates of heritability for general cognitive function are about 20–30%. However, these estimates might increase to about 50% when family-based designs are used to retain the contributions made by rarer SNPs.」（亲核）；历年点估计：Davies 2011=0.51、「40% of the variation in crystallized-type intelligence and 51% of the variation in fluid-type intelligence」（2011 摘要亲核）、Davies 2015=0.29、Trampush 2017=0.22、Sniekers 2017=0.20、Davies 2018=0.25。内部张力如实登记：2011 年小样本 GCTA 给出 40–51%，后来更大样本收敛到 20–30% 带——早期估计本身就在移动；而双生子 50–80% 与 SNP 20–30% 之间的缺口（missing heritability）有稀有变异、LD 不完全、以及——下一项——环境染色三个竞争解释。

家庭内革命（本篇分子层主菜）。标准 GWAS 的效应里混着三种东西：直接遗传效应、人口结构/选型交配、以及父母基因经环境施加的间接效应（genetic nurture）。within-family 设计用同胞间随机分离把它们拆开。Howe et al. 2022（Nature Genetics 54:581–592）：19 队列 178,086 名同胞、25 个性状；「Within-sibship GWAS estimates were smaller than population estimates for height, educational attainment, age at first birth, number of children, cognitive ability, depressive symptoms and smoking.」（摘要亲核）；正文缩水幅度（主笔亲核）：关联估计缩水——EA 47%（95%CI 41–52）、认知能力 22%（6–37）、身高 10%；SNP h² 衰减——EA 76%（0.13→0.04，P=5.3×10⁻²⁶）、认知 44%（0.24→0.14）、身高 17%（0.41→0.34）。Young et al. 2022（Nature Genetics 54:897–905，snipar 方法）：直接效应与标准 GWAS 效应的相关——EA r=0.739（s.e. 0.086）、认知 r=0.490（s.e. 0.086）；「Our results demonstrate substantial confounding bias in standard GWASs for some phenotypes.」（亲核；任务书所引标题对应正式发表版，bioRxiv 2018/2020 预印本标题为「…for genome-wide estimation of direct and indirect genetic effects」，随文注明）。综合判词（Coop & Przeworski 2022, Evolution 76(4):846–853，三句主笔亲核）：教育 PGS 里「direct genetic effects for educational attainment are estimated to account for as little as one fourth of the variance in the PGS of a standard GWAS… The remaining three-fourths reflect a tangled mess—a braid, as Harden refers to it—of genetic and environmental effects.」；「current PGS for educational attainment are neither interpretable nor particularly meaningful.」。这一刀切的方向是双向的：它砍向命定派（「基因效应」里约一半到四分之三可能是环境通道与结构混杂），也砍向虚无派里「GWAS 什么都没找到」的一翼（直接效应非零、rg=0.89 跨年龄稳定、位点富集在脑发育与神经通讯通路——信号真实存在）。

G×E：一个「美国现象」的不稳定复制史。Turkheimer et al. 2003（Psychological Science 14(6)）：国家围产期项目 7 岁双生子，「in impoverished families, 60% of the variance in IQ is accounted for by the shared environment, and the contribution of genes is close to zero; in affluent families, the result is almost exactly the reverse.」（亲核）——贫困层 h²≈0、富裕层 h² 高：h² 是环境的函数，这若稳定成立，本身就是「h² 是常数」读数的死刑。Tucker-Drob & Bates 2016（Psychological Science 27(2)）：跨国元分析，「In U.S. studies, we found clear support for moderately sized Gene × SES effects. In studies from Western Europe and Australia, where social policies ensure more uniform access to high-quality education and health care, Gene × SES effects were zero or reversed.」（亲核）。Figlio et al. 2017（PNAS 114(51)）：佛州 1994–2002 出生队列兄弟姐妹与双生子行政数据，「the largest, most population-diverse consideration of this hypothesis to date. We found no evidence of SES moderation of genetic influence on test scores, suggesting that articulating gene-environment interactions for cognition is more complex and elusive than previously supposed.」（亲核）——「美国现象」在美国最大样本里也没复制出来。G×E 的当前形态：方向有理论引力、证据不稳定、量级与边界未裁（任务书曾指「Fontanillas et al. Nature Human Behaviour 2024」大规模 meta，经子代理多渠道检索未能证实存在，不入账）。

