05 July 2019

Svāhā in The Heart Sutra Dhāraṇī

Jan Nattier showed that the Xīnjīng (T.251) is the original Heart Sutra. All my work on the text confirms this essential observation. It was composed as a digest text (抄經 chāo-jīng) in the period 645-661 CE. Given this, we expect the Chinese version of the text to be consistent with few variations. This is largely true except that the witnesses to the Xīnjīng contain a few character substitutions and other minor variants. Weirdly, the authoritative version of the Chinese Canon, the Taishō Shinshū Daizōkyō, has the least likely reading (and not the one that millions of people chant everyday). In this essay I will focus on a particular transliterated Sanskrit word in the dhāraṇī, i.e. svāhā. I will say a little about the Sanskrit word, but the main focus of this essay will be the plethora of Chinese forms of this word in the earliest witnesses to the text.


The earliest commentaries are ambivalent about the dhāraṇī. In fact, Kūijī (T.1710) declines to comment on it. Woncheuk (T.1711) takes an oblique approach. He notes that the words of the dhāraṇī cannot be translated without losing their magical powers (鬼神). Then he notes that the dhāraṇī contains Buddhist Sanskrit words that can, in fact, be translated and translates them. For example: 揭諦言度度 "gate means: go beyond, go beyond" (33.551c21). The character 度 is used by Kumārajīva (and in the Heart Sutra) to translate samatikramati "go beyond, transcend" and related verbs. Another example is 後莎婆呵此云速疾。"Lastly, svāhā: it means quickly (速疾)" (551c24). This definition is echoed by 法藏 Fǎzàng in his commentary (T.1712).

This turns out to be the standard approach for the dhāraṇī. The words that can be discerned and translated become the basis of various poetic riffs on one's beliefs about Buddhism. Each word becomes a cipher from which one can unpack Buddhist doctrines. The metaphor of crossing a river for transcendence is one of the most popular approaches to unpacking the word gate.

Note that svāhā does not mean "quickly". It is usually analysed as related to the past perfect form of a defective verb √ah, the only surviving form of which is the 3rd person past perfect āha "was said". The first literary occurence of svāhā is in the Yajurveda where it is used to "activate" offerings. An offering to, say, Agni would be made while chanting, angaye svāhā "For Agni. It was well said". The expression has the same illocutionary force as amen in Christian prayers or hitting the <enter> key for computer programming. Svāhā is two syllables and just two akṣara (or characters) in Indian scripts, e.g. स्वा हा. Here  स्व (sva) uses the combining form of (sa) with (va), but the result is one akṣara. Note that the vowel a is inherent in the written akṣara, though for final consonants it can be deleted using the virāma, so that final -s is स्.  The long vowels are indicated by the extra vertical stroke.

Also note that dhāraṇī were probably created in Prakrit or at best Buddhist Hybrid Sanskrit. The -e ending in a context that strongly implies the nominative case points us to this conclusion.

Such poetic attempts at exegesis of the dhāraṇī have an ad hoc feel to them and seldom match up with other attempts. More so than with other aspects of the Heart Sutra, the dhāraṇī presents challenges to the would be exegete. We've some some way since D T Suzuki wrote:
Dhāraṇī is a study by itself. In India where all kinds of what may be termed abnormalities in religious symbology are profusely thriving, Dhāraṇī has also attained a high degree of development..." - D T Suzuki. The lankavatara Sutra. p.223, n.1.
However, I'm not sure we have yet found a positive way to understand the role of magic in Buddhism. Indeed, we seem to be in denial about the place of magic in pre-modern Buddhism, partly, no doubt, because of the outrageous suggestion that Buddhism is a "rational religion". In 2013, I blogged about dhāraṇī in general and about Ariel Glucklich's approach to magic (Why is there a Dhāraṇī in the Heart Sūtra?18 October 2013), which I think is quite a useful way of thinking about it. However, Glucklich remains quite unknown in Buddhism Studies, as far as I can tell, and no one seems to study Buddhist magic, though Jeffrey Kotyk and Bill Mak do study Buddhist astrology.

