Sanskrit, Dravidian, and Munda

Modern distribution of
Indian languages

In this essay I will reiterate some important points made by Michael Witzel about the linguistic history of India. When the first anatomically modern humans reached India ca. 70,000 years ago they almost certainly used language. But all the direct evidence for language is much more recent, the oldest being written forms of language. Comparative linguistics allows us infer a great deal more about the history of language so that we can get a picture of how people spoke long before writing was even invented.

Like many historians I use the term India, or sometimes Greater India, to mean then whole of the sub-continent, taking in the political territories of modern day Pakistan, India, Nepal, Sri Lanka, and Bangladesh. Given that the main languages of North India and Sri Lanka are all modern Indic Languages: Urdu, Panjabi, Hindi, Bihari, Bengali, Nepali, and Sinhala, the modern political divisions belie the common linguistic history they share. However we must be a little cautious. Language, ethnicity, and geography can be independent variables when discussing culture. This essay mainly concerns languages and the speakers of languages. We cannot be sure of the ethnicity of these people.

We know with some certainty that the speakers of Old Indic languages (now represented only by Vedic) came from outside India. This is an unpopular thesis amongst Indian Nationalists who try to make a case for Sanskrit arising in India and spreading out. Some would have us believe it is the original language (Cf Eco 1997). However the relationship of Old Indic with Old Iranian and a variety of other internal evidence show that Indo-Iranian -- an early offshoot from Proto-Indo-European that further split into two sub-families, Iranian and Indic -- was spoken by nomadic peoples of the Southern Central Asia. Old Indic is mostly distinguished from Old Iranian by a few sound changes. Later grammatical forms drifted apart as well, though the attested languages, Vedic and Avestan, were closely related. 

Comparative linguists showed in the late 18th century that Greek, Latin and Sanskrit are all so similar that they must have derived from a common ancestor. That hypothetical languages is nowadays called Proto-Indo-European (PIE) and the language family that it spawned is called Indo-European (IE). PIE also has a Germanic branch giving rise to all the Germanic languages (including English), a Slavonic branch incorporating all the Slavic languages, and takes in many of the languages of Iran and Afghanistan, not to mention Armenian. In addition we have written evidence of a number of now dead Indo-European languages such as Tocharian and Khotanese from Central Asia. By comparing the changes in many languages, linguists are able to describe pragmatic 'rules' which describe how sounds and forms of words change. This procedure has been very successful in some areas. PIE is probably the best example. But the Sino-Tibetan language family also gives a clear view of the proto language that underlies them all. 

There have been efforts of varying success to try to cover all the languages of the world in this way. And this has naturally lead some scholars to propose a further ancient layer of relatedness. So for example there is the conjectured Nostratic proto-language (or macro-family) that takes in Afroasiatic (including the Semitic languages), Kartvelian (Caucasian languages and possibly Basque), Indo-European, Uralic (including Finno-Ugric), Dravidian, Altaic (covering the Turkish, Central Asian, and probably Korean and Japanese), and Eskimo–Aleut. These macro-families are still controversial, though many of the objections are ideological rather than logical.

A major branch of the PIE family is Indo-Iranian taking in languages that were spoken throughout the combined sphere of influence of Persia and India, taking in large swathes of Central Asia. In this essay I will refer to the Indian branch of the PIE or Indo-Iranian as Indic. It has previously referred to as Aryan or Indo-Aryan but these terms have been deprecated because of the racial overtones of the word aryan and the discrediting of old ideas about race. Indic is strictly a linguistic term that gives us no information about ethnicity. We can talk about three phases of Indic: Old - principally attested as Vedic though other variations must have existed (before ca. 500 BCE); Middle - attested by Pāḷi, Gāndhārī, and Apabramsa (ca 500 BCE - 1000 CE); and New or Modern (emerging in the last millennia).

When the speakers of Old Indic crossed the Hindu Kush and entered India, ca 1700-1500 BCE, they met people who spoke languages with a much longer history in Greater India.

There is a whole family of Dravidian languages for example, including Tamil, Telegu, Malayalam, and Kannada. Today the people who speak languages from the Dravidian family are a large minority (about 20%). Some linguists (eg. McAlpin 1974, 1975, 1981) have noted a similarity between Dravidian and the language spoken in ancient Elam, near what is now the border of Iran and Iraq on the Red Sea. Written records of Elamite stretch back to 3000 BCE. McAlpin et al believe that Dravidian speakers split off from Elamite speakers and entered Indian very early, perhaps 4000 BC. Others are more doubtful (Blench 2008), dismissing the evidence for flimsy and pointing out affiliations with other language groups as well. 

Less well known is the Austroasiatic family. This family of languages extends from the North-east of India to Vietnam. One Indian branch of this widely geographically spread out family, is Munda, with several languages spoken in small pockets of India today, but probably more widespread in the past. In Burma there is a strong overlay of Tibeto-Burman languages that descended from the north, but there are still enclaves of Austroasiatic speakers as well. Genetic studies of Austroasiatic speakers suggest that the Austroasiatic language family may have arisen in India and spread east. 

Additionally there are a number of languages in India that appear to be unrelated to any known languages. These language isolates, as they are called, are found in the so-called tribal peoples who seem never to have been assimilated into the mainstream of Indian culture (in other words they were never Brahmanised).

Michael Witzel's exploration of the linguistic history of India begins by establishing his parameters, most important for the purposes of this essay is the periods of composition of the Ṛgveda (1999: 3).

• I. The early Ṛgvedic period: c. 1700–1500 BCE: books (maṇḍala) 4, 5, 6, and maybe book 2, with the early hymns referring to the Yadu-Turvaśa, Anu-Druhyu tribes;

• II. The middle (main) Ṛgvedic period, c. 1500–1350 BCE: books 3, 7, 8. 1–66 and 1.51–191; with a focus on the Bharata chieftain Sudās and his ancestors, and his rivals, notably Trasadasyu, of the closely related Pūru tribe.

• III. The late Ṛgvedic period, c. 1350–1200 BCE: books 1.1–50, 8.67–103, 10.1–854; 10.85–191: with the descendant of the Pūru chieftain Trasadasyu, Kuruśravana, and the emergence of the super-tribe of the Kuru (under the post-RV Parikṣit).

These layers of composition have been established on the basis of "internal criteria of textual arrangement, of the ‘royal’ lineages, and independently from these, those of the poets (ṛṣis) who composed the hymns. About both groups of persons we know enough to be able to establish pedigrees which sustain each other." (1999: 3).

Dutch Indologist, F. B. J. Kuiper, had already identified some 383 words in the Ṛgveda that are not Indic and must be loan words from another language family. We know this because they break the phonetic rules of Indic languages. We can use an example from English to demonstrate this. We have a word ptolemaic, which comes from the Egyptian name Ptolemy. It refers to a particular view of the world as earth-centred. Now we know that ptolemaic cannot be a native English word because English words cannot start with /pt/, and indeed native English speakers cannot easily pronounce this sound combination and tend to just say /t/. It is clues like this that linguists use to identify loan words. And we have to take into account that loan words are often naturalised. Many loans words in English are Anglicized. So another loan word like chocolate has been altered to fit English spelling patterns from an original spelling more like xocolātl, which clearly breaks English phonetic rules. We also have a number of Yiddish loan words like shlemiel, shlep, shlock, shmaltz, shmuck, and shnoozle, etc, that defy, but also to some extent redefine English spelling. Similarly no other Indic language has retroflex consonants (ṭ, ṭh, ḍ, ḍh, ṇ, ṣ), but Old Indic absorbed these from languages it met in India and they became a naturalised aspect of the Indic phonology by the time the Ṛgveda was composed.

It's not always possible to identify where a loan word has come from. But Kuiper and Witzel manage to identify most of the 300 words as belonging to Proto-Dravidian or Proto-Munda, with a few from other language families like Tibeto-Burman.

Perhaps the most striking finding that Witzel gives, repeatedly in his essay, is that in the early Ṛgvedic period there are no loan words from Dravidian, e.g.

"It is important to note that RV level I has no Dravidian loan words at all (details, below § 1.6); they begin to appear only in RV level II and III." (Witzel 1999: 6)

Ṛgvedic loans from Drav[idian] are visible, but they also are now datable only to middle and late Ṛgvedic (in the Greater Panjab), and they can both the localized and dated for the Post-Ṛgvedic texts. (Witzel 1999: 19)

This is an important finding. The landscape of the Ṛgveda is that of modern day Panjab. This is clear, for example, from the names of rivers that are mentioned, e.g. the Kabul, Indus, Sarasvati (now dried up) and Yamuna rivers.

