What is bhakti?

In the Rigveda, the word bhakti appears once.

It is at 8.27.11, in the dative, and a search of the text for the letters b-h-a-k-t-i will not find it. The dative of the stem is bhaktaye, so the string never occurs. Hermann Grassmann’s Wörterbuch zum Rig-Veda, the standard nineteenth-century lexicon of the hymns, published at Leipzig in 1873, lists that single citation and glosses the word without ornament: Vertheilung, Austheilung. Distribution. Apportionment.

Not devotion. Not love. A share.

The root behind it, bhaj, does the same work throughout the hymns. At Rigveda 3.30.7 a man obtains as his portion even the unapportioned household wealth. At 1.24.5, in Stephanie Jamison and Joel Brereton’s 2014 translation, the poets ask Savitar to let them “reach upward to (a portion) of that apportioned by Bhaga”, Bhaga being the deity whose name simply is the share. The parentheses are the translators’, marking words supplied in English which the Sanskrit does not carry. Bhaga, portion, runs through the hymns in something over a hundred places. The vocabulary is one of allotment: what is divided out, and whose name is on which pile.

The root reaches further back than Sanskrit. Proto-Indo-European bhag-, to divide or allot, has among its other descendants the Greek verb phagein, to eat: literally, to have one’s share of the food. At the root of the word, receiving a portion and eating are the same act.

One thing should be said plainly before going further. The sequence this article follows, from apportionment through partisanship to allegiance and finally to love, is an organising reading. It is not a settled result. Scholars of Tamil devotion in particular would object that Tamil bhakti is a thing in its own right and not a later stage of a Sanskrit word’s career, and the historians discussed at the end of this article have spent their careers arguing against tidy single-line accounts of bhakti. The stages below are dated and sourced. The line drawn through them belongs to this article.

Painted faces at a play

The next sighting is in a grammar.

Panini, whom Oskar von Hinüber and Harry Falk place around 350 BCE on the evidence of the coin-terms in his own text, though the field’s range runs from the sixth century to the fourth, has a rule at Ashtadhyayi 4.3.95 consisting of a single word: bhaktih. It is a heading, supplying a semantic condition to the rules that follow it.

Those rules repay reading in order, because they show the range. At 4.3.96 the Kashikavritti supplies, in Srisa Chandra Vasu’s 1896 translation, apupika: “a person who loves cakes.” At 4.3.97 Panini’s own word is maharaja, one who venerates the great king. And at 4.3.98 he names two figures outright: vasudevarjunabhyam vun, the affix coming after Vasudeva and Arjuna in the sense of “this is his object of veneration.”

Whether that Vasudeva is a god is a separate question, and Panini is not the one who answers it. The Kashikavritti, nearly a thousand years later, says the name is that of a particular deity and not the designation of a kshatriya, reasoning that a famous kshatriya would already fall under the next rule. Patanjali may say something similar, but Kielhorn argued in 1908 that the phrase in question is a false reading, and it has been fought over in print ever since, Grierson and Bhandarkar defending the god, Mazumdar and others refusing him.

So the object of this bhakti may be a city, a king, a pastry or, if the commentators have it right, a god. What the heading does not supply is an inward state. The relation it governs is one of attachment or allegiance, and all of those objects sit on one list.

The sharpest early evidence is in Patanjali’s Mahabhashya, the great commentary on Panini, around 150 BCE. Patanjali is solving a grammatical problem about tense, and reaches for performers, pictures and reciters, all of which make a past event present. Then he writes:

kecit kamsabhakta bhavanti kecid vasudevabhaktah | varnanyatvam khalv api pushyanti | kecid raktamukha bhavanti kecit kalamukhah

Some become bhaktas of Kamsa, some bhaktas of Vasudeva. They even take on different colours: some become red-faced, some black-faced.

