ISSN (Print) - 0012-9976 | ISSN (Online) - 2349-8846

A+| A| A-

Census in Colonial India and the Birth of Caste

Caste, as conceived in contemporary academic writings or within the policies of the State, is a new idea produced during the second half of the 19th century in the course of and because of the census operations. Colonial census officials, working with concepts of varna and jati, struggled unsuccessfully to define and classify these into castes on a single pan-India list, where each caste had to be discreet, homogeneous and enumerable. The history of caste enumeration in the Indian census illustrates how difficult it is to capture indigenous social hierarchies and identities under the term "caste". We embark on a new caste census without having addressed many of these challenges.


Census in Colonial India and the Birth of Caste

Padmanabh Samarendra

Caste, as conceived in contemporary academic writings or within the policies of the State, is a new idea produced during the second half of the 19th century in the course of and because of the census operations. Colonial census officials, working with concepts of varna and jati, struggled unsuccessfully to define and classify these into castes on a single pan-India list, where each caste had to be discreet, homogeneous and enumerable. The history of caste enumeration in the Indian census illustrates how difficult it is to capture indigenous social hierarchies and identities under the term “caste”. We embark on a new caste census without having addressed many of these challenges.

The paper has benefited from the comments made by Gautam Bhadra, Neeladri Bhattacharya, Rohan D’Souza, and scholars from the Department of History and Culture, Jamia Millia Islamia and Centre for Historical Studies, Jawaharlal Nehru University. Sangeeta DasGupta has read the manuscript and suggested extensive changes. I am thankful to all of them. The usual disclaimers apply.

Padmanabh Samarendra ( is at the Dr K R Narayanan Centre for Dalit and Minorities Studies, Jamia Millia Islamia, New Delhi.

Economic & Political Weekly

August 13, 2011 vol xlvI no 33

he proposal of the Government of India to conduct a caste

– census has generated intensely conflicting responses

within and outside the Indian Parliament. Those in favour of the proposal argue that the survey would reveal the identity and numerical strength of the castes which suffer from deprivation; this in turn would help the state in reformulating and extending policies, such as that of reservation, which would then ensure the upliftment of the deprived sections. Those arguing against the enumeration of caste believe that the measure contravenes the ideals of citizenship and would foster divisive tendencies within society and the polity of the nation.

Notwithstanding the other differences, both the protagonists and antagonists in this debate stand together when sharing the following premises: they believe that caste is an indubitable reality of Indian social life; that it has been so since the earliest times; and that the census could map this reality and produce castedata. In the present paper I adopt a contrasting stance. I argue that caste, as conceived in contemporary academic writings or within the policies of the state, is a new idea and that the social form imagined through this term never characterised the Indian society. Further, I suggest that this idea of caste was produced towards the end of the 19th century in the course of and because of the census operations.

Caste of the Jati

I begin by submitting that caste is a foreign word. Derived perhaps from the Portuguese “casta”, this word could not have been in circulation for more than past three to four centuries. The would-be critics of this paper might respond by stating that even if the word is new, its connotation is not so. What does caste connote? The scholars, in general, have interpreted caste as an equivalent of varna or/and jati, entities that have been parts of the indigenous traditions.

If it can be demonstrated that the attributes of caste match with that of either varna or jati then the hypothesis that I propose would certainly stand rejected. Yet even a cursory look at the scholarly writings reveals widespread confusion and uncertainty on the issue. For example, M N Srinivas equates caste with jati and not varna. In an article entitled “Varna and Caste” (evidently caste here means jati), he writes: “The varna-model has produced a wrong and distorted image of caste. It is necessary for the sociologist to free himself from the hold of the varna-model if he wishes to understand the caste system” (Srinivas 1962: 66). McKim Marriott and Ronald Inden differ with Srinivas. They do not reject the relevance of the varna model; further they argue that jati does not exclusively refer to caste. “The South Asian word jati”, they write, “refers to a great many kinds of things other than those we mean by the word ‘caste’. It refers to all sorts of categories of things” (Marriott and Inden 1977: 230). Some scholars have tried to sidestep the whole problem by combining the two terms: Susan Bayly writes about the “varna jati norms” (Bayly 2002: 25, see also, 12), while Suvira Jaiswal in her book traces “the origins of the varna caste [jati] structure” (Jaiswal 2000: ix).

Such formulations, I should point out, disregard the widely shared consensus that varna and jati are two dissimilar concepts which implies that together these cannot constitute a singular set of norms or a single structure. So what is caste – varna or jati? Further, when the scholars write about the pan-Indian caste system what are they referring to – a system of varna or that of the jatis? Contrary to the broadly shared belief, the problems faced by the scholars in equating caste with either varna or jati, I believe, allude to the non-equivalence of these terms.

In order to illustrate the distinction between caste vis-à-vis varna and jati, I briefly look into their meanings. Caste has been variously defined as an endogamous, ethnic, occupational, ritualistic or racial division. Underlying all these definitions are present the following two assumptions: that caste actually exists and can be observed in society, and that it has a fixed and uniform boundary implying that communities called caste across India have something in common.

