The Linguistic Survey of India’s Experiment in Mapping Languages, 1896–1927

ABSTRACT The Linguistic Survey of India (LSI), supervised by the Anglo-Irish linguist and civil servant George A. Grierson, surveyed and classified more than seven hundred languages and dialects. An integral part of the state-funded survey was mapping where languages were spoken in India, which resulted in the publication of 45 language maps between 1899 and 1927. As individual maps they are comparable to other thematic maps of the period. As a series of language maps, however, they are inconsistent in scale, colour, use of relief and labels, and depiction of boundaries. This paper argues that the inconsistency in presentation reflected the LSI’s experimental approach to mapping language, trying to reconcile the approximate representation of a complex geographical phenomenon with the colonial state’s expectation for accurate and up-to-date language maps.

The colonial Linguistic Survey of India (LSI), supervised by the Anglo-Irish linguist and member of the Indian Civil Service, George A. Grierson, set out in 1896 to identify, classify and map the languages and dialects of India. 1 Over the course of three decades, the LSI surveyed 179 languages and 544 dialects with the results published non-sequentially in twenty volumes between 1903 and 1928. 2 The stated empirical basis for the LSI were three linguistic 'specimens'-a translated 'Parable of the Prodigal Son', a piece (or pieces) of vernacular prose or verse, and a translated list of 241 common words or phrases-acquired from each language and dialect surveyed.However, an integral but unstated component of the LSI was identifying where a language or dialect was spoken, and who spoke it.The geographical information obtained was featured in descriptions that prefaced the published specimens, locating where a language or dialect was spoken in India and its geographical relationship to other languages.More significantly, this material formed the basis for a series of 45 language maps inserted across the published volumes of the LSI.
This article examines a cross section of the 45 maps produced by the Linguistic Survey of India.Read as individual maps, they echo the types of thematic maps common at the turn of the twentieth century. 3The LSI language maps, however, are inconsistent in scale, colour, use of relief and labels, and depiction of boundaries.This article argues that the inconsistency reflected the LSI's experimental approach to mapping language and was an attempt to reconcile the expectations of a government, which had overcome its reluctance to fund the LSI based on the promise of mapping language, with the difficulty of mapping a geographically complex phenomenon. 4The LSI tested many different approaches as it attempted to find the most appropriate way to visualize and map languages that overlapped, had no clearly defined boundaries, and where many spoke more than one language or dialect.The LSI's experimentation implied to a diverse audience that language maps could only be approximate geographical snapshots of a heterogeneous and mutable phenomenon.
The Linguistic Survey of India emerged at a time when language mapping had become a distinctive feature of European statistical science and cartography.The advent of the enumerative census enabled nations and empires not only to gather a variety of statistical data on their populations but also to use the data to map features such as language. 5Such efforts were intrinsically linked with the development of nationalism in Europe and the growth of imperialism in regions such as South Asia as a means of knowing, classifying and trying to manage new and diverse populations. 6ichard Böckh's Linguistic Map of the Prussian State According to the Census of the Year 1861 was an example of how statistical mapping based on census data could be mobilized for ideological and nationalist objectives, cultivating the idea of a German nation. 7Technological advances in the nineteenth century such as the choropleth map also equipped a generation of mapmakers with new methods for showing more complex data and mapping multiple variables. 8istorians of mapping have noted that many of these (often state) exercises to map populations and languages occurred on the peripheries or boundaries of nations and empires, where centralized power was insecure and contested, in places such as the Alsace (Franco-German border) or the Baltic region and other peripheral areas of the Russian Empire. 9Language mapping was not just the domain of states or governing authorities: learned societies, travellers and private scholars also took advantage of technological advances to make maps of languages for various reasons, including linguistic research.There was a tradition of language maps being produced for philological studies of India, for example, those appearing in the work of John Beames and Robert Needham Cust, among others. 10owever, map historians have scarcely focused on the type of cartographic exercise that this paper addresses: the linguistic or dialect survey.These surveys, which appeared in the nineteenth century, had clear overlaps with statistical exercises such as the census.But they also functioned on more specific linguistic terms, mapping variation and change as much as distribution.Notable examples include Georg Wenker's Sprachatlas des Deutschen Reichs (surveyed between 1876 and 1887) in Germany, and Jules Gilliéron and Edmond Edmont's Atlas linguistique de la France (1902-1910). 11These systematic exercises began to develope common methods of surveying and mapping such as having 'surveyors' in situ translate materials in the local dialect.'The Parable of the Prodigal Son', for example, had been used as a set text in other surveys and would later feature in a survey of Irish dialects. 12Furthermore, linguistic or dialect surveys were often located somewhere between the state and scholarly research, which provides an interesting case study for how mapping and map making navigate these sometimes complementary but often conflicting domains.
