In classical Chinese thought the boundary between an individual, his predecessors, his superiors, his successors and his subordinates was vague. All were part of a single network, able to exert a greater or lesser influence upon one another.
This being so, it was not uncommon for texts to be attributed to people who could not possibly have written them. The Guanzi, to pick just one example, is a lengthy work on statecraft and self-cultivation that was transmitted under the name of the seventh century BCE statesman Guan Zhong for two thousand years, despite the fact that the style and vocabulary clearly shows it to have been written centuries after his death. For the compilers this did not matter: they felt that he would likely have approved of the content, and having their essays published under his name added lustre to their own work.
This being so, many modern scholars argue that we should accept the situation and acknowledge all ancient Chinese texts as being the product of “a collective writing under the name of”, rather than attempting to assign authorship.
The problem with this position is that many texts are not merely essays but algorithms. They describe methods for building self-perpetuating human systems that require precision and completion if one is to replicate them. In an ideal world, each text would only contain what was necessary and sufficient to replicate that system. This is not, in fact, the case. Quite often later compilers added in material that they felt helped to keep the ur-text up-to-date, or added extra functions, or simply suited their own political preferences. Thus, for example, while the core text of the Shang Jun Shu cautions against distributing excessive rewards, the Encouragement of Immigration chapter, written nearly a century later, argues in favour of liberality, circumstances having changed in the meantime. The compilers, believing that the original author(s) - who argued forcefully and consistently for the importance of changing policies to suit the times - would have approved of this U-turn, thus included a text that directly contradicts the core of the book, and the work as a whole is greatly improved for their choice. Nevertheless, its presence in the text can be confusing for anyone unaware of the background to its inclusion.
The goal in conducting authorship analysis, then, is not to prove that certain chapters of an accepted work do not “belong” there - these books were known and accepted as compilations even at the time of composition, and if a chapter is in the accepted text then it belongs there ex officio. However, compilers’ decisions and decision-making processes are not always as clear as in the case of the Encouragement of Immigration chapter. This can be a source of confusion for those readers attempting to use the books as they were originally intended - for building political and economic systems. They can be left unsure whether a given chapter is intended to describe a component of one system, or an optional sub-system, or a different system altogether. In this context, authorship analysis can be an important tool for distinguishing the various different systemic wholes described in a given text. Chapters by a known group or individual are more likely to be intended to constitute a coherent set of instructions for building a given system. If we can cluster texts by author, this helps in their interpretation, and in doing so, we too - in our own way - are following the ancients. These essays may have circulated anonymously, but when Qin Shihuang was considering using them to inform his own political choices he first wished to know who wrote them:
In the absence of a Li Si to provide background information, this article is intended as a demonstration of the qualitative and quantitative methods by which this may be done. It takes as its guinea-pig a chapter of the Han Feizi that has been subject to much scholarly argument: the Jie Lao (“Commentaries on Lao Tzu’s Teachings”), which may be found in parallel text here. This essay will carry out a thorough quantitative and qualitative analysis of the chapter, less to convince other scholars (this being a lost cause) but rather to demonstrate how this may be done in the case of other works, providing a step-by-step approach to text analysis.
The Big Picture
Why, on the face of it, should this not be by the historical Han Fei?
Any initial judgment is necessarily vibes-based. It simply does not sound like the core chapters of the book that are more or less universally recognised as such. Firstly and most obviously is the content. Both the historical Han Fei and the collective that - undoubtedly - wrote under his name during his lifetime were openly and venomously anti-Confucian, referring to Confucians as vermin and traitors. By contrast, the Jie Lao chapter opens with a passage that reinterprets Daoist theory such as to render it compatible with Confucian ideals:
In the Dao De Jing, this shift from spontaneity to benevolence, righteousness and propriety is portrayed as a negative phenomenon - a shift away from primitive simplicity. In this text, however, the emergence of Confucian virtues is not seen as a symptom of social degeneration but rather as the means by which Daoist theory finds expression in the everyday world. It is an ingenious reinterpretation of a foundational text, and would become one of the cornerstones of the Huang-Lao syncretism (of legalist, Daoist and Confucian ideas) adopted by the early the Han dynasty as its semi-official ideology. However, it is also in explicit contradiction with the vision of Daoist thought held by the author(s) of the core texts (“inner chapters”) of the Han Feizi[1], who saw nature as an assemblage of indifferent Darwinian forces that came together to produce optimal outcomes while remaining entirely indifferent to the individuals upon whom they acted, and outright rejected the relevance of the Confucian virtues to this process.
The role attributed to the sage-king by the author of the Jie Lao is also at odds with that described by the author(s) of the inner chapters. The author of the Jie Lao takes a conventionally Confucian view of leadership, arguing that the personal morality adopted by the sovereign will be imitated by his subordinates and thus spread throughout the state[2].
