Around 300 BCE, a teacher from what is now Shandong called Mencius put a thought experiment to his audience. Suppose you see a small child about to crawl into a well. You feel a jolt of alarm and pity. You do not stop to think of a reward from the parents, or of your reputation, or of how much you dislike the noise of crying. So, he said, the feeling is not a product of calculation: it is a sprout, the beginning of goodness, and every human being has it.
A generation later, a philosopher called Xunzi replied that this proves much less than it seems. Feelings are not conduct. Left to themselves, people also quarrel, envy and take what they can; goodness is a skill acquired through teachers, rituals and years of practice, like a warped board straightened with steam and a press. Xunzi's two most famous students then went further: both of them became Legalists, who held that you should not rely on goodness at all.
That is a debate, with sides, evidence and counter-examples, not a collection of proverbs. India produced the same thing at the same scale: schools that argued about how we know anything, whether there is a self, and whether the world is real, in texts written for opponents to read and answer. This piece gives you the main positions on both sides, the reasoning behind them, and the places where each is weakest. It is a map, not a tour of everything.
Confucius (551–479 BCE) left no treatise. His sayings were collected by disciples in the Analects, over a long time. His starting point was practical: society is held together by relationships (parent and child, ruler and minister, friends), and a good person is shaped by how they fulfil those roles, with ren (humaneness, care for others) as the central virtue and li (ritual, courtesy, the forms of respect) as its discipline. The question he left open was the one the next two centuries fought over: where does the capacity for goodness come from?
Mencius (c. 372–289 BCE) argued from the well:
He added a warning: sprouts can be trampled or starved, and they can be ruined by too much help (his farmer who tugged at his seedlings to make them grow, and killed them). So his claim was that human nature is good in its tendency, not that all people are good.
Xunzi (c. 310–c. 235 BCE) used different evidence:
Note the difference in a word. When Mencius says "nature" he means what a thing grows into when it flourishes; when Xunzi says "nature" he means what is there at birth. A good deal of their disagreement is real, though: Xunzi's view says the ruler's job is to build a moral order; Mencius's says it is to protect one that people already want. Each has an obvious weak point. Mencius must explain why so many people go wrong. Xunzi must explain where the human ability to see that ritual is worth following comes from if we start out merely selfish. His answer is that humans have intelligence and the capacity to learn, which Mencius might fairly say is itself a sprout.
Pick a case, and see how the schools in the next section answer it. "Text" means the school's own writings address the case; "reconstruction" means I am working out what its principles imply.
Without JavaScript: on the sheep case, the Analects (13.18) has Confucius say that uprightness lies in fathers and sons covering for each other; the Han Feizi uses the story to argue that private loyalty and public law conflict and law must win. On war, Mozi condemns offensive war as crime on a large scale; Mencius allows only a ruler with the people's support to "punish" a tyrant.
Mozi (c. 470–c. 391 BCE) thought Confucian partiality, caring most for kin and less for strangers, was the root of war and waste. His argument was about consequences. The test of any teaching, he said, is whether it benefits the people: what the ancient models did, what people can observe, and whether it works when applied. Wars of conquest fail this test. Stealing a peach is a crime, killing a man a worse one, and conquering a state is the same crime on a larger scale, praised only because the scale is hard to see (Mozi ch. 17). His remedy was jian ai, usually translated "impartial care" or "concern for all": treat strangers' households as you would your own. Mencius thought this was a rejection of the family itself ("to love without distinction is to have no father", Mencius 3B9). Mohist followers also developed canons on logic, argument and defensive warfare, which makes them the closest ancient China came to a school of logic.
The Legalists started from the opposite pole. Han Fei (c. 280–233 BCE), a student of Xunzi, argued that a state cannot be built on the hope that rulers and officials will be good, since they mostly won't be. It can be built on a design: clear public law (fa), techniques for controlling officials (shu), and the ruler's authority (shi), run by two "handles", reward and punishment. The strongest part of the argument is practical: it asks for institutions that work with ordinary motives. The weak part is that a ruler can use the same machine on his own subjects, and Han Fei himself died in prison in Qin in 233 BCE, a victim of the court intrigue he had analysed. The Qin dynasty that adopted Legalist methods unified China in 221 BCE and fell in 206 BCE. The Han dynasty that followed kept much of its administrative machinery while making Confucian learning the official curriculum: a long-running scholarly description is "Confucian outside, Legalist inside".
Zhuangzi (c. 369–286 BCE) rejected the argument everyone else was having. Rules, rituals and calculations all assume that we can step outside the flow of things and judge. His best-known piece of reasoning is a story. Walking with the logician Huizi on a bridge over the Hao river, he remarked on how happy the fish were. Huizi objected: you are not a fish, so how do you know? Zhuangzi: you are not me, so how do you know I don't know? Huizi: I am not you, so I do not know what you know, and you are not a fish, so you cannot know the fish's joy. Zhuangzi: go back to your first question. You asked how I knew, which assumed I did know. I know it from here, standing on the bridge. It is a joke, and also an argument about what "how do you know" is asking.
