Like all Hindus, somewhere deep down inside me I had assumed that Harijans (Dalits), Gandhi’s supposed “children of God” relegated to the fringes of society, were part of the Hindu community, part of “us.”
— URVASHI BUTALIA, The Other Side of Silence
The partition of British India in 1947 was accompanied by the mass murder of Hindus by Muslims, and Muslims by Hindus. It is estimated that up to a million people were killed (B. Metcalf and Metcalf 2006, 221), often by hand and at close quarters. While collecting oral histories of this vast human tragedy, historian Urvashi Butalia was astonished by evidence of Dalits who remained calm amid the violence around them. They knew no one would touch them, one Dalit woman recalls, because the conflict was between Hindus and Muslims and they belonged to neither group. Her father nevertheless urged her to stay indoors for fear that the Hindus might mistake his daughter for a Muslim (Butalia 2000, 235). Dalits displaced by the violence had no place to go, Butalia later reveals. For “in a war that was basically centered around Hindu and Muslim identities,” she explains, no one set up “camps to help [Dalits] tide over the difficult time. No recourse to government—all too preoccupied at the moment with looking after the interests of Muslims and Hindus, no help from political leaders whose priorities were different at the time” (2000, 238). Could untouchables not have gained admission to Hindu relief camps? At a time when Dalits were excluded from schools for Hindu children-because they were regarded as polluting-and were confined to separate quarters in government prisons for the same reason, this most likely would not have been possible. Even in 2004, in the wake of a devastating tsunami, Dalits were excluded from relief camps that sheltered caste Hindus (Human Rights Watch 2005, 25-29; Anand and Thangarasu 2006; Gill 2007).
Butalia is not the only late twentieth-century scholar surprised to discover that “Hindu” and “untouchable” are understood as contrasting categories by ordinary people, even in the present day. Mary Searle-Chatterjee recalls, “I could hardly believe the evidence of my ears” when she first noticed sweepers in Benares, among whom she was conducting ethnographic research in 1971, refer to “the Hindus” as other to themselves (2008, 189). These were not politicized Dalits who actively rejected the Hindu label in favor of Ambedkarite Buddhism, she explains. “This was the usage of ordinary, non-politicized sweepers” (2008, 189). They were not using the term in a segmentary sense, in which “the Hindus” means caste people in contrast to “Harijans,” and that encompasses both caste people and Dalits in contrast to Muslims. While “segmentary [terms] may be more or less inclusive,” SearleChatterjee explains, “in the case of the ‘low’ caste reference to Hindus as people other than themselves, something more is involved. Even when Muslims were present, sweepers did not shift to referring to themselves as ‘Hindus’” (2oo8, 189).
Butalia’s and Searle-Chatteriee’s sense of surprise is itself unsurprising. As historical anthropologist Joel Lee explains, it reflects the hegemony of the contemporary assumption
that “sweepers” and other Dalits. insofar as they are not formal converts to Buddhism, Islam or Christianity, belong in a taxonomical sense to the Hindu community and should therefore see themselves as Hindus. This assumption follows logically from the most basic lessons that students across the globe learn about Indian society; to be educated in the world today, whether in Tokyo or Chicago or Johannesburg or Delhi, is to know that Hindu society has or had a caste system that classed some people “untouchable,” that therefore “untouchables” belong to Hindu society, that therefore “untouchables” are Hindus… This constitutes commonsense among the educated in urban India and in the academy as well. (2015, 82)
Today this common sense is backed by the force of law. Dalits who do not specifically proclaim themselves Christian or Muslim are legally categorized as Hindu by default. Exceptions like those discovered by Searle-Chatterjee and Butalia were still common in the late twentieth century (e.g., Lynch 1969, 162-63). But in many urban settings at least—including my own field site—Dalits now accept the government’s new, more inclusive definition of Hindu as including people like themselves (Roberts 2015a).