It is tempting to read this as an audience at a performance of the Krishna story, taking sides and painting their faces to show which side they are on, and that is how this article is inclined to take it. The inclination should be marked as one. The passage is among the most disputed in the study of early Indian drama. Weber, Lüders, Keith and Hillebrandt have disagreed over whether it attests staged performance at all, whether the shaubhikas are actors or shadow-play showmen, and whether the parties that divide are the audience or the reciters; Keith gives the face-colours to the reciters rather than the crowd. The colours themselves are a textual variant: Kielhorn prints red first, as above, while Keith adopts the reverse against the bulk of the manuscripts, because his reading of the passage as vegetation ritual needs red to belong to Krishna’s party. What the Sanskrit licenses on any of these readings is narrower and still useful. Bhakta here forms a compound expressing an objective relation, kamsasya bhaktah, Kamsa’s partisans, and it is used of two rival groups at once.

Around the same period the archaeology gives a name. At Besnagar in central India a pillar raised about 113 BCE by Heliodorus, a Greek ambassador from Taxila, carries a dedication to Vasudeva, devadeva, the god of gods. Heliodorus calls himself a bhagavata, which is a term of sectarian affiliation, one belonging to the Bhagavat, rather than a word for an inward state. The inscription attests the Vasudeva cult. It does not tell us what bhakti meant.

The Upanishad that says it once

The verse usually produced as the first religious use of the word is the last verse of the Shvetashvatara Upanishad, 6.23. The text is dated variously: Gavin Flood puts it in the fifth or fourth century BCE, Patrick Olivelle more cautiously in “the last few centuries BCE.”

yasya deve para bhaktir yatha deve tatha gurau | tasyaite kathita hy arthah prakashante mahatmanah

Max Müller’s translation, from the Sacred Books of the East in 1884:

“If these truths have been told to a high-minded man, who feels the highest devotion for God, and for his Guru as for God, then they will shine forth,—then they will shine forth indeed.”

Read what the sentence does. It is a condition on transmission, standing in a pair with the verse before it, which forbids giving the doctrine to one whose passions are unsubdued, or who is not a son or a pupil. It is a rule about who may be taught. It does not say bhakti is a path.

Müller’s own comment is worth having, because it is more interesting than the position usually attributed to him. At page xxxiv of his introduction he writes that the Upanishad, “so far from inculcating bhakti, in the modern sense of the word, it never mentions that word, except in the very last verse, a verse which, if necessary, certain critics would soon dispose of as a palpable addition.” He then asks: “Does that prove the existence of Bhakti as we find it in the Sâṇḍilya-sûtras?” His argument is about meaning rather than manuscripts. He neither asserts nor rejects the interpolation. He is saying the word is there and the thing is not.

Note also that Müller renders para bhakti as “the highest devotion,” which is itself the reading under dispute.

What the Gita asks for

With the Bhagavad Gita, somewhere between 200 BCE and 200 CE, bhakti becomes a way of life with a doctrine attached.

The famous verse is 9.26, and its radicalism is economic. A leaf, a flower, a fruit, water: whoever offers these to Krishna with devotion, he accepts and eats. Against the world of the great sacrifice with its priests and its cattle and its cost, the price of access drops to whatever is growing outside the door. Note what is being offered. Food.

Three verses later comes something stranger. At 9.29 Krishna says he is the same in all beings, that none is hateful to him and none dear, and then says that those who worship him with devotion are in him and he in them. The Sanskrit turns on one adversative particle, tu. God is impartial, and yet. The commentators have argued about that yet ever since.

And at 9.32 comes the verse a modern reader has to be given rather than spared. Krishna says that those who take refuge in him reach the highest goal, and then names them: papa-yonayah, those of sinful birth, along with women, vaishyas and shudras.

The grammar itself is contested. Read one way the list is appositional, and the women, vaishyas and shudras are the sinful-born. Read the other way, which is how Ramanuja’s tradition takes it, the enumeration is separate: women, vaishyas and shudras, and in addition those of sinful birth. The difference decides whether the verse insults the people it admits, and it has not been settled. The site does not adjudicate it. It does think a cornerstone that quotes 9.26 and hides 9.32 is advertising.

Most of what the Gita asks for is decided rather than felt. Man-mana bhava mad-bhakto mad-yaji mam namaskuru, at 9.34 and again at 18.65: fix your mind on me, be my devotee, sacrifice to me, bow to me. Where affection does appear it is marked. At 10.10 Krishna speaks of those who worship him priti-purvakam, with affection going before, and sa me priyah, he is dear to me, returns as a refrain through the later verses of chapter 12.

