In the whole of India, excluding Burma, Brahmans consti- tuted some 15.2 million individuals in 1931 (not counting some 1.25 million "cultivating Brahmans"), or 4.5% of the total and 6.4% of the Hindu population. Only in a relatively small number of districts, mainly in what is today Pakistan and in largely tribal areas, did they account for less than 1% of the population. Districts in which they exceeded a tenth of the population were concentrated in the central Gangetic Plain, the ancient heartland of Indo-Aryan civilization, and in adjacent regions. Especially notable are the high concentra- tions of Brahmans in the western Himalayan districts, a reflec- tion in part of that area's sanctity in Hindu cosmology and its historical function as a place of refuge for high-caste Hindus during the centuries of Muslim rule in the adjoining lowlands. Another refuge area that stands out for its high Brahman con- centration is South Kanara district, to which many Brahmans migrated during the long period of Portuguese rule in the neighboring region of Goa. With the single exception just cited, one is struck by the relative scarcity of Brahmans in peninsular India, where no more than a dozen districts have as much as 5% of their population in that caste group.

Rajputs, the principal group generally regarded as within the Kshatriya varna (i.e., among those who are Hindus), ac- counted for more than 11 million individuals in 1931. Of that number some 9 million were Hindus, about 50,000 Sikhs, and 2.1 million Muslims, presumably descended from Hindu con- verts. On plate X.C.2 Muslim Rajputs are mapped separately from the others. Irrespective of religion, Rajputs are divided into scores, if not hundreds, of separate clans that tend to be hierarchically graded within any given region. Among these clans hypergamous marriages are common (i.e., brides are given to members of clans of higher status than one's own). Although we have attempted on our maps to indicate the names of the principal Rajput clans in those areas in which they are particularly common, we have not begun to do justice to the complexity of their distribution. Most areas in which Rajputs are at all numerous will have representatives of a di- versity of clans; and most large clans are more widespread than our maps suggest.

The concentration of Hindu Rajputs in northern and cen- tral India, and of Muslim Rajputs in Punjab and the United Provinces, is pronounced. It is remarkable, however, that Raj- putana (now Rajasthan), an area that in the popular mind is most closely associated with the martial Rajput tradition and chivalric life-style, has relatively fewer members of that group than many of the former princely states of Central India. In only one state of Rajputana, Jaisalmer, did Rajputs form over a tenth of the total population. The greatest relative strength of Rajputs lies in the western Himalayan region, the refuge area to which—as with the Brahmans—so many of the caste presumably repaired during the period of Muslim rule on the plains. In portions of this region Rajputs actually exceeded half the total 1931 population.

As may be seen from maps (e) and (f), jatis traditionally associated with mercantile occupations are widespread in In- dia and are to be found not only among Hindus, but also among Jains (most of whom are associated with business), Sikhs, and Muslims. Regrettably, the 1931 census treatment of such groups varies considerably from one province, princely state, or states agency to another.

In some areas all or most such castes are grouped under a single heading (e.g., "Vaishya," the varna category to which most mercantile castes are traditionally assigned, in the United Provinces; "Bania" in Bihar and Orissa, the Central Provinces, and Gwalior; "Mahajan" in Jammu and Kashmir). Elsewhere (e.g., Rajputana), a large number of individual castes were enumerated. Typically, we have grouped several castes in ar- riving at the district totals indicated in map (f), and not infre- quently in map (e) as well. The specific names on the map will indicate not only most of the major caste names aggre- gated, where specific data were available on them for 1931, but also a number of additional names (e.g., "Agrahari" and "Baraseni" in the United Provinces) that were not separately listed in the 1931 census but that referred to mercantile castes locatable from previous censuses and other sources.