The varna order, when compared, seems to share the definitiveness of caste: as mentioned in the Sanskrit texts there are four groups within this order the attributes of which are more or less consistent. Yet, unlike caste, the scholars point out, the varna system is text-based and does not exist on the social plain in the same form. No society across the Indian subcontinent is actually divided into merely four groups of brahmin, kshatriya, vaishya and shudra.

The presence of the jatis, on the other hand, can be observed in society. The similarity with caste however ends there. The connotation of jati, in contrast to caste, is far from uniform. For example, in vernacular literature, we come across the Lohar and the Sonar jati (professional communities), the Maratha and the Bangla jati (linguistic or cultural communities), the Hindu and the Mussalman jati (putatively, religious communities), the Munda and the Oraon jati (communities presently registered in the government documents as tribes), the Vaidya and the Bhumihar jati (communities which are endogamous), mardon ki jat and aurat jat (community of men and community of women), etc. Jati thus denotes professional, regional, linguistic, religious, only locally recognisable and even gendered communities. It should also be noted that endogamy, often held as the defining marker of caste, does not uniformly characterise all these communities. Which of these jatis, then, could be called castes?

In this paper I argue that caste is neither varna nor jati though it masquerades as one or the other or both at the same time. And since caste is not the same as the other two, it must be a new category deployed for imagining the Indian society.

Census Compulsions

The birth of caste is directly linked with the census operations in colonial India. Beginning around the middle of the 19th century, the census – because of its methodology and agenda – was a unique project. It was a direct survey of population; instead of surmising or using textual references, the enumerators went to the people with questionnaire to know about their number and attributes. This meant, for instance, that the acceptance of the varna order in the census would depend not on the textual citations that could be mustered in its support but on its observed presence in society. The purpose of the census was to count the population and classify it under different heads – age, sex, religion, caste, occupation, etc.

The conduct of these two interrelated processes depended on the fulfilment of certain preconditions. First, an entity to be counted had to be definite and discrete with no overlapping boundaries. For example, people that were to be enumerated as Hindus must first be defined as such and separated, say, from the Sikhs, Buddhists or animists. Classification, on the other hand, referred to the practice of sorting and arranging the data in different columns and rows to be presented in a tabular form. The first step relevant in this regard was to formulate the classificatory principle/s. The principle/s would be derived from the defining attribute/s of the entity to be classified. For example, if caste was recognised as an occupational division then different types of occupation would be used to name the columns or rows in a table within which the numerous castes enlisted during a survey would be placed. An accurate classificatory table demanded that the name describing a column or a row should match with the nature of data pigeonholed therein.

The census was a serialised pan-Indian project. The data that was collected was put to comparison across the provinces as well the various editions of the census operations. The fulfilment of this very basic exercise hinged on consistency in the use of the classificatory models. Uniformity in the criteria of identification and classification of communities thus was essential for a proper conduct of enumeration. And once again, keeping in view the empirical nature of the project, these criteria had to be drawn not from texts but from the lived social experience. These features of the census, I argue, were instrumental in producing the idea of an empirical and uniform caste.

The assessment of population in colonial India had started from the early years of the 19th century (Cohn 1990: 233-36). These however were indirect appraisals with no uniformity in the method followed. The first census based on a direct survey of population took place on 1 January 1853, in the North-Western Provinces (called the NWP hereafter, the province covered parts of the present day UP). In the following decade census was also held in several other provinces including Oudh and Punjab, though as yet it was neither a regular nor a pan-Indian affair. The enumeration of caste started from the census of the NWP in 1865; it continued to be a prominent part of the colonial census till 1931.

There was no grand design of knowing and controlling the population as generally attributed to the colonial state that inspired the inclusion of caste in the census of the NWP. The objectives were rather limited: caste was expected to identify the communities and “facilitate classification [of the population] into ‘agricultural’ and ‘non-agricultural’ ” (Plowden 1867:1). In addition, information about certain castes was also supposed to help in checking customs like “infanticide” prevailing amongst them (Plowden 1867: 30, 48).

August 13, 2011 vol xlvI no 33

The Difficulty of Counting Castes

The introduction of caste in census was accompanied with serious problems of identification and classification. The enumerators discovered during the surveys that society was not patterned according to the fourfold varna division comprising brahmin, kshatriya, vaishya and shudra. Instead in their localities they encountered jatis – communities with unfamiliar names, uneven status and unalike characteristics. For example, the report on the census of the NWP in 1865 showed the Sikh, Jain, Goshain, Jogee, Sunni, etc, to be sharing the same space in the caste table along with the brahmin, kshatriya, vaishya and shudra (Plowden 1867: 80-81). The caste table in the report on the census of Oudh in 1867 presented a similar picture. The group of “higher caste of Hindu” included “the following thirteen castes”: Brahmin, Bengali, Jat, Jain, Kshatriya, Kayasth, Khatri, Kashmiri, Marwari, Punjabi, Sikh, Sarawaks, Vaishya (Williams 1869: 86).

The critics might assume that the lists cited above contain many errors committed by those British supervisors who were perhaps not acquainted with the social divisions prevailing in this country. To dispel such assumptions I refer the readers to another list, prepared this time by a person who was a native to this land and was also intimately connected with the society he lived in. Between 1658 and 1664, Munhata Nainsi, a high-ranking official in the kingdom of Marwar, Rajasthan, had conducted a survey of households of the locality. The results that were compiled in Marvar ra Paraganam ri Vigat (An Account of the District of Marwar) which, in certain instances, also mentioned the jatis occupying the households. The names of some of these jatis are quite illuminating: Oswal, Agarwal, Darzi, Gujrati, Multani, Mahajan, Mussalman (for excerpts from Munhata Nainsi’s “Account”, see, Peabody 2001: 827-28).