One caution about vocabulary. Describing bhakti as one of three paths, beside knowledge and action, is a later systematisation, and its popular currency owes much to Vivekananda, whose own nineteenth-century scheme was in fact fourfold, adding raja-yoga. The three-path scheme is a useful map. It is not the Gita’s own architecture.

The nine things to do

The most widely repeated scheme in all of bhakti is a list of nine, and almost everything popularly said about it is slightly wrong.

In the Bhagavata Purana at 7.5.23 the boy Prahlada names them: shravanam kirtanam vishnoh smaranam pada-sevanam, archanam vandanam dasyam sakhyam atma-nivedanam. Hearing of Vishnu, singing of him, remembering, attending his feet, worshipping, saluting, serving as his slave, as his friend, and offering up the self.

Two things about that verse repay attention. Seven of the nine are verbal nouns in -ana, names of things done; dasya and sakhya are abstract nouns of relation, servanthood and friendship. And 7.5.24 puts the whole in a conditional with the optative passive kriyeta: if bhakti of nine marks were performed in this way, that is what the speaker considers the best learning. It is not a command. It is still a verb.

And the Bhagavata does not call them navadha bhakti. Its own word at 7.5.24 is nava-lakshana, having nine marks. The word navadha occurs three times in the whole Purana, never about devotion: twice in the Kardama and Devahuti narrative, a ninefold birth and a ninefold self-division, and once at 12.11.31, where Hari is said to be described ninefold by ajaa, the unborn, which the commentators gloss as maya. The third is easy to miss, because sandhi hides it inside navadhoktah, and a search for the letters will not turn it up. The label by which the scheme is universally known today is not the Bhagavata’s, and it is familiar above all from Tulsidas, who writes navadha bhagati in Awadhi in the Ramcharitmanas, in the Aranya Kanda, in Rama’s instruction to Shabari. And his nine are not the Bhagavata’s nine. Tulsidas begins with satsanga, the company of the good, which is a social precondition rather than an act of worship, and in the seventh has Rama count the saints as greater than himself. The Sanskrit list is liturgical. The Awadhi list is moral.

What the south added

The weeping comes later, and it comes from the Tamil country.

Between roughly the sixth and ninth centuries two groups of poet-saints composed an enormous devotional literature in Tamil: the canonical twelve Alvars, who sang to Vishnu, and the canonical sixty-three Nayanmars, who sang to Shiva. Their work is not a development of the Gita’s temperate discipline. It is longing, jealousy, sulking, love-sickness, the grammar of Tamil love poetry turned on a god.

How old it is, we cannot say with precision. Valérie Gillet, introducing a 2014 volume devoted to exactly this problem, states it flatly: there is no firm evidence, internal or external, for an unquestionable chronology of Tamil bhakti, and none of the texts can be assigned a definite date. There is a general consensus pointing to the sixth century. There is nothing firmer than consensus.

The gap between traditional and scholarly chronology here is not a rounding error. Traditional Shrivaishnava reckoning places the first Alvars in 4203 BCE. The epigraphic and literary evidence is generally read as pointing to the sixth century of the Common Era. Nearly five thousand years separate the two accounts, and they are not reconcilable, because they disagree about what these figures fundamentally are.

What can be shown is that the Sanskrit tradition knew where the fire was. The Bhagavata Purana, composed perhaps in the ninth or tenth century, says at 11.5.38 to 40 that in the Kali age devotees of Narayana will be found here and there, kvacit kvacit, “and in the Dravida lands abundantly,” and names among its rivers the Tamraparni, the Kritamala and the Kaveri. A Sanskrit text pointing south, by name, at the water.

Friedhelm Hardy built a large thesis partly on that observation in Viraha-Bhakti (Oxford, 1983): that emotional Krishna devotion is a southern development which the Sanskrit Bhagavata absorbed and carried north. It remains contested. Tracy Coleman has argued against it directly, holding that highly emotional bhakti is already present in Sanskrit literature before the Alvars, and the old anomaly that neither Yamuna nor Ramanuja ever cites the Bhagavata has not gone away. It is the explanation this article finds most economical for why a Sanskrit purana is naming Tamil rivers.

Something you feel, or something you do

Here the tradition breaks its own tidy picture.