Linguistic or dialect surveys have largely been examined by linguists in geographical or mapheavy subdisciplines such as dialectology, yet such work has often disregarded the LSI as a significant example, partly because of its perceived lack of rigour in using untrained surveyors, but perhaps also out of an overly Eurocentric emphasis. 13The last point could be made of the broader study of language mapping that has been overly Europe-and North America-oriented, and there remains much to do on the history of language mapping elsewhere.In other colonial and post-colonial areas similar programmes were undertaken to survey and map languages.In Asia, for example, there was the aborted Linguistic Survey of Burma (Myanmar) and a partly completed survey of Afghanistan. 14ecentring studies of language mapping and surveying also allows us to confront the problematic concept of an ethnolinguistic map in nineteenthcentury cartography. 15Grierson himself pushed back against colonial anthropology's propensity to identify one language with one race or group, taken from European notions of language and nationality. 16As Grierson argued, languages did not neatly align with racial or ethnic categories; languages borrowed from each other, one ethnic group might speak multiple languages, and languages could vary within a particular ethnic group depending on social position. 17The LSI makes this explicitly clear on its maps, referring to the 'localities' or 'areas' where languages were spoken without making any claims regarding the ethnicity or racial identity of the speakers.

Surveying Language
At the 1886 International Congress of Orientalists in Vienna, George Grierson proposed the first 'deliberate, systematic survey of the languages of India, nearer and further'. 18Armed with a unanimous resolution asking the government in India to launch a linguistic survey, Grierson returned to his post in Bihar (then part of Bengal Province) and set out to assemble the necessary intellectual and government support for such a scheme.Concerns about the cost of such an expansive survey and lack of support from provincial governments, however, delayed its start by a decade.Although the 1881 Census, the first to include a language question, revealed serious deficiencies in the state's knowledge of what languages were spoken in India, Grierson had great difficulty convincing the authorities that the LSI should be completely funded and supervised by the state rather than a university or the Asiatic Society in Calcutta.
The 1891 Census provided further evidence of the problem, leading the Census Commissioner to publicly call for a linguistic survey to establish what languages were spoken in India, by whom and in which localities. 19The Linguistic Survey of India, as it came to be known, formally commenced in 1896 after managing to convince the colonial government to fund a more financially restrained scheme.A decisive reason for the government's reversal was the promise of mapping, and making language maps, of India's complex linguistic geography.This was reflected in the financial arrangements imposed on the LSI.Although Grierson had to justify costs and expenses incurred by surveying, the production of language maps was to be financed by the Government of India who ordered the Map Publication Office of the Survey of India to print as many maps-and especially more expensive coloured maps-as the LSI required. 20overnment support ensured the LSI had the necessary personnel and infrastructure to survey and map India's languages and dialects.Various officials, military officers, schoolteachers and other on-the-ground correspondents, both European and Indian, were brought together in an informal network of language 'surveyors' (a term Grierson used inconsistently).Their task was to gather linguistic specimens and information from their localities, which were subsequently edited and proofed by Grierson and numerous experts across India and internationally. 21Maps were produced in collaboration with the Survey of India, using the geographical information provided by the LSI's surveyors.Once coloured and printed, using different techniques, by the Survey of India, the language maps were inserted into their respective volumes, arranged by language family (Indo-Aryan, Dravidian, etc.) and subgroup.Although the printed specimens and descriptive prefaces outlining the history, geography and status of each language or dialect formed the bulk the LSI's published material, the 45 language maps (four in black and white, 41 in colour) functioned as a cartographic reference for India's linguistic geography.
The utility of detailed language maps to the colonial state was obvious.It addressed the deficiency of knowledge about Indian languages and enabled officials to understand and govern better the areas to which they were posted.That explains why the LSI volumes with maps were issued not only to the imperial and provincial governments in India, but also to Oxford, Cambridge and the University of London, the primary training centres for future members of the India Civil Service. 22Perhaps the expectation was that the maps, in addition to the other language materials surveyed, would be used by junior officials to identify which language or dialect they needed to learn before arriving at their post.
