Reconstructing Sanskrit Mantras from Chinese

Between 633-645 the famous pilgrim monk and scholar Xuanzang (602-664) 玄奘 visited around India, becoming an adept scholar and user of the Sanskrit language. He sincerely believed Sanskrit was the language of the gods, and readily pointed out “accented” forms of the language. The character é (meaning "accented") appears 93 times in his travel account the Great Tang Record of Travels to Western Lands 大唐西域記. He also describes the languages of India as follows:

《大唐西域記》卷2:「詳其文字,梵天所製,原始垂則,四十七言也。寓物合成,隨事轉用。流演枝派,其源浸廣,因地隨人,微有改變,語其大較,未異本源。而中印度特為詳正,辭調和雅,與天同音,氣韻清亮,為人軌則。隣境異國,習謬成訓,競趨澆俗,莫守淳風。」(CBETA, T51, no. 2087, p. 876, c9-14) 

Their letters were created by Brahma and have been passed down from their beginnings until now, being forty-seven in number. They combine to form words according to the object [declension?] and shift in use according to the action [inflection?]. It has spread around and branched off, its source being deep and broad. Due to regions and peoples there have been some changes, though the words are generally not different from the original source. Central India is especially proper, their diction being elegant and the same sound as devas with a character sharp and clear, which is a model for people. The neighboring countries have become accustomed to erroneous pronunciation. In their chaotic ways and base nature they do not maintain genuineness.

Such a stated belief on Xuanzang's part however did not encourage many Chinese clerics and scribes to learn Sanskrit in China. While indeed in the time of Xuanzang it was possible to learn Sanskrit to some degree in China, this was not so widespread or alluring it seems throughout the Tang dynasty (618-907) when many texts containing mantras and dhāraṇīs were translated into Chinese from Sanskrit. Although the language might have been divine in Indian Buddhist minds, this did not merit precise use of it on the part of Chinese Buddhist clerics.

Sanskrit and Chinese are fundamentally very different languages. Unlike the former, the latter lacks gender, declension, conjugation and any number of other features common to most Indo-European languages. This was as much the case fifteen centuries ago as it is now.

Unlike the Tibetans, the Chinese never attempted to produce a standardized phonetic script to preserve Sanskrit pronunciation. While indeed some texts employed Indian siddhaṃ script for preserving the proper pronunciation of Sanskrit mantras, the general preference in both ancient and modern times has been to use Chinese characters (hanzi 漢字) for their phonetic values in transcribing mantras and dhāraṇīs.

This leads to distortion of the original sounds because even Middle Chinese (used in the Tang period) – to say nothing of modern Mandarin – was phonetically quite different from Sanskrit. Additionally, the pronunciation of characters differed from region to region (the speech of the capital was considered standard), and moreover changed throughout time. Such developments are actually reflected in the Japanese language. The Japanese imported Chinese characters over the course of several centuries and attempted to preserve the different pronunciations from each period. The system in Japan is as follows:

Go 吳 – Readings from before the 7th / 8th centuries. Possibly from the Korean peninsula or southern China. Often used in Buddhist texts. 

Kan 漢 – Readings from the mid Tang Dynasty (618-907). Generally reflect the pronunciation of Chang'an 長安. 

唐 – Readings from the Song Dynasty (960-1279). Often used in the Zen school. Here refers to China rather than the Tang Dynasty.

This is why a single Chinese loanword in Japan can have multiple pronunciations. For instance, 和尚 (preceptor or priest) is pronounced oshō in Zen, kashō in Tendai, and washō in Shingon. In modern Mandarin in would be héshàng.

The pronunciation of Chinese loanwords in Japanese is actually closer to Middle Chinese than modern Mandarin (this is not necessarily the case with other Chinese dialects however). For example, the character (“to eat”) is pronounced shoku in Japanese and shí in Mandarin. The consonant ending from earlier Chinese has been preserved in the Japanese importation of Chinese. The Japanese language was thus better able to retain approximate pronunciations of Sanskrit mantras and dhāraṇīs. They also continued regularly using the phonetic siddhaṃ script, though not without problems.