场内立场谱与自我设限（全部亲核）。乐观翼：Plomin & von Stumm 2018（Nature Reviews Genetics 19:148–159）：「Recent genome-wide association studies have successfully identified inherited genome sequence differences that account for 20% of the 50% heritability of intelligence.」（亲核）。应用翼与批评：Harden《The Genetic Lottery》（2021）主张承认基因运气是理解不平等的必要输入（普林斯顿大学出版社官方简介）；批评（Coop & Przeworski 2022）：「its central argument mischaracterizes where the field of human genetics stands and what it promises.」（亲核）。共识文件：Meyer et al. 2023（Hastings Center Report 53(S1)，19 人共识、署名含 Turkheimer 与 Harden 本人）：涉及敏感性状且按种族/族裔/祖源分组比较的 SBG 研究「requires a compelling justification to be conducted, funded, or published」（亲核）。EA4 官方 FAQ 的自我设限（红线级，主笔亲核八句）：「Does this study show that an individual’s level of educational attainment (or any other outcome) is determined, or fixed, at conception? Do genes determine the choices we make and who we become? No and no.」；个体预测——「the score fails to predict the vast majority (87%) of variation in years of education across individuals. Many of those with low polygenic indexes go on to achieve high levels of education, and a large proportion of those with high polygenic indexes do not complete college.」；上限——「the additive effects of common SNPs will only ever predict about 20% of the variance in educational attainment across individuals.」；娱乐化警示——「these individual results are not meaningful predictions and should be regarded essentially as entertainment.」；实践应用——「Should practitioners… use the results of this study to make decisions? No. Doing so would be extremely premature and unsupported by the science.」。做这项研究的人自己把命定派的结论一条条预先否决了。

第六、七层合并裁决：分子层 15 年给出的是「确认＋改账」。确认：信号真实（数千位点、跨年龄 rg=0.89、脑发育通路富集），双生子不是孤证——虚无派「整个领域破产」再断一跳。改账：①解释方差从双生子的 50–80% 收缩到 SNP 的 20–30%、PGS 的 4–16%，且 within-family 再把「直接遗传效应」砍去约一半（EA 缩水 47%、认知 22%；EA4 摘要自承直接效应约一半）；②「基因效应」里混着 genetic nurture 与人口结构——父母不给孩子的那 half 基因组也在通过家庭环境推孩子；③PGS 跨祖源断崖（欧裔 13% vs 非裔 2%）＋EA4 官方「cannot be meaningfully compared」——组间推论在分子层同样没有器械。命定派想要的「DNA 判决书」，今天的分子科学自己写明「extremely premature」；虚无派想要的「全是混杂」，又被 rg=0.89 与非零直接效应挡回。

八 升格链审计：命定派四跳 × 虚无派四跳

把前七层合并成两条对称的链条，逐跳出示在案判词。

命定链（每跳都有在案反证）。

  • 跳一：h²≈.8 →「智力八成由基因决定（你这个人）」。判词：方差比≠个体因果份额。h² 是「这群人、此时、此环境分布下」差异的相关结构（Visscher 2008 定义句，亲核）；「phenotypes are not passed on but genes are」（Box 2 误读一，亲核）；同一个体的能力里没有哪 80% 贴着基因标签——正如身高 h²≈.8 不意味着你的身高八成都「在基因里」而两成「在营养里」。断。
  • 跳二：「基因决定」→「干预无用」。判词：四重否证在案——近视 h² 88% 配镜（Dirani 2006＋Bishop 2015）；PKU 单基因病被饮食「virtually eliminated」（NICHD）；弗林效应每十年 3 分且家庭内复现（Flynn 1987／Trahan 2014／Bratsberg & Rogeberg 2018）；教育每多一年 +1~5 分且终身持久（Ritchie & Tucker-Drob 2018）。「heritability does not imply immutability」（APA 1996 p.86，亲核）。断。
  • 跳三：组内 h² → 组间差异有遗传成分。判词：Lewontin 1970 双思想实验→Block 1995「licenses no conclusion」→Schraiber & Edge 2024 数学一般化（「Perfect knowledge of within-group heritability provides no information about between-group heritability」，亲核）；连 Rushton & Jensen 2005 字面承认「no necessary implication」（亲核）。断。（H3 形态——组内证据加旁证的概率提升——是合法假说形态，但它是假说，不是推论完成时。）
  • 跳四：组间均值差 → 群体优劣/价值排序。判词：范畴错误叠数学错误。分布重叠远大于均值差；ASHG 2018「Any attempt to use genetics to rank populations demonstrates a fundamental misunderstanding of genetics」（亲核）；「racial purity」科学上无意义（亲核）。断。