We have scant evidence, however, of the maps being used by trainees or officials to improve governing.More copies of the LSI, printed by the Government of India, were issued to university and research libraries in India, Europe and North America (for example, the British Museum, Harvard University), to Oriental and other learned societies, and to scholars and linguists. 23rierson would privately suggest that the Government of India never settled on a purpose for the LSI or made the most of its results, seeing it as another data collection and language documentation exercise. 24Such ambivalence meant the LSI operated in the 'grey area between officialdom and "private scholarship", balancing both against each other'. 25Experimenting with various methods of mapping language balanced the expectations of the state, which financed the LSI, with the complex geographical picture becoming evident from surveying India.

Experimenting with Maps
In the first substantive study of the LSI, Javed Majeed makes a compelling argument that its strength lies in 'the Survey's acknowledgement of its shortcomings and provisionality which bolsters its authority rather than any expressed sense of the mastery of India as a linguistic region [that] reflects the nature of its work in a complex linguistic environment'. 26The information gathered by the surveyors sketched out India's complex and indefinite linguistic geography, where many languages were mixed and borrowed from each other; where similar languages transcended boundaries or were spoken in seemingly unrelated areas; where many communities spoke two or more languages; and where some languages were spoken by mobile or nomadic communities.
For Grierson, language maps presented approximate geographical delineations of what was in reality fluid or indefinite and difficult to map.As he succinctly explained in the Introductory volume to the LSI, 'Indian languages gradually merge into each other and are not separated by hard and fast boundary lines'.Thus, maps indicating a territorial or bounded geography: must always be understood as conventional methods of showing definitely a state of things which is in its essence indefinite.It must be remembered that on each side of the conventional line there is a border tract of greater or less extent, the language of which may be classed at will with one or other. 27derstandably, linguists have focused on the linguistic implications of the LSI's views on how language operates in space. 28Grierson's statement was, however, focused on reading and interpreting language maps.With the optimistic expectation that the LSI would be used not only by linguists but also by officials, students and laypersons, the statement instructed users that the maps in the Linguistic Survey should not be read or understood as a political, administrative and even topographical map, but as a pragmatic attempt to locate and outline roughly where a language was spoken.Grierson was fully aware that the intended audience for the LSI would have been better acquainted with, and perhaps even trained in using, topographical and administrative maps where boundaries and shaded or coloured areas reflected a political reality (for government, if not for ordinary Indians).
Although Grierson warned against conflating a language map with a political map, the LSI could not disentangle politics because the promise of mapping languages had helped to fund the survey.However, the pronounced variation of size, features, scale, representation and methods used in the series of 45 maps reflected the LSI's pragmatic approach to mapping language.The LSI was attempting to find the most effective way of using available methods to map the particular, sometimes peculiar, geography of the language(s) in question.It experimented with different ideas, methods and techniques to map India's 'complex linguistic environment' and avoided, or at least minimized, the interpretation of 'hard and fast' boundaries.As Majeed points out, the key word deployed in much of Grierson's correspondence about language statistics is 'approximate' and this is evident in the LSI's mapping. 29aps were made in collaboration with the Survey of India, which had the infrastructure and skilled draftsmen needed to prepare and publish them.Grierson presumably had in mind from the beginning which language maps he wanted to publish in the LSI and used base maps supplied by the Survey of India on which to sketch or outline language areas from the data gathered.Like the variation in the scale and content of maps, the map-making process was inconsistent.Grierson would sometimes be involved in choosing which colours to use, yet in other cases he largely deferred to the Survey of India when it came to colour and labelling. 30rierson later admitted that his role in making maps involved drawing 'rough indications of what was required, rather than finished maps', whereas 'the maps for the Survey have, for the past twenty-five years, been prepared for me, without question, by the Survey of India'. 31The first maps were completed as early as 1899. 32Thirty of 45 maps were made by 1903, while the final fifteen were finished between that year and 1927.The length of time it took to produce the final fifteen maps might explain some of the variation, since personnel, methods and technology changed at the Survey of India. 33However, the maps completed between 1899 and 1903 were also diverse in presentation, suggesting length of time was not a significant factor.
Mapping language was not an exact science nor was making a language map.Grierson described his own approach to drafting a language map as a process of trial and error, experimenting on base maps supplied by the Survey of India and other sources.In response to an enquiry from Edward Stanford Ltd. in London asking whether he would like to purchase more copies of their map showing Eastern Europe and Western Asia for a map he was working on, Grierson wrote, 'I am sorry that I have given you so much trouble with this subject, for I find, after various experiments, that the map regarding which you have written will not suit my purpose'. 34By adopting a process of trial and error, Grierson was seeking answers to questions presented by trying to map language: How detailed and at what scale?How to map languages which were spoken both within and beyond India?How to map a multilingual area with many overlapping languages and dialects?How to map the geographical margins of a language?The following part of the paper examines in more detail the different ways in which the LSI mapped India's complex linguistic environment.