I've previously blogged about why I call the Heart Sutra spell a dhāraṇī rather than a mantra (Heart Sutra Mantra. 30 August 2013) and published a refined version of the argument in 2017 (‘Epithets of the Mantra’ in the Heart Sutra), and about the -e ending. Because my thesis on this has not caught on, I want to quickly rehearse the argument that the "mantra" is a dhāraṇī before getting into transliterations of svāhā.

vidyā → dhāraṇī → mantra

The Chinese character 呪 (and the variant咒) means "incantation, spell, curse" and was adapted for a Buddhist context, where it meant dhāraṇī for a time, and then, when mantras were introduced to China in the mid-late 7th Century, it was used again for mantra. And this led to some confusion; crucially, the Heart Sutra spell was called a mantra when, in fact, it is a dhāraṇī.

The section of the epithets was originally something like this 6th Century Sanskrit text:
mahāvidyeyaṃ bhagavan yaduta prajñāpāramitā | anuttareyaṃ bhagavan vidyā yaduta prajñāpāramitā asamasameyaṃ bhagavan vidyā yaduta prajñāpāramitā (Gilgit Ms 146v: 12-13)
Kumārajīva translated:
世尊!般若波羅蜜是大明呪、無上明呪、無等等明呪。(T223, 8.286b28)
Bhagavan, perfection of insight is a great spell, an unsurpassed spell, an unequalled spell.
Note that vidyā is translated as 明呪. However, 200 years later, Xuanzang translates vidyā as 呪 and treats 明呪 as two separate words:
如是般若波羅蜜多是大神呪、是大明呪,是無上呪,是無等等呪,是一切呪王 (T220-ii, 7.156.a17-22: fasc. 429)
Comparing the original and the two translations...
anuttarā vidyā無上明呪無上呪
asamasamā vidyā無等等明呪無等等呪

Xuanzang is famous for the accuracy of his translations, but here he has done us a disservice. Curiously, he adds two epithets of prajñāpāramitā that have no Sanskrit counterpart. Firstly, 神呪 is probably a synonym of 明呪 (vidyā). And secondly, 一切呪王, would probably correspond to sarvavidyārājñī "queen of all the spells", but this term is not found in Sanskrit in this context.

In any case, the trend of treating 明呪 as two words, combined with the influence of Tantra, led to 呪 being translated into English as mantra. In 1863, twenty years before the appearance of a Sanskrit Heart Sutra in England, Rev. Samuel Beal was the first to translate the Heart Sutra into English. He translated 呪 as dhāraṇī, suggesting that his local informants and the Tang Dynasty commentary he was using both supported this reading.

When translating a word like 呪 we have to pay careful attention to the context and especially the era in which the character was used, but also the context in which the source was composed. In this context it means dhāraṇī. In the major Prajñāpāramitā sūtras dhāraṇī is not a magic spell, but the ongoing effects of the insight (see Aṣṭasāhasrikā: Insight and Ongoing Transformation. 1 December 2017). Later, in the Pañcaviṃśatisāhasrikā, it also refers to the Arapacana acrostic. And later still (by some centuries), it came to be associated with apotropaic magic spells consisting of standalone words, repeated and reiterated with different prefixes, and finished with svāhā. These are related in intention to the Pāli paritta texts.

A mantra is ideally situated in the context of a Tantric ritual (with abhiṣekha, mudrā, and maṇḍala) and performs a function within that ritual (On the importance of context for tantra see Abe 1999). The function may be invoking a deity, or a quality (such as peace śānti), or one the magic rites of pacification (śāntikakarman), subjection (vaśyakarman), prospering (puṣṭikarman), or destruction (raudrakarman). A mantra usually begins and ends with magical symbols called bījākṣara or "seed syllables" such as hūṃ, hrīḥ, or dhīḥ; beginning always with the bījākṣara oṃ (not auṃ which is exclusively used in Hindu contexts).

There are, of course, hybrid forms, such as the Tārā mantra oṃ tāre tuttāre ture svāhā which has features of both dhāraṇī and mantra, but it comes from a distinctive Tantric context and performs the function of recollecting the name of the deity (nāmānusmṛti) or recollection of the Buddha (buddhānusmṛti) within the context of a sādhana or ritual including mantra, mudrā, and maṇḍala (speech, body, and mind).

