(De)classifying Arunachal Pradesh Languages:
Reconsidering the Evidence

Table of Contents

Acronyms and Conventions
1. Introduction
2. Data sources
3. Excursus on method
4. The Mö [= Mey, Shertukpen] cluster
5. Bugun and the proposed Kho-Bwa cluster
6. Puroik [=Sulung]
7. Hruso [= Aka]
8. Koro and the possibility of a Siangic phylum
9. Mijiic
10. Mishmic (~ Digarish)
11. Kman [= Miju]
12. Meyor [Zakhring]
13. Synthesis
14. Conclusions
References
Appendix: Comparative Wordlist →

Acronyms and Conventions

"The preceding remarks will have shown there is considerable difference between the North Assam dialects…The home of the North Assam tribes may be considered a kind of backwater. The eddies of the various waves of Tibeto-Burman immigration have swept over it and left their stamp on its dialects."

—Konow in Grierson (1909:572)

1. Introduction

Exactly what Sten Konow (1909) thought about the classification of the languages of 'North Assam', which largely corresponds to the modern-day state of Arunachal Pradesh, may never be clear. His account is both confused and apparently self-contradictory. However, his general conclusion was that these languages were highly diverse and showed evidence of different layers of contact with Tibeto-Burman languages spreading from the north. These perceptions have so far to make much of an impact on the world of Trans-Himalayan scholarship. The implication is that a language can effectively be classified by identifying a few words with likely Trans-Himalayan cognates. This method, while it has a certain Greenbergian charm, has problems which will be discussed at more length in §3.

¹ This is the now widely accepted replacement term for Sino-Tibetan, which was based on a cultural classification on Sinitic, rather than a linguistic analysis.

² This is now also a problematic term, since it evolved to group together all the Trans-Himalayan languages except Sinitic, and this is no longer considered a valid subgrouping. In this paper, the term is not used except in reference to other publications.

The purpose of this paper³ is to take issue with this approach through a re-examination of the problematic languages of Arunachal Pradesh. It proposes we should take seriously the underlying presumption that they are isolates. Moreover, it will suggest that even where languages probably are correctly classified as Trans-Himalayan, we can in part attribute their divergent characteristics to substrates or contact with language isolates now vanished or submerged.

³ Since the publication of Blench & Post (2014), Roger Blench has been able to travel to Arunachal Pradesh every year except during covid. The main focus of research has been Idu and Kman in the northeast of the state. Thanks to Mite Lingi, Hindu Meme and Sokhep Kri for collaboration on dictionaries and grammatical work. However, in 2023, the author was appointed Ethnographic Survey Co-ordinator for Arunachal Pradesh, which presented the opportunity to travel more widely in the state. Research is now focused on the Sajolang [= Miji] of Nafra and the Mö or Shertukpen language of Rupa. My thanks to Rijin Deru and Tshering Thongdok for their assistance in bringing teams together for elicitation work.

2. Data Sources

Much of the data available for these languages does not meet modern standards of documentation. Apart from the recension of sources in Konow, van Driem (2001), Burling (2003) and Bodt (2014, 2021) review the earlier materials. While some languages, like Aka (i.e. Hruso) early drew the attention of scholars, languages like Bugun or Meyor have remained barely known. For the Tani languages, which are certainly Trans-Himalayan, Post (2011) has circulated a modern grammar and dictionary of Galo, a Tani language, and more recently Tangam (Post 2017).

Until recently, the main sources were the 'Language Guides' published by the Research Directorate of the Arunachal Pradesh government in Itanagar, included in the references. These can be supplemented by a few related publications by the Central Institute of Indian Languages, which are in the same descriptive tradition. The function of these books is rather opaque; they are part phrase books, part ethnographic guide and part linguistic description. It is not easy to imagine why one would go to one of the most inaccessible mountainous regions of the world and offer a translation of 'the elephant is the strongest of all animals' (Simon 1976; Hill Miri).

A source for some previously unknown languages is Abraham et al. (2005, 2021) which provides the data according to a wordlist arranged for lexicostatistic coding. Fieldwork between November 2011 and April 2024 has made it possible to improve both the transcription and lexical database for some languages in Arunachal Pradesh as well as critically remapping the area where they are spoken. Despite the critical tone here, the wordlists in most sources are quite substantial and it is usually possible to isolate key morphemes and determine basic sentence structure from the grammar sketch. As a consequence, it is reasonable to say that we should have enough information to classify these languages, or possibly declassify them in the sense of excluding them provisionally from Trans-Himalayan.