It was not always so. Before the twentieth century Dalits were not regarded as Hindu by others, nor did they regard themselves as such. It is conceivable that exceptions may yet be found, but in the absence of clear documentary evidence it is an anachronism to label pre-twentieth-century Dalits Hindu (Searle-Chatterjee 2008, 187; Frykenberg 1997). The only significant context in which “untouchable” castes were treated as Hindu before the twentieth century was the decennial census of the colonial state, which categorized all Indians by caste and religion and which assigned those who did not proclaim themselves Muslim, Christian, or members of another recognized religion to the category “Hindu” by default. In so doing, the colonizer rejected native precedent. Census taking and other enumerative technologies were well established in India’s precolonial states and like later colonial censuses categorized subjects by caste and religion (Guha 2013; Roberts 2015b).
But these precolonial censuses did not recognize untouchable castes as Hindu. As Norbert Peabody has shown in an important paper on precolonial census taking, even as late as 1835 the Hindu kingdom of Marwar conceived the primary division among its subjects as lying, not between Hindus and Muslims, but between the so-called “clean” castes, which included both Muslims and Hindus jātis, and the impure servile castes, namely Dalits, who were understood as distinct from both (2001, 834-36). Unpublished research by Divya Cherian on the same kingdom in the late eighteenth century paints a similar picture. Official guidelines on religious duties and prohibitions in that state “categorically divided its subjects into two types: Hindus (hinduvan) and untouchables (achhep). The latter category consisted of leatherworking castes, nomadic pastoral groups, Muslims (turak), and the sanitation labor castes (halalkhor). Not only were the sweepers not Hindu, they were the antipode of the Hindu: the order made clear that what actions the state required of its Hindu subjects were precisely those that it forbade its untouchable subjects” (quoted in Lee 2015, 120). Cherian’s findings differ from Peabody’s only insofar as the records she unearthed categorize Muslims together with untouchables.
Rupa Viswanath’s research in the Madras Presidency similarly finds that the term Hindu referred, until the early twentieth century, exclusively to those jātis eligible to live in the ūr and expressly excluded those confined to the cēri, the Dalit ghetto (2014c; see also Ebeling 2010). Indeed, the association of Hindu with respectable caste status was so well established in nineteenth-century Madras that Christians and Muslims belonging to the so-called clean castes were sometimes referred to in native discourse as “Hindu Christians” and “Hindu Muhommedans,” to distinguish them from coreligionists of untouchable origin, known as “Pariah Christians” and “Pariah Mohammedans” (Rupa Viswanath, personal communication). And even as late as 1916, Gyan Pandey records that in Chhattisgarh, “to call a man a Hindu convey[ed] primarily that he [was] not a Chamar,” that is, not a Dalit (1993, 246).
British census officials departed from existing usage. Colonial observers had long stereotyped Indian subjects as divided into two distinct and antagonistic religious “communities,” Hindu and Muslim. The latter were portrayed as following the religion of “foreign” invaders who had ruled much of the Indian subcontinent since 12o6, the former as followers of India’s original religion. By playing up alleged conflict between the two, colonizers justified their own rule as bringing peace to the land and as protecting India’s disenfranchised Hindu masses (T. Metcalf 2007, 132-48). Reversing precolonial precedent, the colonial census simply lumped untouchables together with Hindus. This policy met with frequent objections by native census takers, typically high-caste Hindus, who persistently refused to record Dalits as Hindu (Lee 2015, 11o; Mendelsohn and Vicziany 2000, 27-28; Juergensmeyer 1988, 72, cited in Searle-Chatterjee 2008, 191).
Hindu census takers were not alone in rejecting untouchables. Dalits were also banned from Hindu temples; access to sacred Hindu texts was forbidden them; Hindu priests refused them. A distinctly anthropological argument could nevertheless be made for classifying untouchables as objectively Hindu, irrespective of how the classified themselves or were classified by others, on the basis of three criteria: morphological similarities between their cults and those of popular (non-Brahminical) Hinduism, Dalits’ limited participation in village religious festivals, and the fact that Dalits serve Hindus by removing ritually impure substances for them.
