Anyone familiar with Buddhist mantras will be familiar with the number of words that end in -e. They constitute something of a mystery as they don't make sense grammatically or semantically, and explanations of them are obviously ad hoc (i.e. made up on the spot). For instance the Heart Sūtra mantra:
gate gate paragate parasaṃgate bodhi svāhā Compare to this to the dhāraṇī offered by the Medicine King Bodhisattva in the White Lotus Sūtra:
anye manye mane mamane citte carite same samitā viśānte mukte muktatame same aviṣame samasame jaye kṣaye akṣaye akṣiṇe śānte samite dhāraṇi ālokabhāṣe pratyavekṣaṇi nidhiru abhyantaraniviṣṭe abhyantarapāriśuddhimutkule araḍe paraḍe sukāṅkṣi asamasame buddhavilokite dharmaparīkṣite saṃghanirghoṣaṇi nirghoṇi bhayābhayaviśodhani mantre mantrākṣayate rute rutakauśalye akṣaye akṣayavanatāye vakkule valoḍra amanyanatāye svāhā ||
(Vaidya 1960 : 233) Note how many of these words have the -e ending. Kern, the first person to translate the Lotus into English, in 1884, links many of these names to the Great Mother Goddess.
“All of these words are, or ought to be, feminine words in the vocative. I take them to be epithets of the Great Mother, Nature or Earth, differently called Aditi, Prajñā, Māyā, Bhavānī, Durgā. Anyā may be identified with the Vedic anyā, inexhaustible, and synonymous with aditi. More of the other terms may be explained as synonymous with prajñā (eg pratyaveksaṭi), with nature (kṣāye akṣāye), with earth
(dhāraṇī).” (Kern p.371, note 3) For the uninitiated perhaps a brief explanation about inflected languages is in order. Where in English we use prepositions such as "of, for, to, by, with, on, in, from" etc. to indicate the relationship between words in a sentence, inflected languages add different endings to the words. The easiest way will be to show. Let's take a word that has a stem in -a: buddha. (It's actually a past-participle meaning awoken or understood). So somewhat simplistically we could show the endings and their 'meaning':
buddhaḥ - nominative - the Buddha. buddhaṃ - accusative - the Buddha as the patient of a verb: e.g. I saw the Buddha. buddhena - instrumental - by means of, or with the Buddha. buddhāya - dative - to or for the Buddha. e.g. namo Buddhāya - homage to the Buddha. buddhāt - ablative - from the Buddha buddhasya - genitive - of the Buddha; the Buddha's... (possessive) buddhe - locative - in or on the Buddha buddha - vocative - O Buddha. (address or invocation) There are many paradigms like this in Sanskrit. Each noun has dual forms in addition to singular and plural. Masculine, feminine and neuter nouns vary slightly, and stems can end in any monophthong vowel or certain consonants - meaning that there are very many different forms to remember! The -e ending is typically associated with three grammatical forms. In the case of the word 'gata', which is a part-participle and declined like a noun, the possibilities are:
feminine vocative masculine locative (see above) or vocative Gaté therefore most likely means something like "O she who is gone", i.e. it is in the vocative case. This is what Edward Conze thought. The other possibilities are open, but the subject of the Heart Sūtra is Prajñāpāramitā - who is feminine both in gender and grammatically. But when all the words in a mantra, as above, are in the feminine vocative the string of words is not grammatically sensible - they do not make a sentence. The usual explanation is some variation on the idea that these are strings of invocations to deities or qualities. Kern obviously thought something like this and expected feminine vocatives - but note that he is expecting Hindu goddesses in a Buddhist text. To some extent they do appear, but he may not, in 1884, had a very clear idea of the differences between Buddhism and Hinduism