The jatis, these examples indicate, have never been homogeneous though our caste-conditioned minds have failed to appreciate the significance of the situation or even take this into account in the first place. Nonetheless, the census officials were now faced with a serious challenge: which of those diverse jatis enlisted in the census schedules were castes, and how were these to be classified in view of the fact that the varna framework was not considered relevant anymore?

The diverse identity of the jatis precluded the possibility of cataloguing these according to any uniform classificatory principle. The caste table in the report on the census of the NWP, 1865, had several columns within which all the jatis were placed: first there were four columns headed by brahmin, kshatriya, vaishya, shudra, then came the five columns named after the Sikh and four sections of the Jains, following which came the 14 columns of “religious sects”, three columns of “travellers”, and finally one of those “Hindoos whose caste is not known” (Plowden 1867: 81). Being the first such survey where caste was included, the officials supervising the census of the NWP, 1865, could not exclude the varna division from the classificatory table. At the same time, the presence of additional columns named after “religious sects”, “travellers”, etc, implies that the jurisdiction of the text-based order was a lready being compromised. In fact, the varna model did not figure at all in the caste-table of the report on the census of

Economic & Political Weekly

August 13, 2011 vol xlvI no 33

Oudh, in 1867, which employed the following nine columns for the purpose of classification:

Europeans, Eurasians and Native Christians, Higher castes of Muhammadans, Muhammadans converts from the higher castes of Hindus, Lower castes of Muhammadan, Higher castes of Hindu, Lower castes of Hindu, Aboriginal castes, Religious mendicants and Miscellaneous (Williams 1869: 72).

The enumeration of caste during the early census operations generated conflicting trends. The census reports from the initial decades showed that the caste list of every province differed from that of the other because of the presence of heterogeneous jatis obtaining in that area. Further, every province seemed to have adopted a distinct classificatory table keeping in mind the nature of the local jatis. However, the operation of census depended on the ability to compare data. Only by comparing and aggregating caste-data from various provinces could one arrive at all-India figures on the subject. Similarly, comparison and differentiation was necessary to know about the changes in the caste numbers of any province or the country as a whole over different periods of time. Yet, the dissimilarity in the caste lists as well as classificatory tables that arrived from the first decade of caste enumeration seemed to deny any possibility of comparing data.

The Administrative Imperative

The census was faced with an unresolved contradiction: its schedules attested to the heterogeneity of jatis, its processes demanded homogeneity from the communities to be counted. The problem was to assume acute proportions with the inauguration of the pan-Indian decennial census series and the manifold rise in the collection of data.

The pan-Indian decennial census series, which commenced from 1871-72, suffered from several defects. It was not conducted

Classification of Castes during the Census of 1871-72
North Western Provinces Central Province Bengal Madras
1 Brahmans 1 Brahmin 1 Superior castes 1 Priests
2 Rajputs 2 Agriculturists 2 Intermediate 2 Warriors
3 Buniyas 3 Pastoral castes 3 Trading 3 Traders
4 Other castes
of Hindoos 4 Artisans 4 Pastoral 4 Agriculturists
5 Merchants 5 Engaged in preparing 5 Shepherds
cooked food and pastoral
6 Scribes 6 Agricultural 6 Artisans
7 Small traders 7 Engaged in 7 Writers and
personal service accountant
8 Servants and 8 Artisan 8 Weavers
9 Manufacturing
castes 9 Weaver 9 Labourers
10 Mendicants and
devotees 10 Labouring 10 Potmakers
11 Dancers, etc 11 Occupied in selling 11 Mixed castes
fish and vegetables
12 Boating and fishing 12 Fishermen
13 Dancer, musician, 13 Palm
beggar and vagabond cultivators
14 Barbers
15 Washermen
16 Others
17 Outcastes

all over the country on the same day, nor did it cover all the regions of British India. However, its most crucial shortcoming, as evident from the columns used in the reports on different provinces was the want of consistency in the classification of castes. The arrangement of castes as followed in the NWP (Plowden 1873: lxxxviii), Central Province (Neill 1873: 33), Bengal (Waterfield 1875: 21) and Madras Presidencies (Waterfield 1875: 21) is shown in the previous page.

The outcome of the process of classifying castes, it was admitted in a memorandum on the conduct of the first decennial census in India that was presented before the British parliament, was “not satisfactory, owing partly to the intrinsic difficulties of the subject, and partly to the absence of a uniform plan of classification, each writer adopting that which seemed to him best suited for the purpose” (Waterfield 1875: 20). Such a lack of consistency in the classificatory columns, the author of the memorandum regretted, permitted only “a few particulars” (Waterfield 1875: 20) to be aggregated. The very purpose of the census operations was in jeopardy as questions arose about the feasibility of deducing all-India figures from the provincial reports on the issue of caste.

