5D Library › Scientific Methodology, Falsifiability and the Mathematical Statistics of the Replication Crisis · 中文
Chapter 7 Reading the Classics: Seeking Truth from Facts, No Trust without Evidence, No Conclusion from a Single Proof
This chapter reads the discussion of earlier chapters beside several classical Chinese texts. The aim is not to show that the ancients had already discovered falsifiability or statistical inference, but to see how earlier scholars understood the relation between “evidence” and “belief”, and to state plainly that they worked with historical texts and language, with no randomised controlled experiments and no theory of probability sampling or statistical inference. The English renderings of the classical passages are the author’s translations.
Seeking truth from facts: a prince’s library
Source text
河间献王德以孝景前二年立,修学好古,实事求是。从民得善书,必为好写与之,留其真,加金帛赐以招之。繇是四方道术之人不远千里,或有先祖旧书,多奉以奏献王者,故得书多,与汉朝等。(班固 撰,颜师古 注 1962)
颜师古注:务得事实,每求真是也。
(Book of Han, chapter 53, “The Thirteen Sons of Emperor Jing”)
In plain language: Liu De, Prince Xian of Hejian, was enfeoffed in the second year of Emperor Jing’s first reign period. He devoted himself to learning and loved the ancient texts, seeking true conclusions from real things. Whenever he obtained a good book from the people, he had a careful copy made and given to its owner, kept the original himself, and added gifts of gold and silk to attract more. So learned men came from every quarter regardless of distance, and those who had old books left by their ancestors often brought them to present to the prince; he thus acquired as many books as the imperial court. Yan Shigu’s gloss: strive to obtain the facts, and always seek what is truly right.
The framework says this passage comes from a “Biography of Prince Xian of Hejian”, which needs correcting: the Book of Han has no separate biography of that title; the passage is in chapter 53, “The Thirteen Sons of Emperor Jing”, the part of this joint account of Emperor Jing’s thirteen sons devoted to Liu De.
In its original context, “seek truth from facts” (实事求是) describes an attitude to collecting books and to learning: “keeping the original” (留其真), not replacing the original with a copy, which has something in common with what we now call keeping the raw data. It is not a method for testing hypotheses. The four characters later became the Qing textual scholars’ watchword against reading the classics by one’s own lights; this chapter discusses only the original sense and the Qing use.
The records are insufficient: Confucius’s restraint
In plain language: The Master said: I can speak of the rites of Xia, but (Xia’s descendant state) Qi cannot supply the proof; I can speak of the rites of Yin, but (Yin’s descendant state) Song cannot supply the proof. This is because the records and the men of learning are insufficient. If they were sufficient, I could use them to prove what I say. Zhu Xi glosses: “wen means written records; xian means worthy men”, so the “documents” (文献) include both texts and worthies who knew the old usages, roughly written sources and oral testimony.
What stands out in this passage is Confucius’s restraint: he believed he could describe the rites of Xia and Yin, yet would not present his account as settled where the evidence fell short. It connects with the Analects’ “to know what you know and to know what you do not know: that is knowledge” (Weizheng). As we shall see, Dai Zhen used exactly this saying against scholars who drew conclusions from a single piece of evidence. Chapter 7 of volume 26 discussed the Han Feizi’s “to be certain without cross-checking is folly”, which can be read alongside this; it is not repeated here.
How Ming and Qing textual scholars thought about evidence
Chen Di: against “harmonising pronunciations”
In the late Ming, Chen Di argued in his Study of the Ancient Pronunciation of the Mao Odes against the theory of “harmonising pronunciations” (叶音) that had been current since the Song. Some rhyming characters in the Book of Songs did not rhyme in later pronunciation; the harmonising theory’s solution was to read the character, at that point only, with another sound so that it rhymed. Chen Di held this to be wrong: the poets’ pronunciation rhymed in its own day, and the sounds had since changed:
Source text
盖时有古今,地有南北,字有更革,音有转移,亦势所必至。(陈第 撰,康瑞琮 点校 2008)
(Chen Di, preface to Study of the Ancient Pronunciation of the Mao Odes)
In plain language: times differ, ancient and modern; places differ, north and south; characters change their forms and sounds shift; this too is bound to happen.
For each character’s ancient reading Chen Di listed two kinds of evidence: rhymes within the Book of Songs itself, which he called “internal evidence” (本证), and evidence from other pre-Qin and Han texts, which he called “external evidence” (旁证). In the language of Chapter 1, the harmonising theory is a classic “patch after the fact”: each time a rhyme fails, assume that this character was read differently here; the assumption predicts nothing beyond the case it explains, so it can “explain” any failed rhyme. Chen Di’s hypothesis of ancient pronunciation took a risk: if a character had a certain ancient reading, it should be consistent across all its rhymes in the Book of Songs and in other ancient texts. That is a hypothesis counter-examples could refute.