For example, the mantra of the Heart Sūtra (gate gate pāragate pārasaṃgate bodhi svāhā) reads as follows in modern Mandarin:

羯諦羯諦 波羅羯諦 波羅僧羯諦 菩提薩婆訶 

jiēdì jiēdì bōluó jiēdì bōluósēng jiēdì pútí sàpóhē

According to the Heart Sūtra as printed by Eihei-ji 永平寺 (Sōtō Zen) which I picked up at the Japanese temple in Bodhgaya last year, it reads as follows in the Sino-Japanese rendering:

gyātei gyātei hārā gyātei harasō gyātei bōjii sowakā

It is clear that the Japanese pronunciation, which has attempted to preserve Middle Chinese pronunciation, better reflects the Sanskrit.

This is why in my attempts to reconstruct certain mantras' pronunciations from Classical Chinese texts I have to consider the Sino-Japanese readings of characters. The modern Mandarin is just too far divorced from Middle Chinese, which was the form of Chinese used when the mantras in question were transliterated.

This is perhaps one of the most difficult parts of translating Classical Chinese Buddhist texts. In some cases there are siddhaṃ renderings which while often containing inaccuracies are still much easier to decode than ancient Chinese transliterations. Today we have the Chinese Buddhist canon digitalized. The software, like CBETA, fortunately now provides the Chinese and siddhaṃ as it appears in the printed Taishō canon with added romanization:

However, not all mantras will have the siddhaṃ in the primary text. In the absence of it, one has to attempt to reconstruct the Sanskrit using a variety of means.

Later Japanese texts will sometimes provide the siddhaṃ. For example, the Betsu Gyō 別行 (T2476) by Kanjo 寛助 (1057-1125) and the Gyōrin Shō 行林抄 (T2409) by Jōnen 靜然 (12th century). These texts will often, though not always, provide the Chinese characters and a siddhaṃ reading, though this is not always accurate and has to be assessed cautiously. For example, the latter text provides the following rendering for the mantra of Ākāśagarbha Bodhisattva:

namo akaśagarbhaya oṃ ari kamari mori svāhā

Based on the definitions of the terms used in the mantra, my understanding of this mantra would render it as follows:

namo ākāśagarbhāya oṃ ārya kamala mauli svāhā

Bear in mind the pronunciation of a mantra can differ also according to the lineage (in Japanese: ryū ), so traditions will have their own inherited pronunciations.

These last few months I have attempted to reconstruct the mantras for planets as given by Yixing (684–727) in his works. These are found in only a few texts in the Chinese canon. Fortunately, the aforementioned Betsu Gyō does provide siddhaṃ for them, though it is still problematic. For instance, the following is given for the mantra of Mercury:


oṃ vudha nakṣatra svāmina kheduma svāhā

Since the text expressly states this is the mantra for Mercury, we can assume Sanskrit terms for Mercury are likely to appear in the mantra. So, if we do a search for “Mercury” in the Monier-Williams Sanskrit Dictionary (, we find similar sounding terms like budha and induja. It then makes logical sense to read the mantra as follows:

oṃ budha nakṣatra svāmin induja svāhā

Here kheduma / induja would be read in Sino-Japanese (again remember this is reflecting Middle Chinese pronunciation) as keidoma 契弩摩. I would assume this reflects either a mistransliteration of heard Sanskrit, or the speaker was in fact speaking an Indian dialect rather than standard Sanskrit (or it was not their native language and they mispronounced the word). In any case, it simply seems logical to read induja (“son of the moon”) here and not kheduma.

It should be clear then that East Asia – neither China nor Japan – did not attempt to preserve precise transliterations of mantras and dhāraṇīs. This is understandable given that intensive Sanskrit studies with complete grammars were seldom undertaken or available. This stands in contrast to SE Asia where the study of Pāḷi was diligently undertaken and preserved until the present day.