The inclusiveness of map (e) and to a lesser extent of map (f) entailed many more or less arbitrary decisions. Readers with a good knowledge of the caste system may, for example, find the inclusion of several Hindu caste groups, most notably "Shaha" and "Sunri" in northeastern India, peculiar in that those castes, because of their traditional association with dis- tilling, have not been accorded the high ritual status attrib- uted to the other groups named. But it should be noted that such groups have been counted only where there was evidence to support their recent upward mobility and economic suc- cess, a typical concomitant of which is a changeover to a more ritually pure "Sanskritized" life-style. Similarly, one may ques- tion the absence on the map of other castes that combine mer- cantile and other pursuits, for example, "Telis," an oil-pressing caste, a great many of whom have become grocers and petty tradesmen of other kinds. The basis of their exclusion, which is admittedly arbitrary, is that by and large their status remains rather low, despite their mercantile involvement. Finally, we note the unavoidable omission in Mysore and the Bombay Karnatak of quantitative data for the Banajiga Lingayats, a mercantile subgroup within the great Lingayat sect of Hindu- ism, all of whom have been returned as a single caste by the 1931 census. In the light of these remarks it is obvious that, for Hindus in particular, any estimate of the number of indi- viduals in high-status mercantile castes, as of 1931, must be approximate. For what they may be worth, we suggest totals of roughly 6.5 million Hindus, 1.25 million Jains, 850,000 Muslims, and 200,000 Sikhs. Those business castes, of which Oswal is the largest, wherein Jains appear to be more numer- ous than Hindus are indicated by a "(J)" on map (e). There was in 1931 no mercantile caste in which Sikhs outnumbered Hindus; but significant numbers of Sikhs were found among both Aroras and Khatris.³

Although their claims to belong to the Kshatriya varna are disputed, few would question the high status enjoyed by the group of scribal castes whose distribution is depicted in map (g). Their total numbers as of 1931 were just short of 3 mil- lion, the great majority of whom were Kayasthas. Important throughout the Gangetic Plain, they were particularly promi- nent in Bengal, where, in a number of districts, they were the most numerous Hindu caste. In both Calcutta and Chittagong districts they accounted for more than a tenth of the total population.

Sources

L. K. Anantha Krishna Iyer (1909, 1912), (1928–35), (1937–41); J. Bhattacharya (1896/1968); A. H. Bingley (1899); P. D. Bonarjee (1899); W. Crooke (1906); R. E. Enthoven (1920); D. C. J. Ibbetson (1916); India, Census (1901), (1911), (1921), (1931), (listed under Government Documents); E. J. Kitts (1885, listed under General Refer- ences); A. Mitra (1953); H. H. Risley (1891); H. A. Rose (1911–19); R. V. Russell (1916); E. Thurston (1909); see also all sources for plate X.C.1 for Ceylon.

X.C.3 Castes, Tribes, and Other Ethnic Groups, 1931, Arranged by Functional Categories

The five maps of plate X.C.3 are intended to expand the in- formation of plate X.C.1 with respect to the latter's functional groups D, E, F, G, and I and to complement the data of plate X.C.2, which served a similar function for groups A, B, and C. It will be noted that in showing the proportions of the groups named to the total population of each district or princely state we have selected a set of class intervals different from that of plate X.C.1. By studying the two plates together the reader will therefore often be able to arrive at a closer idea of the strength of a particular group in a given region than he could gain from a single map.

The inclusiveness of map (a), "Most Numerous Landholder Castes," is indicated below the title. As we use the term, "Land- holder Caste" is not quite synonymous with "cultivator and/or pastoral castes" (category D of plate X.C.1), in that it also includes Rajputs (category B), who are traditionally landhold- ers but do not themselves practice cultivation. For the pre- dominantly Hindu areas of India, then, as well as those parts of Punjab where Hindus and Sikhs together outnumbered Muslims in 1931, map (a) is an even closer approximation of a map of "dominant castes" than was the map on plate X.C.1, in the text for which that concept has already been discussed.

In contemplating map (a) one is struck by the large, com- pact areas in which particular landholding castes are most numerous: Jats in Punjab and most of neighboring Rajputana, Maratha-Kunbis in virtually the whole of the present area of Maharashtra and a few adjacent districts, Goalas over most of what is now Bihar, and so forth. One is also struck, in con- sidering the quantitative aspects of the distributions, by the fact that the modal class for the strength of the leading land- holding caste is 10–20%. This suggests to us that observations in the anthropological literature about the numerical strength and influence of dominant castes in particular regions are fre- quently exaggerated. Districts in which the proportion exceeds 40% are extremely rare, being limited to Rajput-dominated areas of the western Himalayas and the Maratha–Kunbi-dom- inated area astride the Western Ghats to the north of Goa. A relationship between this fact and the successful military ca- reer of the Maratha leader Shivaji appears obvious.

