Altaic languages

• 1. Introduction

  ›

  The Altaic languages bring together about 66 living tongues from three families, Turkic, Mongolic and Tungusic, spread across the breadth of Asia. Their speakers live from open steppe to forests far to the east. Yet linguists cannot agree on what, if anything, binds them. For a long stretch of the twentieth century, the common answer was that they descend from one shared ancestor. Today most comparative linguists say that answer was mistaken, while a stable minority of scholars still defends it. So what do these languages really share? Are they cousins, or neighbours who spent centuries borrowing from one another? And what did the earliest scribes, travellers and philologists notice that made them suspect a connection at all? The answers run from carved inscriptions to statistical modelling, and the argument is still running.

  46122

• 2. Earliest Written Traces

  ›

  One of the earliest traces of a Turkic language is a horse. Lanhai Wei and Hui Li reconstruct the name of the Xiōngnú ruling house as Alayundluğ, which they gloss as the piebald horse clan. Anna Dybo finds several dozen foreign-sounding Proto-Turkic words in Han Chinese transcriptions. They appear in the Shizi, from 330 BC, and in the Book of Han, from AD 111.

  True Turkic texts come later. The Orkhon inscriptions, carved between 720 and 735 AD, are the earliest known. In 1893 the Danish linguist Vilhelm Thomsen deciphered them, in a scholarly race against his rival Wilhelm Radloff, who worked in Germany and Russia. Radloff, however, was the first to publish.

  The first Tungusic language on record is Jurchen, spoken by the ancestors of the Manchus. A writing system for it was devised in 1119. An inscription using that system survives from 1185.

  For Mongolic, the Stele of Yisüngge, dated to 1224 or 1225, is the earliest written evidence of Middle Mongol. The Secret History of the Mongols followed in 1228. Older traces exist in related languages. The Memorial for Yelü Yanning, written in the Khitan large script in 986, is the earliest Para-Mongolic text. The Inscription of Hüis Tolgoi, discovered in 1975, is analysed as early Mongolic and dated to 604-620. The Bugut inscription reaches back to 584.

  Japanese first shows up as personal names in short Classical Chinese inscriptions of the 5th century, such as the one on the Inariyama Sword. The first substantial text is the Kojiki, from 712. The Nihon shoki, completed in 720, comes next. After it comes the Man'yōshū, from about 771-785, which preserves material roughly 400 years older.

  Early Korean rests mainly on the Hyangga, a collection of 25 poems. Some reach back to the Three Kingdoms period, 57 BC-668 AD, but they survive in a spelling system only as old as the 9th century. From the mid-15th century on, the phonetically precise Hangul script documents Korean copiously, a far fuller record than that 9th-century spelling.

  1716846615

• 3. A Name From the Mountains

  ›

  Nicolaes Witsen's work of 1692 holds the earliest known reference to Turkic, Mongolic and Tungusic as one unified group. It may rest on a 1661 work by Abu al-Ghazi Bahadur, the Genealogy of the Turkmens.

  In 1730 Philip Johan von Strahlenberg published a proposed grouping of the same three. He was a Swedish officer who had travelled through the eastern Russian Empire as a prisoner of war after the Great Northern War. He may not, however, have meant to suggest a closer relationship among them.

  The Finnish philologist Matthias Castrén widened the net in 1844. His broader grouping, later called Ural-Altaic, set Turkic, Mongolian and Manchu-Tungus together as an Altaic branch. Beside it stood the Finno-Ugric and Samoyedic languages as the Uralic branch. Castrén himself preferred the labels Tataric and Chudic. The later names came from geography. The Altai Mountains of East-Central Asia sit roughly at the centre of the three main families' range, and the Ural Mountains gave Uralic its name.

  Since the 1960s the Ural-Altaic hypothesis has drawn heavy criticism, though it survives in some reference works and atlases. Even Sergei Starostin, who accepts a basic Altaic family, discards the Uralic branch entirely.

  The old name still has a defender for a different purpose. Juha Janhunen uses it for the central Eurasian zone of shared grammar and vocabulary. He argues that "speaking of 'Altaic' instead of 'Ural-Altaic' is a misconception, for there are no areal or typological features that are specific to 'Altaic' without Uralic."

  Once the family grew further east, the original three-family proposal was renamed Micro-Altaic after the fact, a retronym.

  2054366

• 4. Korean, Japanese and Ainu

  ›

  In 1857 the Austrian scholar Anton Boller suggested adding Japanese to the Ural-Altaic family. Korean joined the conversation in the 1920s, advocated by G.J. Ramstedt and E.D. Polivanov. By his 1952 book, Ramstedt had rejected the Uralic link but kept Korean. That book made the first comprehensive attempt to find regular correspondences among the sound systems of the Altaic families. Most leading Altaicists, the theory's supporters, have followed his choice ever since.