Scale
The LSI had two scales to consider when making a language map: the scale of the map itself and the taxonomic scale of language to be mapped.The LSI generally classified languages by family and branch, followed by language group or language, and then specific dialects or subdialects.For example, Bengali, with several dialects and subdialects, was classified as part of the eastern group of the Indo-Aryan language branch of the much broader Indo-European family.It made practical sense for the Survey to show the higher classification of family and branch, which generally covered a larger area, on a smaller-scale map of the entire Indian subcontinent, while presenting less widely spoken languages on larger-scale maps showing more detail.
The Introductory volume of the LSI included five synoptic maps that presented an overview of the main language families in India, namely the Indo-Aryan, Dravidian, Austroasiatic and Sino-Tibetan families.The Sino-Tibetan family was allotted two maps, one for the Tibeto-Burman subfamily and the other for what at the time was termed the Siamese-Chinese subfamily.Four of the five maps (except Siamese-Chinese) used a 1 inch to 160-mile base map showing the entire Indian subcontinent, allowing the LSI to convey the vast geographical range of each language family.The Indo-Aryan family, for example, extended from Kashmir in the northwest to Assam in the east and to the Deccan plateau and Hyderabad in southern India (Fig. 1).Colour was used to distinguish three sub-branches of the Indo-Aryan languages.
Most of the LSI's maps, however, used a scale appropriate to the estimated areal distribution of a language group, a language, or a group of dialects being mapped.Bengali, a major Indo-Aryan language in eastern India, was mapped on a 1 inch to 64-mile base map covering an area which today includes West Bengal (India), Bangladesh and surrounding regions (Fig. 2).At that scale the LSI was able to delineate the dialects and subdialects of Bengali, which were identified on the map by red lettering.Less widely spoken languages were accordingly mapped on larger-scale base maps.The map of the Kashmiri language, an Indo-Aryan language spoken in the part of Kashmir around Srinagar, used a 1 inch to 16mile scale that was detailed enough to allow the map user to identify where several minor and mixed dialects were spoken. 35Scale mattered for clarity and detail, which the LSI chose on a pragmatic, ad hoc basis, creating maps adjusted to a language's geography or the linguistics details the LSI wanted to display.The Survey's ad hoc approach to scale is evident in the last of the five synoptic maps showing where the Siamese-Chinese languages were spoken. 36As the name indicates, the languages that formed part of the Siamese-Chinese subfamily were principally found in Burma (then a province of India) along the frontier with China and Siam (now Thailand).Rather than providing a uniform and consistent presentation by using a 1 inch to 160-mile map of the subcontinent, the LSI chose a 1 inch to 128-mile map covering only Burma and the border with China and Siam (Thailand).Such a map proved more useful since it enabled the reader to identify in greater detail the geography of constituent languages in the less-widely distributed Siamese-Chinese subfamily of the Sino-Tibetan languages.
The Dravidian languages map in the Introductory volume illustrated the importance of choosing the right scale and base map, when visualizing the geography of language.Most Dravidian languages were spoken in southern India, but a small number could be found in northern India (Fig. 3).One such outlier was Brahui, spoken in Baluchistan in the western part of the map.Rather than creating a separate inset map for Brahui, the 1 inch to 160-mile map of all India allowed the map user to comprehend the non-contiguous geography of the Dravidian languages and the relative isolation of Brahui.Although such a peculiar geographical distribution may not have mattered to 76 P. Jagessar Imago Mundi 75:1 2023 non-linguists, it had implications for reconstructing the historical development of the Dravidian languages, suggesting scholarly usefulness was important in choosing the right scale to map a language's geography. 37he Dravidian and Indo-Aryan maps, like the other synoptic maps in the Introductory volume, were cartographic summaries of the more detailed large-scale maps of languages.Whereas the map of Bengali (see Fig. 2), for example, used a scale appropriate for showing the geographical distribution of the specific language and its dialects, the synoptic maps were a map of maps, assembled from the other language maps in the LSI.The Introductory volume provided a descriptive summary of the languages and dialects identified by the Survey, and the synoptic maps included in it enabled the user-whether an official or scholar-to acquire an outline of India's linguistic geography before turning to the other language maps for more detail. 38Grierson perhaps envisaged a junior official about to be stationed in a remote corner of Bengal consulting the Indo-Aryan map to determine which Indo-Aryan languages were spoken there (see Fig. 1), and after identifying Bengali as the primary regional language, turning to the Bengali map to ascertain which dialect or sub-dialects were spoken in the district they were assigned to (see Fig. 2).Officials could do the same with other areas and languages of India mapped by the Linguistic Survey.However, the seeming uniformity of the synoptic maps in the Introductory volume masks the distinctiveness of the other maps published by the Survey.The heterogeneity of India's linguistic geography, where one language might be more widely distributed or be spoken alongside a number of other languages, meant the LSI's maps were equally uneven.While underlying features of the base mapnatural phenomena, settlements, administrative boundaries-were often retained as points of reference for the user, features such as relief were also experimented with.

