[Image] Siddhaṃ calligraphy of the 
Lalitavistara Sūtra version of the verse [Updated 30 Aug 2013 with suggestions and corrections from Bhikhhu Ānandajoti (AJ)]

One of the well known archetypal events in the life story of the Buddha is his meeting with Vedic god Brahmā after his Awakening. The episode occurs in Pāli in the Ariyapariyesanā Sutta which is considered by many scholars to contain an archaic version of the life-story which is missing many later elements (see The Buddha's Biography). Here Brahmā is called Brahmā Sahampati 'Lord of All'(?), though elsewhere the epithet Sahampati is dropped.

In the episode, the Buddha, thinking about how he might convey what he has discovered, appears to be reluctant to try to teach it. Brahmā appears to him to ask him to teach, because, though many will not understand, there are some people who will. It is quite an evocative image: the creator god of the Brahmanical religion (of the day) begging the Buddha to teach what we now think of as Buddhism. 
This episode has been studied in detail by a number of scholars, most recently by my colleague Dhīvan Thomas Jones in his article 'Why Did Brahmā Ask the Buddha to Teach?' (2009). It was looking at his article that drew my attention to one particularly interesting verse that recurs in several places. All three versions are in triṣṭubh (P. tuṭṭhubha) metre, which has 11 syllables in 3 measures of 4:3:4.* Those who aren't interested in detailed analysis of grammar and metre can skip to the conclusions which discuss the verse in terms of cladistic thinking.

Aryapariyesanā Sutta. MN 26, PTS i.170 (Ap)
apārutā tesaṃ amatassa dvārā
ye sotavanto pamuñcantu saddhaṃ
vihiṃsasaññī pagunaṃ na bhāsiṃ
dhammaṃ paṇītaṃ manujesu brahme' ti The doors of the deathless are opened for them,
Let those who listen renounce the funeral rites.
Familiar with their vicious minds, I did not speak,
The lofty Dhamma amongst human beings, O Brahma.
Metre

^ - ^ - | - - ^^| - ^ - -
- - ^ - | - ^ - | - ^ - -
^ - ^ - | - ^ ^ | - ^ - -
- - ^ - | - - ^ | - ^ - -

^ = light syllable
- or ^^ = heavy syllable Context tells us that lines cd represent the hesitation to teach and lines ab represent the resolution to teach or as the metaphor has it, to open the doors to the deathless. This verse has counterparts in the Lalitavistara Sūtra and the Mahāvastu, both composed in Buddhist Hybrid Sanskrit and thus giving us an interesting contrast.

Lalitavistara Sūtra (Lv 25.34)
apāvṛtās teṣām amṛtasya dvārā
brahman ti satataṃ ye śrotavantaḥ |
praviśanti śraddhā naviheṭhasaṃjñāḥ
śṛṇvanti dharmaṁ magadheṣu sattvāḥ || The doors the deathless are opened,
Always for those who listen, O Brahma.
Those with faith and peaceful thoughts enter,
The beings of Magadha listen to the doctrine.
Metre

^ - ^ - | - - ^^| - ^ - -
- - ^ ^ | ^ - - | - ^ - -
- ^ - ^ | - - ^^| - ^ - -
- - ^ - | - ^ ^ | - ^ - - Lv has the same triṣṭubh metre as Ap, with minor variations. 

Mahāvastu (Mv iii.319)
apāvṛtaṃ me amṛtasya dvāraṃ
brahmeti bhagavantaṃ ye śrotukāmā
śraddhāṃ pramuṃcantu viheṭhasaṃjñāṃ |  viheṭhasaṃjño praguṇo abhūṣi
dharmo aśuddho magadheṣu pūrvaṃ || The doors to the deathless are opened,
O Brahmā, let those who wish to hear the Blessed One
Give up the funeral rites and harmful thoughts.   Well acquainted with vicious thoughts, unadorned.
Formerly there was an impure Dharma amongst the
    Magadhans
Metre

^ - ^ - | - ^ ^ | - ^ - -
- - ^ ^^| - - - | - ^ - -
- - ^ - | - ^ ^ | - ^ - -

^ - ^ - | - ^ ^ | - ^ - ^
- - ^ - | - ^ ^ | - ^ - - Once again the metre is triṣṭubh, though here extended to five lines. Re the middle measure of line 2 AJ says "is highly unusual, and probably impossible, as three heavies do not appear in the break".

Ap and Lv are more or less identical in line a, taking into account spelling differences between Pāli and Sanskrit. Mv mirrors these two but has the first person instrumental pronoun me instead of the genitive/dative third person plural pronoun tesaṃ/tesāṃ. In Ap and Lv the doors were open "for them" and in Mv "by me". The metric pattern of all three is triṣṭubh, but, whereas in Lv and Ap resolve on heavy syllable as two light (a mṛ and a ma), Mv must take a mṛ as two light syllables. Thus the Mv version fits the metre more naturally that Ap or Lv. In fact if we changed tesaṃ to me in Ap then the metre of the verse would be more symmetrical - lines a and c, and b and d having identical patterns.

In all cases the conjunct consonant dv fails to 'make position' or cause the preceding syllable to be heavy. I'm less sure about Sanskrit metre even than Pāli, but I suspect this is an artefact of the underlying Prakrit in the Buddhist Hybrid Sanskrit. 
In line two Ap and Lv refer to 'those who have ears' (sotavanto/śrotavantaḥ), i.e. those willing to listen, however they diverge quite a bit otherwise. In Lv the statement (marked by 'ti' at the end) is concluded and the Buddha tells Brahma that the way is always (satataṃ) open. This addition creates downstream effects which will be discussed below. Mv changes śrotavantaḥ to śrotukāmā 'those who desire to hear'. Since the Mv poet here has exchanged satataṃ 'always' for bhagavantaṃ, the two light syllables in bhaga resolve as one heavy syllable.
A significant difference creeps into Lv at this point. Where Ap and Mv have pamuñcantu and pramuṃcantu 'they should give up' (in the imperative mood), Lv changes the verb to praviśanti 'they enter' (present indicative) and the meaning of the sentence is changed considerably. What is more, the verb in the plural is in a line with two words in the nominative plural, forcing us to read them as adjectives or predicates of an implied subject 'they': i.e. "those who have faith and not vicious thoughts". Presumably the composer of Lv thought of the faithful entering the doors to the deathless.
In Ap and Mv what should be given up is saddhā/śraddhā. Often we would translate saddhā as 'faith', but here K.R. Norman (2001) has suggested that saddhā refers to the Brahminical funeral rites. These rites, which include being cremated according to specific instructions, are intended to assure the rebirth of the Brahmin in heaven.** Certainly it would be strange if the Buddha were suggesting giving up faith! Ñānamoḷi and Bodhi here have "let those with ears show their faith". They must be reading pamuñcati as "to let loose, give out, emit". The Sanskrit root is √muc - 'to free', and the prefix pra- indicates the forward direct: hence 'let loose'. This is a valid translation also.
The context supports the reading as "renounce the funeral rites" because the deathless or undying (amata/amṛta = Latin im-morta-) is a Vedic idiom. At least in some cases it is precisely the amṛta , i.e. an end to repeated death in saṃsāra that is sought by Brahmins in their funeral rites. This is reinforced because the Buddha is speaking to Brahmā, the creator god of the Brahmins (seen in crude terms anyway). But the ambiguity remains.
Sanskrit viheṭhasaṃjñāṃ is synonymous with Pāli vihiṃsasaññī (though there is a closer Pāli equivalent in viheṭhasaññā) and in the phrase vihiṃsasaññī paguṇaṃ is often rendered as "perceiving trouble". However 'trouble' hardly seems a sufficient translation of vihiṃsa, which means: 'hurting, injuring, cruelty, injury'; whereas viheṭha (from the verb viheṭheti) is 'to be hostile, to oppress, to bring into difficulties, to vex, to annoy, plague, hurt'. In Pāli a saññin is 'one who perceives, a perceiver', however in Lv the compounds ends with -saṃjñāḥ and Mv -saṃjñāṃ neither of which adds the possessive -in ending though it is available in Sanskrit. I would read the Pāli vihiṃsasaññī as 'one whose thoughts are vicious', and the Sanskrit as simply 'thoughts of cruelty'. 
Here Ap and Mv has a word missing from Lv: paguṇa (Skt. praguṇa) 'well acquainted'. In the narrative context we know that the Buddha has been considering whether or not he could teach what he has discovered and some dramatic tension is built up by his first opinion that people won't get it. He is well acquainted with their vicious minds (vihiṃsasaññī paguṇa) thus he concludes it is hopeless to teach them. The lack of this word praguṇa drastically changes the sense of the passage - leaving the poet of Lv with this word viheṭhasaṃjñāṃ to be fitted in somehow.
Where in Ap it seems natural to take vihiṃsasaññī in line c, by including satataṃ in line b, Lv pushes the replacement verb praviśanti into line c. While "entering faith" makes good sense, it requires a further change because the next word is P vihiṃsa-saññī or S. viheṭha-saṃjñā 'vicious thoughts' and if one is 'entering' instead of 'renouncing' then it requires the vicious thoughts to be negated. Hence Lv has na viheṭhasaṃjñā 'not vicious thoughts', or 'peaceful thoughts'. Furthermore Lv has saddhā and viheṭhasaṃjñā in the nominative case when they are in the accusative in both Mv and Ap.
So in Ap saddhā is the patient of the verb pamuñcantu but vihiṃsasaññin forms part of a sub-clause in a different sentence. In Lv na viheṭhasaṃjñāḥ and saddhāḥ are predicates of an implied subject of the substituted verb praviśanti. (Note we should almost certainly read praviśantī to correct the metre. AJ) And in Mv viheṭhasaṃjñāṃ and śraddhaṃ are both patients of pramuṃcantu. Ap seems more natural than either Lv or Mv in this case.
It's possible here that the author of Lv heard pamuñcantu saddhā and thought it could only be understood as 'abandon faith'. And thus emended the verb to praviśanti, and then realised that a further change was required. Holding the changes within a metrically constrained context meant that the changes became even more significant.

Finally where Ap has vihiṃsasaññī pagunaṃ na bhāsiṃ 'familiar with their vicious minds I did not speak'; Mv has 'viheṭhasaṃjño praguṇo abhūṣi' 'familiar with their vicious minds, unadorned'. The difference is between na bhāsim 'I did not speak' (aorist first person singular from √bhāṣ 'to speak') and abhūṣi 'unadorned' (from √bhūṣ 'to adorn'). The word unadorned does not fit here, and the case or conjugation is unclear (if a verb the a- prefix would indicate past-tense rather than negation). Metrically, the final syllable is anceps, i.e. can be heavy or light, and here ṣi is light. This is allowed, however all the other lines end with heavy syllables. The two words bhāsiṃ and bhūṣi are very similar and would have been written in very similar ways in early manuscripts. The Classical Sanskrit past imperfect of √bhāṣ would be abhāṣani or aorist abhāṣi. Given that Mv is written in a Hybridised Sanskrit it's likely here that abhūṣi is a scribal error for Prakrit na bhāsiṃ. Note that Lv makes no mention of the preliminary decision not to teach in this verse.
There are further changes that could be commented on. We could remark on the change from "Magadhans" to "manuja" (men) in Pāli, if that was the direction of change. We might also reflect on the way that the Dharma is worked into the last lines as something the Buddha almost did not speak (P), something the Magadhans listened to (Lv) or something that pre-existed in impure form (Mv). But in terms of the kinds of processes which are at work in the production of variant texts, we have plenty to think about already and I want to offer some concluding thoughts. 

Thinking Cladistically

This verse seems to have existed independently of all the extant written texts. The fact that the metre is triṣṭubh in all three cases above suggests that the verse was originally in this metre, and the similarity of the first line suggests that it must have been in the original, though perhaps with the pronoun me instead of tesaṃ. Probably the third line of Ap (fourth in Mv) was in the original as well.

However, as Dhīvan makes clear in his article what we have here are three versions of a verse that pre-dates all three texts. Lv and Mv are not thought to have evolved from the Pāli. The stemma, or original text, is no longer extant and what we have are three refractions of the original through three different prisms. Where "prism" is a metaphor for culture, language, and the predilections of the author/editor. The study of the differences is interesting, even if it does not contain profound insights into the Dharma. It illustrates an aspect of the nature of the Buddhist literature. 
If we focus on one of the bodies of Buddhist literature (say the Pāli, though I believe the Pāli Canon consolidates multiple lines of textual development) then we can start to see it as 'original' or 'authentic' at the expense of the others. But this is a distortion. While it is true that some texts are more elaborate than others (certainly the hagiography of the Buddha in the Lv and Mv are very much elaborated compared with similar material found in Pāli) we cannot say that one is closer to the original text that others. Mv and Lv are, generally speaking, no less authentic representations of Buddhist thought than Ap is, despite some indications that they might have mangled this particular verse more than Ap. 
The antidote for this hierarchical thinking is to see things in a cladistic way. This way of thinking derives from evolutionary biology. Traditionally, in the wake of (culturally) Christian scientists, we see nature in terms of hierarchies of increasing complexity and perfection. Some of us may see humans, for example, as the top of the evolutionary ladder (as it is sometimes called). Hence also the cult-like interest in apex predators, especially those of a solitary nature such as great white sharks or tigers. However, some evolutionary biologists point our that every living thing we encounter today has been evolving for about 3.5 billion years and thus all organisms are equally evolved. Some of the simplest and seemingly 'primitive' organisms are far better fitted to their ecological niche than we are. When humans as a species are long gone, bacteria that have hardly changed for 3.5 billion years will continue to thrive. The life we encounter can certainly be divided up into categories or 'clades' for the purposes of study, but as we abandon anthropocentrism we can also abandon the false hierarchy we have imposed on species. Indeed Lynn Margulis argued that 'primitive' bacteria are all able to exchange genetic material and thus ought to be considered one species that is massively diversified. Also they always cooperate in colonies of mixed varieties and in many ways are hardly less complex than we are. Genetically speaking the bacteria that hitchhike on our bodies, often playing vital roles in maintaining our bodies, have several orders of magnitude more genes than our own cells. 
We may also say that all forms of Buddhism currently extant are equally developed though some forms have features that others lack. Buddhism thus exists in a variety of clades, but all forms of Buddhism presently being practised are equidistant from the origins of the religion. With texts we know that they were composed at particular times and thus can be formed into rough chronological order, though this is complicated because the act of composition and the act of writing down often occurred several centuries apart. Additionally writing down was not always the end of the changes that occurred in texts. Chronology, even when we can establish it with any certainty, does not necessarily correlate with authenticity or originality.

In the case of the texts studied in this essay, Ap, Lv and Mv were all written down at around the same time. Thus while we can consider them as occupying different clades, the written texts we are looking at all date from roughly the first century BCE. There is also a minority opinion that the Pāli texts date from a much later period. We have Chinese translations of counterparts of these particular texts dating from the second century CE, so it's possible that the written texts were not created until shortly before the transmission to China (possibly for this express purpose?)
Compared with a large body of literature the idea that such picayune details as I have examined here, in a single verse, are important can seem unconvincing. We may want to take in the majestic sweep of the canonical narratives of Buddhism and forget about the minutia. However unless we understand that our the texts, and the very ideas that are contained in them, are subject to various kinds of change over time, and some quite mundane changes at that, then we might make the mistake of seeing this literature in idealistic terms. Typically religious people come to see their texts as eternal or infallible. But texts are never eternal or infallible. They are cultural productions with all the limitations of their human authors and transmitters (including editors and commentators). 
As religieux we might not want to admit to the humanity of our texts. Too many Buddhists want to see the texts as "the Word of the Buddha" in the same way that fundamentalist Christians see the Bible as 'the Word of God'. But these texts are not the word of the Buddha. They are the words of Buddhists. Good words in many cases, beautiful and inspiring. Though in other cases confused, obscure and dull. Perhaps these words were inspired by stories of meetings with the Buddha that were preserved for centuries by devotees. Even so they are third hand at best.

This is not to say that we should not be inspired by the idea that the doors to the deathless are open. I recall the hairs standing up on the back of my neck and a moment awe when I first heard these words spoken aloud (and interestingly from an historical perspective I heard them from someone called Aśvajit). As far as I can see, if we live as though there were 'doors to the deathless' and that they are open to anyone who will harken to the ideals of Buddhism, then we might live well. The way to the deathless is through practising the Buddhist virtues of generosity, kindness, awareness, etc. And if the deathless is just a fairy story, then at least we can be sure that practising these virtues makes the world a better place. We need not become fundamentalists or indulge in the triumphalism that so often disfigures Buddhist discourse in order to practice these virtues, but we will need inspiration and perseverance. And it is often from telling, and listening to, stories that humans derive inspiration and find the courage to persevere.

~~oOo~~

Note

* My thanks to Bhikkhu Anandajyoti for his work on Pāḷi metres and for some pointers on the meter of this verse in correspondence. 
** In fact I over simplify her for brevity's sake. The goals of the rites change over time and summarising them would take too long, and not be of much relevance to the main point.


Bibliography
Anandajyoti. (2006) An Outline of the Metres in the Pāḷi Canon. [Ver. 3.5] Online: http://www.ancient-buddhist-texts.de/Textual-Studies/Outline/index.htm Jones, D. T. (2009) Why did Brahma ask the Buddha to teach? Buddhist Studies Review 26.1 2009: 85–102. Mahāvastu-Avadana. Based on the ed. by Émile Senart, 3 vols., Paris 1882-1897. Online: GRETIL. Norman, K. R. (2001). The Group of Discourses. Oxford: Pali Text Society. The Pāḷi Tipiṭaka (based on the Chaṭṭha Saṅgāyana CD published by the Vipassana Research Institute). Vaidya, P. L. (ed). Lalitavistara. Darbhanga, The Mithila Institute of Post-Graduate Studies and Research in Sanskrit Learning: 1958. Online: Digital Sanskrit Buddhist Canon.