The Heart Sutra is clearly not a Tantric text, though it was adopted as such and included into various sādhana during the Pala Era in India (8th – 12th Centuries). Oṃ was often added to the dhāraṇī in such contexts. In the case of the Xīnjīng we are not yet in the Tantric realm, though as we will see the dhāraṇī comes from a source translated by one of the first Tantrikas to arrive in China.

The Xīnjīng Dhāraṇī

The text of the section that includes the dhāraṇī as it appears in the printed Taishō Edition of the Chinese Tripiṭaka is:
Gù shuō bōrěbōluómìduō” zhòu Jí shuō zhòu yuē
Therefore recite the Prajñāpāramitā dhāraṇī, that is to say the dhāraṇī that says:
揭帝 揭帝 般羅揭帝 般羅僧揭帝 菩提 僧莎訶
jiēdì jiēdì bānluójiēdì bānluósēngjiēdì pútí sēng shāhē
gate gate pāragate pārasaṃgate bodhi svāhā

Note that Taishō does not punctuate the dhāraṇī but there are spaces between the words. Anyone familiar with this text will be struck by the last three characters. This is not what people who chant the Heart Sutra chant. What they chant is 薩婆訶 sà pó hē. A note in the Taishō tells us that in the Song, Yuan and Ming versions of the Tripiṭaka the dhāraṇī does end 薩婆訶 sà pó hē. We can now also say that the Fangshan Stele ends with 莎婆訶:
揭諦 揭諦 般羅揭諦 般羅僧揭諦 菩提 莎婆訶 (my transcription)
jiēdì jiēdì bānluójiēdì bānluósēngjiēdì pútí shāpóhē
I hope to publish an English language study of the Fangshan Stele very soon (2019). There is one other character variation in the Fangshan inscription dhāraṇī, i.e. 諦 for 帝, but the sound and tone are the same. This romanization sà pó hē  for svāhā makes good sense because the two sounds sà pó are intended to represent the consonant cluster svā. This is a standard technique when transliterating Sanskrit using Chinese characters.And the sounds would have been closer in medieval Chinese or Middle Chinese (MC).

A whole Sanskrit Heart Sutra  is preserved in the Tripiṭaka as a Chinese transliteration (T.256. See The Other Heart Sutra). This text is now attributed to Amoghavajra and dated to the early 8th Century. We also have a manuscript of the same text, from the 8th Century, which has some minor variations. It was preserved at Dunhuang and is now held in the British library (manuscript no. Or.8210/S.5648). Note that this text is accompanied by a Chinese text, but the dhāraṇī is not translated. The dhāraṇī in T 256 is quite different from the standard Xīnjīng (T 251):
誐諦 誐諦  播囉誐諦 播囉僧誐諦 冒地 娑嚩賀 
édì édì bōluō'édì bōluōsēng'édì màode suōpóhè
In T.256, svāhā is represented by 娑嚩賀 suō pó hè. Along with the first two characters is the annotation 二合引 "two characters together, long vowel) and with the last is 引 "long vowel". The Pinyin romanization looks off and we need to know that the Middle Chinese was something like sa-ba ha, where saba together represent svā. And this was a very common way of rendering svāhā in Tantric texts (see Jeffrey Kotyk's blog Reconstructing Sanskrit Mantras from Chinese).

Of the two commentaries composed by Xuanzang's associates, Woncheuk's text has 莎婆呵 shā pó hē for svāhā (T. 1711. 33.551c10), while Kuījī's has 莎訶 shā hē (T. 1710. 33.542c8). Another early Tang dynasty commentary by 法藏 Fǎzàng (702 CE) has 薩婆訶 sà pó hē (T.1712. 33.555a6) which is the Fangshan/popular version.

In which case what do we make of 僧莎訶 sēng shā hē as found in the Taishō version of T 251? It seems to me that the character 僧 occurs earlier in the dhāraṇī and could easily have been copied here by mistake (an "eye-skip" or haplography). Correcting this mistake would leave us with 莎訶 shā hē which might pass for a rendering of svāhā, which is what we find in Kuījī's text. More importantly, we also find this transcription in another text that we will meet shortly.