  Nicholas Poppe's 1960 book was in effect a heavily revised version of Ramstedt's volume on phonology, and it set the standard for Altaic studies. Poppe thought Korean's place unsettled and listed three possibilities. Korean might be unrelated but shaped by an Altaic substratum. It might be as closely related to the other three as they are to each other. Or it might have split off before they underwent their characteristic changes.

  Roy Andrew Miller's 1971 book Japanese and the Other Altaic Languages convinced most Altaicists that Japanese belonged too. The resulting Macro-Altaic grouping adds Korean, Jeju, Japanese and the Ryukyuan languages, lifting the total to about 74, depending on what counts as a dialect. In 1990 J. Marshall Unger urged building relationships rigorously from the bottom up. He proposed comparing Tungusic with the common ancestor of Korean and Japanese before seeking links to Turkic or Mongolic.

  Ainu entered the debate in 1962, when John C. Street placed Korean, Japanese and Ainu in one grouping and Turkic, Mongolic and Tungusic in another. He called the whole the North Asiatic family. James Patrie adopted the idea in 1982. Joseph Greenberg posited the same two groupings in 2000-2002 but treated them as independent members of a larger family he named Eurasiatic. Few Altaicists accept Ainu, and no convincing relationship to any other language family has been shown, so it is generally regarded as a language isolate. Some authors have instead tried to tie Japanese to the Austronesian languages.

  In 2017 Martine Robbeets proposed that Japanese, and possibly Korean, began as hybrid languages. In her scenario the ancestors of Turkic, Mongolic and Tungusic lived in northwestern Manchuria. Some of them migrated south into the modern Liaoning province, where an agricultural community with an Austronesian-like language largely absorbed them. Their fusion would have produced proto-Japanese and proto-Korean. In 2022 Zheng Tiang and colleagues replied that the view suffers from fundamental problems, several contradictions, and data that are non-replicable and invalid. They found no empirical support for a Transeurasian family, nor for tying five language families to Neolithic farmers spreading from the West Liao River region.

  Today Macro-Altaic is firmly rejected, even by proponents of Altaic. A 2020 typological study by Yurayong and Szeto found Koreanic and Japonic "still so different from the Core Altaic languages that we can even speak of an independent Japanese-Korean type of grammar." In 2023 Janhunen affirmed that the two groups lack "any significant number of material similarities with the three Micro-Altaic families, or with each other." He added that Pre-Proto-Japonic stood closer to Sinitic than to anything Altaic-related.

  2627303135

• 5. The Case Against

  ›

  Gerard Clauson in 1956, Gerhard Doerfer in 1963 and Alexander Shcherbak led the early attack on even the minimal three-family theory. They held that most shared words and features were borrowings and that the rest were chance resemblances. Other doubts piled up. Supposed cognates turned out to be invalid, and the sound shifts hypothesized for the family could not be found. Doerfer rejected every genetic claim again in 1988.

  Typology gave critics a second line of argument. Clauson, Doerfer and Shcherbak pointed out that strongly suffixing, agglutinative morphology and subject-object-verb word order often occur together in languages generally. Sharing them proves little.

  The vocabulary pattern also troubled them. Turkic and Tungusic share little vocabulary, while Mongolic shares more with each. If all three came from one ancestor, the critics reasoned, losses should strike at random, not only at the geographical edges of the family. The observed pattern looks like borrowing.

  Claus Schönig's 2003 overview sided with these critics. After he accounted for areal effects, the vocabulary that could be genetically inherited shrank to a small number of monosyllabic roots. These included the personal pronouns and a few other pointing and auxiliary words, whose sharing has other explanations. It was not the kind of overlap expected from a genetic relationship.

  In place of common descent, the three critics proposed in 1956-1966 that Turkic, Mongolic and Tungusic form a sprachbund. That is a set of languages whose likenesses come from intensive borrowing and long contact rather than a shared origin.

  Asya Pereltsvaig added an argument in 2011. Related languages normally diverge over time, so earlier forms resemble each other more than modern ones do. Yet the earliest written records of Mongolic and Turkic show the opposite pattern, which suggests they grew more alike through contact.

  It fell to defenders such as Roy Miller and Sergei Starostin to answer.

  6263611110

• 6. The Case For

  ›

  The original reasons for grouping Turkic, Mongolic and Tungusic in a Uralo-Altaic family were vowel harmony and agglutination. Roy Miller, defending Altaic in 1991 and again in 1996, argued that Clauson's and Doerfer's criticisms touch only the word correspondences. In his view, the strongest evidence lies in similar verbal morphology.

  In 1991 Sergei Starostin ran a comparison on a 110-word Swadesh-Yakhontov list and claimed about 15-20% apparent cognates among the proposed members. By pair, the figures were Turkic-Mongolic 20%, Turkic-Tungusic 18%, Turkic-Korean 17%, Mongolic-Tungusic 22%, Mongolic-Korean 16% and Tungusic-Korean 21%. He conceded that the grouping was "older than most other language families in Eurasia, such as Indo-European or Finno-Ugric, and this is the reason why the modern Altaic languages preserve few common elements".

