|New Guinea, Nusa Tenggara (Alor, Pantar, Timor), Maluku Islands|
|Linguistic classification||One of the world's primary language families|
The extent of various proposals for Trans-New Guinea.
Families accepted by Usher
Other families proposed by Ross (2005)
Other Papuan languages
Trans-New Guinea (TNG) is an extensive family of Papuan languages spoken in New Guinea and neighboring islands, perhaps the third-largest language family in the world by number of languages. The core of the family is considered to be established, but its boundaries and overall membership are uncertain. The languages are spoken by around 3 million people. There have been three main proposals as to its internal classification.
Although Papuan languages for the most part are poorly documented, several of the branches of Trans-New Guinea have been recognized for some time. The Eleman languages were first proposed by S. Ray in 1907, parts of Marind were recognized by Ray and JHP Murray in 1918, and the Rai Coast languages in 1919, again by Ray.
The precursor of the Trans-New Guinea family was Stephen Wurm's 1960 proposal of an East New Guinea Highlands family. Although broken up by Malcolm Ross in 2005, it united different branches of what became TNG for the first time, linking Engan, Chimbu-Wahgi, Goroka, and Kainantu. (Duna and Kalam were added in 1971.) Then in 1970 Clemens Voorhoeve and Kenneth McElhanon noted 91 lexical resemblances between the Central and South New Guinea (CSNG) and Finisterre-Huon families, which they had respectively established a few years earlier. Although they did not work out regular sound correspondences, and so could not distinguish between cognates due to genealogical relationship, cognates due to borrowing, and chance resemblances, their research was taken seriously. They chose the name Trans-New Guinea because this new family was the first to span New Guinea, from the Bomberai Peninsula of western West Irian to the Huon Peninsula of eastern PNG. They also noted possible cognates in other families Wurm would later add to TNG: Wurm's East New Guinea Highlands, Binandere in the 'Bird's Tail' of PNG, and two families that John Z'graggen would later (1971, 1975) unite in his 100-language Madang-Adelbert Range family.
In 1975 Wurm accepted Voorhoeve and McElhanon's suspicions about further connections, as well as Z'graggen's work, and postulated additional links to, among others, the languages of the island of Timor to the west of New Guinea, Angan, Goilalan, Koiarian, Dagan, Eleman, Wissel Lakes, the erstwhile Dani-Kwerba family, and the erstwhile Trans-Fly-Bulaka River family (which he had established in 1970), expanding TNG into an enormous language phylum that covered most of the island of New Guinea, as well as Timor and neighboring islands, and included over 500 languages spoken by some 2 300 000 people. However, part of the evidence for this was typological, and Wurm stated that he did not expect it to stand up well to scrutiny. Although he based the phylum on characteristic personal pronouns, several of the branches had no pronouns in common with the rest of the family, or even had pronouns related to non-TNG families, but were included because they were grammatically similar to TNG. Other families that had typical TNG pronouns were excluded because they did not resemble other TNG families in their grammatical structure.
Because grammatical typology is readily borrowed--many of the Austronesian languages in New Guinea have grammatical structures similar to their Papuan neighbors, for example, and conversely many Papuan languages resemble typical Austronesian languages typologically--other linguists were skeptical. William A. Foley rejected Wurm's and even some of Voorhoeve's results, and broke much of TNG into its constituent parts: several dozen small but clearly valid families, plus a number of apparent isolates.
In 2005 Malcolm Ross published a draft proposal re-evaluating Trans-New Guinea, and found what he believed to be overwhelming evidence for a reduced version of the phylum, based solely on lexical resemblances, which retained as much as 85% of Wurm's hypothesis, though some of it tentatively.
The strongest lexical evidence for any language family is shared morphological paradigms, especially highly irregular or suppletive paradigms with bound morphology, because these are extremely resistant to borrowing. For example, if the only recorded German words were gut "good" and besser "better", that alone would be enough to demonstrate that in all probability German was related to English. However, because of the great morphological complexity of many Papuan languages, and the poor state of documentation of nearly all, in New Guinea this approach is essentially restricted to comparing pronouns. Ross reconstructed pronouns sets for Foley's basic families and compared these reconstructions, rather than using a direct mass comparison of all Papuan languages; attempted to then reconstruct the ancestral pronouns of the proto-Trans-New Guinea language, such as *ni "we", *?gi "you", *i "they"; and then compared poorly supported branches directly to this reconstruction. Families required two apparent cognates to be included. However, if any language in a family was a match, the family was considered a match, greatly increasing the likelihood of coincidental resemblances, and because the plural forms are related to the singular forms, a match of 1sg and 1pl, although satisfying Ross's requirement of two matches, is not actually two independent matches, again increasing the likelihood of spurious matches. In addition, Ross counted forms like *a as a match to 2sg *ga, so that /?V, kV, V, V/ all counted as matches to *ga. And although /n/ and /?/ occur in Papuan pronouns at twice the level expected by their occurrence in pronouns elsewhere in the world, they do not correlate with each other as they would if they reflected a language family. That is, it is argued that Ross's pronouns do not support the validity of Trans-New Guinea, and do not reveal which families might belong to it.
Ross also included in his proposal several better-attested families for non-pronominal evidence, despite a lack of pronouns common to other branches of TNG, and he suggested that there may be other families that would have been included if they had been better attested. Several additional families are only tentatively linked to TNG. Because the boundaries of Ross's proposal are based primarily on a single parameter, the pronouns, all internal structure remains tentative.
