Proto-Altaic language

Based on the proposed correspondences listed below, the following phoneme inventory (in IPA notation) has been reconstructed for the hypothetical Proto(-Macro)-Altaic language.

¹ This phoneme only occurred at the beginnings of words.
² These phonemes only occurred in the interior of words.

It is not clear whether /æ/, /ø/, /y/ were monophthongs as shown here (presumably [æ œ~ø ʏ~y]) or diphthongs ([i̯a~i̯ɑ i̯ɔ~i̯o i̯ʊ~i̯u]); the evidence is equivocal. In any case, however, they only occurred in the first (and sometimes only) syllable of any word.

Every vowel occurred in long and short versions which were different phonemes in the first syllable. Starostin et al. (2003) treat length together with pitch as a prosodic feature.

As reconstructed by Starostin et al. (2003), Proto-Altaic was a pitch accent or tone language; at least the first and probably every syllable could have a high or a low pitch.

If a Proto(-Macro)-Altaic language really existed, it should be possible to reconstruct regular sound correspondences between that protolanguage and its descendants; such correspondences would make it possible to distinguish cognates from loanwords (in many cases). Such attempts have repeatedly been made. The latest version is reproduced here, taken from Blažek's (2006) summary of the newest Altaic etymological dictionary (Starostin et al. 2003) and transcribed into the IPA.

When a Proto-Altaic phoneme developed differently depending on its position in a word (beginning, interior, or end), the special case (or all cases) is marked with a hyphen; for example, Proto-Altaic /pʰ/ disappears (marked "0") or becomes /j/ at the beginning of a Turkic word and becomes /p/ elsewhere in a Turkic word.

Only single consonants are considered here. In the middle of words, clusters of two consonants were allowed in Proto-Altaic as reconstructed by Starostin et al. (2003); the correspondence table of these clusters spans almost seven pages in their book (83–89), and most clusters are only found in one or a few of the reconstructed roots.

¹ The Khalaj language has /h/ instead. (It also retains a number of other archaisms.) However, it has also added /h/ in front of words for which no initial consonant (except in some cases /ŋ/, as expected) can be reconstructed for Proto-Altaic; therefore, and because it would make them dependent on whether Khalaj happens to have preserved any given root, Starostin et al. (2003: 26–28) have not used Khalaj to decide whether to reconstruct an initial /pʰ/ in any given word and have not reconstructed a /h/ for Proto-Turkic even though it was probably there.
² The Monguor language has /f/ here instead (Kaiser & Shevoroshkin 1988); it is therefore possible that Proto-Mongolian also had /f/ which then became /h/ (and then usually disappeared) in all descendants except Monguor. Tabgač and Kitan, two extinct Mongolic languages not considered by Starostin et al. (2003), even preserve /p/ in these places (Blažek 2006).
³ This happened when the next consonant in the word was /lʲ/, /rʲ/, or /r/.
⁴ Before /i/.
⁵ When the next consonant in the word was /h/.
⁶ This happened "in syllables with original high pitch" (Starostin et al. 2003:135).
⁷ Before /æ/, /ø/ or /y/.
⁸ When the next consonant in the word was /r/.
⁹ When the preceding consonant was /r/, /rʲ/, /l/, or /lʲ/, or when the next consonant was /ɡ/.
¹⁰ Before /a/, /ə/, or any vowel followed by /j/.
¹¹ Before /j/, or /i/ and then another vowel.
¹² When preceded by a vowel preceded by /i/.
¹³ Before /a/.
¹⁴ Starostin et al. (2003) follow a minority opinion (Vovin 1993) in interpreting the sound of the Middle Korean letter ᅀ as [nʲ] or [ɲ] rather than [z]. (Dybo & Starostin 2008:footnote 50)
¹⁵ Before /u/.
¹⁶ Before /a/, /o/, or /e/.