Topography and Frontiers
Surveying languages along India's remote, mountainous northwestern and northeastern frontiers was a challenge for the Linguistic Survey.Many areas were inaccessible except to well-prepared expeditions, and British administrative control over the frontier remained tenuous in the early twentieth century, meaning the LSI could not rely on government officials to efficiently gather upto-date information. 39Instead, Grierson depended on the reports and personal knowledge of officers and scholars who had traversed the regions, many of whom were capable linguists themselves such as the Anglo-Hungarian archaeologist Aurel Stein. 40tein and other travellers were able to roughly locate for Grierson where languages, and in many cases little-known dialects, were spoken along and beyond the frontiers of India.
The visual representation of topography on Survey of India base maps emphasized the difficulty of acquiring information on frontier languages and dialects.The map of the Dardic languages, for example, enabled the user to appreciate the rugged landscape of northwestern India in which its constituent languages were found (Fig. 4). 41There was also a linguistic rationale for showing relief.Natural phenomena such as mountains and valleys were thought to be a factor behind language change by separating communities over time and could explain variation between dialects. 42The Dardic map not only outlined where the Dardic languages were spoken, but also provided an insight into how the environment might have shaped their linguistic relationship.
Figure 4, however, presented the geography of the Dardic languages as unbroken within a neatly defined area.In reality much of the Indian frontier was sparsely populated, and dialects were unevenly distributed.Making that point was a map of the Ghalchah languages which were spoken by communities at the intersection of Kashmir, Afghanistan and Central Asia (Fig. 5). 43In this case the LSI chose to present the location of the Ghalchah-speaking population on a planimetric rather than a topographic map. Figure 5 therefore indicates which languages are known and approximately where they are located but makes no claims about their A more pressing question was deciding how far the Linguistic Survey of India should map languages that were also spoken in territories outside of India's jurisdiction.Languages invariably crossed artificial political lines and the demarcation of international boundaries rarely accounted for shared linguistic identity, often splitting communities speaking the same language.A prominent example was the Durand line between Afghanistan and the North-West Frontier Province, which split the Pashto-speaking areas between the two jurisdictions. 44The Survey's map of Pashto, however, did not stop at the political boundary between India and Afghanistan but traced in colour where Pashto was spoken well into Afghan territory. 45olitical frontiers were not linguistic frontiers and by mapping where languages such as Pashto were spoken outside of India, the LSI distinguished a 'linguistic India' from a 'political' or 'territorial' India.Surveying Balochi, an Indo-Iranian language spoken in both the British-administered Baluchistan Agency and Qajar Persia, showed not only the lengths to which the LSI went at times to map languages spoken beyond India but also the arbitrariness of deciding how far to map.Dissatisfied with existing geographical knowledge of Balochi, Grierson asked diplomats and officers stationed in eastern Persia to help identify areas where it was spoken by a majority or a minority in order to prepare an accurate and complete map. 46The Political Resident in the Persian Gulf returned a map of South Eastern Persia to Grierson showing an approximate and arbitrary 'straight line from the junction at Minab and Biaban districts on the coast to Nasratabad Sipi', based on the best estimations of agents working in the region. 47rierson made further enquiries to find a more precise linguistic geography with little success.Consequently, he duplicated the Political Resident's straight line on the LSI's map of Balochi as its westernmost boundary in Persia (Fig. 6).The length of time it took to furnish this information-nearly a year and a halfmay have played a part in Grierson's decision to give up waiting for more information and simply reproduce the line.Grierson had sent his request in early 1918, at a time when the Political Agents in eastern Persia were preoccupied with ongoing events related to the First World War and unlikely to prioritize work for the LSI. 48he arbitrariness of the straight line on the Balochi map, however, speaks to inconsistencies in the Linguistic Survey's mapping of languages beyond the frontier.It acknowledges on the map that 'linguistic India' was not confined by international boundaries but was constrained by its own criteria to produce a Linguistic Survey of India only.The LSI did not routinely try to map the geographical limits of languages spoken beyond the frontier.The Siamese-Chinese map, for example, only identified constituent languages spoken in eastern India and Burma at the border with China and Siam.