The Shandilya Bhakti Sutra, a compact treatise of a hundred aphorisms whose date is not fixed, raises the question directly. Is bhakti an action? Because if it is, a difficulty follows: the Veda says liberation does not come by works. Sutra 1.1.7 answers in four words.

na kriya krityanapekshanaj jnanavat

In E. B. Cowell’s translation of 1878: “It is not an action, for, like knowledge, it does not depend on effort.” Svapneshvara’s commentary gives the reasoning:

“This faith cannot properly be an action, because it does not follow an effort of will. Whatever does not follow an effort of will, is not an action, as for instance knowledge ; for knowledge depends on the presence of the proper evidence, and a man cannot, at his pleasure, produce it or not, or alter it.”

Cowell renders bhakti as “faith” throughout, which is his choice and not the only one. The argument stands whichever word is used. You cannot produce an attachment by deciding to.

Now set beside it the most systematic scholastic treatment bhakti ever received. Rupa Gosvami composed the Bhaktirasamritasindhu at Gokula in 1541, a date the text supplies itself in a closing chronogram. At 1.1.10 he announces that he is about to give a lakshana, a formal definition in the technical style of Sanskrit philosophy. Then, at 1.1.11:

anyabhilashita-shunyam jnana-karmady-anavritam | anukulyena krishnanushilanam bhaktir uttama

The highest bhakti is anushilana directed toward Krishna, favourably disposed, empty of any other desire, and uncovered by knowledge, ritual action and the rest.

Anushilana means cultivation. Habituation. Sustained practice. It is from the root shil, cousin to shila, character formed by discipline. It is not prema, love, nor rati, attachment, nor bhava, feeling. Defining the summit of devotion, Rupa reaches for the vocabulary of training.

He means it. His vaidhi bhakti, at 1.2.6, is defined as engagement arising because raga has not been attained, by the command of scripture alone. It is what you do when you do not feel it. He lists its limbs and totals them at 1.2.95: sixty-four. The count belongs to that rule-governed practice specifically, not to devotion as such. And when he comes to raganuga bhakti, the kind that follows spontaneous passion, he says at 1.2.296 that the limbs stated under vaidhi apply there too. Same practices, different engine.

Even the text most often produced as evidence for pure feeling declines to cooperate. The Narada Bhakti Sutra defines bhakti at sutra 2 as parama-prema-rupa, having the form of supreme love. At sutra 76 it instructs that treatises on devotion be studied and the acts that awaken devotion performed. Both sutra collections are pseudepigraphic. J. N. Farquhar placed both in his 1920 chapter on bhakti, A.D. 900 to 1350, observing that each depends on the Bhagavata and that neither mentions Radha, though he called the Shandilya’s affiliations still obscure.

Where the schools break

Put the question to the commentators and four incompatible answers come back.

Shankara absorbs bhakti rather than dismissing it. Commenting on Gita 18.55 he glosses bhaktya as jnana-nishtha, steadfastness in knowledge, and concludes that this is the supreme kind of devotion the Gita means. The structural argument is in his Brahmasutra commentary at 1.1.4: meditation depends on the meditating person and so “may either be performed or not be performed or modified,” whereas knowledge “can therefore not be either made or not made or modified.” Liberation is not a product, so a performance cannot be its direct cause. On Gita 11.54, in fairness, he allows that undivided bhakti enables one to realise and enter God. He is consistent about where he lets bhakti terminate.

Advaita is not one voice on this. Madhusudana Sarasvati, writing in the later sixteenth century, was an Advaitin who argued in his Bhaktirasayana that bhakti is itself the highest human end. Lance Nelson has argued that he takes much of it back in his Gita commentary, where devotion is subordinated to knowledge again.

Ramanuja makes bhakti a cognitive discipline: the word is a synonym of upasana, meditative worship, and meditation is an unbroken stream of remembrance, like a stream of oil. His treatment of Gita 18.66, the verse his tradition calls the carama-shloka or final word, is where the trouble starts. Abandon all dharmas, come to me alone for refuge. In his own commentary Ramanuja gives one reading, writes atha va, “or else,” and gives another. On the first, you continue your disciplines and abandon only the fruit, the ownership and the agency. On the second, sarva-dharman means the impossibly numerous expiatory rites no finite life could complete, and you abandon those and perform prapatti, self-surrender, so that devotion may begin.