To summarise, then, in the various early Witnesses of the Xīnjīng, svāhā in the dhāraṇī takes many forms (with Pinyin romanization).

T.251僧莎訶sēng shā hē
Popular薩婆訶sà pó hē
Fangshan莎婆訶shā pó hē
T.256娑嚩賀suō pó hè
Woncheuk莎婆呵shā pó hē
Kuījī莎訶shā hē

Note 2.9.2019: Huilin’s 一切經音義 Yīqièjīng yīnyì "Pronunciation and Meaning of All the Sūtras", published 807 CE, lists the spelling in T 250 as 僧婆訶 whereas Taishō has 僧莎呵 with no notes.

I now want to introduce a text that was proposed by the late John McRae as a possible source of the dhāraṇī (1988: 107 n.10). [McRae was Jan Nattier's husband]  See Nattier 1992: 211 n 53 for other possible sources.


The Indian bhikṣu, Atikūṭa, whose name we infer from the Chinese 瞿多 (Qú duō) or 阿地瞿多 (Ādìqúduō) was a contemporary of Xuánzàng. He arrived in Chang’an in 651 and was installed in 慈恩寺 Ci’en Monastery by the Emperor. This monastery is also closely associated with Xuánzàng, so it's quite possible that they knew each other. However, Xuanzang seems to have had little interest in the tantric side of Buddhism and was focussed on establishing his Fǎxiāng School (法相宗) of Yogācāra Buddhism.

Atikūṭa translated a number of texts into Chinese, including some of the first Tantric texts (which had already started to arrive in China without the context of Tantric doctrines and practices (e.g. abhiṣekha, sādhana, mantra, mudrā, and maṇḍala). He may have conducted the first abhiṣekha ritual in China. In particular, we are interested in him because, in 654, he translated a text called the 《陀羅尼集經》 Dhāraṇīsamuccaya "Collection of Spells" (T 901). The text was translated at Huirisi 慧日寺 a monastery associated with the 三階教 "Teaching of Three Levels" movement that was proscribed in 600 CE, but favoured by Sui Wendi and later by Wu Zetian. In this text we find the very dhāraṇī that occurs in the Xīnjīng.

Here is the relevant section, laid out horizontally, but preserving the lines in Taishō, Vol 18, p.807, register b, then arranged more naturally, followed by my very literal translation.
[19] 般若大心陀羅尼第十六呪曰。
[20] 跢姪他(一)揭帝揭帝(二)波羅揭帝(三)波囉僧揭
[21] 帝(四)菩提(五)莎訶(六)
[22] 是大心呪。用大心印。作諸壇處一切通用。
[23] 般若小心陀羅尼呪曰。
[24] 跢姪他(一)揭帝揭帝(二)波囉民(彌忍反)揭帝(三)波
[25] 囉若(若冶反)他(四)莎訶(五)
We can usefully arrange this into connected passages. I'm not sure that the numbers add anything, but let's leave them in for now.
呪曰。跢姪他 (一) 揭帝揭帝 (二) 波羅揭帝 (三) 波囉僧揭帝 (四) 菩提 (五) 莎訶 (六)
跢姪他 (一) 揭帝揭帝 (二) 波囉民 (彌忍反) 揭帝 (三) 波 囉若 (若冶反) 他 (四) 莎訶(五)
Spell of the great mind of insight. (prajñā-mahāhṛdaya-dhāraṇī). (16th)
The dhārāṇī says: tadyathā (1) gate gate (2) pāragate (3) pārasaṃgate (4) bodhi (5) svāhā (6)
This is the spell of great heart. Employ the gesture of great heart. Make many magic circles empowering all on the same basis.
Spell of the great mind of insight. (prajñā-cūla-hṛdaya-dhāraṇī).The dhāraṇī says: tadyathā (1) gate gate (2) pāramī (彌忍反) gate (3) paraya (若冶反) tā (4) svāhā (5)
Use the mudra of the lesser mind on the same basis.