But it is unclear why forced ritual service to a cult implies membership in it, and by the criterion of participation many Indian Muslims and Christians are also Hindu, and vice versa (Roberts 2015a, 242-44). As for morphological similarities at the level of practice, these are common also between popular Hinduism and Islam in India, which is why Peter van der Veer, an anthropologist who has studied these extensively, argues that the only valid criteria for group membership are self-definition and acceptance by others. Morphological comparisons at the level of doctrine are reviewed by Viswanath (2o12a), who argues they do not establish common religious identity. But the classification of Dalits as Hindu for census-taking purposes was never purported to rest on objective criteria. Dalits were recorded as Hindu by state fiat. It is thus not surprising that Hindu census takers would refuse to comply with this order, only to have their surveys later “corrected” by higher-ups.
Joel Lee’s ethnographic study of the 2011 Census describes a fascinating historical reversal: the Brahmin census taker he accompanied on rounds recorded untouchables as Hindu even when they themselves told him they were not (Lee 2015, 3-10). What had changed? Since the late nineteenth century Muslims and a Hindu missionary organization known as the Arya Samaj had been competing for converts in the United Provinces and the Punjab. The Aryas focused on converting Christians, Muslims, and wayward Hindus but at this time still regarded untouchables as beyond the pale, and the one or two attempts by renegade Samajists to convert untouchables were met with a strong backlash within the organization (Jones 1976; Adcock 2007).
This began to change when the Morley-Minto reforms were announced in 1909. The franchise was extended, and representation of different communities became tied to demographics. What had been a struggle for cultural preeminence became a competition for sheer numbers (Tejani 2008, 141-43). The inclusion of untouchables within Hinduism merely for purposes of census taking suddenly had very real political implications, and Muslims began to argue that Hindus’ numbers were artificially inflated by the inclusion of untouchables (Mendelsohn and Vicziany 2000, 28; Rao 2009, 131; Sartori 2003). And all over India untouchables themselves demanded to be recognized as a separate element distinct from both groups (Sartori 2003, 171-73; D. Sen 2011; Irschick 1969, 71-71; Viswanath 2014b).
Fairness and accuracy required that the controversial policy of listing untouchables as Hindu be revisited. Census Commissioner E. A. Gait argued in 1910 that Hinduism should be conceived broadly and not be limited to those holding specific beliefs or practices. “A man may believe in the whole Hindu Pantheon, or for that matter, in no god at all—he may sacrifice or abstain from sacrifices—may eat fish and flesh, or abstain from doing so,” and still be legitimately called a Hindu, Gait argued (quoted in Mukerji 1911, v). But he observed that it was “absurd to enter without comment as Hindus persons … who are not regarded [as such] by others, and do not profess themselves to be Hindus,” as previous censuses had done (quoted in Mukerji 1911, v).
The possibility of losing a large portion of their official numbers sharpened the minds of high-caste Hindu leaders, who began at this point to vociferously claim Dalits as fellow Hindus and to accuse the colonial state of conspiring to undermine Hinduism by divide-and-rule tactics. It is true that the British would later seek to capitalize on the refusal by Dalits to recognize the leadership of the high-caste Hindu-led Congress Party (Prashad 1996). But “divide and rule” implies a prior unity, and in the case of Dalits and Hindus the evidence for any such unity is lacking. Interestingly, evidence for precisely the opposite—a lack of both common identity and regular social relations between caste Hindus and Dalits—can be found in the urtext of the argument that the British were subjecting Hindus to divide-and-rule by separating out untouchables. This was a pamphlet entitled A Dying Race, originally published in serial form by the Hindu strategist U.N. Mukerji in 1909. Even as the author accuses Gait’s memo of attempting to create a division between Hindus and untouchables, he elsewhere notes,
It will puzzle most Hindus if they are asked as to the inner life of these “low castes.” Respectable people scarcely trouble themselves about such things. There is a sort of a “Ghetto” … attached to nearly every village, far away, of course, from where the respectable classes live. Nobody belonging to the “high castes” ever thinks of visiting these quarters. Everything about the … people of that class is pollution—their touch is pollution, their presence is pollution, water touched by them is polluted, their very shadow carries infection. These people do a certain sort of work and, when their services are needed, are tolerated to that extent, but they are the “itars”—“the others”—quite apart from respectable people. At other times there is hardly any contact. (Mukerji  1929, 43)
Mukerji’s admission that for Hindus untouchables were outsiders, and that Hindus wanted nothing to do with them, is not presented as an original observation. That Hindus regarded untouchables as beyond the pale was common knowledge; Mukerji’s express purpose was to persuade them to reverse course and enlist untouchables as fellow Hindus. Nationalist scholars have nevertheless treated as an established fact the accusation that Gait set out deliberately to create a division where none had existed. Historian Pradip Kumar Datta calls Gait’s memo a “blatant act of social engineering … designed to encourage the detachment of low castes from the ‘Hindu’ category” and to provoke “low-caste resentment” (1999, 24, my emphasis). Datta offers no evidence for this startling claim, apart from a quote from Mukerji insisting that it is so, an instance of circular reasoning on Datta’s part that usefully illustrates the common ground between the secular liberal and Hindu nationalist on the untouchable question.