3. Excursus on Method

Trans-Himalayan has a curious status as a phylum: long identified by a small set of widespread common lexical items, it has rarely been subject to attempted proof of its genetic unity (Matisoff 2003). In the languages further west such as Kiranti, many exhibit complex verbal morphology, suggesting the possibility that this was a feature of the proto-language. However, this model depends heavily on the internal structure attributed to the phylum. If the ancestors of Trans-Himalayan moved eastward, they would have gradually reduced this morphology, resulting in the monomorphemic structures in many branches. Indeed, this lack of morphology in many branches is problematic, since the similarity of some lexemes to those in other phyla, notably Daic [Tai-Kadai] and Hmong-Mien, has been responsible for a long history of discarded macrophyla proposals (for discussion of these, see van Driem 2008). Leaving aside constructs such as Sino-Austronesian and Sino-Caucasian, the membership is assumed to be broadly as characterised in Bradley (2002). Recent years have seen the publication of low-level reconstructions (e.g. Sun 1993; Mortensen 2003; VanBik 2007; Wood 2008; Button 2011; Pelkey 2011; Mortensen 2023) which is useful, but a long way from the goal of demonstrating the unity of the phylum.

The classification of Arunachal Pradesh languages in the literature is essentially an extension of this model—if a number of lexemes resemble reconstructed Trans-Himalayan, they are assigned to a particular branch. For example, the Ethnologue entry for Puroik says; "Lexical similarity: 22% with Eastern Miji, 20% with Western Miji, 20% with Bugun, 17% with Sartang." This statement is apparently based on Abraham et al. (2005), although despite close inspection it is not visible in their tables. Abraham used a wordlist of 210 items, so this represents some 42 shared items, although the extent to which any of them are actually cognate is never considered.

Not unreasonably, such a low figure has made scholars reluctant to accept their genetic affiliation. One strategy is to attribute the low figure to retention of putative substrate lexicon, sometimes euphemised as 'archaic'. Thus, van Driem (2001:530) says; "It would appear that these [= Kho-Bwa] languages have preserved archaic Trans-Himalayan vocabulary that has otherwise been lost everywhere else or innovated very differently and unrecognisably." The supposition is that these forms were once part of Common Trans-Himalayan and have been lost in other branches through lexical replacement. This is certainly a possible scenario, but it makes it difficult to understand why this type of explanation would not be applicable to many cases of apparent divergence, and would undermine what is generally presumed about the unity of Trans-Himalayan.

In recent years, an alternative model has become popular, seeing divergent lexicons as resulting from unknown substrates. Thus, Burling (2003: 181) says of Bugun; "If it has a Tibeto-Burman component at all, it seems to be very deeply submerged. It is an open question whether Bugun has a larger Tibeto-Burman component which is simply obscured by a large substrate from another language, or whether it represents some fundamentally different linguistic tradition that has borrowed a bit of Tibeto-Burman vocabulary." Even where membership of Trans-Himalayan is credible there can still be evidence for substrates of an unknown affiliation. For example, the Tani languages are usually considered to pass the test of Trans-Himalayan membership in terms of numbers of cognates and at least some regularity of correspondences. Nonetheless, they incorporate significant amounts of divergent vocabulary whose source is unknown.⁵ Indeed, in the Milang language, which is usually considered Tani on the basis of a large number of cognates, a high percentage of cognates seemingly has a substrate of a quite different character on which a Tibeto-Burman structure has been superimposed (Post and Modi 2011; also see §8).

⁵ Sun (1993:173) wrote that "beyond the most fundamental core vocabulary, the peculiarity of the Tani lexicon becomes painfully apparent, making it extremely difficult to track down reliable extra-Tani cognates of the PT roots proposed [here]. This means that exhaustively tracing the PT initial and rhyme distinctions back to plausible PTB sources is presently quite impossible."

The core data for this paper is the comparative wordlist given in the appendix table. It tabulates the lexemes for a variety of basic terms in Arunachali languages (excluding the regions bordering Myanmar) and aligns them with the most Common Tibeto-Burman (CTB) starred forms quoted from Matisoff (2003). Apparent cognates are coded in yellow, while other more local cognate sets are assigned other colours. This provides a convenient rapid visual impression of both the correspondences with commonly accepted Trans-Himalayan and the relationships between individual languages.

4. The Mö (= Mey, Shertukpen) Cluster

Mö or Shertukpen constitutes a small family of languages spoken in the valley of the Tengapani River south of Bomdila in West Kameng district. The name Shertukpen is a construct, from the settlement of Shergaon and 'Tukpen', a Monpa name for the people of Rupa town. The correct name for the Rupa is Mö and their language is Mö nyuk. A related set of lects are spoken by the Sartang, a people also called 'But Monpa' (Dondrup 2010; Bodt 2021). Two other related languages which were formerly classified under the general rubric of 'Monpa' are Duhumbi [= Chug] and Khispi [= Lish], spoken in isolated villages north of Dirang, within the Central Monpa area. Dutta (2007) includes a brief comparative wordlist of Lish and Tawang Monpa in his monograph on Central Monpa, notes its differentiation from both, but makes no comment on its possible affiliation. The Khispi live in the village of that name and in Gompatse. Fresh field data collected in November 2011 make it clear that Chug, Lish and Gompatse all are essentially the same language. The people of these settlements deny any connection with the Mey of Rupa and Shergaon. Overall, Rupa consists of three major subsets, as shown in Table 1.