The LSI's map of Tibetan dialects acknowledged they were spoken in Tibet by shading the Himalayan frontier between India and Tibet, but it does not indicate how far into Tibet these dialects might be found or delineate different forms of Tibetan. 49Grierson could justifiably point to limited Western knowledge of Tibet, but the LSI's understanding of linguistic geography was contingent upon having knowledgeable actors available.
It appears that Grierson did not choose to correspond with local scholars or British officials in China, Siam or Tibet to survey and map languages beyond India's northern and eastern frontier as he had in Afghanistan and Persia.Grierson's own bias for Kashmiri and other languages spoken in the North-West Frontier, in which he had a great expertise and interest, could have influenced these decisions. 50

Multilingual Areas
A fundamental problem for the LSI was how to convey on a map the diversity of languages in a given area.The dialect areas outlined on the map of the Bengali language (see Fig. 2) were also locations in which non-Bengali languages were spoken, while bilingualism and even multilingualism were not uncommon throughout India.As the consul for Seistan and Kain in Persia reminded Grierson in their correspondence about the Balochi map, while 'Balochi is the mother tongue of the Baluch tribes, it is also spoken by the Brahui tribes'. 51In trying to map a geography which was, in Grierson's owns words, 'indefinite', the Linguistic Survey used many different innovative methods to map languages that were not neatly circumscribed but were spoken in areas with multiple languages or where languages were mixed. 52The Balochi map simply used different shades of the same colour to distinguish areas where it was spoken by a majority (deep red and deep blue) and areas where it was in the minority or spoken by scattered communities (pale red and pale blue) (see Fig. 6).
A more common method, however, was to use text and labels to identify language admixture or areas where multiple languages were spoken.For example, 'Western Bengali also spoken here' is written in the southwest corner of the Bengalilanguage map just outside the area identified in blue as the principal Bengali speaking region (see Fig. 2).This label informed the map user that the western Bengali dialect was spoken by scattered communities or as a second language in an area identified as part of another Indo-Aryan language, Oriya.Shaded areas seldom indicated the maximum distribution of a language, but instead designated where it was principally spoken.The LSI's map of Assamese included a label in the area south of the principal Assamese-speaking area informing the user that 'Assamese & Mayang spoken sporadically here'. 53he LSI's maps used labels and text to emphasize the multilingual geography of India where languages and communities overlapped.The LSI used labels such as 'Bihari also spoken here', 'Mixed Oriya and Bengali' and 'Bihari (Magahi) also spoken here', on a map showing the intersection of Bengali, Oriya and Bihari as an expedient method of condensing a complex linguistic situation for the map user. 54Yet such labels and text also sustained the LSI's language of approximation, reinforcing Grierson's view that languages could not be mapped discretely or precisely nor should language maps be read as a conventional map.
The 'Aryan Languages and Dialects Spoken in East Chota Nagpur' map experimented with other map-making techniques (Fig. 7).Here the mapmaker combined four different methods using colour, labels, dots and explanatory notes in the legend. 55Magahi and Bhojpuri (both part of the Bihari languages), along with Oriya and Bengali are assigned separate colours which are maintained when labelling dialects (thus Saraki is written in orange as a Bengali dialect).The map, unusually, also employed dots to represent mixed linguistic communities.For example, the legend indicates that red dots on a yellow background showed the 'area in which Bhojpuri (Nagpuria) is the main language, but in which Bengali (Saraki) is also spoken'. 56o other map in the LSI used this many techniques or tried to map language as precisely as the East Chota Nagpur map, a decision which can be explained by the complexity of the map itself.The elaborate description for each dotted area did not make it any easier to identify the languages spoken, while the crowded labelling made it more difficult to read.In trying to create a more detailed language map that better reflected the linguistic reality of the area, the East Chota Nagpur map pushed the boundaries of pragmatism and user-friendliness too far.