Mukerji’s pamphlet was reprinted countless times, and its arguments are repeated by Hindu nationalists to the present day (Bhatt 2001, 62.-68). Its core message was that Hinduism was in a demographic struggle with Islam, a struggle in which Hindus were literally in danger of becoming biologically extinct. It also provided the strategic blueprint that would become a central feature of Hindu nationalism from that day onward-namely, that the very survival of Hinduism in the face of a putative Muslim threat (and later a Christian one) depended on its ability to incorporate Dalits and tribals within its fold. The necessity of integrating untouchables within Hinduism would become a key plank in the program of Hindu sangathan (consolidation/organization).
By far the most important proponent of untouchable integration was Swami Shraddhanand, an Arya Samaj leader whose mission to the untouchables was inspired by a personal meeting with U.N. Mukerji in 1912 (Datta 1999, 22). Mukerji’s enduring influence is evident in the title of Shraddhanand’s 1924 tract, Hindu Sangathan: Saviour of the Dying Race, a text that reproduces and expands upon Mukerji’s core argument. A “constant refrain” of the swami’s writings and speeches on the untouchable question, as Joel Lee’s study of Shraddhanand’s corpus reveals, was the worrying prospect of Dalits emerging as an autonomous political force in the Indian landscape (Lee 2015, 141). Perhaps equally alarming to Shraddhanand was the threat of Dalits converting to Christianity or Islam. Thus the swami openly warns that Dalits who convert to Islam “will become equal to Hindus …. They will not depend on Hindus, but will be able to stand on their own legs”; those who convert to Christianity will “dream of entering its halls of governance” (quoted in Lee 2015, 140, 142). The key to preventing this, according to Shraddhanand, was eliminating the divisive practice of untouchability from the collective Hindu body.
As Lee observes, “Shraddhanand insisted that the danger … of [Dalit] autonomy from the Hindus could only be defused if Hindus radically curtailed the regime of disabilities they imposed on untouchables. Further, he maintained that [checking] this autonomy, [by] bringing the untouchables to accept Hindu leadership, equated with the neutralization of the Muslim and Christian threat, and was an essential, sine qua non … for the manufacture of a Hindu nation” (Lee 2015, 143). For Shraddhanand the Hinduization of untouchables was not merely a Hindu communal cause but a national one. In his writings the good of the Hindu community was indistinguishable from India’s struggle for national independence: “The uplift of the untouchables and their assimilation in the Hindu polity is the very plinth on which alone the edifice of free India can be constructed” (quoted in Lee 2015, 143). As we will see, the mission of ending untouchability would play an identical role for Gandhi. For Gandhi, too, it was essential to the strength of both Hinduism and the nation—even to the extent that the good of the one was often presented as indistinguishable from the good of the other.