Textual Definitions and Field Data

Before moving on to the census of India in 1881, I briefly underline those procedures of enumeration that played a critical role in constituting a new idea of caste. The understanding of caste in colonial literature in the early decades of the 19th century was primarily text-based. The colonial officials like William Jones and Henry Colebrook, writing from towards the close of the 18th century, considered Sanskrit texts as the authentic sources of knowledge about the Hindus. Hence these officials, also called the Sanskritists, despite being aware of the presence of the jatis, treated the text-derived varna order to be representing the original and the authentic caste system.

The empirical approach of the census led to a fundamental change in the outlook: the focus shifted from text to people. The social space was marked as the habitat of the real; the verification of truth came not from the pages of the Sanskrit tomes but the experience of the lived lives of men and women. As per the new norms, the varna model was put through the empirical test, and rejected. Society was populated not only with brahmins, kshatriyas, vaishyas and shudras. Simultaneously, the questioning of the credibility of the propagator of this model – Manu – started. Let me cite here some of the officers associated with the census of different provinces in 1872. Referring to the “Code of Menu and some of the Puranas [that] profess to give an account of the institution of castes”, Cornish (1874: 121, 122), who supervised operations in the Madras Presidency, commented: “It is plain that in a critical inquiry regarding the origin of caste we can place no reliance upon the statements made in the Hindu sacred writings. Whether there was ever a period in which the Hindus were composed of four classes is exceedingly doubtful.” Similarly C F Magrath, the officer entrusted with the compilation of castes from Bihar, stated, “it was necessary, if the classification was to be of any use, that the now meaningless division into the four castes alleged to have been made by Manu should be put aside” (cited in, Beverley 1872: 155).

The critique of Manu, in a broader sense, was addressed to the text-based exposition of the Indian society as presented by the Sanskritists. Yet, as Trautmann points out, the legacy of William Jones, Henry Colebrook and later Max Muller and John Muir continued to figure in and configure the terms of investigation at a time when anthropological interpretations had begun to displace the text-based versions of caste (Trautmann 1997: 26, 191206). Furthermore, in the face of the presence of the divergent jatis during the surveys, the anthropological theories found corroboration of the idea of caste as a single entity in concepts drawn from Sanskrit texts. Hence, the allusion to the ideas of varna samkara (narrowly interpreted as miscegenation) or the separate function of brahmin, kshatriya, vaishya and shudra when caste was interpreted respectively as an endogamous or occupational division.

Thus the textual categories, though disputed, could not be discarded during the empirical shift in the understanding of caste. At the same time the bases and the connotations of these categories were radically altered. For example, while the varna model was refuted, some of its components could still find a place in the census reports. The brahmins and the Rajputs were indeed included in the classificatory table of the NWP in 1872. However, the inclusion took place not before the physical presence of these communities could be verified. Thus F S Growse who had helped Plowden in preparing the caste table of the NWP, wrote that in the province the brahmins numbering around 3,234,342 were “still a living entity” and the kshatriyas were “adequately represented in modern speech by the word Thakur, or Rajput” though the Vaishyas and the shudras had “completely disappeared” (cited in, Plowden 1873: lxxix).

The integrity of the varna order was broken; hollowed out of the earlier connotations, the constituting groups when individually appropriated were suffused with new meanings. The census, I believe, was accompanied with a process of empiricalisation of textual traditions. In the course of enumeration, the components of the varna order, i e, brahmin, kshatriya, etc, were empiricalised: attributed a visible and verifiable body. Caste, henceforth, was recognised necessarily as an empirical category. It is in the context of empirical inquiry, I should also add, that the academic tradition of counterpoising varna and jati in terms of textual vs real (or for that matter, ideal vs actual, original vs contemporary) started. The tradition continues in the writings on caste even today.

Refining the Definition of Caste

Let us return to the census of India in 1881, which also happened to be the first synchronous survey of population to be conducted in the country. The question that had assumed seminal significance for the project was that of maintaining uniformity in the classification of castes. The source of the problem lay in the nature of jati: the multiple connotations of jati did not bear out the official perception of caste having a fixed and uniform boundary. Thus the Secretary of State, who had forwarded in 1877 a set of measures to the Government of India aimed at avoiding the shortcomings of the preceding census, complained that often a man would state that he was

by ‘caste’ a ‘marhatta’ when he is a Kunbi of the Marhatta nation; that he is by ‘occupation’ a carpenter when he is a cultivator of the

August 13, 2011 vol xlvI no 33


carpenter caste; that he is by ‘religion’ a Brahmin, when he is a

Brahmin by caste.1

This divergence of jatis to which was added the dimension of regional variations had earlier led the provincial enumerators to use different classificatory tables. Ignoring such ground realities in order to ensure statistical computations, the authorities supervising the Census of India in 1881 insisted that castes be uniformly classified under the following heads: brahmin, Rajput and Other Hindus (Plowden 1883: 277).