The ISO codes are rather unsatisfactory. Shergaon and Rupa are sufficiently distinct as to warrant their own codes. Sartang is divided into five lects, although they are all mutually intelligible (Bodt 2021). As to Khispi and Duhumbi, they are extremely close and would elsewhere be regarded as dialects of one another.

Linguistic literature on the form of Mö spoken in Rupa town is sparse. The short description by Dondrup (1988) is based on the Shergaon dialect, while Grewal (1997) includes some sentences in the dialect of Rupa. The main source is Jacquesson (2015), which has considerable issues, due to its idiosyncratic transcription. Bodt (2014) is a literature review, and Boro (2024) has published a preliminary phonology.

The text of Abraham & Kara (2021) treats Sartang, Duhumbi (their Chug) and Khispi (their Lish) as separate languages. This is not supported by the comparative wordlist in Table 2, which shows that, allowing for variations in transcription, Khispi and Duhumbi are hardly even dialects of one another. Bodt (2020) is a grammar of Duhumbi, written according to modern linguistic norms. Surprisingly, Rupa is quite distinct from the language of Shergaon. The Sartang forms given below are based on newly transcribed field data.⁶ Where the Mö cluster item resembles reconstructed CTB, the line is shaded.

⁶ Roger Blench would like to thank the Gaonbura of Rahung for recording a wordlist of Sartang on January 18th, 2011.

Table 2 shows that Shergaon, Rupa, Sartang, Khispi and Duhumbi form a dialect complex and that resemblances to reconstructed CTB are sporadic.

5. Bugun and the Proposed Kho-Bwa Cluster

The Bugun language [bgg; glottocode bugu1246] is spoken in West Kameng district of Arunachal Pradesh. The Bugun, also known as Khowa, live in some ten villages, were estimated to number 800 in 1981, but current estimates put them at around 1700 speakers.⁷ The Bugun language has been barely documented. The only published source is the orthographic Dondrup (1990) which should be used with care; some phonetically transcribed data appears in the Appendix to Abraham et al. (2005) and Madhumita Barbora of Tezpur University has recorded a wordlist and sample sentences as part of an unpublished study of the phonology. Lander-Portnoy (2013) is a thesis based on recorded material. Data for this paper was recorded from Martin Glo, secretary of the Bugun Welfare Society, who is a native of Chittu village, in Tenga in January 2011.

⁷ Glottolog lists no less than six lects of Bugun, which seems unlikely. These are village names.

Pandey (1996) is part descriptive ethnography, part hagiography, and again should be used with care. Despite being a small ethnolinguistic group, the Bugun are quite active in promoting their culture with an active Bugun Youth Association. Bugun may be the only language in this region to have contributed a loanword into English. The Bugun liocichla (Liocichla bugunorum) is an endemic bird species first described in 2006. Vanessa Cholez (pers. comm.) has completed a dissertation (2024) on the sociology of the Bugun, but it is currently not in circulation.

Inasmuch as Bugun is mentioned at all, it is assumed to be Trans-Himalayan (e.g. Ethnologue 2024). van Driem (2001:473) originally referred to unpublished and unavailable work by Roland Ruttger relating Bugun to the Shertukpen cluster suggesting this and names the resultant grouping 'Kho-Bwa'. This has been enthusiastically promoted by Bodt (2019, 2020, 2022) and even used in experiments to predict relatedness using algorithms (Bodt & List 2019; Wu et al. 2020). However, the existence of this construct is far from proven, and I suggest here it may be a chimaera. Bugun people are often able to speak Mö as a language of intercommunication and the similarities between the two may simply be borrowings. Table 3 suggests neither a regular relationship between Bugun and Mö, nor a strong resemblance to reconstructed CTB. Some words show relations of near-identity, for example 'head', 'water' and 'leaf'. /ʧ/ is conserved in 'eat' and 'liver' but /ʃ/ corresponds to /ʧ/ in 'mouth'.

On the broader question of whether Bugun and the Mey cluster are Trans-Himalayan, neither language shows many cognates with CTB and some of those are doubtful or possible loans, such as 'pig' and 'iron'. The low number of Tibeto-Burman cognates could just as easily be explained by borrowings as by genetic affiliation. Table 3 shows some of the typical resemblances and a more in-depth search would be likely to uncover great numbers. I have marked the CTB form in the second column; it shows that only a very few forms are shared with the Bugun-Mey pair.

