Introduction

This Translation

The Udānavarga, with its 33 chapters and (in this edition) As will be seen in the commentary to the text, there are a number of verses I omitted that appeared in Bernhard’s edition, either because no akṣara at all were visible, or too few to add anything substantial in the translation. These omissions are noted as they occur. 1061 verses (including irreconcilable variants), is by far the largest extant collection of Dharmapada-type verses. It vastly outnumbers the other major collections that we have now. By comparison, the Patna Dhammapada has 22 chapters with 414 verses, and the Pāḷi has 26 chapters with 423. Brough estimated that there were approximately 540 verses Brough, pg 17. in the partly lost Gāndhārī Dharmapada, of which only some 350 are available to us now, the rest having been lost. There may have been as many of 27-28 chapters in that text. Brough, pg 14.

This large collection is attributed to Ven. Dharmatrāta in the Chinese and Tibetan traditions. Dharmatrāta seems to have written a commentary on the collection, which does not appear to be extant now. He was known as an Abhidharma master who lived around the 2nd century CE, He is sometimes cited in the Mahāvibhāṣa. and belonged to the Dārṣtāntika branch of the Sarvāstivāda. While the attribution is traditional and persistent, modern scholarship often regards the Udānavarga as a composite, evolving anthology rather than the work of a single author. The association with Dharmatrāta may reflect his role as a redactor, commentator, or simply lend authority to a pre-existing collection.

Both the Tibetan and the Chinese texts have been translated previously; the Tibetan was translated by Rockhill, William Woodville, trans. (1883). “Udānavarga: a collection of verses from the Buddhist canon compiled by Dharmatrāta being the Northern Buddhist version of Dhammapada, transl. from the Tibetan of the Bkah-hgyur, with notes and extracts from the commentary of Pradjnāvarman.” Prof. Charles Willeman has translated “The Chinese Udānavarga, a collection of important odes of the law Fa chi yao sung ching”.

I first worked on the Udānavarga while preparing A Comparative Edition of the Dhammapada in 2004. Later I published the text along with two studies in 2006. After completing my translation of the Patna Dhammapada in 2017, I had intended to make a translation of the Udānavarga. That intention was stalled when I learned that my good friend and colleague at the International Buddhist College, Prof. Lobsang Jamspal, was undertaking a translation from the Tibetan, referenced against the Sanskrit. Tragically, his work was never completed, and he passed away in 2024, and so I have taken up and now completed that work.

This is the first complete translation of the Sanskritised Prakrit (hereafter Sanskrit) text as far as I am aware of into English, although Peter Skilling has translated the first chapter (Anityavarga) previously. It was reproduced in his collection Buddha’s Words for Tough Times. Unfortunately the translation of that chapter was not available to me.

Here I have collected and translated all the Pāḷi parallels which had at least a pādayuga in common. They span a wide variety of texts in Pāḷi, besides the Udāna (nearly all the Udānas, including the prose Udānas, have verse parallels giving a total of 101) and Dhammapada (451 parallels), we also find verses from the following collections: Aṅguttaranikāya (10), Dīghanikāya (1), Dhammapadaṭṭhakathā (1), Itivuttaka (32), Jātaka (25), Majjhimanikāya (8), Nettippakaraṇa (5), Saṁyuttanikāya (47), Suttanipāta (48), Theragāthā (43) and Therīgāthā (1). I am not listing all the parallels in Pāḷi, just one, to show that it exists. There are also sometimes parallels in Sansritised Prākrit or Gandhārī, but they are not noted here. If no parallel is noted it means I am not aware of a parallel in Pāḷi.

To accurately highlight the parallels and, crucially, the often-subtle variations between the Sanskrit and Pāli recensions, I have adhered strictly to the grammar of each version. This granular, comparative approach allows us to see not just the shared teachings, but the distinct lexical, metrical, and doctrinal choices made within each tradition. Having said that, as always, I have tried to keep everything as readable as possible. I have used italics to highlight the differences between texts, as in the following (6.8 + SN 1.23:2):

And as can be seen from this example the Sanskrit is coloured in blue (text) and red (translation); while the Pāḷi is in purple (text) and green (translation), which serves to make each easily identifiable.