Map (b), "Most Numerous Muslim Sub-Groups," presents a distribution pattern strikingly different from that of the map just discussed. In both northwest and northeast India, now 3 It should be noted that the Sikhs, and less strongly the Jains, in theory reject the ideas of varna and jati and claim that all are equal in the faith. Nevertheless, caste distinctions have been made in both sects and still are prevalent, although in recent times many Sikh and Jain leaders are striving to eliminate them. Pakistan, Kashmir, and Bangladesh, areas that were largely blank on map (a), one finds large blocks of territory, roughly coterminous with linguistic regions (see map X.B.1), in which the most numerous Muslim ethnic groups (Jats in Punjab, Sammo Sindhis in Sind, Sheikhs in Bengal, etc.) typically comprised a percentage of the total population much greater than that of the leading Hindu landholding caste in what is today India. Proportions below 20% were relatively rare, be- ing confined almost exclusively to Punjab. In other provinces or princely states proportions over 40% were the rule, a num- ber of districts actually registering proportions above 80% in a single group. Thus, while the predominantly Muslim areas of South Asia were ethnically diverse, the degree of diversity encountered therein was markedly less in 1931 than in the mainly Hindu areas. Since independence, the local reduction in ethnic diversity in both Pakistan and Bangladesh has un- doubtedly been very pronounced. At the same time, however, there has probably been a hightened awareness within Paki- stan of the sharp ethnic differentiation from one province to another. This was obviously the case in the period leading to the separation of Bangladesh, and it appears to be true also in the now-truncated state.

Of the dozen Muslim ethnic groups named on map (b) in the non-Muslim majority areas of India and Ceylon, the Sheikhs are by far the most widespread and most numerous. Putatively of Arabian origin, the Sheikhs of South Asia are one of four groups collectively regarded as constituting the social category known as "Ashraf" (roughly, "gentry"), the other three groups being the Saiyads (discussed above, plate X.C.2, map (a)); Mughals, of Central Asian (Mongol) ori- gin; and Pathans, of Afghan origin. Given the relative impacts of invaders from Arabia, Central Asia, and Afghanistan on Indian history, one wonders about the grossly disproportionate representation of their respective alleged descendants in the Muslim population of the subcontinent. This, however, is a matter on which we shall not speculate.

Of the remaining Muslim groups, several are occupationally specialized (e.g., the mercantile Memons and Bohras and the Jolahas, a jati of weavers); one, the Meos, is a converted tribal groups; and at least two, the Mapillas of Malabar and the Cey- lon Moors, are said to be descendants of Muslims arriving peacefully by sea and intermarrying with the indigenous pop- ulation. The strength of the most numerous Muslim group in the non-Muslim majority areas rarely is as high as a fifth of the total population, and generally it is under a tenth. One is impressed with the large number of districts, in light green, in which Muslims as a whole constitute more than 5% of the total population, but in which no single group attains that percentage.

Map (c), "Ethnic Burmans," relates to all those ethnic groups described by the 1931 census as "racially" Burman, irrespective of their religion (though the vast majority are Buddhists), as contrasted with all other indigenous groups within Burma, whom we have arbitrarily classified as tribes and considered on map (e). The two maps obviously com- plement one another, jointly covering all of Burma's 1931 population save for the immigrant Indians, Chinese, and Euro- peans. The preponderance of a single group, the Burmans proper, in the Irrawaddy core region of the country, where they typically exceed four-fifths the total population, is strik- ing, as is the relative scarcity of Burmans in certain peripheral regions. One should also note the several distinctive coastal regions where "racially" Burman groups predominate, in which local subgroups, Arakanese, Merguese, and so forth, are more numerous than the Burmans proper. The implications of this pattern with respect to the achievement of national unity in contemporary Burma are fairly obvious.

Map (d), which portrays "The Most Numerous Depressed Castes" among the combined Hindu and Sikh populations of India as of 1931, is, on the whole, somewhat simpler in pat- tern than map (a), largely because of the very large area of northern India over which a single caste, the Chamars, pre- dominates. Traditionally leatherworkers, the Chamars are in practice mainly agricultural laborers, as are most of the other caste groups mentioned, whether or not they have some other traditional occupation. Studying map (d) together with plate X.C.1 one notes that, apart from the Chamars and, to a lesser degree, the Madigas and Paraiyans in what are today Andhra Pradesh and Tamil Nadu, it is rare for any single depressed caste to be the most numerous single ethnic group in a district of India. In only a handful of districts or princely states does such a caste constitute over a fifth of the total population. While proportions over a tenth are common, one also notes a rather large number of districts wherein depressed castes as a whole exceed 5% of the population (cf. plate X.A.8), but no single caste attains that percentage.

The "Most Numerous Tribes" of South Asia, exclusive of those Muslim groups sometimes described as tribal, are de- picted on map (e). The areas covered on this map are for the most part hilly or mountainous and largely forest-covered, ob- viously serving as regions of refuge for aboriginal groups un- able to compete effectively against the more highly civilized