6. Puroik [= Sulung]

The Puroik language [suv: glottocodes sulu1241 and west2872] is spoken by a few thousand people in East Kameng and Lower Subansiri districts in Arunachal Pradesh and adjacent parts of Tibet. Previously known as 'Sulung', this name has now been rejected by the community as pejorative. The ethnography of the Puroik is described in Stonor (1952) and Deuri (1982). Their basic subsistence system is hunting and gathering with a significant dependence on the sago palm, Metroxylon, rather as in Melanesia. All forms of agriculture appear to be recent innovations. The Puroik were formerly in a serf-like relationship with the Tani-speaking Nyishi, for whom they collect cane and labour on farms. Puroik were still being officially liberated as late as 2001 (see appended documents in Remsangphuia 2008:102-102).

The most well-described of the three Puroik varieties is the dialect of Chayangtajo circle, East Kameng, where Sanchu is the biggest and best accessible Puroik village. This variety is described in the major published sources, Tayeng (1990), Li (2004), Remsangphuia (2008) and Soja (2009). The Chinese work is summarised in Matisoff (2009) and reviewed in Jackson (2003). Matisoff (2009) has an appendix to a paper on the persistence of Tibeto-Burman roots, compares Puroik materials from Li (2004) with his CTB roots and claims numerous cognates. Many of these require the eye of faith but it is notable that there are more resemblances than are evident in southern forms which often have quite different lexemes. Since the Tibetan Puroik apparently also speak Tibetan as a second language, this should make us suspicious at the least. Work by Lieberherr (2017), assuming it is published, is likely to give us a much improved description of Puroik. The dialect of Kojo-Rojo is spoken in two, possibly three villages (Kojo, Rojo, Jarkam), and is different but mutually intelligible with the dialect of other villages in Lada circle. The third dialect is Bulu, west of Kojo-Rojo. Map 1 (from Lieberherr 2015) shows the locations of these dialects as well as neighbouring languages mentioned in this paper.

Although listed both as Kho-Bwa and 'possibly Austroasiatic' in earlier versions of the Ethnologue (e.g. 2013), the arguments for this are elusive. Concerning the classification of Puroik, a footnote to Sun (1993: fn. 14) says; 'Sulung is a newly discovered distinct Tibeto-Burman language showing remarkable similarities to Bugun, another obscure Tibeto-Burman language spoken to the west of the Sulung country.' This is a considerable exaggeration, and later, reviewing the Chinese source, Sun (1992) assumes that Puroik is Trans-Himalayan, he is pessimistic about finding the evidence for cognates. The most detailed examination of these similarities, as well as a discussion of the dialect situation of Puroik is Lieberherr (2015). This paper takes on the challenge presented in Blench & Post (2014) to show that the apparent cognates with Trans-Himalayan are in fact evidence of genetic affiliation and not simply borrowings. His method is slightly idiosyncratic, since he compares Puroik with reconstructed Kuki-Chin (VanBik 2009) rather than CTB. Kuki-Chin is certainly an authenticated branch of Trans-Himalayan, and Lieberherr's arguments are coherent. As a consequence, I regard the argument for a Trans-Himalayan affiliation as generally convincing and my prior scepticism as refuted.

A separate question is whether Puroik, Bugun and the Mö cluster form a linguistic group. If so, they would then all be Trans-Himalayan. Apart from the numerals and some body parts, the general form of the vocabulary is highly divergent and suggests the Kho-Bwa construct is not meaningful. Table 4 shows a preliminary table of lexical similarities, including Mö cognates (shaded):

7. Hruso [= Aka]

Hruso [hru: glottocode hrus1242] is also called Aka, although this name means 'painted', presumably a reference to facial tattooing which was formerly practised by these people. This name is still used in official publications such as Ethnologue, and Hruso is the self-name which has begun to come into use within the community. Aka is sometimes confused with Koro, which, although also called 'Aka' by the local Miji people, has now been assigned a distinct glottocode (koro1316). Hruso has been the subject of a considerable literature. A reference to the Hruso people dates to Robinson (1855). The first linguistic information is due to Needham (1886), reprised in Campbell (1874) and in Konow (1909). These sources contain substantial wordlists, although in some cases they are remarkably transcribed, while in Konow, the transcription of other languages in his survey is of a high standard. The first substantial monograph is due to Schubert (1964) and a second version of this is reprinted in Shafer (1966/1967:245-277). Modern sources include the orthographic grammar by Simon (1993). More recently, D'Souza (2015, 2018, 2021) is a detailed phonology of Hruso. The author has recorded Hruso from Sera Saring in Serba on several occasions since 2011.⁸

⁸ Roger Blench would like to thank Serwa for being an enthusiastic and patient informant for Hruso.