It has sometimes been possible to adhere fairly closely to the Sanskrit word order (1.35):

And again (1.36):

But many times the need for compression and ellipsis in the text has made this impossible. For example (17.7): na te taranti saritāṁ, tīrṇā medhāvino janāḥ. If we use word order as a guide, we would get a meaning: they are not (na) crossing the waters, the intelligent people have crossed over; but the context of the verse makes it clear that what it actually means is almost the opposite (17.7):

Usually, I have translated pādayuga by pādayuga in this translation, but very occasionally it is not possible. Here, the subject of the verse, the bones (asthīni) only occurs in the second pādayuga, whereas the demonstrative pronoun which points to it, is in the first pādayuga. As we can’t make this work fluently in English we have to take the two lines together (1.5).

I have added a grammatical analysis to each of the verses so that the student can see why they mean what they mean. I have also segmented the analysis, so that is possible to see the components of the words, and especially the compounds.

In the analysis of the nouns and pronouns I list gender, declension, number, and where necessary the type of pronoun and whether it is a past or present particple. With the verbs, I give person, number and tense, where relevant. I also identify absolutives and indeclinables. As an example (8.1):

A-bhūta-vādī (m. nom. sg.) the false-speaker; narakān (m. acc. pl.) purgatory; upaiti (3rd. sg. pres.) undergoes; yaḥ (m. nom. sg. rel. pron.) one who; ca (indec.) and; api (indec.) also; anyat (n. acc. sg.) other; ācarati (3rd. sg. pres.) performs; iha (indec.) here; karma (n. acc. coll. sg.) deeds; ubhau (m. nom. num. du.) both; hi (indec.) indeed; tau (m. dem. pron. nom. du.) those two; pretya (abs.) after death; samau (m. nom. dual) the same; niruktau (m. nom. du. pp.) are declared; nihīna-dharmau (m. nom. du.) of base things; manujau (m. nom. du.) humans; paratra (indec.) in the next world.

I have added notes to the translation to explain difficult grammatical features that appear in the text, and while I provide the Pāli commentarial exegesis to illuminate that tradition’s understanding, I have avoided importing it into the translation of the Sanskrit verses. This separation respects the distinct exegetical traditions; the Theravādin commentaries emerged in a different historical and geographical context than the Sarvāstivādin milieu in which the Sanskrit Udānavarga was compiled and transmitted.

Types of Verses

As mentioned I have previously analysed the metres (chandas) of the Udānavarga, and I initially had an idea to analyse the rhetoric (alaṅkāra) in this translation, but after looking into this I decided that the classical categories, which grew up around classical and court poetry, did not seem to apply very well. I have therefore looked at the verses afresh according to their actual form and function, and divided them into the following four categories.

Simple Statements (stm.)

This is by far the common form that the verses take, and I count 794 of these verses having this form. These verses constitute the doctrinal bedrock. Their power lies not in commanding action but in reframing perception. By stating “this is how things are” (yathābhūtaṁ), they seek to overwrite the learner’s innate, mistaken assumptions about permanence, pleasure, and self.

The ethical imperative is often implicit because, in the Buddhist worldview, wisdom (prajñā) naturally and necessarily leads to ethical reorientation (śīla). They present an impersonal, almost factual diagnosis of the human condition, leaving the emotional and ethical response to arise organically in the discerning student. This category is an engine of wisdom.

A verse may say for instance (Ud 1.3):

The implication is that, if you see that this is the truth, you should indeed strive to still conditioned things.

Or, again (2.6):

Which clearly implies that we should be unconcerned, wander forth, and abandon sense desires.

Others of this kind, do not have an implied ethical imperative about them, but simply describe reality, helping us to reorientate our view of the world (1.6):

or again (1.7):

Although these types of verses cannot be said to have an imperative about them, still, I think, we can recognise the truth of what is being said, and that these things are important to understand, from a practice point of view, and they may give rise to saṁvega, a sense of urgency, which is a key emotion in the spiritual life, and may work in some cases precisely by withholding an explicit moral, and presenting a stark reality.

Admonitory Verses (adm.)

These are more direct verses, saying that we should (optative mood) or must (imperative mood) do something, or that something is to be done (gerundive mood). They occur much less than the simple statement type of verse, I count 177 such verses in this collection. These are not arbitrary commands but prescriptions that logically follow from the truths declared in the simple statements. They are the training rules (śikṣā) that mould behavior and mind in accordance with reality. We have to look at these carefully, because although the imperative is always admonitory in mood, the same cannot be said for the optative or the gerundive.

