How Do We Know What the Buddha Taught? Texts, Scrolls, Stones

The Buddha wrote nothing. Neither did anyone who heard him. For at least two centuries after his death his teaching existed only in the memories of monks and nuns who recited it to one another, and the first time a Buddhist text can be seen as a physical object is on birch bark in Gandhāra, around the turn of the era.^[1]​^[2] Every later page of this series makes claims about what came before that. This one explains how those claims are made and where they stop.

The short answer: early Buddhism is reconstructed by comparing parallel versions of the same texts preserved by different schools in different languages, and by tying them to a small set of firm dates. Ashoka’s inscriptions (c. 250 BCE) are the first dated Buddhist evidence. Gandhāran manuscripts (first century BCE onward) are the first physical texts. Chinese translations carry their own dates (the earliest Mahāyāna one is from 179 CE).^[3] Archaeology at Lumbini and other sites adds what the texts leave out. No single method is decisive. Where they agree, this series says so; where they do not, it says that too.

The problem: a teaching kept in memory

The canon describes itself as recited, not written, and so the first fact about every early Buddhist text is that it is the end product of a long oral transmission.

The tradition’s own account is the First Council at Rājagaha, held soon after the Buddha’s death, where Ānanda recited the discourses and Upāli the monastic rules. Most historians treat that council as a founding story rather than an event, but the picture it gives of the process is accurate: the texts were held by groups of specialist reciters (bhāṇaka), each responsible for one collection, and kept stable by communal recitation (§3).^[4]​^[5]​^[1]

Oral transmission leaves marks that are still visible in the texts:

• Repetition. Stock passages recur word for word across hundreds of discourses, and lists are everywhere: the four truths, the five aggregates, the eight path factors, the twelve links. These are memory aids, not padding.^[6]
• Formulaic frames. Nearly every discourse opens evaṃ me sutaṃ, “thus have I heard,” and places the Buddha at a named town or park. The frame gives the reciter a location to hang the text on.
• Variation in the details. The same teaching can appear with different settings, different audiences or a different number of items in a list. That is what independent oral lines produce.^[6]​^[7]

Writing came late. Indian writing is not securely attested before Ashoka (c. 250 BCE), apart from contested finds at Anuradhapura in Sri Lanka, and there is no sign that monks wrote down texts for some time after it became available.^[8]​^[9]​^[10] Consensus The question is not whether the canon was transmitted orally, but how well.

Three canons, many languages

No single “original” Buddhist canon survives. What survives is several partial canons of different early schools, preserved in different languages, which can be compared.

The Pali canon is the best known in the West, partly for an accident of history: it was the first complete early canon European scholars could read in an Indian language, and the Pali Text Society was founded in 1881 to publish it.^[12] That gave it an authority it never claimed for itself. It is one school’s canon, closed and written down in Sri Lanka, with its own later additions.^[11]

The Chinese Āgamas are the essential counterweight. The four collections correspond to the four main Pali Nikāyas: the Dīrgha Āgama to the Dīgha Nikāya, the Madhyama to the Majjhima, the Saṃyukta to the Saṃyutta and the Ekottarika to the Aṅguttara. They were translated from the canons of other schools, mostly the Dharmaguptakas and Sarvāstivādins, which split from the ancestors of the Theravāda in the third and second centuries BCE.^[7] Consensus

The comparative method: what all versions share

Material that appears in the Pali Nikāyas, the Chinese Āgamas and the Gāndhārī or Sanskrit fragments is probably older than the split between the schools that preserved them. Material found in only one version is probably later.^[7]​^[13]

This is the most important tool in the field, and it works the same way as comparing manuscripts of any ancient text. If two schools that separated in the third century BCE both preserve the same discourse with the same core argument, the simplest explanation is that the discourse existed before they separated.