Loan words from the earliest period are from the Austroasiatic language family, meaning that the people living in this area when the Vedic speakers arrived, spoke a variety of proto-Munda. This is important because it is believed that the people living in this area were the descendants of the collapsed Indus Valley Civilisation (IVC). They had scattered as the climate became much drier and caused their large scale cities to be unlivable. The IVC had disappeared by 1700 BCE. If the people of the Panjab ca 1500 BCE spoke a variety of proto-Munda, this strongly suggests that the people of the IVC also spoke an Austroasiatic language, rather than is usually supposed, a Dravidian on even Indic language. Indian nationalists often assume that the IVC spoke Sanskrit, but this was never plausible. Interestingly the very name we have for the north of this region, Gandhāra, is itself an Austroasiatic loan word.

It's often suggested that because there are northern pockets of Dravidian speakers, with whom the Vedic speakers presumably interacted, that Dravidian was once considerably more widespread and perhaps that the language of the IVC was Dravidian. The loan words in the Ṛgveda argue against this view. The north-western pockets of Dravidian could be isolated populations left behind by the migration of Dravidian speakers into Southern India from Mesopotamia. Those in the North-East are more consistent with a previously larger territory, but if they were ever on the Ganges Plain they were forced out of it completely, leaving remnant populations only as far north as mountain ranges on the southern edge of the Ganges Valley.

Conclusion

The picture that emerges is that Old Indic speaking people crossed the Hindu Kush in small numbers and met people who spoke a form of proto-Austroasiatic; and then later, perhaps as they penetrated further into the sub-continent, people who spoke proto-Dravidian languages. The Dravidian speakers, themselves probably immigrants had lived in India for some thousands of years already, displacing and assimilating even earlier waves of human migrants. The pockets of people who speak language isolates, not related to any known language, have presumably lived in India for a very long time. Indeed they often pursue a hunter-gatherer lifestyle that reinforces this impression.

Other authors have suggested that the Old Indic speakers had the advantage of superior technology and this led them to dominate the original inhabitants. We can't really know how it happened at this distant time, but in any case Indic languages came to dominate the North of India - from Afghanistan to the Ganges Delta. Again it is worth repeating that language, culture, and location may not be correlated. To the extent that we can make comparisons, there were a few surviving similarities between the people who composed the Ṛgveda and those who composed the Avesta. But in many respects their cultures had diverged along with their languages. Zoroastrianism was the major innovation in Iran, although the dates of the founder are difficult to pin down, the most likely scenario places him a little after the Ṛgveda. Based on informal comments by Michael Witzel, I have argued for an trickle of Iranian tribes entering India ca. 1000-800 BCE, who ended up settling on the margins of the Central Ganges city states of the second urbanisation, especially Kosala and Magadha (Attwood 2012). Genetic studies suggest that though their language came to be spoken throughout the Panjab and down into the Ganges Valley, that the Vedic speakers contributed little to the gene pool, which is remarkably homogeneous in India. The genetic contribution is far less striking than we might imagine by patterns of culture or language family (Attwood 2012).

This poses a difficulty for Indian Nationalists who want Sanskrit to be the mother tongue of India (I'm not sure how they fit Dravidian into the picture) and for it to have originated within the subcontinent. People with this view often express their hatred of Michael Witzel, referring to him in extremely uncomplimentary terms. But as rational people we have to follow the evidence and allow it to guide us to conclusions, even when these are uncomfortable for us. And the evidence is abundantly clear in this case. If any language is the mother tongue then it is probably Proto-Austroasiatic, the ancestor of the modern Munda and Austroasiatic languages. Sanskrit developed from Indo-Iranian, initially somewhere in Greater Iran, then was carried into India with Vedic speaking migrants. Since we know they were nomadic cattle herders (unlike, say, the Śākyas who were settled agriculturalists) they may have made the journey up the Khyber Pass seeking greener pastures.

In Attwood (2012) I tried to show that certain important features of early Indian Buddhist culture could be tied to Zoroastrianism and/or Iran. Unfortunately all too often the history of the region is divided into Indian and Iranian by academics. And thus I fear that many connections between the two regions have been overlooked. The connections that are evident seem to demand more attention from suitably qualified scholars. We know a great deal about the interactions of Greece and Persia, but far too little about relations between Persia and India.

~~oOo~~

Bibliography

Attwood, Jayarava. (2012) Possible Iranian Origins for Sākyas and Aspects of Buddhism. Journal of the Oxford Centre for Buddhist Studies. 3.

Blench, Roger (2008) Re-evaluating the linguistic prehistory of South Asia. Toshiki OSADA and Akinori UESUGI eds. 2008. Occasional Paper 3: Linguistics, Archaeology and the Human Past. pp. 159-178. Kyoto: Indus Project, Research Institute for Humanity and Nature.

Eco, Umberto. (1997) The Search for the Perfect Language. London: Fontana Press.

McAlpin, David W. (1974) Toward Proto-Elamo-Dravidian. Language 50: 89-101.

McAlpin, David W. (1975) Elamite and Dravidian: Further Evidence of Relationship. Current Anthropology 16: 105-115.

McAlpin, David W. (1981) Proto Elamo Dravidian: The Evidence and Its Implications. American Philosophy Society.

Witzel, Michael. (1999) Substrate Languages in Old Indo-Aryan: Ṛgvedic, Middle and Late Vedic. Electronic Journal of Vedic Studies. 5(1): 1–67.

New Heart Sutra Manuscript

1st folio of EAP 676/2/5.
For hi-res visit EAP

I made a surprise return to the world of manuscripts and the Heart Sutra this week. On Monday I saw a tweet that told me the British Library had uploaded images of more than 200 Buddhist Sanskrit manuscripts (in fact some of them are not Buddhist, but that's another story). The collection in question is called: Survey of Buddhist Sanskrit manuscripts in the possession of Vajrayana Viharas and Newar Buddhist families in Lalitpur in the Kathmandu valley, Nepal (EAP676). I scanned through the titles, not expecting anything very interesting but came across this:

EAP676/2/5: Ārya Pañcaviṁśatikā Prajñāpāramitā Mantranāma Dhāraṇī.

Now. Pañcaviṁśatikā Prajñāpāramitā is Sanskrit for the 25,000 line Perfection of Wisdom, literally it is The Perfection of Wisdom which is 25,000-fold. Not 'verses', note, since only one Prajñāpāramitā text is in verse, the Ratnaguṇasamcayagāthā, and it is not referred by the number of verses. These texts with numbers in their names are all in prose and the number of lines they take up depends on the length of the line and the handwriting of the scribe. What really peaked my interest is the last word, dhāraṇī, because the Heart Sutra is sometimes referred to in Nepalese manuscripts as the dhāraṇī of the Pañcaviṁśatikā Prajñāpāramitā.

Bhujimol Script.
(mantra begins on line 4)

This text is in Bhujimol Script which is one of the Nepalese 'hooked' scripts - they have a curl on top instead of a line. I have learned this before, even have read the Heart Sutra in Sanskrit in this script (one of the Cambridge Manuscripts) so I could make out enough of it to be confident that it was a long-text Heart Sutra. I set about transcribing the whole text and then comparing it to Conze's critical edition (both his selected text and variations). Since this manuscript is in a private collection in Nepal I thought it unlikely that Conze could have seen this one or used it in his edition. And this turns out to be the case. This manuscript has not been studied before to the best of my knowledge.

The manuscript is in reasonably good state of preservation, the paper is not blemished, mouldy or torn. There is some smudging of ink, but for the most part even the smudged akṣaras are readable. A quick note about the word akṣara. This word refers to a unit of Indic writing. In this type of text akṣara are not written joined up, each stands alone, meaning there are no word breaks. An akṣara represents a graphic unit rather than a spoken syllable, usually a single vowel, or one or more consonants followed by a vowel (we say pad ma, but write pa dma). Since each akṣara is a single graphic unit they are often called 'letters' though I think this is also imprecise. I favour using the Sanskrit term since it is more precise. A few akṣaras in our text are difficult to read because of the handwriting, or smudging, but can be inferred from the context and what remains on the page. Only one akṣara was completely un-guessable, since it is both obscure and seems to have been added by the scribe in error and is out of context. Some marks on the paper appear to have been deliberate, but are of uncertain meaning or function.

Regards the title I find I disagree with the catalogue record. The text's colophon - the source of the title of a Sanskrit ms, reads:

| āryyapañcaviṃśatika prajñāpāramitā samāptaḥ | 

Given the huge numbers of omissions I think it reasonable to conjecture an omitted word between prajñāpāramitā and samāptaḥ, namely hṛdaya. This occurs in almost all of the other manuscripts. Most importantly nowhere on the manuscript is the title: Ārya Pañcaviṁśatikā Prajñāpāramitā Mantranāma Dhāraṇī (nor does Conze list this precise variant in any of his witnesses). My title for the work would be Ārya-pañcaviṃśatika-prajñāpāramitā-hṛdaya. Though many of the other Nepalese mss. do refer to the text as a dhāraṇī, and it's very common to find it written along with the Uṣṇīṣavijaya Dhāraṇī.

The text is highly error ridden and there would be no point simply sharing an unedited transcription, so I have produced a diplomatic edition.  I have commented on the text at more than 140 places, which means on average I have commented on one word in two: a remarkable level of corruption given the supposedly sacred nature of the text. Some errors seem to have been introduced by the scribe. For example the sibilants (ś ṣ s) are frequently mixed up; l and r are exchanged without pattern; anusvāra (ṃ) and visarga (ḥ) are added and subtracted without pattern. Another regular error in this ms. is vya for ve. This kind of error is corrected in my edition. Nepalese manuscripts also regularly write sattva as satva, and have ryya for rya and rvva for rva, though EAP676/2/5 is inconsistent. These errors are not corrected, since they appear to be features rather than bugs. Word endings also vary without pattern and these have been normalised, following Conze 1967. Occasionally extra syllables have been added, and at one point a phrase seems to be copied twice (the scribe's eye may have drifted between lines). These appear in my edition with a strike-through. More often syllables, words and phrases have been omitted. All changes have been noted.

Of course it's possible that our scribe was simply copying a manuscript that was already corrupt. Thus these scribal errors can become generational. These types of errors become apparent when comparing many mss. And they help to establish relatedness amongst manuscripts. If two mss. have the same scribal error, the chances are they were copied from a single mss. which contained that error. We presume that the regular doubling of yya in ārya was an early scribal error that has since been propagated across Nepal as a feature.

One thing to note in comparing any text to Conze's critical edition, is that he sometimes simply ignores witnesses that are very corrupt even though he lists them as variations. I have found about half a dozen omissions great and small to date. As such, his critical notes are incomplete to an unknown degree, and unreliable. 

Some of the changes are not errors but editorial decisions. For example the usual way of saying "at that time" is tasmin samaye in the locative case. This is also seen in the phrase ekasmin samaye 'at one time". In the first para, after the Buddha goes into samādhi, Avalokiteśvara becomes the main protagonist and he is introduced using tena samayena which also means 'at that time' but uses the instrumental case. However in EAP676/2/5 some previous editor has changed this back into the locative. Such changes are not 'corrected' except where a scribal error has crept in. This change is also seen in a number of mss., using Conze's enumeration: Nd, Ne, Ni and Nm. Thus when we come to try to figure out the lineage of all the surviving mss., something I hope to do some day, then we'd expect these mss. that share this feature to be more closely related to each other than to other mss. Similarly for the idiom "tasmāc chāriputra" This spelling comes from the action of sandhi rules on the two words tasmāt śāriputra. Our text prefers tasmāt tarhi kulaputra or tasmāt tarhi śāriputra. This kind of change couldn't be the ms. equivalent of a typo. Someone must have decided this was "better". In this our text is like about half of the Nepalese mss., and the Tibetan. Conze's reference to the Tibetan is to a single unnamed source. As Jonathan Silk has shown two distinct recensions with many variations exist in printed editions of the Tibetan canon. Thus in comparing the texts we'd have to use Silk's critical edition of the Tibetan to see which recension our text is like.

Despite the sheer weight of errors and editorial changes in this text we still ought to be able to show how it relates to other texts. The generational scribal errors and editorial changes are like a DNA finger-print for the texts. However it's already clear that EAP676/2/5 is not linearly related to any of the other Heart Sutra mss. It sometimes follows one, and at other times another. The relationship between the Nepalese manuscripts appears to be complex. And EAP676/2/5 also has some unique features. For example in the Heart Sutra mantra the scribe writes gate 2 to indicate gate gate. This appears to be a unique feature amongst Conze's witnesses. The mantra itself gives us some sense of the level of corruption. As written it is:

| auṃ gate 2 pāraṃgate pārasaṃgate bodhisatva svāhā |

Note that auṃ is unusual in a Buddhist ms. In the second mantra (see below) we have the usual oṃ. Perhaps the scribe was used to seeing auṃ and we might conjecture that they also copy Hindu texts. I've put a strike through akṣaras added in error. The misplacing of anusvāra (ṃ) is a feature of these mss. They are added and subtracted seemingly at random. The scribe has added -satva to bodhi. This mistake is not found in other Heart Sutra mss.

Another unique feature of this ms. is that it is followed by another mantra in Lantsa or Ranjana script, by what appears to be the same scribe (it's difficult to tell). The mantra is introduced with an invocation of Avalokiteśvara and I was able to locate the exact mantra in the Sādhanamālā. This section also uses the numeral 2 to indicate repetition, and most of the numerals are surmounted by a chandra-bindu.

|| oṃ nāma śrī āryyāvvalokiteśvarāya bodhisatvāya mahāsatvāya mahākāruṇikāya | tadyathā | oṃ cala 2 cili 2 culu 2 hulu 2 mulu 2 huṃ huṃ huṃ huṃ huṃ phaṭ phaṭ phaṭ phaṭ phaṭ padmahastaya svāhā ༔ || iti āryyamokṣamantra nāma dhāraṇi samāptam | ༓ | ༓ |

A text like this was unlikely to have been made to be read. If it was the reader would be left very confused. Instead it was most likely produced as a act of merit by a pious Buddhist. Professional scribes churn these auspicious objects out and while we assume they do not deliberately make mistakes, neither are they checking for errors or making corrections. The manuscript might also be an object of worship. The dhāraṇī, like it's early predecessor the parittā, is a form of apotropaic magic and possession of such an object no doubt protects the holder (other Mahāyāna texts make this function of the text explicit by promising protection to anyone who would recite or copy the text). On the other hand it's entirely possible that the patron and/or the scribe had memorised the text and used it in their daily practice. Chanting in Sanskrit or Pali, by devotees who know nothing of the language, is a feature of Buddhist liturgy.

The edition with footnotes is available from my website: EAP676-diplo-edition.pdf. This is still a draft, since no one else has yet looked at it. It's painstaking, fiddly work and, ironically, errors are easily overlooked.

If nothing else this shows the value of putting manuscripts online. They become available to a much broader range of people - including feral scholars such as myself. High quality scans mean I never have to touch the fragile ms. itself and risk further damaging it. I can see everything I need to sitting at my desk. With larger manuscripts there is no longer a risk of mixing up leaves. I've spent a happy week examining this text. I'm grateful to everyone involved in putting this manuscript online, especially the British Library Endangered Archives Project and the owner of the ms. Shanker Thapa.

My other Heart Sutra essays, some graphics and my working bibliography are listed here (or under the tab at the top of the page).

~~oOo~~

Thanks for reading and all the best for 2015.

Manomaya Kāya: Other Early Texts

Essay no.400.

For the Nikāyakāra (the authors of the Pali Nikāyas) it was devas in the rūpadhātu (or their meditative equivalents) who possessed bodies (kāya) made by the mind, or were a mind-made group (kāya). Devas were supernatural beings, but they had the advantage of being part of the existing mythology of Buddhism (and the Ganges Valley generally). Devas were already culturally contextualised. Devas think, speak and interact with beings in the manussaloka (often with the Buddha) and thus, by the Buddhist understanding, they need to be embodied, to have a body endowed with senses, to possess all of the skandhas. We've seen (Manomaya Kāya: Pali Texts) that the Pali words for this are "rūpiṃ... sabbaṅga-paccaṅgiṃ ahīnindriyaṃ".

The antarābhava appears to be a new category of existence outside the universally accepted threefold model of the cosmos consisting of: kāmadhātu, rūpadhātu and arūpadhātu. Nor is it one of the five (later six) rebirth destinations (gati; sugati/durgati). These facts lie at the heart of the arguments of the sects that reject antarābhava. Like devas, beings in the antarābhava are conceived of as having cognition and thus they must be embodied in some form. Buddhists who believed in an antarābhava seem to have adapted the existing idea of a deva with a manomaya kāya to help explain the mode of existence in that state. This new mode of existence, outside of other models like the dhātus or gatis, implied a new ontology.

This essay will survey some of the non-Pali early Buddhist texts to see what use Buddhists made of manomaya in conjunction with antarābhava.


Samyuktāgama

One of the earliest references to this new ontology is in a Chinese translation of the Saṃyuktāgama (SĀ; the counterpart of the Pāḷi Samyuttanikāya). Lee ascribes this text to the Kāśyapīya Sect, which is "doctrinally close to the Sarvāstivāda". Bucknell (2011) tells is that SĀ is widely considered to have been translated in the period 435-443 CE from a Sanskrit Saṃyuktāgama brought to China from Sri Lanka". (Note there is another translation of the SĀ in Chinese)

佛告婆蹉。眾生於此處命終。乘意生身生於餘處。T99.244b03-05.

"When a sentient being exhausts the life-force in the present life, they ride (乘)
on a mind-made body (意生身) to be reborn in another place." (Adapted from Lee 2014: 70, Chinese from Radich 246 n.543)

The Pali counterpart of this passage (SN 44.9; iv.400) is also interesting, though it does not mention manomaya kāya. As noted in Arguments For and Against Antarābhava, it forms an essential part of arguments by Sujato and Piya Tan for the existence of an antarābhava, since it says:

Yasmiṃ kho, vaccha, samaye imañca kāyaṃ nikkhipati, satto ca aññataraṃ kāyaṃ anupapanno hoti, tamahaṃ taṇhūpādānaṃ vadāmi.

With respect to that, Vaccha, at the time when the body is relinquished, and a being is not arisen in certain kāya, I call that fuelled by craving.

I discussed how we might understand this passage in context in my earlier essay on antarābhava. Vaccha is asking about people who are not reborn and I said:

"To then read the question about rebirth in temporal terms, as explaining a time gap between bodies (kāya) is to misunderstand the metaphor. The question, really, is about what drives a person (satta) from body to body."

The idea of a satta (Skt sattva) going from body to body is consistent with Brahmin eschatology and we guess from his name (Vaccha is a Brahmin clan name) and the drift of his questions that Vacchagotta is a Brahmin. I also noted that "taṇha is always the upādāna for bhava" and that it cannot be considered specific to this case. I noted that the idea of a gap between lives may well have been Vaccha's and the Buddha simply failed to dispute it. Vacchagotta frequently pesters the Buddha and other bhikkhus with questions about ontological issues: "is there a self?" or "does the tathāgata exist after death?" and so on. In SN 22.10 (the next sutta) the Buddha refuses to answer his questions about the existence of self because any answer would have confused the Brahmin. Often such questions are said to be avyākata 'without explanation', by which the Buddha seems to mean that they can't be answered with certainty, only with speculation and he doesn't speculate, but that in any case they are irrelevant to the task at hand (Cf Cūlamālunkya Sutta MN 63).

In the Saṃyuktāgama texts antarābhava is one of four modes of existence (caturbhava) (Lee 2014: 70):

• rebirth (生有 = upapattibhava)
• life (本有 = pūrvakālabhava)
• death (死有 = maraṇabhava)
• between (中有 = antarābhava)

Antarābhava here seems to be well developed as an idea and as a state of existence has the same status as life; with death and rebirth as transitions between states - but all four having the same label bhava. We see here the beginnings of the Tibetan system of six bardos (which adds the states of dreaming and meditation). The antarābhava is no longer only a transition phase between kāmadhātu and rūpadhātu, as we saw in the discussion of the antarāparinirvāyin (Arguments For and Against Antarābhava).

In the《阿毘達磨大毘婆沙論》or *Mahāvibhāṣā, an encyclopaedic and influential Sarvāstivāda Abhidharma text which survives only in Chinese translation (T.27, no. 1545), we see, probably for the first time the equating of a number of terms: antarābhava, manomaya, gandharva and saṃbhavaiṣin (literally: 'one who seeks birth'). The Mahāvibhāṣā re-interprets manomaya to mean: "[beings in the antarābhava] are born complying with the mind" (Lee 2014: 74) and further to include "beings at the beginning of kalpas, all the beings of the intermediate existence (antarābhava), [the devas of] the pure form realm (rūpadhātu) and the formless realm (arūpadhātu), and the transformative bodies (*pariṇāma-kāya)." This seems to build on categories very like the ones we find in Pali. The category of pariṇāmakāya is obscure here, but taken up by some Mahāyāna texts. Vasubandhu, in his Abhidharmakośabhāṣya, continues the tradition of linking antarābhava, manomaya, gandharva and saṃbhavaiṣin.



Ekottarāgama

The word manomaya occurs in some Sanskrit fragments of the Ekottarāgama (EĀ; Tripathi 1995) which is a counterpart of the Pali Aṅguttara Nikāya. Here EĀ proposes two kinds of manomaya kāya which depend on the conduct of the being, i.e. an ethicized version of the manomaya kāya in SĀ.

yo 'sau bhavati strī vā puruṣo vā duḥśīlaḥ pāpadharmāḥ kāyaduścaritena samanvāgato vāṅmanoduścaritena samanvāgatas tasya kāyasya bhedād ayam evaṃ rūpo manomayaḥ kāyo 'bhinirvartate tadyathā kṛṣṇasya kutapasya nirbhāsaḥ andhakāratamisrayā vā rātryā yeṣāṃ divyaṃ cakṣuḥ suviśuddhaṃ ta enaṃ paśyanti || EĀ 18.51 || 

That woman or man of bad conduct and evil-nature, endowed with bad behaviour of the body and bad behaviour of speech and mind, with the breaking up of the body [at death] they give rise to (abhinirvartate) a form, a mind-made body. It has the appearance of a black blanket or the blinding darkness of night, which they see with the purified divine eye.

yo 'sau bhavati strī vā puruṣo vā śīlavān kalyāṇaadharmāḥ kāyaduścaritena samanvāgato vāṅmanahsucaritena samanvāgatas tasya kāyasya bhedād ayam evaṃ rūpo manomayaḥ kāyo 'bhinirvartate tadyathā śuklasya paṭasya nirbhāsaḥ jyotsnāyā vā rātryā yeṣāṃ divyaṃ cakṣuḥ suviśuddhaṃ ta enaṃ paśyanti || EĀ 18.52 ||

That woman or man of ethics and good-nature, endowed with good behaviour of the body and good behaviour of speech and mind, with the breaking up of the body [at death] produce a form, a mind made body. It has the appearance of white cloth or a moonlit night, which they see with the purified divine eye.

This phrasing is both similar to and different from Pali counterparts. The image appears to be absent from the Pali. The closest we get is this:

So kāyena duccaritaṃ caritvā vācāya duccaritaṃ caritvā manasā duccaritaṃ caritvā, kāyassa bhedā paraṃ maraṇā apāyaṃ duggatiṃ vinipātaṃ nirayaṃ upapajjati.

So kāyena sucaritaṃ caritvā vācāya sucaritaṃ caritvā manasā sucaritaṃ caritvā, kāyassa bhedā paraṃ maraṇā sugatiṃ saggaṃ lokaṃ upapajjati. (SN i.93)

Having behaved badly with the body, behaved badly with the voice, and behaved badly with the mind, with the breaking up of the body after death, he goes to a bad destination, a state of suffering, reborn in hell.

Having behaved well with the body, behaved well with the voice, and behaved well with the mind, with the breaking up of the body after death, he goes to a good destination, reborn in heaven.

The idea of a dark and bright manomaya using much the same terminology (highlighted in bold) is reflected much later in Asaṅga's version of the antarābhava in his Bodhisattvabhūmi (Chapter 3.6; Cf Wayman 1974: 233)

dvābhyām ākārābhyāṃ tamaḥ-parāyaṇānām ayam evaṃ rūpo manomayo 'ntarābhavo nirvartate | tadyathā kṛṣṇasya kutapasya nirbhāsaḥ andhakāra-tamisrāyā vā rātryāḥ | tasmād durvarṇā ity ucyante | 

Because of the two modes [of action] thus a form which is mind-made in the interim state is produced filled with darkness, just like the appearance of a black blanket or the blinding darkness of night. Because of that they call it “inferior”. 

ye punardvābhyām ākārābhyāṃ jyotiḥ-parāyaṇās teṣām ayam evaṃ rūpo manomayo 'ntarābhavo nirvartate | tadyathā jyotsnayā rātryā vārāṇaseyakasya vā sampannasya vastrasya | tasmātsuvarṇā ityucyante | (Dutt 269 Wohihara ed 390-91)

Because of the two modes [of action] a form which mind-made in the interim state is produced which is filled with light, just like a moonlit night or excellent cloth that comes from Vārāṇasi. Therefore they call it “superior”. 

The context shows that this passage is also referring to duścarita and sucarita. One leads to a bad destination (durgati-gāmina) the other to a good destination (sugati-gāmina).


Mahāvastu & Lalitavistāra

This pair of texts, composed in Sanskrit, are often seen as transitional between early Buddhist and Mahāyāna texts. Sumi Lee doesn't go into these seminal texts or their use of manomaya kāya, perhaps because their treatment of manomaya kāya is similar to the Pali. In Sanskrit we do begin to see manomayakāya as a compound.

In the Mahāvastu (Mhv) there is a retelling of the beginning of a new epoch of the cosmos (cf DN i.17). The first beings to come into existence are self-luminous, move through the sky (antarīkṣa), are mind-made, feed on rapture, are in a state of bliss, and can move about as they wish. (svayaṃprabhāḥ antarīkṣacarā manomayā prītibhakṣāḥ sukhasthāyino yenakāmaṃgatāḥ Senart 1.338; cf. Jones 285-286). A little further on when these beings fall from this refined state due to greed, they lose the state of being a mind-made group (manomayakāyatā). Thus the usage in Mahāvastu closely reflects the Pali usage. Manomaya kāya refers to devas and is used in conjunction with this old parody of Brahmanical notions of cosmogony (which may have been hypostatised by this time).

The phrase manomaya is used just twice in the Lalitavistāra Sūtra (Lv). Firstly it is used in a gāthā:

atha khalu sunirmito devaputro rājānaṃ śuddhodanam upasaṃkramyaivam āha—manomayam ahaṃ śrīmadvaśma tad ratanāmayam |
bodhisattvasya pūjārtham upaneṣyāmi pārthiva || Lal_6.18 || [Vaidya 46]

Then indeed a beautifully formed divine child approached King Śuddhodanam and said:
I will offer a mind-made, glorious jewelled mansion;
As an act of worship of the bodhisattva, O King.

My translation here follows the Dharmachakra Translation Committee translation of the Tibetan in taking śrīmad-vaśma to mean "a glorious mansion". My dictionaries have no word vaśma[n]; it may be a hyper-Sanskritisation of vasman 'nest' (from √vas 'to dwell'). Here manomaya must mean 'made by the mind' in the sense of 'mental' or 'imaginary'. Compare ratanā-maya 'made of jewels' or 'jewelled'. Note that Bays translation obscures the presence of the word manomaya.

Secondly manomaya is once again used with reference to some devas, in this case devakanyā or the girls of the devas. They are described as divya-manomaya-ātmabhāva-pratilabdha (Vaidya 36) terms familiar from the discussion of manomaya especially DN i.197-202. The term attapaṭilābha 'acquired self' (Skt ātmapratilabdha) which Buddhaghosa glossed as attabhāva 'a state of self' (Skt ātmabhava) and the sutta describes as having three types: oḷārika, manomaya, and arūpa; with the second being associated with the rūpadhātu. So the Lv adjective means 'having the acquired state of self of a divinity' though what this means in practice is not clear. Thus in these transitional period texts we are not seeing an association with antarābhava or the afterlife at all.


Conclusion

One important point to make with respect to the antarābhava and manomaya kāya is that the Āgama texts reflect the view of the sect who preserved them and the Nikāya texts (largely) reflect the Theravāda view. My inclination is to explain this presence and absence as the addition of antarābhava to the texts of those who believed in it.  Of course, it is impossible to eliminate the possibility that the antarābhava has been retrospectively expunged from the Pali texts. This however suggests proactive editing on a much larger scale than I have ever encountered. It seems more likely that as time went on new ideas were added in, than that old ideas were expunged. Buddhist texts tend to be quite conservative of old ideas. Presumably some ideas were introduced and subsequently died out. For some of these we no doubt have stubs in the Canon - brief mentions with no follow up.

Our non-Pali witnesses further confuse the situation: the Chinese Āgama translations are from the 4th or 5th century CE and at least in the texts I've studied, section 5 of the Madhyāgama for example, show a higher degree of standardisation and homogenisation than the Pali texts. The Sanskrit translations are similar to the Pali texts, but also distinct in many ways, suggestive of long separation between Sanskrit- and Pali-using sects.

What we do know is that some lines of development across the spectrum of early Buddhist thought included references to a manomaya kāya in the antarābhava and others only mention manomaya kāya with respect to the psycho-cosmology of the deva realms. The development of the idea of manomaya kāya in Buddhism seems to go like this:

• Devas in the rūpadhātu are a manomaya (ni)kāya (group).
• Meditators in the fourth jhāna magically create (abhinir√mā) a manomaya kāya (body) which is rūpin (out-of-body experiences?)
• Non-returners (anāgāmin) transitioning from the kāmadhātu to the rūpadhātu do so in a manomaya kāya (body).
• The advent of antarābhava leads to all beings having (or "riding") a manomaya kāya in the interim between death and rebirth.
• Antarābhava and manomayakāya are equated, along with gandharva.

In these early Buddhist texts there are two distinct metaphysics reflecting a binary split over the time interval between death and rebirth. The sole surviving representatives of those who claimed no interval, the Theravādins, dealt with this issue as part of their comprehensive response to the problems in the tradition, as part of their Abhidhamma: a stream of citta moments, connected in up to twenty-four ways (paccayatā), according to certain restrictions (niyama) making no initial ontological distinctions. In this and related schools manomaya kāya largely remains a description of beings in the rūpadhātu well into the common era. And this does not stop Theravādins on the ground believing in an interim state, or beings in an interim state, or subtle bodies.

On the other hand if one stipulates that it takes some appreciable time for rebirth to occur, then certain questions arise. In particular we want to know what form of existence one has between death and rebirth, since non-existence is nonsensical. The proposed explanation had to avoid the trap of eternalism by not being nicca, sukha, and attan, but it had to explain the connectivity and continuity (in karmic terms). Additionally, for most Buddhists, especially in the ancient world, it could not conflict with scripture. Proponents of antarābhava had to invent a whole new field of inquiry and vocabulary for it. The previous unrelated term manomaya formed part of this new metaphysics.

In the next essay in this series we'll look at some Mahāyāna sources to get a flavour of how the idea developed in those (many and various) milieus. A comprehensive survey is neither within my means or skill level, but by looking at some influential texts, Śāntideva's anthology of sūtra texts, Śikṣasamuccaya, and Vasubandu's Bhāṣya, we can at least get a sense of how some of the more prominent Buddhists of later periods viewed manomaya kāya.

~~oOo~~

Bibliography

Aurobindo (2004) The Upanishads: Kena and other Upanishads. Sri Aurobindo Ashram Publication Dept.

Bucknell, Roderick S. (2011) ‘The Historical Relationship Between the Two Chinese Saṃyuktāgama Translations.’ Chung-Hwa Buddhist Journal. 24:35-70.

Dharmachakra Translation Committee. (2013) The Play in Full: Lalitavistara. [Ārya-lalitavistara-nāma-mahāyāna-sūtra] http://read.84000.co/browser/released/UT22084/046/UT22084-046-001.pdf

Jones, J J. (1949) The Mahāvastu. Vol. 1 Luzac. https://archive.org/stream/sacredbooksofbud16londuoft#page/286/mode/2up

Lee, Sumi. (2008) 'The Philosophical Meaning of Manomaya-kāya.' 2008 Korean Conference of Buddhist Studies. http://www.skb.or.kr/down/papers/129.pdf [pages not numbered]

Lee, Sumi. (2014) 'The Meaning of ‘Mind-made Body’ (S. manomaya-kāya, C. yisheng shen 意生身) in Buddhist Cosmological and Soteriological systems'. Buddhist Studies Review. 31(1): 65-90.

Radich, Michael David. (2007) The Somatics of Liberation: Ideas about Embodiment in Buddhism from Its Origins to the Fifth Century C.E. [PhD. Dissertation]. https://www.academia.edu.

Senart, Émile. (1897) Mahavastu-Avadana. 3 vols., Paris 1882-1897. http://gretil.sub.uni-goettingen.de/gretil/1_sanskr/4_rellit/buddh/mhvastuu.htm

Suzuki, Daisetz Teitaro. (1930) Studies in the Lankavatara Sutra. Routledge.

Tripathi, Chandra Bhal (1995) Ekottarāgama-Fragmente der Gilgit-Handschrift, Reinbek 1995 (Studien zur Indologie und Iranistik, Monographie 2). http://gretil.sub.uni-goettingen.de/gretil/1_sanskr/4_rellit/buddh/ekottaru.htm

Vaidya, P. L. (1963) Saddharmalaṅkāvatārasūtram. The Mithila Institute of Post-Graduate Studies and Research in Sanskrit Learning. Darbhanga.

Wayman, Alex (1974) 'The Intermediate-State Dispute in Buddhism' in Buddhists Studies in Honour of I. B. Horner. D Reidel: 227-239.

The Second "Hidden" Kātyāyana Sūtra in Chinese

Mahākātyāna
Stele, Korea.

This text is "hidden" because even though it has been translated into English (Choong 2010), it has not been discussed in relation to the other versions of the text so far as I'm aware. What tends to happen is that when the text is mentioned, scholars think of the Pāli version or the Sanskrit passage cited by Candrakīrti in his commentary on Nāgārjuna's Mūlamadhyamaka Kārikā which mentions the Kātyāyana Sūtra (MMK 15.7). I'm hoping to give some prominence to the other versions of which two are in Chinese.

The Pali Kaccānagotta Sutta (SN 12.15 = KP) is quoted verbatim in the Channa Sutta (SN 22.90; iii.132-5) and as such is of little interest except that when a text is cited by another text we get a sense of relative dating: it implies chronology. In the Chinese Saṃyuktāgama, the counterpart of the Channa Sutta (CC; SĀ 262 = T 2.99 66c01-c18) also quotes the Chinese counterpart of the Kātyāyana Sūtra (KC; SĀ 301), but in this case the text is different in some interesting ways. And thus we have a fourth version of the text: KP, KS, KC and now CC.

Most significant is how the two Chinese versions deal with a partic-ularly difficult paragraph that in Pali and Sanskrit reads:

KP: dvayanissito khvāyaṃ kaccāna loko yebhuyyena atthitañceva natthitañca. Upayupādānābhinivesavinibandho khvāyaṃ, kaccāna, loko yebhuyyena. Tañcāyaṃ upayupādānaṃ cetaso adhiṭṭhānaṃ abhinivesānusayaṃ na upeti na upādiyati nādhiṭṭhāti ‘attā me’ti.

“Generally, Kaccāna, this world relies on a dichotomy: existence and non-existence.” Usually, Kaccāna, this world is bound to the tendency to grasping and attachment. And he does not attach, does not grasp, is not based on that biased, obstinate tendency of the mind to attachment and grasping: [i.e.] “[this is] my essence”.

 ~~~

KS: dvayaṃ niśrito ’yaṃ kātyāyana loko yadbhūyasāstitāñ ca niśrito nāstitāñ ca | Upadhyupādānavinibaddho ’yaṃ kātyāyana loko yad utāstitāñ ca niśrito nāstitāñ ca | etāni ced upadhyupādānāni cetaso ’dhiṣṭhānābhiniveśānuśayān nopaiti nopādatte nādhitiṣṭhati nābhiniviśaty ātmā meti |

“Generally, Kātyāyana, this world relies on a dichotomy: it relies on existence and non-existence. This world, which relies on existence and non-existence, is bound by attachments and grasping. If he does not attach to these, does not grasp, is not based on or devoted to the biased, obstinate tendency of the mind to attachments and grasping: “[this is] my essence”.

~~~

Notes

The syntax here is tortuous and in addition contains some distracting word play. The nouns in the green section are from the same roots as the verbs in the orange section. Both Chinese versions replicate this same structure. It's possible that the nouns and verbs are meant to be understood as linked: upāyaṃ with na upeti; upādānaṃ with na upādiyati and so on, but at this stage I'm unsure. The Sanskrit is more difficult to parse because of the "if" (ced) and the Pali seems like a better reading for not having it. 

Note that P "attā me" & Skt "ātmā me" appear to be references to the formula often used with reference to the skandhas. Here wrong view would be of the form:

rūpaṃ etam mama, eso’ham amsi, eso me attā ti samanupassati.
He considers form: “it is mine”; “I am this”; “this is my essence”.

Our text hints that the duality of existence (astitā) and non-existence (nāstitā) arises from the same wrong view. Indeed seeing experience in terms of existence and non-existence is probably at the heart of interpreting it as "mine", "I" or "my essence". 

The Saṃyuktāgama text translated into Chinese was evidently similar to the Sanskrit of KS and carried out by Guṇabhadra in the 5th century CE. Even non-Chinese-readers will see there are similarities and differences in the two Chinese versions of this paragraph, which I've marked up using the same colour scheme as above for comparison.

KC: “世間有二種依， 若有、 若無，為取所觸； 取所觸故，或依有、或依無。若無此取者。心境繫著。使不取、不住、不計我。

KC: “Among the worldly (世間) two categories are relied on: being and non-[being]. Because of having grasping the touched, they either rely on being or non-being. If he is not a seizer of that , he doesn’t have the obstinate mental state of attachment; he doesn’t insist on, or think wrongly about ‘I’.”

~~~

CC: 『世人顛倒依於二邊，若有、若無，世人取諸境界，心便計著。迦旃延！若不受、不取、不住、不計於我，

CC: “Wordly people (世人) who are topsy-turvy (顛倒) rely on (依) two extremes (二邊): existence (若有) and non-existence (若無). Worldly people (世人) generally (諸) adhere to (取) perceptual objects (境界) [because of]  because of biased, obstinate tendency of the mind (心便計著) Kātyāyana: if not appropriating (受), not obtaining (取), not abiding (住), not attached to or relying on ‘I’...”

~~~

The first difference is in interpreting Skt/P. loka. KC translates 世間 while CC has 世人: "in the world" versus "worldly people". CC adds that the worldly people are 顛倒 "top-down" or "upside-down" or "topsy-turvy". Choong translates "confused", which is perfectly good, but there's a connotation in Buddhist jargon of viparyāsa (c.f. DDB sv 顛倒) which refers to mistaking the impermanent for the permanent and so on.

KC and CC both translate niśrito/nissito as 依. But they again differ in how they convey dvayam: KC 二種 "two varieties" and CC 二邊 "two sides". The character 邊 often translates Skt. anta which is significant because the word crops up later in the text in the Sanskrit and Pali: Skt:

ity etāv ubhāv antāv anupagamya madhyamayā pratipadā tathāgato dharmaṃ deśayati |

Thus, the Tathāgata teaches the Dharma by a middle path avoiding both these extremes.

KC and CC both use 二邊 in this latter passage. It makes more sense to refer to "two extremes" (antau; in the dual) early on if that's what's talked about later, especially when by "later" we mean just three sentences later. Thus CC provides better continuity than KC.

The next part of this section is where the two texts differ most markedly.

KC: Because of having grasping the touched (取所觸), they either rely on being or rely on non-being (或依有、或依無). If [he is] not a seizer of that (若無此取者), he doesn't have the obstinate mental state of attachment (心境繫著).

CC: worldly people generally (境 ) adhere and attach to 計著 objects of the mind (界心). Kātyāyana: if not appropriating (受), not obtaining (取), not abiding (住), not attached to or relying on “I”...

(Choong "Worldlings become attached to all spheres, setting store by and grasping with the mind.")

In KC we have some confusion around the phrase 取所觸. In Choong's translation of KC (40) he wants to have it mean “This grasping and adhering" but that's not what it appears to say and in any case no dictionary I have access to translates 觸 chù as ‘adhere’ or anything like it. On face value, and taking into account Buddhist Chinese it says "grasping what is touched": 取 = Skt. upādāna; 所 = relative pronoun; 觸 = Skt. sparśa < √spṛś 'touch'. In other words Guṇabhadra seems to have made a mistake here. I think Choong is tacitly amending the text to correct it, probably based on reading the Pali.

Elsewhere KS seems to be defective: KP has upay(a)-upādāna-abhinivesa-vinibandha ‘bound by the tendency to attachment and grasping’ whereas KS has upadhy-upādāna-vinibaddho, missing out abhinivesa, which doesn't really make sense. Upadhi is out of place here and probably a mistake for upāya. It may be that the source text for KC was also defective. 

Note that CC has abbreviated the text. The green section of KC repeats some of the first red section, but CC eliminates the repetition and makes the paragraph easier to read overall. 

The Chinese texts both run on to include the next section, although it's clear from KP and KS that the next part is a separate sentence. 

Conclusion

"In short, when reading any given line of a Chinese Buddhist sūtra—excepting perhaps those produced by someone like Hsüan-tsang, who is justifiably famous for his accuracy—we have a roughly equal chance of encountering an accurate reflection of the underlying Indian original or a catastrophic misunderstanding."

Jan Nattier. A Few Good Men. p.71

As a warning this might be slightly overstated for effect and it is qualified by Nattier who says that multiple translations make it easier for the scholar. But it's often true that in order to really get what a Chinese text is on about, one must use the Indic (Pāḷi, Saṃskṛta, Gāndhārī) text as a commentary. This is partly because Buddhist Chinese is full of transliterations and jargon. Words are used in ways that are specific to a Buddhist context and must be read as technical terms. Buddhist Chinese very often uses something approximating Sanskrit syntax (Chinese is an SVO language while Sanskrit is SOV). The paragraph we have been considering is a good example of this phenomena as the Chinese apes the syntax of the Sanskrit. 

It's hard to avoid the conclusion that KC and CC were translated by different people and that the translator of CC did a slightly better job than the translator of KC. So perhaps the named translator, Guṇabhadra, was a sort of editor-in-chief working with a team? This was a common way of creating Chinese translations. Or perhaps he translated the same passage twice and did it differently each time? Though this seems less likely. 

By comparison with the Pāli Tipiṭaka we expect KC and CC to be identical, as the quotation of KP in the Channa Sutta is verbatim. The fact that they are not raises questions about the source text for the Samyuktāgama translated in Chinese. Having different translations into Chinese is valuable because it is precisely where KC is difficult that CC is different and arguably clearer. But perhaps the different translations are because the source text itself was different? KS is different from KP in other ways, and different from citations in later literature. This points to a number of versions of the text being in circulation of which we have a sample in the various canons.

So often the Chinese Tripiṭaka contains little that conflicts with the Pāḷi Tipiṭaka. But sometimes, as in this case, the differences are instructive, especially where versions in Sanskrit and/or Gāndhārī survive. We're now starting to see the treatment of Pali and Chinese versions of texts side by side in articles about early Buddhism. No doubt the publication of canonical translations into English, which has begun, will facilitate this. Certainly Early Buddhism is no longer synonymous with Theravāda and Pāḷi.

My close reading of all four Kātyāyana texts is slowly becoming a journal article. A subsequent project will be to explore the many citations of the text in Mahāyāna Sūtras. Exact citations or mentions of the same idea can be found in at least the Aṣṭasāhasrikā Prajñāpāramitā Sūtra  and the Laṅkāvatāra Sūtra, and also in Nāgarjuna's Mūlamadhyamakakārikā and especially in Chandrakīrti's commentary on MMK, Prasannapāda. Thus the text and the ideas in it were foundational to the Mahāyāna and provide an important thread of continuity, between the first two great phases of Buddhist thought.

~~oOo~~

Roots of the Heart Sutra

Edward Conze was of the opinion that the oldest layer of the Prajñāpāramitā textual tradition is probably the  first two chapters of the Ratnaguṇasaṃcayagāthā (Rgs). He sees it as closely related to the Aṣṭasāhasrikā Prajñāpāramitā Sūtra (8000 line Perfection of Wisdom Text; Aṣṭa) but conjectures that verse precedes prose. Other scholars have conjectured that the verse is actually a post hoc summary of the prose. At present there is insufficient information to decide one way or the other. In some manuscripts the Rgs is included as a chapter of a larger Prajñāpāramitā text (8k and 100k), inevitably numbered "chapter 84".

While we don't have very old manuscripts of the Rgs, we do have one of the Aṣṭa in Prakrit from the 1st Century (carbon dated to between 47 and 147 CE) and thus probably from the end of the 1st Century CE. We also have a very early (179 CE) Chinese translation by the Scythian translator, Lokakṣema. We now know that the Aṣṭa was composed in Prakrit in Gandhāra which solves some of the existing problems of where to locate the early Prajñāpāramitā tradition. Rgs by contrast was not translated in Chinese until the 10th century by Faxian 法賢 (991CE), 《佛母寶德藏般若波羅蜜經》Fúmǔ-bǎodécáng-bōrěbōluómì(duo)-Jīng (T 229). This means that it played no part in the understanding and development of Perfection of Wisdom thought in China. This Chinese version is, according to Yuyama (1976), probably the most corrupt of all the versions. It is closer to Recension A, but still very different in many places. This may be due to Faxian's "free (or perhaps bad) translation." (xl)

As we now know, the Heart Sutra or Hṛdayaprajñāpāramitā is mostly comprised of some quotes from a Chinese Large Perfection of Wisdom Text equivalent to the Sanskrit Pañcaviṃśatisāhasrikā Prajñāpāramitā Sūtra. The most likely candidate is Kumārajīva's translation 《摩訶般若波羅蜜經》Móhēbōrěbōluómì Jīng by Kumārajīva (T 223; 404 CE). We also know that the larger texts in 18,000, 25,000 and 100,000 lines are simply an expansion of the Aṣṭa. It is possible to trace many of the cited passages from the Hṛdaya back into the Pañcaviṃśati; but we ought to be able to go one step further and trace some of them back into the Aṣṭa and Rgs. This essay will revisit an aspect of the roots of the Hṛdaya in the Prajñāpāramitā literature, particularly for the epithets passage:

Tasmāj jñātavyam prajñāpāramitā mahāvidyā anuttaravidyā 'samasamavidyā. 

In Jan Nattier's watershed article (1992) she notes a few examples of this genealogical approach with respect to the epithets passage. Nattier, in note 54a, explains that the epithets are epithets of prajñāpāramitā itself and that the word mantra is an erroneous back translation of 明咒 míng zhòu (= vidyā), with confusion arising because 咒 zhòu is used in the same capacity and even appears alongside 明咒. I conjecture that the confusion was exacerbated because the composition of the early Prajñāpāramitā texts occurred well before Tantric Buddhism, while the translation of the Hṛdaya from Chinese into Sanskrit occurred post-tantra and the reading of 咒 was.

An interesting comparison is with the six-syllable mantra: oṃ maṇipadme hūṃ. Alex Studholm's study of the Kāraṇḍavyūha Sūtra (Kvs) reveals the early history of this mantra. I've written two previous essays on Studholm's conclusions:

• The origins of oṃ maṇipadme hūṃ.
• The Meaning of oṃ maṇipadme hūṃ.

Throughout the Kvs the "mantra" is never referred to as a mantra, but only and always as the ṣaḍakṣarī mahāvidyā (61) "the six-syllabled great techne". Vidyā is a difficult word to find a good English translation for in this context. Techne is a Greek word which has almost exactly the same range of meaning, i.e. 'art, craft, skill', and is the source of our word technology. I'm trying it out in this essay, but I wouldn't insist on it. My only stipulation would be that we cannot translate vidyā as "science" because it is anachronistic. The word "science" properly applies to the systematic empirically based descriptions of the world, tested through conjecture and refutation, which began in the late 15th century in Europe. There is an archaic use of the word science which does cross over with vidyā, but it is archaic. And in the present there is too much confusion over the distinctions between science and religion to be slack in how we use the words. Better a neologism than an anachronism if we communicating to a present day audience. Indeed if anything the argument ought to go the other way. We might, as some do, translate vidyā as 'spell' or 'magic'. The main idea here is the specialist knowledge of Buddhist practices which allow us to gain first-hand, experiential knowledge of how the perceptual situation creates duḥkha. This has a magical flavour in many texts, and can result in extra-sensory perceptions, but there is nothing of science here. Vidyā is religious knowledge.

As we know, mahāvidyā is one of the epithets of prajñāpāramitā also. In the Chinese Heart Sutra, it is 大明咒  dà míng zhòu, where 大 means 'great' and 明咒 = vidyā. Due to the ambiguity of the characters and the confusion over the original meaning this is often translated as "great bright mantra". See e.g. Mu Soeng's recent translation and commentary on the Chinese version. Soeng, like Red Pine, simply ignores Nattier's findings. In the Chinese 咒 = vidyā also. Hence the change I have suggested in the Sanskrit: that "mantra" (qua word) is eliminated from the text and replaced everywhere with vidyā. For the argument in more detail see: Heart Sutra Mantra Epithets.

In Nattier's article she notes that the epithets passage has been traced by Nobuyoshi Yamabe in various Chinese translations of Pañcaviṃśati and Aṣṭa. I have been working on a comprehensive list of the counterpart passages in the basic Prajñāpāramitā texts in Sanskrit and Chinese. There are in fact only two passages, but they recur across two languages and multiple translations of multiple texts and so number about two dozen in total. I hope to publish this information before too long. Using electronic searches, I also came across a counterpart in the Rgs. In Chinese the passage is (T 8.229 678.a4-5):

大明般若諸佛母，
能除苦惱徧世界，
所有三世十方佛，
學此明得無上師。 

This great techne of perfect wisdom is the mother of all Buddhas,
Able to remove distress in all world spheres,
All the Buddhas of the three times and the ten directions,
Schooled in this techne are the supreme masters.

We can identify it as a counterpart because of its place in the sequence; its use of mahāvidyā in connection to prajñāpāramitā; and because it connects the vidyā with the buddhas of the three times and ten directions. The latter also links it to the second of two possible Pañcaviṃśati passages that are the basis of the Hṛdaya epithets (and the more likely of the two). In this translation mahāvidyā is rendered as 大明 dà míng (line 1), while vidyā in line 4 is simply 明 míng, in line with the practice of the day.

There are two main recensions of the Sanskrit Rgs and I have used recension A from the critical edition published by Akira Yuyama. Vaidya's edition is based on Recension B mss and has only minor differences in this verse which are discussed below. Conze describes the metre as "irregular Vasantatilaka" though my sources suggest that this is a 14 syllable metre, with a caesura after 8, i.e.:

– – ∪ | – ∪ | ∪ ∪ – ॥ ∪ ∪ – | ∪ – –

(∪ light; - heavy; ॥ ceasura)

Rgs has 13, 14 and 15 syllable lines, with 15 seeming to be the norm and the verse we're looking at is quite regular. The final pattern of Rgs 3.5 does match the post-caesura portion of Vasanatilakā, i.e. | ∪ ∪ – | ∪ – – .

Note 21 Nov 2014. In a comment Kevin suggests reading the initial two light syllables ∪ ∪ a one heavy –. This would result a perfect match with the pattern found in references:

(∪ ∪) – ∪ | – ∪ | ∪ ∪ – || ∪ ∪ – | ∪ – – 

The text is not really in Sanskrit, but rather in a Prakrit that has been Sanskritised to some extent. It's quite near the Prakrit end of the Buddhist Hybrid Sanskrit spectrum, though this verse looks more or less like Sanskrit. Conze's translation is based on his "corrected" edition of Recension B, published in Russia by E. Obermiller and (not infrequently) on the Tibetan edition (Yuyama 1976: xlvii). I will compare both, but favour Recension A as found in Yuyama (1976). The text of the Rgs (A) verse which corresponds to the Chinese above is:

mahavidya prajña ayu pāramitā jinānāṃ |
dukhadharmaśokaśamanī pṛthusattvadhātoḥ ||
ye ’tīta’nāgatadaśaddiśa lokanāthā |
ima vidya śikṣita anuttaravaidyarājāḥ || Rgs 3.5 ||

∪∪ –  * | – ∪ | ∪ ∪ – || ∪ ∪ – | ∪ – –  (* see discussion)
∪∪ – ∪ | – ∪ | ∪ ∪ – || ∪ ∪ – | ∪ – –
    – – ∪ | – ∪ | ∪ ∪ – || ∪ ∪ – | ∪ – –∪∪ – ∪ | – ∪ | ∪ ∪ – || ∪ ∪ – | ∪ – –

This perfection of wisdom of the Jinas is a major techne;
In the realm abounding in beings, whose nature is suffering, grief, and darkness.
The world protectors of past and future, in the ten directions, who;
Trained in this knowledge, are the unexcelled kings of the knowledgeable.

The overall similarity of content shows that these two are the same passage, and the numbering also accords. That said some interesting differences occur as we would expect from the introductory comments based on Yuyama's observations about the state of the Chinese text.

The first line in particular marks this passage out as not being Classical Sanskrit. In Sanskrit it would probably read: mahāvidyā prajñā ayaṃ pāramitā jinānām | Such differences might appear trivial to the non-philologist, but all to often it is these differences in vowel length and final syllables that make the difference between languages. One can see that, here at least, the metre is fairly regular and in fact the Classical Sanskrit spelling would spoil the metre!

In line 1 measure 2 the ya in vidya does not make position. It is followed by pra which ought to make it heavy. The Pāli spelling would be paññā so it may be that the Prakrit speaking author pronounced prajña more like Pāli i.e. /pa/ rather than /pra/ and thus thought of ya as light. Or it may be an adjustment metri causa. Also the caesura occurs after the pā of pāramitā which is quite poor poetry. The Chinese (line 1) introduces the phrase 諸佛母 'mother of all the Buddhas' as an epithet of prajñāpāramitā. This may relate to the time period of the translation. A translation of Aṣṭa from 985 CE gives the title as《佛母出生三法藏般若波羅蜜多經》Fúmǔ-chūshēng-sānfǎcáng-bōrěbōluómìduō-jīng. 佛母出生三法藏 means something like: "The mother of the Buddhas that gives birth to the casket of the three Dharmas"; while 般若波羅蜜多 is the standard transliteration of prajñāpāramitā. And note that the title of Rgs also contains the phrase 佛母 'mother of the Buddhas'. This is an epithet of prajñāpāramitā even in early texts, but seems to have become more important by 10th century in China. 

Line 2 is interpreted in a more positive sense in the Chinese version. The Sanskrit merely notes the characteristics of saṃsāra (misery and grief) while the Chinese insists that the vidyā is able to relieve the suffering. This change is consistent with parallel changes in how the Buddha was seen in the Mahāyāna that I noted in my recent article in the Journal of Buddhist Ethics (21): Escaping the Inescapable: Changes in Buddhist Karma. What we see is the Buddha becoming more like a messiah in his ability to intervene in human affairs to benefit people. In particular in the revised (Mahāyāna) versions of the Śrāmaṇyaphala Sūtra, King Ajātaśatru, who has killed his father, is excused from a lengthy stay in Hell simply by meeting the Buddha. It seems that the idea that a Buddha, or in this case prajñāpāramitā, was unable to intervene was unacceptable to Mahāyānavādins. As was the idea that a Buddha might not return to this world as a saviour despite having won liberation from it.

Line 3 shows minimal variation between the two versions, except that the epithet lokanāthā 'protectors of the world' is exchanged for 佛 = buddhas. Since Faxian was translating in verse he may well have made a change like this for the sake of preserving the metre (metri causa). Chinese verse is restricted not only in the number of characters, but sometimes in the pattern of tones also, though we don't have reliable information about tones from Middle Chinese as far as I know.

Line 4 in Prakrit uses a simple play on words that might have been confusing in Chinese. One who trains in the vidyā becomes a supremely knowledgeable (vaidya) king. Here vaidya derives either from vidyā and means 'of or related to vidyā'; or from veda with much the same meaning. In any case someone who "knows" is "knowledgeable". (Note that "Vaidya" is the surname of one of the editors of the Sanskrit text.) In Chinese however two cognate words like vidyā and vaidya would be represented by the same character: 明 míng. Since this might be confusing, Faxian opts for 師 shī, which less ambiguously (and more concisely) conveys the sense of "mastery" and "expertise". Faxian does however retain 無上 wú shàng as a rendering of an-uttara, both meaning 'none higher' or 'unexcelled'. 

Conclusion

If Conze is correct then this mention of prajñāpāramitā qua vidyā in Rgs may be the original passage. However the chronology is complex. Between the composition of Rgs, the copying of the ms. that recension A is based on, and the Chinese translation ten centuries have passed. These texts are known to change over time. For example the Sanskrit parallel in extant Aṣṭa is much more elaborate than the version in extant Pañcaviṃśati suggesting that the manuscript of Aṣṭa is later, though overall we observe the opposite. In general Pañcaviṃśati is a development of Aṣṭa but apparently different parts of the texts evolved at different rates.

So any differences may be due to differences in the text rather than changes introduced by the translator. For all we know Faxian might be absolutely true to the text he had before him. However there are indications of adaptation to both cultural assumptions and to metric necessity.

Most commentaries on the Heart Sutra project late, synthetic views back onto the text. Thus for the recently published Zen commentaries  (e.g. Red Pine or Mu Soeng) the text is almost a tabula rasa onto which the ideas of Zen are inscribed, drawing on the Sūtra's status for authentication. Conze's commentary, influenced no doubt by D T Suzuki, takes a similar approach. Commentators like Pine and Soeng set out to tell a story about Zen based on the Heart Sutra. They seem unaware of any tension between the story they wish to tell and the text itself; and blithely gloss over any inconsistencies. Neither have any time for Jan Nattier's discovery. Pine talks himself out of having to take it seriously on spurious grounds, while Soeng notes the article and then proceeds as if it were never written. The religious story they wish to tell overrides any inconvenient historical or philological facts. And yet both writers are praised for being "scholarly".

For some scholars this genealogical approach to re-used passages is texts is intrinsically interesting. The re-use of passages and texts is a distinct subject for study in Indology (see Elisa Freschi on academia.org, her blog, and Indian Philosophy Blog). In the case of the Hṛdaya however it also opens up an entirely new way of reading the text (an hermeneutic). We can see that the Heart Sutra is thoroughly rooted in the early Prajñāpāramitā texts and that at the very least we need to allow that the author of the text (working between about 400 CE and 650 CE) had that kind reading in mind. It is possible that we see in the Heart Sutra an epitome of early Mahāyāna/Prajñāpāramitā thought rather than a legitimation of later readings which synthesise many elements of Mahāyāna thought and manifest as new forms of Buddhism like Zen or Gelugpa.

In which case we ought to be looking at the characteristic ideas of the early Prajñāpāramitā texts for the foundation concepts that allow us to understand the Heart Sutra. This exploration is a large part of what I will be doing over the next few years. Clearly there are many continuities with a particular stream of early Buddhist ideas and practice. For example, meditation practices such as the skandha reflections provide continuity. And an obvious discontinuity was the rejection of Abhidharma Realism, particularly the Sarvāstivāda ideas that were expressed in pursuit of a solution to the problem of Action at a Temporal Distance (I've yet to discover a specifically Prajñāpāramitā solution to this problem, though Nāgārjuna's solution may give us hints about what to look for). In any case there is a great deal of research on early Mahāyāna that modern commentators pay lip-service to, but fail to incorporate into their narratives. I'd like to revise the story of the Heart Sutra, particularly in the light of Nattier (1992).

~~oOo~~

Bibliography

Chinese text from the CBETA version of the Taishō Edition of the Chinese Tripiṭaka. 

Falk, Harry and Karashima, Seishi. (2012) A first‐century Prajñāpāramitā manu-script from Gandhāra - parivarta 1 (Texts from the Split Collection 1). ARIRIAB XV, 19-61. Online: https://www.academia.edu/3561115/prajnaparamita-5

KIMURA Takayasu (2010). Pañcaviṃśatisāhasrikā Prajñāpāramitā. Vol. I-1, Tokyo: Sankibo Busshorin 2007. Online: http://fiindolo.sub.uni-goettingen.de/gretil/1_sanskr/4_rellit/buddh/psp_1u.htm [Input by Klaus Wille, Göttingen, April 2010].  

Nattier, Jan. (1992). The Heart Sūtra : a Chinese apocryphal text? Journal of the International Association of Buddhist Studies. Vol. 15 (2), p.153-223. 

Studholme, Alexander (2002). The Origins of oṃ maṇipadme hūṃ: A Study of the Kāraṇḍavyūha Sūtra. State University of New York Press. 

Vaidya, P. L. (1960) Aṣṭasāhasrikā Prajñāpāramitā. The Mithila Institute of Post-Graduate Studies and Research in Sanskrit Learning. Also online: http://www.dsbcproject.org/node/8242 

Yuyama, Akira. (1976) Prajñā-pāramitā-ratna-guṇa-saṃcaya-gāthā (Sanskrit Recension A). Cambridge University Press.