In the optative mood we find, for instance (3.18):

This could also be classified as a simple statement (craving is the source of suffering), but I have not counted it as such, in order to keep the main types separate. Therefore if a verse contains an admonition, it is simply classified as admonitory alone. The same applies to the other categories also.

As mentioned above not all optatives are admonitory in function. Some are simple statements, such as (3.17):

The imperative mood seems to stress the necessity of action in a much stronger form, often signaling an escalation in urgency (saṁvega), suitable for rousing the negligent (4.33-34):

Again, because of the middle two lines, this could be classed as a simple statement, but the admonitions at the beginning and the end of the verses override that classification here.

The gerundive mood can also have an advisory tone to it, as in the following (12.9):

Another case may be cited (31.2):

As with the optative, the gerundive does not necessarily have an imperative sense though. The following is a simple statement (4.19):

Rhetorical Questioning (rh.q.)

This is an important device which is used on 48 occasions, and is a profoundly effective pedagogical tool. Unlike a statement, which can be passively received, a rhetorical question actively engages the listener’s mind, forcing them to internally search for—and thereby own—the answer. The form often breaks a didactic monologue, creating a moment of self-interrogation. Sometimes we simply get a list of questions (1.4):

The questions “Why this laughter?” and “Do you not seek the light?” are not requests for information but are accusations and provocations. They mirror the function of a teacher challenging a student. Of course, this again has ethical implications, suggesting that we should, indeed, seek for the light.

Note that sometimes, like the optative, rhetorical questions are combined with simple statements, whereby a statement of fact is made, and then a question is raised (5.1):

These are classified as rhetorical questions, and not as simple statements.

Declarative Verses (decl.)

Some of the verses in this collection are declaratives from a first person standpoint. First-person declarations, especially those attributed to the Buddha or an arhat, serve a triple function:

For instance, there is a simple declaration (2.1):

This is definitely aspirational, not indicating completion, but knowing how to get there. These may be combined with other categories—such as with rhetorical questionings—the Buddha says (21.1):

These verses I classify as declarative, even though they culminate with a question, because the verse serves to establish the speaker’s credentials, grounding the entire text’s authority in lived experience rather than abstract philosophy.

Many of these sorts of statements are attributed directly to the Buddha, which are spoken in the singular number; but they may also be spoken in the plural, and be attributed to Arahats (30.43):

The plural declarations expand this from the singular Buddha to the enlightened community (Āryāsaṅgha), showing the path’s repeatable success.

These are the main types of verses that I have identified in this collection, and while these categories are useful for analytical clarity, we must note that the verses often blend modes. This very blending mirrors the integrated nature of the Buddhist path: insight (stm.) inspires effort (adm.), which is guided by self-inquiry (rh.q.), culminating in a transformed state of being (decl.).

Together, these forms create a complete pedagogical cycle, designed not just to inform but to transform the person who engages with them. This framework, I feel, could be used for analysing other didactic verse collections, such as the Dhammapada or the Gāthā literature generally, where form is inextricably linked to soteriological function.

I have made an exclusive categorisation of these four, so that they are classified as one or the other, but not both. There is a hiearchy involved, first declarative types, second rhetorical questions, third admonitory, and then simple statements.

Figures of Speech

What was said just above does not apply to the figures of speech which I now take up for examination. The two main categories I have identified here are similes and analogies. As these are figures of speech, they do not clash with the above types, and so the verses may be classified as both simple statements, etc. and containing similes or analogies, as many of them are.

Similes (sim.)

Similes are used in around 250 of the verses, their usefulness in verse lies in how they can illustrate the truths being announced. Indeed they seem, in very few words, to conjure up in the mind a memorable picture, forming an association, which may even come to mind, when the listener sees various things in the course of their everyday life. Their prevalence underscores a core teaching method of the Buddha: truth is best grasped not through doctrine alone, but through resonant, visualisable images that are easy to retain in the memory and re-contextualise everyday perception.

Sometimes this is done with a figure drawn from the natural world. For instance a ripe fruit falling from its stalk must have been a very familiar scene in the Buddha’s day, and the listener is being trained here to see the Dharma in the world around them: every ripe fruit becomes a memento mori. (1.10):

The figures do not need to be drawn from nature, but sometimes well-known objects in industry can be used to illustrate a point (2.10):

Sometimes more than one simile is used in a verse. Here we have two very different pictures presented to us, of milk, which illustrates what we might call karmic latency, because it is not usually immediate in effect, but gradually somes about; and of fire, showing that deeds done can be like a hidden, but equally dangerous fire, covered over, for now by ashes (9.17):

Sometimes the simile is quite complex, almost like a small story:

The complex, narrative simile in the verse above functions as an archetypal map: it charts the journey from effort (pramāda) to elevated insight (prajñāprāsāda), rendering the internal, spiritual ascent as a vivid, external drama. The embedded metaphor (the palace of wisdom) shows that these figures are fluid, working in concert to build understanding.