The classic demonstration is the Mahāparinibbāna Sutta, the account of the Buddha’s last months. Ernst Waldschmidt printed its Sanskrit and Tibetan versions in parallel with the Pali, together with a German translation of the Chinese, in 1950–51. They agree on the outline: the journey north, the last meal, the death at Kusinārā, the division of the relics. They disagree on episodes and speeches, which accumulate in some versions and not others.^[14]​^[15] Bhikkhu Anālayo has since done the same for every discourse in the Majjhima Nikāya and its parallels.^[7]

The results point in a consistent direction:

• The doctrinal core is shared. The four noble truths, dependent origination, the five aggregates, not-self, the eightfold path and the four meditative absorptions (jhāna) appear across the schools in recognisably the same form.^[13]​^[7] Consensus
• The Vinaya rules are mostly shared. The schools’ lists of monastic rules (pātimokkha) agree on the serious offences and differ mainly in the minor rules of etiquette, which is why schools could split over discipline without splitting over doctrine (§3).^[16] Consensus
• Narrative grows. Biography, miracle and cosmology expand from version to version. The life of the Buddha is the clearest case (§2).^[14]​^[5] Consensus
• Abhidharma does not match. Each school’s systematic philosophy is its own, which is why the Abhidharma is dated after the splits.^[17] Consensus

The limit. Agreement shows a text is pre-sectarian, which means roughly before the third century BCE. It does not show the Buddha said it. The shared core could have been shaped during the century or more between his death and the first schism. How much of the core goes back to the Buddha himself is the real argument, and it is taken up below.^[13]​^[18] Majority

The language problem: Pali was not the Buddha’s dialect

The Buddha taught in the eastern Ganga plain, probably in a local Middle Indo-Aryan dialect. Pali is a literary language with a more western character, and the canon shows signs of having been translated into it.^[11]​^[1]

Two kinds of evidence point the same way:

• Eastern forms left behind. Scattered through the Pali canon are forms typical of eastern dialects, such as nominative singulars in -e where Pali normally has -o. Linguists since Heinrich Lüders have read these as fossils of an older eastern version that was translated into Pali but not completely.^[1]​^[11] Majority
• The tradition’s own name. Pāli originally meant “text,” the canonical line as opposed to the commentary. Theravāda tradition later called the language Māgadhī, the speech of Magadha, and identified it with the Buddha’s own. Linguists do not accept the identification.^[1]​^[11] Consensus

One Vinaya passage is often quoted on the question. Two monks of brahmin family complain that monks of different backgrounds are corrupting the Buddha’s words with their own dialects, and propose putting them into Vedic-style metre (chandaso). The Buddha refuses:

anujānāmi bhikkhave sakāya niruttiyā buddhavacanaṃ pariyāpuṇitun ti.

“I allow you, O Bhikkhus, to learn the word of the Buddhas each in his own dialect.” (trans. T. W. Rhys Davids and Hermann Oldenberg, 1885)^[19]​^[20]

What it proves. That early Buddhists thought about language and refused to sacralise one. The exact meaning of sakāya niruttiyā is disputed: Buddhaghosa read it as “in the Buddha’s own language,” which he took to be Māgadhī, while most modern scholars read it as “in each person’s own dialect.” Some recent translators follow Buddhaghosa: Bhikkhu Brahmali renders the ruling “You should learn the word of the Buddha using its own expressions.”^[19] Either way the contrast with Vedic religion is sharp. The Veda was preserved syllable by syllable in one sacred language; Buddhism allowed translation from the start, and that is part of why it could later move across Asia.^[1] Majority

Manuscripts: the oldest physical Buddhist texts

The oldest Buddhist manuscripts anywhere are birch-bark scrolls from Gandhāra, in today’s northwest Pakistan and eastern Afghanistan, written in the Gāndhārī language and Kharoṣṭhī script between about the first century BCE and the second century CE.^[2]

Until 1994 the earliest Buddhist manuscripts known were Sanskrit fragments from Central Asia, several centuries younger. That year the British Library acquired twenty-nine birch-bark fragments that had been buried inside an inscribed clay pot, probably in eastern Afghanistan. Richard Salomon dated them to about the first century CE on grounds of script and language.^[21] More collections followed: the Senior scrolls, the Bajaur collection (found in 1999) and the Split collection. Together they comprise dozens of scrolls, the oldest Indian manuscripts of any kind then known.^[2] Consensus