Vowel harmony is pervasive in the languages attributed to Altaic: most Turkic and Mongolic as well as some Tungusic languages have it, Korean is arguably in the process of losing its traces, and it is controversially hypothesized for Old Japanese. (Vowel harmony is also typical of the neighboring Uralic languages and was often counted among the arguments for the Ural–Altaic hypotheses.) Nevertheless, Starostin et al. (2003) reconstruct Proto-Altaic as lacking vowel harmony. Instead, according to them, vowel harmony originated in each daughter branch as assimilation of the vowel in the first syllable to the vowel in the second syllable (which was usually modified or lost later). "The situation therefore is very close, e.g. to Germanic [see Germanic umlaut] or to the Nakh languages in the Eastern Caucasus, where the quality of non-initial vowels can now only be recovered on the basis of umlaut processes in the first syllable." (Starostin et al. 2003:91) The table below is taken from Starostin et al. (2003):

¹ When preceded by a bilabial consonant.
² When followed by a trill, /l/, or /lʲ/.
³ When preceded or followed by a bilabial consonant.
⁴ When preceded by a fricative (/s/, /ʃ/, /x/).

Length and pitch in the first syllable evolved as follows according to Starostin et al. (2003), with the caveat that it is not clear which pitch was high and which was low in Proto-Altaic (Starostin et al. 2003:135). For simplicity of input and display every syllable is symbolized as "a" here:

¹ "Proto-Mongolian has lost all traces of the original prosody except for voicing *p > *b in syllables with original high pitch" (Starostin et al. 2003:135).
² "[...] several secondary metatonic processes happened [...] in Korean, basically in the verb subsystem: all verbs have a strong tendency towards low pitch on the first syllable." (Starostin et al. 2003:135)

Starostin et al. (2003) have reconstructed the following correspondences between the case and number suffixes (or clitics) of the (Macro-)Altaic languages (taken from Blažek, 2006):

/V/ symbolizes an uncertain vowel. Suffixes reconstructed for Proto-Turkic, Proto-Mongolic, Proto-Korean, or Proto-Japonic, but not attested in Old Turkic, Classical Mongolian, Middle Korean, or Old Japanese are marked with asterisks.

This correspondences, however, have been harshly criticized for several reasons: There are significant gaps resulting in the absence of etymologies for certain initial segments: an impossible situation in the case of a genetic relationship; lack of common paradigmatic morphology; in many cases, there are ghosts, invented or polished meanings; and word-list linguistics rules supreme, as there are few if any references to texts or philology.[14][15]

There are also many reconstructions proved to be totally false. For instance, regarding Korean, Starostin et al. state that Middle Korean genitive is /nʲ/, while it actually was /s/ in its honorific form, and /ój/ or /uj/ as neutral forms.[16]

In addition, some "cognates" are visibly forced, like the comparison between Turkish instrumental -/n/, -/ɯn/, -/in/ and Japanese locative /ni/. A locative postposition expresses an absolutely different meaning to that of an instrumental, so it is evident that both of them are not related whatsoever. The same applies for Japanese /ga/ and Proto Tungusic /ga/. The first of those particles expresses genitive case, while the second is the partitive case, which bear no resemblance of meaning at all either. A different kind of issue is that of the Old Turkish genitive /Xŋ/ (where "X" stands for any phoneme) and Old Japanese genitive /no/. Although they share the same consonant, the fact that the former is a vowel plus a consonant, and the second is a fixed set of the consonant /n/ plus vowel /o/ makes the fact that those two are cognates extremely unlikely.[17]

The table below is taken (with slight modifications) from Blažek (2006) and transcribed into IPA.

As above, forms not attested in Classical Mongolian or Middle Korean but reconstructed for their ancestors are marked with an asterisk, and /V/ represents an uncertain vowel.

There are, however, several problems with this proposed list. Aside from the huge amount of non-attested, free reconstructions, some mistakes on the research carried out by Altaicists must be pointed out. The first of them is that Old Japanese for the first person pronoun ("I", in English) was neither /ba/ or /a/. It was /ware/ (和禮), and sometimes it was abbreviated to /wa/ (吾). Also, it is not a Sino-Japanese word, but a native Japanese term. In addition, the second person pronoun was not /si/, but either /imasi/ (汝), or /namu/ (奈牟), which sometimes was shortened to /na/. Its plural was /namu tachi/ (奈牟多知).[18]

The following table is a brief selection of further proposed cognates in basic vocabulary across the Altaic family (from Starostin et al. [2003]). Their reconstructions and equivalences are not accepted by the mainstream linguists and therefore remain very controversial.