In the absence of any siddhaṃ for a mantra in transliterated Chinese, the scholar is left to start a long process of guesswork. In some cases, there are parts of the mantra which are identical to other parts of known mantras. However, this will not always be the case as I've discovered.

Sometimes it is quite easy to figure it out, such as this mantra for Indra:


oṃ indrāya svāhā

Sometimes much of the mantra is simply unclear despite some parts being discernible (the underlined words indicate a guess on my part as to the possible pronunciation):


namo ratna trayāya namo suma sarva nakṣatra rājāya saḥ tu dhi pa āḥ lakaraya tad yathā dumati padumati sa piṅ ni kha se svāhā

Sometimes just looking at the Chinese I have been able to decode the Sanskrit without any major issues:


oṃ sarva nakṣatra samaye śrī ye śāntika kru svāhā

As noted above, this is one of the most challenging aspects of translating Classical Chinese Buddhist texts. There are fortunately some scholarly volumes available today which provide decoded mantras, though again these have to be approached with caution. Nevertheless, as a translator when I have to translate these mantras I do appreciate seeing what others have produced. I am neither a Sanskritist nor Vajrayāna specialist, so I would hopefully have experts to defer to, though in many cases I am left to figure things out alone.


Jayarava said...

Hi Indrajala

This is a fascinating subject. You mention some volumes which decode mantras. Could you post some references? I'm particularly interested in the dhāraṇīs that Jan Nattier mentions in her article on the Heart Sutra. There's one in T 12.387, for example, that reads

竭帝 波利竭帝 僧竭帝 波羅僧竭帝波羅卑羅延坻 三波羅卑羅延坻 婆羅 婆羅 波沙羅 波娑羅 摩文闍 摩文闍 遮羅帝 遮羅坻 波遮羅坻 波遮羅坻 三波羅遮羅坻 比提 嘻利 嘻梨 薩隷醯 薩隷醯 富嚧 富嚧 莎呵

I can make out parts of it, but not the whole thing. The opening part is very similar to the Heart Sutra dhāraṇī.

Jeffrey Kotyk said...

Hi Jayarava

See the following:

Lin Guangming 林光明, ed. Xinbian Dazang quan zhou 新編大藏全咒. The New Edition of All Mantras in Mahāpiṭaka. Mahāpiṭaka-mantramāla-purṇāṅga-nava-saṃskaraṇaṃ. 18 vols. Taipei: Jiafeng chubanshe/Mantra Publisher, 2001.

There's also this site, unfortunately only in Chinese:

It doesn't seem like your dhāraṇī in question has been dealt with in the online world yet unfortunately.

I can definitely discern parts of it as well, but a lot of it is unclear to me.

One other problem is that earlier periods used different forms of transliteration, like this one.

In the Tang it was more standardized to some extent, or at the very least you had a single translator and his students doing many texts, such as Amoghavajra, so there was consistency.

Also, during the Tang a lot of texts were translated in Chang'an, so there was perhaps a common understanding of how to transliterate Sanskrit with Chinese characters. There were manuals on Sanskrit pronunciation and transliteration entirely in Chinese.

That being said, they really never attempted to faithfully preserve the pronunciation of anything.

Nowadays in Chinese temples they still insist on reciting everything using modern Mandarin readings of these transliterations, so it ends up being garbled gibberish. Almost nobody you ask will know what they mean, but they'll memorize long pages of them (and it is mandatory that you do so as well in some Chinese Buddhist seminary programs).

Jayarava said...

Thanks anyway.

I'm surprised to see you saying that "they really never attempted to faithfully preserve the pronunciation of anything." because I would have thought that was exactly what the elaborate transliterations were about. Also the continued use of Siddham would seem to argue against lack of concern for correct pronunciation.

My impression of Kūkai (though it's been a while since I looked at his stuff) was that he highly valued correct pronunciation of mantras. And this is why Siddham is still part of the Shingon curriculum.