Despite all efforts to ensure statistical consistency on the various issues connected with the enumeration of caste, discrepancies persisted in the census. For instance, in the enumerator’s schedule, the class of Hindu had been subdivided into castes, and still the total number of Hindus, in 1881, did not match with the sum total of caste-persons (The numbers respectively being 187,937,450 and 188,121,772; see Plowden 1883: 277). This difference in number occurred because certain communities, though classified as castes, had not been counted as Hindus. The conflicts in number cast a shadow of doubt on the very authenticity of the census project. It was increasingly being acknowledged that to count it was first necessary to know what caste was. In 1882, W C Plowden, the outgoing census commissioner, recommended that in every province “some officer who has a taste for, and a knowledge of, archaeological research” should be deputed to compile information about caste. The “advantage of having such information at hand at the next census”, he concluded, needed “no comment”.2 Responding to the recommendation, the Government of Bengal appointed H H Risley, in 1885, to conduct a survey of castes and occupations of the people of the province.3

The Ethnographic Move

I briefly digress here to ask why the empirical surveys, which underlined the divergence of jatis and exposed the mismatch between the social situation and textual representation of society leading to a questioning of the relevance of the varna model, failed to dislodge the belief of the colonial officials in the existence of a pan-Indian caste structure? Several reasons can be cited here.

The first relates to the burden of history. In the body of knowledge about Indian society that the census officials had inherited from the Sanskritists, the existence of caste had already been recorded. Further, the Sanskritists used to address both varna and jati as caste; so, even when the continuance of the varna order was doubted the jatis remained present to be counted as castes.

The second reason is linked to the functioning of the state. The census operations launched by the colonial state produced a social map of the country reconstituting pan-Indian identities of caste, tribe, Hindu, etc, within its very format. The design of these operations replicated the model of the administrative edifice of India: the district census reports were compiled to produce a provincial report; the provincial reports together, in turn, generated the general report on the census of India. Implicit in the format was the assumption about the universality of caste; that castes from different parts of the country could be added up and presented in a master table. The obligation on the census officials was to uncover the essence of caste, to abstract those defining

Economic & Political Weekly

August 13, 2011 vol xlvI no 33

features on the basis of which castes across the regions could be identified, counted, compared and classified. Thus in the wake of the questioning of the text-based varna model, attempts were made in the course of the successive census operations to sift and identify castes from amidst the multifarious jatis and construct for these an alternative pan-Indian classificatory grid.

The conclusion of the census of 1881 marked a watershed in the history of the project in India. Till now the collection and compilation of data were understood to constitute the field of enquiry of census. However, the persistent problems in statistical computations had rendered the continuation of this project with its older focus untenable; the colonial state under the circumstances was compelled to make an ethnographic move.

The results of the survey of Bengal undertaken by Risley were published in 1891 in four volumes entitled the Tribes and Castes of Bengal; subsequently, under the generic title of Tribes and Castes glossaries of communities were compiled and published for the North-Western Provinces and Oudh (1896), Southern India (1909), Punjab and North-Western Provinces (1911) and the Central Provinces (1916). Generally, the glossaries came prefaced with a long essay where the respective authors attempted to explicate caste: its origin, history, salient characteristics, etc. These essays aimed to investigate and discover the defining feature/s on the basis of which castes could be identified and separated from amongst the diverse jatis living in society.

Three brief qualifications need to be added here. First, when formulating their respective anthropological explanations of caste, the colonial scholars did not and could not exclude the concepts found in the Sanskrit texts. Second, prior to the beginning of the ethnographic surveys sponsored by the provincial governments, some census officials like James Bourdillon and Denzil Ibbetson did try to investigate caste. However, such attempts were borne out of personal curiosity and were not parts of the colonial state’s initiative. Finally, despite consenting to conduct a survey of tribes and castes in certain provinces, classification remained the primary concern of the colonial state. Accordingly, the state tried to limit the investment, both material and academic, in anthropological researches and keep it germane to the specific problem. In short, while the moves made by the state in the post 1881 phase were ethnographic, the same cannot be said about its nature. Nonetheless, change had come to mark the project of census; unlike their predecessors, census commissioners in the following decades, before embarking on the exercise of classification, engaged in clarifying what caste was.

Status or Function

The section entitled “The Distribution of the Population by Race, Tribe, or Caste” in the general report on the census of India in 1891 opened with an analysis of caste. Caste, wrote J A Baines, the census commissioner, referred to “status or function” that was perpetuated by “inheritance and endogamy” (Baines 1893: 182). Despite signalling function as the basis of caste, the author, when it came to elucidating the genesis of the institution, shifted to race. The origin of caste, he believed, was “distinctly racial” (Baines 1893: 183); it was the result of the struggle of “the Arya” to keep out “the dark races” (Baines 1893: 183).

The wavering emphasis on function and race in the writing of Baines was caused by the two differing interpretations of caste, operational at this time. Denzil Ibbetson, William Crooke and John Nesfield had offered a functional explanation, while Herbert Risley, in the course of his survey of Bengal, had propounded a racial theory of caste. The parallel influence of these theorisations can also be seen in the act of classification. In order to ensure uniformity in returns, caste groups, declared Baines, were arranged in the census tables “mainly” on the basis of “function” (Baines 1893: 188-89). Yet, even a cursory look on such arrangements reveals that the census commissioner had also used the category of race to name the classificatory columns (Baines 1893: 188, 207-08).

The lack of consistency was thus the chief defect in the explanation of caste as presented in the general report on the census of 1891. Risley, who became census commissioner of India for the 1901 edition of the project, observed later that the report prepared by his predecessor provided only “a patchwork classification in which occupation predominates, varied here and there by considerations of caste, history, tradition, ethnical affinity, and geographical position” (Risley 1903: 538).