Our focus has been on the Chinese rendering of svāhā. I noted that the canonical Heart Sutra has the problematic phrase: 僧莎訶 sēng shā hē. Now, in the Dhāraṇīsamuccaya we see that Atikūṭa transcribes svāhā as 莎訶, just as Kuījī's text does.

Note also that if the dhārāṇī comes from the Dhāraṇīsamuccaya then it may have lent its name to the Xīnjīng. It is called 般若大心陀羅尼  Bōrě-dàxīn-tuóluóní = Prajñā-mahā-hṛdaya-dhāraṇī.


The obvious question is, did Atikūṭa copy this dhāraṇī from the Heart Sutra or did the "author" of the Heart Sutra borrow from the Dhāraṇīsamuccaya? The Dhāraṇīsamuccaya does associate the dhāraṇī with a minimal context, in that it is associated with a mudrā and empowerment (adhiṣṭhāna).
Employ the mahāhṛdaya-mudrā. Make many maṇḍalas empowering (adhiṣṭhāna) all on the same basis.
The language used here suggests a tantric context, though, at best, it is partial. Looking at the Heart Sūtra we usually presume that the epithets section is what contextualises the dhārāṇī. As discussed above, I have shown that this is not the case and the epithets are not a reference to the dhāraṇī, which stands alone (decontextualised) in the Heart Sutra. This leads me towards seeing the Dhāraṇī-samuccaya as the immediate source of the dhāraṇī (though as a samuccaya or "anthology", it cannot be the original source).

One thing to note is that the early Tang appears, from looking at the character changes here, to be a time of phonetic change. For example, Kumārajīva has 般若波羅蜜 (MC banya baramit) allowing the final consonant of 蜜 to carry the sound of , whereas Xuanzang opts for 般若波羅蜜多 (banya bara mitda) suggesting that the final consonant is on its way out already. It is completely absent in Mandarin: 蜜 mì.

If we accept the Dhāraṇīsamuccaya as the source then we can make fairly straight-forward stemma diagram by looking for single character additions or substitutions. Note that this diagram indicates sequence rather than chronology. I will discuss chronology below.

↙                 ↘
Woncheuk         popular 
莎婆呵             薩婆訶
Huilin (T 250)                   
(807 CE)                     

This gives us a neat explanation of the variant readings. The Dhāraṇīsamuccaya transcription 莎訶 shāhē (MC saha) is transmitted directly to the text used in Kuījī's commentary. At some point a scribe makes a (haplographic) mistake, adding an extraneous character from earlier in the dhāraṇī giving us 僧莎訶 sēngshāhē (MC seung saha). This version must circulate and is adopted as the preferred reading in the Taishō, but the editors note other editions consulted have the popular transcription. It's not the weirdest thing in the Taishō, but given how popular the text is and which version is chanted, one wonders why they adopted the defective reading for their main text.

The Fangshan Stele, the oldest dated Heart Sutra (661 CE), has a three character transcription of svāhā, i.e. 莎婆訶 shāpóhē (MC sabaha). This was created by inserting 婆 (ba) between 莎 and 訶, to better represent svā. As noted, using two characters to represent Sanskrit conjuncts is common. 

Woncheuk's text alters the final character: 訶 → 呵.  i.e. 莎婆呵 shāpóhē (MC sabaga or sabaxa). Again, this change may have been inspired by changes in Middle Chinese pronunciation.  The pronunciation of 呵 was probably already moving towards the modern pronunciation of . On the other hand, the initial 莎, which I conjectured was changed for the same reasons, remains in this version. We know that different regions experience different sound changes and different rates of change, so perhaps some of that is reflected here.

The popular version of the Heart Sutra is created from the Fangshan version by substituting the initial character 莎 → 薩. This may be because the pronunciation of 莎 was moving from sa towards the modern shā and was starting to sound wrong. 

Amoghavajra also wanted a three character transcription to better capture the consonant cluster svā and opts for 娑嚩賀 suō pó hè, which is an entirely new transcription.

It would be interesting to see where the other character variations showed the same pattern (I may do some more on this).