Most TNG languages are spoken by only a few thousand people, with only seven (Melpa, Kuman, Enga, Huli, Western Dani, Makasae, and Ekari) being spoken by more than 100,000. The most populous language outside of mainland New Guinea is Makasae of East Timor, with 100,000 speakers throughout the eastern part of the country. Enga is the most populous Trans-New Guinea language spoken in New Guinea, with more than 200,000 speakers. Golin, Sinasina, Mid Grand Valley Dani, Kamano, and Bunaq have between 50,000-100,000 speakers (Galela of Halmahera, usually not classified as Trans-New Guinea, also has between 50,000-100,000 speakers.) All other Trans-New Guinea languages have fewer than 50,000 speakers.
The greatest linguistic diversity in Ross's Trans-New Guinea proposal, and therefore perhaps the location of the proto-Trans-New Guinea homeland, is in the interior highlands of Papua New Guinea, in the central-to-eastern New Guinea cordillera where Wurm first posited his East New Guinea Highlands family. Indonesian Papua and the Papuan Peninsula of Papua New Guinea (the "bird's tail") have fewer and more widely extended branches of TNG, and were therefore likely settled by TNG speakers after the proto-language broke up.
Ross speculates that the TNG family may have spread with the high population densities that resulted from the domestication of taro, settling quickly in the highland valleys along the length of the cordillera but spreading much more slowly into the malarial lowlands, and not at all into areas such as the Sepik River valley where the people already had yam agriculture, which thus supported high population densities. Ross suggests that TNG may have arrived at its western limit, the islands near Timor, perhaps four to 4.5 thousand years ago, before the expansion of Austronesian into this area. Roger Blench associates the spread of Trans-New Guinea languages with the domestication of the banana.
The classification here follows Wurm, and includes some later modifications to his 1975 proposal. Wurm identifies the subdivisions of his Papuan classification as families (on the order of relatedness of the Germanic languages), stocks (on the order of the Indo-European languages), and phyla (on the order of the Nostratic hypothesis). Trans-New Guinea is a phylum in this terminology. A language that is not related to any other at a family level or below is called an isolate in this scheme.
Trans-New Guinea phylum (Wurm 1975)
('Family-level' groups are listed in boldface)
As of 2003, William A. Foley accepted the core of TNG: "The fact, for example, that a great swath of languages in New Guinea from the Huon Peninsula to the highlands of Irian Jaya mark the object of a transitive verb with a set of verbal prefixes, a first person singular in /n/ and second person singular in a velar stop, is overwhelming evidence that these languages are all genetically related; the likelihood of such a system being borrowed vanishingly small." He considered the relationship between the Finisterre-Huon, Eastern Highlands (Kainantu-Gorokan), and Irian Highlands (Dani - Paniai Lakes) families (and presumably some other smaller ones) to be established, and said that it is "highly likely" that the Madang family belongs as well. He considered it possible but not yet demonstrated that the Enga, Chimbu, Binandere, Angan, Ok, Awyu, Asmat (perhaps closest to Ok and Awyu), Mek, Sentani and the seven small language families of the tail of Papua New Guinea (Koiarian, Goilalan, etc., which he maintains have not been shown to be closely related to each other) may belong to TNG as well.
Ross does not use specialized terms for different levels of classification as Donald Laycock and Stephen Wurm did. In the list given here, the uncontroversial families that are accepted by Foley and other Papuanists and that are the building blocks of Ross's TNG are printed in boldface. Language isolates are printed in italics.
Ross removed about 100 languages from Wurm's proposal, and only tentatively retained a few dozen more, but in one instance he added a language, the isolate Porome.
Ross did not have sufficient evidence to classify all Papuan groups. In addition, the classification is based on a single feature - shared pronouns, especially 1sg and 2sg - and thus is subject to false positives as well as to missing branches that have undergone significant sound changes, since he does not have the data to establish regular sound correspondences.
Although Ross based his classification on pronoun systems, many languages in New Guinea are too poorly documented for even this to work. Thus there are several isolates that were placed in TNG by Wurm but that cannot be addressed by Ross's classification. A few of them (Komyandaret, Samarokena, and maybe Kenati) have since been assigned to existing branches (or ex-branches) of TNG, whereas others (Massep, Momuna) continue to defy classification.
Ross removed 95 languages from TNG. These are small families with no pronouns in common with TNG languages, but that are typologically similar, perhaps due to long periods of contact with TNG languages.
Trans-New Guinea phylum (Ross 2005)
Trans-New Guinea phylum (Pawley and Hammarström 2018)
Groups and isolates considered by Pawley and Hammarström (2018) as having weaker or disputed claims to membership in Trans-New Guinea (some of which they suggest may ultimately turn out to be Trans-New Guinea, but further evidence is needed):
Groups and isolates sometimes classified as Trans-New Guinea, but rejected by Pawley and Hammarström (2018) as Trans-New Guinea:
Timothy Usher has reconstructed lowel-level constituents of Trans-New Guinea to verify, through the establishment of regular sound changes, which purported members truly belong to it, and to determine their subclassification. In many cases Usher has created new names for the member families to reflect their geographic location. Much of his classification is accepted by Glottolog (though his names are not, as Glottolog invents its own names). As of 2020, his classification is as follows, including correspondences to the names in earlier classifications. He expects to expand the membership of the family as reconstruction proceeds.
Trans-New Guinea phylum (Usher 2020)
These branches may cluster together (the southwestern branches, for example, may group together), but the details are as yet unclear.
The families from the Ross and Glottolog classifications that are not included are Kaure, Pauwasi, Engan, Chimbu-Wahgi, Madang, Eleman, Kiwaian, Binanderean, Goilalan, and the several Papuan Gulf families. Usher only includes families that have a regular reflex of the 2sg pronoun, so there may be additional TNG families that have changed their pronouns.