The census of India in 1901 witnessed the most comprehensive attempt made by any census commissioner yet to understand caste. Using anthropological concepts and the data relating to physical features of local population collected during the survey of Bengal and later from a few other provinces, Risley forged his racial theory of caste. He presented the initial version of this theory in 1891 in an essay entitled “Caste in Relation to Marriage”, though it should be mentioned here that he was not the first scholar to discuss the idea of race in the Indian context. The genesis of caste, according to this explanation, lay in the “fact” (Risley 1903: 555) of racial difference. Briefly, when the tall, fair and sharp-nosed Aryans entered India from the north, they encountered the short, dark and stub-nosed Dravidians. In order to prevent intermixture of blood and loss of purity of their race, the Aryans instituted restrictions on marriage with the local race. From such restrictions evolved the structure that we know as caste (Risely 1903: 555).

The responsibility of the census commissioner, however, was not to provide an academic explanation about the origin of caste in the remote past. The onus rather was to specify the criteria that could be used to identify and classify castes in the present. Once again Risley was confronted with the same question that had tormented his predecessors. What was caste? And once again Risley was presented with the same answer: the jatis on the ground were far from homogeneous. Let us hear the census commissioner himself dwelling upon the dilemma:

In a country where the accident of birth determines irrevocably the whole course of a man’s social and domestic relations… one thing that he may be expected to know with certainty, and to disclose without much reluctance, is the name of the caste… to which he belongs.

Yet when asked about his caste, wrote a dismayed census commissioner (Risley 1903: 537), a respondent might actually give the name of

an obscure caste… a sect… a sub-caste… an exogamous sept… a hypergamous group… may describe himself by… occupation or… the province or tract of country from which he comes.

So, which name from among these was that of caste? Circumstances demanded clarity on the issue; to count and collate, it was necessary to fix the identifying marker. Hence, in the census report of 1901, appeared a “definition” (Risley 1903: 517) of caste. Caste had several characteristics, Risley wrote; it “may be defined as a collection of families… associated with a specific occupation; claiming common descent”. But more than anything else, caste was “almost invariably endogamous” (Risley 1903: 517). Alongside Risley also declared that castes should be classified according to the “principle” of “social precedence” (Risley 1903: 538).

A Patchwork of Definitions

Neither the criterion of identifying castes nor that of classifying these could deliver the desired results. Endogamy could not be the defining marker of caste; as we know a large number of jatis, perhaps a majority of these, are not actually endogamous. For example, within the fold of the Kurmi jati in Bihar were/are present the following groups: Awadhia, Chanaur, Ghamela, Jaiswar, Kachaisa, Ramaiya, etc. Erroneously called “sub-castes” by many sociologists, these are the groups that actually practise endogamy: thus, an Awadhia would not be inclined to marry a Ghamela or a Chanaur. A similar situation prevails in the case of the Kamma jati of Andhra Pradesh which encompasses the below named groups: Gampa, Illuvellani, Godajati, Kavali, Vaduga, Pedda, Bangaru. Going by the academic criterion of endogamy, Awadhia or Illuvellani and not Kurmi or Kamma should be treated as a caste and yet doing so would completely violate the popular opinion on the subject. Hence, Risley’s definition of caste remained inapplicable.

Risley also failed in his attempt to classify castes on a pan-Indian scale according to the principle of “social precedence”. The divergence in the nature of jatis naturally implied the absence of a singular scale to measure status or construct hierarchy. The census commissioner himself later observed: “Castes can only be classified on the basis of social precedence. No scheme of classification can be framed for the whole of India” (Risley 1903: 557). No pan-Indian classificatory table was presented in the report on the census of 1901.

Risley had criticised the classificatory tables containing the results of the census of 1891 as a “patchwork”, yet even a small sample from the census report prepared under his guidance reveals a similar situation. Thus some of the names included in the caste table of the report on the census of Bengal in 1901 were the following: Chamar, Halwai, Baniya, Madrasi, Marwari, Manipuri, Burmese, Chinese, Japanese, Bengali, Maratha, Sikh, Baishnab (Bairagi), Buddhist, Munda, Santal, Oraon, Ahir, Kurmi, Barna-sankar, etc (Gait 1902: 192-266). Tentatively these communities can be described as professional, regional (both from within and outside the country), linguistic, religious/sectarian, tribal (according to the records of the state), and only locally known groups. Risley indeed was able to compile a list of castes made up of the heterogeneous jatis as his predecessors and successors had done.

In fact, in a similar manner the state as well as the academicians in contemporary India also has been producing caste lists. However the question that must be asked is the following: what is

August 13, 2011 vol xlvI no 33


it that is common between, say, Chamar, Marwari, Munda and Kurmi that qualify these to be parts of a list of castes. How can we decide whether Baniya is a caste or not? The only factor that seems to be uniting the names that form the list of castes cited above is that these are seen to belong to different communities. Jati, in the first place, then, signifies a community. Caste on the other hand is also imagined as a community but a community of a particular type. The particularity of caste is inscribed in the features attributed to it: a definite and singular hierarchy with perhaps the brahmin at the top and the so-called “untouchables” at the bottom, a specific set of practices relating to endogamy or commensality, etc. These particularities do not uniformly apply to the jatis and this is why jati as understood in indigenous traditions and society is not the same as caste.