虚无链（对称，同样每跳在案）。

  • 跳一：「IQ 只测应试能力/书本知识」。判词：跨域硬终点在案——死亡率（Whalley & Deary 2001：15 分→存活 0.79；Calvin 2011：1SD→-24%，110 万人）、发病率（Gottfredson & Deary 2004 苏格兰近全人口）、终身稳定 r=.66（Deary 2004/2013）。Gottfredson 1994 第 1 条 「It is not merely book learning, a narrow academic skill, or test-taking smarts.」（亲核）。Boring 1923 名句是操作定义的方法论自觉，被脱离语境当成「自认循环」使用（俄勒冈学位论文的语境警示在案）。断。
  • 跳二：「g 是因子分析的幻觉/实体化谬误」。判词：全正相关矩阵是数据不是模型（1904 年观察到、460+ 数据集复现）；「g 该被实体化成什么」确实未裁——但提出者本人与主流都拒绝实体化（Spearman 不把 g 当物，Schneider & McGrew Box 4.1；APA「consistent with, but does not prove」）。Gould 对因子分析的指控被 J. B. Carroll 当面驳回（[「he still fails to understan[d]」](https://www.arvindguptatoys.com/arvindgupta/mismeasures-of-man.pdf)）；Gould 自己的 Morton 正典案例被 Lewis 2011 反指控且三方未决（第二节）。把「g 的本体地位未裁」读成「g 的数据不存在」，是虚无派自己的升格。断。
  • 跳三：「双生子方法论全破产」。判词：EEA 与 MISTRA 批评真实登记（第四节），但①Polderman 2015 跨 17,804 性状 h² 均值 49%——IQ 不是孤例；②分子方法不经双生子假设独立收敛（GCTA、跨年龄 rg=0.89、within-family 直接效应非零——第七节）；③Devlin 1997 的模型敏感性（narrow .3 vs broad .5）说明点估计该带区间，而非信号为零。断。
  • 跳四：「测验是优生工具 → 测验无信号」。判词：遗传谬误。挪用史每一拍都真（第二节），但①发明人原意是安置不是排名（Binet 1905 四句，亲核）；②挪用者自己有公开退却链（Goddard 环境限定、Terman 1924、Brigham 1930「without foundation」）；③「测验驱动 1924 法案」未过国会记录核查（S&H 1983，亲核）；④唯象数据（信度、终身稳定、死亡率关联）不因这段历史改变一个数字。断。

九 双向裁决：命定未立 × 虚无未立

命定未立。「你的智力在你出生时被基因写死」这一命题，在其最弱的可读形式（「群体差异的方差结构」）上是真的，在它被讲述的每一个强形式（个体命定度／干预无用／组间天定／价值排序）上都不是科学的结论——而是对一个方差参数的连续四次越权宣读。做分子研究的人自己写着「No and no」「extremely premature」「essentially as entertainment」；做行为遗传学的人自己写着「heritability does not imply immutability」；连 1994 年那份最常被命定派引用的联署声明都写着组间问题「no definitive answer」。命定论不是这个领域的结论，是这个领域一百二十年来反复试图甩掉的读数方式。

虚无未立。「IQ 测不到任何真实的东西、整个领域是优生学伪装」这一命题，同样未立。器械信度 .9x、11 岁与 80 岁相关 .66、死亡风险 1SD→-24% 且儿童期 SES 调整几乎不动；g 的全正相关地基跨 460+ 数据集；双生子信号跨设计收敛并被分子层独立确认；弗林效应、教育、收养、铅四个环境锚说明环境真实可塑——但没有一条支持「测验无信号」，它们支持的是「信号真实且边界清晰」。挪用史必须披露（包括两边的升格），披露之后，数据还在原地。

诚实空位（四项，挂起不裁）：①g 的本体地位（统计结构之上是什么，神经/机制层面未裁）；②组间差异的因果构成（APA 1996「no adequate explanation」至今；分子器械尚未够格）；③missing heritability 的剩余分解（稀有变异 vs LD vs 环境染色的配比未裁）；④G×E 的稳定形态（Turkheimer→Tucker-Drob→Figlio 一正一分一负，未裁）。

十 对称三向红线 · 关键来源 · 灵魂句

不升格（防命定）：不把 h² 读成个体命定度；不把「高 h²」读成「干预无用」；不把组内 h² 外推组间；不把组间均值差读成群体优劣；不把 PGS 读成个体预测器（EA4 FAQ：个体层面「essentially as entertainment」）；不把「遗传性状的直接效应存在」读成「教育的基因决定论」。