The scrolls show three things that the later canons could not:

• Early texts existed in more than one language. One British Library fragment is a Gāndhārī version of the Rhinoceros Sūtra, the poem known in Pali as the Khaggavisāṇa Sutta (Sutta Nipāta 1.3). Its verses correspond closely to the Pali but are not a translation of it.^[22]​^[21] The Pali canon was one version among several. Consensus
• Schools had their own collections. The pot that held the British Library scrolls is inscribed as a gift to the Dharmaguptakas, one of the early schools whose Vinaya and Dīrgha Āgama survive in Chinese.^[21] Majority
• Mahāyāna texts are almost as old as the earliest physical texts. A Split collection manuscript of the Perfection of Wisdom in 8,000 Lines (Aṣṭasāhasrikā Prajñāpāramitā) has been radiocarbon dated to roughly 47–147 CE.^[23]​^[24] Physically, Mahāyāna sūtras and early discourses appear in the record at nearly the same time, though most scholars still hold that the discourses are older (§6). Consensus on the date

The Gandhāran finds have a provenance problem that should be stated plainly. Almost none came from a controlled excavation. They reached the antiquities market from a border region at war, and their exact findspots are reported, not known. That affects what can be said about where they were used, not whether they are genuine or how old they are.^[2]

The Pali manuscripts are much younger. The tropical climate of Sri Lanka and Southeast Asia destroys palm leaf, so almost every Pali manuscript is from the eighteenth or nineteenth century.^[11] The oldest known is four Vinaya folios in the National Archives in Kathmandu, copied in the eighth or ninth century.^[25]​^[26] Older than any of them are twenty gold leaves inscribed with Pali passages, buried in a reliquary at the Pyu city of Śrī Kṣetra in Myanmar around the fifth or sixth century.^[27] Consensus

When the Pali canon was written down

Sri Lanka’s chronicles say the Pali canon and its commentary were written down in the reign of Vaṭṭagāmaṇi Abhaya, in the first century BCE, after a period of famine and war.^[28]​^[29] The chronicles do not name the place; the Sinhala Pūjāvaliya (thirteenth century) and Nikāya Saṅgrahaya (fourteenth) put it at the Aluvihāra monastery.^[1]

The oldest chronicle, the Dīpavaṃsa, gives the event in two verses:

piṭakattayapāḷiñ ca tassā aṭṭhakatham pi ca mukhapāṭhena ānesuṃ pubbe bhikkhū mahāmatī || hāniṃ disvāna sattānaṃ tadā bhikkhū samāgatā ciraṭṭhitatthaṃ dhammassa potthakesu likhāpayuṃ ||

“Before this time, the wise Bhikkhus had orally handed down the text of the three Piṭakas and also the Aṭṭhakathā. At this time, the Bhikkhus who perceived the decay of created beings, assembled and in order that the Religion might endure for a long time, they recorded (the above-mentioned texts) in written books.” (trans. Hermann Oldenberg, 1879)^[30]​^[31]

What it proves. That the Mahāvihāra tradition itself remembered a long oral period followed by a deliberate decision to write, prompted by fear of loss. The date is often given as 29 BCE, which comes from older reckonings of the Sri Lankan king list. Vaṭṭagāmaṇi’s reign is now usually placed in the first century BCE without that precision.^[29]​^[1] Majority

What was written then was not exactly the canon now printed. The commentaries compiled by Buddhaghosa in the fifth century CE show that the canon was still being arranged, and a few texts (the Abhidhamma’s Kathāvatthu, parts of the Khuddaka Nikāya) were added or expanded after the schools separated.^[11] Consensus

Inscriptions: Ashoka and the first hard dates

Ashoka’s inscriptions, carved around 250 BCE, are the first dated evidence for Buddhism anywhere, and the only evidence from within roughly 150 to 200 years of the Buddha’s death.^[32]​^[33]

The absolute date comes from Rock Edict 13, which names five Hellenistic kings whose reigns are known from Greek sources.^[34]​^[35] Four Ashokan inscriptions bear directly on Buddhism:

• Rummindei pillar (Lumbini). In his twentieth regnal year Ashoka came in person and worshipped, the inscription says, because “here the Buddha was born, the Sakyamuni” (hida budhe jāte sakyamunī ti, rendered literally here), and freed the village of the bali tax, leaving it liable to only an eighth share.^[36] The earliest evidence that a specific place was identified as the birthplace (§2).
• Nigali Sagar pillar. In his fourteenth year Ashoka doubled the size of the stupa of a past Buddha, Konākamana.^[37] By c. 250 BCE Buddhists already believed in a line of earlier Buddhas and built monuments to them.
• Calcutta-Bairāṭ edict. Ashoka addresses the saṅgha and names seven texts he wants monks, nuns and lay followers to hear and reflect on often.^[38]
• The schism edicts at Sarnath, Sanchi and Kauśāmbī order that monks or nuns who split the saṅgha be made to wear white and live apart.^[39] Divisions were already a problem (§3).

The Bairāṭ list deserves attention, because it is the only place where a dated source names Buddhist texts:

What it proves. Named texts existed in Ashoka’s time, and several can still be identified in the Pali canon.^[38]​^[40]​^[41]​^[5] The canon’s core is not a late invention. Majority

What the edicts leave out. Ashoka never mentions nirvana, not-self, the four noble truths or meditation. He speaks of moral conduct, generosity and reaching heaven (svaga).^[42] The silence is not evidence that the doctrines did not exist. The edicts are public addresses to an empire, mostly to lay people, and they show what one powerful lay Buddhist thought ordinary people needed. That is itself a finding: from the start, Buddhism for householders was largely about merit and good rebirth, not liberation.^[42]​^[33] Majority

Archaeology: what the texts leave out

Archaeology shows Buddhism as it was practised, which was often not as the texts prescribe. Gregory Schopen made the argument that changed the field.

For most of the twentieth century scholars reconstructed early Buddhism from canonical texts and treated inscriptions and monuments as secondary. In a 1991 essay, Schopen argued that this preference was inherited from Protestant assumptions that the true religion lies in scripture. When inscriptions are read first, a different picture appears: monks and nuns who owned property, gave large donations in their own names and made merit for their dead parents, practices the texts either ignore or seem to forbid.^[43] Majority

The donor inscriptions at Sanchi and Bharhut, from the second and first centuries BCE, bear this out. A large share of the donors are monks and nuns, many of them identified as reciters of the texts or as “knowing the five collections” (pacanekāyika).^[43]​^[44]​^[45] The texts describe renunciants who own nothing; the stones record renunciants who paid for railings and gateways. Consensus on the inscriptions; Majority on the interpretation

Excavation also tests the stories directly. At Lumbini, Robin Coningham’s team dug beneath the Maya Devi temple in 2011–13 and found a sequence of brick structures with an earlier timber one below, which they dated to the sixth century BCE.^[46] What that does and does not show about the Buddha’s dates is taken up in §2.

Dated translations: the Chinese clock

Chinese translations often carry dates and translators’ names, which makes them a clock for Indian texts: if a text was translated in a known year, its Indian original existed by then.

The first securely attributed translators worked in Luoyang in the second century CE. An Shigao, a Parthian who arrived around 148, translated meditation and Abhidharma texts.^[47]​^[48] Lokakṣema, from the Kushan lands, translated Mahāyāna sūtras; his version of the Perfection of Wisdom in 8,000 Lines is dated 179 CE.^[3]​^[49] Consensus

The method has a weakness. Chinese catalogues later attached famous names to anonymous translations, so an attribution is only as good as its catalogue. Jan Nattier’s guide sorts the genuine Han-dynasty translations from the misattributed ones, and is the basis for most dating of early Mahāyāna.^[49] Beside the attribution problem is a translation problem: a Chinese version shows what an Indian text looked like in a given century, but filtered through a translator’s vocabulary and choices (§7).