¹ Contains the Proto-Altaic dual suffix -/rʲV/: "both breasts" – "chest" – "heart".
² Contains the Proto-Altaic singulative suffix -/nV/: "one breast".
³ Compare Baekje */turak/ "stone" (Blažek 2006).
⁴ This is in the Jurchen language. In modern Manchu it is usiha.
⁵ This is disputed by Georg (2004),[20] who states: "The traditional Tungusological reconstruction *yāsa [ = /jaːsa/] cannot be replaced by the nasal-initial one espoused here, needed for the comparison." However, Starostin (2005)[21] mentions evidence from several Tungusic languages cited by Starostin et al. (2003). Georg (2005) does not accept this, referring to Georg (1999/2000) and a then upcoming paper.[21]

In the Indo-European family, the numerals are remarkably stable. This is a rather exceptional case; especially words for higher numbers are often borrowed wholesale. (Perhaps the most famous cases are Japanese and Korean, which have two complete sets of numerals each – one native, one Chinese.) Indeed, the Altaic numerals are less stable than the Indo-European ones, but nevertheless Starostin et al. (2003) reconstruct them as follows. They are not accepted by the mainstream linguists and are controversial. Other reconstructions show little to no similarities in numerals of the proto-languages.

¹ Manchu /soni/ "single, odd".
² Old Bulgarian /tvi-rem/ "second".
³ Kitan has /tʃur/ "2" (Blažek 2006).
⁴ -/uː/- is probably a contraction of -/ubu/-.
⁵ The /y/- of /ytʃ/ "3" "may also reflect the same root, although the suffixation is not clear." (Starostin et al. 2003:223)
⁶ Compare Silla /mir/ "3" (Blažek 2006).
⁷ Compare Goguryeo /mir/ "3" (Blažek 2006).
⁸ "third (or next after three = fourth)", "consisting of three objects"
⁹ "song with three out of four verses rhyming (first, second and fourth)"
¹⁰ Kitan has /dur/ "4" (Blažek 2006).
¹¹ Kitan has /tau/ "5" (Blažek 2006).
¹² "(the prefixed i- is somewhat unclear: it is also used as a separate word meaning 'fifty', but the historical root here is no doubt *tu-)" (Starostin et al. 2003:223). – Blažek (2006) also considers Goguryeo */uts/ "5" (from */uti/) to be related.
¹³ Kitan has /nir/ "6" (Blažek 2006).
¹⁴ Middle Korean has /je-(sɨs)/ "6", which may fit here, but the required loss of initial /nʲ/- "is not quite regular" (Starostin et al. 2003:224).
¹⁵ The Mongolian forms "may suggest an original proto-form" /lʲadi/ or /ladi/ "with dissimilation or metathesis in" Proto-Mongolic (Starostin et al. 2003:224). – Kitan has /dol/ "7".
¹⁶ /ɖirkup/ in Early Middle Korean (タリクニ/チリクヒ in the Nichūreki).
¹⁷ "Problematic" (Starostin et al. 2003:224).
¹⁸ Compare Goguryeo /tok/ "10" (Blažek 2006).
¹⁹ Manchu /dʒiri/, /dʒirun/ "a very big number".
²⁰ Orok /poːwo/ "a bundle of 10 squirrels", Nanai /poã/ "collection, gathering".
²¹ "Hundred" in names of hundreds.
²² Starostin et al. (2003) suspect this to be a reduplication: */kɯr-kɯr/ "20 + 20".
²³ /kata-ti/ would be expected; Starostin et al. (2003) think that this irregular change from /k/ to /p/ is due to influence from "2" /puta-tu/.
²⁴ From */nʲam-ŋu-/.
²⁵ Also see Tümen.
²⁶ Modern Korean – needs further investigations