An unspoken assumption here is that the mantras are written in Classical Sanskrit. This is the language which was formalised and polished (ie 'saṃskṛta' - literally "made complete") in about the 4th century BCE, and became the standard language for literary and religious compositions in India. Given that some of the texts, the Heart Sūtra for one, are written in Classical Sanskrit this seems at first glance a reasonable assumption. However we know that Buddhists before the Gupta Empire (4th - 6th centuries CE) wrote in a Sanskritised version of the Prakrit, or spoken dialect, locally spoken. This literary language which is now known as Buddhist Hybrid Sanskrit actually shows massive variations over time and place. The name Buddhist Hybrid Sanskrit (BHS) was coined by Frank Edgerton who wrote a grammar and dictionary for it. Many well known sūtras were written in BHS including the Mahāvastu, Lotus, Golden Light, Gaṇḍhavyūha, laṅkkāvatātra, Sukhāvatīvyūha (larger and smaller), the Large Perfection of Wisdom, and the Diamond. Śantideva's Śikṣāsamuccaya which is made up of quotes from many Mahāyāna sūtras is almost entirely in BHS. In addition, notably, Buddhists reverted to BHS during the Tantric period after the 6th century.

BHS is full of irregularities particularly in the grammatical endings. In Classical Sanskrit for nouns with stems ending in -a the nominative singular is -aḥ (e.g. devaḥ, the god, or the king), whereas is Pāli the ending is -o, (e.g. devo), while in Māgadhī the nom sg. was -e (deve). The variety of BHS nom sg. endings for nouns with -a stems found in extant manuscripts includes: -o, -u, -ū, -a, -ā, -aṃ, and -e. The -e ending is also used for the vocative singular as in Sanskrit. In the case of the Gāndhārī Prakrit, at least in written form, the variety of nom. sg case endings has been described as "bewildering", and it seems as though final vowels may have been de-emphasised to the point of almost disappearing in speech, which caused confusion amongst scribes (Salomon p.130-131)

So if the mantras were written in Prakrit or perhaps BHS then we might suspect that they were simply words in the nominative singular. It might better explain the long lists of words such as the Lotus Sūtra example quoted above, although the Heart Sūtra mantra might still best be seen as an invocation. If one is stringing together words then the most basic form is usually the nominative singular.

But why would Sanskrit texts preserve a form that is aberrant from the point of view of Classical Sanskrit grammar? To answer this I cite the example of a Prakrit feature that is preserved in many Sanskrit texts, over quite a long period of time and including in indirect borrowings.

The Arapacana alphabet is the alphabet of the Gāndhārī prakrit. We know that at least from the first couple of centuries of the common era it was used as a mnemonic device, where each letter stands for a key word that is used in a line of verse of a poem. Most extant examples are either obviously a practical reminder about a meditation practice, or derive from one of these. Having been composed in Gāndhārī, perhaps as a stand alone poem*, it was imported into Sanskrit texts such as the Lalitavistara Sūtra and the Large Perfection of Wisdom Sūtra**. In the case of the Lalitavistara a version exists which was fully Sanskritised, but there is also a version in Chinese translation which retains the Gāndhārī order (see Brough 1977). In the Large Perfection of Wisdom Sūtra all known versions retain the Gāndhārī order. In the Gaṇḍhavyūha Sūtra the Gāndhārī order is retained but the phonetic connection between the alphabet and the keywords is lost. In the Mahāvairocana Abhisaṃbodhi Tantra (MAT) the alphabet is Sanskrit, but the vowels except 'a' are left off in imitation of the Gāndhārī alphabet (which only uses one sign for initial vowels, which is then modified by diacritics to make all the other vowels). It's reasonably obvious that the source for the alphabet in the MAT is the Large Perfection of Wisdom Sūtra, which means that it is twice removed from Gāndhārī, a language not spoken in India for several centuries by the time the MAT was composed! Perhaps not surprisingly there was a streak of conservatism by Buddhists when composing texts, especially with regard to mantra.