His school later divided over which of the two readings to canonise. The Vadakalai position holds that some act on the soul’s part is required; the Tenkalai position holds that God saves on his own initiative. The doctrinal divergence is medieval, running between Pillai Lokacharya and Vedanta Desika in the thirteenth and fourteenth centuries. The two named sub-sects are later. On Srilata Raman’s account the labels come into use from the late seventeenth century and harden into rival sects only under nineteenth-century litigation in the British courts.

The famous shorthand is later still. The images of the cat, who carries her kitten, against the monkey, whose infant must hold on, appear in Monier-Williams’ dictionary, which cites as its authority nothing but his own English book of 1883. Raman thinks the simile of relatively recent coinage, and traces the earliest mention she can find to a Tamil text, the Palanatai Vilakkam of Tirumalicai Anna Ayyangar, no earlier than the nineteenth century. The divergence is medieval. The cat and the monkey are modern.

Madhva will not let devotion dispense with knowing. His definition, at Mahabharata-tatparya-nirnaya 1.86, is love preceded by knowledge of God’s greatness, firm, surpassing everything else, and by that alone is there liberation, and not otherwise.

And the Gaudiya school of Bengal, which formed around Chaitanya (c. 1486 to 1534) and was given its theology by the six Gosvamis of Vrindavan after his death, inverts the scale entirely. For Krishnadasa Kaviraja, prema is a fifth human goal above moksha itself, and in the Chaitanya-charitamrita the bliss of liberation does not amount to a drop of it. Rupa lists among bhakti’s marks that it makes liberation look small. At 1.2.22 he goes further:

bhukti-mukti-spriha yavat pishachi hridi vartate | tavad bhakti-sukhasyatra katham abhyudayo bhavet

So long as the witch of desire for enjoyment and for liberation dwells in the heart, how can the happiness of devotion arise there?

The desire for liberation sits in the same sentence, under the same word, as the desire for pleasure. Both are the witch.

Devotion without a form

A definition of bhakti as devotion to a personal God would exclude some of the most widely sung devotional poetry in India, and it excludes it on a principle the poets themselves attacked.

The traditions distinguish saguna devotion, directed to the divine with qualities and form, from nirguna devotion, directed to the divine without them. David Lorenzen puts the paradox plainly, introducing a volume on bhakti religion in north India: this is heartfelt devotion to a God with no attributes and no definable personality at all. The distinction is not an Orientalist import. John Stratton Hawley has traced its work in the early manuscript anthologies of Hindi devotion, in the same volume Lorenzen edited.

The nirguna poets are not marginal. Kabir, the weaver of Banaras, and Ravidas, the leather-worker, both have compositions in the Guru Granth Sahib: Ravidas contributes forty or forty-one poems, by the usual counts, and is one of thirty-six named contributors to the Sikh scripture. Their devotion refuses form and refuses the apparatus that comes with form. Hawley quotes a Kabir poem whose last line prefers hell to Kashi. He gives it through Purushottam Agrawal, who rejects reading it as a nirguna preacher scoring off the holy city, and Hawley’s own note allows that the line may run differently again. The refusal is there. What it refuses has been argued about.

This matters for how the word itself has been defined. Krishna Sharma argued in 1987, from a doctoral thesis submitted at London in 1964, that the identification of bhakti with Vaishnava monotheism is an artefact of nineteenth-century European scholarship, and that an impersonal God, a non-dual account of reality and an emphasis on knowledge can be the legitimate constituents of a bhakti tradition too. Her target was the assumption, which she traced to R. G. Bhandarkar and to Wilson, Weber, Lorinser and Monier-Williams, and above all to Grierson, that devotion requires a person to be devoted to.

What devotees actually do

None of the above is what a reader will see on a Tuesday evening in a temple.

Worship begins with darshana, from the root drish, to see. The devotee goes to see the deity, and the deity sees back. Diana Eck’s study of the subject turns on that reciprocity: the gods are held to keep their eyes open. The same word names the six schools of Indian philosophy, each of them a way of looking.