For collections of similes see especially the chapters on Water (17), Flowers (18), and Horses (19), but note that they occur throughout the text.

Analogy (ana.)

Similar to similes are analogies, and I find analogies in over 60 of the verses in this collection. But whereas similes have a word of comparison (yathā, iva, vā, -vad), in an analogy these sorts of words are missing, and you have to draw your own conclusions, which might even engage the listener more actively.

To illustrate this the following doesn’t say transmigration is long, just as, or like the night is long, or the league is long, you are left to draw that conclusion for yourself (1.19):

The above verse refers to something familiar from anyone’s experience of sleep or walking, which can then be applied to the chief teaching: that transmigration can be very long indeed.

Analogies, like similes, may be drawn from nature also, note how these function something like diagnostic models of behaviour. The farmer knows that fields are ruined by weeds, and then can apply that to how people are similarly ruined by passion, hatred, etc. We also see how the moral lesson of giving to the virtuous is encouraged here, because that will have great fruit (another obvious metaphor) (16.16):

Sometimes there is a thin line between analogy and metaphor, as in the following verse (4.5):

or in the fragrance of the good which pervades all directions (6.16):

Sometimes analogies can be quite opaque, as in the following (6.14):

The commentary on the Pāḷi parallel to this verse says that it is about the covering up and revealing in confession of offences against the monastic code. This shows how the verse could set up an explanatory situation, where the recitor would proceed to use the commentary to open up the meaning.

Collectively, these figures transform the Udānavarga from a simple treatise into a reconfiguring of perception. They equip the learner not just with statements to believe, or advice to follow, but with a new way to see the world, and to always be immersed in Dharma. A smith at his forge, a covered vessel, a field overrun—these cease to be merely themselves and become constant reminders of the workings of the mind, karma, and ethical cultivation.

Metaphors

We have mentioned metaphors above, and while similes and analogies illustrate truths through explicit or implicit comparison, metaphors operate by a more direct form of identification. For instance, they do not say “the disciplined mind is like an island”; they say the wise one “makes an island” (4.5). This linguistic fusion enlivens the verse with exceptional concision, collapsing the distance between the concrete image and the abstract principle. The metaphor does not just describe; it reconstitutes reality in doctrinal terms.

I have identified similes and analogies alongside the verse numbers. I have not done this with metaphors, as they rarely govern the meaning of the verse, as the former do. But here I will list some of them, to highlight them and give an idea of how they work:

The palace of wisdom; the army of death; the root of misery, Death’s bait; the fragrance of virtue; his speech is sweet; one crosses the flood; the darts of ignorance; the cutting of thorns of craving; the weapon of wisdom; has great fruit; is established on the (far) shore; cutting off Māra’s flowers; the drum of the Deathless; leading out of the forest; the taste of solitude, etc., etc. These serve to enliven the verse, in a way that is similar to similes and analogies, but is even more concise.

The metaphors therefore are not mere ornaments. They form a coherent vocabulary that maps the spiritual landscape:

Whereas a simile might ask the listener to contemplate a relationship, a metaphor seeks to change perception instantly. Hearing of “the drum of the Deathless,” the mind does not compare; it is presented with a new, potent object that is resonant and calling for attention. These metaphors, woven throughout the text, build a dense network of associative meaning, making the teachings cohesive and memorable. They are the linguistic embodiment of the Buddha’s skill-in-means (upāya-kauśalya), using the world’s imagery to point to the way beyond it.

Notice how figures of journey and refuge spatialize the path from suffering to liberation: Crossing the flood, reaching the far shore, leading out of the forest. Conflict and protection function in a similar way: The army of Death, Māra’s flowers, the weapon of wisdom, the drum of the Deathless frame the spiritual life as a struggle, arming the practitioner with specific, personified forces. Then we have what must have been very familiar to the listener, agriculture and sensation: The root of misery, great fruit, the fragrance of virtue, sweet speech, the taste of solitude translate ethical and meditative qualities into immediate, sensory experiences, suggesting their natural, cultivable, and palpable qualities.