Shafer (1947) is mainly devoted to a discussion of the classification of Hruso. He considered Hruso was Trans-Himalayan on no very good grounds, both because of the poor quality of trancriptions and because his discussion conflates Hruso proper with Miji. Cognates with Trans-Himalayan languages are very few and involve sometimes highly ad hoc rules. Koro Aka is quite unrelated to either of these, as the appendix table shows; its affiliation is dealt with in §8. Of all the languages considered here, Hruso has the fewest roots that can plausibly be related to Trans-Himalayan. Shafer's arguments are fairly weak, and it is more credible to treat these as regional borrowing than evidence for genetic affiliation.

8. Koro and the Possibility of a Siangic Phylum

The references to two Aka languages has been a source of confusion since the earliest period and indeed there are two languages under this label, Hruso proper and Koro. The Koro [jkr: glottocode koro1316] people live in East Kameng district, in Bana and surrounding hamlets, east of Thrizhino on the Seppa road. Although claims were made for its 'discovery' in 2010 (Anderson & Murmu 2010), presumably as a consequence of financing by National Geographic, a grammar sketch of this language appears in Grewal (1997) and lexical data can be extracted from Abraham et al. (2005/2021). Further lexical data was collected by the author in December 2011.⁹

⁹ Roger Blench would like to thank the headman, Somo Yamde, for taking time to record a sample of Koro in Yangse village in November 2011.

A brief comparison with Hruso quickly shows that the two have virtually nothing in common, as was also stated by Anderson & Murmu (2010). However, strikingly, Koro does share a number of lexemes with Milang, a language far to the east in Siang district usually identified as Tani (Sun 1993: §3). Milang is spoken in three dispersed villages in East Siang district (Modi 2008). With the exception of Tayeng (1976) almost nothing has appeared in print on this language. Sun noted its divergent character, but treated it as an early branching of Tani. Milang is characterised by both divergent lexicon and highly irregular correspondences with the rest of Tani (Modi 2008, Post and Modi 2011). The hypothesis here is that Milang was a non-Tani language that came under heavy and repeated Tani influence. Milang is spoken a considerable distance from Koro, so shared lexicon is unlikely to be the result of contact. The proposal, set out in more detail in Post & Blench (2011), suggests there was once a chain of languages, tentatively named Siangic, stretching between West Kameng and the Siang river, whose presence can be detected both in Koro, in the substrate lexicon of Milang and in irregularities in other Tani languages now spoken in the intervening area (Table 5).

¹⁰ Seemingly only when negated in Koro.

Figure 1 represents the configuration of the proposed Siangic phylum. If this argument is correct, then Siangic is a small phylum which is distinct from Trans-Himalayan. Milang underwent heavy cultural influence from Tani (Adi and Padam in particular), making it appear a divergent Tani language, but underlyingly it is unrelated.

9. Mijiic

The Miji language (also Sajolang¹¹, Dimai [= Dhimmai]) should not be confused with Dhimal in Nepal. Simon (n.d.) reports that it is spoken in some thirteen villages around Nafra in West Kameng district and that the population was 3,549 in the 1971 census. Fresh fieldwork was undertaken among the Miji of Nafra in February and March 2024¹² which revealed serious errors of elicitation in previous work, which mixed forms from two lects. This is now corrected in more recently circulated documents.

¹¹ The name Sajolang has been widely adopted in Nafra in reference to Western Miji, but is not in use among the Eastern Miji in Lada circle.

¹² Thanks to Rijin Deru who both acted as an informant and arranged to drive me to villages to record shrines and other ritual sites.

Miji is divided into two lects, representing Nafra and Lada circles, effectively west and east. Weedall (2021) has a more complex division of western Miji into subdialects. Whether these would be naturally mutually comprehensible is hard to determine, since almost all speakers seem to know the equivalent words in the other dialect.

Until recently, the existence of a language known as Bangru [no ISO glottocode, bang1369] was only rumoured. However an ethnographic thesis and subsequent report confirms that this is a genuine ethnic group, but also that it is a language related to Miji (Ramya 2011, 2012). Ramya's transcriptions are orthographic, but the underlying forms are easily seen when compared to Miji. Blench (2015) circulated comparative data on Miji and Bangru in support of this assumption. Bodt & Lieberherr (2015) have published a wordlist of Bangru based on the CALMSEA list, with analysis and comparisons with Miji and Hruso. While this represents a major advance on Ramya, much more remains to be done on Bangru. Separately, I compiled a wordlist with the aid of Mrs. Chera Mekia Mili and family, now resident in Itanagar. Mrs. Mili grew up in a monoglot household, but later married a Nyishi husband and came to speak the language as the main language of her home.