Puja itself is shaped as hospitality. The sixteen services, in P. V. Kane’s reckoning, run: invite the deity, offer a seat, water for the feet, water of welcome, water for sipping, a bath, garments, the sacred thread, sandal-paste, flowers, incense, a lamp, food, obeisance, circumambulation, and finally leave to depart. Kane adds in the same breath that in different works the items differ, and Gudrun Buehnemann puts it harder: there are hardly two lists of upacharas that are identical. Paul Thieme argued, in a 1957 paper on the word itself, that the whole structure derives from the Vedic reception of a guest. Kane resists the cognate claim on the very next page from the one quoted here: an extension, he says, and not a borrowing from an alien cult. In a temple the deity is a resident rather than a guest, and so is woken in the morning instead of invited.

The food offered is naivedya. What is returned, having been taken and tasted by the deity, is prasada, and it is materially the same food and ritually a different thing.

Here an attractive idea has to be given up. Prasada does not belong to the vocabulary of shares. It is from pra plus the root sad, to settle, and its semantic centre is clarity, which is Monier-Williams’ first sense, then serenity, then graciousness; Apte’s dictionary, which leads with favour and kindness, lists the edible sense sixth. The returned-portion logic is carried by other words, chiefly yajna-shishta, the remnant of sacrifice, which Gita 3.13 says those who eat are freed from fault, and ucchishta, the leftover, which the Atharvaveda devotes an entire hymn to deifying. The share sense is already inside bhakti itself, and it does not need prasada to carry it.

What prasada does is worth more than the etymology anyway. At Jagannath in Puri the offered food is eaten together at the Ananda Bazar without regard to caste, suspending for as long as the meal lasts the commensality rules that organise nearly everything outside it. The suspension has limits worth stating. The temple admits only Hindus, and the two refusals people remember are not simply that rule at work: when Gandhi came to Puri in 1934 the question in play was the admission of Harijans, and Ambedkar was refused in 1945.

Then there is sound. Kirtan is congregational and antiphonal, a call and a response; bhajan can be one voice alone. Chaitanya’s contribution to Bengal was public sankirtana, singing in the street. The eight verses of the Shikshashtaka are the one composition the tradition ascribes to him, and in them he writes that there is no rule of time for remembering the divine names. The ascription is older than the collection: Rupa Gosvami already has all eight in the Padyavali, singly, out of order and filed under ordinary topical rubrics, though each of them already ascribed to him, and they are first gathered into a set in the Chaitanya-charitamrita. S. K. De allowed the eight verses as the possible exception to a blunter judgement, that attributing any specific work to Chaitanya is more a matter of pious belief than a positive historical fact. The Varkaris of Maharashtra recite the Haripath daily and walk to Pandharpur. A name costs nothing and requires no priest, which is why singing is the practice that opens furthest.

The saint who entered as a brahmin

The claim most often made for bhakti in the modern period is social: that it dissolved caste.

The poetry supports it. Appar opens a hymn nam arkkum kutiyallom, which Indira Viswanathan Peterson renders “We are slaves to no man, nor do we fear death”; the tradition makes it his reply to a summons from the Pallava court. Devara Dasimayya asks whether the outcaste, last in line, held with his outgoing breath the stick of his tribe. Akka Mahadevi, in A. K. Ramanujan’s translation, says of herself that she has lain down with the Lord and lost caste.

Then read what the hagiographies do with their outcaste saints.

Tiruppan Alvar, of the Panar, sings to Ranganatha from the far bank of the Kaveri because, the hagiographies hold, he may not approach. What happens next comes down in two tellings, and they are not the same story. In Garudavahana Pandita’s Sanskrit Divyasuri Caritam there is no stone and no dream at all: Lakshmi hears the singing and says the singer is not fit to be kept outside, Ranganatha tells the sage Loka Saranga to seat him on the chariot of his shoulders, the Alvar is unwilling, and the sage carries him by force. In the later Tamil Guruparamparai the sage comes to the river for water, throws a stone that draws blood from the Alvar’s forehead, the sanctum doors shut, and the Lord gives the shoulder-command in a dream that night. Most retellings run the two together.

Either way he is carried, so that his feet never touch the temple ground. Eleanor Zelliot reads that as the device it looks like: the devotee is granted his sight of the Lord, the floor stays unpolluted, and the brahmin takes the pollution onto himself. A commentarial tradition reads it the other way and calls him munivahana, he who had a sage for his mount; Vedanta Desika wrote a commentary on his hymn under that name. Both things are true of it at once: the ground stays unpolluted, and a brahmin carries an outcaste into the sanctum on his shoulders.