Concision and Ellipsis

One of the striking characteristics of the verse texts is the concision with which things are stated. Most of the verses are in the Śloka metre, giving 32 syllables if the verse is—as it normally is—made up of four lines; or 48 syllables if it is an extended six-line verse. The other popular metrical form is the Triṣṭubh, which is usually four lines of eleven syllables each. This usually gives somewhere between ten and twenty words within which to express a thought sequence.

This pressure for economy gives rise to a dense, elliptical style where grammatical elements must often be supplied by the reader. Understanding this is essential for correctly interpreting the doctrine being conveyed. The verse often provides a skeleton of meaning which the listener, trained in the Dharma, is expected to flesh out.

This sometimes leads to a very condensed form of expression, and in many of these verses we find that words that would be stated twice in prose, are only written once, and must be extrapolated to fill in meaning.

Negatives

As an example of how this works with negatives, we can look at this verse, where we first get a straightforward statement, and then na (not) is applied to two verbs (26.2):

That is very straightforward, I think. A more complex example if where there are two negatives, that are applied three times, as in the second line of this verse (28.30):

If we don’t apply the last of the negatives twice we end up with doctrinal confusion, and it would say: one should not (na) take comfort thinking, it has a result. There is no comfort of course in realising your bad deeds have a result in the end.

Here the negative has to be applied to both the past participle (a noun) and the optative (a verb) (8.9):

We should note here that because of concision word order is often not a guide to meaning, or can actually confuse meaning, if we are not aware of this fact. For example, the negative at the beginning of the second pādayuga affects the verb at the end of it (22.17.i):

We saw above that a single negative might affect both a noun and a verb. Here the negative affects both the verb (caret) and the manner (dharmeṇa) (13.7):

Verbs

Sometimes it is not negatives that are applied twice, but verbs. This is especially so when similes are being used (25.13):

Note that in the last line the negative is again applied twice.

Another example (4.10):

It isn’t always necessary, or useful, to reduplicate the verb in a simile, as an example, this makes perfect sense as it is (1.18):

Here, in English, as in the Sanskrit, we do not need to repeat the verb wastes away for the meaning to be complete and understandable.

Another case is at 18.1c-d:

We will look at this verse again below.

An interesting case is this one at 2.10, where both the verb nirdhamen (should remove), and the noun malam (impurities) must be understood twice:

Notice, by the way, the nuanced difference between should remove and would remove in this verse.

Nouns and Pronouns

Nouns can also be reduplicated in meaning (22.3).

We also see the same phenomena with pronouns, as in (22.12.ii):

This elliptical style is not a flaw of the gāthā style but a feature. It demands active engagement, turning the listener from a passive recipient into a collaborator who must unpack the verse, and supply what is missing. This very process of unpacking—supplying the missing negatives, verbs, and nouns—ensures the teaching sinks in more deeply. The text gives a concise statement, which is sometimes like a puzzle, and the listener must diligently work it out and apply it to the complexities of mind and world.

Unstated Agent

Another kind of ellipsis is where the actual agent of the verse goes unstated. Compare (3.15):

We can see from the above that when the specific agent is left out, then it has the effect of universalising the teaching. This isn’t for one person, but applies to all. Another example (3.2):

Omission of the Copula

It is very common in both Sanskrit and Pāḷi verse texts that the copula (bhavati/hoti/asti, etc.) is omitted, and has to be understood from the context. These occur too frequently to be noticed in each case, so I just give a few examples here. For instance (4.15ab):

Or this (2.5):

or again (9.2):

We could also interpret this as an implied optative:

Other tenses may be involved also, compare (7.5c-d) where the future tense is implied:

Collective Neuter Singulars

A recurrent stylistic feature of the gāthā form, closely tied to its concision, is the use of the neuter singular noun in a collective sense. This metrical device allows a single, often shorter, singular form to represent a plural or abstract collective concept, satisfying syllable constraints while maintaining semantic breadth. This construction is standard for abstract or mass nouns denoting things that are often plural or aggregate in reality: karmam (deeds), dhanam (wealth), malam (impurities), pāpam, puṇyam (bad and good acts, or demerit and merit), sukham, duḥkham (pleasures and pains, or happiness and suffering).

For example (20.1):

Here saṁyojanaṁ is a singular, but sarvaṁ (all) ensures its meaning as a plural.