Did the Buddha exist? The historicity question, stated fairly

Most scholars hold that the Buddha was a historical person whose teaching can be partly recovered. A minority argue that nothing reliable about him can be known.^[50]​^[18]

Majority, with a real dissent.

The sceptical case. David Drewes argues that two centuries of scholarship assumed a historical Buddha without proving one. The earliest texts were fixed long after the events, they are full of miracle and myth, and the method of peeling away the legend to leave a core assumes there is a core. On his reading, we know nothing about the historical Buddha, including whether he existed.^[18]

The majority case. Alexander Wynne and Oskar von Hinüber answer that the early texts contain exactly the kind of detail invented scripture does not: the names of the Buddha’s two meditation teachers, Āḷāra Kālāma and Uddaka Rāmaputta; minor characters and places that served no later interest; awkward episodes, such as the first person he met after his awakening walking away unconvinced; and political and legal arrangements that fit the fifth century BCE and not later ones.^[50]​^[51]​^[52] Add Ashoka at Lumbini 150 years or so later, honouring a named person at a named village, and the simplest explanation is a real teacher remembered by a real community.

Where this series lands. With the majority. The Buddha is treated as a historical person who lived in the fifth or early fourth century BCE,^[32]​^[53] whose core teaching is substantially preserved in the material common to the early schools. The biography is treated as a separate question, and most of it as later (§2). Where a claim depends on knowing what the Buddha himself said rather than what the early community taught, it is tagged accordingly.

What “Buddhism” means in this series

The word “Buddhism” is a European coinage of the early nineteenth century. Buddhists called their tradition the dhamma (Sanskrit dharma), “the teaching,” or the Buddha-sāsana, “the Buddha’s instruction.”^[12]

“Buddhism” as the name of a single world religion with one founder and many national forms was assembled in the nineteenth century, when European scholars realised that the religions of Sri Lanka, Tibet, China and Japan all traced themselves to the same Indian teacher. Before then those traditions were often unaware of one another, and Western observers sometimes took them for separate religions (§12).^[12]​^[54] Consensus

That does not make the category false. The traditions really do share a founder figure, a body of texts and a monastic institution, and they knew themselves as heirs of the Buddha. But it means a history of Buddhism is a history of many lineages that only later came to be seen as one thing, much as the Hinduism series found for “Hinduism” (Hinduism §0).

Reading conventions used throughout

• Dates are BCE/CE, given as ranges where they are uncertain, with a confidence tag.
• Pali and Sanskrit appear in italic transliteration, then translation, then what the passage is being used to prove. Chinese and Tibetan sources are cited by their standard catalogue numbers.
• Spelling. Names are given in their Pali form for the early period (nibbāna, dhamma, Gotama) and in Sanskrit for Mahāyāna and later India (nirvāṇa, dharma, Gautama), with the common English form alongside the first time.
• Citations use standard abbreviations: DN (Dīgha Nikāya), MN (Majjhima Nikāya), SN (Saṃyutta Nikāya), AN (Aṅguttara Nikāya), Sn (Sutta Nipāta), Dhp (Dhammapada), Vin (Vinaya), T (the Taishō edition of the Chinese canon). MN 26 = Majjhima Nikāya, discourse 26.
• Confidence tags: Consensus Majority Contested Speculative Political

Further reading

• Richard Salomon, The Buddhist Literature of Ancient Gandhāra (Wisdom, 2018) — every Gāndhārī collection, with translations.
• Oskar von Hinüber, A Handbook of Pāli Literature (De Gruyter, 1996).
• K. R. Norman, Pāli Literature (Harrassowitz, 1983).
• Bhikkhu Anālayo, A Comparative Study of the Majjhima-nikāya (2 vols., 2011).
• Gregory Schopen, “Archaeology and Protestant Presuppositions in the Study of Indian Buddhism,” History of Religions 31 (1991).
• Jan Nattier, A Guide to the Earliest Chinese Buddhist Translations (2008).
• David Drewes, “The Idea of the Historical Buddha,” JIABS 40 (2017), and Alexander Wynne, “Did the Buddha Exist?”, JOCBS 16 (2019) — read them as a pair.