Andrew Hsiu kindly forwarded to me Li (2003) which is a phonology and wordlist of the Bangru spoken in Tibet which was unknown to all previous researchers. Li includes phonemes that seem to have disappeared from the Bangru of India, and also envisages a more elaborate tonal system. This may represent a more conservative form of the language and Bangru in India has undergone heavy influence from the dominant Nyishi language.

Bangru is undoubtedly related to Miji, as the numerals in the appendix table show, although it has quite a divergent lexicon. In some cases, this is due to borrowing from Nah and Nyishi, both Tani languages. Figure 2 shows a tree of the relationships of Mijiic languages.

There is a persistent idea in the literature that Miji is related to its eastern neighbour Hruso and that there exists a subgrouping 'Hrusish'. More surprising are the cases where Bangru shows similarities to Hruso. This idea seems to derive from Shafer (1947) but is perpetuated in Bodt (2014) and Bodt & Lieberherr (2015). The evidence for Hruso and Miji having a distinctive relationship seems to be based on a very few similarities, for example the words for 'sun', 'eight' and 'nine', which appear to be exclusively shared. The great majority of basic vocabulary appears to be quite different. Miji, Hruso and Koro share a great deal culturally, but show virtually no linguistic resemblances, except for sporadic loanwords, as might be expected between any two neighbouring languages. This can be clearly seen in the appendix table. The conclusion must be that the purported Hrusish group is spurious, and Hruso is a language isolate.

Miji has long been classified with the 'North Assam' languages and this is generally repeated in subsequent publications (e.g. Bradley 1997). However, there is no data in Konow (1909) and Shafer (1947) may be the first citation of linguistic material. As is now a rather familiar pattern, resemblances to Trans-Himalayan are confined to a few lower numerals and some basic body parts. Otherwise, few Miji lexemes resemble reconstructed Trans-Himalayan forms. I therefore argue that the Mijiic languages constitute a small independent phylum.

10. Mishmic (~ Digarish)

It has been proposed that there is a Mishmic group consisting of two related languages, Idu (Luoba in Chinese sources) and Tawrã (= Taraon, Digaru, Daruang in Chinese sources). This group has sometimes been known as Digarish, or alternatively grouped together with Kman (= Miju) as Midźuish, a denomination which may go back to Shafer (1955). It is certainly the case that the Idu and the Tawrã refer to themselves in local English as 'Mishmi'. In modern times, however, it is the Tawrã and the Kman who regard themselves as a single culture, despite the fact that their languages seem to have almost nothing in common. There is a plethora of local publications which compare phrases in both languages. Quite how this cultural convergence came about is unclear, but see speculations in Blench (2024). Kman is treated in the following section as a distinct language whose classification remains unclear.

The Idu [clk: glottocode idum1241] are also known as Chulikata [= Chulikotta, Sulikota], Midu [= Ida, Midhi], Yidu Luoba, Lhoba [Chinese terms]. It is unfortunate that the ISO code is based on the pejorative term Chulikata, now discouraged. The earliest reference to the Idu language is in Brown (1837) and some material can be found in Campbell (1874) and Konow (1909). Pulu (2002) is a brief orthographic introduction to Idu in the characteristic style of Arunachal Pradesh Government publications. Idu has also been described from the Chinese side [under the name Lhoba], notably in Sun et al. (1980), Sun (1983a,b, 1999) and Ouyang (1985).

The Tawrã [mhu; glottocode: diga1241] are also known as Darang 达让僜, Daruang, Deng, Digaro, Digaru, Mishmi, Taaon, Taraon and Taying. The name 'Digaru' (vernacular name of a major river) is often used in English conversation. Records of Tawrã go back at least to Robinson (1856). Needham (1886) gives a comparative wordlist of Tawrã, Kman and Tibetan. Recensions of existing data are given in Campbell (1874) and Konow (1909). Modern publications with a 'practical' orientation include Chakravarty (1978) and Pulu (1991). Chinese scholars have also worked on Tawrã, which they call 'Daruang'. The Tawrã language has been briefly described in Sun (1999) and Jiang et al. (2013) is an extended grammar of Tawrã in Chinese. Evans & Manyu (2021) is a phonology of Tawrã in India prepared for the purposes of Bible translation, so its reliability is hard to gauge.

Whether Idu and Tawrã actually form the genetic unit claimed in the literature is questionable. They clearly share a significant amount of lexicon in some semantic fields. For example, Table 6 shows the lower numerals, which suggest a strong relationship.

By way of contrast, Table 7 compares Idu and Tawrã terms for body parts, few of which have anything in common.