My paraphrase of Zhuangzi ch. 17 ("Autumn Floods"). The joke is that "how" can mean "by what method?" or "from where?"; translators and scholars still argue about which move wins.
Without JavaScript: Zhuangzi says the fish are happy. Huizi: you are not a fish. Zhuangzi: you are not me. Huizi: so neither of us can know the other's mind. Zhuangzi: you asked "how do you know?", so you already took it that I knew; I know it standing on the bridge.
The Daodejing, attributed to Laozi (whose historicity is unclear; the oldest excavated manuscripts, from about 300 BCE, were found at Guodian, in a different order from the received text), makes a related move in a calmer voice. Its idea of wu wei, "non-forcing", is not laziness: it is the claim that the best action responds to the situation without strain, like a skilled butcher whose knife never touches bone. The Daoist argument against Confucian virtue is that explicit moral codes only appear once natural care has been lost (Daodejing 18).
When Buddhism entered China (around the first century CE) it brought a metaphysics Confucian thinkers had not worked out for themselves: what the world is made of and what a self is. From about the eleventh century, Confucian philosophers answered with a metaphysics of their own, now called Neo-Confucianism. Zhu Xi (1130–1200) gave it its classic form. Everything has two aspects: li, a pattern or principle (what makes a thing the kind of thing it is, and what makes it good), and qi, the stuff the pattern is realised in. A cracked jar and a perfect one have the same li but different qi. To become wise you should "investigate things" (gewu): study texts, history and the world until the principles in them become clear to you. Zhu Xi selected and commented on the Four Books (the Analects, the Mencius, the Great Learning and the Doctrine of the Mean). From 1313 they were the core of the civil-service examinations, and they stayed there until the examinations were abolished in 1905.
Wang Yangming (1472–1529) found the programme did not work, and his story is a good one. As a young man he and a friend decided to investigate the principle in the bamboo in front of the pavilion. The friend gave up on the third day, exhausted. Wang kept at it for seven, fell ill, and concluded that he lacked the ability to become a sage. Years later, exiled to the remote Longchang, he had an insight that became his philosophy: principle is not out there to be found; the mind itself is principle (xin ji li). We have an innate knowing of right and wrong (liangzhi, "good knowing"), and the work is to stop obscuring it. From that came his most quoted claim, the unity of knowledge and action: if you truly know that you should help someone, you are already helping; knowing without acting is not yet knowing. The standard objection is that this makes it hard to say anyone ever knowingly does wrong. Wang's answer is that selfish desire clouds the knowing itself.
These two are still taught as a pair: Zhu Xi's cautious, reading-based inquiry against Wang's immediate moral intuition. Neo-Confucianism shaped Korea's Yi Hwang (Toegye, 1501–70) and Japan's Tokugawa-era schools as well.
Indian philosophy has an unusual feature: schools were defined by how they answered the question "what counts as a source of knowledge?" (pramāṇa). The six "orthodox" (āstika) schools are those that accept the authority of the Veda, in some sense; the Buddhist, Jain and Cārvāka schools do not, and are called nāstika. The six are not a list of religions but of research programmes:
| School | Best known for |
|---|---|
| Nyāya | Logic and epistemology: four sources of knowledge, and a theory of inference |
| Vaiśeṣika | Categories of what exists; atomism |
| Sāṃkhya | Two ultimate realities: puruṣa (awareness) and prakṛti (matter and its three strands) |
| Yoga | Patañjali's practical psychology, on top of Sāṃkhya's metaphysics |
| Pūrva Mīmāṃsā | How to interpret the Vedic texts; a theory of language and of duty (dharma) |
| Vedānta | The nature of brahman and the self: non-dualist (Advaita, Śaṅkara, c. 700 CE), qualified non-dualist (Rāmānuja, 11th–12th c.), dualist (Madhva, 13th c.) |
Nyāya says there are four sources of knowledge: perception, inference, comparison (analogy) and testimony. Its model of inference is the five-part argument: (1) the hill has fire; (2) because it has smoke; (3) wherever there is smoke there is fire, as in a kitchen; (4) this hill has smoke of that kind; (5) therefore it has fire. The crucial element is the third: the vyāpti, the "pervasion" that links the two. This is where the fight is. How could you know that wherever there is smoke there is fire? You have not seen every case.
That is the exact question the Cārvāka or Lokāyata school raised. Their texts are lost, and we know them mainly through opponents' summaries, an important caution. The position reported is that perception is the only reliable source: inference gives only probable guesses, because the universal link cannot be checked in all cases, past and future. A later sceptical text, Jayarāśi's Tattvopaplavasiṃha (around the eighth century), pushes the doubt against every source of knowledge. The Nyāya reply is that the Cārvāka also relies on inference when arguing that inference fails, and that we do establish universal links by repeated co-occurrence with attention to the absence of exceptions. It did not settle the dispute, and the problem of induction in the West is its cousin.
The Buddhist logician Dignāga (c. 480–540 CE) turned this into a test that you can run. A reason (hetu) supports a conclusion only if it meets three conditions: (1) it is present in the subject (the hill); (2) it is present in at least some sapakṣa, cases which have the property to be proved (fire); (3) it is absent from all vipakṣa, cases which lack it. Crossing "present in all / some / none of the similar cases" with "present in all / some / none of the dissimilar cases" gives his nine-cell "wheel of reasons". Only two cells pass. Dharmakīrti (c. 600–660) refined the system and defined valid knowledge as "non-deceiving", that is, as guiding action without failing.