Gu Yanwu: “mining copper from the mountain”
In the early Qing, Gu Yanwu held that “reading the Nine Classics begins with examining the script, and examining the script begins with knowing the sounds” (letter to Li Zide), making script and phonology the foundation for understanding the classics (顾炎武 著,华忱之 点校 1959). In one letter he compared some scholars’ works of his day to recasting old coins: “the ancients mined copper from the mountain; people today buy old coins, call them scrap copper and use them to fill the mould”, while his own Record of Daily Knowledge was “copper mined from the mountain”, obtained by “reading morning and night, searching over and over” (letter to a friend, no. 10). In today’s terms, he insisted on primary sources and original evidence rather than second-hand conclusions.
Dai Zhen: ten-tenths seeing and the single proof
Source text
寻求而获,有十分之见,有未至十分之见。所谓十分之见,必征之古而靡不条贯,合诸道而不留余议,巨细毕究,本末兼察。若夫依于传闻以拟其是,择于众说以裁其优,出于空言以定其论,据于孤证以信其通;虽溯流可以知源,不目睹渊泉所导,循根可以达杪,不手披枝肄所歧,皆未至十分之见也。以此治经,失「不知为不知」之意,而徒增一惑,以滋识者之辨之也。(戴震 著,汤志钧 校点 1980)
(Dai Zhen, letter to Yao Nai)
In plain language: what one finds by searching may be “ten-tenths seeing” or something short of it. Ten-tenths seeing must be verified in the ancient texts and hang together without exception, accord with principle and leave nothing open to dispute, examine every detail great and small, and attend to root and branch alike. But to guess something is right from hearsay, to pick the better of many opinions, to settle a conclusion with empty words, or to believe a thing holds generally on a single piece of evidence: even if following the stream one may infer the source without having seen where the spring rises, or following the root one may reach the treetop without parting every branch to see how it forks, none of these is ten-tenths seeing. To study the classics this way loses the meaning of “to know what you do not know”, and only adds one more confusion for the discerning to sort out.
The passage lists four inadequate kinds of reasoning: from hearsay, by picking among opinions, by empty words and from a single proof. “Relying on a single piece of evidence to believe it holds generally” (据于孤证以信其通) is the closest Qing original to the saying “no conclusion from a single proof”. In his Evidential Commentary on the Meanings of Terms in the Mencius Dai Zhen applied the same standard to moral principle: “only what all minds affirm alike may be called principle and righteousness; what has not reached such agreement remains a person’s opinion, not principle, not righteousness”; and “nothing is worse than being blinkered and thinking oneself wise, following one’s opinions and holding them to be principle” (戴震 著,汤志钧 校点 1980). What he opposed was taking a personal “opinion” for universal “principle”.
Wang Niansun and Wang Yinzhi: consistent here, valid elsewhere
Wang Niansun and his son Wang Yinzhi of Gaoyou were famous for collation and philology. In the preface to his Explanations of Particles in the Classics and Commentaries, Wang Yinzhi says his readings of function words must “fit when weighed against the passage itself and hold when tested in other chapters” (揆之本文而协,验之他卷而通) (王引之 撰 1984): an interpretation must make sense not only in the sentence where it arises but also wherever the same word appears in other texts. This is a systematic cross-check: an interpretation that works in only one place may simply have been tailored to that place.
The source of “no conclusion from a single proof”
Source text
一曰:凡立一义,必凭证据;无证据而以臆度者,在所必摈。……三曰:孤证不为定说。其无反证者姑存之,得有续证则渐信之,遇有力之反证则弃之。四曰:隐匿证据或曲解证据,皆认为不德。五曰:最喜罗列事项之同类者,为比较的研究,而求得其公则。(梁启超 著,朱维铮 导读 1998)
(Liang Qichao, Intellectual Trends in the Qing Period, summarising the style of the Qing “orthodox school”)
In plain language: first, whoever puts forward a claim must rest it on evidence; what is guessed without evidence must be rejected. Third, a single piece of evidence does not make a settled conclusion; where there is no counter-evidence the claim is provisionally kept, with further evidence it is gradually believed, and in the face of strong counter-evidence it is abandoned. Fourth, hiding or distorting evidence is regarded as unethical. Fifth, they loved to set out cases of the same kind side by side for comparative study and so arrive at general rules.