So there is a possibility here: if the mantras were in fact composed in Māgadhī or Buddhist Hybrid Sanskrit, or some other dialect where a nom sg. in -e was used, then it is likely that the original form of the mantra would have been retained even as the text itself was Sanskritised. It might even have stayed in that form when borrowed by other texts. This means that the form in the mantras could be nominative and strings of words in nom sg., or at least intended to be. Perhaps a closer examination of the words, without the assumption of Classical Sanskrit might lead to a better understanding. For the moment the puzzle remains, and my conjecture though plausible is not a final answer to the problem - if anything I may have muddied the waters!

* we're still waiting for a full translation and analysis of a fragment of manuscript containing the earliest known version of the Arapacana. See the publications page of the Bajaur Collection of Buddhist Kharoṣṭhī Manuscripts for a preliminary report.

** The Large Perfection of Wisdom Sūtra has versions in 18,000, 25,000 and 100,000 lines which are distinguished chiefly by the number of repetitions and the thoroughness of spelling out variations on a theme. The chief feature of the version in 100,000 lines is the lack of the use of "etc" or "and so on". Conze has published an English translation largely based on the 25,000 line version, but which draws freely on the others due to the "execrable state" of the manuscripts.

Reading
Boucher, Daniel. 1998. Gāndhārī and the early Chinese Buddhist translations reconsidered : the case of the Saddharmapuṇḍarīkasūtra. Journal of the American Oriental Society., 118 (4), p.471-506. Brough, John. 1977. The arapacana syllabry in the old Laita-vistara. Bulletin of the School of Oriental and African Studies. 40 (1), p.85-95. Vaidya, P.L. 1960. Saddharmapuṇḍarīkasūtram (Buddhist Sanskrit Texts, 6). Darbhanga : The Mithila Institute. Online: www.sub.uni-goettingen.de. Conze, Edward. 1975. (trans.) The large sutra on perfect wisdom ; with the divisions of the Abhisamayālaṅkāra. (Delhi, Motilal Banarsidass : 1990) Edgerton, Franklin. 1953. Buddhist Hybrid Sanskrit Grammar and Dictionary. New Delhi : Munshiram Monoharlal, 2004. Kern, H. 1884. The saddharma-pundarīka or the lotus of the true law. (Delhi : Motilal Banarsidass, 1980) (1st pub. 1884 Sacred books of the East Vol.21) Nattier, Jan. 2005. A few good men : the Bodhisattva Path According to the Inquiry of Ugra (Ugraparipṛcchā). (University of Hawaii Press) Salomon, Richard. 1999. Ancient Buddhist scrolls from Gandhāra : the British Library Kharoṣṭhī Fragments. London : The British Library. See also my bibliography of publications on the Arapacana.


30 July 2010.
See also Some Additional Notes for a summary of: Cohen, Signe. "On the Buddhist Hybrid Sanskrit/Middle Indic Ending "-e" as a 'Magadhism'," Acta Orientalia Vol. 63 (2002), p.67-9.
13 Aug 2015

I was discussing the correct way of writing a Fudō mantra and realised something about the mantras in the Mahāvairocana Abhisaṃbodhi Tantra. MAT is probably the first proper tantric text. The mantra in question is: namaḥ samantavajrāṇāṃ hāṃ. Most of the mantras begin with either this or namaḥ samantabuddhāṇāṃ. The Sanskrit is slightly peculiar. Namaḥ means 'bow, pay homage, etc' (from namati 'he bows'). Samanta means 'universal, all, entire'. The slightly weird bit is the grammatical inflexion. Usually we bow *to* something as in namo buddhāya. The inflection is the dative singular case, "homage to the buddha". Here the inflection is the genitive plural. samantavajrāṇāṃ means 'of all the vajras'. So it looks like it says "homage of all the vajras". Really we want namaḥ samantavajrebhyaḥ "homage to all the vajras". So this is in fact a very interesting thing. Because in Prakrit (like Pāḷi), unlike Sanskrit, the dative merged with the genitive. So this proves that the mantras in the MAT were composed in Buddhist Hybrid Sanskrit. And this bolsters the argument here that we should consider the -e ending to be Prakrit, or Buddhist Hybrid Sanskrit, rather than Pāṇinian Sanskrit.