In the midst of lengthy academic expositions and scores of classificatory tables, caste had a troubled presence in the pages of the census reports. No exhaustive list of castes could ever be prepared for any province, let alone for the country as a whole, every such list was completed only by adding columns named as “other castes”, “castes not specified/not known”, etc; no list was ever submitted without questions being asked whether those enlisted were really castes; no two reports on the census of India ever matched in the way these classified castes; no inventory of caste was ever compiled without the presiding census commissioner expressing misgivings about the whole project.

The enumeration of caste continued given its deemed administrative relevance. However, no sooner had the pan-Indian census started that the state was compelled to try and trim down its engagement with the subject. During the census of 1881 it was decided that only those castes having a minimum numerical strength of 1,00,000 would be classified; after 1901, following the failure of Risley to construct a pan-Indian classificatory table, the practice of classifying castes in census itself was given up, castes from now onwards were enlisted alphabetically; in 1931, the last general counting of castes took place. Thus, after more than six decades of enumeration, the census authorities failed to provide the criteria to identify caste and classify these groups. Yet, the very structure of the census first, generated and then, sustained the belief that caste was a uniformly definable and empirically verifiable entity. These two premises continue to configure the varying interpretations of caste today.


The idea of caste, as conceived in contemporary academic writings and within the policies of the state, was produced in the course of the census operations in colonial India. In the textbased explanations offered by the Sanskritists in the early decades of the 19th century, the brahmin, kshatriya, vaishya and shudra were treated as authentic castes; jatis enjoyed no distinct conceptual status, the assumption being that these could be subsumed, at least theoretically, within the varna order.

The onset of the census operations from the middle of the 19th century signalled fundamental changes. Empirical surveys showed the varna division to be non-existent; the focus now shifted onto the numerous jatis populating the social space. The project entailed that the diverse jatis be first counted, and then

Economic & Political Weekly

August 13, 2011 vol xlvI no 33

classified within a new pan-Indian template. The obligation was novel and unparalleled. The jatis, along with their assumed numerical strength, had been enlisted earlier too; however these communities stood in such lists as discrete units. The summing up of number in the census, on the other hand, was possible only if the entities counted were made comparable. The compulsion, under the circumstances, was to find the feature/s common to the otherwise divergent jatis, so that these could be defined and demarcated uniformly.

In search of the defining features, the state started investing in ethnographic surveys and was gradually drawn into a nexus, which was not always complementary, with the western academic complex. Consensus eluded (and still eludes) any definition of caste. Nevertheless, the ethnological investigations in the context of statistical requirements eventually produced a dogma – of caste being an empirically verifiable entity with a uniform and fixed boundary across the country. Once delimited, caste was endowed with a singular history of origin (racial, occupational, etc), a set of common practices, and a structure. Its pan-Indian architecture included the varna names. Though much doubted on empirical grounds, these categories, because of being known beyond a locality, could still serve the state’s agenda of classifying the jatis. However, in the course of their selective appropriation, brahmin, kshatriya, vaishya and shudra were first empiricalised and then interpreted to be merely denoting ranks. Boundary and hierarchy thus became the two dimensions characterising the new caste system conceived in the context of counting and classification.

The image of caste moved from the pages of the census reports into the domain of the state and the wider academia. The census was an official project and its impact on the state policies was only understandable. Further, the census remains till date the only non-sample based all Indian survey. Hence, it constituted the ground from which it became possible to talk of an empirical and uniform pan-Indian caste. It is not a mere coincidence, as Cohn (1990: 241-42) has pointed out, that most of “the basic treatises on the Indian caste system written during the period 1880 to 1950” were by men entrusted with the supervision of the census operations either at the provincial or all-India levels. Denzil Ibbetson, J A Baines, H H Risley, E Thurston, E A Gait, J H Hutton are a few of those names that can be cited in this regard.

Caste, I believe, cannot be equated either with varna or jati. The components of the varna order are not indeterminate; yet, these are not empirically verifiable. The presence of jatis, on the other hand, can be observed; however, these never had a singular and uniform identity in the Indian society. Caste thus is fundamentally different from both varna and jati; yet, because of its associations with both varna-names and jati-practices struck in the course of the census operations, it has been misconceived as a component of indigenous society.

To illustrate this, let us explore what the expression “brahmin caste”, so commonly used in both academic and everyday parlance, could mean. Translated textually, “brahmin varna” cannot denote a group of people physically existing, though, that is the idea the word caste labours to convey. Its empirical rendering as “brahmin jati”, on the other hand, would be fallacious – there has never been nor could there ever be a jati in Indian society called production, a pan-Indian caste system in its empirical avatar apbrahmin. Kanyakubja brahmin, Maithil brahmin, Namboodiri peared initially towards the close of the 19th century in the docubrahmin, Chitpavan brahmin, etc, are some of the groups re-ments of the state. Hence, the use of the category of caste in place garded in the census reports as brahmin castes. Significantly, in of varna and jati, in historical explanations of the Indian society all these names, the word brahmin neither comes alone, nor as a or in the framing of policies by the state in contemporary times, prefix. It is conjoined as a suffix and actually appears as part of can only be misleading. the identity of the diverse jatis. I should perhaps add before I conclude that I am not trying to