If we accept this explanation of the variants then we can narrow the window during which the Heart Sutra was composed to 654 – 661 CE. The terminus post quem is now defined by the translation of the Dhāraṇīsamuccaya in 654 and terminus ante quem is still the date on the Fangshan Stele, 13 March 661.

These dates are at least not inconsistent with the earliest literary reference to the Xīnjīng. As I wrote a year ago:
On the 5th day, 12th Month of Yǒnghuī 6 (永徽六年十二月五日) i.e., 6 January, 656, Xuánzàng sent a letter to the emperor celebrating the birth of a new prince the month before. He wrote in a letter: "I dare to offer a copy of the Prajñā Heart Sutra in gold letters, one scroll and a case." (輒敢進金字《般若心經》一卷并函 T 50.272b.12). This piece of information is quite a big deal because this is the earliest literary reference to the Heart Sutra with a precise date that I know of. It is five years before the earliest physical evidence (661 CE) and definitely before Xuánzàng began translating Prajñāpāramitā texts in ernest (660 CE). (Critical Thinking in Buddhist Historiography)
The Taishō edition has added punctuation here, and it's quite likely before they got hold of it, the text read 金字般若心經, i.e. "gold-lettered Prajñā-hṛdaya-sūtra". The Biography by Huili and Yancong records Xuanzang translating a text called 金剛般若經 (i.e. Vajracchedikā Sūtra) for Taizong around the time he died (T 2053. 50.259a13-a28). And I note the graphical similarity to the name of the text we have just been discussing, i.e. 金字般若心經. Indeed, in the previous essay on my blog, I noted that in Chinese, vajra is translated as 金剛 (jīngāng), i.e. "gold hard" (not that gold is particularly hard).

This might not be such an issue, except that we know that the Xīnjīng was a synthetic construction and that the Sanskrit text is a Chinese forgery. Thus, what appears to be a "gold-lettered Heart Sutra" could conceivably have originally been a reference to the Vajracchedikā.

This sequence is, of course, conjectural. Occam's razor gets us to the most likely answer, not the truth. And we have to be careful about drawing further inferences. For example, we may think it likely that the Xīnjīng evolved in this way because each step only requires a single character change.

僧莎訶 ← 莎訶 → 莎婆訶 → 莎婆呵 &  薩婆訶

In addition, we associate the texts with various people or objects, e.g. Kuījī, Woncheuk, and the Fangshan Stele. We would be mistaken, however, to infer that Kuījī was commenting before the Fangshan Stele. The only inference we can draw is that the text he was commenting on predated the Fangshan Stele. Similarly, the fact that Woncheuk was commenting on a later iteration is not an indication of even the relative chronology of the commentaries.


A forensic comparison of the Heart Sutra continues to unexpectedly throw up interesting and previously unknown data points which contribute to our understanding of the history of the text. Imagine what is lurking in our texts waiting to be discovered! 

I think we can tentatively say that the Dhāraṇīsamuccaya is the likely source of the dhāraṇī. The fact that assuming this helps to explain the variant transcriptions of svāhā by linking them through single character changes is significant. A lot more painstaking work will be required to see if other variant readings. But if it's true, then we can say that the Heart Sutra was composed 654 – 661 CE.

At this point we have a sequence of development, but not an absolute chronology. 



Abé, Ryūichi (1999). The Weaving of Mantra: Kūkai and the Construction of Esoteric Buddhist Discourse. New York: Columbia University Press.

Attwood, Jayarava. (2017). ‘Epithets of the Mantra’ in the Heart Sutra. Journal of the Oxford Centre for Buddhist Studies, 12, 26–57. http://jocbs.org/index.php/jocbs/article/view/155

Glucklich, Ariel. (1997). The End of Magic. New York, Oxford University Press.

McRae, John R. (1988). "Ch'an Commentaries on the Heart Sûtra: Preliminary Inferences on the Permutation of Chinese Buddhism". Journal of the International Association of Buddhist Studies 11, no. 2: 87-115. Online: http://www.blogger.com/blogger.g?blogID=19327107


17.8.2019. 僧莎訶 is only found in T251 and a later text T 849 (a tantric text related to the Mahāvairocana cycle of texts). 
Related Posts with Thumbnails