Again here (2.5):

The variety of different kinds of bonds ensures that bandhanam is a collective singular.

And this (1.23):

Where karma (sg.) is being used to indicate the collection of deeds one has built up, as one doesn’t go to the next life on the basis on one deed.

The frequent use of the neuter singular with a collective sense (e.g., karmam, dhanam, palitam, puṇyam, pāpam, sukham, duḥkham, pramāṇam, lābham, viśeṣam, malam) is a standard feature of the gāthā style, serving the metrical need for concision where a plural form would be more syllabically cumbersome (18.1):

Where the commentary to the Pāḷi parallel specifically says Yamalokaṁ (sg.) means the fourfold lower world; sadevakaṁ means the human world together with the Deva worlds; and that the Dhammapadaṁ is comprised of the 37 things on the side of Awakening. The commentary doesn’t say that pupphaṁ specifically means flowers (pl.), but it does seem reasonable to translate it as such.

Compositional Techniques and Stylistic Features

Beyond its doctrinal content, the Udānavarga exhibits distinct compositional patterns and stylistic traits that reflect its origins as a crafted, memorizable anthology. These features—from large-scale structural repetitions to minute phonetic adjustments—offer insight into the methods of its compilers and the oral milieu from which it emerged.

Keyword Substitution

A fundamental compositional technique in the Udānavarga is the generation of multiple verses through the systematic substitution of key doctrinal terms. Standard sequences—most notably involving rāgam (passion), dveṣam (hatred), moham (delusion), mānam (conceit), lobham (greed), and tṛṣṇā (craving)—create thematic clusters (e.g., 705.i–710; 711.i–716). This method suggests a pedagogical intent, reinforcing core concepts through patterned repetition. Notably, the Pāli parallels often contain only the seed verse, indicating that the Udānavarga’s compilers may have elaborated a simpler original into a more exhaustive doctrinal list.

Systematic Variation and Chapter Construction

This technique of variation is employed on a larger scale to structure entire chapters. The Peyālavarga (Ch. 33) is built from intricate variations on just four core verses. Similarly, the Ātmavarga (Ch. 23) presents 14 successive variations (23.12–26) on a single seed verse (23.11), each exploring a different implication of self-protection:

with a different last quarter each time.

We also have the Uraga verses in this collection, separated into two sequences, the first 18.21–18.21F is another example of the rāgam, dveṣam, etc. sequence discussed above; the second runs from 32.55–32-80 (26 verses). In the latter the final pādayuga is the same each time, with differing opening lines. This makes the total number of Uraga verses 33, almost double the 17 verses in the Pāḷi, where it occurs, not in a verse collection, but as the opening chapter of the Suttanipāta.

Recombination

Another way in which verses could be assembled, from what were probably existing verses is through recombination of parts of a verse to form a new verse. Compare this verse (4.27):

Here we have parts of two Dhammapada verses (Dhp 31ab + Dhp 327c-d).

The following verse is also made up of parts from other verses, but in this case, they are from widely differing collections, being assembled from Dhp 31ab + Thag 2c-d (4.28):

Note that Dhp 31 is paralleled in full in verse 4.29, and the Theragāthā verse finds a full parallel at 28.8.

There are many of these recombinations of parts—usually of pādayugas—throughout the text.

Fillers and Particles

Metrical Fillers

The strict demands of syllabic metre (chandas) sometimes forced composers to insert semantically empty particles to complete a line. Over 160 such instances occur in this text. Words like tu, hi, or ca often serve this purely metrical function, as in 18.12 where hi and tu add no discernible meaning but fulfill the required syllable count:

Vowel Separators

In Pāḷi all words end in either a vowel or niggahīta (ṁ), and there is no restriction of vowels meeting at the ends and beginnings of words. So it is perfectly acceptable to write:

Where apatthāni ends in a vowel and alāpūne begins with one; and kāpotakāni ends in a vowel and aṭṭhīni begins with one.

In classical Sanskrit, a word ending in a vowel typically cannot be followed directly by a word beginning with a vowel (hiatus); they must coalesce or be separated. While the Udānavarga shows 146 instances of allowed hiatus (a sign of its Middle Indo-Aryan substrate), it more frequently inserts particles like hy (181x) and tv (66x) to bridge the gap. This reflects a conscious, though inconsistent, effort to conform to Classical Sanskrit phonotactics that may reflect either incomplete Sanskritisation, or possibly layers of composition in the text.