Nandanar, in Sekkizhar’s twelfth-century Periya Puranam, lives among the Pulaiyar farmhands and supplies the leather and the gut for temple drums, and he cannot enter the temple at Chidambaram. The Lord tells him in a dream to enter fire, and tells the brahmins of Tillai to prepare it. They are struck with fear and say they will do as they are bidden. He walks in and rises from the pit with matted hair and a sacred thread on his chest, goes with the brahmins to the hall where the Lord dances, and is seen no more. The outcaste saint is admitted after being remade as a brahmin, in a fire prepared on the god’s own instruction by brahmins who were afraid of it. Lynn Vincentnathan, collecting versions of the story around Chidambaram, found the two communities telling it differently: caste Hindu versions treat Nandanar as an anomaly the fire resolves, a brahmin or a god somehow encased in an untouchable body, while Paraiyar tellings have no difficulty believing in the piety of an untouchable who stayed one, because for them there is nothing to explain.

In 1936 Gandhi argued that the scriptures should be read as the saints had read them, and asked whether a religion professed by Chaitanya, Jnanadev, Tukaram, Tiruvalluvar, Ramakrishna Paramahansa, Raja Ram Mohan Roy, Devendranath Tagore, Vivekananda “and a host of others who might be easily mentioned” could be so utterly devoid of merit as Ambedkar had made out. B. R. Ambedkar answered him, and the answer was published as an appendix to the second edition of Annihilation of Caste in 1937:

“Firstly, none of the saints ever attacked the Caste System. On the contrary—they were staunch believers in the System of Castes.”

“They did not preach that all men were equal. They preached that all men were equal in the eyes of God—a very different and a very innocuous proposition, which nobody can find difficult to preach or dangerous to believe in.”

The masses, he adds, were taught that a saint might break caste but the common man must not, so the saint never became an example and remained a pious man to be honoured.

Two things have to be said alongside that. Ambedkar was answering Gandhi, and the saints he actually names in that passage are Marathi: Jnanadev, who he says moved heaven and earth to have his brahmin status recognised, and Eknath, who dined with untouchables and did it, on Ambedkar’s reading, not out of any opposition to caste but because he felt the pollution could be washed away by a bath in the Ganges. He is not analysing Sekkizhar or the Alvars. Nor is he answering the whole of the list he was given: half of Gandhi’s names are nineteenth-century reformers rather than medieval saints, and the reply narrows the ground to the saints before it answers on it. And in 1954 he declared himself a devotee of the Buddha, of Kabir, who is a bhakti poet, and of Phule. His quarrel was with what the saints were being made to prove, and the sweep of “none of the saints ever” sits in tension with the verses quoted a few lines above, though not with all of them: a poem saying the breath carries no caste is exactly the equality in the eyes of God that he calls innocuous.

Gail Omvedt read the same material and concluded differently. In Seeking Begumpura she takes Ravidas’s Begumpura, the city without sorrow, with no taxes and no toil, no property, no king and no temple, where all are equal and anyone may walk anywhere, as genuine political imagination rather than consolation, and reads the anti-caste saints as a continuous intellectual tradition running down through Phule and Periyar to Ambedkar himself. On her reading the poetry is doing the work Ambedkar says it failed to do, and doing it in a register politics had not yet invented.

One point belongs to the record rather than to either side. The oppression modern audiences remember from the Nandanar story, the cruel landlord and the impossible task, is largely absent from Sekkizhar. It enters with Gopalakrishna Bharati’s musical drama, printed in 1861.

A phrase with a history

There is one more thing to say, and it concerns the words “bhakti movement.”

In April 1907, in the Journal of the Royal Asiatic Society, George Grierson wrote:

“Suddenly, like a flash of lightning, there came upon all this darkness a new idea. No Hindu knows where it came from, and no one can date its appearance; but all the official writings which describe it and which can be dated with certainty were written long after the Christian era. This new idea was that of bhakti.”

