Andamanese Languages:
Lexicostatistic Comparison

The purpose of the present contribution is to study several topics: A. Mutual relations of the individual Andamanese languages. B. The relationship between Onge-Jarawa and Great Andamanese. C. Chronological estimations regarding the disintegration of the various groupings' respective ancestral languages.

Table 1: South Andaman & Middle Andaman languages

Table 2: North Andaman & Little Andaman languages

1. The Andamanese classification by Manoharan (1983, 92)

Tree-diagram 1:

2. The Great Andamanese classification by Hammarström, Forkel & Haspelmath 2019 (Glottolog 4.0)

<https://glottolog.org/resource/languoid/id/grea1241>

Tree-diagram 2:

3. Andamanese classification by Ethnologue22 (2019)

<https://www.ethnologue.com/subgroups/andamanese>

Tree-diagram 3:

4. Quantitative classification by Comrie & Zamponi (2019)

Probably the first quantitative classification of the Great Andamanese languages was realized by Comrie & Zamponi (2019). Applying the 200-word-list, they have obtained the following pairwise similarity percentages (p. 42):

These results may be projected into tree-diagrams according to two strategies, first applying the partial averages of the obtained percentages, second applying the minimal percentages from every partial group.

The method, using the partial averages of percentages, leads to Tree-diagram 4a.

Tree-diagram 4a (partial averages method):

The method, using the minimal values of percentages, leads to Tree-diagram 4b:

Tree-diagram 4b (minimal values method):

5. The present study: recalibrated glottochronology

The present study aims to add to the Andamanese classification a chronological dimension, applying so-called 'recalibrated glottochronology,' which was developed by Sergei Starostin in the end of the 1980's.

Applying the method of partial averages, these figures can be projected into Tree-diagram 5a:

Tree-diagram 5a (partial averages, with chronological estimates):

If the lowest percentages are preferred, the result is slightly different: Tree diagram 5b:

Tree-diagram 5b (minimal values, with chronological estimates):

Discussion of Results

Let us compare all models of classification of the Andamanese languages introduced here. The discussed models are designated according to the numbers of the corresponding tree-diagrams, i.e. models 1, 2, 3, 4a, 4b, 5a, 5b. With the exception of models 2 and 4, which classify only the languages of the Great Andaman, all remaining models agree in the first separation of the South/Little Andaman branch, represented by Kora and Jeru, and maybe also Sentinelese. The contemporary language called Great Andaman is more or less identical with Kora (97.73% common items in the 100-word-list), followed by Jeru (95.16% in the 200-word-list).

Models 1, 2 and 3 agree on the positions of Bo vs. Cari, Jeru & Kora; according to models 1 and 3 Kede belongs together with Kol, Juwoi & Puchikwar, while model 2 classifies Kede together with Bo, Cari, Jeru, Kora. On the other hand, the newly generated model 5 connect Bo with Juwoi, Kol, Puchikwar (74.73%, 75.81%, 78.42% respectively, against 49.47%, 43.18%, 39.56%, 37.64% for Bo vs. Cari, Jeru, Great Andaman & Kora respectively), and Kede with Cari (73.40%), and further with Kora//Great Andaman & Jeru (55.88%//60.00% & 60.14%, against 51.06%, 51.06%, 46.32% for Kede vs. Kol, Juwoi, Puchikwar respectively, all on the basis of the 100-word-list). The close position of Kede to Jeru (81.36%) and Chari (78.43%) is also supported in model 4, generated on the basis of 200-word-list. The relatively high score of Bo vs. Kede, 61.17%, is significantly lower than Kede vs. Cari (73.40%; both in 100-word-list) and may be explained from the direct Bo-Kede neighborhood along the western coast, while the Kede and Cari languages were separated just by the Bo and Jeru languages.

The second diagnostic feature in all classifications is the position of the Bea-Bale cluster. Models 2, 3, 4a, 5a agree on the closer relationship of Bea-Bale with Juwoi, Kol, Puchikwar (plus Bo according to the model 5), while models 1, 4b and 5b connect first Cari-Kora-Jeru and Kol-Juwoi-Puchikwar and only their common ancester with Bea-Bale with regard to the lowest result between Bea(-Bale) and Cari-(Kora-)Jeru. For construction of tree-diagrams the method of partial averages is more adequate than the method of lowest percentages of the shared cognates. In the cases of models 1, 2, 3, it is not known which approaches to classification were applied.

On the other hand, models 4a & 5a represent in principle the same topology of the Great Andamanese tree-diagram - they differ only in the number of the languages studied. The chronological data of model 5a offer estimations of when the processes of divergence culminated. On the other hand, the chronological estimations developed in model 5b indicate the hypothetical beginnings of divergence. This means that the dates 5180 BCE and 4210 BCE are supposed to determine the probable times when the disintegration of the Andamanese protolanguage began, and when it culminated, respectively. This time interval is comparable with the dates of disintegration of several other language families, applying the same glottochronological procedure: Mon-Khmer - 5100 BCE; Sino-Tibetan - 5090 BCE; Austronesian - 4800 BCE; Tai-Kadai - 4310 BCE; Indo-European - 4340 BCE (all George Starostin 2015, 568) or 4670 BCE (Sergei Starostin, Workshop on the chronology in linguistics, Santa Fe 2004). These results give evidence that it is not necessary to divide the Andamanese languages into two independent language families. But it is also necessary to stress that the comparison of Onge-Jarawa and the Great Andaman (sub-)families is only tentative and preliminary, namely for two reasons: (a) Lexical data of both, Onge and Jarawa, are incomplete: in the Onge and Jarawa 100-word-lists there are 33 and 20 missing items respectively; (b) The regular phonetic correspondences between both of the (sub-)families are not yet established and every lexical match is determined only intuitively. There were only several attempts to formulate basic sound rules between the Great Andamanese languages, which remain symptomatically unpublished, namely Alfredo Trombetti 1922-23, 409-18 (referring to the unpublished ms. Studi di fonologia andamanese of his former student, Emilia Pilla, from 1921); Timothy Usher (ms. 2003; he kindly sent to the author a synopsis of phonetic correspondences on June 21, 2014) and Juliette Blevins (forthcoming). It is important to mention that old records from the end of the 18th cent. till the early 1960's are imperfect, only later descriptions recognize e.g. retroflexive dentals or aspirate labials.