Euphonic Particles

A stricter and more revealing rule governs the nasalised vowel anusvāra (ṁ). In the Udānavarga, an anusvāra is almost never followed directly by a vowel—a restriction absent in Pāli. To avoid this, particles like hy (61x) and tv (18x) are again inserted. The near-total consistency of this rule indicates it was a core compositional principle, more rigorously enforced than the rule against simple vowel hiatus. This phonetic shaping points to a developed, scholastic environment concerned with the precise sonic texture of the recited verse.

There are some very rare exceptions, but I only find two, one at 13.6 (vivekaṁ anubṛṁhayet), and another in the Uddāna, which is almost certainly later than the main text. This shows that this rule was applied much more strictly than the need to separate vowels was.

Together, these characteristics—formulaic repetition, systematic variation, metrical filling, and phonetic refinement—reveal the Udānavarga as a carefully engineered textual corpus. Its form served the dual purpose of ensuring mnemonic stability for transmission and creating a ritually potent, phonetically polished artifact for recitation.

About the Text

The Sanskrit text of the Udānavarga presented here is based on the critical edition by Franz Bernhard (Udānavarga, Band I, Abhandlungen der Akademie der Wissenschaften in Göttingen, Philologisch-Historische Klasse, Dritte Folge, Nr. 54, Vandenhoeck & Ruprecht: Göttingen, 1965). I am grateful to Dr. Bernhard’s sister and literary heir, Mrs. Sobotzik, for her kind permission to reproduce this edition for the present translation.

To ensure consistency with the digital presentation standards of this website and to enhance readability, I have introduced the following formal modifications to Bernhard’s edition. These changes affect presentation only, not the substance of the text:

1. Uncertain readings are marked with red coloured text.
2. Lacunae are marked by square brackets, not by a hyphen (upa[ ]nāni; not upa – nāni).
3. Verses which had 2 irreconcilable versions have been given separate numbering (as 98.i, 98.ii), rather than printed side by side.
4. The numbering of the chapter titles has been changed from Roman to Arabic numerals (e.g. 33: Brāhmaṇavarga, not XXXIII: Brāhmaṇavarga), which is the way the numbers appear at the end of the chapters in the printed text itself.

In a handful of instances, my interpretation of the context or assessment of variant readings has led me to adopt a reading different from Bernhard’s constituted text. The following five emendations have been made for the purposes of this translation. A full discussion of the rationale for each is provided in the annotation to the relevant verse:

1.9: Reading Garbh(e hy) eke (with Ms. EU1) instead of Garbha eke.
3.9: Preferring (pra)vardhante to the manuscript reading nivardhante.
27.30: Adopting paśyat(h)āyāntaṁ to supply the needed imperative form, rather than paśyata.
30.8: Writing susaṁvidhāne, na (’)saṁvidhānam, marking the ellipsis (’) omitted by Bernhard but required by the sense.
33.1: Emending notkuṭukaprahāṇaṁ to notkuṭukapra(d)hāṇaṁ for semantic reasons, despite manuscript support for the former.

All of these readings are marked—as here—with brackets, so that they easily identifiable.

Acknowledgements

While making this translation I extensively used my own previous translations of verse texts, including the Dhammapada, the Patna Dhammapada, Udāna, Jātaka and Buddhanīti Saṅgaho among others, which formed a basis for this translation. I also translated many verses for the first time.

I discussed each translation with DeepSeek AI, which has very good knowledge of Pāḷi, Sanskritised Prakrit, as well as classical Sanskrit (accessed from July–December 2025). This often led to corrections and emendations, but I also sometimes had to hold my ground and show it why I believed my translation was correct.

I used DeepSeek AI also to generate the initial grammatical analysis, which was then discussed and corrected where necessary. Overall it was an instructive experience, and I learned a lot through it, it was like having a benevolent, polite and well-read teacher always available on hand.

I am very grateful to Prof. Dharmavardhana-Jñānagarbha with whom I discussed some difficult points. Despite his being both Rector and Dean of the Faculty of Graduate Studies, and his busy schedule having teaching work at the International Buddhist College in Sadao, Thailand, he dedicated much time looking at points I needed help with.

Of course, if there are any mistakes remaining, that is my fault entirely. I would be grateful to readers who notice such to contact me at anandajoti@gmail.com, so corrections can be made.

Ānandajoti Bhikkhu,
December 2025