Or was it more like getting a good enough approximation - like Roman script without diacritics?

Jeffrey Kotyk said...

The transliteration into Chinese was only loosely approximate, and I would assume in the Tang the Chinese and Indians were both well aware of this. Nevertheless, although siddham was an option, I often have gotten the impression East Asians just preferred to transliterate things into Chinese.

One reason I say this is that Kūkai's journal, which I viewed at a museum in Japan some time ago, reveals him transliterating Sanskrit words into Chinese. He does this in his Chinese works (not all though) as well, rather than using siddham. Also, the extant texts available in modern canons show that only a portion have siddham. The general preference has been to use Chinese exclusively.

Also, bear in mind that Japanese use of siddham is not without its problems. For one thing, they use Japanese phonetics (basically katakana readings) of mantras, and from my understanding these vary from tradition to tradition. Also, the texts I've seen with siddham often display constant errors, like bajra for vajra or vudha for budha. This might have been later scribal errors, but then that just goes to show how precision was not a priority. In China it was much the same. Siddham was paired with Chinese characters for approximate pronunciation.

So, there were scribal problems plus the pronunciation was only kept approximate in Japan. Some scholars also have reservations about Kūkai's abilities in Sanskrit. He might have read siddham and knew a lot of vocabulary, but that doesn't mean he knew Sanskrit.

Classical Japanese readings of siddham would have been like their katakana pronunciation of English words. Chocolate cake becomes chokoreeto keeki, for instance.

It works sort of, but it isn't precise.

LY said...

In Taiwan, I've heard chanting in Taiwanese (Hokkien). I asked for Latin pinyin texts, and was handed a book that had the pinyin for Taiwanese! Not sure if that was the chanting liturgy though...

Would it be worthwhile to get the pronunciation in Cantonese, Hakka, etc.? For me, it would drive me nuts.

Jeffrey Kotyk said...


The romanization of Chinese will normally be in Mandarin as they recite things in Taiwan generally in Mandarin, not Taiwanese or Hokka. Likewise Cantonese romanization wouldn't be widely available.

As far as mantras are concerned, I'd say find the romanized Sanskrit and learn what they mean. Reciting them as Chinese is meaningless.

Unknown said...

Hi there,

Thanks so much for this fascinating post! I'm wondering what the function of the 1-3 character phrases or notations are in parenthesis - for example,the (二合) notation in the mantra you transcribed: 唵(引)母駄曩乞殺(二合)怛羅(二合)娑嚩(二合)弭曩(名位)契弩摩娑嚩(二合引)賀.

Also - I've always wanted to ask someone why, in East Asian Buddhist ritual manuals, mudra descriptions refer to the two hands of the deity or practitioner with the phrase 二羽? It's a strange euphemism, or perhaps I'm just missing something. Thanks so much for any information you can provide!

Jeffrey Kotyk said...

Hi Waverly


引 indicates a long vowel.
二合 is a phonetic combination of the last two characters.

So 怛羅 = tra.

I'm not sure about the phrase 二羽, but it could be a translation of pakṣa, which has a number of meanings including limb:


(H2) pakṣá [p= 573,2] [L=113061] m. (ifc. f(ā or ī). ) a wing , pinion (in one passage n.) RV. &c
[L=113062] a symbol. N. of the number two Var. Hcat.
[L=113063] a feather , the feathers on both sides of an arrow (cf. gārdhra-p°)
[L=113064] the fin of a fish (cf. nis-tvak-p°)
[L=113065] the shoulder
[L=113066] the flank or side or the half of anything RV. &c
[p= 573,3] [L=113067] the side or wing of a building AV.
[L=113068] the wing or flank of an army MBh. Hariv.
[L=113069] the half of a lunar month (the first half from new moon to full moon was called pūrva or āpūryamāṇa , later śukla or śuddha ; the other half apara or apa-kṣīyamāṇa , later kṛṣṇa or tāmisra ; each fortnight consists of 15 tithis or lunar days called prathamā , dvitīyā &c ) Br. Gr2S3rS. MBh. Var. &c
[L=113070] a side , party , faction
[L=113071] multitude , number , troop , set , class of beings
[L=113072] partisan , adherent , follower , friend (śatru- " the enemy's side " or " a partisan of the enemy " ; mahā- , " one who has many adherents ") MBh. Ka1v. &c
[L=113073] side i.e. position , place , stead (°kṣe ifc. instead of or by way of) ib.
[L=113074] quantity (» keśa-)
[L=113075] one of two cases or one side of an argument , an alternative (°kṣe , " on the other hand " , with atra , " in this case " , pakṣā*ntare , " in the other case ") Pa1n2. Sch.
[L=113076] a point or matter under discussion , a thesis , a particular theory , a position advanced or an argument to be maintained (cf. pūrva- , uttara-)
[L=113077] an action or lawsuit Ya1jn5. Sch.
[L=113078] (in logic) the proposition to be proved in a syllogism Tarkas. Bha1sha1p.
[L=113079] any supposition or view , motion , idea , opinion (mukhyaḥ pakṣaḥ , " an excellent idea " S3ak. Sch.) MBh. Ka1v. &c
[L=113080] the sun Sa1y. on RV. iii , 53 , 16
[L=113081] N. of sev. men VP.
[L=113082] (in alg.) a primary division or the side of an equation in a primary division
[L=113083] the wall of a house or any wall L.
[L=113084] an army L.
[L=113085] favour L.
[L=113086] contradiction , rejoinder L.
[L=113087] the ash-pit of a fire-place L.
[L=113088] a royal elephant L.
[L=113089] a limb or member of the body L.
[L=113090] the feathers of the tail of a peacock , a tail L.
[L=113091] proximity , neighbourhood L.
[L=113092] a bracelet L.
[L=113093] purity , perfection L.
(H2B) pakṣá [L=113094] mfn. = pācaka , bādhaka Sa1y. on RV. vi , 47 , 19. [cf. OGerm. fahs ; Angl.Sax. feax.]

TPK said...
This comment has been removed by the author.
TPK said...

Hi Indrajala,
Most of the Chinese mantras in the esoteric section of Taisho Tripitaka were reconstructed to Sanskrit in the Vietnamese translations. You can read at:
Mật Tạng Bộ 1 (848 ~ 917) [CBETA T.18]
Mật Tạng Bộ 2 (918 ~ 1029) [CBETA T.19]
Mật Tạng Bộ 3 (1030 ~ 1198) [CBETA T.20]
Mật Tạng Bộ 4 (1199 ~ 1420) [CBETA T.21]

TPK said...

竭帝 波利竭帝 僧竭帝 波羅僧竭帝波羅卑羅延坻 三波羅卑羅延坻 婆羅 婆羅 波沙羅 波娑羅 摩文闍 摩文闍 遮羅帝 遮羅坻 波遮羅坻 波遮羅坻 三波羅遮羅坻 比提 嘻利 嘻梨 薩隷醯 薩隷醯 富嚧 富嚧 莎呵
The Chinese text of Jayarava Attwood can be transliterated into Sino-Vietnamese as follow:
And the Sanskrit reconstruction based on Sino-Vietnamese:
(Note: Sino-Vietnamese is the pronunciation of Chinese characters in Vietnamese which is said to be similar to pronunciation of Chinese in Tang Dynasty)

Jeffrey Kotyk said...


Thank you very much for this. I was unaware this had been done before. I'll have to compare their reconstructions with what I managed to do. I generally look at the Japanese preserved readings, but I suppose the Vietnamese are equally worth examining!

TPK said...

For converting Chinese characters to Sino-Vietnamese pronunciation you can use the online tool at:
> Choose [Phiên âm 翻音] Tab
There are two options: Âm Hán Việt and Pinyin. The former is Sino-Vietnamese.