Given this divergence by semantic field, making any definitive statement about the relatedness of Idu and Tawrã is problematic. Blench (2024) argues that the two languages are ultimately unrelated, and that there has been intense bilingualism at an unknown period in the past, which resulted in the convergence of the numerals. The grammar of Tawrã is poorly known, so it is difficult to compare it with Idu, which is relatively well-described. Perhaps further in-depth studies will clarify the situation.

Whether Idu and Tawrã are Trans-Himalayan remains doubtful. The appendix data table shows that there are few evident cognates with reconstructed CTB. The pattern is much the same as noted for other languages, i.e. lower numerals, some basic body parts, sun, moon etc. I submit this is inadequate to accept as evidence for membership of Trans-Himalayan.

11. Kman [= Miju]

The Kman people [mxj: miju1243] live in villages around Tezu in the extreme northeast of Arunachal Pradesh.¹³ Alternative names for the Kman include Eastern Mishmi, Geman Deng, Kaman, Miju. The first record of Kman appears to be Robinson (1856) which is quite accurate for the period, and his transcriptions are recognisable today. Orthographic publications on the Indian side are Das Gupta (1977) and Boro (1978) which are said to be 'practical' although the transcription of Kman is highly inaccurate by modern standards. Despite the small number of speakers on the Chinese side of the border, there have been several publications on 'Geman', the Chinese version of the name. These include Sun (1991, 1999) and most importantly, Jiang et al. (2013) which is a full-length description of the language. Kman has undergone an intriguing development in terms of its orthography; a local system of writing used for communication on Facebook has developed which is also used in a children's book (Dai et al. 2013). Separately, lexical guides have been published covering both Kman and the neighbouring Tawrã [Taraon] language (Krisikro 2006; Tawsik 2014) although the orthography bears no resemblance to any other publication. The Kman people have been the subject of an anthropological monograph (Dutta 2012).

¹³ Fieldwork among the Kman began in 2015 and has continued until 2024 in collaboration with Sokhep Kri. A preliminary Android dictionary has been released in the community.

The evidence for the affiliation of Kman to Trans-Himalayan is more convincing than for Idu and Tawrã. The appendix table shows many more clear cognates with reconstructed CTB, suggesting a closer affiliation. Another aspect of Kman which shows analogies with more established Trans-Himalayan languages is in the morphology of the verb, verb stem alternation which has numerous parallels in regional languages. In the case of Kman, number marking is indexed to the head through nasal alternation in the stem. This process is only applied to a subset of verbs and is thus not predictable. The output is also atypical, since the result is verbs with final N+C sequences which do not occur elsewhere in the phonology. In Kman grammar, these are verbal nouns or gerunds, since they primary occur with action verbs. A sample of these gerunds is given in Table 8.

This type of alternation, which is sporadic and unpredictable, can be compared to verb stem alternation in other Trans-Himalayan languages, for example Tangsa (Morey 2018) and Kuki-Chin (Bedell et al. 2023). This type of morphosyntax is far less likely to be borrowed than the lexicon.

12. Meyor [Zakhring]

The Meyor language [zkr: glottocode zakh1243], also known as Zakhring, is spoken in Lohit District, Walong and Kibithoo circles, Arunachal Pradesh. In 2001 there were some 376 speakers scattered in fifteen hamlets. On the Arunachal Pradesh side, the major published source on the language is Landi (2005) although Jacquesson (2001) includes some data on pronouns. Lĭ & Jiang Di 李大勤，江荻 (2001) is a brief overview of the 'Zha' language. Sun (1999) has comparative tables of language he calls 'Dza' which he relates to 'Geman' [Kman] and assumes it is a 'mixture' with Tibetan. Geman (together with Idu-Tawrã) is said to be Kachinic, although no evidence is presented for this. Both Ethnologue and Glottolog put Kman and Meyor together as a subgroup of Trans-Himalayan, but the evidence for this is thin.

Landi (2005: 164 ff.) notes the similarities to Kman, although he conflates genuine shared cognates due to common CTB inheritance with borrowings. Nonetheless, some useful observations can be extracted from his tables. Table 9 shows a sample of lexical items where Kman and Meyor share a common root. The Dza data shows more cognates with Kman than the material in Landi (2005).

These items are suspiciously similar, whereas Kman and Meyor otherwise have many completely divergent roots, suggesting borrowing. Considerably more Meyor basic lexicon is related to the Brokpa languages, such as Memba and the Senge cluster (represented in the appendix table by Tawang and Memba dialect). These languages have a relatively high proportion of Tibeto-Burman roots, preserved in a constellation very close to the hypothetical proto-form.