Appendix 1: Summary of Sound Correspondences among the Languages of Great Andaman

Emilia Pilla (apud Alfredo Trombetti 1922-23, 409-18)

Lexical correspondences established by her lead to the following sound rules:

Timothy Usher: Tentative consonantal correspondences

Timothy Usher: Tentative vocalic correspondences

Blevins (forthcoming)

Note: The idea of Blevins (2007) about the Austronesian links of the Onge-Jarawa protolanguage should be carefully verified in perspective of the Great and perhaps Common Andamanese protolanguage. The Austronesian influence cannot be excluded, but probably came later, e.g. a hypothetical trajectory of the migration wave bringing Malagasy to the island of Madagascar could have included the Andamanese Islands. It is possible to speculate that disintegrations of several partial protolanguages in the period 30-300 CE or 30-150 CE according to models 5a and 5b, respectively, may reflect these events.

Appendix 2: Language Maps

Source: <http://www.andamanese.net/Grammar_Notes.html>

Source: Blevins 2007, 157

Appendix 3: The Indo-Pacific Hypothesis

The Indo-Pacific hypothesis proposes a genetic relationship among pre-Austronesian languages in the insular area between the Indian and Pacific Oceans. These include the languages of the Andaman Archipelago; the Indonesian islands Timor, Alor and Pantar; North Halmahera; all Papuan (i.e. non-Austronesian) languages of New Guinea; Tasmania; and the islands to the west of the (mostly Austronesian-speaking) Solomon Islands, as that hypothesis was proposed by Joseph H. Greenberg (1971-2005). Jonathan Morris (2008) speaks about Alfredo Trombetti as the forefather of the Indo-Pacific hypothesis, but Trombetti more accurately played the role of a godfather since he was working from the results of his two predecessors, Hans Schnorr von Carolsfeld and Riccardo Gatti. These (also cited by Morris), and especially the latter, were the true forefathers of the hypothesis.¹

Both Schnorr von Carolsfeld and Gatti concentrated on the Australian languages, searching for potential relatives outside the Australian continent. In comparanda summarized by Schnorr von Carolsfeld it is possible to identify 20 lexical parallels between Australian and Andamanese languages, with three possible Papuan cognates among them. Gatti collected as many as 170 Australian-Andamanese lexical parallels, with 21 possible Papuan cognates. Among these Australian-Andamanese parallels are also two hypothetical Tasmanian cognates, but without Papuan counterparts.

Greenberg (1971, 2005) excluded the Australian languages and focused on comparing the various Papuan languages to the Andamanese and the extinct Tasmanian languages. Altogether he collected 35 Andamanese-Papuan lexical parallels. In addition, he identified three common pronominal bases, as well as the velar suffix of the past tense, which are common to both the Andamanese and several groups of the Papuan languages.

Wurm (1975: 927-929) generally accepted Greenberg's Andamanese-Papuan comparisons, though he believed it more probable that their similarities were due to the existence of a substratum.

More critical was Pawley (2009), and for this reason his careful evaluation of Greenberg's results is instructive. From the corpus of 84 Indo-Pacific lexical comparisons collected by Greenberg, where the Andamanese data were represented in 35 items, Pawley (2009: 167) chose the 23 most promising lexical comparisons, including six hypothetical cognates in the Andamanese languages and five in the Tasmanian languages. He remains skeptical, citing other specialists in the field who hold a similar opinion (Pawley 2009, 160):

Why has the Indo-Pacific hypothesis received little attention from specialists in the relevant language groups? In the four decades since Greenberg's main publication on this subject there have been a handful of brief assessments by specialists, consisting of just a few sentences, and all have rejected the evidence as unconvincing (among these are Laycock 1975; Pawley 1998, 2005; Ross 2005).

But Pawley's categorical dismissal of the hypothesis is weakened by his omission of the work of Timothy Usher, who focuses his attention on reconstruction of the partial protolanguages of the individual language families, including the reconstruction of the protolanguage of the Great Andaman languages.² Usher offers a quite new conception of the genealogical classification of all non-Austric languages of the Indo-Pacific region, dividing them into two great super-groups, and noting their mutual interactions and relations:³

The East Papuan phylum is classified by Usher (2002: 66) as follows:

East Papuan classification (Usher 2002):

The conclusions of Timothy Usher should naturally be verified, but his proposal allows for the inclusion of a greater number of reconstructed protolanguages of individual Papuan families, and so the conditions for the use of the standard comparative method are thereby improved.

Acknowledgements

The author would like to express his sincere thanks for provision of inaccessible materials by Maarten Mous, Timothy Usher and Harald Hammarström. Great gratitude also belongs to John D. Bengtson for his corrections of my English.

Source Abbreviations