  The 2003 Etymological Dictionary of the Altaic Languages, by Starostin, Anna Dybo and Oleg Mudrak, expanded those lists to 2,800 proposed cognate sets. Its authors tried hard to separate loanwords between Turkic and Mongolic, and between Mongolic and Tungusic, from true cognates. They also point to words found in Turkic and Tungusic but absent from Mongolic. The book lists 144 items of shared basic vocabulary, among them eye, ear, neck, bone, blood, water, stone, sun and two.

  The dictionary also proposes sound-change laws. In its reconstruction, Proto-Altaic itself lacked vowel harmony, even though most daughter languages have it today. Instead, various vowel assimilations between first and second syllables arose separately in Turkic, Mongolic, Tungusic, Korean and Japonic.

  Newer statistical work has given the idea fresh support. Robbeets and Bouckaert used Bayesian phylolinguistic methods in 2018 to argue that the three core families cohere with Japonic and Koreanic, a set they call Transeurasian. In 2021 Robbeets and colleagues argued that early Transeurasian speakers were farmers in Northeastern Asia who became herders only later.

  Also in 2021, Kassian and colleagues compared specially built 110-item word lists for Proto-Turkic, Proto-Mongolic, Proto-Tungusic, Middle Korean and Proto-Japonic. The results gave partial support to the Altaic macrofamily, with Korean seemingly excluded. Ancient contact remained a plausible explanation, but the team judged it less likely for the matches between Turkic and Japonic. The substantial distances between those two, they argued, make a genealogical link the better explanation.

  5455565859

• 7. Homeland and Age

  ›

  Little is known of the prehistory of the peoples who speak these languages. For Indo-European, Uralic and Austronesian, scholars can frame substantial hypotheses about origins. For the proposed Altaic family, much remains to be done.

  Some scholars have placed a shared Uralic and Altaic homeland in the Central Asian steppes. Juha Janhunen has a tighter proposal. He places the ancestors of Turkic, Mongolic, Tungusic, Korean and Japanese in a relatively small area covering present-day North Korea, Southern Manchuria and Southeastern Mongolia. Even so, he doubts that Japanese belongs with Altaic. András Róna-Tas remarked that any link between Altaic and Japanese must be more remote than the link between any two Indo-European languages.

  S. Robert Ramsey was similarly cautious about Korean and Japanese themselves. He wrote that their genetic relationship, "if it in fact exists, is probably more complex and distant than we can imagine on the basis of our present state of knowledge."

  Supporters once dated Proto-Altaic to around 4000 BC. They now place it around 5000 BC or 6000 BC. That would make it older than Indo-European, put at around 3000 to 4000 BC by mainstream hypotheses. It would also make it far younger than Afroasiatic, dated from c. 10,000 BC to 16,000 BC depending on the source.

  Genetics has entered the picture as well. Chaubey and van Driem link the spread of ancient Altaic-speaking communities to the early Holocene dissemination of the paternal lineage haplogroup C2 (M217). It appears to track Turkic, Mongolic and Tungusic speakers. It accounts for 11% of Korean paternal lineages and even less of Japanese ones, so they conclude that neither Japanese nor Korean is a "Father Tongue". They suggest the marker may still trace the arrival of Altaic language in the Japanese archipelago, where it persists at just 6%.

  476798036

• 8. A Quarrel Without End

  ›

  Stefan Georg, in 2004 and 2005, and Alexander Vovin, in 2005, criticised Starostin's dictionary. Vovin's stance carries a twist. He had been an advocate of Altaic through 1994-2001, and later became a critic, publishing against it in 2005-2010 and 2017.

  Defenders answered in print. Starostin replied in 2005, Blažek in 2006 and Robbeets in 2007. Dybo and G. Starostin followed in 2008. In 2010 Lars Johanson echoed Miller's 1996 rebuttal and called for a muting of the polemic.

  That same year Robbeets and Johanson proposed retiring the family label in favour of a broader Transeurasian superfamily. Altaic would survive as a subgroup of Turkic, Mongolic and Tungusic, beside a second group called Japano-Koreanic, made of Japonic and Koreanic. They gave four reasons. The new term would end confusion over which languages Altaic includes, soften the polarization between Pro-Altaists and Anti-Altaists, and drop the reference to the Altai mountains as a potential homeland. It would also be broader in applicability, because the suffix -ic implies affinity while -an leaves room for an areal hypothesis.

  The theory still has institutional backing. It is supported by the Institute of Linguistics of the Russian Academy of Sciences. It also remains influential as a substratum of Turanism, a multiethnic nationalist movement that has used a hypothetical common linguistic ancestor as part of its basis.

  Because older handbooks still list Altaic as a family, references to it keep percolating into modern sources, long after most linguists moved on.

  4547465253