Landi also compares Meyor to Turung (Singpho), a Jingpho language spoken in this region, but his comparisons are all either only doubtfully cognate or are CTB and thus not relevant as evidence. Scott Delancey (p.c.) has also presented evidence for some striking grammatical similarities with Kuki-Chin type languages, in the area of pronominal indexing. Yet an examination of the lexicon using the data posted in STEDT did not produce a single example of a specific link to the Naga/Kuki-Chin languages. Meyor must definitely be left unclassified at present and indeed represents a broad problem for the usual procedures of historical linguistics.

13. Synthesis

The emerging synthesis is quite strikingly at variance with the received and published classifications. Arunachal Pradesh languages are extremely diverse, not only in relation to common Trans-Himalayan but also to one another. As we learn more about their grammar, it is clear that although they are synchronically isolating languages with eroded morphology, they reveal remarkable and isolated traits, quite unlike their neighbours. Idu, for example, has a complex verbal system with lengthy stacked extensions, similar to Bantu languages of Central Africa.

The usual explanation for a low incidence of CTB vocabulary is the gradual loss of lexemes over time. What is striking here is that the presumably innovative forms—the diachronically secondary forms, according to the received view—are both (a) far greater in number than the attested CTB forms and (b) not (or not obviously) relatable to any other known language. The implications of this linguistic model for proto-historical reconstruction are extreme, and should be made plain: we are asked to believe that individual Tibeto-Burman language groups repeatedly encountered populations which so overwhelmed them that they adopted forms from these mystery languages on such a scale that the overwhelming majority of their lexicons were wholly replaced, and that these mystery languages subsequently died out, leaving only the previously marginal genetically Tibeto-Burman languages to reflect their past existence in the form of an overwhelmingly massive substrate. Why precisely this model is more persuasive than one in which it is rather a suite of non-Tibeto-Burman languages which, coming into contact with different Tibeto-Burman languages at various points in their history, adopted a handful of Tibeto-Burman forms, remains to be demonstrated. The model adopted here is more in tune with modern contact linguistics, assuming borrowing unless inheritance is demonstrated.

In the light of this, Table 10 (next page) synthesises the new proposals presented here, omitting a detailed listing of Tibetic, Tani and Naga languages.

This represents a fairly radical departure from the conventional view of these languages. In another way, this is far from surprising. Arunachal Pradesh is highly dissected, remote and inaccessible and was bypassed by major East-West trade routes. That language isolates should have persisted here long after they were assimilated elsewhere in SE Asia is quite credible. The challenge for the future will either be to build on these hypotheses or disprove them on the basis of improved evidence.

14. Conclusions

The impetus behind this paper is the re-examination of the evidence for a Trans-Himalayan affiliation proposed for the languages of Arunachal Pradesh, in the light of the practice of repeating the work of previous scholars without an evaluation of the actual data. The conclusion is that a number of languages or clusters could well be isolates, and that the Trans-Himalayan roots they do evince may well be borrowings. Obviously, each topic requires a full-length paper, and these will be undertaken in due course, especially as better-transcribed data becomes available. Meanwhile, provisionally we may well consider Konow's summary quoted in the epigraph to be a useful image. Arunachal Pradesh consists of a chain of isolated languages, which have been on the southern edge of the core Trans-Himalayan area. A plethora of different contact situations have allowed both lexical borrowing and sometimes striking grammatical and phonological restructuring. But perhaps it would be useful to begin considering this region as more similar to the Amazon or Northeast Asia than Tibet.

In view of this, the languages of Arunachal Pradesh should be treated as a major priority on a global scale. Languages such as Basque and Burushaski have attracted high levels of scholarly interest over many decades precisely because of their status as language isolates. Those in Arunachal Pradesh have been completely bypassed. Moreover, although these languages are presently still spoken, their populations are small and pressure to switch to Hindi, promoted in both the media and via the school system, is growing. Probably by no coincidence, Arunachal Pradesh is also a major centre for biodiversity, something which attracts worldwide attention and resources. It is suggested that the little-known languages of Arunachal Pradesh should be given similar priority due to their uniqueness and endangered status.

Appendix: Comparative Wordlist

Comparative Wordlist for Arunachal Pradesh Languages

The table below presents lexical data across 16 Arunachal Pradesh language varieties: Koro, Mö, Bugun, Puroik, Hruso, Miji, Milang, Tawrã, Idu, Kman, and several Tibetic varieties (Dirang, Tawang, Memba, Meyor, Nah). Color coding indicates proposed cognate sets. CTB = Comparative Tibeto-Burman reconstruction.

Legend: Yellow = Mö cluster cognates; Orange = Bugun-Puroik connections; Blue = Tawrã-Idu-Kman; Cyan = Milang connections; Pink = Tibetic loanwords