The imperative of Hinduizing the untouchable was eventually endorsed, in theory if not always in practice, by Hindu organizations across northern India and from Bengal to Bombay (Prashad 1996). The major bases of support for this movement were in towns and cities, among modernizing Hindu organizations like the Arya Samaj (Jones 1976; Adcock 2014), and among politically minded Hindu reformers keen on establishing India as a Hindu nation (Bayly 1998). But it was by no means universally accepted. Orthodox Hindus remained deeply opposed, wanting nothing to do with those they regarded as untouchable. Opposition was also widespread among rural Hindu elites (Jones 1976). And in South India, where Muslims were not perceived as a threat, programs for Hinduizing the Dalit found few takers.
The Congress Party passed its first resolution condemning the practice of ritual untouchability, seen by advocates of Hindu sangathan as the principal barrier to Dalits’ inclusion within the Hindu fold, only in 1917. But it took no concrete steps on this matter until the 1930s (Prashad 1996, 553). As early as 1920, M. K. Gandhi, the party’s paramount leader, proclaimed the eradication of untouchability as essential to swaraj, and in 1921 claimed he had in fact opposed untouchability since childhood (Zelliot 2010, 153). Gandhi did not act on this conviction until the 1930s, however, despite urgings by Shraddhanand to join him (Lee 2015, 145), and despite multiple opportunities to lend support to autonomous Dalit struggles, including invitations from Dalit activists, which he repeatedly declined (Ambedkar 1946, 251-59). According to J. T.F. Jordens, Shraddhanand’s sole supporter of any note within Congress until the 192os was G. D. Birla, one of India’s leading industrialists, whose vast wealth bankrolled Shraddhanand’s operation (1981, 165, cited in Lee 2015, 143-44).
As for Dalits themselves, some reacted with enthusiasm to the prospect of Hinduization (shuddhi), which by the second decade of the century many embraced as an opportunity for social advancement. And just as it was Dalits who first approached Christian missionaries, and not the reverse, with demands to be converted (Viswanath 2014b; Webster 2.009), so too did they begin to approach the Arya Samaj (Rawat 2011; Adcock 2014, 48-50; Lee 2015). But finding the promise of full inclusion as equals illusory, Dalits began to turn away from shuddhi by the 1920s (Lee 2015, esp. 150-53) and, simultaneously, to assert their independence from the Congress Party (Prashad 1996, 552).
This is an excerpt from Nathaniel Roberts’ To Be Cared For: The Power of Conversion and Foreignness of Belonging in an Indian Slum (pp. 124–31).
 Butalia (2000, quoted in Lee 2015, 81). I thank Joel Lee for directing my attention to the writings of Butalia and Searle-Chatterjee and for providing the analysis of them (Lee 2015, 81–82) that the next three paragraphs repeat and expand upon.
 This contrastive usage remained commonplace in confidential Government of Tamil Nadu reports on anti-Dalit atrocities as late as the 1970s. Caste folk responsible for these attacks are referred to in these reports simply as “the Hindus” in contradistinction to their victims, who are distinguished as “the Harijans” (Rupa Viswanath, personal communication, September 2015).
 The idea that untouchables were always regarded as beyond the pale of Hinduism has been challenged by Arvind Sharma (2015). Sharma argues that the common understanding of untouchables as being outside the fourfold varna system is wrong. According to him untouchables should instead be thought of as a special subcategory within the shudra varna, a category he terms the excluded shudra. If untouchables were formerly included within the varna system, the argument seems to go, then by definition they must have been Hindu. But Sharma has previously argued that varna was a classificatory system that extended to the entire world, and therefore that all the world’s people were originally regarded as being within it (1992, 179). If consistently followed, the logic of Sharma’s argument would compel us to accept not only that untouchables were originally Hindus but that Chinese, Greeks, and Persians were too. Apart from this implausible implication, Sharma’s claim that untouchables were regarded as Hindus in ancient times rests on a faulty methodology. Rarified theoretical texts accessible to only a tiny cohort of Brahmin intellectuals provide no direct window into ancient social reality and tell us nothing about how ordinary people classified themselves and others.