The four-character saying “no conclusion from a single proof” (孤证不立) is often presented as the words of some Qing scholar. On checking, it is closer to Liang Qichao’s summary of the Qianlong–Jiaqing scholarly style in Intellectual Trends in the Qing Period (“a single piece of evidence does not make a settled conclusion”, 孤证不为定说) than to any Qing scholar’s own words. Liang also wrote that this school’s “fundamental method of scholarship lay in ‘seeking truth from facts’ and ‘no trust without evidence’”.
Hu Shih’s ten characters
In “The Methods of Qing Scholars”, Hu Shih summed up the Qing method as “bold hypotheses, careful verification” and held it to share the spirit of modern science (胡适 著 1921). The saying often attributed to him, “say as much as your evidence allows” (有一分证据说一分话), circulates in wordings that do not agree (for example “with one part of evidence you may say only one part; only with three parts may you say three”); this book could not find that exact form in Hu’s writings and treats it as a paraphrase, not a quotation.
Reading together: resonance and boundaries
Set beside the earlier chapters, these texts show some genuine resonance:
- Against patches after the fact. Chen Di’s criticism of harmonising pronunciations is logically the same as Chapter 1’s criticism of “explaining it away afterwards”: an assumption made only to explain one anomaly, and predicting nothing else, is not evidence.
- Cross-checking from several sources, and counter-evidence. “A single piece of evidence does not make a settled conclusion … in the face of strong counter-evidence it is abandoned” and “hold when tested in other chapters” both require a conclusion to be consistently supported by several independent sources, and grant that counter-evidence can overturn it, in the spirit of Chapter 1’s falsifiability and Chapter 5’s independent replication.
- Full reporting. “Hiding or distorting evidence is regarded as unethical” is the same ethic as the demand of Simmons et al. in Chapter 3 to report all variables, all conditions, and results with and without exclusions.
- Grading the strength of a view. Dai Zhen’s division of views into “ten-tenths seeing” and less is close in spirit to this series’ separate marking of evidence, calculation and hypothesis.
The boundaries must be stated just as clearly:
- Different objects. Textual scholarship deals with historical texts, scripts and sounds; its “evidence” is textual testimony; it has no control groups, no random assignment, and cannot test a conclusion by running an experiment again.
- No language of probability. The distinction between “a single proof” and “many proofs” is qualitative. Textual scholarship had no tools for quantifying the strength of evidence, nor statistical methods for the possibility that several pieces of evidence are not independent; several ancient books may corroborate one another simply because they were copied from the same source.
- It had its own garden of forking paths. The method of phonetic loans is powerful, but if the standard “similar sounds may substitute” is too loose, almost any two characters can be shown to be interchangeable, and such explanations also become able to “explain anything”. This is the author’s methodological observation, not a criticism of any particular scholar.
- No “the ancients already knew”. The Qing scholars proposed no theory of falsifiability and no statistical inference; Hu Shih’s comparison of their method with modern science is a twentieth-century interpretation. What we can say is that, in their own field, they developed scholarly norms that opposed subjective conjecture, required corroboration from several sources and granted the force of counter-evidence, and that these norms resonate with the spirit of modern empirical science.
Chapter summary
“Seek truth from facts” comes from the Book of Han’s “Thirteen Sons of Emperor Jing” and originally described Prince Xian of Hejian’s way of collecting books, “keeping the original”; Yan Shigu glossed it as “strive to obtain the facts and always seek what is truly right”. The Doctrine of the Mean’s “no trust without evidence” concerns the conditions under which rites and authority win the people’s trust, and its “verification” includes ghosts, spirits and sages. Confucius’s restraint over insufficient records, Chen Di’s criticism of harmonising pronunciations, Gu Yanwu’s “copper mined from the mountain”, Dai Zhen’s judgement that “believing a thing holds generally on a single piece of evidence” falls short of “ten-tenths seeing”, Wang Yinzhi’s “hold when tested in other chapters” and Liang Qichao’s summary “a single piece of evidence does not make a settled conclusion” all reflect a scholarly style that opposed conjecture and required corroboration from several sources. They resonate with modern empirical science, but their object was historical texts, without experiment or probabilistic inference. The next chapter gathers the book’s tools into a checklist and a trial protocol.
References
Fan, Y. (2026). Scientific Methodology, Falsifiability and the Mathematical Statistics of the Replication Crisis: From Popper's Demarcation and p-Hacking to Ioannidis's Theorem and Pre-Registration, Read Beside Classical Chinese Evidential Scholarship [科学方法论、可证伪性与可重复性危机的数理统计] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23782104 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).