Language Boundaries
The use of labels and dots on the East Chota Nagpur map also functioned to minimize the incorrect interpretation of 'hard and fast' language boundaries that Grierson warned of.For the LSI, boundaries were artificial conventions influenced by cartographic norms, when in reality there was a liminal space-a language borderland-where a language gradually becomes less common or changes through contact with other languages. 57rierson experimented with an alternative method of capturing the liminal area between languages on the Kachin dialects map (Fig. 8). 58igure 8 used a series of graduated dots to show that dialects gradually merge into other linguistic areas, avoiding the artificial representation of a hard boundary.The use of dots also set the map apart from the emerging use of the isogloss in European dialect mapping. 59Grierson did not replicate this technique on any other LSI map.Technical or financial reasons may have limited this method to one map, but the other 44 maps suggest that the LSI preferred to experiment with methods and techniques more commonly employed in colonial map making such as labels and explanatory notes in the legend.
Because the Kachin dialect map successfully circumvented the appearance of 'hard and fast' boundaries, it raises the question why the LSI did not consistently avoid drawing language boundaries on its maps.The Indo-Aryan (see Fig. 1) and Bengali language (see Fig. 2) maps appear to demarcate where each language or dialect is and is not spoken, as do many other maps in the LSI.Grierson argued in the Introductory volume that the LSI could not circumvent the 'conventional methods' of mapping and map making, including the drawing of boundaries.Yet the maps of the Ghalchah languages (see Fig. 5) and Kachin dialects indicate that the draftsmen at the Survey of India could make maps for the LSI that minimized, if not fully removed, the appearance of 'hard and fast' boundaries.
The depiction of boundaries on maps had little to do with technical expertise, cartographic conventions or the cost of using other methods.The LSI managed to publish maps of languages without 'hard and fast' boundaries.Grierson was also clear in other instances that political maps differed from language maps because of their emphasis on boundaries. 60Nevertheless, Grierson remained mindful that one of the main reasons why the LSI managed to receive government funding was the promise of mapping where languages and dialects were spoken for the benefit of training and governing.It is easy to understand then why boundaries appear on language maps: to clearly demarcate language areas so officials could quickly identify which languages were spoken in the vicinity of their deployment.Grierson evidently thought his maps could be useful for governing as much as for research, and he willingly offered to supply a map showing the limits of the Western Pahari dialects for the Deputy Commissioner of Simla's office. 61There is, however, little evidence that the LSI and its maps were extensively used by the government that funded it.This perhaps played into Grierson's belief the state never fully utilized the LSI, although it allowed his survey to operate in a 'grey area' where experimentation could happen.By experimenting with existing methods and new techniques with varying degrees of success, the LSI conveyed through its maps the limits of knowing precisely where languages and dialects were spoken.This paper has examined only a selection of maps, but the unevenness of scale, colour, use of relief and labels, and the depictions of boundaries are reflected across all the Linguistic Survey maps.Although languages are mapped in different ways, the LSI remained consistent on what a language map constituted: an approximate representation of an indefinite geography.The inconsistency in presentation was a consequence of trying to find the most appropriate way to map and visualize a geographically complex phenomenon, while delivering on the promise of mapping which convinced the government to fund an exercise it was reluctant to fully support.While the maps may originally have been commissioned for officials and civil servants in India, they nevertheless reached a significantly wider international audience of scholars, students and other interested parties through the distribution of the LSI to libraries and learned institutions. 62Tacitly acknowledging that language maps were likely to be interpreted by users more familiar with conventional topographical and administrative maps, Grierson included an instructive statement in the Introductory volume explaining what a language map represents.He warned his readers that language boundaries and the maps on which they feature, 'must always be understood as conventional methods of showing definitely a state of things which is in its essence indefinite', a statement perhaps foreshadowing Brian Harley's critique of the power of maps operating behind a mask of scientific neutrality. 63rierson recognized that language maps were not geographically precise-they mapped something known approximately and liable to change -yet the inclusion of that statement showed an awareness that language maps were prone to multiple interpretations and could be used to make distorted claims.Although we cannot know for certain, it is possible that the statement was added in response to one map-the Oriya language map-being mobilized in the 1910s and 1920s by a language movement in Orissa as conclusive evidence for a defined and bounded linguistic territory, exactly as Grierson warned against. 64'enquête linguistique de l'Inde comme expérience de la cartographie des langues, 1896-1927 L'enquête linguistique de l'Inde (Linguistic Survey of India, LSI), supervisée par le linguiste et fonctionnaire anglo-irlandais George A. Grierson, a recensé et classé plus de sept cents langues et dialectes.L'objectif principal de l'enquête financée par l'État était de cartographier les langues parlées en Inde, conduisant à la publication de 45 cartes linguistiques entre 1899 et 1927.À titre individuel, ces cartes sont comparables à d'autres cartes thématiques de la période.En tant que série de cartes linguistiques, cependant, elles sont incohérentes en termes d'échelle, de couleur, d'utilisation du relief et des toponymes, et dans leur représentation des limites régionales.Cet article soutient que cette incohérence de présentation reflétait l'approche expérimentale du LSI en matière de cartographie de la langue, qui cherchait à concilier la représentation approximative d'un phénomène géographique complexe avec les attentes de l'État colonial pour des cartes linguistiques précises et à jour.