不虚无化（防虚无）：不把「g 的本体未裁」读成「g 是幻觉」；不把「h² 被误读」读成「h² 不存在」；不把双生子的模型敏感性读成全破产；不把挪用史当成擦除唯象数据的橡皮；不把「校正后 .54 有争议」读成「预测效度为零」；不把弗林效应读成「测验什么都没测」（终身稳定 r=.66 与代际漂移同存一库）。

不污名亦不神圣化：研究者个人立场、资助来源与数据命题分开审计（Pioneer Fund 谱系披露入权重，但不作反证；SPLC 转述未交叉核实者不直接引用）；叙事双方的失真都登记（Kamin 省略 Goddard 限定／S&H 属修正派史学；Gould 被反指控／Lewis 2011 非终局）；历史人物的退却链如实记录（Terman 1924、Brigham 1930、Goddard 1928）。高风险声明（复述定调）：本篇不评估任何个人或群体的价值，不作教育、临床、移民或社会政策建议；它只做一件事——把科学主张与其推论结构放上承重架。

关键来源（定锚级）：Neisser et al. 1996《Intelligence: Knowns and Unknowns》（APA task force 报告，全篇第一定锚）｜Visscher, Hill & Wray 2008, Nat Rev Genet（遗传率概念与五条误读）｜Binet 1905 一手英译｜Spearman 1904｜Dickens & Flynn 2001, Psychol Review（悖论与乘子）｜Flynn 1987, Psychol Bulletin｜Schraiber & Edge 2024, PNAS（组内→组间非法的数学封顶）｜Bouchard et al. 1990, Science（MISTRA）＋Joseph 2022＋Bouchard 2023｜Howe et al. 2022, Nat Genet（within-family）｜Okbay et al. 2022（EA4）＋EA4 官方 FAQ｜ASHG 2018 声明｜Snyderman & Herrnstein 1983, Am Psychol（1924 法案核查）｜Lewis et al. 2011, PLoS Biology（Gould-Morton）。

灵魂句：遗传率回答的只是「在这群人、这个年代、这个环境分布里，个体差异与基因型差异相关几成」；它对你一个人是什么、对另一群人差多少、对未来能改多少，三个问题都保持沉默——命定派替它抢答了全部三个，虚无派则没听见它确实回答了第一个。 而那台被双方轮流抢夺的仪器，从 Binet 的桌面到今天三百万人的 GWAS，始终只在做同一件事：测量，然后等待有人忍住不把读数读成命运。

机制裁决红队风第九十二篇 · 对称双向红队第八十七篇 · 全库 149 篇。调研纪律：六捆并行一手调研（唯象地基／遗传率方法论／叙事考古／弗林效应与环境／组间差异证据·最高规格纪律／分子遗传学与G×E）＋主笔亲核十八条独立渠道约九十处逐字比对（PubMed efetch 38 篇摘要批量逐字比对·Neisser 1996 APA 先行版镜像·Visscher 2008 PDF pdftotext·Binet 1905 与 Spearman 1904 Classics 全文·Jensen 1969 全文 PDF·EA4 FAQ PDF·Mainstream 1997 重印版 PDF·Neisser 1997 答复文 PDF·Rushton-Jensen 2005 正式版 PDF·ASHG 2018 Wayback PMC 镜像·Schraiber-Edge 2024 EuropePMC 全文 XML·Coop-Przeworski 2022 PMC·Wicherts 2004 全文 PDF·Snyderman-Herrnstein 1983 PDF·Saretzky 1982 ETS 备忘录 PDF·Nisbett 2005 评论 PDF·Flynn 1987 PsycNet Wayback·Dickens-Flynn 2001 Brookings 作者自撰·Gottfredson-Deary 2004 APS 摘要页）全部逐字一致，不凭记忆，引用带链接；扫描件无文本层或多源转引入账处逐条注明。任务书五处指针纠错随文注明（Calvin 2011=IJE 非 BMJ·Bouchard 2013=TRHG 非 Current Directions·Bratsberg-Rogeberg=1962-1991 出生队列·Dickens-Flynn 2006=1972-2002 非 1947 起·Duncan 2019 DOI 尾号 11112-0）；「Fontanillas et al. NHB 2024」未能证实存在不入账；Young 2022 以 Nat Genet 正式版入账；Scarr-Weinberg「Genetic differences and malleability」系 1983 年文。与自私基因篇的去重分界见定调声明。

标签: