First, I did try to apply Lachman’s principles of textual criticism used in classical text studies. Specifically, I have tried to determine a possible genealogical relationship between the witnesses, draw a stemma, and reconstruct the earliest archetype of the text. However, the task was obstructed by a number of factors.

The first difficulty was in deciding on genealogically significant variants. In classical studies, for constructing a stemma, only variants that cannot be easily altered into other occurring variants are to be used (Salemans 1996). Yet, in Pāḷi texts, a variation in one grapheme can result in a meaningful word (for example ca and ce, satā and sadā, anuttaraṃ and anuttamaṃ) that stands all chances to remain the same in the future, however, chances are that another scribe could by mistake change it to another occurring reading. When such readings and obvious mistakes that would tempt a scribe knowing Pāḷi to correct them are not accounted for, the remaining variations appeared to be too irregular for drawing a stemma.

Another difficulty for establishing the genealogical relationship between the witnesses was caused by at times occurring a large number of variations for a particular variant location. For example, the third pāda of the second verse of the nigamana has eight different readings, and the third pāda of the next verse has ten variant readings.

The biggest obstacle, however, was the amount of cross-contamination between the manuscripts. Scribes likely used more than one manuscript when copying the text since when grouping the manuscripts according to variation for separate variant locations individually the resulting groups are never the same. This observation is confirmed by digital analysis of the tradition using Stemmatology R library.

The library has a function for evaluating the contamination of the tradition. When the PCC.contam function was run on the matrix of ten most meaningful variant locations from the colophon, it revealed a high level of contamination. Out of the ten variant locations eight conflicted. The figure bellow visualizes them.

The nodes represent variant locations and edges, the conflicts. Thus the more edges connect a node to the others, the larger is the number of conflicts. As it can be seen five of the variant locations have five conflicts. Digital stemmatological analysis, therefore, confirmed an impression I got from the analysis of collation graphs, i.e. that the tradition is highly contaminated.

Even though I was not able to arrive at a complete stemma, the analysis of all mistakes showed that two pairs of the manuscripts are closely related. The manuscript C shares many mistakes with the manuscript I (for instance, instead of reading ‘byappathānañhi gatiyo’, they read ‘byāsadhānañhi satiyo’; instead of ‘kaṇhadāsena’, ‘kaṇḍadāpane’; instead of ‘sumatinā’, ‘samutina’; and instead of ‘pālayantu’, ‘pālantu’), suggesting that they may be either copied from the same hyparchetype, or from one another. Another pair of manuscripts that had common mistakes and seemed to be related was K and L. Since these pairs had a common prototype, I used for the edition only the one that had fewer mistakes.

Not being able to arrive at a stemma, taking a route of reconstructing the earliest archetype of the text became impossible. That was the reason for deciding to approach editing eclectically.

I took the second oldest manuscript, I, as the base text primarily because of its age and since it appeared to have fewer orthographic mistakes than the oldest dated manuscript. Yet, I do not provide a faithful transcript of that manuscript as the main text since having a text with mistakes would hardly be useful for a reader. Instead, I substitute the readings of I that appear to be erroneous as to orthography, grammar, or semantics with a variant from (an)other witnesses(es) that complies with modern Pāḷi grammar conventions. Erroneous readings in other witnesses are not recorded in critical apparatus. When a witness disagrees with I only about compounding words, I do not record it, but follow I or the commentaries when they do comment on the words in question and their lemmas mutually agree. Only the variants that give possible meaning are retained, and their spelling is normalised.

At times when there were two or more possible variants for one variant location, I consulted Burmese and Sinhalese printed editions as well as the manuscripts of the Abhidh-av commentaries to see what the commentators were reading, and my choice would follow the lemma in a commentary. On rare occasions when the lemma in the commentaries or in the editions of a commentary differs, it is also registered in the apparatus. The editorial choice in those instances was arbitrary.

As D. Kosambi mentioned in the preface to his edition of the Visuddhimagga, the spelling of some words is always different in Burmese and Singhalese versions of his text (Warren and Kosambi 1999, p.vii). Given that the Visuddhimagga is one of the main sources of the Abhidh-av, that it was written in South India and most likely arrived to Myanmar through Sri Lanka, in this edition, I give preference to Sinhalese spellings the way they are reflected in E^e and confirmed by the table of differences in spelling provided by D. Kosambi (ibid., p.615-16).

Other editorial interventions to be mentioned are punctuation, numbering and titles. The only punctuation mark that is found in the manuscripts is daṇḍa, a vertical line that mostly indicates the second hemistich of a verse or the end of a stanza. The punctuation I use in this edition does not replicate that but uses a system more common for texts printed in Roman script. The main purpose of introducing punctuation is to enhance the readability of the text helping the reader to navigate through it. The only punctuation mark in verses is an inverted comma indicating elision of a vowel. In prose, the full stop is used to indicate the end of longer sentences. When the sentences are very short and are related a comma or a semicolon is used instead. The lists are introduced by a colon and their items are separated by a comma or a semicolon when the list item forms a sentence. Single quotation marks are used to indicate the beginning and the end of a quotation. A dash separates members of a longer compound. The first letter of proper nouns is the only letter capitalised. All verses are numbered. A number in square brackets indicates the page number of the E^e.

Manuscripts do not indicate beginnings of chapters. Instead, they indicate their end (which is a usual phenomena for Pāḷi texts) with a phrase ’iti Abhidhammāvātare [chapter name] nāma [chapter number] paricchedo’. This is, of course, kept in the edition. In addition, however, following modern conventions, beginnings of new chapters are also indicated. To help the reader to navigate through the text headings for a smaller sections are also introduced. Thus whatever headings there are in the text, they were added by ythe editor.

Since digital format does not put on an editor a pressure of ink- and space-saving, I provide the notes in a positive format, i.e. explicitly naming the witnesses that have the same reading as the edition, and not just those that have alternative readings. The variant locations are underlined with a dotted line and accompanying notes appear on hovering them. A note first provides a list of witnesses testifying the reading in the text, which ends with a square bracket. After that, alternative readings with their witnesses or any other additional information is provided. Each witness has its own color as an additional marker, and by clicking on the list of witnesses a display of their description can be called.

The decribed approach to editing results in a text that has a minimalistic critical apparatus. Recording all the variations would be rather a distraction than a help to a reader. I, however, have considered those few, who may like to replicate the editorial process and provide a collation of witnesses with all the variants in a form of graphs produced by CollateX and TRAViz.

In preparing the edition I used CollateX and TRAViz applications, each of them for distinct purposes. The graph generated by TRAViz is very visual and interactive. The inconvenience about using it, though, is the input format and its inability to work with a larger corpus of a text. For this reason, I used it for evaluating the quality of the manuscripts and their relation only. Bellow are the links to a few fragments of the collation.

TRAViz graphs:

• Introduction and Chapter 1 (v. 1-10)
• Chapter 1 (v. 11-20)
• Chapter 1 (v. 21-24)
• Colophon

To prepare the input for obtaining a CollateX graph, on the other hand, is easy, and it can work with large amount of data. Thus I used it as the main collation tool. Bellow are the links to the 'never-ending graphs’ of the collation. Collations of the first chapter and the colophon include the C, J, and L witnesses, that are excluded from the final edition.

CollateX graphs:

• Introduction
• Chapter 1
• Colophon

Sources used for the edition

K

Copied in 1804, this Burmese script palm-leaf manuscript is the oldest dated manuscript I was able to obtain. It is held at the Universities’ Central Library in Yangon, Myanmar, under accession number 3738. The library digitalised this and other manuscripts from their collection at my request and provided me with perfectly readable images. The text comes in a bundle together with its newer commentary, the Abhidhammatthavikāsinī. The leaves are 51$*$6.5 cm in size and usually have ten lines of rather clear writing. The title in the upper right corner of recto sides is Abhidhammāvatarajāko. A total of 51 leaves are numbered in the upper left corner of recto sides beginning with ka and ending with ñe. Even though being the oldest among available to me manuscripts it is abundant with irregularities and mistakes common to most Burmese Pāḷi manuscripts, such as i for ī, ā for o, a for ā, omission of syllables and niggahīta, ṭ for ṭṭh, sometimes di for ti, tta for ttha, and ttha for tta. Immediately after the text come first ten verses of the Namakkārapāḷi, which are followed by a colophon in Burmese stating that the copying of the commentary on abhidhamma was completed on the 1st day after the full moon of the tenth month in 1185 Sakkarāja (1804 CE).

I

Copied in 1845, this Burmese script palm-leaf manuscript is the second oldest dated manuscript that I was able to obtain. It is held at the Universities’ Central Library in Yangon, Myanmar under accession number 3922. The text comes in a bundle together with its two commentaries. The leaves are 48$*$6 cm in size and usually have ten lines of clear round writing. The title in the upper right corner of recto sides is Abhidhammāvatāra. A total of 46 leaves are numbered in the upper left corner of recto sides beginning with ka and ending with ghau. Mistakes are few, common ones being i for ī, the omission of niggahīta, and sometimes mu for pu. The colophon does not come immediately after the text but after the old and the new commentaries. The first one declares the sacredness of letters with a common formula ‘akkharaṃ ekamekañ ca buddharūpasamaṃ siyā’ and expresses an aspiration for the attainment of the threefold wisdom. The second one states that the date of completion of copying was the 2nd day in the half-month of the waxing moon of the fourth month in 1207 Sakkarāja (1845 CE), and expresses an aspiration for the attainment of nibbāna.

E

This Burmese script digitized palm-leaf manuscript from Fragile Palm Leaves Project Works collection provided by Buddhist Digital Resource Center with identification number W1FPL3771 and downloaded from tbrs.org was copied in 1864. The images are of average quality. The text is bundled with three other texts: the Samohavinodanī, the Nāmarūpapariccheda, and the Khemāpakaraṇa. A total of 46 leaves are numbered in the upper left corner of recto sides beginning with da and ending with pā: and usually have 10 lines of somewhat crammed handwriting with frequent corrections. It does not distinguish between i and ī or tta and ṭṭa. The colophon does not come immediately after the text but after the last text of the bundle stating that the copying was completed on the 6th day in the half-month of the waxing moon of the eighth month in 1226 Sakkarāja (1864 CE).

B

This Burmese script digitized palm-leaf manuscript from U Pho Ti Library in Thaton, Mon State, Myanmar, with identification number 761 was copied in 1862. It has been well described in Pruitt, Ousaka and Kasamatsu (2019, p. 363). Most frequent mistakes are ya for ra, bba for ppa and ppha, the omission of niggahita and sometimes entire words. The text is followed by an unusual verse in Pāḷī ‘vamitabbaṃ vamittayaṃ | ramitabbañ ca taṃ buddhaṃ || dhammamaggaṃ namāmataṃ | mahitabbaṃ gaṇaṅ guruṃ ||’ followed by its nissaya, lengthy aspirations and the date of completing copying, which was the 4th day in the half-month of the waxing moon of the first month in 1224 Sakkarāja (1862 CE) (Pruitt, Ousaka and Kasamatsu (2019), however, record the date as 1242).

D

This Burmese script digitized palm-leaf manuscript from Fragile Palm Leaves Project Works collection provided by Buddhist Digital Resource Center with identification number W1FPL2036 (tbrs.org) was copied in 1879. The text is bundled with its new nissaya, which was written at the beginning of the nineteenth century by Salin Sāsanāpain (Calṅasāsanāpuilṅa) Sayadaw. A total of 52 leaves are numbered in the upper left corner of recto sides beginning with ka and ending with ñū and usually have 9 lines of mostly clear handwriting. The title in the upper right corner of recto sides is Abhidhammāvatārapāḷi. Besides such common mistakes as writing i for ī, ta for ṭa, or omitting niggahīta, it sometimes confuses ya and ra. Immediately after the text follows a lengthy colophon that starts with the common Pāḷi formula declaring the sacredness of letters: ‘akkharaṃ ekamekañ ca | buddharūpasamaṃ siyā || tasmā hi paṇḍito paso | likheyya piṭakattayaṃ || ahaṃ vandāmi sabbadā’. Then the scribe first dedicates the merit acquired in Burmese. After, in a form of Pāli nissaya, he expresses an aspiration for the attainment of nibbāna and acquiring various qualities of the great disciples of the Buddha. The exact date of completion of copying the manuscript, which comes in a colophon after the Abhidhammāvatāranissaya, is the 10th day in the half-month of the waxing moon of the tenth month in 1240 Sakkarāja (1879 CE).

F

This Burmese script digitized palm-leaf manuscript from kelasa.org collection (identification number 1918) was copied in 1915. The quality of the pictures is not uniform, some are of higher resolution than others. The text is bundled with ten other abhidhamma texts: the Paramatthavinicchaya, the Khemāpakaraṇa, the Nāmācāradīpaka, the Saccasaṅkhepa, the Rūpavibhāga, the Rūpārūpavibhāga, the Nāmarūpapariccheda, the Abhidhammatthasaṅgaha, the Saṅkhepavaṇṇanā, and the Apheggusāradīpani. A total of 43 leaves are numbered in the upper left corner of recto sides beginning with ka and ending with ghè. They usually have 12 lines of mostly clear handwriting. The title in the upper right corner of recto sides is Abhidhammāvatārapāḷi. Mistakes are not too many, among them short for long vowels, occasional omission of niggahita, and ba for pha. Throughout the first chapter, it spells kriya as kiriya, even in verses where the additional vowel breaks the meter. According to the colophon, which comes immediately after the text, the copying completion date is the 1st day in the half-month of the waning moon of the fourth month in 1277 Sakkarāja (1915 CE). On the right margin of the last leaf of the bundle are stated donors’ names and the town where the copying was done, Dedaye (De:daè).

A

This Burmese script digitized palm-leaf manuscript from U Pho Ti Library in Thaton, Mon State, Myanmar, with identification number 589 is the youngest among the dated manuscripts. It was copied in 1928. It has been described in Pruitt, Ousaka and Kasamatsu (2019, p. 264). Mistakes are only occasional. The title in the upper right corner of recto sides is Abhidhammāvatārapāḷi. Immediately after the text follows the colophon stating that the copying of the manuscript was completed on the 1st day after the full moon of the ninth month in 1290 Sakkarāja (1928 CE). On the left margin of the first and the last leaves of the text are stated donor’s name and the town where the copying was done, Thaton (Sathuṃ).

G

Undated Kham script palm-leaf manuscript from Pāli Manuscript Collection in the Royal Asiatic Society with identification number Ras 34, the digital copy of which available for download from archive.org. It has been described in Filliozat (1999, p. 46). The copying date is not stated, but Filliozat (ibid.) conjectures that it might have been the manuscript that was presented to the Society in 1858 by King Mongkut (Rama IV). Six bundles of 183 leaves of text in total usually have five lines of clear writing, and a page number on the left of recto sides beginning with ka and ending with ttaḥ. The title written on the 1st leaf of each bundle is braḥ pāṭha abhidhammavatāra phuk. Mistakes are rare. Free space of the last leaf is filled with the ‘Iti pi so’ formula up to akāliko. No colophon.

H

Undated Burmese script digitized palm-leaf manuscript from Fragile Palm Leaves Project Works collection provided by Buddhist Digital Resource Center with identification number W1FPL527 (tbrs.org). The text comes in a bundle together with its two commentaries. A total of 43 leaves are numbered in the upper left corner of recto sides beginning with ka and ending with gho and usually have 11 lines of clear handwriting. The title in the upper right corner of recto sides is Abhidhammāvatārapāḷi. Mistakes are only occasional. One colophon in Burmese at the end of the bundle with the dedication of merits.

PTS

This Roman script printed edition of the Abhidh-av published in 1915 was prepared by A. P. Buddhadatta. It is the first text of the first volume of ‘The Buddhadatta’s Manuals’ and occupies 138 pages. Buddhadatta does not describe his sources but does mention that by the end of his work he got hold of a Burmese manuscript and has supplied variant readings from it. Being Sinhalese, the editor most likely used a Sinhalese manuscript or manuscripts. There are some indications that he silently edited the text. For example, when all manuscripts read -vatthu where one would expect plural, Buddhadatta reads -vatthūni (Abhidh-av 9$_{7$), or, when all other witnesses in the verse 28 have an ambiguous cattārirūpisu, which according to the meaning should be resolved as cattāri plus arūpisu, he has an unambiguous cattārarūpisu (Abhidh-av 6₁). The text is accompanied by a thin critical apparatus. After the text, is attached the appendix with references to the Nikāyas that Mrs. Rhys Davids was able to trace, and some parallel passages supplied by A. P. Buddhadatta.

CS

This Burmese script printed edition of the Abhidh-av published in 1962 represents the text recited at the sixth Buddhist council. The text occupies the first 186 pages of the book and is followed by three other works on abhidhamma: the Nāmarūpapariccheda, the Paramatthavinicchaya, and the Saccasaṅkhepa. In footnotes alternative Burmese and Sinhalese variants, and, on very rare occasions, references to Tipiṭaka are provided. The romanized version of the text is available at tipitaka.org, however, without variants and references.

Sources examined but not used for the edition

C

This Burmese script digitized palm-leaf manuscript from Fragile Palm Leaves Project Works collection provided by Buddhist Digital Resource Center with identification number W1FPL640 (tbrs.org) was copied in 1905. It shares mistakes with I and adds more. Thus it must be either a descendant or a lower quality sibling of I. This was the main reason for excluding it from the edition.

J

This Burmese script palm-leaf manuscript from the Universities’ Central Library in Yangon, Myanmar under accession number 3922 was copied in 1913. It contains a large number of mistakes, mostly omitting syllables, words, and phrases. Handwriting is unclear. These were the main reason for excluding it from the edition.

L

This Burmese script palm-leaf manuscript from the Universities’ Central Library in Yangon, Myanmar under accession number 9714 was copied in 1869. The main reason for excluding it from the edition was an overwhelmingly large amount of mistakes and its apparent relationship with K.

Secondary sources

Old commentary

Ṭ^Pb

The electronic edition of the Abhidh-av-pṭ, CSCD Tipiṭaka (Roman).

Ṭ^Pm

This is a Burmese script 43 palm leaves long manuscript of the Abhidh-av-pṭ bundled with I. For its source, physical description and date of copy, see I.

New commentary

Ṭ^Ab

The electronic edition of the Abhidh-av-aṭ, CSCD Tipiṭaka (Roman).

Ṭ^Ac

Two volumes of Sinhalese edition of the Abhidh-av-aṭ, the Abhidhammatthavikāsinī, edited by A.P. Buddhadatta and printed in 1961 by Anula-printers, Colombo.

Ṭ^Am

This is a Burmese script 202 palm leaves long manuscript of the Abhidh-av-aṭ bundled with I. For its source, physical description, and date of copy, see I.

An important factor that should determine any translation is the purpose behind the work. Thus the first question to address here is: why to translate the Abhi-āv?

One possible answer is: to create a useful tool, a commentary for other scholars who are reading the text in its original language. This audience most likely is already familiar with abhidhamma and may be interested in the text as a witness of the development of the abhidhammic thought rather than as a source from which they could learn the subject matter of the book.

Another possible answer is: to make the text accessible to those who cannot read Pāli, but are interested in the subject, and would like to actually learn abhidhamma, that is to say, to use the book as the author has intended it.

The first purpose calls for a source-oriented translation that is meant to be read together with the source. The translator aims at an adequate translation that tries to reproduce the source text in the target language while sticking to the literary norms of the source text as close as possible. The second purpose would require a target-oriented translation that would be able to stand alone and be accessible to a modern educated reader. Here the translator aims at an acceptable translation while complying with the literary norms of the target language.

Rather than choosing between the two approaches, I have decided to adopt both by producing two translations. In this case, a source-oriented translation would be a natural first step in the process of translation. In this step, the meaning of every separate word is accounted for and tried to be represented in the target text. The commentaries and parallel texts are consulted in order to establish the exact meaning of the words, phrases, sentences, and paragraphs. If these meanings do not come through in translation naturally then missing information is added in the footnotes. This approach is susceptible to heavy annotation, which, on the one hand, could be seen by a scholastic audience as a useful source of information, and on the other hand, as a distraction and information overload by an abhidhamma student. Moreover, when a highly technical language of abhidhamma is coupled with syntactical norms of the source language the text is prone to become ‘translationese’ understandable only to a specialist reader. To make the text accessible to a wider readership, I take the next step.

That next step is an expression of higher levels of meaning in the clearest way possible so that the meaning would be easily apprehended and retained by someone learning abhidhamma. This step produces the text that can perform the function of a textbook. This is an approach requiring more creativity. It focuses on the intended meaning of each passage rather than on the meaning of the words. To keep the reader focused on the text when something needs further explanation it is expanded on directly in the text rather than in a footnote.

When translating for an abhidhamma student I always question a possible author’s intention behind a particular stanza or passage and aim for the translation to fulfil that intention. The prose passages usually are rather straight forward and even source-oriented translation is adequate enough for a student-reader. Stanzas, however, seem to have different purposes.

Some of the stanzas laconically summarize the preceding prose part of the text and have a mnemonic function. In those instances, the translation likewise aims to produce a text that would be clear and easily remembered. Other stanzas are also a summary, but of passages from the source texts that Buddhadatta was working with, and their meaning becomes apparent only when the sources are read. In such cases, translation may use such procedures as addition and explication thus becoming wordier than the original. In other stanzas, the author is displaying his poetic abilities employing such poetic figures as alliterations, twinings, similes, etc. On those occasions, translation tries to reproduce those as much as possible.

Thus the translation of stanzas is primarily determined by their individual purpose. The role of form, in this case, is relegated. Whenever a particular function of the text could be satisfactorily fulfilled while retaining the verse-form it was done so. When it was not possible, however, translation was done in prose. This may be considered inconsistent by some, nevertheless, on occasions when prioritising form obscures the meaning and obstructs learning, it seems logical to relegate it. The way translation informs the reader about the form of the original text is numbering. To indicate that the original is written in verse both Burmese and PTS printed editions of the text number their verses. The same was done in the translation.

One could ask why to produce two translations? If translator’s aim is to revive the text, making an old, forgotten abhidhamma book available to students again, then what St. Jerome would call ‘sense-for-sense’, Vinay and Dabernet ‘oblique translation’, and Newmark ‘communicative translation’, i.e. the text produced in the second step would be enough and would serve the purpose. That is true, however, a‘word-for-word’, ‘direct translation’, or ‘semantic translation’ has to be produced before arriving at the student-friendly target text anyway. Of course, one could simply discard it, however, knowing that some do not even consider ‘free’ translation to be a translation and given that these days publishing a text does not have to be on paper, I have decided to retain it and make it available here in a digital form.

The two resulting texts have different purposes and thus require different approaches to reading them. The first one, as already mentioned, is there to help a reader who is struggling with the source text, thus it is expected that the two are read together. Whenever Buddhadatta’s source could be traced the reference is given in a note. When the translation calls for further clarifications they are also given in notes. Thus reading notes is essential for understanding this translation fully.

The second translation can be read on its own. As it is intended for an abhidhamma student who may be or may not be familiar with the subject, the most common technical terms when they first occur are given both in Pāli and English, and to help the student to internalise Pāli terminology for his present and future studies, the Pāli term may be repeated later with the English translation or without it.

Abhidhamma as a subject operates many technical terms and concepts through which it explains the reality at its ultimate level. To follow its explanations one has to understand and remember their meaning. Memorising concepts and most important numbers is a part of the abhidhamma studies that cannot be disregarded even in our digital age when information is always there at the reach of our fingertips. The Abhi-āv, in particular, classifies s in various ways and often explains how different classifications can be related to each other. To be able to follow such explanations a student has to have a sufficient grasp over what is being classified and the classifications overall when the author is connecting them to each other. If a student is not clear about them then there is no way they will be able to understand what the author is explaining. The entire fourth chapter, for example, is dedicated to classifications of consciousness. To understand the exposition a student has to be clear which and how many are the wholesome, unwholesome, resultant, and functional consciousnesses. Without such knowledge a passage like: ‘The rootless are eighteen: fifteen wholesome and unwholesome resultants, and three functional in sense-pleasure sphere’ or like: ‘The two mentioned above have five functions, four functions have the eight, three functions, nine; two, two functions; remaining have one each’ will be meaningless. I would thus encourage a reader who is studying abhidhamma to read the text aloud and several times so that at least some of the information is retained in the mind. Indeed, reading abhidhamma and learning it are two different processes. To facilitate the learning process the second translation is accompanied by margin notes highlighting the topic of discussion and important numbers connected with it.