This contrasts with the inner chapters, which argue that the throne should be a lieu vide - that the Emperor should refrain at all costs from using his office to express personal preferences. Instead, he should content himself with establishing and enforcing general rules, being a mere rubber stamp to decisions made within the mechanistic technocracy over which he rules. However, this did not mean that - as the author of the Jie Lao suggests - a politician should strive to repress his own desires, merely to keep a close and objective eye upon them. Multiple commentators have noted the crypto-Buddhist tone of the Jie Lao when addressing physical compulsions:
This contrasts sharply with the attitude taken by both Daoist and legalist texts during the Warring States era. Both schools advised against allowing one’s assessments to be coloured by partiality, but neither advised the kind of radical renunciation adopted by ascetics in later years. The Dao De Jing merely notes laconically that if one wishes to manipulate the world according to one’s desires it is advisable not to let them bias one’s observations of it, while the Han Feizi suggests that an intelligent ruler dealing with his concubines should “enjoy their personal attractions and not listen to their requests”. Indeed, this quiet utilitarian respect of the power of physical desire pervades the legalist texts - a population of monks, after all, would be difficult to incentivise with material rewards and sanctions. As Lundahl puts it:
However, while the author of the Jie Lao cautions his readers against the pitfalls of desire, he himself displays a preoccupation with physical security and longevity that is difficult to reconcile with the inner chapters - and, indeed, the known details of the historical Han Fei’s life[3]. As seen above, the author displays a fascination with the magic-medicine nexus that appears more characteristic of the Han dynasty - the period when this enduring national pastime really came into its own - than of the Warring States. During the years in which the inner chapters were composed, a cynical proto-scientific atheism prevailed among the literate classes - indeed, the superstitious hypochondria of the First Emperor was notable largely because they were unusual for their context. It was only after the establishment of political stability in the early second century BCE that such ideas would become the norm. By contrast, the historical Han Fei displayed a reckless chutzpah in his work and a willingness to put his life on the line that is at odds with the philosophy of the Jie Lao, for whose author tranquil longevity was the highest good:
All in all, the content of this chapter appears to be a better reflection of Han dynasty ideas than Warring States ones. The Huang-Lao interpretation of the Dao De Jing and the interest in mystical-medical interventions in particular are seldom seen in works produced pre-unification, but are widespread afterwards. However, there are other indicators in the text that argue both in favour of and against a Han dynasty dating for the text.
Style and Vocabulary
From intellectual trends it is a short jump to stylistic ones. The chapters generally believed to be by the historical Han Fei share an extremely distinctive style, and this is not present in the Jie Lao. While it is extremely difficult to describe literary style in a scientific manner, it is not impossible.
The sentences in the inner chapters are long and complex, using strings of thuses and therefores to trace out multi-stage dialectic arguments. A parallel in Western literature might be Alexis de Tocqueville, whose sentences tend to be long but clearly structured using logic operators. The original text was not punctuated, but when punctuation is added for modern readers it clearly reflects the run-on quality of the text. If we take a selection of 1000-character extracts from chapters likely to have been written by the historical Han Fei, they are remarkably consistent, with 33-34 full stops. By contrast, the Jie Lao has a massive 50.
A more comprehensive stylometric analysis can be done using text similarity algorithms such as, for example, the “dendrograms” code from this repository (not written by me - all credit goes to Joseph Lim and others at Lexikat). This can be used to cluster the various chapters of the existing Han Feizi corpus according to their mutual similarity. Having first prepared 1000-character excerpts from those of the inner chapters long enough to permit this[4], the code then strips out all characters not appearing in all extracts (an approximative but simple way to minimise the effect of topic upon the outcome) and calculates the vectors between them according to a Manhattan distance algorithm, using this to build up clusters of most similar texts as well as pairwise similarity scores:
Unsurprisingly, it is the Wang Zheng chapter - one that follows a completely different format to the rest of the book - that is the notable outlier. The fact that the code indicates as much is a reassuring sign that it is working. However, the results regarding the Jie Lao chapter are not particularly conclusive. The code has not assigned it into the large left-hand clade to which it has assigned the majority of the inner chapters, but neither is it entirely isolated - indeed, it has been placed in a sub-clade with the Difficulties in Persuasion, a chapter that has historically been believed to be the work of Han Fei himself.
However, it is possible to zoom in even further, considering the specific terms used by the author of the Jie Lao. In this case two three words/phrases are of particular interest: 端 (straight), 諸侯 (feudal lords) and 兼有天下 (uniting All-Under-Heaven).
One particular incidence of the character 端 in the chapter is particularly interesting:
The context in which 端 is used here implies that it is standing in for the character 正, something which would normally imply that the text was written in the years following the death of the First Emperor, when this character was subject to a naming taboo. (Qin Shihuang’s given name is usually rendered 政 today, but at the time 政 and 正 were used interchangeably.) Things are complicated, however, by the fact that 正 itself appears earlier in the text, footloose and un-tabooed:
Were the two sections written at different times, one before and one after the death of the First Emperor? Possibly the use of 端 was simply a coincidental stylistic choice rather than a ritual obligation? Maybe the author was simply a bit haphazard about the observation of naming taboos (as was occasionally the case in texts of the age)? Perhaps it was substituted in by a later copyist? It is impossible to say with certainty.