In a similar fashion, the question “what is caste?” routinely suggest that there was no hierarchy or discrimination in society asked in classrooms and beyond generates an ambiguous sense. before the birth of “caste”. Both varna and jati, which have been Is the question about varna or is it about jati? These two, I have different and yet interacting parts of indigenous traditions since repeatedly emphasised, are not identical questions and therefore ancient times, carry their respective notions of hierarchy. In fact, anticipate different answers. Caste, hence, is an idea of recent I believe that by avoiding the generalised structure of hierarchy origin that emerged by displacing the text-based varna order as presented within the caste system, we could understand better on the one hand and suppressing the multifariousness of the jatis the specific constituents of authority that have been operating in on the other. Though there was no prior design shaping its Indian society.


1 Asia, Pacific and Africa Collection, British Library, London (hereafter, APAC), Statistics and Commerce Department (hereafter, SCD), L/E/2/84, Register 5393, No 34, March 1880.

2 APAC, SCD, L/E/7/73, Register 521, No 1840, August 1882.

3 APAC, SCD, L/E/7/73, Register 521, No 91 ½, January 1884.


Baines, J A (1893): General Report on the Census of India, 1891 (London: Eyre and Spottiswoode).

Bayly, Susan (2002 [1999]): Caste, Society and Politics in India from the Eighteenth Century to the Modern Age (Cambridge: Cambridge University Press).

Beverley, H (1872): Report of the Census of Bengal ( Calcutta: Bengal Secretariat Press).

Cohn, Bernard S (1990 [1987]): “The Census, Social Structure and Objectification in South Asia” in his An Anthropologist among the Historians and Other Essays (Delhi: Oxford University Press), 224-54.

Cornish, W R (1874): Report on the Census of the Madras Presidency, 1871, with Appendix (Madras: The Government Gazette Press).

Gait, Edward A (1902): Census of India, 1901: The Lower Provinces of Bengal and their Feudatories, Vol VI A, Part II: The Imperial Tables (Calcutta: Bengal Secretariat Press).

Jaiswal, Suvira (2000): Caste: Origin, Function and Dimensions of Change (Delhi: Manohar).

Marriott, McKim and Ronald Inden (1977): “Towards and Ethnosociology of South Asian Caste Systems” in Kenneth David (ed.), The New Wind: Changing Identities in South Asia (The Hague: Mouton), 227-38.

Neill, J W (1873): Report on the Census of Central Province (Calcutta: Bengal Secretariat Press).

Peabody, Norbert (2001): “Cents, Sense, Census: Human Inventories in Late Precolonial and Early Colonial India”, Comparative Studies in Society and History, 43(4): 819-50.

Plowden, W C (1867): Census of the North Western Provinces, 1865, Vol I: General Report and Appendices, A, B, C, and D (Allahabad: Government Press, North Western Provinces).

  • (1873): Census of the North Western Provinces, 1872, Vol I: General Report and Statements and Tables (Allahabad: North Western Provinces’ Government Press).
  • (1883): Report on the Census of British India taken on the 17th of February 1881, Vol I (London: Eyre and Spottiswoode).
  • Risley, H H (1891): “Caste in Relation to Marriage” in his The Tribes and Castes of Bengal: Ethnographic Glossary, Vol I (Calcutta: Bengal Secretariat Press).

    – (1903): Census of India, 1901, Vol I: India, Part I: The Report (Calcutta: Bengal Secretariat Press).

    Srinivas, M N (1962): “Varna and Caste”, in his Caste in Modern India and Other Essays (Bombay: Asia Publishing House).

    Trautmann, Thomas R (1997): Aryans and British India (New Delhi: Vistar).

    Waterfield, Henry (1875): Memorandum on the Census of British India of 1871-72, presented to both Houses of Parliament by Command of Her Majesty (London: George Edward Eyre and William Spottiswoode).

    Williams, J C (1869): The Report on the Census of the Oudh, Vol I: Report (Lucknow: Oudh Government Press).

    REVIEW OF WOMEN’S STUDIES April 30, 2011
    Women and Water: Issues of Gender, Caste, Class and Institutions – Maithreyi Krishnaraj
    Questioning Masculinities in Water – Margreet Zwarteveen
    ‘They Are Not of This House’: The Gendered Costs of Drinking Water’s Commodification – Kathleen O’Reilly
    Caste, Gender and the Rhetoric of Reform in India’s Drinking Water Sector – Deepa Joshi
    Women and Decentralised Water Governance: Issues, Challenges and the Way Forward – Seema Kulkarni
    For copies write to: Circulation Manager, Economic and Political Weekly, 320-321, A to Z Industrial Estate, Ganpatrao Kadam Marg, Lower Parel, Mumbai 400 013. email:

    August 13, 2011 vol xlvI no 33

    Dear Reader,

    To continue reading, become a subscriber.

    Explore our attractive subscription offers.

    Click here


    To gain instant access to this article (download).

    Pay INR 200.00

    (Readers in India)

    Pay $ 12.00

    (Readers outside India)

    Back to Top