Choose a reason, or build one. Then check which of the nine cells it lands in.
Assumes the reason is present in the subject. Presets follow the standard textbook examples, which assume the conclusion's opposite-class is as stated; Indian logicians argued about the examples as well as the test.
Without JavaScript: a reason is good only if it appears in some similar cases and in no dissimilar ones; smoke for fire passes. "Sound is eternal because it is produced" fails because produced things are all non-eternal.
Nāgārjuna (c. 150–250 CE), the founder of the Madhyamaka ("Middle Way") school, wrote the Mūlamadhyamakakārikā, "Fundamental Verses on the Middle Way". Its target is the idea that things have svabhāva, an intrinsic nature that makes them what they are independent of anything else. His own position is that nothing has one. This is called emptiness (śūnyatā), and it is easily misheard: it does not mean that nothing exists. Things exist, as dependent on causes, parts and the way we name them. He says emptiness is itself dependent arising, and that it is "the middle way" (24.18).
The method is a form of argument by elimination. In the first chapter he asks how a thing could arise, and lists the only options:
The four options are the catuṣkoṭi, or tetralemma. Scholars disagree about how to read it. Some say Nāgārjuna is only showing that a particular view is incoherent while holding no view himself; others, such as Jay Garfield and Graham Priest, argue that he accepts some contradictions as true. The classical objection is: "if everything is empty, so is your argument, so why should I accept it?" Nāgārjuna's reply (Vigrahavyāvartanī) is that an argument does not need intrinsic nature to work; his own claim is empty too, and that is consistent.
Take a sprout and a seed. Press each option to see the objection, and watch the options run out.
Compressed from the first chapter of the Mūlamadhyamakakārikā and its commentaries. This is a reading; the verse itself is terse.
Without JavaScript: from itself, then arising is pointless; from another, then anything from anything; from both, both faults; from neither, anything at random. So nothing arises with a nature of its own.
The Jains, whose tradition names Mahāvīra (a contemporary of the Buddha, with dates debated, about the sixth to fifth century BCE) as its last great teacher, drew an epistemological consequence from the thought that reality is complex. Anekāntavāda, "non-one-sidedness", holds that any real thing has endless aspects and that a single statement captures only one. The story of blind men touching an elephant (told in Buddhist, Hindu and Jain sources) is the usual illustration. From it they derived syādvāda, the doctrine of "in some respect": every claim should be made relative to a standpoint. The classical form has seven statements. For a pot: in some respect it is; in some respect it is not; in some respect it both is and is not; in some respect it is inexpressible; and then the three combinations of "inexpressible" with the first three.
This is not "everything is true". The "respects" are specific: the pot exists as this clay, here, now, in this shape, and does not exist as cloth, elsewhere, at another time, in another shape. Supporters (the philosopher B. K. Matilal among them) read it as a system for relativising predicates and avoiding contradictions. Critics, from Śaṅkara to Buddhist logicians, said it fails to say anything definite. It is a good test of your own view of what "contradiction" means.
The seven statements (saptabhaṅgī) as usually presented in textbooks; the standpoints are the Jain list of own and other substance, place, time and form.
Without JavaScript: the pot is, in respect of its own clay, place, time and form; it is not, in respect of anything else; both in turn; and "inexpressible" when both are meant at once, since one word cannot say both.
China and India each built traditions of argument. Mencius and Xunzi disagreed about human nature, Mozi and the Legalists attacked partial loyalty and weak institutions from different sides, Zhu Xi and Wang Yangming split on whether principle is in the world or in the mind, Nyāya and the Cārvāka fought over the universal link in inference, Dignāga turned a reason into a test, Nāgārjuna argued that nothing arises with a nature of its own, and the Jains made relativity of standpoint into a logic. Read them as you would Plato and Hume: state the argument, then look for the weakest premise.
Try this: pick one belief you hold firmly and write it as a five-part Nyāya argument (claim, reason, example, application, conclusion). Then ask whether your "wherever…" link is really true of every case.
Sources. Analects 13.18; Mencius 2A2, 2A6, 3B9, 7B2; Xunzi ch. 23 ("Human nature is bad"); Mozi chs. 17–19; Han Feizi; Zhuangzi chs. 3 and 17; Daodejing chs. 18 and 31 (Guodian manuscripts, c. 300 BCE); Zhu Xi and Wang Yangming via Chan, A Source Book in Chinese Philosophy (1963) and the Stanford Encyclopedia entries; Four Books in the examinations from 1313 to 1905. Nyāya Sūtras, Dignāga's Pramāṇasamuccaya, Dharmakīrti's Pramāṇavārttika, Jayarāśi's Tattvopaplavasiṃha; Nāgārjuna, Mūlamadhyamakakārikā chs. 1 and 24, trans. Garfield (1995); Matilal on Jain logic. Dates of ancient figures are approximate and disputed.