Further complicating the matter is the presence of the term 諸侯 (feudal lords) in a context that clearly implies them to be independent foreign powers (as was the case prior to unification and to a degree during the Chu-Han contention) rather than the ruler’s direct subordinates (as was the case during the early Han dynasty).
Those who favour an earlier date for the text argue that this means that it must have been composed during the Warring States period, since by the Han dynasty the threat of neighbouring lords had been more or less neutralised. However, it the way in which the terms are used here - preceded by 夫, a particle often used to introduce generalities - suggests that this passage is intended to be an ahistorical exposition of theory rather than advice to a particular ruler. The major part of the Dao De Jing was composed during a time at which conflict between feudal principalities was a defining feature of the age, and even after the establishment of the Han dynasty there was no guarantee that such a situation would not return. The use of 諸侯 here is therefore a curiosity but not a conclusive proof of an early date.
Something similar can be said about the use of 兼有天下 (uniting All-Under-Heaven). While its presence in the text could imply a pre-unification date for some or all of the chapter, the context in which it is deployed suggests it is being used to describe a general principle rather than a specific historical goal.
While unifying All-Under-Heaven was no longer a matter of concern for the Han dynasty leadership, it is not difficult to imagine that an author who had lived through the later years of the Warring States period and/or the Chu-Han contention would continue to treat it as a current affair worth theorising about.
Conclusion: Situating the Text
Currently there are almost as many opinions on this text as there are readers. Chen Qiyou argued that it was written by the historical Han Fei at a relatively early date (possibly the 260s BC), Bertil Lundahl comes down on the side of an anonymous author writing during the reign of Qin Shihuang, while Sarah Queen hedges her bets and suggests that it could have been written by Han Fei at the end of his career or someone else working at roughly the same time. Each explanation has data in favour and against.
My own preference would be to assume a Han dynasty date for all or most of the text. The Huang-Lao ideals, the fascination with mystical-medical matters and the use of 端 all point in this direction, while the counter-indicators provide no proofs of an earlier date that cannot convincingly be countered. However, I happily admit that the evidence is far from conclusive, and other readers may well come to the opposite conclusion based upon the same data.
Does this mean that we are no further forward, then? I would argue not. We now know precisely what we do and do not know, and why. Moreover, we now have a solid method at our disposal for tackling other similar problems in the future. More tenuously, but also more interestingly, we can take a guess at the significance of the chapter. If this was indeed written in the early Han dynasty, it can be seen not merely as a means of conveying information, but as a means of protecting the rest of the book. In a world that was growing increasingly hostile to texts that rejected the official Han Confucian-Daoist orthodoxy, the inner chapters of the Han Feizi - with their open contempt for Confucianism - were at risk. The Jie Lao chapter, however, is a clever demonstration of the ways in which the earlier chapters can be made to fit in with the mainstream ideology. Thus, in such a context, the composition and inclusion of the chapter is not merely a contribution to the philosophical debate, but a demonstration of astute political strategising.
[1] There is a certain amount of dispute regarding which chapters should be considered the core. It is possible to take a historical perspective (those chapters attributed to Han Fei by other authors) or a literary one (the first five books plus possibly a few other likely candidates such as the Five Vermin). For the purposes of this exercise I am using the first five books, just because this relies upon fewer subjective judgments.
[2] Lundahl goes into more detail on the role of Confucian ideas in this chapter. He notes that Chen Qiyou has suggested that the Jie Lao may be an early work, showing the influence of Han Fei’s time spent studying Confucianism under Xun Kuang, however, he also points out that the author of the Jie Lao is more Mencian than Xunzian in his views, and hence this seems an unlikely explanation. (It would also contradict the historical/stylistic elements that suggest a later date for the composition of the chapter.)
[3] Lundahl makes the same argument using the Wen Tian chapter, which purports to record a conversation between Han Fei and a follower in which he justifies his decision to risk his life in the defence of his ideas. However, this chapter is itself most likely a later confection added to the corpus after Han Fei’s death, so it is not necessarily to be relied upon.
[4] Based upon prior experience, length tends to have a disproportionate effect upon results when dealing with short classical Chinese texts, and the statistical mechanisms usually used to mitigate this in modern languages either have no effect or actively worsen the problem (due to the fat-tailed word frequency distribution). Consequently, it is easier and more effective to control rigorously for length prior to analysis than to correct for it afterwards. Similarly, while it is possible to take such excerpts from every chapter in the book long enough to allow for it, it output is so complicated as to be almost unreadable. Moreover, certain chapters - such as the Persuasions and Criticisms series - are likely to be compilations from multiple sources in any case, making it difficult (Criticisms) or impossible (Persuasions) to include them in any such analysis.