Two things sit just outside that quotation. The first is what Grierson says next: “Religion was now no longer a matter of knowledge. It became a matter of emotion.” And then, over the page, that bhakti “requires a personal, not an impersonal, God.” That is the assumption Krishna Sharma spent a book attacking, stated in one line by the man she was attacking. The second is the essay’s actual thesis, announced in its first paragraph: that Hindu devotion was absorbed from the early Nestorian Christians, whom he localises a page later in southern India. It is a position Hawley notes Grierson himself later modified.

Grierson called it a reformation and, elsewhere in the same essay, a revolution. He did not call it a movement.

The English words were already loose. Sophia Dobson Collet, writing about the Brahmo Samaj of India in 1873, could already call its first noteworthy phase “what has been called ‘the Bhakti movement’”, and she made the phrase a section heading in the Brahmo Year-Book for 1877; Farquhar uses the phrase of Maharashtra and of Gujarat in the 1920 Outline, and F. E. Keay of Hindi poetry in the same year. What was missing was not the words. It was the idea that they named one thing, moving across a subcontinent. That is what came through Hindi. Hazariprasad Dvivedi carried Grierson into Hindi across three books between 1936 and 1942. In the Bhumika of 1940 he takes up Grierson’s article on bhakti for Hastings’ Encyclopaedia of Religion and Ethics, and there Grierson’s “revolution” becomes andolan. In Kabir of 1942 it is the lightning passage above, and there it reads yah bhakti ka andolan hai: this is the bhakti movement. A descriptive noun had become a historical actor. Ramchandra Shukla’s periodisation of Hindi literary history supplied the dates, and andolan entered his own account only with the revised edition of 1940. The dates were then read backwards as a social programme.

John Stratton Hawley traced this in A Storm of Songs (Harvard, 2015), and he is careful in a way worth preserving. He does not argue that bhakti was invented, or that the Alvars did not exist, or that nothing happened. He argues that the unified, pan-Indian, single movement framing is largely modern, and prefers to speak of a network. There is a fine irony in his findings. Krishna Sharma had attacked Grierson nearly three decades earlier for the phrase “bhakti movement.” Hawley reports that the words are nowhere to be found in Grierson’s prose, and that he could not trace the concept to any of his writings either. He wonders whether she had been reading Grierson through Dvivedi.

The verse usually quoted for the movement’s geography belongs to the same story. In the Bhagavata Mahatmya, which is attached to the Padma Purana though the oldest manuscripts carry it as a text of its own, and which Hawley dates to the latter half of the seventeenth century or the early years of the eighteenth, on dated manuscripts that burst onto the scene in 1714, Bhakti appears as a woman and gives her life: born in the Dravida country, grown in Karnataka, here and there in Maharashtra, and grown old in Gujarat, where heretics cut her limbs. Then she reaches Vrindavan and becomes young again.

It is usually repeated with the ending removed, and Hawley’s point is that Vrindavan is not the ending either. Reading the born-in-Dravida passage apart from the rest, he argues, is what led earlier scholars to Vrindavan; the story culminates at Haridvar, and he wants Haridvar kept in view. He declines to hand the text to the Vallabhites or the Chaitanyites, saying it is hard to read as a document belonging to either sect. What he reads in it is “a north looking self-consciously to the south for its legitimacy.” Its phrase kvacit kvacit reuses the older Bhagavata verse about the Tamil rivers. That last observation is this article’s and not his; he does not make it.

Under a vow

Which returns us to the question the sources will not settle, and to the one text that gets close.

Rupa Gosvami, having insisted that the practice is effortful and indispensable, writes at 1.2.2 that what it accomplishes is the manifestation in the heart of a bhava that is eternally already accomplished. Nitya-siddha. Always there.

On that account the discipline does not manufacture the love. It uncovers it. Which concedes Shandilya’s point, that affection cannot be produced by an act of will, without surrendering a single one of the sixty-four limbs. It is Rupa’s answer rather than the tradition’s, and Shandilya’s school would not accept it.

And there is the verse people reach for when they want bhakti wild. Bhagavata Purana 11.2.40: his heart melted, he laughs aloud, weeps, cries out, sings, and dances like a madman, heedless of the world.

The verse opens on one compound. Evam-vratah. He whose observance is thus.

The man dancing like a madman is under a vow.