Fig. 2 .
Fig. 2. 'Map Illustrating the Dialects and Sub-dialects of the Bengali Language', in George A. Grierson, ed., Linguistic Survey of India, vol.5:1, Indo-Aryan Family, Eastern Group I (Calcutta, 1903), folded, facing p. 11.Scale 1 inch = 64 miles.36 × 26 cm.Note the text in the Oriya-language area in the southwestern corner, which states 'Western Bengali also spoken here'.(Courtesy of the Digital South Asia Library, University of Chicago.)

Fig. 3 .
Fig. 3. 'Map Showing the Distribution of the Dravidian Languages of India', in George A. Grierson, ed., Linguistic Survey of India, vol.1:1, Introductory (Calcutta, 1927), folded, facing p. 81.Scale 1 inch = 160 miles.36 × 54 cm.The area where Andhra is spoken is shown in blue, the Dravida Group in yellow, and the Intermediate Group in green.Note the curious geographical isolation of Brahui (red) in what was then the Baluchistan Agency.(Courtesy of the Digital South Asia Library, University of Chicago.)

Fig. 4 .
Fig. 4. 'Map Illustrating the Area in Which the Languages of the Dardic Sub-family Are Spoken', in George A. Grierson, ed., Linguistic Survey of India, vol.8.2: [Indo-Aryan Family.North-Western Group] (Calcutta, 1919), folded, facing p. 1. Scale 1 inch = 64 miles.36 × 54 cm.The Kho-wari, or Khowari, dialects (green) are not included on the map, whereas those of the Kafir Group (blue) and Dard Group (yellow) are given.Uniquely, the map explicitly describes the boundaries marked as 'only approximate'.(Courtesy of the Digital South Asia Library, University of Chicago.)

Fig. 5 .Fig. 6 .
Fig. 5. 'Map Illustrating the Localities in Which the Ghalchah Languages Are Spoken', in George A. Grierson, ed., Linguistic Survey of India, vol. 10, Languages of the Eranian Family (Calcutta, 1921), folded, facing p. 456.Scale 1 inch = 32 miles.34 × 36 cm.The Ghalchah languages are identified in red; other languages are in blue.The map covers a region with some of the highest mountain ranges on the Indian frontier with Afghanistan and Central Asia, making its planimetric style even more incongruous.(Courtesy of the Digital South Asia Library, University of Chicago.)

Fig. 7 .
Fig. 7. 'Aryan Languages and Dialects Spoken in East Chota Nagpur', in George A. Grierson, ed., Linguistic Survey of India, vol.5:2, Indo-Aryan Languages, Eastern Group II (Calcutta, 1903), facing p. 140.Scale 1 inch = 32 miles.36 × 26 cm.The four language zones are Magahi (red), Bhojpuri (yellow), Bengali (orange) and Oriya (blue).Dots are used to show mixed linguistic areas, e.g., red dots on orange background indicate the 'area in which Bengali is main language but in which Magahi (including Kurmali) is also spoken'.(Courtesy of the Digital South Asia Library, University of Chicago.)

Fig. 8 .
Fig. 8. 'Map of the Dialects of the Kachin Group of the Tibeto-Burman Family', in George A. Grierson, ed., Linguistic Survey of India, vol.3:2, Tibeto-Burman Family II (Calcutta, 1903), facing p.499.Scale 1 inch = 32 miles.34 × 36 cm.Specific Kachin dialects are written in red text.The map uses a series of graduated dots to avoid the appearance of 'hard and fast' language boundaries.(Courtesy of the Digital South Asia Library, University of Chicago.)